Sea Ice: Physics and Remote Sensing [2 ed.] 1119828163, 9781119828167

Table of contents :
Cover
Title Page
Copyrigt Page
Contents
Preface
Acknowledgments and Recognitions
Chapter 1 Introduction
1.1 Background
1.2 Canada and the Arctic: Historical and Community Synopsis
1.3 The Fascinating Nature of Sea Ice
1.4 Sea Ice in Research and Operational Disciplines
1.4.1 Sea Ice in Physics
1.4.2 Sea Ice in Climatology
1.4.3 Sea Ice in Meteorology
1.4.4 Sea Ice in Oceanography
1.4.5 Sea Ice in Marine Biology
1.4.6 Sea Ice in Marine Navigation
1.4.7 Sea Ice and Offshore Structures
1.4.8 Sea Ice as A Transportation Platform
1.4.9 Sea Ice in Relation to Solid Earth Sciences: Rocks and Plate Tectonics
1.5 Sea Ice and Remote Sensing
1.6 Motivation for the Book Writing
1.7 Organization of the Book
1.8 References
Chapter 2 Ice Physics and Physical Processes
2.1 Prior to Freezing: About Freshwater and Seawater
2.1.1 Molecular Composition of Water
2.1.2 Seawater Salinity
2.1.3 Seawater Density
2.2 Phase Diagram of Sea Ice
2.3 Initial Ice Formation
2.3.1 Freezing Processes in Freshwater and Seawater
2.3.2 Initial Formation of Ice Crystals and Frazil Ice
2.4 Sea Ice Growth
2.4.1 Lateral Ice Growth
2.4.2 Vertical Ice Growth (Congelation Ice)
2.4.3 Superimposed Ice
2.4.4 Thermodynamic Ice Growth
2.4.4.1 Simplified Models of Sea Ice Growth
2.4.4.2 Effect of Snow On Sea Ice
2.4.4.3 Effect of Oceanic Heat Flux
2.4.4.4 Effect of Surface Ablation
2.5 Processes in Ice
2.5.1 Compositional (Constitutional) Supercooling At the Ice–Water Interface
2.5.2 Dendritic Ice–Water Interface and Entrapment of Brine Within Sea Ice
2.5.3 Grains and Subgrains In Sea Ice
2.5.4 Brine Pockets Formation, Contents and Distribution In Sea Ice
2.5.5 Salinity Loss During Sea Ice Growth
2.5.5.1 Initial Rapid Salt Rejection At the Ice–Water Interface
2.5.5.2 Subsequent Slow Salt Rejection from the Bulk Ice
2.6 Ice Deformation
2.6.1 Rafting of Thin Ice
2.6.2 Ridging of Thick Ice
2.6.3 Formation of Ice Rubble Field
2.6.4 Fractures in Ice Cover
2.7 Ice Decay and Aging
2.7.1 Ice Decay
2.7.2 Ice Aging
2.8 Sea Ice Classes
2.9 Sea Ice Regimes
2.9.1 Polynyas
2.9.2 Pancake Ice Regime
2.9.3 Marginal Ice Zone and Ice Edge
2.9.3.1 Marginal Ice Zone
2.9.3.2 Ice Edge
2.9.4 Ice of Glacier Origin
2.10 References
Chapter 3 Sea Ice Properties: Data and Derivations
3.1 Typical Values of Sea Ice and Snow Physical Parameters
3.2 Temperature Profiles in Ice and Snow
3.3 Bulk Salinity and Salinity Profile
3.3.1 Bulk Salinity
3.3.2 Salinity Profiles
3.4 Density of First-Year and Multi-Year Ice
3.5 Volume Fraction of Sea Ice Constituents
3.5.1 Brine Volume Fraction
3.5.2 Solid Salt Volume Fraction
3.5.3 Pure Ice Volume Fraction
3.5.4 Air Volume Fraction
3.5.5 Temperature Dependence of Volume Fractions of Different Components
3.6 Thermal Properties
3.6.1 Thermal Conductivity of Sea Ice
3.6.2 Thermal Conductivity of Snow
3.6.3 Specific Heat of Sea Ice
3.6.4 Latent Heat of Sea Ice
3.7 Dielectric Properties
3.7.1 Dielectric Constant of Brine
3.7.2 Dielectric Mixing Models
3.7.3 Field Measurements of Dielectric Constant
3.8 References
Chapter 4 Laboratory Techniques for Revealing the Structure of Polycrystalline Ice
4.1 Relevant Optical Properties
4.1.1 Polarized Light
4.1.2 Birefringence or Double Refraction of Ordinary (Ih) Ice
4.1.3 Optical Retardation
4.1.4 Interference Colors for White Light
4.2 Ice Thin Sectioning Techniques
4.2.1 Hot-plate Techniques for Thin Sectioning of Ice
4.2.2 Double-Microtoming Technique for Thin Sectioning of Ice
4.2.3 Double-Microtoming Technique for Thin Sectioning of Snow
4.2.4 Precautions for Thin Sectioning by DMT
4.2.5 Optimum Thickness for Thin Sections of Ice and Snow
4.3 Viewing and Photographing Ice Thin Sections
4.3.1 Laboratory and Hand-Held Polariscope
4.3.2 Cross-Polarized versus Parallel-Polarized Light Viewing
4.3.3 Scattered Light and Combined Cross-Polarized/Scattered Light Viewing
4.3.4 Circularly Polarized Light and Rapid Crystallographic Analysis
4.4 Advanced Techniques for Revealing Fine Crystallographic Microstructural Features
4.4.1 Sublimation of Ice and Sublimation Pits
4.4.2 Etching Processes
4.4.2.1 Thermal Etching of Microtomed Ice Surfaces
4.4.2.2 Chemical Etching and Replicating Ice Surfaces
4.5 References
Chapter 5 Polycrystalline Ice Structure
5.1 Terms and Definitions Relevant to Polycrystalline Ice
5.1.1 Special Thermal State of Natural Ice
5.1.2 General Terms for Structural Aspects of Ice
5.1.3 Basic Terms and Definitions
5.2 Morphology of Ice
5.2.1 Forms of Ice Crystals
5.2.2 Miller Indices for Hexagonal Ice
5.2.3 Growth Direction of Ice Crystals
5.2.4 Ice Density in Relation to Crystalline Structure
5.3 Structure- and Texture-Based Crystalline Classification of Natural Ice
5.3.1 Freshwater Ice Classification of Michel and Ramseier
5.3.2 Extending Crystallographic Classification of Freshwater Ice to Sea Ice
5.3.3 Crystallographic Classes of Natural Ice
5.3.3.1 Granular or Snow Ice (T1 Ice)
5.3.3.2 Randomly Oriented (S4) and Vertically Oriented (S5) Frazil Ice
5.3.3.3 Columnar-Grained with c Axis Vertical (S1 Ice)
5.3.3.4 Columnar-Grained with c Axis Horizontal and Random (S2 Ice)
5.3.3.5 Columnar-Grained with c Axis Horizontal and Oriented (S3 Ice)
5.3.3.6 Agglomerate Ice with Discontinuous Columnar-Grained (R Type Ice)
5.3.3.7 Ice of Land-Based Origin
5.3.3.8 Platelet Sea Ice
5.3.4 Stereographical Projection (Fabric Diagram) of Natural Polycrystalline Ice
5.4 Examples of Crystallographic Structure of Natural Sea Ice
5.4.1 Crystallographic Structure of Seasonal Sea Ice
5.4.1.1 Frazil Ice (S5 Type)
5.4.1.2 Columnar-Grained Ice (S3 Type)
5.4.1.3 Agglomeration of Various Crystallographic Structures
5.4.1.4 Air Entrapment in Seasonal Ice
5.4.2 Crystallographic Structure of Perennial Sea Ice
5.4.2.1 Hummock Ice
5.4.2.2 Melt Pond Ice
5.5 Biomass Accumulation at the Bottom of the Ice
5.6 Information Contents in Polycrystalline Ice Structure
5.6.1 Geometric Characteristics of Crystalline Structure
5.6.2 Geometric Characteristics of Brine Pockets in First-Year Ice
5.6.3 Geometric Characteristics of Air Bubbles
5.7 References
Chapter 6 Major Field Expeditions to Study Sea Ice
6.1 The Arctic Ice Dynamic Joint Experiment (AIDJEX)
6.2 Mould Bay Experiments 1981–1984: Stories that Were Never Told
6.2.1 Site, Resources, and Logistics
6.2.2 Sea Ice Conditions
6.2.3 Aging of Sea Ice: from FYI to MYI
6.2.4 Interface Between Old and New Ice in Second-Year Ice Profile
6.3 High Arctic Experience with Ice of Land Origin
6.3.1 Ward Hunt Ice Shelf and Hobson's Choice Ice Island Experiment
6.3.2 Multi-Year Rubble Field Around the Ice Island
6.4 Labrador Ice Margin Experiment (LIMEX)
6.5 Sea Ice Monitoring and Modeling Site (SIMMS) Program
6.6 The Surface Heat Budget of Arctic Ocean (SHEBA)
6.7 The Norwegian Young Sea Ice Experiment (N-ICE)
6.8 Marginal Ice Zone (MIZ) Experiments
6.9 Ice Exercise by Us Navy
6.10 The Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC)
6.11 References
Chapter 7 Remote Sensing Fundamentals Relevant to Sea Ice
7.1 General Principles of Satellite Remote Sensing
7.2 Electromagnetic Wave Properties and Processes
7.2.1 Polarization of EM Wave
7.2.2 Reflection, Transmission, Absorption, Scattering, and Emission
7.2.2.1 Reflection and Fresnel Model
7.2.2.2 Transmission
7.2.2.3 Absorption and Scattering Losses
7.2.2.4 Emitted Radiation (Re-radiation)
7.2.3 Brightness Temperature and Emissivity
7.2.4 Penetration Depth
7.3 Optical Sensing
7.4 Thermal Infrared Sensing
7.5 Microwave Remote Sensing
7.6 Imaging Radar Sensing
7.6.1 Imaging Radar Principles
7.6.1.1 Radar Equations and Spatial Resolutions of RAR and SAR
7.6.1.2 Coherency and Polarization of Radar Signals
7.6.1.3 Radar Scattering Mechanisms
7.6.2 Multichannel SAR
7.6.3 SAR Polarimetry: Formulation and Derived Parameters
7.6.3.1 Formulation of Polarimetric Measurements
7.6.3.2 Polarimetric Parameters Derived from the FP SAR Data
7.6.3.3 Linking Radar Scattering Mechanisms to Ice Features
7.6.3.4 Age-Based versus SAR-Based and Scattering- Based Sea Ice Classification
7.7 Scatterometer Systems
7.8 Altimeter Systems
7.9 Radiative Processes in Relevant Media
7.9.1 Atmospheric Influences
7.9.1.1 Influences of Atmosphere on Optical and Infrared Observations
7.9.1.2 Atmospheric Correction for Passive Microwave Observations
7.9.2 Seawater
7.9.2.1 Seawater in the Optical and Thermal Infrared Data
7.9.2.2 Seawater in the Microwave Data
7.9.3 Snow on Sea Ice: Physical and Radiative Processes
7.9.3.1 Snow in Optical and Thermal Infrared Data
7.9.3.2 Snow in the Microwave Data
Effect of Dry Snow Depth
Effect of snow density
Effect of Snow Grain Size and Ice Layering
Effects of Snow Wetness
7.10 References
Chapter 8 Satellite Sensors for Sea Ice Monitoring
8.1 Historical Synopsis of Remote Sensing Satellites for Sea Ice
8.2 Optical and Thermal Infrared Sensors
8.3 Modern Passive Microwave Sensors
8.4 Modern Imaging Radar Sensors
8.5 Scatterometer Sensors
8.6 Altimeter Sensors
8.7 References
Chapter 9 Radiometric and Scattering Observations from Sea Ice, Water, and Snow
9.1 Optical Reflectance and Albedo Data
9.2 Microwave Brightness Temperature Data
9.3 Radar Backscatter
9.3.1 Backscatter Databases from Single-Channel SAR
9.3.2 Dual Polarization Data
9.3.3 Fully Polarimetric Data
9.4 Emissivity Data in the Microwave Bands
9.5 Microwave Penetration Depth
9.6 References
Chapter 10 Retrieval of Sea Ice Surface Information
10.1 Mechanically Generated Surface Deformation
10.1.1 Rafted Ice
10.1.2 Ridged, Rubble, and Brash Ice
10.1.3 Kinematic Processes: Convergence, Divergence, Shear, and Vorticity
10.1.4 Cracks and Leads
10.2 Thermally Induced Surface Features
10.2.1 Surface Melt
10.2.1.1 Optical Observations
10.2.1.2 Passive Microwave Observations
10.2.1.3 Active Microwave Observations
10.2.1.4 Airborne Photography
10.2.2 Frost Flowers
10.3 Meteorologically Driven Surface Features
10.3.1 Polynya Identification and Properties
10.3.2 Snow Depth
10.4 References
Chapter 11 Retrieval of Sea Ice Geophysical Parameters
11.1 Sea Ice Type Classification
11.1.1 Ice Classification from Optical and TIR Systems
11.1.2 Ice Classification from Passive Microwave Data
11.1.3 Ice Classification from SAR
11.1.3.1 Ice Classification from Single-Channel SAR
11.1.3.2 Ice Classification from Dual-Channel SAR
11.1.3.3 Ice Classification from Polarimetric SAR Data
11.2 Sea Ice Concentration
11.2.1 Ice Concentration from Optical and TIR Images
11.2.2 Ice Concentration from Coarse-Resolution Microwave Observations
11.2.2.1 NASA Team (NT) Algorithm
11.2.2.2 The Enhanced NASA Team (NT2) Algorithm
11.2.2.3 The ASI Algorithm
11.2.2.4 ECICE Algorithm
11.2.2.5 Intercomparison of PM Algorithms
11.2.2.6 Sources of Error and Sensitivity of Ice Concentration Algorithms
11.2.2.7 Assessment of Ice Concentration Results Against Ice Charts
11.2.3 Ice Concentration from Fine-Resolution SAR
11.3 Sea Ice Extent and Area
11.4 Sea Ice Thickness (SIT)
11.4.1 SIT from TIR Observations
11.4.2 SIT from PM Observations
11.4.3 SIT from Altimeter Observations
11.4.4 SIT from SAR Observations
11.5 Ice Surface Temperature (IST)
11.5.1 IST from TIR Observations
11.5.2 IST from PM Observations
11.6 Sea Ice Age
11.7 Sea Ice Motion and Kinematics
11.7.1 Methods of Ice Motion Tracking
11.7.1.1 Motion Tracking Using Image Features
11.7.1.2 Motion Tracking Using Individual Sea Ice Floes
11.7.2 Operational Ice Motion Products
11.8 References
Chapter 12 Modeling Microwave Emission and Scattering from Snow-Covered Sea Ice
12.1 The Need for Modeling Microwave Emission and Scattering from Snow-Covered Sea Ice
12.1.1 The ECMWF Workshop and Large-Scale Sea Ice Modeling
12.1.2 Gross Features of Forward Models
12.2 Radiative Transfer and Modeling Approaches for Sea Ice Thermal Microwave Emission
12.2.1 Dense Media Volume Scattering
12.2.2 Sea Ice Emission Models
12.2.3 Sea Ice Backscatter Models for Level Ice
12.2.4 Sea Ice Backscatter Models for Ridged Ice
12.3 The Input to a Forward Model
12.3.1 Primary Input Parameters
12.3.2 Secondary Input Parameters
12.3.3 Tertiary Input Parameters, Volume, and Surface Scattering
12.4 Example of the Implementation of an Altimeter Model to Study the Impact of Saline Snow on the Backscatter
12.5 Example of Combining Atmospheric, Ocean, and Sea Ice Emission Models to Simulate the Noise in Sea Ice Concentration Estimates
12.5.1 Snow in the Emission Models
12.5.2 The Combined Sea Ice Thermodynamic, Atmospheric, Ocean, and Sea Ice Emission Models
12.6 Inverse Modeling
12.7 References
Chapter 13 Impacts of Climate Change on Polar Ice
13.1 The Inconvenient Truth of Global Warming: How is it Manifested in The Polar Region?
13.2 Sea Ice Regimes in the Two Polar Regions
13.2.1 Geographic Differences Between the Two Polar Regions and Their Impacts on Sea Ice
13.2.2 Differences in Sea Ice Characteristics Between the Two Polar Regions
13.3 Changes of Polar Sea Ice in Response to Global Warming
13.3.1 The Arctic and Antarctic Ice Extent
13.3.2 The Arctic and Antarctic Ice Thickness and Volume
13.3.3 The Arctic Sea Ice Age
13.3.4 The Arctic Sea Ice Dynamics
13.3.5 The Antarctic Icebergs
13.4 Coupling Between Polar Sea Ice and Environmental Factors.
13.4.1 Interaction of the Arctic Sea Ice with the Environment
13.4.1.1 Atmospheric Factors that Contribute to Changes in the Arctic Sea Ice
13.4.1.2 Enhanced Arctic Warming due to Changes of Sea Ice Cover
13.4.1.3 Arctic Warming due to Sea Ice Advection Out of the Arctic Basin
13.4.1.4 Interaction of the Arctic Sea Ice with Wind
13.4.1.5 Mutual Interactions Between the Arctic Sea Ice Cover and Oceanic Forcing
13.4.2 Interaction of the Antarctic Sea Ice with the Environment
13.4.2.1 Interaction of the Antarctic Sea ice with Atmospheric Factors
13.4.2.2 Interaction of the Antarctic Sea Ice with Oceanic Forcing
13.4.2.3 Interaction Between the Antarctic Sea Ice, Ice Shelves, and Icebergs
13.5 References
Index
EULA

Citation preview

Sea Ice

Sea Ice Physics and Remote Sensing Second Edition Mohammed Shokr

Retired Senior Scientist Meteorological Research Division Environment and Climate Change Canada Canada

Nirmal K. Sinha

Retired Senior Scientist Institute for Aerospace Research National Research Council of Canada Canada

This edition first published 2023 © 2023 John Wiley & Sons, Inc. Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, (201) 748-6011, fax (201) 748-6008, or online at http://www.wiley.com/go/permission. Trademarks: Wiley and the Wiley logo are trademarks or registered trademarks of John Wiley & Sons, Inc. and/or its affiliates in the United States and other countries and may not be used without written permission. All other trademarks are the property of their respective owners. John Wiley & Sons, Inc. is not associated with any product or vendor mentioned in this book. Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at (800) 762-2974, outside the United States at (317) 572-3993 or fax (317) 572-4002. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic formats. For more information about Wiley products, visit our web site at www.wiley.com. Library of Congress Cataloging-in-Publication Data applied for Hardback ISBN: 9781119828167 Cover Design: Wiley Cover Images: Courtesy of Nirmal K. Sinha and Mohammed Shokr Set in 10/12pt Times New Roman by Straive, Pondicherry, India Front cover images: the main central photo shows the Canadian icebreaker Pierre Radisson, operated by the Canadian Coast Guard, in the north water polynya (Sarvarjuaq to Inuit in Canada) north of the Baffin Bay. A few scientists were exploring a possible sampling site on the apparent thin sea ice in late April 1998. The photo was taken by M. Shokr while joining the expedition. The bottom left image is a photograph of a thin section of young ice in a calm bay in the Labrador Sea, eastern Canada, showing the crystallographic structure of the ice. The photo was taken by M. Shokr in March 1996. More details are given in section (5.3.3.4) The center bottom image is an artistic view of a European radar satellite orbiting the earth. The bottom right image, showing the letters “ICE”, illustrates the principle of interference colors for optical retardation, which articulates the “anatomy” of sea ice (i.e., the crystalline structure). The word ICE has been cut from vertical thick sections of directionally solidified columnar-crystal ice, frozen from distilled water. More information is presented in section (4.2.5). The image was prepared and photographed by N.K. Sinha. Back cover image: this was also taken by N.K. Sinha and it shows the images of the two co-authors of this book reflected off the dark glasses of a fellow engineer during a field expedition in the Canadian Arctic in May 1993 (M. Shokr at the left side and N.K. Sinha at the right side).

To the two most caring women in my life, my Late mother and my wife; my three sons who taught me more than I have ever taught them; and to all those who inspired me to finish this book knowing that they will not read it.

Mohammed Shokr If I was successful as a scientist, it was only because of the unconditional support I received from my wife, Supti Sinha, and my three daughters, Priya, Roona, and Shoma, who often helped me in my cold laboratory. Shoma was born while I was camping on an old ice floe near the North Pole during my long, long absence from home.

Nirmal K. Sinha

CONTENTS Preface .................................................................................................................................................................... xv Acknowledgments and Recognitions .................................................................................................................... xvii 1. Introduction......................................................................................................................................................... 1 1.1 Background.................................................................................................................................................1 1.2 Canada and the Arctic: Historical and Community Synopsis .....................................................................4 1.3 The Fascinating Nature of Sea Ice .............................................................................................................. 8 1.4 Sea Ice in Research and Operational Disciplines ..................................................................................... 12 1.4.1 Sea Ice in Physics .......................................................................................................................... 12 1.4.2 Sea Ice in Climatology...................................................................................................................13 1.4.3 Sea Ice in Meteorology ..................................................................................................................14 1.4.4 Sea Ice in Oceanography ..............................................................................................................15 1.4.5 Sea Ice in Marine Biology..............................................................................................................16 1.4.6 Sea Ice in Marine Navigation ........................................................................................................17 1.4.7 Sea Ice and Offshore Structures .....................................................................................................19 1.4.8 Sea Ice as A Transportation Platform ............................................................................................. 20 1.4.9 Sea Ice in Relation to Solid Earth Sciences: Rocks and Plate Tectonics ........................................ 21 1.5 Sea Ice and Remote Sensing ..................................................................................................................... 22 1.6 Motivation for the Book Writing ............................................................................................................... 24 1.7 Organization of the Book.......................................................................................................................... 25 1.8 References................................................................................................................................................. 27 2. Ice Physics and Physical Processes ................................................................................................................... 29 2.1 Prior to Freezing: About Freshwater and Seawater ................................................................................... 30 2.1.1 Molecular Composition of Water .................................................................................................. 30 2.1.2 Seawater Salinity............................................................................................................................31 2.1.3 Seawater Density ...........................................................................................................................32 2.2 Phase Diagram of Sea Ice ......................................................................................................................... 33 2.3 Initial Ice Formation.................................................................................................................................. 33 2.3.1 Freezing Processes in Freshwater and Seawater ............................................................................ 33 2.3.2 Initial Formation of Ice Crystals and Frazil Ice...............................................................................35 2.4 Sea Ice Growth ......................................................................................................................................... 37 2.4.1 Lateral Ice Growth ......................................................................................................................... 37 2.4.2 Vertical Ice Growth (Congelation Ice) ...........................................................................................38 2.4.3 Superimposed Ice .......................................................................................................................... 39 2.4.4 Thermodynamic Ice Growth..........................................................................................................40 2.4.4.1 Simplified Models of Sea Ice Growth.............................................................................. 41 2.4.4.2 Effect of Snow On Sea Ice............................................................................................... 45 2.4.4.3 Effect of Oceanic Heat Flux ............................................................................................ 46 2.4.4.4 Effect of Surface Ablation ................................................................................................ 46 2.5 Processes in Ice......................................................................................................................................... 47 2.5.1 Compositional (Constitutional) Supercooling At the Ice–Water Interface ......................................50 2.5.2 Dendritic Ice–Water Interface and Entrapment of Brine Within Sea Ice........................................ 51 2.5.3 Grains and Subgrains In Sea Ice ....................................................................................................53 2.5.4 Brine Pockets Formation, Contents and Distribution In Sea Ice..................................................... 54

vii

viii

CONTENTS

2.5.5 Salinity Loss During Sea Ice Growth ............................................................................................. 58 2.5.5.1 Initial Rapid Salt Rejection At the Ice–Water Interface.................................................. 59 2.5.5.2 Subsequent Slow Salt Rejection from the Bulk Ice.......................................................... 61 2.6 Ice Deformation ........................................................................................................................................ 67 2.6.1 Rafting of Thin Ice ......................................................................................................................... 69 2.6.2 Ridging of Thick Ice.......................................................................................................................70 2.6.3 Formation of Ice Rubble Field ....................................................................................................... 73 2.6.4 Fractures in Ice Cover.................................................................................................................... 74 2.7 Ice Decay and Aging ................................................................................................................................ 76 2.7.1 Ice Decay ...................................................................................................................................... 76 2.7.2 Ice Aging ....................................................................................................................................... 80 2.8 Sea Ice Classes.......................................................................................................................................... 84 2.9 Sea Ice Regimes ........................................................................................................................................ 85 2.9.1 Polynyas ........................................................................................................................................86 2.9.2 Pancake Ice Regime ......................................................................................................................90 2.9.3 Marginal Ice Zone and Ice Edge....................................................................................................92 2.9.3.1 Marginal Ice Zone........................................................................................................... 93 2.9.3.2 Ice Edge........................................................................................................................... 94 2.9.4 Ice of Glacier Origin......................................................................................................................95 2.10 References................................................................................................................................................. 99 3. Sea Ice Properties: Data and Derivations ....................................................................................................... 107 3.1 Typical Values of Sea Ice and Snow Physical Parameters ......................................................................107 3.2 Temperature Profiles in Ice and Snow .................................................................................................... 108 3.3 Bulk Salinity and Salinity Profile ............................................................................................................. 113 3.3.1 Bulk Salinity.................................................................................................................................115 3.3.2 Salinity profiles ............................................................................................................................116 3.4 Density of First-Year and Multi-Year Ice .................................................................................................121 3.5 Volume Fraction of Sea Ice Constituents ................................................................................................ 123 3.5.1 Brine Volume Fraction.................................................................................................................123 3.5.2 Solid Salt Volume Fraction ..........................................................................................................124 3.5.3 Pure Ice Volume Fraction ............................................................................................................124 3.5.4 Air Volume Fraction ....................................................................................................................124 3.5.5 Temperature Dependence of Volume Fractions of Different Components ..................................125 3.6 Thermal Properties ..................................................................................................................................126 3.6.1 Thermal Conductivity of Sea Ice..................................................................................................126 3.6.2 Thermal Conductivity of Snow ....................................................................................................129 3.6.3 Specific Heat of Sea Ice...............................................................................................................131 3.6.4 Latent Heat of Sea Ice..................................................................................................................133 3.7 Dielectric Properties ............................................................................................................................... 134 3.7.1 Dielectric Constant of Brine.........................................................................................................136 3.7.2 Dielectric Mixing Models ............................................................................................................136 3.7.3 Field Measurements of Dielectric Constant .................................................................................142 3.8 References...............................................................................................................................................146 4. Laboratory Techniques for Revealing the Structure of Polycrystalline Ice..................................................... 149 4.1 Relevant Optical Properties ....................................................................................................................151 4.1.1 Polarized Light.............................................................................................................................151 4.1.2 Birefringence or Double Refraction of Ordinary (Ih) Ice.......................................................... 153 4.1.3 Optical Retardation......................................................................................................................155 4.1.4 Interference Colors for White Light..............................................................................................157 4.2 Ice Thin Sectioning Techniques .............................................................................................................. 158 4.2.1 Hot-plate Techniques for Thin Sectioning of Ice .........................................................................159 4.2.2 Double-Microtoming Technique for Thin Sectioning of Ice ........................................................159

CONTENTS

4.3

4.4

4.5

ix

4.2.3 Double-Microtoming Technique for Thin Sectioning of Snow ....................................................161 4.2.4 Precautions for Thin Sectioning by DMT.....................................................................................163 4.2.5 Optimum Thickness for Thin Sections of Ice and Snow .......................................................... 163 Viewing and Photographing Ice Thin Sections ....................................................................................... 164 4.3.1 Laboratory and Hand-Held Polariscope ......................................................................................165 4.3.2 Cross-Polarized versus Parallel-Polarized Light Viewing .............................................................168 4.3.3 Scattered Light and Combined Cross-Polarized/Scattered Light Viewing ....................................169 4.3.4 Circularly Polarized Light and Rapid Crystallographic Analysis ..................................................172 Advanced Techniques for Revealing Fine Crystallographic Microstructural Features............................. 173 4.4.1 Sublimation of Ice and Sublimation Pits ......................................................................................173 4.4.2 Etching Processes.........................................................................................................................176 4.4.2.1 Thermal Etching of Microtomed Ice Surfaces.......................................................... 179 4.4.2.2 Chemical Etching and Replicating Ice Surfaces ............................................................ 183 References...............................................................................................................................................188

5. Polycrystalline Ice Structure ........................................................................................................................... 191 5.1 Terms and Definitions Relevant to Polycrystalline Ice ............................................................................192 5.1.1 Special Thermal State of Natural Ice ...........................................................................................192 5.1.2 General Terms for Structural Aspects of Ice.................................................................................193 5.1.3 Basic Terms and Definitions ........................................................................................................194 5.2 Morphology of Ice ..................................................................................................................................197 5.2.1 Forms of Ice Crystals....................................................................................................................197 5.2.2 Miller Indices for Hexagonal Ice .................................................................................................198 5.2.3 Growth Direction of Ice Crystals .................................................................................................199 5.2.4 Ice Density in Relation to Crystalline Structure ...........................................................................199 5.3 Structure- and Texture-Based Crystalline Classification of Natural Ice ................................................... 200 5.3.1 Freshwater Ice Classification of Michel and Ramseier.................................................................200 5.3.2 Extending Crystallographic Classification of Freshwater Ice to Sea Ice .......................................202 5.3.3 Crystallographic Classes of Natural Ice........................................................................................203 5.3.3.1 Granular or Snow Ice (T1 Ice) ....................................................................................... 203 5.3.3.2 Randomly Oriented (S4) and Vertically Oriented (S5) Frazil Ice................................... 204 5.3.3.3 Columnar-Grained with c Axis Vertical (S1) Ice ........................................................... 205 5.3.3.4 Columnar-Grained with c Axis Horizontal and Random (S2 Ice) ................................. 207 5.3.3.5 Columnar-Grained Ice with c Axis Horizontal and Oriented (S3 Ice)........................... 211 5.3.3.6 Agglomerate Ice with Discontinuous Columnar-Grained (R Type Ice) ......................... 211 5.3.3.7 Ice of Land-Based Origin .............................................................................................. 212 5.3.3.8 Platelet Sea Ice.............................................................................................................. 213 5.3.4 Stereographical Projection (Fabric Diagram) of Natural Polycrystalline Ice ................................214 5.4 Examples of Crystallographic Structure of Natural Sea Ice ..................................................................... 216 5.4.1 Crystallographic Structure of Seasonal Sea Ice ............................................................................217 5.4.1.1 Frazil Ice (S5 Type) ....................................................................................................... 217 5.4.1.2 Columnar-Grained Ice (S3 Type)................................................................................... 218 5.4.1.3 Agglomeration of Various Crystallographic Structures .................................................. 220 5.4.1.4 Air Entrapment in Seasonal Ice ..................................................................................... 220 5.4.2 Crystallographic Structure of Perennial Sea Ice ...........................................................................221 5.4.2.1 Hummock Ice ............................................................................................................... 223 5.4.2.2 Melt Pond Ice................................................................................................................ 226 5.5 Biomass Accumulation at the Bottom of the Ice .....................................................................................230 5.6 Information Contents in Polycrystalline Ice Structure ............................................................................. 232 5.6.1 Geometric Characteristics of Crystalline Structure.......................................................................232 5.6.2 Geometric Characteristics of Brine Pockets in First-Year Ice .......................................................236 5.6.3 Geometric Characteristics of Air Bubbles ....................................................................................242 5.7 References...............................................................................................................................................244

x

CONTENTS

6. Major Field Expeditions to Study Sea Ice........................................................................................................ 249 6.1 The Arctic Ice Dynamic Joint Experiment (AIDJEX) ................................................................................250 6.2 Mould Bay Experiments 1981–1984: Stories that Were Never Told....................................................... 252 6.2.1 Site, Resources, and Logistics ......................................................................................................252 6.2.2 Sea Ice Conditions .......................................................................................................................254 6.2.3 Aging of Sea Ice: from FYI to MYI ...............................................................................................259 6.2.4 Interface Between Old and New Ice in Second-Year Ice Profile .................................................260 6.3 High Arctic Experience with Ice of Land Origin .....................................................................................262 6.3.1 Ward Hunt Ice Shelf and Hobson’s Choice Ice Island Experiment..............................................262 6.3.2 Multi-Year Rubble Field Around the Ice Island............................................................................265 6.4 Labrador Ice Margin Experiment (LIMEX) ............................................................................................... 266 6.5 Sea Ice Monitoring and Modeling Site (SIMMS) Program .......................................................................268 6.6 The Surface Heat Budget of Arctic Ocean (SHEBA)................................................................................270 6.7 The Norwegian Young Sea Ice Experiment (N-ICE) ................................................................................272 6.8 Marginal Ice Zone (MIZ) Experiments.....................................................................................................274 6.9 Ice Exercise by Us Navy .........................................................................................................................277 6.10 The Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC) ..........................278 6.11 References...............................................................................................................................................280 7. Remote Sensing Fundamentals Relevant to Sea Ice ........................................................................................ 283 7.1 General Principles of Satellite Remote Sensing....................................................................................... 284 7.2 Electromagnetic Wave Properties and Processes ....................................................................................289 7.2.1 Polarization of EM Wave .............................................................................................................290 7.2.2 Reflection, Transmission, Absorption, Scattering, and Emission ..................................................292 7.2.2.1 Reflection and Fresnel Model ....................................................................................... 293 7.2.2.2 Transmission ................................................................................................................. 295 7.2.2.3 Absorption and Scattering Losses .................................................................................. 296 7.2.2.4 Emitted Radiation (Re-radiation).................................................................................... 296 7.2.3 Brightness Temperature and Emissivity........................................................................................297 7.2.4 Penetration Depth........................................................................................................................299 7.3 Optical Sensing....................................................................................................................................... 300 7.4 Thermal Infrared Sensing ........................................................................................................................303 7.5 Microwave Remote Sensing....................................................................................................................305 7.6 Imaging Radar Sensing............................................................................................................................308 7.6.1 Imaging Radar Principles .............................................................................................................308 7.6.1.1 Radar Equations and Spatial Resolutions of RAR and SAR............................................ 309 7.6.1.2 Coherency and Polarization of Radar Signals .................................................................... 311 7.6.1.3 Radar Scattering Mechanisms ....................................................................................... 312 7.6.2 Multichannel SAR ........................................................................................................................313 7.6.3 SAR Polarimetry: Formulation and Derived Parameters ................................................................... 315 7.6.3.1 Formulation of Polarimetric Measurements................................................................... 316 7.6.3.2 Polarimetric Parameters Derived from the FP SAR Data ............................................... 317 7.6.3.3 Linking Radar Scattering Mechanisms to Ice Features .................................................. 320 7.6.3.4 Age-Based versus SAR-Based and Scattering-Based Sea Ice Classification ........................ 321 7.7 Scatterometer Systems.............................................................................................................................322 7.8 Altimeter Systems.................................................................................................................................... 323 7.9 Radiative Processes in Relevant Media................................................................................................... 325 7.9.1 Atmospheric Influences ...............................................................................................................325 7.9.1.1 Influences of Atmosphere on Optical and Infrared Observations ................................. 325 7.9.1.2 Atmospheric Correction for Passive Microwave Observations...................................... 328 7.9.2 Seawater ......................................................................................................................................330 7.9.2.1 Seawater in the Optical and Thermal Infrared Data...................................................... 330 7.9.2.2 Seawater in the Microwave Data .................................................................................. 331

CONTENTS

xi

7.9.3

Snow on Sea Ice: Physical and Radiative Processes ................................................................333 7.9.3.1 Snow in Optical and Thermal Infrared Data ............................................................335 7.9.3.2 Snow in the Microwave Data...................................................................................336 7.10 References....................................................................................................................................341 8. Satellite Sensors for Sea Ice Monitoring....................................................................................................... 349 8.1 Historical Synopsis of Remote Sensing Satellites for Sea Ice.................................................................349 8.2 Optical and Thermal Infrared Sensors...................................................................................................352 8.3 Modern Passive Microwave Sensors .....................................................................................................353 8.4 Modern Imaging Radar Sensors ............................................................................................................355 8.5 Scatterometer Sensors ...........................................................................................................................358 8.6 Altimeter Sensors ..................................................................................................................................359 8.7 References.............................................................................................................................................360 9. Radiometric and Scattering Observations from Sea Ice, Water, and Snow................................................. 363 9.1 Optical Reflectance and Albedo Data ..................................................................................................364 9.2 Microwave Brightness Temperature Data .............................................................................................370 9.3 Radar Backscatter .................................................................................................................................376 9.3.1 Backscatter Databases from Single-Channel SAR .....................................................................378 9.3.2 Dual Polarization Data .............................................................................................................384 9.3.3 Fully Polarimetric Data .............................................................................................................387 9.4 Emissivity Data in the Microwave Bands ..............................................................................................395 9.5 Microwave Penetration Depth ..............................................................................................................403 9.6 References.............................................................................................................................................407 10. Retrieval of Sea Ice Surface Information...................................................................................................... 411 10.1 Mechanically Generated Surface Deformation .....................................................................................412 10.1.1 Rafted Ice.................................................................................................................................412 10.1.2 Ridged, Rubble, and Brash Ice.................................................................................................413 10.1.3 Kinematic Processes: Convergence, Divergence, Shear, and Vorticity ....................................417 10.1.4 Cracks and Leads.....................................................................................................................421 10.2 Thermally Induced Surface Features .....................................................................................................428 10.2.1 Surface Melt.............................................................................................................................428 10.2.1.1 Optical Observations ...............................................................................................428 10.2.1.2 Passive Microwave Observations .............................................................................432 10.2.1.3 Active Microwave Observations ..............................................................................434 10.2.1.4 Airborne Photography ..............................................................................................437 10.2.2 Frost Flowers............................................................................................................................438 10.3 Meteorologically Driven Surface Features ............................................................................................442 10.3.1 Polynya Identification and Properties ......................................................................................442 10.3.2 Snow Depth .............................................................................................................................444 10.4 References.............................................................................................................................................448 11. Retrieval of Sea Ice Geophysical Parameters ............................................................................................... 453 11.1 Sea Ice Type Classification ...................................................................................................................454 11.1.1 Ice Classification from Optical and TIR Systems......................................................................456 11.1.2 Ice Classification from Passive Microwave Data .....................................................................457 11.1.3 Ice Classification from SAR......................................................................................................458 11.1.3.1 Ice Classification from Single-Channel SAR .............................................................460 11.1.3.2 Ice Classification from Dual-Channel SAR ...............................................................461 11.1.3.3 Ice Classification from Polarimetric SAR Data .........................................................467 11.2 Sea Ice Concentration...........................................................................................................................471 11.2.1 Ice Concentration from Optical and TIR Images......................................................................472 11.2.2 Ice Concentration from Coarse-Resolution Microwave Observations......................................473 11.2.2.1 NASA TEAM (NT) Algorithm ....................................................................................475

xii

CONTENTS

11.3 11.4

11.5

11.6 11.7

11.8

11.2.2.2 The Enhanced NASA Team (NT2) Algorithm ...........................................................476 11.2.2.3 The ASI Algorithm ....................................................................................................478 11.2.2.4 ECICE Algorithm.......................................................................................................479 11.2.2.5 Intercomparison of PM Algorithms...........................................................................486 11.2.2.6 Sources of Error and Sensitivity of Ice Concentration Algorithms ............................490 11.2.2.7 Assessment of Ice Concentration Results Against Ice Charts ....................................493 11.2.3 Ice Concentration from Fine-Resolution SAR...........................................................................496 Sea Ice Extent and Area ........................................................................................................................498 Sea Ice Thickness (SIT)..........................................................................................................................501 11.4.1 SIT from TIR Observations .......................................................................................................503 11.4.2 SIT from PM Observations .......................................................................................................506 11.4.3 SIT from Altimeter Observations ..............................................................................................510 11.4.4 SIT from SAR Observations......................................................................................................514 Ice Surface Temperature (IST) ...............................................................................................................517 11.5.1 IST from TIR Observations .......................................................................................................517 11.5.2 IST from PM Observations .......................................................................................................520 Sea Ice Age ...........................................................................................................................................522 Sea Ice Motion and Kinematics ............................................................................................................524 11.7.1 Methods of Ice Motion Tracking..............................................................................................526 11.7.1.1 Motion Tracking Using Image Features....................................................................526 11.7.1.2 Motion Tracking Using Individual Sea Ice Floes ......................................................528 11.7.2 Operational Ice Motion Products.............................................................................................532 References.............................................................................................................................................533

12. Modeling Microwave Emission and Scattering from Snow-Covered Sea Ice............................................... 541 By Rasmus Tage Tonboe 12.1 The Need for Modeling Microwave Emission and Scattering from Snow-Covered Sea Ice ..................541 12.1.1 The ECMWF Workshop and Large-Scale Sea Ice Modeling ....................................................542 12.1.2 Gross Features of Forward Models...........................................................................................542 12.2 Radiative Transfer and Modeling Approaches for Sea Ice Thermal Microwave Emission.....................543 12.2.1 Dense Media Volume Scattering .............................................................................................543 12.2.2 Sea Ice Emission Models..........................................................................................................543 12.2.3 Sea Ice Backscatter Models for Level Ice .................................................................................544 12.2.4 Sea Ice Backscatter Models for Ridged Ice ..............................................................................545 12.3 The Input to a Forward Model ..............................................................................................................545 12.3.1 Primary Input Parameters .........................................................................................................545 12.3.2 Secondary Input Parameters ....................................................................................................546 12.3.3 Tertiary Input Parameters, Volume, and Surface Scattering .....................................................546 12.4 Example of the Implementation of an Altimeter Model to Study the Impact of Saline Snow on the Backscatter............................................................................................................................................547 12.5 Example of Combining Atmospheric, Ocean, and Sea Ice Emission Models to Simulate the Noise in Sea Ice Concentration Estimates .............................................................................................548 12.5.1 Snow in the Emission Models ..................................................................................................549 12.5.2 The Combined Sea Ice Thermodynamic, Atmospheric, Ocean, and Sea Ice Emission Models .....................................................................................................................................549 12.6 Inverse Modeling ..................................................................................................................................552 12.7 References.............................................................................................................................................553 13. Impacts of Climate Change on Polar Ice...................................................................................................... 557 13.1 The Inconvenient Truth of Global Warming: How is it Manifested in The Polar Region? ....................560 13.2 Sea Ice Regimes in the Two Polar Regions ...........................................................................................562 13.2.1 Geographic Differences Between the Two Polar Regions and Their Impacts on Sea Ice ........562 13.2.2 Differences in Sea Ice Characteristics Between the Two Polar Regions...................................564 13.3 Changes of Polar Sea Ice in Response to Global Warming ..................................................................565 13.3.1 The Arctic and Antarctic Ice Extent .........................................................................................565

CONTENTS

xiii

13.3.2 The Arctic and Antarctic Ice Thickness and Volume...................................................................569 13.3.3 The Arctic Sea Ice Age.............................................................................................................572 13.3.4 The Arctic Sea Ice Dynamics...................................................................................................575 13.3.5 The Antarctic Icebergs .............................................................................................................576 13.4 Coupling Between Polar Sea Ice and Environmental Factors................................................................577 13.4.1 Interaction of the Arctic Sea Ice with the Environment............................................................578 13.4.1.1 Atmospheric Factors that Contribute to Changes in the Arctic Sea Ice ....................578 13.4.1.2 Enhanced Arctic Warming due to Changes of Sea Ice Cover ..................................578 13.4.1.3 Arctic Warming due to Sea Ice Advection Out of the Arctic Basin..........................580 13.4.1.4 Interaction of the Arctic Sea Ice with Wind............................................................................582 13.4.1.5 Mutual Interactions Between the Arctic Sea Ice Cover and Oceanic Forcing..........584 13.4.2 Interaction of the Antarctic Sea Ice with the Environment.......................................................585 13.4.2.1 Interaction of the Antarctic Sea ice with Atmospheric Factors .................................585 13.4.2.2 Interaction of the Antarctic Sea Ice with Oceanic Forcing.......................................586 13.4.2.3 Interaction Between the Antarctic Sea Ice, Ice Shelves, and Icebergs......................587 13.5 References.............................................................................................................................................590 Index...................................................................................................................................................................... 595

PREFACE During the winter months, sea ice plays the most important role in the daily life of coastal communities of the cold regions of the earth. As for the Arctic regions, especially Canada and Russia, the recent trend of sea ice retreat under global warming is leading to new opportunities for the extension of navigational season and exploitation of natural resources. Growing interest in sea ice for socioeconomic purposes has also increased awareness of its role in affecting the global environment, the climate system, and most importantly the human aspects. The advent of remote sensing techniques from Earth-observing satellites, as the major tool for obtaining global record of sea ice, is making it necessary to re-examine sea ice as a material with updated knowledge and approaches. In the customary description of snow, fresh-, and seaice, often the basic sciences and measurement techniques used are assumed to be known yet not covered or explained in details. These are mostly related to a few physical aspects: (1) the thermodynamic high-temperature state (being close to the melting point) of these materials, (2) physics of solidification governing the development of microstructural features especially in sea ice, (3) the impact of mobility and ageing of sea ice on its geometrical and physical properties, and (4) the birefringent properties of ice crystals that are used extensively for revealing internal substructures. By the same token, the customary description of the applications of remote sensing usually leaves out details of the electromagnetic wave interaction with ice and its snow cover, particularly the impact of the snow on the measured reflection, radiation, and radar scattering. This book is written with the hope of filling these gaps and to making a bridge between the physics and the remote sensing communities. This is spot-on whether the readers are working on sea ice in the two primary polar cryosphere or on snow/ice of the secondary cryosphere of the Himalayas and other mountain ranges. The advent of satellite remote sensing and its applications to sea ice, in the 1970s, fulfilled a wish by some of the pioneers of ice research in the 1950s. That was to develop instruments suited for recording the evolution and dynamics of sea ice in the polar environment. Space-borne sensors operating in the visible, infrared, and passive microwave emerged but it wasn’t until the introduction of the Synthetic Aperture Radar (SAR), in 1978, onboard Seasat when the sea ice community realized how revolutionary this tool would be for ice reconnaissance and parameter retrieval at fine resolutions.

Work started in the 1980s to study the feasibility of developing space-borne imaging radar systems. The work bore fruits in the 1990s, when the European the Europe Remote Sensing Satellite (ERS) and the Canadian RADARSAT-1 were completed. Data from those satellites and a bunch of earlier satellite-borne visible, thermal infrared, and passive microwave sensors opened many opportunities for individuals from different disciplines to join the interesting field of remote sensing of sea ice. Individuals from the ice physics and mechanics disciplines also came forward to link their knowledge to remote sensing observations. The first author of this book considers himself lucky when opportunities emerged for him in Environment Canada, starting in 1988, to get involved in many projects to demonstrate applications of imaging radar data in support of the Canadian sea ice monitoring program. Later he expanded his experience by developing applications using microwave (active and passive) and other remote sensing data categories. Opportunities also emerged for him to participate in field expeditions in the Arctic and the eastern seas of Canada and be in direct touch with the fascinating world of sea ice. The second author has spent long career in scientific research of high temperature material, of which ice has been at the core of his research. He has contributed to development of advanced knowledge on crystallographic structure of natural ice. He also considers himself lucky when opportunities emerged for him to get involved in many Arctic R&D activities after he joined the National Research Council of Canada in January, 1975. Both authors combined their experience in sea ice physics and remote sensing to produce this book. The idea of the book started in 2010 when the authors felt the importance of combining knowledge in these two fields in a single publication. They were also keen to present their unpublished data collected and analyses conducted over many years of collaborative work. In many ways this is an entirely new book on natural floating ice in oceans with an emphasis on sea ice. The goal is to describe and explain the principles of physics and chemistry of sea ice and its space-borne observation using a suite of remote sensing imaging systems. The reader needs only to flip through the pages to note that the general appearance related to the format and illustrations are different from any technical book. The subjects are treated with a multidisciplinary and, to some extent, transdisciplinary approach. Underlining assumptions

xv

xvi

PREFACE

are that the reader/user will come with wide ranging engineering backgrounds—civil, mechanical, environmental, electrical/electronic, etc. or sciences—physics, chemistry, geography, oceanography, climatology, and even social sciences and anthropology related to human settlements. The style used, therefore, is to describe complex subjects in an understandable way so that the reader will not only gain a working knowledge of basic science related to the area of study, but also applications of ice information retrievals using remote sensing tools. It is also expected that the reader will cultivate an appreciation of the human aspects required to carry out the investigations and, of course, the real-life operational issues encountered in shipping, fishing, etc. This book is intended to summarize experiences acquired by international researchers and operational sea ice communities. Focus is placed on experiences gained in Canada, of which its northern sea water is covered by about one million square kilometers of sea ice in winter on an average. The book is also intended to reach out to a variety of sea ice audiences interested in different aspects of ice physics, mechanics, remote sensing, operational monitoring, climatic impacts, etc. Information covers sea ice processes operating at a range of spatial scales from micro- and macro-scales (such as brine entrapment in sea ice within high energy paths of crystallographic structure) all the way to the synoptic scale of ice motion across the Arctic-wide domain. The first edition of this book was published in June 2015. It contained ten chapters in addition to the Introduction: five on ice physics, four on remote sensing of sea ice, and one providing a historical account of the national ice monitoring service in Canada. In this new edition some of the content and layout has remained the same, while significant changes have also been made. Chapters are not free-standing. Instead, links are established as much as possible between information about the different themes of the chapters. More melding of physics and remote sensing has been included.

In this edition, the chapter on the Canadian ice monitoring is removed while three new chapters have been added. After the Introduction (Chapter 1), Chapters 2 to 6 constitute the ice physics part of the book. They present key physical properties and processes suitable to anyone with a background in physics and mathematics. However, some topics may be considered advanced. Chapter 4 on laboratory techniques for revealing the polycrystalline structure of ice has been rewritten with inclusions of more microphotographs to reveal fine features of ice microstructure. Chapter 6 on major field expeditions has been expanded to include information on international expeditions rather than focusing on Canadian field programs. In some sections of the ice physics part the presentation has been made more detailed and coherent to clarify new concepts. This includes entrapment of inclusions at subgrain boundaries and the etching and replica techniques that reveal microdetails of sea ice crystallographic structure. Chapters 7 to 11 constitute the remote sensing part of the book. They are devoted to remote sensing basic principles, platforms, and most importantly retrieval of ice and snow parameters from active and passive imaging sensors, with emphasis on microwave systems. New material is introduced in this part to capture advancements in new satellite sensor technologies and analysis methods. Chapter 8 is a new chapter that covers the suite of satellite sensors most useful for sea ice and snow monitoring. Chapter 12 is another new chapter (written by Rasmus Tonboe), which covers radiative transfer approaches in modeling microwave emission and scattering from snow-covered sea ice. Chapter 13 (the third new chapter) addresses key differences between the responses of sea ice to climate change in the two polar regions, with links to their different geographic and environmental factors. Mohammed Shokr and Nirmal K. Sinha 22 December 2022

ACKNOWLEDGMENTS AND RECOGNITIONS When the first author started his collaboration with the Canadian Ice Service (CIS) in 1988 as a scientist employed by then Environment Canada (now Environment and Climate Change Canada), the staff was very supportive and welcoming of his input in processing the remote sensing images of sea ice. More importantly, they granted him opportunities to learn about the operational ice monitoring program. Part of the learning process was participation in a “round robin” reconnaissance flight, which was an annual series of flights to generate single span shots of ice conditions across the eastern section of the Arctic. That was the author’s first observation of sea ice, though from some 10,000 feet above the surface. Shortly after, the author was involved for many years in field work on the Arctic and Labrador Sea ice to sample and measure ice and snow properties and crystalline structure. He would like to express his deepest appreciation to members of the CIS team who supported his efforts to develop knowledge about sea ice operational aspects: John Falkingham, Bruce Ramsay, Terry Mullane, Mike Manore, Dean Flett, Matt Arkett, and Dr. Roger De Abreu. Ken Asmus offered great help in field and laboratory experiments. The author would also like to express his sincere thanks to the team of scientists and support staff who supported his work during several field expeditions. The generous support of many individuals led the author to develop not only knowledge but passion for sea ice and the cold region environment. For that, he would like to acknowledge the support of the Late Dr. David Barber of the University of Manitoba, and extend thanks to Dr. Simon Prinsenberg of the Department of Fisheries and Ocean Canada, Drs. Garry Timco and Michelle Johnston of the National Research Council of Canada and all members of the field expeditions with whom he shared work and experience during the decade of the 1990s. Dr. Venkata Neralla and Mr. Roop Lalbeharry of Environment Canada inspired the first author to complete this book. Dr. Shawn Turner of Environment Canada reviewed parts of the remote sensing material and Dr. Walter Meier of the National Snow and Ice Data Center (NSIDC, USA) kindly reviewed Chapter 13 of this edition. The second author would like to convey his sincere appreciation to innumerable persons who helped him in the field and to recognize the financial, technical, and strategic supports provided by several Canadian and International organizations since 1975. George Hobson, Director of

Polar Continental Shelf Project (PCSP) in Canada from 1972 until 1988, provided moral, financial and strategic support to the author for almost ten years (1977–1986). This allowed him to train an Inuit Team on carrying out the year-round investigations on growth, structure, and engineering properties of sea ice at Eclipse Sound near Pond Inlet in Baffin Island. Special thanks are due to Mr. M. Komangapik, Mr. S. Koonark, and Mr. S. Koonoo for their efforts in collecting scientific data in the harsh climatic conditions of the Arctic and to Ms. D. Komangapic, Ms. S. Akoomalik, and Ms. J. Arnakallak for performing tedious measurements in the Pond Inlet field laboratory and tabulating the results, and to Ms. I. Kilukishak and Mr. M. Komangapik for laboratory assistance in Mould Bay. The author also wishes to express his sincere thanks to Mr. Hermann A.R. Steltnar and his wife Mrs. Sophie Steltnar of Arctic Research Establishment (ARE) at Pond Inlet, for organizing and managing the field activities of the Inuit team and compiling the field data. For studies related to multi-year (MY) sea-ice rubble field at the “Hobson’s Choice” Ice Island, the author is obligated to Mike Schmidt, the PCSP camp manager for providing all the logistical support and to Drs. Ed Stander and Paul Barrette, respectively, for their assistances in conducting the microstructural analysis of MY sea ice and Ward Hunt shelf ice. Drs. Lorne Gold, Robert (Bob) Frederking, Robert Gagnon, Garry Timco, Mohammed Sayed, and Mary Williams of NRC provided encouragements, guidance, and useful critical comments for more than three decades. In fact, Bob Frederking introduced the author to the world of sea-ice engineering. He led the author, literally by hand, to his first field experience at Strathcona Sound, Baffin Island during the extremely cold polar environmental conditions of 24-hours dark “polar nights” of November–December, 1975. Due to the unavailability of any remotely sensed images, the only view of sea ice the author had during this mission was whatever was visible under the illumination of headlights of trucks positioned on the newly constructed dock. The truck had to be turned to get different points of view. Ice cores were taken by suitably pointing the headlights to the spot for sampling. Very special regards go to Dave Wright of the Institute for Research in Construction of NRC who actually taught the author the refined techniques of “hot-plate melting” for making near-perfect thin sections of “fresh-water ice”, fabricating a small field polariscope with built-in

xvii

xviii

ACKNOWLEDGMENTS AND RECOGNITIONS

light source and working with him during the first few, but crucial field trips in the High Arctic in 1976–1977. Ron Jerome of the same institute provided technical assistance in almost all phases of laboratory and field investigations. This included the design of the large-field polariscope, 100-mm-diameter “feather-weight” fiber-glass ice-core auger, the NRC borehole indentor (BHI) and the portable cold laboratory used at Pond Inlet, Resolute and on board of icebreakers. The author is indebted to Donald Martin and the staff of the National Capital Commission (NCC) of Ottawa for securing the entire Dow’s Lake and providing technical assistance during the days of testing for the evaluation (and modifications) of these tools over several years, before they were taken to the Arctic. During the Mould Bay experiments of 1981–1985, the following persons were involved and provided assistance to the author: Dr. Rene’ O. Ramseier, Ken W. Asmus, Larry B. Solar, Doug Hagen, and David M., Susan Digby and Dr. Chuck Livingstone of CCRS, EMR; David Lap and Catherine Bjerkelund; Drs. Tom Grenfell and Dr. Donald K. Perovich, and Arne Hanson. The author is also grateful to Drs. Michelle Johnston and Caizhao Zhan for their undivided attentions as PhD candidates during the many years of field experiments.

On the subject of extending the double microtoming technique, originally developed for sea ice to investigate the microstructural aspects of snow morphology, the author is thankful to Dr. Pramod Satyawali and Maj. Gen. Satya Sharma of Snow and Avalanche Study Establishment (SASE), India. He is also indebted to Dr. Alex Klein-Paste of the Norwegian University of Science and Technology (NTNU), Trondheim and Mr. Armann Norheim of the Norwegian Civil Aviation (AVNR), Oslo for refining the methods and applying to snow on runways. Vance Garbowski and Howard Massecar, helicopter pilot of Vancouver-based Quasar provided field supports in Mould Bay and Baffin Bay. The author is indebted to Howard for rescuing and saving his life, on 22 June 1984, in northern Baffin Bay during a tsunami storm that shattered the 1.5-m-thick sea ice sheet and forced the author, David Lapp, and an Inuit polar-bear watcher to take shelter on a bergy bit. Otherwise, the tales of this author with respects to the human aspects of Arctic research would not have been told in this book. Mohammed Shokr Nirmal K. Sinha December 2022

1 Introduction

1.1 Background ................................................................................ 1

Sea Ice in Marine Navigation........................................ 17 Sea Ice and Offshore Structures..................................... 19 Sea Ice as A Transportation Platform ........................... 20 Sea Ice in Relation to Solid Earth Sciences: Rocks and Plate Tectonics........................................................ 21 1.5 Sea Ice and Remote Sensing ..................................................... 22 1.4.6 1.4.7 1.4.8 1.4.9

1.2 Canada and the Arctic: Historical and Community Synopsis..... 4 1.3 The Fascinating Nature of Sea Ice ............................................. 8 1.4 Sea Ice in Research and Operational Disciplines...................... 12 1.4.1 Sea Ice in Physics........................................................... 12 1.4.2 Sea Ice in Climatology................................................... 13 1.4.3 Sea Ice in Meteorology .................................................. 14 1.4.4 Sea Ice in Oceanography ............................................... 15 1.4.5 Sea Ice in Marine Biology ............................................. 16

1.6 Motivation for the Book Writing ............................................. 24 1.7 Organization of the Book ......................................................... 25 1.8 References................................................................................. 27

1.1. BACKGROUND

the variations in the boundaries of the polar region are very small and negligible. The secondary cryospheric regions include the Alps, Andes, Himalayas, Rockies, etc. Among the secondary cryospheric regions, the Himalayan belt covers and affects the largest effective area of human habitation. Climate change has been affecting all the cryospheric regions of the world, and the effects can be directly observed and quantified using airborne and space-borne remote sensing as well as land-based instruments. Remotely sensed images of the land- and ocean-based snow and ice information are paramount in understanding the state of health of the earth for sustainability of life. Other methods such as ice core analysis of ice caps and ice shelves, composed mainly of snow in different phases, are also used. After all, snow is the messenger of the sky, and ice is the answer to the cold climate. Sea ice covers most of the oceanic surface of the primary cryospheric area of the global surface. While not noticeable by the majority of the population of our planet, sea ice observations provide a powerful tool for quantifying climate change and the health of our only home. The world of sea ice encompasses the polar region, particularly the Arctic basin and a belt around the continent of Antarctica. Out of the 71% of the earth’s surface that is covered by ocean, about 7%–15% is covered by sea ice at certain times

Our world is divided into five regions according to the position of the sun throughout the year: a tropical region around the equator, two temperate regions, and two polar regions. On two equinoxes, 21 March and 23 September, the sun is directly over the equator and the sun’s rays reach both the North and South poles. On 21 June (the Summer Solstice), the sun is directly over the Tropic of Cancer (about 23.5 N) in the northern temperate region, and on 22 December (the Winter Solstice) it is positioned directly over the Tropic of Capricorn (about 23.5 S) in the southern temperate region. In the two polar regions, mostly relevant to the material in this book, the sun never sets in their summer and never rises in their winter. The Arctic region (or zone) containing the North polar region (with latitudes greater than “about” 66.5 N) and the Antarctic region (or zone) containing the South polar region (having latitudes greater than “about” 66.5 S) are the primary cryospheric regions of the world. The 66.5 angle comes from the tilt of the earth’s rotation axis (23.5 ), such that 90 – 23.5 = 66.5 . Recall that cryosphere comprises all regions where water exists in solid form. Although the latitudes of the Arctic and Antarctic circles depend on the earth’s axial tilt, which fluctuates slightly with time (about 2 over a 40,000-year period),

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 1

2

SEA ICE

(more in the winter and less in the summer). That is equivalent to 5%–10% of the earth’s surface. About 37% of the total oceanic surface is covered by sea ice at one time or another. Sea ice area in the Arctic varies between a minimum of about 4 million km2 in September to a maximum of about 15 km2 in March. The corresponding figures for the Antarctic are 3 million and 18 million km2 in February and September, respectively. However, the maximum volume of the sea ice cover in the Arctic (about 0.05 million km3) is nearly twice the maximum volume in the Antarctic. This is because the mean thickness of sea ice is 3 m and 1.5 m in the Arctic and Antarctic, respectively. Sea ice can develop very smooth or very rough surfaces. It can be soft or hard, a bare-surface or snow-covered, stagnant (fastened to the shoreline) or mobile pack ice, stiff and silent or crushing with loud noise. It exhibits seasonal variations to which life in the polar regions is closely adapted. In the Arctic region, sea ice starts its growth in September/October and reaches its maximum in February/ March, when it covers the entire Arctic basin. This trend is reversed during the summer, and the ice extent reaches its minimum in September. In the Antarctic, the annual fluctuations range between a minimum in February to a maximum in August/September, when ice extends to latitudes between 55 –65 South. For a limited time during the summer months, certain areas of the polar waters in the Arctic zone are used extensively by ships (ice-strengthened or escorted by icebreakers) where the floating bodies of new and old sea ice and icebergs can prove hazardous. The expected reduction of sea ice extent, the reduction of the navigationally hazardous old ice, and the increase in the duration of summer melting season will certainly increase marine activities in these areas. No doubt, the Arctic waters, particularly the legendary Northwest Passage (NWP) that passes through the Canadian Arctic Archipelago (CAA) and the Beaufort Sea, will be used more in the future for shipping goods between Asia, North America, and Europe. Sea ice extent in both the Arctic and Antarctic averages the same, about 15 million km2, during winter. However, because the mean thickness of sea ice in the Arctic is larger, the maximum volume of sea ice cover in the Arctic (about 0.045 million km3) is nearly twice that of the Antarctic. In summer, ice extent shrinks significantly to about 50% of the winter coverage in the Arctic. Nearly 90% of the sea ice coverage disappears by the end of the summer in the Antarctic. Ice that melts completely during the summer is called “seasonal ice” or “annual ice.” If the ice melts only partially, then the part that survives until the next winter and growth season is called “perennial ice.” This can be second-year ice (SYI) or multi-year ice (MYI), depending on how many summers the ice has survived.

As a major component of the cryosphere, sea ice influences the global ocean and atmosphere in a profound manner. Its continuous interaction with the underlying oceans and the overlaying atmosphere leaves major impacts on weather, climate, and ocean current systems. Moreover, ice in one form or the other plays a significant role in the daily life of communities inhabiting the cold regions of the earth. Sea ice in particular influences the coastal areas in most of the circumpolar nations of the Northern Hemisphere. It affects, to a lesser extent, a few countries in the Southern Hemisphere. Of all the countries of the world, Canada has the longest coastline as well as the largest reservoir of fresh-water lakes and rivers with floating ice in them annually at least for half of the year. Except for Alaska, practically all the areas north of the 49 N in North America belong to Canada. While sea ice plays a major role in areas above 60 (north or south), it does not affect areas below that latitude except in the Hudson Bay, Labrador Sea, and the Gulf of St. Lawrence in Canada, and relatively speaking, to a lesser extent in the Baltic Sea, Gulfs of Bothnia and Fin in Europe, the Sea of Okhotsk, north of Japan, and Bohai Bay in China. Above about 35 N in Eurasia and North America, most of the streams, rivers, and lakes (e.g., Black Sea, Sea of Azo, Caspian Sea in Eurasia, and the Great Lakes in North America, to name a few among thousands) have some ice cover each winter. In fact, severity of winters in North America is often measured in terms of ice coverage of the five Great Lakes (Lake Superior, Lake Michigan, Lake Huron, Lake Erie, and Lake Ontario). In spite of the fact that sea ice covers vast areas of sea surface of the earth, most of the people of the primary cryospheric regions of the world have not seen it or are even aware of it. That is because most people, even within the cold regions of the earth, live far from the areas affected by sea ice. Other than a few thousand multinational scientific observers and a few annual visitors, nobody lives in the South Polar Zone (beyond the Antarctic Circle). Only a few small communities of the Falkland Islands and Argentina consider the Antarctic region their home. On the other hand, beyond the Arctic Circle in circumpolar areas of Alaska, Canada, Norway, and Russia, perhaps a few million people live. This is incomparable to the nearly 3200 million people living in Afghanistan, Bangladesh, China, India, Nepal, Pakistan, and Tibet, who are indirectly affected by the Himalayan cryosphere, but sea ice does not exist in those regions. It is not uncommon for people who live away from the circumpolar boundaries to be confused between sea ice and icebergs. Yet, general awareness about sea ice has been growing as public information about the decline of sea ice in the Arctic with its positive economic impacts and negative environmental impacts is spreading. This

INTRODUCTION 3

book, though not oriented to serve as a popular science document, provides scientific information with explanations that may hopefully expand the domain of interest in sea ice and attract a number of young scientists to pursue studies about its physical aspects as well as its detection using space-borne remote sensing technologies. The Arctic Basin consists of primarily the Canadian and the Eurasian subbasins (for details on these two basins, see Chapter 3 in Weeks, 2010). It is extremely difficult to obtain extensive sea ice data in these areas because of the remote locations and extreme climatic conditions in which ice exists. Until the beginning of the twentieth century information about sea ice was mainly gathered and used by the local people who lived in the subarctic regions. Later, increasing information was obtained from ship sighting and harbor icing records, but the purpose remained to assist the very limited number of marine operations. However, since the end of World War II in 1945 and the beginning of the Cold War, there had been a significant increase in human activity in both the polar regions, and in particular the Arctic. Numerous weather stations equipped to gather scientific information and military bases with airports and radar lines were constructed in Canada, Alaska, and Greenland. Although some of the supplies for the construction and maintenance of these bases were transported by aircraft, ice-strengthened ships escorted by icebreakers were extensively used during the summer melt seasons. Submarines and buoys have also been used to gather data for sea ice in the Arctic Ocean. Russia has a longer history of record on measurements of sea ice, dating back to several centuries. Russia (actually the Soviet Union) played an important role in increasing the awareness of the Arctic outside its boundaries. Scientific interest in Arctic sea ice grew fast during the Cold War era as nuclear submarines of the United States and Soviet Union used the Arctic Ocean basin as a prime area to launch ballistic missiles. Funds were made available by the United States for scientists to launch field studies in the Arctic for the first time in the 1950s. In fact, many US-based scientists together with their Canadian counterparts carried out their investigations on sea ice using facilities available in Canada. It is appropriate to mention here that the first English language book on physics of ice that covers significant sections on sea ice was written in Montreal, Canada by Pounder [1965] and incidentally, the first comprehensive English language book on the physics of glaciers, the source of icebergs, was also written in Ottawa, Canada by Paterson [1969]. The number of field studies dedicated to sea ice measurements in the Arctic peaked in the 1980s and 1990s. The work was motivated by the use of the space-borne remote sensing, particularly radar sensors. There was a need for on-sight field observations and measurements

of sea ice characteristics to validate interpretations of satellite images and support information retrieval from the images. The scope of field programs has broadened later as knowledge accumulated and more scientific questions have risen. Today, major scientific field campaigns in the Arctic are highly multidisciplinary, and a few campaigns used icebreaker platform frozen in the ice for a few months or full year to study the ice and snow seasonal cycle. This has become a more efficient way to collect a wide range of data and link physical, biological, and meteorological aspects of the ocean-ice-atmosphere system. It should be noted that field campaigns have produced wealth of information on now-covered sea ice but when the ice is thick enough to walk on safely. Ice thinner than 15 cm is not safe to sample while walking on, although this ice is probably the most important for weather and climate studies. It instigates the strongest interaction between the warm ocean and cold atmosphere in winter. The photograph in Figure 1.1 shows the procedure of cutting a sample of thin ice (about 50 mm thick) using a gangway descending from an icebreaker while the operator is attached to a harness. The ice surface appears to be covered with frost flowers. Obviously, information was lost during this sampling process due to brine drainage from the sample. Only crystallographic information could be preserved. Usually, thin ice can be studied in

Figure 1.1 Sampling of 0.05 m thick sea ice in the Baffin Bay in May 1998 by the author (Shokr) using a gangway lowered from an icebreaker (photo by K. Asmus, Canadian Ice Service).

4

SEA ICE

facilities of outdoor laboratories, such as the one at the Cold Region Research and Engineering Laboratory (CRREL) in Hanover, New Hampshire, USA, or University of Manitoba, Winnipeg, Canada. By mid-1990s (after the end of the Cold War), interest in Arctic sea ice shifted from being military-, security-, or offshore-industry-driven to being environmentally-driven. New concerns for the region are now comprised of issues such as environmental conservation, including nuclear waste and other pollution issues, protecting the livelihood of the Arctic’s inhabitants and species, and most importantly identifying sea ice as an indicator and result of climate change. As a result of Arctic sea ice being a strong indicator of climate change, its monitoring has triggered an increase in funds by many countries to conduct more research. The Arctic basin connects the Atlantic and Pacific oceans. Therefore, if the ice diminishes or is replaced by thinner (navigable) ice, marine navigation routes (Northwest or Northeast passage) may open. This potential scenario will have a great positive economic impact on Canada, Russia and all countries that will use these navigation routes. Antarctica is the land of glaciers and ice shelves from which icebergs calve and float within the surrounding sea ice. In fact, sea ice around Antarctica is home for numerous icebergs. Antarctic sea ice has not responded to the recent episode of climate change as much as the Arctic ice has. Nevertheless, the region has received attention lately because of the impact of climate change on glaciers and ice shelves. This impact has been manifested in the increasing number of icebergs calving from ice shelves. While numerous small icebergs have calved (and will continue to calve in the future), the number of major icebergs has also increased. The largest iceberg had recently calved from the Ross Ice Shelf in 2000. It was 295 km long and 37 km wide, larger than many small countries of the world. Mutual interaction between icebergs and the surrounding sea ice is the most important impact of climate change on Antarctic sea ice (section 13.4.2.3). Snow cover plays an important role in the thermodynamic evolution of sea ice. Although it accounts roughly for only 10% of snow/ice volume, its properties differ sharply from their equivalent properties of sea ice. Two familiar examples are the albedo and thermal conductivity. The albedo of dry snow is above 0.9, which is much higher than the albedo of bare first-year sea ice surface (about 0.52). Increased albedo allows less sunlight to penetrate the surface. Therefore, snow-covered ice and the underlying seawater receive less sunlight. Equally important, the thermal conductivity of the snow is one order of magnitude less than that of sea ice. This means that snow can thermally insulate sea ice and slow down its growth. The above two factors cause a delay of sea icer melting in the spring in spite of the increase in air temperature.

Sea ice-related information is rather scattered over vast areas of interdisciplinary study fields (e.g., physics, chemistry, materials science, remote sensing, climate, oceanography, cryosphere, marine structure and operation, marine biology, and, not the least, civil, mechanical, and naval engineering, related to coastal and offshore engineering). Publications related to the physics and remote sensing are also scattered and not limited to the English language. When written in languages other than English, they obviously become not readily available. Some of the familiar books that address physics and geophysics related to sea ice are Pounder [1965], Paterson [1969], Untersteiner [1986], Wettlaufer, Fasj and Untersteiner [1999], Petrich and Eicken [2009], and Thomas [2017]. A comprehensive description of sea ice physics in addition to a historical background of the Arctic and Antarctic regions explorations is presented in Weeks [2010]. Books that cover remote sensing of sea ice with some coverage on sea ice physics include Hall and Martinec [1985], Haykin et al. [1994], Carsey [1998], Jefferies [1998], Jackson and Apel [2004], Sandven and Johannessen [2006], Reese [2006], Johannessen et al. [2007], Comiso [2010], and Johannessen et al. [2020]. A few notable review papers on remote sensing of sea ice include Sandven [2008], Breivik et al. [2010], Kwok [2010], Meier et al. [2011], and Heygster et al. [2012].

1.2. CANADA AND THE ARCTIC: HISTORICAL AND COMMUNITY SYNOPSIS Except for small areas of persistent ice-free conditions, called polynyas, virtually all the northern oceanic areas of the second largest country in the world, Canada, freeze each winter. Historically, however, Canadian north is divided loosely to High Arctic and Low Arctic regions. The definition is based on various environmental and biological characteristics. Tundra is most common in the Low Arctic while polar barrens dominate the High Arctic. In Canada, the High Arctic is marked approximately by a circle parallel to the latitude of about 72 N and spans between the longitudes of 75 W and 125 W, and consists of numerous islands within the CAA. These islands are separated from those of the Low Arctic by a few connected sea waterways: Lancaster Sound, Barrow Strait, Melville Sound, and McClure Strait (Figure 1.2). These are hostile waterways for human activities because of the mobile pack ice in many areas. In general, the High Arctic used to be the “no-man’s territory” until about 1958; but its vast area was traveled by explorers from Europe, particularly Norway, only during the late nineteenth and early twentieth century. Naturally, the Inuit communities of the polar region of Canada concentrated their movements in the Low

INTRODUCTION 5 120°

110°

100°

80°

90°

70°

60°

. unt is

H Ward

D NA

AR

AXEL HEIBERG

CT

IC

EA

BOR

DEN

N PR PAT INCE RIC K IS .

Mou

Isacren

ELLEF RINGNES IS.

CAU STR RE eA lfre AIT d

ILLE IS .

D

NORWEGIA N BAY

Rea P o

int

Sac h

IS.

sH

arb

our

UND

SEN

GU

LF

Holm

an

VICTO

RIA ISLAN

D

IN FF BA Y BA

DEVON IS. CROKER BAY

Resolute

PRINCE OF WALES IS.

CK L TO NE IN N CL CHA

Mc

AM

AQ

NA

A QA

D JONES SOUN

BATHURST

VISC O BARROW MEL UNT VILL STRAIT E SO Pell P U oint STEFANND SSON IS.

S

E

GR

Grise Ford

MELV

BAN K

L EN

ISLAN

AMUND RINGNES IS

LOUGHEED IS

ay

Mc

Pri nc

IS.

MACK ENZIE KING IS.

ld B

pe

IS.

D

AN

E K AN N I BAS

Eureka

ND L ANCASTER SOU

SOMERSET IS.

GULF OF BOOTHIA

DEUR PEN. BRO

OC

MERE

ELLES

Cape Bourne

CA

ROB ESO N

IAN

rt

Ale

SIN BA

CHA NNE L.

130°

T BYLO IS.

ik

Nanisiv

BERNIER

Bond level e Clyd

BAY

IN BAFF

BOOTHIA PEN.

Figure 1.2 Map of the Canadian Arctic Archipelago showing locations of major islands and the weather stations (operational and decommissioned) operated by Environment and Climate Change, Canada.

Arctic region, particularly the Lancaster Sound and southern areas. The present-day settlements of Resolute Bay (74.72 N, 94.97 W) and Grise Fiord (76.42 N, 82.90 W), shown in Figure 1.2, were established by the federal government of Canada in late 1950s to move Inuit communities to live in the High Arctic, but did not succeed as planned. Even today, the Inuit communities of the newly formed territory of Nunavut prefer to live primarily in the Low Arctic. Nunavut was created to join the 10 provinces and two other territories of Canada on 1 April 1999. Before that it was part of the Northwest Territories. Historically speaking, the islands of the High Arctic remained isolated from human activities, except for the explorers and the adventurers, until the unrest in Europe spilled over to the north Atlantic. World War II changed the situation significantly. Interest in the Arctic as a

strategic and possible economic region was boosted, but even then, the climatological observations from the Canadian Arctic were scanty and inadequate for any meaningful analysis. Some geophysical and meteorological data were collected by the explorers and some useful data could be extracted from records kept by expeditions that attempted to find the NWP. These observations were, however, inadequate for accurate climatological studies because information was collected on an opportunity basis over short periods. The records rarely extended over a period longer than a year. Moreover, the observations made at different localities often were made in different years. The results were not comparable because of the absence of continuity in the mode of data collection. The advent of the Cold War shortly after the end of World War II, led to the creation of strong interests in Canada and particularly in the United States for the

6

SEA ICE

establishment of a network of Arctic Stations. There was a general recognition that the weather in both Canada and the United States is dominated to a large extent by the Arctic air masses. Accurate forecasting in these two countries requires detailed knowledge of the weather pattern of the polar region. Consequently, it was realized that observations from the Canadian Arctic would increase knowledge of the circulation of the earth’s atmosphere and permit an extension of the period of reliability of weather forecasts. It was emphasized that year-round observation stations have to be established for carrying out regular observational programs directly linked with the investigation of Arctic meteorological problems. On 12 February 1946, the United States Congress approved the Arctic Project and eventually a suitable basis of cooperation between the governments of Canada and the United States was developed for the Canadian High Arctic. These two countries reached a working agreement on 27 February 1947, for the establishment and operation of five weather stations in the Canadian High Arctic. This agreement was originally made for a period of five years. Consequently, five strategically placed permanent weather stations were built on islands within the Canadian Archipelago. These stations included Alert and Eureka on Ellesmere Island, Isachsen on Ellef Rignes Island, Mould Bay on Prince Patrick Island, and Resolute on Cornwallis Island (see locations in Figure 1.2). Following the first five years of project, subsequent agreements affirmed that these five stations should continue to be operated jointly by Canada and the United States in accordance with the specifications agreed to at the Joint Arctic Weather Stations Conference, which was held annually. According to this agreement the Atmospheric Environment Services (AES) of Canada (now Meteorological Service of Canada–MSC), part of Environment and Climate Change Canada (ECCC), provided all permanent installations and approximately half the staff, including an officer in charge who was responsible for the overall operation of the station. The advancements in rocketry led to the dawn of the space race and made the world smaller and more easily accessible. A sudden thrust in the space race was imposed by the successful launching of the Soviet Union’s satellite “Sputnik” in October, 1957 and directly forced and indirectly enhanced interests in the north polar zone. The launching of the first man-made satellite suddenly proved the world to be small and the polar regions within reach, but more importantly, levied a need for obtaining an indepth knowledge of the terrain in the north beyond the Arctic Circle—the High Arctic and the Canadian Basin. In 1958, the United States lost the race for launching the first man-made satellite, but succeeded in completing the much-publicized first underwater crossing of the

Arctic Ocean using the submarine Nautilus. Although the Soviet Union (USSR) was prying the oceans with their fleet of nuclear submarines, no claim was made about any under-ice activities around the geographic North Pole by the USSR. In a very direct manner, the under-ice activities and the space race forced Canada to increase the awareness of the Canadian Arctic Island and implement new measures to strengthen the Canadian sovereignty in all these islands of the High Arctic. This led to the establishment of two permanent human (Inuit) settlements, as mentioned earlier, Resolute at 74.72 N and Grise Fiord at 76.42 N. The United Nations Conference in 1958 on the Law of the Sea, which extended the resource and exploration rights of maritime nations on their continental shelves to a depth of 200 m, acted as the catalyst for Canada for undertaking multidisciplinary scientific exploration of the north. Consequently, the Canadian Government established the Polar Continental Shelf Project (PCSP) under the ministry of Energy Mines and Resources or EMR (recently renamed as Natural Resources Canada or NRCan) in 1958. The most important aspects were the geophysical mapping of the High Arctic—areas of the land and the ocean (Figure 1.2) stretching from Alaska to Greenland and from the Arctic Circle to the North Pole. The important field of investigations included recording the magnetic and gravitational data required for the space program. Nonetheless, the Canadian Government decided not to invest heavily in building up its armed forces for the purpose of maintaining Canadian presence all over the Canadian High Arctic. Instead of sending armed forces personnel to the High Arctic, Canada always used a pool of scientists and experienced field workers. Since then, the Canadian sovereignty of the High Arctic is essentially maintained by the labor of love of Canadian scientists (as well as their scientific collaborators from other countries). The birth of the new territory of Nunavut in 1999, ranging from mountains and fiords on the eastern shores of Baffin and Ellesmere islands, through the lakes and tundra of the Barrens on the mainland, to the plateaus and cliffs of the Arctic coast, is changing the course of history of the Canadian Arctic. Since 1959, the PCSP (Polar Shelf for short) with permanent base camps (shelters or shacks with rudimentary facilities) at Resolute on Cornwallis Island and Tuktoyaktuk in the Mackenzie Delta provided logistic support services to the scientists. This organization provides room and board (actually excellent nourishing food for the body and the soul) at the base camps, supplies land vehicles designed for all types of terrain, special field equipment, and responds to the Arctic fieldworker’s greatest expense and concern: safe, efficient air transport (helicopters and aircrafts), and an excellent radio communications

INTRODUCTION 7 Table 1.1 Weather stations in the Canadian Arctic region (currently operational or decommissioned). Weather station

Latitude N

Longitude W

Start

End

Territory

Alert Eureka Isachsen Grise Fiord Mould Bay Resolute Nanisivik Pond Inlet Sachs Harbour Holman Clyde

82.522 79.983 78.783 74.417 76.233 74.717 72.983 72.682 72.000 70.733 70.486

62.285 85.933 103.533 82.950 119.333 94.970 −84.617 −77.969 −125.267 −117.783 −68.517

1913 1947 1948 1973 1948 1947 1976 1975 1955 1941 1933

Cont. Cont. 1978 1977 1997 Cont. 2011 Cont. Cont. 1969 Cont.

Nunavut Nunavut Nunavut Nunavut NWT Nunavut Nunavut Nunavut NWT NWT Nunavut

network. Civilian research scientists and engineers of Canada, of wide-ranging disciplines, played and keep playing the most significant roles in advancing polar science and crucial roles in flying the nation’s flag in the High Arctic, the Canadian Basin. and the Arctic Ocean since 1959. The weather stations in the Arctic region (most of them are in the High Arctic) are listed in Table 1.1. These stations were established by the Canadian Government and some were operated jointly with the US military. Only six stations are currently operational (out of about 900 stations operated by the MSC across Canada): Eureka, Alert, Resolute Bay, Pond Inlet, Sachs Harbour, and Clyde. The locations of all stations are marked in Figure 1.2. When all stations were operational, they formed a network for providing weather services which was considerably significant by the Arctic standards. A detailed historical account of development and expansion of meteorological facilities in the Arctic is given by Smith [2009]. Sea ice, in one form or the other plays a significant role in the daily life of communities inhabiting the cold regions of the world. The presence of ice in the seas is a welcoming sight for the people of the northern Canada, the Inuit. Those are indigenous people with common culture and language inhibiting the Arctic and subarctic regions of Canada, Greenland as well as Alaska. Actually, the people from the Baffin Island migrated and settled in Greenland, and share the same cultural aspects of the northern communities in Canada. The town of Qaanaaq (77.49 N, 69.38 W), also known as Thule or New Thule, is the northernmost permanent human settlement in the world, located in northwestern Greenland. For its population of nearly 650, like the other communities of the Low Arctic of Canada, the sea ice is home. Inuit communities have developed many stories, myths, and profound knowledge about sea ice. Prior to the onset of outside pressures in 1960, life in the north continued both on land and the ice depending on the season. Inuit used to have temporary villages of igloos on

Figure 1.3 Using discarded sea ice samples from N.K. Sinha (after bending strength tests), the children of Qaanaaq, Greenland, posed proudly to show off one of their creations—an Inuit sculpture or “inukshuk” (photo of N.K. Sinha, from National Research Council Canada, Publication, Sphere, No. 4, 1994).

the ice. Children were born on the ice. People would live and travel on the ice, albeit slowly, without the fastmoving snowmobiles introduced in the 1970s. Consequently, the Inuit had developed some “age-based” terminologies for sea ice types (some are introduced in the next section). In short, the frozen seas have always been a very welcome feature and still are, as can be felt from Figure 1.3. A child in Qaanaaq showed the co-author of this book (Sinha) a symbolic gesture of working together (Figure 1.4). Hunting has been at the core of activities of the Inuit communities for generations. They hunt seals, caribou, musk-ox, and polar bears among other animals. Polar bear hunting used to be a major economic activity before it was regulated to save this species. It is still practiced though under strict regulations (quota proportional to the local bear population). In fact, the Canadian Arctic

8

SEA ICE

Figure 1.5 Drying polar bear skin in the village of Resolute Bay in the Canadian Arctic. The picture was taken in May 1995 (photo by M. Shokr).

Figure 1.4 Qaanaaq child proudly proving practical use of a core-hole and a sense of sharing (photographed by N.K. Sinha, March 1994).

is one of the most successful areas for bear hunting. Inuit use polar bear meat as a source of protein, vitamin A, and iron; and bear skin to make warm clothes and blankets. Bear hunting involves scanning a wide expanse of ice on a dog sled and once a polar bear is spotted, the hunter releases the dogs to corner the bear, and then kill it with a harpoon. Figure 1.5 shows a polar bear skin drying in the sun in Resolute Village, Canadian Arctic. In spite of the significant changes in the life of the Inuit communities in the last fifty years, day-to-day life is still intertwined with sea ice. However, with the advent of microwave communications systems, television signals reached the people of the north during the middle of the seventies. Saturdays became the days of the “Hockey-night of Canada” popularized by the Canadian Broadcasting Corporation (CBC). In early days of his research, based at Pond Inlet, the co-author (Sinha) found it impossible to get anybody to assist him on the ice during the days of hockey-nights. In fact, they would invite him to spend the time with them, share their food, hear their stories, and learn the intricacies of handling pucks on the ice. It is extremely important, therefore, for sea ice scientists and engineers, to work together with the local people of the north and share information and experience. The Government of Canada allocates millions of dollars through programs to support and develop the

indigenous communities in the Arctic region. Projects include tasks to mitigate impacts of climate change, maintain clean air and water, and warrant healthy environment. The government acknowledged the people of the north by issuing a series of stamps on the Arctic, of which a sample is shown in Figure 1.6. In Canada, those communities are called First Nations and now they have more say in their affairs.

1.3. THE FASCINATING NATURE OF SEA ICE Most people living in cold countries, where snow and ice are part of the most familiar of natural phenomena, do not think much of scientific importance of these natural materials. We never realize that the solid state of water, in all of its forms, is actually a unique and the most fascinating natural crystalline material. Floating sea ice, in particular, is a very complex material. It features four readily noticeable and interesting characteristics. First, it is a composite material that encompasses three phases of matter: solid, liquid, and gaseous, depending upon temperature. Second, it exists in nature at temperatures very close to its melting point. In fact, the ice–water interface at the bottom of floating ice covers is always at the melting point. Third, it floats simply because it has lower density than the density of its melt (i.e., the liquid from which it solidifies). Moreover, snow deposits on floating ice sheets add to the complexities of the ice regime. Fourth and certainly the most important aspect of floating ice covers (both freshwater and sea ice) is the fact that they act like blankets and protect marine life in lakes, rivers, and oceans.

INTRODUCTION 9

Figure 1.6 First-day cover issued by Canada Post in 1995 featuring a series of stamps on the Arctic and a photograph by N.K. Sinha (National Archive Canada No. C-24520) of NRC borehole indenter system (measuring ice strength) on sea ice in Resolute Bay (“Canada 95,” The Collection of 1995 Stamps, published by Canada Post Corporation, pp. 40–41).

The first two characteristics make floating ice highly responsive to changes in atmospheric temperature, especially when it is thin. Its physical and radiometric properties, as well as the properties of a possible overlaid snow cover, change in order to maintain a state of thermal equilibrium between the ice and the atmosphere. The third characteristic is responsible for the flotation of ice on its melt and therefore moving in response to wind and oceanic current unless it is shore fast (called “landfast”) or becomes grounded in relatively shallow waters. When landfast ice is subjected to tidal actions, it produces cracks, “ice hinges” and rubbles parallel to the shorelines. Sea ice is considered to be the fastest global-scale solid material moving upon the earth’s surface. Given the complex nature of sea ice composition, its thermal state, and mobility, it is important to understand the processes involved in its formation, growth, desalination and deformation as well as its decay. This should help to demystify the descriptions found in literature about ice, and sea ice in particular, as apparently peculiar, bewildering, confusing, puzzling, baffling, etc.

The heterogeneous and multiphase composition of sea ice arises because the salts and gases that dissolve in seawater cannot be incorporated into the lattice (polycrystalline) structure of sea ice. This structure is made up of pure ice crystals, leaving salts to be included within the interstices of the solid ice matrix in the form of liquid brine. Gases are also included in the form of gaseous bubbles. Other impurities such as microalgae, non-organic deposits, and trace elements may also exist. A characteristic process that follows from this multiphase composition is the brine drainage (which takes other impurities with it) into the underlying ocean water. This process takes place at a rate that depends on the ice permeability and temperature. It continues throughout the lifetime of the ice cover causing the bulk properties of the ice to be continuously changing. Since ice exists in nature at temperatures of only a small fraction below its melting temperature. Therefore, from the geophysical and materials science point of view, it is considered to be a high-temperature material. High thermal state of ice in nature and related implications are

10

SEA ICE

discussed in detail, in section 5.1.1. Even a very cold Arctic air temperature, T, of 233 K (or −40 C) for pure ice is equivalent to a normalized (called homologous) temperature Th(=T/Tm) of 0.85 Tm, where Tm is the melting point of 273 K. This is only 15% below the melting point. Even floating ice sheets rarely attain such low temperatures. Snow covers act like blankets that keep the ice rather warm even in the middle of the winters. A realistic average temperature of a snow-covered floating ice sheet, say −10 C (263.15 K) is only about 3.7% below its melting point. Unquestionably, this has to be considered as an extremely high temperature. This important aspect of natural ice is normally overlooked by the communities of ice scientists, engineers, or environmentalists. An important implication of this feature is the rapid and drastic changes of surface and bulk properties of ice cover in response to variations in meteorological conditions (temperature and precipitation). These variations often take place during a very short time frame (a few days or even hours), especially during the early formation and onset of melt periods. This is similar, albeit occurring at a much shorter scale, to long-term changes in the landscape of the earth’s solid crust (existing at homologous temperatures, less than 0.15 Tm) in response to changes in the underlying viscous mantle. In both cases, changes are caused by external or internal stimuli but the landscape of the earth’s crust evolves over tens of thousands of years because it exists at homologous temperatures far below those of ice. Ice has lower density compared to its melt (fresh or saline water). Most liquids shrink as their temperature approaches the freezing point because the molecules tend to move slower. At the freezing temperature the molecules of the crystalline solids are usually tightly packed, exhibiting higher density. Water and ice are exceptions because of the polar nature of their molecule that develops hydrogen bonding between them. When water is cooled to near its solidification temperature, hydrogen bonding causes molecules to rearrange into lattice structure with “open gaps” within the lattice. This causes a decrease of water density below its maximum density of 1000 kg/m3 at 4 C. The density of pure ice is 917.6 kg/m3 at 0 C and it increases slightly with decreasing temperature, reaching 934.0 kg/m3 at −180 C. Due to this unique feature, ice floats on water. This sustains marine life under the ice cover as seawater never freezes all the way to the ocean floor. Nevertheless, the floating ice becomes mobile under wind and ocean current forces and that causes continuous opening, closing, thickening, and deforming the ice pack. It represents hazards to marine operations. One of the fascinations of ice, and sea ice in particular, that has attracted many researchers in the field of ice physics (including the authors of this book), is the microstructural feature. This is the study of ice “anatomy”; namely

its crystallographic structure that reveals much information about its multiphase components of solid ice, liquid (brine) inclusions, and gas bubbles. There is much to learn about ice from its microstructure photographs when they are produced in the best clarity and colorful quality. The microstructure reveals information about the water quality from which ice was developed, mechanical stresses to which ice was exposed to, oceanic and atmospheric conditions that prevailed during its growth, continuation or interruption of the growth, metamorphism of the snow overlaying the ice, and so on. Without microstructural information revealed by photographs of thin section of sea ice, one could only make an educated guess about the sea ice composition and growth history. Much of the information in the ice physics part of this book is supported with macro- and micrographs of ice and forensic type of investigations. These are photographs of very thin sections (less than 1 mm), with very smooth top and bottom surfaces parallel to each other, obtained by using a method called double-microtoming technique (DMT), presented in detail in section 4.2.2. When a thin section with thickness in the range of about 0.5 to 0.7 mm is viewed and photographed using polarized light, different crystals exhibit different colors. The colorful micrographs reveal the hidden beauty of nature’s art. The microstructural details also provide information related to complex structure–property relationships. Out of admiration for its beauty, the authors of this book decided to use the terms “ice-rich” or “ice-covered” water instead of the commonly used term “ice-infested water.” To the people interested only in the economic aspects of making short-cuts of marine routes from the Atlantic Ocean to the Pacific by making through the NWP, sea ice has been a formidable obstacle and the water was considered to be “infested” with ice. Moreover, ice-covered areas were considered, for example, by the British explorers in the nineteenth century as something that stood in their way and complained about the ice forming too early or not breaking up. While working on a project related to the interactions of ice with the newly constructed dock at Nanisivik in Borden Peninsula of Baffin Island [Frederking and Sinha,1977], the author (Sinha) came in contact with several persons from the nearby Inuit community of Arctic Bay (Ikpiaqjuk) and Pond Inlet (for location, see Figure 1.2). They were working at a lead–zinc mine under construction. The author learned a few simple terms of sea ice in Inuktitut (ee-nook-tee-tut) language. This is the indigenous language spoken by the Inuit in Canadian Arctic. The word means, “in the manner of the Inuit.” A few examples of sea ice names in Inuktitut language are given below. Qinu (Kai-nu) The darker frazil or grease ice, seen during the very beginning of a winter, is known as qinu.

INTRODUCTION 11

No specific distinctions are made between frazil ice and grease ice. In general, the frazil refers to the stage when thin and tiny plates or needles of ice are suspended loosely in the water. Grease ice refers to the stage when the frazil crystals have coagulated sufficiently to form a soupy or oily layer on the surface. Sikuaq (se-kwak) As the ice grows thick enough to support a person to walk on it, the name is changed from qinu to sikuaq. The ice thickness is around 80 mm or more. This stage and up to a thickness of 0.30m, is termed as “Young” ice according to the World Meteorological Organization (WMO). Tuvaq (Tovak) Ice forms first around the shorelines. If it continues to stay attached with the land and continues to grow, then it is called, tuvaq, which is equivalent to landfast ice according to WMO terminology. Sinaaq (Sin-ak) Edge of fast-ice of ice floe in open sea. One has to approach the edges of landfast (tuvaq) extremely carefully. The edges may crack and, therefore, always very unsafe to walk. The children will never be allowed to play and go close to the sinaaq. Siku (Siko) When the sea ice cover is sufficiently thick to support a group of people, it is called siku. This term is equivalent to the WMO term first-year ice (FYI), which is assigned to ice with thickness greater than 0.30 m. Ivuniit (Ivoniet) During travel on the siku, or even going from the shore to level ice, often ridges or rubbles of ice are encountered. They are caused by winds and/or currents or tidal waves. These obstructions are named as ivuniit. Qavvaq (kav-ak) Old, salt-free, white or translucent sea ice is named as qavvaq and is equivalent to MYI. Nilak (Nilak) Freshwater for drinking. When an ice cover is thin, a small variation in thickness results in large variations on the load it can carry, especially for sea ice. Moreover, relatively larger variations in ice thickness occur during the early periods of winter, when the ice cover is thin. Most drowning or break-through accidents occur when the ice is thin or not sufficiently thick to carry the load placed on the ice covers. In reality the introduction of snowmobiles have not only significantly increased a range (and speed) of activities among the young Inuit, but also, unfortunately, increased injuries and deaths, including icerelated drownings. Consequently, efforts in making rescue operations become very challenging [Lifesaving Society, 1998]. Anyway, as pointed out and illustrated earlier in Figure 1.2, it is not easy to sample thin sea ice, such as qinu or sikuaq, for ice physics investigations. But one does not require an icebreaker either to collect samples of thin sea ice. The main problem is the fact that most ice scientists living in areas remote from the Arctic and Antarctic sea ice are neither familiar with the bearing capacities of newly grown sea ice in nature nor are they equipped (mentally and physically) to go over thin

ice. Above all, unless one lives in the areas of interest and is familiar with the formation of ice, year after year, it is impossible to know the characteristics of the ice cover, particularly its thickness variability a few days after the freeze-up. Floating young sea ice cover, of thickness < 0.01 m, impose severe physical limitations on field measurements and sampling. Prior to 1977, therefore, very little effort has been made to examine young ice (YI) as compared to numerous field investigations carried out on mature FYI. For that reason, most of the early studies on the initial growth of saline ice came from laboratory investigations. Realizing the practical limitations—strategically, financially, as well as for the knowledge base—the author (Nirmal Sinha) decided to work closely with the local inhabitants of the Baffin Island. The goal was to tap and learn from the Inuit knowledge about sea ice. No question, the Inuit elders were the experts of sea ice. Due to their hunting experience on sea ice in certain localities for many years and their long-term observations on ice covers, they have developed profound and rich understanding of the marine environment. Toward the end of 1976, an opportunity came to him in Pond Inlet to build long-term relationships with the local residents. Although the locals were the experts of ice, they never witnessed the inner beauty of ice crystals that can be seen in thin sections of ice under polarized light. Using home-made portable polariscope, the author introduced the Inuit elders, the young, and the children to the wonderful world of colors of ice crystals. Almost instantaneously he developed loving relationships of the local people. In fact, they were bringing their ice samples, including pieces from an iceberg grounded in Eclipse Sound. The bonds resulted in conducting the long-term studies on annual sea ice in Eclipse Sound, described in Sinha and Nakawo [1981] and Nakawo and Sinha [1981, 1984]. A deep knowledge on the growth and distribution of ice thickness in areas of oceans close to human settlements helps in assessing wide-ranging characteristics of ice, its bearing capacity and other engineering properties of annual ice elsewhere including pack ice. Microstructurally, sea ice is significantly different from freshwater river and lake ice. It is also different from other floating ice types of land-origin, such as icebergs or landattached ice shelves and floating ice islands. The latter forms have significantly larger volume than sea ice and therefore can easily be distinguished even if they are surrounded by sea ice. Icebergs and ice islands protrude a few meters above the sea level. About 90% of all icebergs encountered in Canadian waters are calved from the glaciers of Western Greenland. The annually observed icebergs in the Baffin Bay and Labrador Sea range between 10,000 and 40,000. The Canadian Ice Service (CIS) produces charts of icebergs in the Canadian East Coast water, south of 60 N. The information is mainly

12

SEA ICE

that is heavily reinforced to withstand the pressures induced by sea ice. 1.4. SEA ICE IN RESEARCH AND OPERATIONAL DISCIPLINES Sea ice is an important component in a few study and application disciplines including physics, climatology, meteorology, oceanography, marine biology, marine navigation, and marine and offshore structural engineering. Each community has developed its interest in certain aspect(s) of sea ice, yet all of them draw from certain fundamental physical or geophysical properties of sea ice as explained below. This section highlights the need for sea ice information in a few scientific and operational disciplines. 1.4.1. Sea Ice in Physics

Figure 1.7 A tilted iceberg in Baffin Bay, June 1984, revealing underwater bulbous bow extension of the hull (top), and a modern ship with bulbous (bottom) (iceberg photo by N.K. Sinha).

based on visual observations from ships and aircrafts and also occasionally obtained from satellite imagery. Icebergs are fascinating to watch. An example of a nontabular iceberg in northern Baffin Bay is shown in Figure 1.7. They can provide interesting lessons to naval architects regarding Mother Nature’s efficient design of bulbous bows. A bulbous bow, in the language of naval architects, is an extension of the hull just below the waterline. It has many subtle shape variations, but it is basically a rounded front portion that flares out slightly as it blends into the traditional displacement hull construction. These forward protrusions are about twice as long as the width of the base. They would usually not extend forward past the top of the bow. This can be seen in Figure 1.7. The iceberg in the figure is shown in a slightly tilted position such that the underwater bulbous bow is revealed. The basic principle of bulbous bow is to create a low-pressure zone to eliminate the bow wave and reduce drag. Hulls built with bulbous bow sections are common today for seagoing ships. Under certain conditions, depending on speed and sea-state, the bulbous bow is very efficient at redirecting forces of hydrodynamic resistance and drag. Icebreaking ships do have a special shape of bulbous bow

The inner structures of any material that cannot be recognized by human eyes are known as microstructure or crystalline structure. While large single crystals of pure ice can be found in low temperate glaciers or can be prepared in cold laboratories, the usual icy objects found in nature or fabricated for human consumption consist of aggregate of many crystals. Geophysicists and materials scientists are interested in ice not only because it is transparent or translucent, but also its commonly occurring polycrystalline structure is similar to crystalline minerals that exist in nature, e.g., common metals, alloys, and ceramics. The use of knowledge about ice microstructure for understanding and demystifying the behavior of several alloys at high temperatures was highlighted and pioneered in an early study by Tabata and Ono [1962]. The extremely high thermal state of sea ice adds even more interest in sea ice physics to scientists and engineers of high-temperature material. This subject is covered in a few parts of this book and in more detail in the recently published book by Sinha and Sinha [2022]. Since ice has an accessible melting temperature, it could provide a model of how other materials will behave at very high temperatures near their melting points. Physicists have applied theories and experimental techniques that describe the behavior of snow and ice around their melting temperatures to high-temperature materials, such as ceramics and advanced alloys, used inside jet engines or gasturbine engines (examples include titanium-based and nickel-based superalloys). As pointed out earlier, sea ice exists at temperatures about 5% of, or less than, its melting point. This is significantly more than the maximum temperature allowed for the operation of man-made, nickel-based, directionally solidified columnar-grained superalloys. Such superalloys are used in blades of modern gas-turbine

INTRODUCTION 13

engines (jet or power-generating engines) and end-casings of rocket engines. They were introduced in jet engines during the late 1960s [Duhl, 1987]. Physicists, ceramicists, and high-temperature metallurgists are therefore interested in applying lessons learned from studying structure–property relationship of ice directly to complex high-temperature, titanium- and nickel-based superalloys [Sinha, 2009]. They can apply those lessons to model thermomechanical behavior of gas-turbine engine materials, such as stress-relaxation processes and strain-rate sensitivity of yield strengths [Sinha and Sinha, 2011]. The above information is substantiated by the fact that sea ice is a prime example of a natural material that exhibits impurity entrapment very similar to binary alloys. With this in mind, the crystallographic structure of sea ice exhibits grain- and subgrain-scale substructure very similar to high-temperature metallic alloys and ceramics. These features are key factors in determining many properties of the material including thermal conductivity, microwave emissivity, dielectric constant, and mechanical strength. 1.4.2. Sea Ice in Climatology Climatologists are interested in sea ice because its thermal and optical properties are important inputs to climate models. Its extent and thickness have strong influence on the high- as well as mid-latitude large-scale circulations of the atmosphere and ocean. Moreover, Arctic sea ice has long been thought to be a primary indicator of global warming. In general, sea ice plays a key role within the climate system as explained in section 13.3. Thermal properties of sea ice determine the thermodynamic interaction between the cold atmosphere and the warm underlying ocean. This is demonstrated in two processes: the heat exchange between the two media, and the heat of fusion released or acquired when ice freezes or melts, respectively. The thermal conductivity of sea ice is relatively low (2.25 W/m.K) compared to thermally conductive metals (e.g., 205 W/m.k of aluminum or 401 W/m.K of copper). When covered with snow, the effective conductivity becomes even lower (thermal conductivity of snow is between 0.1–0.25 W/m.K, depending on its density and wetness). Therefore, the presence of the ice cover on ocean surface reduces the ocean–atmosphere heat exchange even in the presence of thin ice [Maykut, 1978,1982]. This reduces the moisture transfer from the ocean to the atmosphere, the momentum transfer from the atmosphere to the ocean, and the exchange of chemical constituents between the two media. The heat of fusion of water (defined as the heat required for transforming one gram of water at freezing temperature to ice) is one of the highest of all substances. When seawater freezes, it releases approximately 80 cal/g of heat

to the atmosphere. This is the energy released as the hydrogen bonding between the ice molecules is established. It is equivalent to energy required to raise the temperature of one gram of water from 0 C to 80 C. When ice melts, it absorbs the same energy to produce water at the same temperature. This represents a huge amount of heat exchange between the ocean and the atmosphere that affects weather systems and eventually regional climate. It certainly influences regional and perhaps global climate. Motion of floating ice sheets may cause divergence, hence fractures in ice cover, or convergence, hence surface deformation (ridges, rubble, or fractures) (section 2.6). The winter heat flux from fractures (particularly large leads) to the atmosphere can be two orders of magnitude larger than the heat flux through an adjacent thick ice cover [Maykut, 1982]. However, this sudden increase may not last for a long time as water in these openings usually freezes quickly. Lüpkes et al. [2008] found that even a few percent of open water or refrozen thin ice within extensive ice cover in the Arctic may increase the heat flux significantly, hence the local air temperatures by several degrees. In general, any openings in the ice cover in the polar regions (e.g., leads and polynyas) are considered to be important climatic features and their distribution, along with the distribution of thin ice, is particularly important to the regional and possibly global climate scenarios. The primary optical property of sea ice that affects weather and climate systems is albedo (the ratio of reflected to incident optical radiation). Albedo from snow-covered ice (between 0.4 and 0.8) is significantly higher than albedo from seawater (≈ 0.06). This causes more sunlight to be reflected and therefore the ocean will not be as warm as it would be in the absence of ice. That is how the ice-covered areas in the polar regions are climatologically cold. A possible decrease in the area of polar sea ice, caused by global warming, will cause more sunlight to be absorbed, hence amplify the warming (a positive feedback loop). In the past 40 years, surface air temperature in the Arctic has increased four times faster than the globe [Rantanen et al., 2022]. In relative terms, more increase in winter temperature has been observed compared to increase in summer temperature. Climatologists are therefore interested in information about the ice extent in the polar regions and the frequency and distribution of leads, melt ponds, and polynyas. Climatologists monitor closely the interannual variability of sea ice extent in both polar regions. While the extent has not changed much in the Antarctic regions, it has decreased significantly in the Arctic in the past few decades. Minimum ice extent in the Arctic usually occurs in mid-September. So far, the lowest ice extent on record occurred on 5 September 2012 (3.39 × 106 km2), while the second and third lowest occurred in September 2020 and

14

SEA ICE

2007 with extent of 3.74 × 106 km2 and 4.10 km2, respectively. The ice extent in the Arctic has recently been reduced by 40%–50% of its average value in the 1980s. This has also been accompanied by reduction in ice thickness. A significant drop of about 0.6 m in MYI thickness, and consequently overall ice thickness, has been observed since 2005. During the first decade of this century the decrease in Arctic ice thickness has occurred at a rate of about 0.17 m per year [Kwok and Sulsky, 2010]. All these data highlight the importance of sea ice climatology (especially in the Arctic) at this time. More information on the response of polar sea ice to the recent climate change is presented in section 13.3. Early global climate models predicted retreat of polar ice as the current anthropogenic global warming continues [Kattenberg et al., 1996]. However, the current climate models have underestimated the rate of ice retreat and thinning in the Arctic. According to the Intergovernmental Panel on Climate Change (IPCC), only a quarter of climate models have successfully simulated the Arctic sea ice decline observed by satellite data. Stroeve et al. [2007] found that the rate of reduction of ice extent and thickness estimated from satellite remote sensing observations is four times faster than model calculations. Modelers indicate that this is mainly due to a lack of accurate representation of ice conditions in the models. For example, sea ice type distribution are not represented (just ice/no-ice is represented), and the heat flux from snow-covered sea ice, which is exposed to different thermodynamic processes, is usually simplified using parameterization to represent its complex physics. 1.4.3. Sea Ice in Meteorology In order to restore a temperature balance between the cold polar atmosphere in winter and the much warmer ocean, the ocean releases heat to the atmosphere. This process triggers what is known as atmospheric oscillation (AO), which is a climate index of the atmospheric circulation of the Arctic. AO is responsible for generating most weather systems over north-latitude areas. This includes the recurring low-pressure systems associated with storms and precipitation. Obviously, the presence or absence of sea ice in the polar regions affects these regional weather systems, directly or indirectly (this is much more pronounced in the Arctic). The effect is forecasted using weather models where a collection of physical processes is integrated into a consistent framework that can be used to forecast weather and predict future climatic conditions. The traditional representation of sea ice information in weather/climate models is achieved by feeding ice information at certain time steps. This approach was used in the early 1970s to initialize weather forecast models [Stensrud, 2007]. At a basic level, the required information

at each grid cell should include ice concentration, but additional data such as ice thickness, surface roughness, and snow cover are also recommended. Other ice data such as albedo, surface emissivity, and surface fluxes can be used in parameterized forms. However, it should be noted that parameterization of those parameters can be challenging because of the heterogeneity and the rapidly changing behavior of ice surface. Recent weather models incorporate a sea ice module (or a coupled iceocean model) to represent fundamental dynamic and thermodynamic processes. So far, incorporation of sea ice information in weather models is limited to the identification of the grid point as being exclusively composed of water or ice, regardless of ice type or surface features (e.g., leads, ridges, etc.). However, ice type and thickness are desirable pieces of input even at a coarse grid resolution. What complicates the issue further is the need for snow information, which is highly variable; spatially and temporally. One of the challenges in representing sea ice information in weather models is the spatial and temporal variation of the sea ice cover in terms of its physical parameters (particularly temperature, emissivity, and albedo) and surface features (leveled, deformed, or fractured ice). While the surface air temperature over the Arctic sea ice is fairly cold and stable in winter (between −30 C and −40 C), the open-water areas in leads, polynyas, or open sea beyond the ice edge, have a surface temperature at −1.8 C (the freezing point of seawater) or higher. Accurate characterization of the spatial distribution of ice surface temperature is crucial to improving weather forecast, particularly during winter, in areas where leads become closely distributed within the ice cover. The same applies in summer when melt ponds, with their substantially less albedo than surrounding ice also become closely distributed. Growing scientific evidence has pointed to the influence of the current decline in Arctic sea ice on atmospheric phenomena within and beyond the Arctic. Budikova [2009] found that changes in Arctic sea ice interact with processes related to atmospheric wind and temperature fields, as well as to thermodynamic and radiative processes connected with water vapor, clouds, and aerosol feedbacks. Cassano et al. [2014] used climate model simulation to explore the effect of the increasing open water in the Arctic on Earth’s atmosphere. They found that the increase in open water during the fall of 2007 led to an increase in atmospheric temperature. Higher temperatures in the Arctic thus cause a decrease in the pole-to-equator temperature gradient, which in turn creates a weaker jet stream and less storminess in the midlatitudes. More on the interaction of Arctic sea ice with atmosphere is included in section 13.4.1.1. To conclude this section, it is worth repeating that sea ice conditions, while of interest in their own right, are

INTRODUCTION 15

North Atlantic deep water

Pacific Ocean

Gulf Stream Pacific Ocean

Atlantic Ocean

Indian Ocean

Figure 1.8 Major routes of the thermohaline circulation that circulates cold water (blue) and warm water (red) between the deep and surface ocean worldwide. The asterisk marks the location of Odden ice where warm surface water prevents the expansion of the ice cover.

important information for improving weather forecast models, both regional and global. It is also worth noting that the term “meteorology” encompasses sea ice according to some definitions. For example, the CIS, a division of the MSC, is the leading authority for information about ice in Canada’s navigable waters. Weather and ice services operate under the same federal department of ECCC.

1.4.4. Sea Ice in Oceanography During the winter months of the northern hemisphere, when the oceans in the Arctic region freeze, sea ice in the Antarctic region melts. A reversal in this process occurs during the summer months. Freezing and thawing of the two primary cryosphere work like a seesaw of oceanic current—like a narrow board pivoted in the middle (the equatorial region in this case) such that as one goes up, the other goes down. Freezing and thawing of annual sea ice in the primary cryosphere regions affect the movement of ocean water in two ways: by rejection of salt to the underlying water during sea ice formation and growth, and by melting large volumes of ice which is drifted away from the polar areas during the summer. The first process initiates what is known as thermohaline oceanic circulation, while the second disturbs it. Thermohaline circulation (also known as the Great Ocean Conveyor Belt) is a large-scale pattern of seawater motion around the world driven by vertical density gradient. The adjective thermohaline is composed of two syllables: thermo, which refers to temperature, and haline which refers to salt content. Changes in water density are associated with changes in its temperature (colder water is denser). Both are caused mainly by the process of salt rejection at a high rate during

the initial sea ice formation and a lower rate throughout lifetime of the ice (see section 2.5.5 for details). Figure 1.8 shows the major routes of the thermohaline circulation worldwide. Since the saltier water under the ice is heavier than the deep water, a density gradient is developed that causes water to sink at the location of sea ice formation and growth. This triggers the circulation in the oceans around the world. The cold and dense polar water starts to move along the ocean bottom toward the equator, while warm water from mid-depth to the surface level travels from the equator back toward the poles. Much of the world oceans’ deep and bottom water is believed to be formed in polar latitudes as a result of sea ice formation and growth. Obviously, major changes in the amount of newly formed sea ice (e.g., decline due to global warming) can disrupt normal ocean circulation [Maykut, 1978 and Carmack, 1986]. Figure 1.8 also shows the location of formation of what is known as Odden sea ice in the Greenland Sea (Odden is the Norwegian word for headland). This is the crucial location for the initiation of the thermohaline circulation. The circulation takes the form of a large tongue that extends over 250,000 km2 [Comiso et al., 2001] and it can rapidly expand or shrink over a period of a few days. In some winters it persists for months, while failing to form in others. The thermohaline circulation is closed in the Greenland Sea as the warm water moving from the south near the ocean surface reaches that high-latitude area and starts to cool. The cooler water becomes dense and therefore sinks. The circulation continues afterward. This cycle may slow down if significantly less salty water (or freshwater) is provided. In this case, the sinking of seawater will be activated only by the colder and denser water at the surface. The cycle may be even interrupted if much less

16

SEA ICE

sea ice is formed or the atmospheric temperature is maintained at the warm side so that warmer water coming from the south will not be cooled enough to make it denser. This anomaly is called the “great salinity anomaly.” It was discovered during the late 1960s and early 1970s in the Nordic Seas (which includes the Greenland, the Norwegian, and the Iceland Seas). It continued to be observed in the 1980s and 1990s. A notable study that addresses this phenomenon is presented in Häkkinen [1999]. The question that would be raised at this point is: what would give rise to the major impulse of freshwater that causes this anomaly? As mentioned before, excessive rain, snow, or river runoffs are the possible answers. Yet, excessive melting of glaciers and icebergs is a much more serious cause. It may reduce the formation of sea ice. The impact of this anomaly is yet to be assessed. Normally, due to ocean currents and winds, Arctic sea ice moves southward from the Arctic basin to the Greenland Sea through Fram Strait (located between Greenland in the West and Svalbard in the East). As it continues its journey south and when the melt season approaches, the sea ice starts to melt. The melt decreases the salinity of the seawater near the surface. Excessive melting of sea ice may lead to the same effect of decreasing seawater salinity as the melting of ice shelves and icebergs. Once again, this anomaly will slow down the thermohaline circulation. Therefore, any significant increase in the temperature of the Nordic Seas or excessive melt of ice shelves is a possible scenario that may negatively impact the thermohaline circulation. 1.4.5. Sea Ice in Marine Biology Sea ice provides habitat, shelter, breeding, feeding, nursery, and hunting grounds for a variety of species, ranging from microorganisms (alga and bacteria) to birds and marine mammals (polar bear, walrus, and a few species of whales and ice-associated seals). The survival of the animal populations is related to the presence of areas of ice-free waters within sea ice cover. These areas provide migration routes and sources of abundant under-ice food reserve for animals. A notable example of such areas is polynyas. These are areas within an extensive ice cover that contain open water and thin ice even in the middle of winter when atmospheric temperatures are very low. Sea ice also supports the habitats of many different fish as it maintains the seawater at a warm enough temperature for their survival. A retreat of sea ice will decrease the available platforms that birds and mammals use to rest on and from which to hunt. There have been concerns that this may lead to significant loss in the population of those species. Animals that depend on sea ice for their survival prefer different habitats within the ice cover. For example,

Figure 1.9 Seal breathing hole in first-year fast ice in Lancaster Sound, Canadian Arctic (photo by M. Shokr, May 1990).

bearded seals and walruses prefer areas of thin or broken ice cover in relatively shallow water, because their main food source is benthic invertebrates. Walruses and belugas live along leads within pack ice. Ringed seals prefer fast-ice because of its stability for the successful rearing of their pups and the sufficient snow cover for construction of birth lairs. They usually congregate along the edges of the ice and they use holes in the ice to breathe (Figure 1.9). Polar bears also use this ice type since they mainly feed on ringed and bearded seals. Traditionally, Inuit hunters also use the breathing holes for locating and getting their kills for food as well as for the furs. Harp seals live and give birth to their pups on heavy ice floes within ice-edge zones (Figure 1.10). In winter, Arctic foxes venture onto the sea ice to scavenge for remains of seals killed by polar bears. Laidre and Regehr [2017] present a summary on Arctic marine mammals. A review of the impacts of Arctic sea ice decline on marine creatures and human activities is presented in Meier et al. [2014]. As for the Antarctic, Weddell seals spend most of their time beneath fast ice (they can undertake dives for at least 82 minutes), while coming onto the sea ice surface only to rest and have their pups. Leopard seals and crabeater seals are species of deep pack ice. However, since leopard seals are predators and cannot move easily on sea ice, the crabeater seals often rest on rougher ice where they are safe from predation. Emperor penguins, found only in the Antarctic, prefer the fast ice zones. In general, marine biologists confirmed that animals prefer ridged ice as an attractive habitat. Obviously, ice information is needed

INTRODUCTION 17 (a)

(b)

Figure 1.10 Calving season for seals on sea ice in the Labrador Sea: (a) mothers and (b) a puppy (photos by N.K. Sinha, March 1989).

to identify the suitable habitat for each species. Information should include key surface features, not physical properties, of the ice. Remote sensing observations that can be used to identify required features such as fast ice, ridges, leads, ice edge, etc. (section 10.1). Aside from the animals that use sea ice as their platform, many microorganisms (at the bottom of the marine food chain) find their habitats within the volume of the sea ice. A notable review on this subject is presented in Horner [1985]. When seawater freezes, high concentration of salty water (brine) forms immediately under the ice–water interface and is also entrapped between the ice crystals, filling a network of pockets and channels (section 2.5.4). Brine provides the necessary nutrients to host extensive algal, phytoplankton (microalgae), and bacterial communities (Figure 5.42). Ice algae contribute considerably to the total primary production in the Arctic (25%) and in the Antarctic (20%) [Legendre et al., 1992]. About 200 species are found in Arctic sea ice, the most common among them is diatoms. These microorganisms sustain the secondary production and that, in turn, supports life of all marine animals including fish, birds, seals, bears, and many others, all the way to the killer whales. In addition to nutrients, the survival of these organisms hinges upon the availability of light that penetrates the ice sheet to reach the depth, where sufficient nutrients are available for organisms to grow. This is known as photosynthetically active radiation (PAR). Since sea ice absorbs most of the incident light and the snow cover reflects most of it, the PAR at 1 meter depth of an ice sheet represents only 1%–5% of the visible spectrum that penetrates the ice sheet. Incorporation of microorganisms from the water column into the sea ice may occur during the ice

formation. However, high concentrations of microalgae have been observed during the springtime within the interstices of the lower margin of sea ice floes. In early spring the sun starts to shine after the end of long polar nights, and the PAR reaching the bottom of the ice sheet (can be 2 m thick) would be sufficient for the algal growth. Microalgae tend to concentrate in the bottom ice layers because it is more favorable microhabitat than the upper layers. This is due to less stressful temperature and higher salinity environment. Nevertheless, microalgae are also often found in a thin layer of seawater immediately under the water–ice interface. For at least 1–3 months, ice algal blooms enhance and extend biological production in polar waters. Depending largely on climatic and environmental variability, biomass accumulation of sea ice algal populations eventually depends upon the duration of their growth season. More information on biomass accumulation at the bottom of the ice is presented in section 5.5. 1.4.6. Sea Ice in Marine Navigation Sea ice plays vital roles for naval transportation of goods and supplies during winter months in ice-rich waters of many northern countries. Some major areas of impacts are the Gulf of Bothnia (Sweden and Finland), Northern Sea and Okhotsk Sea (Russia), the Baltic Sea (Eurasia), the Gulf of St. Lawrence and Labrador Sea (Canada), and the Bohai Bay (China). Safe marine navigation in these water bodies requires timely information on ice accompanied with meteorological data. Technical products to map ice extent, types, strength, and surface features are produced regularly by a few national ice centers in northern countries to serve marine navigation.

18

SEA ICE

Charts are mainly generated based on interpretation of remotely sensed images, but meteorological and climatic data are incorporated to support the interpretation. In Canada, the Canadian Coast Guard (CCG) provides recommendations on safe navigation routes. Their icebreakers provide escort services to ships transiting Canadian ice-covered waters. The CCG receives ice information from a few sources, of which the CIS is the prime source. Icebreaker superintendents have the updated conditions of the prevailing ice conditions and the anticipated trend of ice motion. Therefore, they are well-positioned to provide reasoned advice regarding the best routes for the ships to pursue. Usually, the masters of the ships report to the CCG before entering an area of ice-covered water. This ensures a continuous monitoring of the ship’s position by CCG operators. If icebreaker support becomes necessary, it can be provided with minimum delay. Records of sea ice concentration and thickness obtained from remote sensing data during the past two decades

have confirmed a trend of Arctic ice cap thinning and shrinking. This may lead to opening a seaway connecting the Atlantic and the Pacific oceans for marine navigation for longer time of year. The dream of many early European explorers, which occupied their adventures for more than 400 years, may soon be realized by the merit of climate change. The economic impact of this scenario is significant. Ice-free future sea routes will reduce the number of days of marine transportation between Europe and Asia and double the vessel fuel efficiency. More importantly, commercial marine transportation will have more access to natural resources in the Arctic. Two routes are expected to open for summer marine navigation as sea ice retreats and will link the Atlantic and Pacific oceans: The Northwest Passage (NWP) and Northeast Passage (NEP). The latter encompasses the Russian Northern Sea Route (NSR). Figure 1.11 displays the two routes overlaid on a bathymetry map of the Arctic Ocean and surrounding seas. The NWP encompasses a few routes

Figure 1.11 Expected ice-free Northeast and Northwest passages that connect the Atlantic and Pacific oceans. The outer boundary of the map is 50 latitude. The location of the north magnetic pole (as of 2008) and north geographic pole are marked. The average location of the north magnetic pole moves in loops at around 40 km/year.

INTRODUCTION 19

that pass through the CAA region. The CAA has always been a major impediment to the fully opening of the NWP, because thick ice from central Arctic advects into it and consolidates during winter. The NEP, on the other hand, is a set of marine routes that run along the coast of Siberia, making use of a few straits through the islands of the Russian Arctic. Sea ice will continue to be monitored in the future to spot any long-lasting opening of the routes.

1.4.7. Sea Ice and Offshore Structures For offshore structures—both floating and fixed, including near shore structures such as docks and ports, bridges and vessels in ice-rich waters—the influence of sea ice is probably the most significant factor to be considered by the designers and structural engineers. Figure 1.12 shows a part of the 13-kilometer Confederation Bridge, the world’s longest bridge over ice-covered water. It connects two maritime provinces in Canada: New Brunswick and Prince Edward Island. Structural frameworks and their mass must be able to withstand the local and overall forces exerted by moving ice. The structures must be protected from encroachments of ice by effective management plans. Satellite remote sensing is an effective monitoring tool. The ice could be seasonal or perennial, thin or thick, and leveled or deformed. Ice movements could be continuous or intermittent. Icebergs and bergy bits are also hazardous for marine structure. Marine structures in ice-rich waters for energy industry are required for construction of environment-friendly drilling platforms, transportation of oil and gas from such remote parts of the globe, and avoidance of oil spills and clean-up in case of accidents.

During ice-structure interactions, there is only one choice. Ice must fail, not the structure. This idea is illustrated in Figure 1.13, which depicts a notch in an SYI floe spotted in northern Baffin Bay. The notch was made during a dedicated ramming event by the ice breaking, bulk carrier, MV Arctic in June 1984. During this planned ice breaking expedition, old ice floes were instrumented and ice characteristics were recorded before conducting the tests. The CCG icebreaker Louis S. St-Laurent also participated in this expedition (shown in the background in Figure 11.13). Soon after these ice breaking trials, the icebreaker joined USCG Polar Sea to become the first North American surface vessels to reach the North Pole on 22 August 1994. Louis S. St-Laurent was launched in June 1966 and joined the SS Manhattan expedition in the NWP during the two weeks, 8–22 September 1969, together with

Figure 1.13 Notch in second-year sea ice floe made by ice breaking bulk carrier, MV Arctic, with Canadian Coast Guard Ship (CCGS) icebreaker, Louis S. St-Laurent in the background, Baffin Bay North, June 1994 (photo by N.K. Sinha).

Figure 1.12 The Confederation Bridge, Canada, connecting the two Canadian provinces of New Brunswick and Prince Edward Island. Ice cones are subjected to ice loading [Lithuanian Geotechnical Society].

20

SEA ICE

two other icebreakers: the USCG Northwind and Staten Island. Louis S. St-Laurent is classed as a heavy Arctic icebreaker, and is the largest icebreaker and flagship of the CCG. There are a number of excellent books on ice-structure interactions and ice engineering, e.g., Cammaert and Muggeridge [1988], Sanderson [1988]; and Jones et al. [1991]. Up to date information on the subject of icestructure interactions can be found in the journal Cold Regions Science and Technology (CRST). Moreover, the regular conferences of the Offshore Mechanics and Arctic Engineering (OMAE) of the American Society of Mechanical Engineering (ASME), and the Port and Oceans under Arctic Conditions (POAC), provide current status of developments in ice engineering. These conferences also provide platforms for disseminating research results, new developments, and novel concepts in the multidisciplinary field of polar science and technology. 1.4.8. Sea Ice as A Transportation Platform During the long winters in many cold regions of the earth, there are more possibilities for finding frozen lakes or sea ice surfaces than land-based areas for aircraft operations with little or no preparations. Usually FYI, sheets in fjords are relatively free from ridges and rubble fields. This is particularly the situation where there are mountains or hilly areas on both sides and thereby protected from severe wind effects. For the Arctic and the

Antarctic, sea ice in refrozen leads within the pack ice also provides ready-made sites for emergency operations for search and rescue, especially with short take-off and landing (STOL) aircraft, such as Canada’s famous Twin Otters. Once landed using a light STOL aircraft, measurements on the ice conditions can be performed and evaluated for operations of larger airplanes [Sinha et al., 1996]. Figure 1.14 illustrates a set of plots for weight vs ice thickness, as functions of flexural strength of ice, recommended for operations of aircraft. The bearing capacity of an ice sheet can be affected more by ice quality than by ice thickness. Reliable estimates of ice strength can be made by experienced ice specialists through observations of the ice type, quality, temperature, and surface uniformity. This may be supplemented by field measurements of ice bending strength, described in detail in Appendix A of Sinha et al. [1996]. These estimates can then provide the basis for decisions concerning use of the airstrip for unlimited movements, or allowing loads in excess of the maximum recommended for limited use. It is emphasized that an engineering analysis, including a detailed survey and investigations of the ice cover, should be made by a qualified ice specialist to approve a runway for an unlimited number of landings. For practice, simulated emergency landing by a Search and Rescue (SAR) aircraft of the Canadian Armed Forces were made in many areas of the High Arctic. An example of landing made on a refrozen lead in first-year sea ice is shown in Figure 1.15. Based on the charts in Figure 1.14,

10,000 Allowable ice flexural stress (kPa) 200

Minimum Effective Ice Thickness (mm)

400 600 800 1000 1500 2000

1,000

Minimum 250 mm required for aircraft operation

FIGURE 1 MINIMUM ICE THICKNESS FOR LIMITED AIRCRAFT MOVEMENTS

100 10

100

1,000

10,000

Gross Weight of Aircraft (kN)

Figure 1.14 Minimum thickness recommended for landing aircraft on sea ice as functions of flexural strength [from Figure 1 in Sinha et al., 1996].

INTRODUCTION 21

Figure 1.15 Search and Rescue (SAR) Twin Otter of the Canadian Armed Forces on a snow-covered first-year sea ice refrozen lead near Beechey Island, Nunavut, Canada (photo by N.K. Sinha).

history was made by a Canadian airline, First Air, by landing a fully loaded (64,640 kg) Boeing B727 jet aircraft on first-year sea ice in Frederick Hyde Fjord in northern Greenland at about 83 11 N, 29 50 W [Pole, 1995]. The landing location was 29 50 W, but at that high latitude, the longitudes cover significantly less distances. The longitude of the 30 km long fjord was between 28 W and 32 W. Also, it was the most northerly point on sea ice cover that a B727 jet aircraft had ever landed and operated on a commercial basis. In all, 16 landings were made in 7 days from 19 to 25 May, 1994 [Sinha, 1995]. The ice strip was 100 m wide and 2.5 km long. The ice thickness varied from 2.24 m at one end to 2.41 m at the other end of the runway. The ice was thus extremely uniform, and about 50% of the surface was absolutely snow-free. Actually, even the snowcovered areas had very little snow. Consequently, from the air the entire fjord was turquoise in color. An example of unloading the aircraft is shown in Figure 1.16, which shows clearly the snow-free surface of the ice.

Figure 1.16 Boeing 727 on first-year sea ice in Frederick Hyde Fjord, Greenland in May 1994; the foreground shows snow-free surface (photo by N.K. Sinha).

Frozen seas also provide vital links for surface transportation between the communities and hunting opportunities in the Arctic and subarctic areas of many northern countries, especially Canada. However, the use of ice covers for transportation is not limited to the waters covered with sea ice. Ice surfaces of frozen lakes and rivers are routinely used for making temporary runways, winter roads for transportation of goods, and recreational trails and winter sports throughout Canada. Increase in tourisms in the Arctic is also opening new opportunities for use of sea ice covers for transportation. Snowmobiles replaced the dog teams almost 50 years ago, but nothing could replace the supreme yet elegant design of the traditional Inuit sledges. The only difference is that the sledges are pulled by the snowmobiles. Dog teams and rides on sea ice have become the sources of tourist attractions (Figure 1.17).

1.4.9. Sea Ice in Relation to Solid Earth Sciences: Rocks and Plate Tectonics In a recently published book [Sinha and Sinha, 2022 (Chapter 11)], the authors present unfamiliar similarities between plate tectonics and sea ice dynamics. Sea ice floes float on its melt of seawater, while plate tectonics, massive slabs of solid rocks of both continental and oceanic lithosphere, float on the asthenosphere, i.e., the “sea of magma” in the lower mantle. The interactions of floating ice with the underlying water operate at temporal scales of hours, days, and weeks. On the other hand, the interactions of floating tectonics with the underlying mantle operate at time scale of millions of years. The two processes appear to be similar albeit the huge difference of the time scale. This has invoked the idea of exploring similarities between the earth’s crust fragmenting, interacting, and drifting while floating on its magma and sea ice floes drift and interactions while floating on seawater. The underlying theme is that the interactions of ice floes (or sheets), in the form of diverging, converging, and transformation, can be observed and recoded in finite scales of time and space. This can be achieved using insitu, airborne, or space-borne measurements. On the other hand, the structure of the upper mantle (comprising lithosphere and asthenosphere) can only be imagined from sketches of geophysicists. Hence, observations of sea ice rafting, ridge-building, and rubble formation can be used to support the work of the geophysicists because sea ice processes bring out a vivid analog of the earth’s crust. Different ice floes can be considered as tectonic plates, and their physical aspects such as thickness, rate of movements, deformation, and material characteristics can be quantified before and after the interactions. This is precisely how they can

22

SEA ICE (a)

Pond Inlet, 1978 (b)

Qaanaaq, 1994 Figure 1.17 (a) Dog team at Pond Inlet (Canadian Arctic) with Bylot Island in the background in 1978 (photo by N.K. Sinha), and (b) dogs simply ignoring author’s commands at Qaanaaq in 1994 (photo by the master of the dog team).

be useful to support the understanding of plate tectonics activities. It is estimated that collision between the northward moving Indian Plate and the stationary Eurasian Plate, which began 50 million years ago, created the Himalayan-Tibetan orogeny. The relative motion of these two major plates is rather complex, like commonly observed “finger rafting” in case of sea ice sheets. The bending, uplifting, and breaking of ice floes in front of the icebreaker’s bow were not unlike subduction zones: sliding and bending of the oceanic Pacific plate or the Juan de Fuca micro-plate interacting with the North American plate. In a similar way, the experimental observations on the vertical deflection of a floating ice sheet subjected to a few hours of loading, can be compared to a few decades of tectonic plate deflection caused by the man-made reservoir, Shivaji Sagar Lake, that started in 1962. Earthquakes began to occur in the vicinity of this reservoir formed by Koyna Dam in Maharashtra state, India. These earthquakes could be due to the reservoir induced deflection of the crust. The Elasto-Delayed-Elastic-Viscous (EDEV) [Sinha and Sinha, 2011] is a predictive material model for “quasi constant-structure” primary creep. It is used in developing an algorithm for predicting stress–strain diagrams at constant strain rates. Can geophysicists use

EDEV methodology to this living example of plate deflection and generation of earthquakes? There is no possibility to directly investigate the indentation processes occurring inside the earth’s lithosphere at various depths and temperature ranges. However, in-situ indentation tests in ice at high homologous temperatures (see definition in section 5.1.1) are routinely performed for evaluation of rate-sensitive mechanical properties for many ice-engineering applications. The in-situ evaluation procedures are known as borehole jack indenter (Figure 1.16). Knowledge gained from observations on cracking patterns, spalling effects, the degree and coverage of damage level during in-situ ice tests may be used for an understanding of indentation and micromechanical processes in plate tectonics. Finally, microscale features of grains and intergranular and intragranular activities in polycrystalline solids, including ice, may become comparable or analogous to inter- and intraplate activities of earth’s lithosphere material.

1.5. SEA ICE AND REMOTE SENSING Polar sea ice exists in remote locations under extreme climatic condition. Therefore, space-borne remote sensing is the only tool that can be used to monitor this ice at

INTRODUCTION 23

extensive spatial scale and reasonable revisit time. This potential was realized about 50 years ago. In the Preface of the notable book on Antarctic sea ice [Jeffries, 1998], the editor of the book refers to a few statements made by Lyn Lewis and Willy Weeks in a report published in 1971 by the Scientific Committee on Antarctic Research. One statement underlines a thought prevailed up to the 1950s that Antarctic sea ice is probably not any different than Arctic ice, so why bother studying it? But realizing that this may not be true to some extent, the authors of the report added “It is clear that future work will depend critically on the logistics available to allow surface observations beyond the fast ice edge at all seasons of the year. Of almost equal importance will be the development of instruments and recording equipment suited for use in the polar environment” [Lewis and Weeks, 1971]. One year later, in 1972, that need was realized when the first US space-borne microwave instrument for Earth observations; the Electrically Scanning Microwave Radiometer (ESMR) was launched on 10 December 1972. The foresight of the authors was true. ESMR was a milestone instrument for sea ice observations in the polar regions, following the success of previous satellites that used optical sensors. Satellite data of sea ice represent one of the longest Earth observation records from space. Since the early 1970s, many different sensors were developed and used effectively in routine surveillance of sea ice. Historical synopsis of satellite microwave remote sensing for sea ice is presented in section 8.1. Nearly half of this book is about applications of remote sensing data for retrieval of sea ice information. The basic premise of this endeavor is the significant difference in physical and radiometric properties between sea ice and open water. To a lesser extent, differences between properties of ice types also exist. They have been utilized (in a broad sense) to identify major ice types (e.g., FYI and MYI) and major surface features (e.g., leveled versus ridged ice) in remote sensing data (chapters 10 and 11). Satellite sensors operate in different bands of the electromagnetic spectrum: optical, infrared, and microwave (passive and active) (see sections 7.3–7.6). Information from these bands complement each other. Different sensors provide information at different scales. Optical sensors discriminate between sea ice and open water based on their contrast in albedo. Thermal infrared sensors use the difference in the physical temperatures (ice surface is usually colder than water surface and perennial ice is usually colder than seasonal ice). Passive microwave sensors use the difference in microwave emissivity between water and ice (water is radiometrically colder than ice). Radar sensors use the difference in backscatter, though it does not uniquely identify the water surface, as winddriven water surface roughness causes wide variability

of backscatter (section 9.3.1). In many situations, remote sensing data can hardly uniquely identify ice types because of the overlap between the radiometric or scattering values from different surfaces. However, integration with ancillary information has been used to improve the retrieved information. Many sea ice parameter retrieval algorithms have been developed using observations from different sensors. Examples include ice surface temperature, retrieved from thermal infrared sensors; ice concentration and extent from passive microwave; ice types from passive and active microwave; and ice kinematics, ice drift, and deformation from radar imagery data. The two most commonly used categories of sensors are active microwave (radar), which produces imagery data at fine resolutions of a few meters or tens of meters but with narrow swath (a few tens of meters), and passive microwave, which produces images at a coarse resolution of a few kilometers or tens of kilometers yet with very wide swath (a few thousands of meters). Various applications of sea ice information retrieval from remote sensing are introduced in Chapters 10 and 11. Ice parameter retrieval methods from remote sensing data have advanced. However, the use of remote sensing data in operational sea ice monitoring programs still depends on visual analysis of satellite imagery data. The reason is the strict requirement of the operational analysis for robustness. Algorithms may not be reliable under all possible ice conditions, but the operational environment requires nothing less than reliability and robustness. Visual analysis of the satellite imagery data is certainly subjective, but the operators can incorporate many pieces of information that cannot be incorporated in a quantitative retrieval algorithm. Examples include climatic information, recent history of the ice field, records of meteorological data, in addition to other heuristic rules used by expert ice analysts. The approach of visual analysis of satellite images to retrieve operational sea ice information will probably continue for many years, until a fleet of satellite constellation sensors is developed to acquire optical, thermal, and microwave data simultaneously. When the potential of SAR applications for sea ice was realized in the 1980s and the early development of spaceborne SAR systems started in the 1990s (namely, ESA’s ERS and the Canadian RADARSAT systems), funds to support development of new applications peaked. They were partly used to launch several field studies to relate ice and snow properties to the measured backscatter, and bring together researchers from different scientific disciplines to secure comprehensive plans for the data use. An unnoticeable positive development stemmed from such rising interest. That was the engagement of many researchers who came from several study fields (electrical engineering, electromagnetic wave modeling, computer

24

SEA ICE

sciences, physics, geography, etc.) into the realm of sea ice applications. Many researchers who had spent their career in developing hardware instruments or software methodologies for remote sensing data collection and analysis found themselves, at one point or another, involved in testing or validating their ideas using sea ice. Sea ice was one of the first major applications of remote sensing, especially microwave sensing, in polar regions. A few of those people later developed interest in sea ice as a phenomenon, and pursued serious research work in this field. Remote sensing tools for sea ice applications continue to develop to fill gaps in information. Ice thickness, for example, is a key parameter, which is needed by operational ice community as well as the climate modeling community. Altimeter systems have been developed to fill this gap (section 7.8). One issue that still hinders the retrieval of sea ice information from remote sensing data is the influence of the snow cover when metamorphosing under varying meteorological conditions, hence suppressing the contribution of the underlying ice. 1.6. MOTIVATION FOR THE BOOK WRITING This book combines information on two aspects of sea ice: physics and remote sensing. Research communities from the two disciplines do not usually interact; in general, the active members may not be aware of the details about each other’s realm of interest or work. On one hand, when geophysicists and materials scientists embarked on intensive studies of sea ice in the 1950s, they soon realized the need for tools to monitor ice conditions at larger scales. The tools of airborne and the space-borne remote sensing eventually became available by mid-1970s. On the other hand, when the remote sensing researchers started their work to retrieve sea ice parameters from the satellite observations, they too realized the importance of knowledge on physics of sea ice in order to support the interpretation of the data and the retrieval of ice information. Today, it is common to see basic information on ice physics cited in the literature of remote sensing, but it is not common to see the opposite. This book furnishes an opportunity for the two research communities to meet and perhaps learn a bit more about each other’s work. Knowledge about sea ice physics is relatively well established while knowledge about remote sensing of sea ice is still, and will continue to be, in developing mode in order to fulfill requirements for more crucial ice parameters retrieval needed to model the polar ice cover under the current trend of global warming. An example is the development of sensors and technologies for ice thickness estimation at spatial and temporal scales satisfactory to the operational needs. The remote sensing information has an edge over the information on ice physics, because the former is widely available on the internet but the latter

is not. Unsurprisingly, at this time of information explosion if the required information is not readily available on the internet, the common notion is that it does not exist! This book is an attempt to restore a balance between the two subjects in a single presentation, although the vast areas of engineering physics of ice remained untouched here. The book provides the basic physics of sea ice for the ice remote sensing researchers and operational communities, to develop better physical insight into the subject of their study/operations. It also provides a reasonable scope of applications of ice remote sensing for the ice physicists and geophysicists to comprehend the potential and limitations of the remote sensing applications. The argument about the importance of a satisfactory level of knowledge about sea ice physics for remote sensing research applies laterally to research in other icerelated disciplines such as engineering, climatology, and glaciology. However, judging from several recent publications, it seems that researchers in these fields tend to remain confined within the level of knowledge about ice physics that they have been circulating since the 1970s and 1980s. One of the reasons for this tendency is the fact that those researchers usually come from different scientific backgrounds. For example, ice-related engineering problems are dealt with by engineers trained mainly in the broad fields of civil, mechanical, or aerospace engineering, to whom, the ice is a “nuisance.” Glacier ice and to some extent sea ice is studied mostly by the geographers trained in the vast aspects of geography, but not necessarily physics. None of these communities, except for the physicists, treat snow and ice as extremely hightemperature materials. The promise of this book is to present a fair amount of sea ice physics, and outline links with remote sensing in ways to comprehend the physicsbased retrievals of ice information from the data. This should help researchers to set more realistic goals and informed plans for their work. Remote sensing of sea ice has been conducted by researchers from a wide range of scientific background that encompasses computer sciences, environmental sciences, geography, and electrical and civil engineering. The book is an attempt to bring out some progress made in the field of ice physics that could expand the scope of knowledge of the researchers and operators in the different aspects of sea ice remote sensing, and hopefully influence their work. The book is intended to reach out to a variety of sea ice audiences who study different aspects of the ice phenomena related to physics, remote sensing, mechanical behavior, climatic impacts, operational monitoring, etc. For many readers who seek applications of remote sensing, it is not crucial to gain deep understanding of ice as a material. For that reason, the authors have tried to

INTRODUCTION 25

present the material in a simplified and gradual manner whenever possible so the reader can gain the amount of information that suits his/her purpose. To facilitate this presentation approach, definitions are introduced constantly with cross referencing, treatments of both physics and remote sensing material start with basic principles, enough number of illustrations is used, derivation of commonly used equations is presented, and many references are provided. In short, the book is intended to be of educational value.

1.7. ORGANIZATION OF THE BOOK Materials related to physics of ice are presented in Chapters 2 to 6; while remote sensing materials are presented in Chapters 7 to 12. Chapter 13 focuses on the impacts of climate change on polar sea ice. It demonstrates uses of remote sensing in revealing those impacts. Chapter 2 is about the physics and physical processes of sea ice. It starts with a section on key properties of freshwater and sea water followed by a section of phase diagram of sea ice. Freezing mechanisms and thermodynamic ice growth are addressed in two sections, where thermodynamic growth are introduced to estimate ice growth rate under different atmospheric, oceanic, and snow conditions. Key processes in sea ice are presented in a dedicated section to address compositional supercooling of sea water, dendritic ice–water interface, and entrapment of brine within ice crystals. Two important processes are discussed in details: the brine drainage throughout the lifetime of sea ice and the grain and subgrain boundaries in crystalline structure. The importance of the subgrain boundaries in determining mechanical and thermal properties of sea ice are highlighted. Sections on ice deformation, decay, and aging are later introduced. The chapter is concluded with two sections on sea ice classes and regimes. The former addresses ice classes based on different criteria and the latter discusses the ice regimes of polynyas, pancake ice, ice edge, marginal ice zone, and forms of floating ice of land-origin. Chapter 3 presents information on basic physical properties of sea ice. It starts with typical values of key properties of ice and snow that are commonly used in radiation and climate modeling studies. This is followed by presenting detailed measurements of temperature, salinity and density profiles in sea ice cover. Calculations of the volume fraction of the four sea ice constituents: brine, solid salt, pure ice and air are presented in a dedicated section. Thermal properties: conductivity, specific heat and latent heat are introduced in another section. The last section addresses the dielectric constant of sea ice from theoretical and measurements viewpoints. It includes detailed derivation of a dielectric mixing model for

two-phase sea ice composed of pure ice (host medium) a single inclusion (brine or air). Chapter 4 addresses laboratory techniques for revealing the polycrystalline structure of ice. Following an introduction containing information on polarized light, polarizing sheets, large field-of-view polariscopes and the birefringent properties of ice, the chapter moves on to present detailed techniques for preparing thin sections of ice and snow. Special emphasis is given to DMT. The procedures for examining and photographing thin sections under polarized and scattered light, or their combinations, are also presented. The chapter is concluded with a section on advanced techniques for revealing fine crystallographic microstructural features. This includes descriptions of sublimation etch bits, thermal and chemical etching, and the dual processes of etching/replicating technique. This is resent in conjunction with using scanning electron microscopy (SEM) for examinations of subgrain boundaries and substructures involving line defects or dislocations within the sea ice polycrystalline structure. Chapter 5 presents detailed information on the polycrystalline structures of freshwater ice and sea ice. It starts with definitions of basic terms related to polycrystalline ice, and then covers briefly the morphology of ice microstructure. This is followed by classification and characteristics of ice crystalline structure based on extending crystallographic classification of freshwater ice to sea ice, which is the main theme of the chapter. A section follows to present many examples of crystalline structures of freshwater and sea ice using photographs of thin sections prepared with the DMT method. Examples show different crystallographic classes of natural ice including structures of seasonal sea ice type (granular, frazil, columnar, platelet ice structure with entrapped inclusion), and perennial sea ice (hummock and melt pond ice). Crystalline structure provides clues on meteorological and oceanic conditions under which ice has been formed and grown. Information on biomass accumulation at the bottom of the sea ice is presented in a dedicated section, and the chapter is concluded with a section on geometric characteristics of microstructural features, e.g., crystal size, brine pockets, and air bubble. Chapter 6 is a new chapter introduced in this second edition. It introduces major field expeditions to study sea ice in the Arctic. Most of them were international campaigns. They are arranged chronologically, beginning with AIDJEX experiment in 1975 and ending with the recent MOSAiC experiment in 2019–2020. More focus is placed on two Canadian field programs, the Mould Bay experiment (1981–1984), and a field experiment in 1982 on MYI attached to an ice shelf in the Arctic. Since only limited information from these two experiments has been published, the chapter disseminates scientific, technical, and human experience information under the title

26

SEA ICE

“stories that were never told.” In general, the chapter shows how the objectives of more recent campaigns are built on experience from previous campaigns. Chapter 7 marks the beginning of the remote sensing part of the book. The scope of the material in this chapter is designed to appeal to researchers and users of remote sensing data, who want to develop quick acquaintance with issues of physics of radiation that are outside their domain of experience. It starts with a few general principles of satellite remote sensing, and then a section on general properties and processes of electromagnetic waves (namely reflection, transmission, absorption, scattering, and emitted radiation) follows. The section also introduces theoretical material on brightness temperature, emissivity and penetration depth. Four sections follow about optical, thermal infrared, passive microwave, and active microwave (radar) remote sensing. Focus is placed on imaging radar sensing, since this is the prime data source for operational sea ice monitoring. This includes radar equations, coherency, polarization, and scattering mechanisms of radar signals. Two subsections are dedicated to multichannel and polarimetric SAR. The radar polarimetry part includes formulation of polarimetric measurements and derived parameters. Linking those parameters to the scattering mechanisms is addressed and a recommendation to develop scattering-based to replace age-based ice classification is justified. The chapter also includes two sections on scatterometer and altimeter systems, followed by a final section on radiative processes in the atmosphere, ocean, and snow cover. These three media affect the satellite observations. The material in this chapter provides sufficient background to facilitate comprehending the information presented in Chapters 9–11. Chapter 8 includes information about current satellite sensors used for sea ice and snow monitoring and parameter retrieval. It starts with a historical synopsis of satellite remote sensing, then advances to introduce major sensors used in sea ice applications from optical, thermal infrared, passive microwave, imaging radar, scatterometer, and altimeter systems. Chapter 9 presents data sets on radiative and scattering properties of sea ice, its snow cover and surrounding open water. This includes visible and near infrared reflectance, albedo, as well as microwave brightness temperature, emissivity, penetration depth, and radar backscatter from single-channel, dual-channels, and polarimetric SAR. Most of the data are obtained from several satellite observations, but airborne and ground measurements are also used. The data can serve as benchmarks to guide the process of ice parameter retrieval from remote sensing data. Chapter 10 is the first of two chapters that present methods of sea ice parameters/features retrieval (including snow cover) from remote sensing data. The chapter

addresses retrieval of ice surface information. This is presented under three categories. The first includes mechanically-generated surface deformation aspects (rafting, ridging, rubble, brash ice, cracks and leads, as well as kinematic ice processes). The second includes thermally-generated surface features (stages of surface melt and frost glowers), and the third is about meteorologically-generated features (polynya and snow depth). For each feature in each category, the most suitable sensors are identified and the most common retrieval methods are described. The advantage of combining observations to enhance the retrieved parameters is demonstrated. Chapter 11 addresses retrieval of sea ice geophysical parameters. This includes ice types, concentration, extent, thickness, surface temperature, ice age, and ice motion and kinematics. Whenever possible, methods for each parameter are grouped according to the type of observation; namely optical, thermal infrared, passive microwave, and radar. Ice type is an important operational parameter and it can be retrieved from all categories of remote sensing observations. Special focus is placed on retrieval of ice concentration, as this has been the most successfully retrieved and commonly used parameter in sea ice monitoring programs and climate modeling studies. Detailed descriptions of four commonly used methods are presented. Some critical issues that affect the accuracy of parameter retrieval are also presented. Ice thickness retrieval from TIR, PM altimeter, and SAR systems is introduced. Methods of ice surface temperature using TIR and passive microwave are covered. Products of sea ice age are presented briefly. The section on sea ice motion and kinematics cover methods of motion tracking and operational products. The information in this chapter demonstrates the potential and limitations of sea ice remote sensing applications. Chapter 12, introduced in this new edition, is on modeling microwave emission and scattering from snowcovered sea ice. It is written by Dr. Rasmus T. Tonboe of the Technical University in Denmark. The chapter describes the use of forward model inversion for estimating sea ice snow cover. It starts with gross features of forward modeling and then presents briefly key features of radiative transfer and modeling of microwave emission and scattering from sea ice. A section is dedicated for input to forward models (primary, secondary, and tertiary input parameters). Another section addresses the implementation of an altimeter model to study the impact of saline snow on the backscatter. Noise in sea ice concentration from using combination of thermodynamic, atmospheric, ocean and sea ice emission models is presented. The chapter is concluded with a brief presentation of inverse modeling. Chapter 13 is another new chapter in this second edition. It covers impacts of climate change on sea ice in

INTRODUCTION 27

the two polar regions. Comparison of the impacts between the two regions is a central theme of this chapter. The chapter starts by delineating differences between the two polar regions in terms of geography and main sea ice features. It then summarizes recent changes in sea ice (e.g., extent, thickness, volume, age, dynamics and icebergs) in response to the current trend of global warming. As the response is different between the two polar regions, with insignificant changes observed in the Antarctic region, reasons are searched by exploring couplings between sea ice and environmental factors in each region. Interactions of sea ice with atmospheric and oceanic forcing in each region are presented, and interactions between Antarctic sea ice and ice shelves are highlighted. 1.8. REFERENCES Breivik, L.A. et al. (2010) Remote sensing of sea ice: peer review community white paper, In: Proceedings of OceanObs09 Conference, Venice, Italy, September, 2009, vol. 2, ESA Publication WPP-306, doi:10.5270/OceanObs09. Budikova, D. (2009) Role of Arctic sea ice in global atmospheric circulation: a review, Global and Planetary Change, 68, pp. 149–163. Cammaert, A.B. and Muggeridge, D.B. (1988), Ice interaction with offshore structures, New York: Van Nostrand Reinhold, 432 p. Carsey, F.D., ed. (1998) Microwave Remote Sensing of Sea Ice, Geophysical Monograph, vol. 68, Washington, DC, USA, American Geophysical Union, 462 p. Carmack, E.C. (1986) The circulation and mixing of ice-covered waters, In: Untersteiner, N., ed. Geophysics of Sea Ice, New York: Plenum, pp. 641–712. Cassano, E.N. et al. (2014) Atmospheric impacts of an Arctic sea ice minimum as seen in the Community Atmosphere Model, International Journal of Climatology, 34, pp. 766–779. Available from: doi: 10.1002/joc.3723. Comiso, J. (2010) Polar oceans from space, Atmospheric and Oceanographic Sciences Library, vol. 41, New York: Springer-Verlag, 522 p. Comiso, J.C. et al. (2001) Seasonal and interannual variability of the Odden ice tongue and a study of environmental effects, Journal of Geophysical Research, 106(C5), pp. 9093–9116. Available from: doi:10.1029/2000JC000204. Duhl, D.N. (1987) Directionally solidified superalloys, In: Sims, C.Y., Stoloff, N.S. and Hagel, W.C., eds., Superalloys II, Chapter 7, New York: John Wiley, pp. 189–214. Frederking, R. and Sinha, N.K. (1977) Ice action on wharf at Strathcona Sound, In: Proceedings of 4th International Conference on Port and Ocean Engineering under Arctic Conditions (POAC 77), Memorial University of Newfoundland, St. John’s, NFL, pp. 707–717. Häkkinen, S. (1999) A simulation of thermohaline effects of a great salinity anomaly, Journal of Climate, 12, pp. 1781–1795. Hall, D.K. and Martinec, J. (1985) Remote sensing of ice and snow, New York: Chapman and Hall Ltd, NY 10001, 189 p.

Haykin, S. et al. (1994) Remote sensing of sea ice and icebergs, Wiley Series in Remote Sensing and Image Processing, vol. 17, New Jersey: John Wiley and Sons, ISBN-13: 978-0-47155494-3. Heygster, G. et al. (2012) Remote sensing of sea ice: Advances during the DAMOCLES project, The Cryosphere, 6, pp. 1411–1434. Available from: doi:10.5194/tc-6-1411-2012. Horner, R., ed. (1985) Sea ice biota, Boca Raton, FL: CRC Press, ISBN-13: 978-0-849-36578-2. Jackson, C.R. and Apel, J.R., eds. (2004) Synthetic Aperture Radar: marine user’s manual, US Department of Commerce, National Oceanic and Atmospheric Administration, Washington, DC, ISBN-13: 978-0-160-73214-0, 464 p. Jefferies, M.O., ed. (1998) Antarctic sea ice: physical processes, interactions and variability, Antarctic Research Series, 74, Washington, DC, USA, American Geophysical Union, 407 p. Johannessen, O. et al. (2007) Remote sensing of sea ice in the Northern Sea route: Studies and applications, Springer-Praxis Books in Geophysical Sciences, ISBN- 10: 3-540-24448-4. Johannessen, O.M. et al., eds. (2020) Sea Ice in the Arctic: Past, Present and Future, Springer Polar series. Springer International Publishing. ISBN-13: 978-3-030-21300-8. Available from: doi: 10.1007/978-3-030-21301-5, 575 p. Jones, S.J. et al., eds. (1991) Ice-structure interaction, IUTAM Symposium, 1989, St. John’s, Canada, Berlin Heidelberg: Springer-Verlag, 738 p. Kattenberg, A. et al. (1996) Climate models—Projections of future climate, In: Houghton, J. T. et al., eds. Climate Change 1995: The Science of Climate Change, Cambridge, UK: Cambridge University Press, pp. 285–357. Kwok, R. (2010) Satellite remote sensing of sea-ice thickness and kinematics: a review, Journal of Glaciology, 56(200), pp. 1129–1140. Kwok, R. and Sulsky, D. (2010) Arctic Ocean sea ice thickness and kinematics: satellite retrievals and modeling, Oceanography, 23(4), pp. 134–143. Laidre, K.L. and Regehr, E.V. (2017) Arctic marine mammals and sea ice, In: Thomas, D.N., ed. Sea ice: third edition, Chapter 21, Wiley Blackwell, pp. 516–533. Legendre, L.L. et al. (1992) Ecology of sea ice biota 2. Global significance, Polar Biology, 12, pp. 429–444. Lewis, E. L. and Weeks, W.F. (1971) Sea ice: some polar contrasts in Antarctic ice and water masses, In: Deacon G., ed. Scientific Committee on Antarctic Research, pp. 22–34. Lifesaving Society (1998) ICE—the winter killer, A Resource Manual About Ice Safety, Ice Rescue, Lifesaving Society, National Office, 287 McArthur Avenue, Ottawa, Ontario, Canada K1L 6P3. Lüpkes C. et al. (2008) Influence of leads in the sea ice on the temperature of the atmosphere boundary layer during polar night, Geophysical Research Letters, 35(L03805). Available from: doi:10.1029/2007GL032461. Maykut, G.A. (1978) Energy exchange over young sea ice in the Central Arctic, Journal of Geophysical Research, 83(C7), pp. 3646–3658. Maykut, G.A. (1982) Large-scale heat exchange and ice production in the central Arctic, Journal of Geophysical Research, 87, pp. 7971–7984.

28

SEA ICE

Meier, W.N. et al. (2011) Sea ice, In: Meier, W. et al., eds. Snow, Water, Ice and Permafrost in the Arctic (SWIPA): Climate Change and the Cryosphere, Chapter 9, Arctic Monitoring and Assessment Programme (AMAP). Meier, W.N. et al. (2014) Arctic sea ice in transformation: A review of recent observed changes and impacts on biology and human activity, Reviews of Geophysics, 51, pp. 185–217. Available from: doi:10,1002/2013RG000431. Nakawo, M. and Sinha, N.K. (1984) A note on the brine layer spacing of first-year ice, Atmosphere-Ocean, 22(2), pp. 193–206. Nakawo, M. and Sinha, N.K. (1981) Growth rate and salinity profile of first-year sea ice in the high Arctic, Journal of Glaciology, 27, pp.315–330. Paterson, W.S.B. (1969) The physics of Glaciers, Oxford: Pergamon Press, 250 p. Petrich, C. and H. Eicken (2009) Growth, structure and properties of sea ice, In: Thomas, D.N. and Dieckmann, G.S., eds. Sea ice: 2nd edition, Wiley-Blackwell, pp. 22–77. Pole, K. (1995) Stars on ice, Wings Magazine, No. 1, pp. 28–29. Pounder, E.R. (1965) The Physics of ice, Oxford: Pergamon Press. Rantanen, M. et al. (2022) The Arctic has warmed nearly four times faster than the globe since 1979, Communications Earth and Environment, 3 (168), doi.org/10.1038/s43247-022-004983 | www.nature.com/commsenv Reese, W.G. (2006) Remote sensing of snow and ice, CRC Press, 123 p. Sanderson, T.J.O. (1988) Ice mechanics: Risks to offshore structures, London: Graham and Trottman, 253 p. Sandven, S. (2008) Sea ice monitoring in European Arctic Seas using a multisensor approach, In: Barale and Gade, eds. Remote Sensing of the European Seas, Springer Science and Business Media B.V., pp. 487–498. Sandven, S. and Johannessen, O.M. (2006) Sea ice monitoring by remote sensing, In: Gower, J., ed. Remote Sensing of the Marine Environment, Manual of Remote Sensing, Third Edition, Chapter 8, vol. 6, Maryland, USA: American Society for Photogrammetry and Remote Sensing, pp. 241–283. Sinha, N.K. (1995) Sea Ice Landing Strip for Boeing-727 in Northern Greenland, In: Proceedings of 14th International Conference on Offshore Mechanics and Arctic Engineering (OMAE), 18–22 June, 1995, Copenhagen, Denmark,

American Society of Mechanical Engineers (ASME), New York, vol. 4, pp. 273–279. Sinha, N.K. (2009) Stress exponent and primary creep parameters using single specimen and strain relaxation and recovery test, Materials Science and Engineering: A, vol. 510–511, pp. 450–456. Sinha, N.K. and Sinha, S. (2011) High-temperature yield strength and its dependence on primary creep and recovery, Materials Science and Engineering: A, 528(16–17), pp. 5366–5378. Sinha, N.K. and Sinha, S. (2022) Engineering physics of high temperature materials: metals, ice, rocks, and ceramics, New Jersey, USA: Wiley, ISBN-13: 978-1-119-42048-4, 432 p. Sinha, N.K. and Nakawo, N. (1981) Growth of first-year ice in Eclipse Sound, Baffin Island, Canada, Canadian Geotechnical Journal, 18, pp.17–23. Sinha, N.K. et al. (1996) Floating ice thickness for aircraft operations, Report published by Civil Engineering Services, Engineering and Maintenance, Safety and Technical Services, Transport Canada. Smith, G.W. (2009) Weather stations in the Canadian north and sovereignty, Journal of Military and Strategic Studies, Spring 2009, 11(3), pp. 1–63. Stensrud, D.J. (2007) Parameterization schemes: keys to understanding numerical weather prediction models, Cambridge University Press, ISBN-13: 978-0-521-86540-1, p. 137. Steenfelt, J. (2016) Ice loads on structures in the Baltic environment, In: Proceedings of 13th Baltic Sea Geotechnical Conference, Lithuania, 22–24 September 2016. Available from: doi:10.3846/13bsge.2016.018. Stroeve, J.C. et al. (2007) Arctic sea ice decline: faster than forecast, Geophysical Research Letters, 34(L09501). Tabata, T. and Ono, N. (1962) On the crystallographic study of several kinds of ice, Low Temperature Science Series 20, pp. 199–214 (text in Japanese). Thomas, D.N., ed. (2017) Sea Ice: third edition, New Jersey, USA: Wiley Blackwell, ISBN-13: 978-1-118-77838-8, 644 p. Untersteiner, N., ed. (1986) The geophysics of sea ice, NATO ASI Series B, vol. 146, New York: Plenum Press, 1196 p. Weeks, W.F. (2010) On sea ice, Fairbanks, Alaska, USA: University of Alaska Press, ISBN-13: 978-1-602-23079-8. Wettlaufer, J.S, Fasj, J.G. and Untersteiner, N., eds. (1999) Ice physics and natural environment, NATO ASI Series, vol. 56, London: Springer, ISBN-13: 978-3-642-64226-5, 355 p.

2 Ice Physics and Physical Processes

2.1

2.2

Prior to Freezing: About Freshwater and Seawater................ 30 2.1.1 Molecular Composition of Water................................ 30 2.1.2 Seawater Salinity ......................................................... 31 2.1.3 Seawater Density ......................................................... 32 Phase Diagram of Sea Ice....................................................... 33

2.3

Initial Ice Formation .............................................................. 33 2.3.1 Freezing Processes in Freshwater and Seawater .......... 33 2.3.2 Initial Formation of Ice Crystals and Frazil Ice .......... 35

2.4

Sea Ice Growth ....................................................................... 37 2.4.1 Lateral Ice Growth ...................................................... 37 2.4.2 Vertical Ice Growth (Congelation Ice)......................... 38 2.4.3 Superimposed Ice......................................................... 39 2.4.4 Thermodynamic Ice Growth........................................ 40 2.4.4.1 Simplified Models of Sea Ice Growth............41 2.4.4.2 Effect of Snow On Sea Ice .............................45 2.4.4.3 Effect of Oceanic Heat Flux ..........................46 2.4.4.4 Effect of Surface Ablation .............................46 Processes in Ice ....................................................................... 47 2.5.1 Compositional (Constitutional) Supercooling At the Ice–Water Interface..................... 50 2.5.2 Dendritic Ice–Water Interface and Entrapment of Brine Within Sea Ice............................ 51 2.5.3 Grains and Subgrains In Sea Ice.................................. 53 2.5.4 Brine Pockets Formation, Contents and Distribution In Sea Ice................................................. 54

2.5

2.6

2.7

2.8

Development of a sea ice cover at any previously ice-free location encompasses a number of phases; (1) nucleation of ice crystals and initial formation including consolidation that depends on the prevailing weather and atmospheric conditions, (2) lateral growth followed by vertical growth, known as “ice congelation,” (3) ice motion and deformation stimulated by wind and current, (4) melt and decay, and finally (5) ice aging, which features sea ice that has survived at least one summer melt. Wind, ocean current and wave-induced motion and associated strain (hence stress) stimulate deformation in different forms. While the majority of sea ice is mobile (floating

2.5.5 Salinity Loss During Sea Ice Growth .......................... 58 2.5.5.1 Initial Rapid Salt Rejection At the Ice–Water Interface ..................................... 59 2.5.5.2 Subsequent Slow Salt Rejection from the Bulk Ice .........................................................61 Ice Deformation ..................................................................... 67 2.6.1 Rafting of Thin Ice ...................................................... 69 2.6.2 Ridging of Thick Ice.................................................... 70 2.6.3 Formation of Ice Rubble Field.................................... 73 2.6.4 Fractures in Ice Cover ................................................. 74 Ice Decay and Aging .............................................................. 76 2.7.1 Ice Decay..................................................................... 76 2.7.2 Ice Aging ..................................................................... 80 Sea Ice Classes ........................................................................ 84

2.9

Sea Ice Regimes...................................................................... 85 2.9.1 Polynyas ...................................................................... 87 2.9.2 Pancake Ice Regime..................................................... 90 2.9.3 Marginal Ice Zone and Ice Edge ................................. 93 2.9.3.1 Marginal Ice Zone .........................................93 2.9.3.2 Ice Edge.........................................................94 2.9.4 Ice of Glacier Origin.................................................... 95

2.10

References............................................................................... 99

on its own melt), grounded and/or fast ice attached to the shore, ice wall, ice shelf or iceberg is also possible. This chapter starts with relevant information about water (freshwater and seawater), which affect the ice formation (section 2.1) and the phase diagram of sea ice (section 2.2). It then addresses the subject of initial ice formation (section 2.3) and the lateral and vertical ice growth (section 2.4). Thermodynamic ice growth models are presented in section 2.4.4. Within the vertical growth part, the processes of entrapment of inclusions (salts and air) within the sea ice and the continuous desalination (brine drainage) are covered (section 2.5). Related to salt

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 29

30

SEA ICE

entrapment in sea ice, the process of compositional supercooling is presented in section 2.5.1 and the important features of the dendritic sea ice–water interface (which actually discriminate between sea ice and freshwater ice) is covered in section 2.5.2. Discussions of the concepts of grain and subgrain boundaries within the crystallographic structure of sea ice are presented in section 2.5.3 and brine pockets formation in section 2.5.4. Here, important links to mechanical properties of saline ice are introduced. Most of the information is obtained from analyzing micrographs of sea ice thin sections. Sea ice loses salinity through a process known as brine drainage throughout its life time. This process has two parts, rapid desalination at the ice–water interface as the interface propagates into the underlying seawater, and slow desalination as the ice continues to grow. The two processes are covered in section 2.5.5. The chapter then continues with a section on ice deformation (section 2.6) followed by a section on ice decay and aging (section 2.7). Ice decay refers to the disintegration of the ice floes/sheet and then the complete melting during summer melt season. Ice aging refers to the age of ice (in years) when it survives one or more summer melt period. Ice classes are addressed in section 2.8 and a few commonly known ice regimes, namely polynyas, pancake ice, marginal ice zone, ice edge, and ice of land origin (i.e., icebergs and ice islands) are presented in section 2.9. 2.1. PRIOR TO FREEZING: ABOUT FRESHWATER AND SEAWATER Since ice originates from water, it would be reasonable to start the presentation with relevant properties of freshwater and seawater. Freshwater lakes and rivers contain very little dissolved impurities (less than 0.5% by mass) and their effects on the growth of ice can be neglected for all practical purposes. That is not the case for seawater. The two properties of seawater that are directly related to sea ice formation are its salinity and density. 2.1.1. Molecular Composition of Water The isotopic composition of water in case of freshwater ice may not be considered as important for ice formation process, but it may play some role in case of seawater and ice formed from it. It is relevant because snow cover on sea ice may have a slightly different isotopic composition than the seawater. Consequently, this property may be used in distinguishing snow ice, formed from solidification of snow saturated with meltwater, from frazil ice formed from the seawater (section 5.3.3). The isotopic composition obtained from glacier ice cores at decadal resolution

are routinely used to infer climate record because the landbased ice bodies developed from snow depositions in the past [Von Grafenstein et al., 1999]. A brief account of the molecular composition of water, including its isotopic composition, is introduced in the following. The atomic and the molecular structure of water in all its phases (liquid and solid) have intrigued scientists for many centuries. Although the chemical formula of water is very simple, its atomic structure, depending on temperature and mechanical or electrical forces is extremely complicated and difficult to study. The stoichiometric composition of water has been investigated by numerous investigators over a very long period of time and has been documented thoroughly in many reports and books. It was known for a long time that if two atoms of hydrogen are combined with one atom of oxygen under the same temperature and atmospheric pressure, then a molecule of water vapor forms. According to Avogadro’s hypothesis this leads to the basic molecular structure of pure water consisting of two atoms of hydrogen and one atom of oxygen, or simply H2O. The ratio of the combining volumes of oxygen (O2) and hydrogen (H2) at 0 C and a pressure of 0.76 m of Hg is O2/H2 = 1/2.00288. Pure water, however, may be divided into several types depending on isotopes of oxygen and hydrogen. Hydrogen has three isotopes: 1H (known as H), 2H (known as deuterium, D), and 3H (known as tritium), which is radioactive and its half-life is 12.5 years. Oxygen, on the other hand, has six isotopes 14O, 15O, 16O, 17O, 18O, and 19O (oxygen 16O is simply known as O). The three isotopes of 16O, 17O, 18O are stable, whereas 14O, 15O and 19O are radioactive but short-lived. The combination of oxygen and hydrogen atoms leads to six stable isotopes of water formed from stable atoms of hydrogen and oxygen. The isotopic content of water in oceans, lakes, and rivers varies depending upon its origin. It contains about 99.73% of 1H216O, 0.2% of 1H218O, 0.04% 1H217O, and 0.03% of 1H2H16O [Shatenshtein et al., 1960]. The last one is simply termed HDO. However, if both the hydrogen atoms in this compound are deuterium (2H), i.e., the molecule is 2H216O or D2O, then the compound is known as heavy water. Based on the above percentages ordinary water, which is generally referred to as H2O is mainly composed of 1H216O. The isotopic water composition has an impact on evaporation and consequently on the composition of snowflakes in the upper atmosphere. At a given temperature, the evaporation of the two major isotopes of water, 1 H216O and 1H218O, depend on their masses. Naturally, the heavier water evaporates at a slower rate than the lighter one. As the temperature rises, the proportionality of these two varieties of water increases. Water containing the heavy oxygen 18O evaporates more under warmer

ICE PHYSICS AND PHYSICAL PROCESSES 31

atmospheric temperature, making higher percentage of the relevant water isotope in the falling snow. This allows for the retrieval of paleoclimatic information to understand the evolution of the cryosphere. Although this information is lost when snow falls on land, rivers, lakes, or oceans, it is maintained when snow is deposited on areas that could sustain it for long periods. Such areas include ice caps and sheets which are formed from the massive accumulation of snow on flat plateaus such as those found in several islands in Canada (i.e., Ellesmere Island, Penny Icecap in Baffin Island), Greenland, and Antarctica. A chemical analysis of deep ice cores extracted from ice caps on top of flat bedrock, where ice had settled with no side movement, provides insight into the climatic conditions of the period related to the analyzed section of the core. This can be achieved by measuring δ(18O), which is expressed as the fractional difference between the ratio 18 O/16O in the sample and the ratio in “standard mean ocean water” (SMOW) measured in percent. In general, the values of δ(18O) in ice cap snow are negative. Since 18 O is heavier than 16O then water molecules containing the former evaporate proportionately less than the latter. Nevertheless, under warmer atmospheric temperatures, more molecules containing 18O evaporate. This means that higher negative values of δ(18O) found in the precipitated snow are likely associated with a warmer climate although there are also a number of non-temperature effects that can alter this parameter [Hobbs, 1974]. Since each depth segment of an ice cap core has originated in the past, chemical analysis of glacier ice cores at different depths can then be used to identify relatively warmer periods of snow deposition. Because snow has a lower content of 18O than 16O compared to seawater, granular ice, formed from slush of snow saturated with meltwater, can be chemically distinguished from frazil ice formed from slush containing seawater and frazil crystals. These two types of ice may not be identified unless forensic type of microstructural analyses are carried out on both horizontal and vertical sections of the ice specimen. Jeffries et al. [1994] used both crystalline structure and oxygen isotopic composition analysis of sea ice cores from the Ross Sea in the Antarctic to determine the amount of snow that contributed to the development of sea ice. In fact, for first-year ice (FYI) in Okhotsk Sea, Toyota et al. [2007] found measurable differences in the values of δ(18O) for snow, frazil and columnar ice. 2.1.2. Seawater Salinity Seawater holds the same isotopic water composition as fresh water, yet it contains a considerable percentage of dissolved salts and gases. Water salinity is the most relevant property to sea ice formation, composition, and growth. It is usually measured as the ratio of the weight

of salts (in grams) dissolved in 1000 g of seawater. Hence, it is usually presented in parts per thousands (ppt or ‰). Alternatively, oceanographers define ocean salinity in terms of the Practical Salinity Unit (PSU), which is the conductivity ratio of a seawater sample to the conductivity of a solution of Potassium Chloride (i.e., the standard solution for measuring electrical conductivity). The salinity measure in ‰ is adopted in this book. Freshwater contains salinities less than 0.5‰, while brackish water bodies have salinities more than that of freshwater, but less than the usual salinity of seawater. Thus brackish water is defined as water having salinities in a wide range, between 0.5‰ to 29‰. It is usually found in estuaries, inland seas, or lakes. According to this nomenclature, saline water can be defined as water with salinity ranging from 29‰ to 50‰. Brine is a term that describes water with salinity higher than 50‰. Seawater salinity is 35‰ on average in most marine areas though slightly higher values (36‰) are observed in some regions in the Atlantic Ocean and Indian Ocean and slightly less values (34‰) are observed in the polar region because of less evaporation. Within the Canadian Archipelago, water salinities are often found to be in the range of 30‰. Salinity of seawater is affected by the relative amount of precipitation and evaporation as well as the mixing with fresh water in the vicinity of river mouths. If a large river is emptying its water into a sea, the local seawater salinity is significantly less than the typical value of 35‰. Variation of salinity is one criterion for seawater classification. It is also revealed in different mechanisms of sea ice formation as well as crystallographic structure and physical properties of ice. The primary salt dissolved in seawater is sodium chloride (NaCl), but other salts exist such as sodium sulfate (Na2SO4.10H2O), magnesium sulfate (MgSO4) and magnesium chloride (MgCl2.12H2O and MgCl2.8H2O). The ionic proportion of chloride, sodium, sulfate, and magnesium in these compounds (regardless of the water salinity) is close to 55.03%, 30.59%, 7.68%, and 3.68%; respectively. The rest of the ionic proportion; namely 3.02%, is composed of calcium (Ca), potassium (K), bromine (Br), cobalt (CO) and other elements in negligible amounts. Solutes in seawater are rejected during the process of solidification. Most are rejected to the water at the ice– water interface. Some are entrapped as inclusions within the ice mass, known as brine pockets (section 2.5.4). They are located, as will be shown later, along boundaries and sub-boundaries of ice crystals. The entrapped brine naturally remains in sea ice in thermal equilibrium with the surrounding ice. The temperature of the ice determines the salinity of the liquid in the brine pockets. As the temperature continues to decrease, different solutes in the brine start to precipitate at different temperatures.

32

SEA ICE

Precipitation of different salts occurs at different temperatures as shown in Table 2.1. 2.1.3. Seawater Density Another property of fresh water, which is quite relevant to the onset of ice formation, is its density. Much like any other material, water is subject to thermal expansion as its temperature rises, or contraction as the temperature decreases. However, the behavior of freshwater is very interesting at temperatures close to its freezing point in the range between 4 C–0 C. Lowering the temperature below 4 C causes the density not to increase, as expected, but to decrease as shown in Figure 2.1. The maximum density (at 4 C) is 999.972 kg/m3. This property affects the ice formation and growth processes as explained later. Another property of water, though not unique to it, is the fact that it can be supercooled far below the normal freezing point without solidification. As the water is supercooled below the normal freezing point, the density continues to decrease with the decrease in temperature even through the supercooling status as shown in Figure 2.1. This behavior affects the freezing mechanism

of freshwater and seawater, as will be seen later in the next section. Dissolved salts in water depress the temperature of the maximum density below that of freshwater (about 4 C) as well as the freezing point. For seawater, both the temperature of maximum density and the freezing point of salty water decrease almost linearly as the water salinity increases, as illustrated in Figure 2.2. At a critical salinity of 24.69‰, the two temperatures are equal to −1.32 C. For salinities higher than the critical value, the freezing temperature becomes higher than the temperature of maximum density. At salinity of 35‰, the two temperatures are −1.88 C and −3.5 C; respectively. It is appropriate to mention here that the density of seawater at the surface of the ocean varies between 1,020 and 1,029 kg/m3 depending on the salinity of the water. This is more than the density of freshwater ice, which is about 917 kg/m3. Yet, the density of newly formed sea ice is only marginally less than 1000 kg/m3 due to high brine content. This is lighter than the saline water and that is how sea ice

1001

999.87 kg/m3

999.97 kg/m3 at 3.98°C

1000

Salt name Calcium carbonate Sodium sulfate Magnesium chloride Sodium chloride Magnesium chloride Calcium chloride

Composition

Precipitating Temp( C)

CaCO3 6H2O Na2SO4 10H2O MgCl2 8H2O NaCl 2H2O MgCl2 12H2O CaCl2 6H2O

−2.20 −8.20 −18.0 −22.9 −36.8 −55.0

Density, kg/m3

Table 2.1 Major salts in sea ice and their precipitation temperatures.

999 998 997

Supercooled region

996 995 –10

0

10 Temperature, °C

20

Figure 2.1 Temperature dependence of density of freshwater; maximum density occurs at 4 C.

Adapted from [Weeks and Ackley, 1982].

4 Temperature of maximum density

Temperature (°C)

3

Freezing temperature

2 1 0 –1 –2 24.69 (‰) –3 –4 0

2

4

6

8

10

30

12

14

16

18 20 22 24 26 28 30 32 34

Salinity (‰)

Figure 2.2 Dependence of the temperature of maximum density and the freezing point on water salinity; the two curves intersect at the salinity of 24.69 ‰.

2.2. PHASE DIAGRAM OF SEA ICE From the viewpoint of materials science, natural sea ice is a very complex material. In order to understand the physics of sea ice, one must have an appreciation of the so-called sea-ice phase diagram. A phase diagram illustrates the relative amounts and composition of the phases (ice, brine, and precipitated solid salts) that coexist at different temperatures as a given volume of seawater freezes and its temperature decreases. The reader must consult Weeks [2010] for a general understanding of this topic. Only a brief description is given here. A comprehensive phase diagram for “standard” sea ice was developed by Assur [1958] and is presented in Figure 2.3. It shows the weight ratio of each component: pure ice, salts, and liquid brine at any temperature when the three phases are in equilibrium. Temperatures for precipitation of different salts are also shown. For all practical purposes, a phase diagram for the binary mixture of sodium chloride and water is sufficient to represent the salt contents in sea ice. Only that part of the phase diagram is presented in Figure 2.4. If brine exists in sea ice at −10 C, for example, some water in the brine mixture has to freeze in order to bring the mixture to its equilibrium concentration point; i.e., to a salinity of 135‰ as shown in Figure 2.4. As the temperature of the mixture decreases further, more water molecules continue to solidify until the temperature reaches −22.8 C,

500

2

-6H

1000

3

Ca CO

1000

500

50

? O 2

O 2

H

O

+

MgCI2-12H2O

SALTS

Mg

CI

2

H2O Na

KCI

2

–

-12 H

CI

KC I

NaCI-2H2O MgCI -8H O 2 2

5

-

CI –

Mg

++

+

+

+Ca +K + rest

–10

10 5

BRINE

SO4

1 0

100

Na2SO4-10H2O

CI-

2H

CI 2 -8

Ca3CO-6H2O

Na

Mg

H2O

50

10

ICE

2

4

SO 2

Na

100

BRINE

Weight ratio (g/kg)

-10

H

O

ICE

–20 –30 –40 Temperature (°C)

–50

1 –60

Figure 2.3 Sea ice phase diagram [Assur, 1958 / U.S. Department of the Army / Public Domain]. 0 NaCI - H2O solution (binary mix)

–5

Temperature (°C)

floats on its melt. As brine in sea ice drains the ice becomes even lighter. Air is also dissolved in seawater. Through the constant stirring of the sea surface by wind and waves, gases are transferred from the atmosphere to the water. The three common gases that make up 99.96% of the atmosphere are nitrogen (78.08%), oxygen (20.95%), and argon (0.93%). The sum of their total percentages in the dissolved air in seawater is 98.5% (62.6% nitrogen, 34.3, oxygen, and 1.6% argon) [Pilson, 2012]. The solubility of oxygen in water is relatively higher than nitrogen. This results in a lower nitrogen content in water and hence ice. Carbon dioxide represents only 0.035% of atmospheric gases but it constitutes a relatively larger percentage (1.4%) of the total air dissolved in seawater. Marine organisms may be factors that contribute to the higher CO2 percentage. As the temperature or the salinity of seawater increases, the amount of gas that ocean water can dissolve decreases slightly. When seawater freezes, air is segregated and entrapped in the form of pockets inside brine at the pockets or separately within the intercrystalline boundaries (grain boundaries). More details are given in section 2.5.4.

O

ICE PHYSICS AND PHYSICAL PROCESSES 33

–10 Usual salinity of sea water 35‰

–15 –20

Eutectic point of NaCI = –22.8°C

–25 0

50

100

150

200

250

300

Salinity (‰)

Figure 2.4 Phase diagram of binary mixture consisting of NaCl and H2O.

which is the eutectic temperature of NaCl. At the eutectic temperature all sodium chloride precipitate with no liquid left in the binary mixture. 2.3. INITIAL ICE FORMATION 2.3.1. Freezing Processes in Freshwater and Seawater As mentioned above, the water salinity is the controlling factor that governs the temperature dependence of both the freezing point and the density of the water. Consequently, the freezing process differs between fresh, brackish, and seawater, though they all freeze when the water surface is cooled down to or below its freezing temperature but with a remaining condition. Freezing requires the presence of a nucleus around which ice crystals can form. In case of

SEA ICE

natural water bodies nuclei could be dust particles, snowflakes, frozen water droplet, or any type of impurities deposited at the upper surface of the water. In absence of such nuclei, water can remain in liquid phase at temperatures well below its freezing point. For pure water, this phenomenon is called supercooling. For impure water, such as seawater, brackish, or even freshwater with its negligible quantity of impurities, it is known as constitutional or compositional supercooling (section 2.5.1). Constitutional originates from constituents—the ingredients of impure water. Although by definition, pure water freezes at 0 C under normal atmospheric pressure, it can be supercooled under that pressure down to about −42 C. Considering the relatively large quantity of dissolved salts in seawater compared to fresh water in lakes and rivers, the extent of constitutional supercooling in seawater is probably limited to a few hundredths to tenths of a degree below the freezing point [Weeks and Ackley, 1982]. The processes of ice formation in freshwater and seawater are described in the following. In freshwater, as the water surface cools, the cooler layer at the top becomes denser and therefore sinks, allowing warmer water to rise. This vertical convection continues until the surface temperature reaches the critical temperature of maximum density at about 4.0 C. At this temperature the convection essentially stops. Further cooling of the surface layer makes it less dense and therefore remains at the surface. The surface temperature then responds faster to further drop in atmospheric temperature until it reaches the freezing temperature of 0 C. At this point ice crystals start to form around appropriate nuclei. It should also be pointed out here that fresh water (depending on its purity and the absence of nucleating agents) can be supercooled well below that temperature if it is not mechanically disturbed. The temperature of the water under the newly formed ice away from the interface remains isothermal at 4.0 C. It gradually cools as a result of the heat exchange at the ice–water interface, thus the freezing continues. The density of clear freshwater lake ice is slightly less than about 917kg/m3, or close to that of single crystal of pure ice. This is about 10% less than the density 999.8 kg/m3 for pure water at 0 C. Seawater, being a solution, undergoes a markedly different freezing process, particularly when the salinity exceeds the critical value of 24.69‰. It can be seen in Figure 2.2 that for salinities above this critical value, the temperature of the maximum density becomes lower than the freezing temperature of the seawater solute (opposite to the situation of fresh water). This means that as the surface of the seawater cools, the denser cooler layer sinks and a vertical convection current is created similar to the case of freshwater. The difference, however, is that this current does not stop before the surface reaches the freezing point. The freezing point of water with salinity 35‰ is −1.8 C, which is higher than the critical temperature, of

Mixing layer Colder water

~50 m

34

Continues to sink

Pycnocline zone Extremely stable No vertical convection

Figure 2.5 Illustration of the lower limit for convection current in seawater in the Arctic.

maximum density for the solution (−3.5 C). This means that the entire depth of the seawater body must be cooled to the freezing temperature before any formation of ice. But this does not happen in nature. The convection current is usually limited to the upper 50–200 m depth, called the mixing layer of the ocean, as can be visualized from Figure 2.5. Many areas of the ocean are stratified with increasing salinity and hence become denser toward the bottom. When the colder and denser upper area sinks, the increased density of that layer may not be as high as that of the lower layer, known as a pycnocline zone that features a rapid increase of density with depth due to continuous change in its temperature /or salinity (Figure 2.5). This stratification imposes a lower limit to the vertical convection in the mixing layer. This limit is about 50 m in the Arctic Ocean, which has slightly less salinity than the Atlantic Ocean. Any further cooling of the surface below the freezing temperature of the isothermal mixing layer will cause surface freezing. Differences in the freezing mechanisms in freshwater and seawater can then be summarized as follows. In the case of freshwater, the convection current stops when the surface temperature reaches about 4 C (the value corresponding to maximum density). Therefore, there is a time lag between the end of the convection current and the actual start of freezing when the surface temperature reaches the freezing point. This does not happen in the case of seawater as the convection current stops at the moment when the surface temperature reaches the freezing point, depending on water salinity. Consequently, in order for ice to form in seawater more heat has to be removed compared to that required for fresh water. In other words, ice formation in seawater is relatively slower

ICE PHYSICS AND PHYSICAL PROCESSES 35

than freshwater. As a rule of thumb, sea ice forms faster under any of the following three conditions: (1) in areas of low salinity such as those near the mouths of rivers; (2) in areas where there is no ocean current such as fjords, bays, and straits; and (3) in areas of shallow waters near coasts or over shoals or banks where only a small body of water must be cooled. 2.3.2. Initial Formation of Ice Crystals and Frazil Ice As mentioned above, attaining the freezing point is not sufficient for water to freeze because the formation of ice crystals requires nuclei to start solidification. In primary cryospheric regions, small ice nuclei are almost always present in the atmosphere. These ice particles are deposited continuously on seawater surfaces to initiate nucleation of crystals at the surface. If the water surface is very calm and the air temperature decreases slowly to develop only small temperature gradient, the surface water may be supercooled and crystals become ready to be nucleated. Once nucleated, the pure single crystal of ice takes a spherical shape. This shape minimizes the surface energy because the surface-to-volume ratio is a minimum for a sphere. The spherical shape, however, is not sustained due to the anisotropic growth rates in different planes of the hexagonal lattice of ice crystals. The crystal starts growing parallel to its basal plane with c axis normal to this plane (definitions are given in section 5.1.3 and Figure 5.4). It takes the form of a discoid because the growth in the basal plane is preferred in hexagonal crystals such as those of ice as will be described in detail in section 5.2. Under calm conditions, without any mechanical or thermal disturbance, each discoid floats at the surface. Each tiny disc floats with its flat surface parallel to the water surface, i.e., the horizontal plane. Since the major surfaces of these discs are parallel to the basal plane, the c axis of these crystals tends to be oriented in the (a)

vertical plane (Figure 5.4). Each of the floating discs, therefore, is oriented favorably for growth in the horizontal plane. If the initial ice crystals are nucleated around snowflakes, for example during snow flurry activities in the air, then those crystals nucleated with flakes whose surfaces (basal plane) is parallel to the water surface tend to grow faster than the others. Of course, this is not the case when snow fall covers the water surface. In those situations, the initial growth processes are complex. The crystal will be in a best position to grow vertically if nucleated by a snowflake with its flat surface (basal plane) oriented in the vertical direction (c axis horizontal) while falling on the water. Knowledge of the early stages of nucleation of sea ice crystals at the snow-free surface of saline water is available from early studies conducted in the laboratories [Weeks and Assur, 1963; Weeks and Lofgren, 1969; Cox and Weeks, 1975]. Under calm conditions, due to the favorable orientation mentioned above, ice discs continue to grow laterally up to 2–3 mm depending on the salinity of the water [Weeks and Ackley, 1982]. The freezing front becomes wavy as the disc size increases. Initially, circular boundaries assume tree-like or dendritic structures by developing “arms” around their peripherals as shown schematically in Figure 2.6. The term dendritic comes from dendrology—the study of trees. This is also called stellar shape, which is true for both freshwater and seawater ice but due to different mechanisms. In freshwater ice, with low concentration of solutes, formation of dendrites is primarily thermally induced phenomenon. The temperature gradient around the crystal is asymmetric due to complications of heat flow. This is mainly due to the instability at the growing front induced by increase in size of the discs. A small perturbation at the ice–water interface ends up in even more supercooled liquid so the interface becomes more unstable. In saline water, on the other hand, initiation of dendritic arms is a salinity-related (in addition of being

(b)

Figure 2.6 Stages of dendritic or stellar growth of a freshly nucleated circular discoid in (a) saline water and (b) after the fragile arms of the stellar break off to form the frazil needles (shown in black).

36

SEA ICE

temperature-related) phenomenon. As a discoid continues to grow it pushes the salt-rich solute to its boundaries. Salt concentration starts to build up around the disc with varying distribution leading to varying degrees of supercooled water. This causes instabilities in the freezing such that any further crystal growth takes place anisotropically (though still in the horizontal plane). The growth direction follows the path of supercooled water, i.e., the direction of least resistance for growth. Details on this aspect of growth are provided in the section 2.4.2. The change from discoidal to stellar crystal shape (Figure 2.6a) is accelerated under rapid cooling when the solute is rejected at a rate fast enough to cause significant variation of the salt distribution in the vicinity of the crystal in the plane parallel to the major plane of the discoid. The arms of the stellar (dendritic) shape are very fragile with an average diameter around 2.5 mm [Weeks, 1959]. Moreover, those arms may grow thicker because they are surrounded by supercooled water and may not be able to sustain their own weight and consequently break. Eventually, they start to break off and form needle-shaped crystals as illustrated in Figure 2.6b. This process is also triggered by the wave actions that tilt and bend the delicate discs to fracture into segments. This type of fragmentation of thin ice discs occurs in both freshwater and seawater but is more pronounced in seawater. Early observations [Suzuki, 1955] showed that a fewer number of needles are observed in case of freshwater ice. The needle-shaped ice crystals are known as frazil ice. They may be geometrically oriented or randomly oriented and usually confined to the thin supercooled layer at the surface [Hallett, 1960]. However, turbulent wave action also causes inclined discs and needles to sink into the thick supercooled mixed layer. The ocean wave may tend to compress those needles and orient them with their long axis parallel to the vertical plane. The spatial distributions of the discoid, the broken discoid, and the needle-shaped crystals are usually sporadic and loose. The frazil needles are typically less than 1 millimeter in thickness and a few millimeters to a few tens of a millimeter in length. Their aggregation forms what is known as frazil or grease ice [Weeks and Ackley, 1982]. Reviews on frazil ice in rivers and ocean can be found in Osterkamp [1978] and Martin [1981]. When frazil crystals cluster together over a large area of a water surface, they dampen surface motion and give the ocean surface a greasy or oily appearance. It is a soupy layer on the ocean surface that does not reflect much light, hence gives the surface a dark matt appearance. It is known as “grease ice,” though it is not actually solid ice. Ocean currents cause frazil crystals to herd into streaks or pile up downwind against the edges of floating ice floes to depths that can be in the order of 1 meter [Martin, 1981]. Figure 2.7 is an example of frazil ice

Figure 2.7 Frazil ice blowing into bands and accumulated next to an ice floe (from the collection of Dr. Pablo Clemente-Colon, National Ice Center).

blowing into bands and accumulating against the edge of an ice floe. Due to their elongated shape, frazil needles may not create enough buoyancy force to keep them floating, so they may be carried by the turbulent currents to areas below the water or ice surfaces. It is, therefore, not uncommon for frazil crystals to exist as suspended elements in deeper waters or solid masses below ice sheets. When frazil crystals form a layer at the ice–water interface at the bottom of ice sheets, the crystalline growth habit of the ice cover, such as columnar structure, is interrupted. However, as will be seen later in section 5.4.1.1, crystals of frazil ice may also act as seeds for new growth of columnar-grained ice. During the early part of the ice growth seasons, frazil crystals in fjords and channels can also be herded and pushed by the wind toward the shores which eventually consolidate to form thick layers of vertically oriented crystals. Microstructure and strength properties of this type of frazil ice have been examined in detail by Sinha [1986] and a brief description of the structural aspects of this type of frazil ice is given in section 5.3.3.2. Before closing this section, it should be mentioned that the initial ice cover can also develop from snow deposition on open water surface and the eventual freezing of watersaturated snow. Relatively thin ice covers can also be thickened by the solidification of flooded snow overlaying the existing ice cover. This commonly occurs when the snow cover becomes heavy enough to depress the ice surface below its freeboard. The influx of seawater through the permeable snow saturates the snow mass which subsequently freezes into what is known as granular or snow ice. Microstructural aspects of this ice are presented in section 5.3.3.1.

ICE PHYSICS AND PHYSICAL PROCESSES 37

2.4. SEA ICE GROWTH Understanding the processes of sea-ice growth, particularly during the early freezing season, is important both from the climatic and remote sensing viewpoints. Ice grows mostly thermodynamically due the temperature gradient between the colder temperature of the atmosphere and the warmer seawater. However, ice can thicken mechanically as a result of its convergence while drifting. Upon collision of ice floes, their edges raise. More seriously, crushed ice blocks may pile to form ridges. This geometrical feature extends above the freeboard of the ice (forming the sail of the ridge) and under the freeboard (forming the keel). The sail is typically 1–3 m high and the keel is 3–5 times deep (section 2.6.2). In this section, only thermodynamic growth of sea ice is addressed. Thermodynamic growth of lake ice is addressed in several publications, e.g., Kheyrollah-Pour et al. (2017), Kirillin et al. (2011), Wang, Key, Liu [2010], and Duguay et al. (2003). The rate of thermodynamic growth of sea ice depends mainly on three factors that can be measured: air temperature, ice thickness, and snow cover. Other factors include the solar radiation, wind conditions, and the density and albedo of snow. Some of these factors are often difficult to quantify. Wind and ocean current are also active factors because they determine the mobility of floating ice, hence possibility of ice thickening. The apparent macroscopic and crystallographic form of ice at any growth stage, particularly during the early stage, is determined mainly by the oceanic conditions, whether calm or turbulent. After the initial formation of ice discoid with its dendritic boundaries as described earlier, the ice growth commences laterally as the minute ice discoid grow sidewise or frazil crystals herd horizontally. The growth will then proceed vertically in the direction of maximum heat flow. During this process, brine is entrapped within the sea ice volume and rejected to the underlying seawater through mechanisms discussed in section 2.5.5. 2.4.1. Lateral Ice Growth The general characteristics of thin ice cover depends on the state of the ocean surface, namely whether it is quiescent or turbulent. Sea ice grows laterally following two different scenarios, depending on wind and oceanic conditions. The first scenario pertains to ice growth in relatively calm atmospheric and oceanic conditions. In this case, ice discoid ice and frazil crystals or frazil streaks continue to grow sideways until they touch each other and cover a large surface area of the water. Also, under a quiescent water surface, frazil crystals can be herded to form streaks of ice. The ice particles eventually consolidate if cold temperature persists and the water surface continues to be calm. These processes lead to the formation of nilas, a thin

Figure 2.8 Dark and light “nilas” in Hudson Strait, Canada during mid-November, photograph taken from altitude of about 700 m (courtesy: Canadian Ice Service).

but extensive elastic crust of ice, easily bending on waves and swell [MANICE, 2005]. While nilas has a matte surface, it is initially translucent and has a dark appearance but it takes on a gray appearance as the sheet continues to thicken. It is common to divide nilas into dark nilas (up to 5 cm in thickness and dark in color) and light nilas (more than 5 cm in thickness and lighter in color). The two types are shown in Figure 2.8. Nilas may grow up to the thickness range of 10–20 mm if the water surface remains calm, but this rarely happens in open seas. In nature, even under relatively calm weather conditions there could be oceanic currents and lowamplitude waves generated by convections or very light breeze. Nilas are usually broken into large pieces (a few meters to tens of meters wide). While floating and moving, the fractured pieces may slide over each other to form what is known as surface rafting (Figure 2.9). This phenomenon is discussed in more detail in section 2.6.1 but it suffices here to mention that, in general, rafting is a characteristic of nilas and thin ice (< 15 cm). In fact, rafting is used to identify thin ice and nilas in nature as well as remotely sensed images. The second scenario of lateral ice growth is observed under wind and turbulent ocean surface conditions. This is commonly seen in open seas where the water surface is usually more turbulent than water in enclosed bays. When the water surface is rough at the time of initial ice formation, turbulence will not allow consolidation of the herded frazil crystals into nilas. Instead, it causes frazil to undergo cyclic compression following the wave action. If compressed enough, herded crystals may bond with each other due to the freezing of water between them. Eventually, they may take the form of discs that may start as slush before the solidification. When solidified, they are called pancake ice. Figure 2.10 shows pancake ice floating but not yet forming an ice sheet. The ocean waves are

38

SEA ICE

Figure 2.9 Nilas with rafting and cracking in Nares Strait, north of Baffin Bay, in April 1994 (photo by N.K. Sinha).

Figure 2.11 Pancakes with diameters up to 5 m in Davis Strait in November 1982; note (a) the amalgamation of two pieces which is manifested by the intrapancake ridge along the line of connection, and (b) the fractured pancake (photo by N.K. Sinha).

heavy interaction (intrapancake ridges). When water between pancakes freezes, the lateral ice growth is confirmed. Conglomeration of pancakes may form extensive ice sheets but this may not be visible in nature if covered by snow. However, it can be detected in Synthetic Aperture Radar (SAR) images. It appears bright because of the significant surface roughness.

2.4.2. Vertical Ice Growth (Congelation Ice)

Figure 2.10 Freely floating and partially consolidated pancakes, with diameters of 0.5 to 1.0 m, damping the ocean waves in Davis Strait (photo by N.K. Sinha).

apparent in the figure, indicating their action in preventing the formation of a continuous ice sheet. The name pancake is used because the discs are round or oval in shape. The diameter ranges from a few fractions of a meter to a few meters. It is common to see the formation of small and thin pancakes during the early freezing period. Often their edges are raised as a result of rotation and collision against each other by the ocean-induced motion. Pancakes will eventually consolidate and this allows for the lateral growth of sea ice cover as shown in Figure 2.11. The figure also illustrates a few features of pancake ice. Large pancakes may be formed by consolidation of smaller ones. Some pancakes may break due to wave actions. Pancakes usually have raised edges as a result of their

As the lateral ice growth continues (either through formation of nilas or consolidation of pancakes), the ice sheet may eventually have no room for further growth and starts to grow vertically down along the direction of the maximum heat flow from the underlying water to the atmosphere. This process is known as ice congelation. In metallurgy, it is known as directionally solidification (DS) crystallization. In fact, turbine blades used in the hottest sections of gas turbine engines (e.g., jet engine) are exclusively made of nickel-based super alloys using the process of directional solidification [Sinha, 2009]. Nilas and pancake ice are usually composed of frazil or granular ice crystals. Congelation ice starts to grow under this layer. Structurally, these two layers have distinguishable fabric. The differences lie in the solidification process involved. Granular or frazil ice in the transition layer develops after the interparticle liquid freezes as the heat continues to flow from the bottom to the top of the ice sheet. The columnar grains formed during congelation are significantly larger than the grains of frazil or granular ice. They are shaped like a pencil with their length along the vertical direction. The grain shape, orientation, boundaries, sub-boundaries, and the geometry of the entrapped inclusions (liquid brine and gases) are affected

ICE PHYSICS AND PHYSICAL PROCESSES 39

by the history of ice growth. The growth depends on the prevailed meteorological conditions (including oceanic). Congelation ice is common in the Arctic but not in the Antarctic. There are innumerable examples that can be cited to confirm the sea ice congelation in the form of columnar crystal structure in almost all sea ice regions in the Arctic, including relatively narrow channels, fjords bays, and in fast ice. Studies that confirm this aspect include Weeks and Lee [1958] at Hopedale, Labrador; Weeks and Hamilton [1962] at Point Barrow, Alaska; Pounder [1965] for Button Bay off Hudson Bay; Peyton [1963] at Barrow, Alaska; Nakawo and Sinha [1981] in Eclipse Sound, Baffin Island; Weeks and Ackley [1982] in Alaskan coast; Shokr and Sinha [1994] in Resolute Bay, Nunavut, and Timco and Frederking [1996] in Canadian Beaufort Sea. Congelation ice is also found in undeformed second-year ice (SYI) [Bjerkelund et al., 1985; Sinha, 1985b] as well as multi-year ice (MYI) [Shokr and Sinha, 1994; Sinha, 1991]. Gow et al. [1987] collected cores from ice floes in pack ice in Farm Strait and found that they typically contain 75% congelation versus 25% frazil ice. Meese [1989] confirmed the presence of congealed sea ice in the Beaufort Sea. Naturally, congelation ice is not expected to exist in pressure ridges or rubble fields containing damaged and crushed ice, except within the fragments or blocks of ice inside the ridges or rubbles [Sinha, 1991]. Weeks [2010] mentions that exact percentages of frazil versus congelation ice separated on a regional basis do not appear to be available. Generally speaking, after the initial ice formation, ice growth usually favors the formation of congealed ice in the Arctic. There are remarkable differences in the conditions for sea ice growth between the Arctic and the Antarctic (the reader may refer to the Antarctic map in Figure 13.7 while going through the following discussion). Practically, all the area of sea ice formation in the Antarctic feature seasonal or temperate ice, equivalent to the ice in subarctic zone. Treshnikov [1966] estimated that about 75% of the Antarctic ice melts during the summer, compared to about 25% in the Arctic (note the old date of the study). In the Antarctic, older sea ice survives essentially in Weddell Sea and some areas of the Ross and Bellingshausen Seas. Moreover, the vast perimeter of the annual ice around the continent of Antarctica is actually within the South Temperate Zone (STZ), not really in the south polar region (i.e., south of 66.6 S), and can be considered primarily as marginal ice zone (MIZ). This MIZ in the STZ is subjected to strong winds of the Southern Ocean and the velocity shear across the Antarctic polar front [Wadhams, 1986]. The persistent stormy conditions associated with high winds cause the water surface to be almost always turbulent. Naturally, the Antarctic ice is expected to feature less congelation ice and more frazil

ice. Gow et al. [1987] measured the ratio of frazil to congelation ice in the Weddell Sea during the austral summer of 1980 and found that the average values were 57% and 43%, respectively. However, they reported significant variations in these ratios from floe to floe. For example, they found floes containing between 6% and 90% congelation ice and between 3% and 100% frazil ice. Lange and Eicken, 1991, reported that sea ice floes of all ages in the Weddell Sea in the Antarctic are dominated by granular ice crystals of frazil origin. However, in a field study in the western area of the Ross Sea, Jeffries and Adolph [1997] found that thermodynamic thickening of the ice in the inner pack ice was dominated by congelation ice growth. They observed that 65% of the ice in the inner pack (up to 400 km from the coast) was congelation ice with a mean thickness of 200 mm, while 22% of ice in the outer pack (>800 km from the coast) was frazil ice with a mean thickness of 120 mm. The authors attributed the preponderance of congelation ice in the inner pack ice due to a less stormy environment. Worby et al. [1998] suggested that at least 39% of the ice volume in the East Antarctic is columnar, 47% frazil, 13% snow ice, and 1% other types. As part of the Japanese Antarctic Climate Research (ACR) Program, a 2-year study of atmosphere/sea-ice/ ocean interaction processes off Queen Maud Land and Enderby Land, Kawamura et al. [1995] performed extensive investigations of sea ice from 1990 to 1992. They examined the structure and texture of sea ice, along with measurements on vertical distribution of oxygen isotope concentration, δ(18O) at 16 stations in Lützow-Holm Bay (a 220 km bay in the northeast Antarctica). The ice thickness varied in the range of about 2 to 3.5 m. Significant variation in the amount of granular and mixed columnar/granular ice was noticed. The thickness of this layer varied a great deal, between 0.5 to 2.5 m, but all the cores exhibited columnar structure at the bottom. For more information on Antarctic sea ice structure the reader should check Worby et al. [1998]. Up-to-date information on Antarctic sea ice is available through literature from the National Institute of Polar Research in Tokyo and the Low Temperature Science Laboratory of Hokkaido University in Sapporo, Japan. The above discussions include information from field studies conducted in the 1980s and 1990s. This information has not been updated because field work on sea ice physics has declined in the past two decades. 2.4.3. Superimposed Ice Superimposed ice is formed on existing ice surfaces as a result of one of the following three processes: freezing of rain on existing ice surface, refreezing of ice surface or snow melt, and freezing of water-logged snow. The second scenario occurs when the ice or snow surface melts due to

40

SEA ICE

warm atmospheric temperature during winter and then refreezes when cold atmospheric temperature resumes. The third scenario occurs when the load of snow cover is sufficient to depress the growing ice sheet below its freeboard so that the water floods the surface and later refreezes. In both cases the freezing may proceed either from top down or from bottom up depending on the temperature difference between the ice surface and the atmosphere. From the viewpoint of the crystallographic classification of natural ice, superimposed ice is usually categorized as snow or granular ice. Incorporation of snow into the sea ice reduces the δ(18O). Therefore, it is possible to differentiate between snow ice and other structurally-different ice types (of seawater but not snow mixture origin) using a threshold on the δ(18O) value. Superimposed ice is not common in the Arctic though it is observed in subarctic areas such as the Labrador Sea, but the ice covers are significantly thinner than those in other areas. However, this type of ice is frequently observed in the Antarctic. In the southern hemisphere, the annual sea ice develops primarily around the coastal areas surrounding the continent of Antarctica, except for the areas covered by the Ross Sea, the Amundsen Sea, the Bellingshausen Sea and the Weddell Sea, all within the western Antarctic (and remarkably just west of the International Date Line and/or longitude of 0 ). Since the coastal line of this continent follows essentially the Antarctic Circle with a latitude of about 66.6 S, the sea ice regime is confined within a narrow belt between 60 S and 66.6 S, except, of course, the western Antarctic. The area, north of the Antarctic Circle but beyond the coastline of the continent in the western Antarctic, is occupied largely, not by sea ice, but by a collar of freshwater ice in the form of shelf ice. There are a number of huge ice shelves, for example, Ross Ice Shelf, Ronne Ice Shelf, and Thwaites Ice Tongue. Thus, strictly speaking, the Antarctic sea ice regime is not within the south polar region. Since it is found in STZ, this should be named as STZ ice, not “polar ice.” The STZ ice is naturally expected to be thinner due to the warmer atmospheric temperature of the STZ. Additionally, heavier snowfall in the STZ (primarily drifted from the land to the ocean surfaces) compared to that in the Arctic also dampens the growth of sea ice. No wonder, the Antarctic sea ice is thinner than its counterpart in the Arctic. Here, of course, we mean the annual sea ice as the “undeformed” FYI grown in oceans without the adverse effects of storms and severe wind. This speculation is confirmed by Weeks [2010] who concluded that the thickness of undeformed ice in the Antarctic rarely exceed 1 m, which is remarkably thinner than1.5–2.5 m typically measured in the Arctic. Only a few field studies have been conducted in the past to identify the superimposed ice in the Antarctic. Worby et al. [1998] observed that more than 50% of thin ice

surfaces in the Eastern Antarctic were flooded. The surface flooding is caused by a number of processes that include surface deformation and ice breakup, wave penetration in ice, snow loading, or upward rejection of brine from the ice subsurface layer [Perovich and RichterMenge, 1994]. In the western Antarctic, namely in the Ross and Amundsen Seas, Jeffries et al. [1994] found that superimposed snow ice varied from 13% to 43%, whereas frazil ice averaged between 25% and 55%. They combined textural analysis with measurements on the stable isotope δ(18O) to discriminate between snow ice and frazil ice, similar to work of Kawamura et al. [1995] in LützowHolm Bay. The application of differential isotope, δ(18O), method to discriminate between snow ice and frazil ice is certainly unique and very powerful. However, the technique for using such isotopic procedures requires accessibilities with laboratories equipped with rather sophisticated and expensive equipment, and highly trained operators. Moreover, only a limited number of samples can be analyzed because of high expenses involved. An alternate powerful method can be performed in field laboratories. Since the microstructures of granular snow ice and frazil ice are significantly different from each other, the two types of ice can be identified readily using the doublemicrotoming technique (section 4.2.2). This involves preparation of thin sections from ice cores or blocks in conjunction with measurements utilizing thermal etching, chemical etching and replicating in addition to polarized light. Replicas can be made in the field and can also be readily examined with optical microscopes. If desired, the replicas can be examined later with scanning electron microscopes (SEM). These techniques are introduced in sections 4.2 through 4.4. Minimum of two sections (i.e., horizontal and vertical) are required for proper discrimination between the two types. 2.4.4. Thermodynamic Ice Growth Thermodynamic ice growth entails increasing thickness in response to the negative energy budget between ocean and the ice sheet. The ocean is the source of heat that is transferred to the atmosphere through the ice. As mentioned earlier, mechanical thickening of ice entails piling up of broken ice along the edges of ice floes due to several processes such as rafting, ridging, and rubble pile-up. These processes result in much larger thickness than that typically caused by thermodynamic growth. Thermodynamic growth continues throughout the entire freezing season and may extend to next seasons if ice survives summer melt. In this case, layers of ice are added at the bottom of the perennial ice. An example of the growth of a MYI floe was observed in the Mould Bay, Canadian western High Arctic [Sinha, 1986]. Here,

ICE PHYSICS AND PHYSICAL PROCESSES 41

thermodynamic ice growth of the 5.1 m thickness was observed and the surface that separated ice growth at the ice bottom in successive growth seasons showed seamless columnar-grained ice structure (Figure 6.17 in section 6.2.4). Ice continues to thicken until it reaches a state of thermodynamic equilibrium with the atmosphere. This happens when FYI thickness reaches 2–3 m in the Arctic and 1–2 m in the Antarctic ice [Whitman, 2011]. Thermodynamic growth rate is mainly controlled by three factors: (1) the severity and duration of cold air temperatures, (2) snow accumulation on the surface, and (3) ice thickness. Other factors include proxy for longwave and latent heat fluxes, ocean heat flux, shortwave flux in spring and summer, and the heat conduction through ice and snow with its sensitivity to the brine volume and geometrical configuration of brine pockets. Persistent cold temperature and absence of snow accelerate the thickness growth. As a rule of thumb, fast growth rate leads to more brine retained within the ice crystals (i.e., higher ice salinity), and slow growth rate allows for larger crystal growth and more brine drainage (i.e., less ice salinity). Data on FYI growth are available from many studies. Extremely rapid freezing can occur in a lead when the ambient air temperature is very low. The Canadian Ice Service (CIS) confirmed (through personal communications) that under a steady air temperature of –25 C, a thin skin of ice would thicken quickly to reach 100 mm during the first 24 h. It may take a few weeks to reach the stage of mature FYI with thickness greater than about 1.2 m [Canadian Coast Guard, 1999]. At warmer temperatures below but closer to the freezing point (e.g., Labrador Sea, Bohai Bay and Sea of Okhotsk), it would take 3–4 weeks for ice to grow up to the stage of thin FYI (0.3 m thick). In general, the initial ice growth rate in the Arctic can be as high as 3–6 cm/day, depending on the region and weather conditions. Recall that the growth rate during the initial ice formation period can be an order of magnitude higher than the rate when ice started to mature (say > 50 cm thick). Snow on top of the ice cover slows the ice growth. Growth and salinity profile of annual sea ice, in conjunction with microstructural investigations, have been studied extensively for several years in Eclipse Sound near Pond Inlet, Baffin Island, Canada [Sinha and Nakawo, 1981, for 1977–1979; Nakawo and Sinha,1981, for 1977–1978). They reported rates between 5 mm/day (i.e., 0.021 cm/h) and 15 mm/day (0.0625 cm/h). Another extensive series of long-term study program, from 1981 to 1985, was carried out in Mould Bay, Prince Patrick Island, Canada. Growth of ice was recorded at several stations across the 7.8 km wide bay. The details of this project are given in section 6.2. As an example, a growth rate of 20 mm/day was noted during the first 10 days of growth, without any snow cover, in September, 1981.

However, for the first 200 days, after the beginning of freezing, during the 1981–1982 winter season, the average growth rate varied from 7.5 mm/day to 9.5mm/day. The differences in the growth rate were caused by the differences in the accumulated snow depth. In-situ data on growth rate of thin ice in the polar regions are scarce not only because it is unsafe to walk on thin ice but also the early freezing period coincides with the approach of the polar night. In a laboratory experiment of sodium chloride ice to simulate sea ice, Cox and Weeks [1975] observed the ice growth rate under constant ambient temperature of −20 C. The authors found that during the first 4 days, the ice grew to 26 cm thick (with an average rate 6.5 cm/day). Within the following 4 days the growth rate dropped to 2.5 cm/day, then to 0.3 cm/day averaged over the next 10 days. As for Antarctic data, Melnikov [1995] conducted insitu measurements in the western Weddell Sea in 1992 during the US–Russian Ice Station Weddell 1 Expedition in the Antarctic and found much higher growth rates of 3.8 mm/h for ice up to 90 mm thick (19–20 May), 1.3 mm/h for thicker ice up to 280 mm thick over the next 8 days, and only 0.3mm/h during 81 days of observations on ice between 0.42 m and 0.97 m thick (18 March to 7 June). No recent studies have been undertaken to update information of the Antarctic sea ice thickness growth rate. Recent studies focus on sea ice extent as it is readily available from satellite observations and researchers are more interested in comparing it against the declining extent of the Arctic ice. 2.4.4.1. Simplified Models of Sea Ice Growth While sea ice grows thermodynamically during the freezing season, surface ablation and superimposed ice directly affect the thickness. Several other factors affect the growth process; hence indirectly affect the thickness estimates. While ice growth rate can be obtained from field and laboratory work as described above, equations are needed to estimate the thickness in terms of meteorological and oceanic parameters (at least in terms of surface air temperature). Modeling approach fulfills this need. Work on modeling of sea ice growth commenced in the 1800s with analytic methods and has become a classical problem in sea ice geophysics. Considering the vertical growth of sea ice, it has been traditionally treated as one-dimensional process. A comprehensive model should account for physical mechanisms and factors that affect the growth. This includes snow cover properties and processes (including depth), oceanic heat flux, radiative balance at the surface and surface wind, among other minor factors. A full-blown ice mass balance model, which involves local ice growth and melt, requires solving a coupled, non-linear set of equations describing thermodynamic

42

SEA ICE

growth and melt, surface energy balance, and the transport and redistribution of ice thickness. Most of the existing models use assumptions to simplify the formulations and produce reasonable estimates valid under specific conditions. Models have been developed to simulate the development of sea ice thickness yet with different levels of sophistication (Maykut and Untersteiner, 1971; Semtner Jr., 1976; Bilello, 1961; Sinha and Nakawo, 1981; Steele and Flato, 2000). Most of those models incorporate one-dimensional heat transfer process but with different assumptions to account for selected factors that affect the growth. In this section, a simple model based on heat balance at the ice–water interface is presented. This model neglects a few factors that are explained in the following section, most importantly the radiation balance at the ice or snow surface (section 2.4.4.4). More comprehensive model is presented in section 11.4.1. Spatial and temporal variations of ice thickness at any given location usually involve using timely satellite data. For example, Wang, Key, Liu [2010] combined thermodynamic formulation and optical satellite data to estimate the spatial distribution of lake and sea ice thickness. Hu et al. [2018] used a numerical model to produce the spatial distribution of ice thickness in the Canadian Arctic Archipelago, which features thick ice, with use of operational ice charts for verification. Sea ice growth models are reviewed in Leppäranta [1993]. The authors elaborate on physical aspects of sea ice growth and present analytic solutions for different environmental conditions. Aside from modeling, Lebedev [1938] developed a simple equation to estimate sea ice thickness in terms of the accumulated freezing degree days (FDD). The FDD is defined as the sum of the average daily sub-freezing degree days (below 0 C for freshwater and –1.8 C for seawater) for a specific period. For example, if the total number of days with air temperature below freezing is 3 and the average daily temperature is –2.8 C, –3.8 C, and –5.8 C, then for seawater with freezing point of –1.8 C, the FDD becomes the sum of the daily temperatures after subtracting the freezing temperature; i.e., 1 + 2 + 4 = 7 C. The equation relates the thickness h (in cm) to FDD as follows: h = 1 32 FDD

0 58

(2.1)

Obviously, this is an empirical equation, which has no physics basis. This renders it unreliable for scientific investigations but it is used for practical purposes. For example, it is used in the CIS ice monitoring program to obtain a rough estimate of ice thickness during its early growth phase (leveled nilas ice sheets). The simplest ice growth model, based on heat balance at the ice–water interface, is presented here. At this interface, latent heat is released as a result of freezing. It is the multiplication of the latent heat of freezing of ice Li, the

Ta

Air hs

Snow

hi

Fc

Water

Fw

Tb Ice

Tw

Figure 2.12 Idealized heat fluxes in a snow-covered floating ice sheet.

ice density ρi, and the rate of increase in ice thickness. The amount of latent heat released should be balanced by the two heat influences: (1) the upwelling heat flux Fw from the warm ocean to the colder ice interface, and (2) the conductive upward heat flux Fc through the bulk of the ice and snow. The latter is instigated by the difference between the colder air temperature and the warmer ocean temperature. Figure 2.12 is an idealized schematic diagram showing a section of floating sea ice of thickness hi with snow cover of depth hs. The air temperature is Ta, the temperature at the snow base is Tb and the water temperature at the ice–water interface is Tw. The energy balance at the ice–water interface takes the following form, following the convention of positive fluxes in the upward direction, ρi Li

dhi dt

= Fc − Fw

(2.2)

The LHS of the equation represents the latent heat of fusion of ice. Fw is usually neglected (reasons are explained in section 2.4.4.3). The heat conduction through the entire ice and snow volume [Fc in equation (2.2)] can be written in terms of the net snow/ice conductivity ksi, assuming linear temperature gradient within the volume, F c = k si

∂T Ta − Tw = ksi ∂z H

(2.3)

where, H is the total thickness of the ice and snow (hi + hs). The continuity of heat flux throughout the ice and snow layers and into the atmosphere (Figure 2.12) leads to the following equation: F c = − T a− T w

1 hi hs + + ki ks ka

(2.4)

where, ka is the effective heat transfer coefficient between ice surface and atmosphere, and ki and ks are the thermal conductivity of ice and snow, respectively. Substituting this equation into equation (2.2) after neglecting Fw yields: dhi ρ Li = − T a− T w dt i

1 hi hs + + ka ki ks

(2.5)

ICE PHYSICS AND PHYSICAL PROCESSES 43

By assuming constant values of all heat transfer coefficients k, the integration of the above equation produces an expression for the ice thickness: ρ i Li 2 hs 1 hi + ρi Li + 2ki ks ka

T w− T a dt

hi =

(2.6)

This equation can be used to estimate the ice thickness hi. The integral (Tw − Ta)dt is the same as FDD upon discretization. If the atmospheric temperature Ta becomes stable and equal to the ice surface temperature, then ka➔ ∞; then equation (2.6) can be written as ρi Li 2 h + 2k i i

ρi Li hs ks

hi = FDD

(2.7)

In absence of snow, this equation can be further simplified to calculate hi as: h2i =

2ki T w − T a ρ i Li

(2.8)

14

14

12

12

(Tw–Ta) = 5 (Tw–Ta) = 10

10

(Tw–Ta) = 15

Ice thickness (cm)

Ice thickness (cm)

The quadratic equation (2.7) is used to calculate ice thickness for different number of FDDs, given different snow thicknesses. Figure 2.13 shows results from using ρi = 917 kg/m3, Li = 334.9 × 103 J/kg, ki = 2.0 and

ks = 0.25 J/m.K.s. The FDD in these data are calculated using constant values of (Tw − Ta) through successive days. The point for day 1 in each graph in Figure 2.13 is generated using the thermal conductivity of seawater (0.6 J/m.K.s) instead of thermal conductivity of sea ice. The plots clearly show the blanketing effect of the snow for higher air temperatures. Note the pronounced sensitivity of the temporal growth of ice thickness to the snow depth during the accumulation of the first 100 mm of snow. The growth of sea ice in the Eclipse Sound near Pond Inlet, Baffin Island, Canada, during the two winter seasons 1977–78 and 1978–79 was also determined from equation (2.7) and compared to field measurements [Sinha and Nakawo, 1981]. Results are shown in Figure 2.14 as a function of the accumulated FDD. Dates given at the top of the figure indicate the time of the season. Calculations were based on the constant snow thickness of 11.4 cm measured in the field and appropriate ice and snow conductivities. Equation (2.7) produces reasonable results in general but underestimates ice thickness during the early part of the season and overestimates it toward the end. The disagreement at the end of the period could be attributed to increasing solar radiation.

10 8 6 (Tw–Ta) = 5 (Tw–Ta)=10

4 2 0

6 4

0 0

1

2 3 4 Freezing degree-days

5

6

7

0

1

14

10

(Tw–Ta) = 15

3

12

(Tw–Ta) = 5 (Tw–Ta) = 10

10

(Tw–Ta) = 15

Snow thickness = 100 mm

Ice thickness (cm)

12

(Tw–Ta) = 5 (Tw–Ta) = 10

2

4

5

6

7

Freezing degree-days

14

Ice thickness (cm)

8

2

(Tw–Ta) = 15

Snow thickness = zero

Snow thickness = 50 mm

8 6 4 2

Snow thickness = 150 mm

8 6 4 2

0

0 0

1

2

3

4

Freezing degree-days

5

6

7

0

1

2

3

4

5

6

Freezing degree-days

Figure 2.13 Dependence of ice thickness on freezing degree days for different snow thickness and air temperature, calculated by using equation (2.7).

7

44

SEA ICE December January 1977–78 November December January 1978–79 October November

February February

March

April March

April

Ice thickness, hi (cm)

150

Theory 1977–1978 100

Theory 1978–1979

50

0

1977–1978 (Freeze-up date: 26 oct. 1977) 1978–1979 (Freeze-up date: 22 oct. 1978)

0

1

2

3

Freezing degree days ×

4

5

103

Figure 2.14 Growth of sea ice in Eclipse Sound measured during winters of 1977–1978 and 1978–1979, compared to calculations. Dates given at the top of the figure indicate the time of the season. [Adapted from Sinha and Nakawo, 1981].

The model of equation (2.7) is known to overestimate the ice thickness during the initial freezing period (when the ice is thin, namely < 15 cm). That is because it assumes thermal equilibrium between the air and ice surface, which is not the case when ice is thin. As air temperature cools below the freezing point of sea ice, the ice starts to form but the latent heat keeps the ice surface temperature higher than air temperature. More importantly, in presence of wind the warmer ice surface of thin ice releases heat to the air through sensible heat flux. Over thin ice, this flux is one or two orders of magnitude larger than that over thick ice [Maykut, 1978]. For this reason, a modification to equation (2.2) was introduced in Ashton [1989] and verified using in-situ measurements. It also ignores Fw and assumes linear heat conduction (Fc) through the ice depth. The original equation can then take the form, ρ i Li

dhi dt

= − ki T w− T s hi

(2.9)

where, Ts is the ice surface temperature. The sensible heat flux Qia from the ice surface to the air is expressed in the form of a bulk heat transfer coefficient Hia applied to the difference between the ice surface and air temperature, Qia = H ia T s− T a

(2.10)

If the conduction heat term [the RHS of equation (2.9)] is assumed to be equal to the heat released to the

atmosphere [the RHS of equation (2.10)], then Ts from equation (2.10) can be substituted into equation (2.9) to calculate the rate of ice thickness growth as follows: dhi = dt

1 ρi Li

Tw − Ta hi 1 + ki H ia

(2.11)

This equation can be integrated with the boundary condition hi = 0 at the onset of freezing, which results in the following expression for ice thickness: 2k i hi = FDD + ρ i Li

ki H ia

2 05

−

ki H ia

(2.12)

This model improves the estimate of thin ice thickness although it does not account for the snow [for the improvement, see Figure 1 in Ashton, 1989]. It is possible to extend the model to account for the snow. When ice becomes thick, the term with FDD in equation (2.12) becomes much larger than the other terms and the equation converges to the form of equation (2.8) without the snow. This situation occurs at ice thickness around 30 cm (upon testing the two equations by the first author of this book). Values of Hia depend on the wind speed [Ashton, 1989]. For still air, Hia is found to be around 10 W/m2. C. For nominal wind in most of the Arctic region, a value of 20–25 W/m2. C is suitable and for severe wind conditions (e.g., > 50 km/h), a value of 30 W/m2. C is more appropriate. Examination of the

ICE PHYSICS AND PHYSICAL PROCESSES 45

sensitivity of hi to Hia shows a bias of 2.3 cm between using Hia = 18 and 20. This is a negligible difference for thick ice but not rightly so for thin ice, say V > 2 0 × 10 − 6 cm s

10

15 20 25 30 35 Ice growth rate (mm/day)

40

45

50

Figure 2.31 Dependence of calculated ratio of salt concentration retained in ice to that in seawater (Kδ) on growth rate. Calculations are based on equation (2.28) with K0 = 0.12 and δ/D = 42000, and from equations (2.30)–(2.32).

(2.29)

This equation was examined by Cox and Weeks [1975] using data obtained during a test on artificial ice grown in a copper container and cooled at the upper surface using a mixture of dry ice and alcohol. However, the experiment resulted in growth rates of up to 24 cm/day, which far exceeds the rates observed in nature. Growth rates in the laboratory experiments are usually much higher than those encountered in the field [Rohatgi and Adams, 1967; Lofgren and Weeks, 1969]. A nominal maximum growth rate of sea ice in nature is between 1.5 and 3 cm/day [Nakawo and Sinha, 1981; Weeks and Ackley, 1982; Legendre et al., 1991; and Krembs et al., 2002]. When ignoring the data from very slow ice growth rates, least squares fit gives K0 = 0.26 and δ/D = 7243 s/cm. Equation (2.28) was also examined by Nakawo and Sinha [1981] using data obtained from naturally-grown ice in Eclipse Sound, Nunavut, Canada. Least Squares fitting of the data gave K0 = 0.12 (corresponding to ln(1/Kδ − 1) = 1.7) and δ/D = 42000 s/cm. Figure 2.31 shows the plot of Kδ versus V on the basis of equation (2.28) using K0 = 0.12 and δ/D = 42000 s/cm. These values are also recommended by Weeks [2010]. The concentration of the salt retained in sea ice increases monotonically as the rate of ice growth increases. The figure also includes a plot of Kδ determined from a model presented in Cox and Weeks [1988]. This is a forward model based on their laboratory experiments of saline ice desalination during growth. It entails the following equations. K δ = 0 12, for V < 2 0 × 10 − 6 cm s

05

(2.30) (2.31)

Kδ =

0 26 for V > 3 6 × 10 − 5 cm s 0 26 + 0 74 exp − 7243V (2.32)

The discontinuity observed in Figure 2.31 at a growth rate of 31 cm/day using this set of equations is due to the transition between equations (2.31) and (2.32) that takes place at the threshold of 3.6 × 10−5 cm/s, which is equivalent to 3.11 cm/day. One aspect regarding salt rejection at the ice–water interface is the oscillating behavior of brine entering and leaving the skeletal layer at the bottom of the ice. The brine in the grooves between the dendrites at the interface oscillates with seawater imported from below, i.e., periodic oscillations usually exist between outflow of brine and inflow of seawater. Weeks [2010] described the two forces that cause the flow oscillation, namely the buoyancy force that pushes the seawater upward into the channels and the opposite pressure gradient that derives the brine out of the channels. The author relates their ratio to the radius of the brine channel and the brine volume. The oscillating behavior of brine in the skeletal layer was observed by Martin [1970] while studying the interaction between salt water and the underlain layer of fresh water. The author observed oscillations appear as a downward jet of salt water, followed by an upward jet of fresh water. In a later study Eide and Martin [1975] made some quantitative observations about the same oscillations that take place at the end of brine channels near the ice interface using a laboratory experiment setup. They froze simulated seawater in very thin plastic cells (3 mm thick by 350 mm wide and 500 mm deep) when cooling was applied

ICE PHYSICS AND PHYSICAL PROCESSES 61

only from above. They followed the oscillations for eight hours and noted that the duration of the downward flow was much longer than that of the upward flow. The most regular oscillation period was roughly one hour, with an 8–15 min inflow of saline water and an approximately 45 minutes outflow of ejected brine. It should be noted, however, that in the Eide and Martin [1975] experiment, the cooling was applied from the top against the gravity, which plays a big role in the desalination process. Moreover, the experiment was carried out by observing water between two closely spaced plastic plates to simulate the grooves at the sea ice dendritic interface. In this case, surface tension between the plastic and the water plays a vital role in the shape of the meniscus and complicates the whole situation. 2.5.5.2. Subsequent Slow Salt Rejection from the Bulk Ice The reader of this section should be informed that some material requires checking definitions and explanations in Chapter 4 about laboratory techniques for revealing structure of polycrystalline ice. The brine entrapped within the bulk of FYI ice sheet continues to drain slowly throughout the entire growing season. Cox and Weeks [1975] developed a simple empirΔS i ical equation to determine the rate of salinity loss in Δt sea ice as a function of brine volume Vb and temperature gradient in the vertical direction ΔT/Δz, ΔS i ΔT ΔT = 1 68 × 10 − 5 − 3 37 × 10 − 7 V b Δt Δz Δz (2.33) The slow desalination process occurs as a result of one or more of the following five mechanisms (1) brine pocket migration (also called brine diffusion), (2) brine expulsion, (3) gravity drainage, (4) brine flushing, and (5) brine mobility through subgrain boundaries. While all these processes cause brine to drain downwards to the bottom of the ice sheet, the first two mechanisms are known to cause some brine to go upwards to the ice surface as well. The fifth mechanism, proposed by the second author of this book, plays a role though not recognized in conjunctions with the other well-recognized mechanisms. The first four mechanisms are adequately described in the literature, based on the prevalent and widely held notions on sea ice as a cryogenic material [Untersteiner 1968; Weeks and Ackley 1982; Notz and Worster, 2009; Weeks, 2010, and Hunke et al., 2011). The fifth mechanism has not been recognized by the scientific community because of two aspects. Though logical, they are not easy to comprehend. The first and the foremost aspect is the high-temperature state of ice in nature. It is not easy to believe that, “ice is

hot.” The second is related to microstructural details of sea ice beyond the macroscopic views seen in thick sections in naked (unaided) eyes or thin sections under polarized light. Subsequently, desirable attention was not paid to the roles played by intergranular and intragranular, and hence transgranular, activities. Nonetheless, physics and mechanics of sea ice related to the substructure of grain boundaries (intergranular) and subgrains and their boundaries (intragranular) are well developed as briefly presented earlier (e.g., in Figure 2.22). Brine Pocket Migration Historically, brine pocket migration was believed to be responsible for the desalination of sea ice. This concept was introduced by Whitman [1926] and later led to the development of the theory of liquid zone migration described by Weeks and Ackley [1982]. It implies “marching” of pockets through the ice mass by diffusion. This is caused by the temperature gradient through the ice sheet, which is colder at the top and warmer at the bottom in winter. Since the brine pockets are usually elongated in the vertical planes along the length of columnar grains, the temperature gradient is also manifested within the brine pockets. It lends itself into colder and higher salt concentration in the upper end of the pocket and warmer with lower salt concentration near the lower end. Consequently, the solute diffuses from the upper to the lower end. This causes simultaneous freezing at the upper end of the pockets and melting at the lower ends. Hence, all the brine pockets essentially migrate downward. This trend may reverse in the early spring when ice temperature becomes warmer at the top of the ice sheet. Based on the theory of the liquid zone migration through solid crystals, Seidensticker [1965] obtained the following simple equation for estimating the velocity of migration of the brine pocket v toward the warmer side of the ice, v = 1 46

D mC

(2.34)

where, D is the diffusion coefficient of the salts in the seawater, C is the salt concentration in the brine, and m is the slope of the liquidus curve in the phase diagram. If the temperature at the location of the brine pocket is −6 C, for example, the theory gives a migration velocity 14.0 μm/h or equivalent to almost 10 mm per month. It should also be noted that brine pockets can migrate upward as well during the early spring when the temperature gradient is reversed. Untersteiner [1968] came to the conclusions that, during the lifetime of FYI, brine pockets migrate by an average of 20 mm downward near the ice surface, with an almost equal migration upward during spring. Weeks [2010] presented an evaluation of equation (2.34) in terms of migration velocities observed at different ice temperatures.

62

SEA ICE

Advection velocity of brine pockets caused by diffusion is modeled using the assumption of sea ice being a twocomponent porous medium (pure ice and brine) and presented in Notz and Worster [2009]. They found that the speed of salt diffusion depends on the temperature gradient in the bulk sea ice and independent of the geometric distribution or interconnectedness of the brine inclusions. They estimated also the typical advection velocity of brine pockets to be 10−9 m/s for a typical temperature gradient of 10 C/m and a brine salinity of about 100‰. These results confirmed observations obtained by Harrison [1965] and Jones [1974]. In any case, brine migrates with very small velocities and, therefore, this mechanism may not be considered as an effective contributor to sea ice desalination. Brine Expulsion Brine expulsion was suggested by Bennington [1967]. It is linked to the decrease in bulk ice temperature and the associated decrease in the volume of the brine pockets due to the solidification of more water inside the brine pockets. Consequently, salinity of the brine remaining in the pockets is expected to increase in order to maintain the phase equilibrium. Since ice density is less than that of water, the frozen ice inside the pocket occupies an approximately 10% greater volume than that of water from which it was formed. This mechanism was thought to increase the pressure inside the pocket, which may ultimately cause cracks around the pocket through which brine is expelled. While most of the expelled brine is expected to flow downwards into the warmer ice medium, some may be expelled upwards in ice near the top surfaces. The brine expulsion mechanism is well covered in the literature [Weeks and Ackley, 1982; Petrich and Eicken, 2009; Notz and Worster, 2009; and Weeks, 2010]. Cox and Weeks [1988] applied equations, developed in Cox and Weeks [1986], to calculate the change in salinity of a given layer in sea ice due to the brine expulsion mechanism. The work was motivated by an interest to estimate brine expulsion from ice immediately after their extraction in the field. Cores are then usually exposed to colder air temperature (particularly in the polar regions) and that is how brine expulsion is believed to become effective, thus spoil the salinity measurement. During cooling from temperature T1 to temperature T2 (T1 > T2) the amount of brine expelled from an ice layer is measured by the ratios of the ice salinities (S1(T1)/S2(T2)) and brine volume Vb(T1)/Vb(T2) according to the following equations (ratios are always < 1 and definition of brine volume is presented in section 3.5.1): 1 − ρ1

Si T 2 Sb T 2 = Si T 1 Sb T 1

and

i

ρb T 2 c exp ρb T 1 ρi S b T 1 − S b T 2 (2.35)

1

Vb T2 S b T 1 ρi c = exp Vb T1 Sb T 2 ρi S b T 1 − S b T 2

(2.36)

where Sb is brine salinity, ρb is brine density, ρi is pure ice density (918 kg/m3), and c is a constant equal to dρb/dT, which is 0.8 kg/m3. On the other hand, during warming of ice (T2 > T1), the above two equations take the form Si T 2 =1 Si T 1

(2.37)

Vb T2 S b T 1 ρb T 1 = Vb T1 S b T 2 ρb T 2

(2.38)

Brine salinity can be determined using the sea ice phase diagram (Figure 2.1) and brine density can be determined based on brine salinity using equation (3.22). In general brine expulsion is more pronounced in the upper layer of sea ice, particularly in thin ice. Cox and Weeks [1975] concluded from their laboratory experiments that brine expulsion was only important during the first few hours of ice growth and later became a minor desalination process relative to gravity drainage. Notz and Worster [2009] calculated the brine velocity caused by brine expulsion and found that it was less than the typical growth rate of sea ice at the ice–water interface. Therefore, brine expulsion can only lead to redistribution of brine within the ice sheet but not to drain from the sheet. The hypothesis that freezing-induced pressure may cause cracks along the wall of the brine pocket, in the bulk of an ice cover, is questionable. This is based on the knowledge gained on the physics related to the rheological properties of ice as a material at high thermal states and the rapidity of stress relaxation processes at high temperatures. This statement, however, is in need for some clarifications. In this respect, an effort will be made next to elaborate on thermal states of sea ice that will also be useful in understanding material science of sea ice. Of particular importance is the application of etching techniques for revealing structural damages, if any, in ice. Realistically speaking, the bottom temperature of floating sea ice sheets is always close to about −1.8 C. In-situ measurements of vertical temperature distribution in sea ice during the entire growth season, for a number of years, indicated that the surface temperature at snow–ice interfaces rarely goes below about −20 C even during the coldest time of the winter [Nakawo and Sinha, 1981]. Thus, the ice cover floating on its melt (water) exists at temperatures close to its melting point. As such, ice in nature always exists as crystalline solids at extremely high thermal states in comparison to that of its liquid state (see section 5.1.1 for details). For saline sea ice, the primary structural parts are always made of pure ice because the ice lattice does not practically allow any impurities. For the main component of sea ice covers, an average ice sheet temperature, T, of

ICE PHYSICS AND PHYSICAL PROCESSES 63

brine pockets in response to mechanical or thermal stresses. Common damage features of deformation in ice are the formation of tilt boundaries, dislocation pile ups and polygonization [Sinha, 1978a] (see definitions in section 5.1.3). However, damage features that can be associated directly with brine pockets, without any ambiguities, were never seen by the authors (N.K. Sinha) in undeformed young and FYI covers when examined shortly after sampling in field laboratories. Structurally damaged features were also not seen later following transportation and storage, and naturally subjected to unavoidable thermal changes and hence thermally induced localized strains due to the shrinking or enlargements of brine pockets. In fact, Figure 2.22 exemplifies subgrain boundaries of lattice mismatch but no etched features are directly associated or linked to the isolated brine inclusions. However, stress (strictly strain) concentrations around brine pockets induced by externally applied mechanical forces applied during strength tests have been routinely observed as described below. Figure 2.32 is a photomicrograph of a thermally etched horizontal section of columnar-grained, S3 type, FYI in a specimen that was subjected to low strain-rate mechanical loading during biaxial compression tests. This micrograph exemplifies an area practically inside one grain in which rows of brine pockets were trapped in basal planes. The c axis, normal to the basal planes, is indicated by . Here, the brine and air inclusions appear as darker objects. The layers of these inclusions can be seen as rows parallel to the basal planes, but they are not necessarily in straight lines. This can be visualized from the schematics presented earlier in Figures 2.19 and 2.21. Signs of stress (strain) concentrations surrounding these inclusions are revealed by the short straight lines. These lines (cross section of planes) are actually grooved in the planes and

Tilt boundary

−5 C (268 K) is equivalent to a homologous temperature, Th of 0.982 (section 5.1.1). This temperature, therefore, is only 1.83% below the melting point. A relatively cold temperature of −20 C, near the top surfaces, is equivalent to only about 7.3% below the melting point. At these high temperatures, mechanically induced stresses (strains) are relaxed extremely rapidly even if the pressure is applied “instantaneously,” say in less than 0.01 s [Sinha and Sinha, 2011]. Strains are accommodated by the elasto-delayedelastic-viscous (EDEV) mechanisms. Initial extreme rapidity in stress relaxation process (in fractions of one second) is governed by the high-temperature mechanism of delayed-elasticity governed by the shearing of grain and subgrain boundaries. However, in natural sea ice thermally induced strains are never developed rapidly, let alone instantaneously or at fractions of one second, mainly due to the thermal inertia of ice and the insulating effect of snow covers. Under snow cover, the sea ice surface is not sensitive to sudden and sharp decreases of atmospheric temperature, even under drastic drop from –5 C to −35 C (section 3.2). Thus, the changes in the vertical temperature distribution inside the ice mass occur very slowly. Consequently, due to the stress relaxation processes, stress concentrations cannot develop sufficiently high to produce cracks around the brine pocket through which brine can be expelled. Moreover, as can be seen in the schematic of Figure 2.21b and actual micrograph of Figure 2.22, brine pockets often exist along the subgrain boundaries that can also accommodate changes in the shape of the brine pockets. This is due to the fact that water molecules along these boundaries are at higher energy levels than those of the ice crystals because of small lattice mismatches at the boundaries between neighboring subgrains. Additionally, brine pockets often contain air bubbles. The air bubbles can also help in accommodating any expansion due to freezing of water molecules in the pockets. Thus, development of cracks at the brine pockets as suggested by the brine expulsion mechanism is rather remote. The above conclusion has been substantiated experimentally by the application of thermal etching (section 4.4.2.1) to delineate grain and subgrain boundaries as well as deformation that induces gross changes in the microstructure such as tilt boundaries or polygonization (often called recrystallization). Thermal etching of ice surfaces is possible only because ice exists at extremely high temperatures and water molecules can sublimate or directly transform from the solid state to the vapor state (section 4.4.1). Submicroscopic features related to the deformation induced by crystalline defects, such as dislocations, dislocation cells, pile up of dislocations, etc., can be revealed by chemical etching and replication described in section 4.4.2.2. These experimental techniques proved vital for performing forensic type of investigations of damages in ice lattice around

1 mm

Figure 2.32 Photomicrograph of a thermally etched surface, normal to the length of a columnar grain, after biaxial strength tests; “small-angle tilt boundaries” visible as straight lines parallel to c axis are linked to inclusions as stress concentration points (photo by N.K. Sinha).

64

SEA ICE

made visible as protruding up, with a three-dimensional effect, by a special arrangement of the source of light of the optical microscope. It can be seen that the lines are parallel to the c axis of the grain, an indication that they correspond to “small-angle tilt boundaries,” a term used in metallurgy for segments of small mismatch of the c axis on the two sides of the lines in case of the hexagonal lattice. In simple words, tilt boundaries form when a crystal is subjected to bending moments and the dislocations move to accommodate the strain, a process commonly known as recovery. Tilt boundaries can be recognized by the fact that they are straight segments, not curved, and parallel to the c axis. Dislocations in the basal planes can easily move (called slip) along the basal planes in ice. They can also leave the basal plane and climb to the next planes. This climbing mechanism is strictly a high-temperature phenomenon and can be monitored during their climbing motions as shown by Sinha [1987b]. Slipping basal dislocations can be blocked by internal obstacles. One of the most common obstacles is the grain boundaries. Pile up of dislocations also appear as straight segments on thermal etching, but since they are produced by slipping dislocations in the basal planes, the etched lines are at right angle to the c axis [Sinha, 1978a]. Major segments of subgrain boundaries in sea ice also correspond to the lattice mismatch or the angular difference in the orientations (a)

of c axis of the neighboring subcrystals. They are simply called as “low-angle boundaries.” Figure 2.33a exhibits an area near a junction of grains inside a specimen deformed due to failure under biaxial confinement. This shows an area of higher localized deformation and formation of cells or the beginning of disintegration of the inner lamellar structure of grains. With the increase in deformation, the number density of cells increases due to breakdown of the subgrains. As the strain increases, the inner structure of columnar-grained sea ice deteriorates to polygonized state and becomes unrecognizable to untrained eyes. Completely polygonized ice may be mistaken as granular ice. This is demonstrated in Figure 2.33b. It shows an area of the ice that is partially polygonized. This was particularly selected to show a few rows of brine pockets in a small segment on the left. During deformation, brine and air pockets are also forced to move and coagulate. A few coalesced pockets can also be seen in the micrograph. Note also the presence of both slip lines and tilt lines, in criss-cross form on the left. Going back to the topic of brine migration upward would be appropriate to include here. Relatively high salinities are commonly observed at the surfaces of thin ice as well as the base of the snowpack. The presence of high salinity can be tasted readily, and naturally this is a common knowledge among the Inuit people of the north. Martin [1979] reported surface salinities as high as 95‰ (b) 1 mm





1 mm

Figure 2.33 Thermally etched surface of biaxially deformed sea ice exhibiting (a) tilt boundaries, cells and grain/ subgrain boundaries at a junction of several grains with different orientations of c axis and (b) polygonized area inside one grain (micrograph by N.K. Sinha).

ICE PHYSICS AND PHYSICAL PROCESSES 65

from newly formed sea ice in the Beaufort Sea. The prime mechanism responsible for salt concentration at thin ice surface is the brine migration to the surface from layers immediately below the surface. Drinkwater and Crocker [1988] reported snow base salinities up to 80‰ in the Labrador Sea ice. They suggested the idea of “wicking up” brine by snow from the highly saline thin ice surface through capillary action. Moreover, when frost flowers grow on the surface of thin ice (section 10.4) they usually wick up brine from the surface and therefore have extremely high salinity that can reach 100‰ or more [Martin, Ducker, Fort, 1995]. Boxe and Saiz-Lopez [2009] discussed the effect of brine layers at the snow/ice interface on the polar environment and its effect on the trace gases within the snowpack and into the atmosphere in particular. Brine Drainage by Gravity Gravity drainage of brine is known to be the prime desalination mechanism where the greater weight of the brine relative to that of the seawater tends to drive the brine out of the ice–water interface. Once the sea ice thickness exceeds 50 or 100 mm a network of so-called brine drainage channels with their tributaries are formed. These are vertical or near-vertical conduits through which plumes of brine drain into the underlying seawater. Hence, gravity drainage requires the presence of a fairly dense interconnected network of brine channels with tributaries in order to carry brine by gravity action. The existence of this network was observed and described in Lake and Lewis [1970]. They concluded that horizontal migration of brine toward preferred drainage areas were analogous to the catchment area of a river and its tributaries. The frequency of occurrence of channels was roughly once every 33 cm2 in the case of thin ice ( 100 mm). This gives an average distance of 32.4 mm between brine channels for thin ice and 134 mm for thick ice. The brine channel formation and geometric characteristics of these channels were addressed in several studies [Bennington, 1967; Martin, 1979; Weeks and Ackley, 1982; Wakatsuchi and Saito, 1985; and Wakatsuchi and Kawamura,1987]. In a laboratory experiment on the growth of simulated sea ice in a 0.5 m deep freezing cell cooled from above, Eide and Martin [1975] observed vertical brine drainage channels with diameters of 1–3 mm and associated them with smaller feeder channels extended throughout the ice thickness. They also observed that the diameter of brine channels at the ice–water interface is much narrower than higher up in the bulk ice, so that the channel has a bottle neck at the interface. This is different from the drainage system in glacier ice which shows a reverse trend. Eide and Martin [1975] developed a

qualitative theory based on the difference in pressure head between the brine in the ice and the seawater under it to explain the formation of the channels and the onset of a convective instability. The latter explains the existence of the “neck” at the ice–water interface. Cole and Shapiro [1998] examined ice sheet in Elson lagoon, northern Alaska, during the winter of 1993–1994. They reported networks extended through the entire thickness of 0.3 m YI sheet. They did not observe any channel that extended completely through the ice sheet when it was 1 m thick. The initial population of channels terminated within the sheet, either randomly or at a specific depth. The typical length of the channel was found to be in the range of 0.3–0.5 m. Gravity drainage is enhanced under the following conditions. (1) The channel’s diameter is large. Larger diameters reduce the viscous drag of brine flow hence enhancing the drainage. (2) The local temperature gradient within the ice that surrounds the channel is steep enough to result in more expulsion of brine from the pockets to feed into the channel. (3) The pressure associated with channel formation is relatively low to facilitate the flow of brine into channels. (4) The intensity of the channels and their tributaries that originate from the pockets is high. Higher intensity increases permeability, and therefore drainage. Since all major salts dissolved in seawater precipitate at a temperature of −22.8 C or less, there cannot be any measurable brine drainage in the ice below this temperature. At temperatures above −22.8 C brine drainage is expected to increase with the increase in ice temperature. Cox and Weeks [1975] stated that the rate of gravity drainage depends on the brine volume and temperature gradient of the ice. As either factor increases, the rate of change of salinity due to gravity drainage also increases. The authors also concluded that the gravity drainage is more active when the brine volume fraction is higher than 0.05, i.e., the volume of pure ice fraction is less than 0.95. Wettlaufer, Worster, Huppert [1997] employed the mush Rayleigh number Ra, which is defined as follows, Ra = h − z

gρl βS br, z ϕυ, max kμ

(2.39)

where (h − z) is the distance between the ice–ocean interface located at depth h and the level z in the ice, g is the acceleration due to gravity, Sbr, z is the salinity of the brine and ρlβSbr, z represents the difference between the density of seawater and that of brine at level z, k is thermal diffusivity, μ is the dynamic viscosity of the liquid, ϕυ, max is the maximum solid volume fraction and (ϕυ, max) is the effective permeability of the ice as a function of ϕυ,max.

66

SEA ICE

Brine Flushing Brine flushing is a commonly known mechanism of ice desalination [Untersteiner, 1968]. This process is normally active after the growth season and during the beginning of the melt season in late spring and summer. It is mainly caused by melting of snow and ice at the top of ice sheets. The hydrostatic pressure overhead of brine and meltwater results in percolation through the ice sheet, mainly vertically and also horizontally [Eicken et al., 2004]. In this case, salinity of the upper 0.5 m of the ice sheet is usually reduced to less than 1‰ down from a typical value of 4–7‰. Eicken et al. [2004] also uses Darcy’s law to calculate the scale of brine flushing downward as a result of meltwater percolation. Notz and Worster [2006] used a onedimensional model to simulate summer time flushing of surface melt. Most of the salt rejection due to the brine flushing mechanism results from the vertical percolation of surface melts. Notz and Worster [2009] ran the same model to simulate the evolution of melt pond percolation during the course of 1 hour into 0.9 m thick sea ice with a bulk salinity of 10‰ (these numbers may be too high to be realistic) and surface temperature of −5 C increasing linearly to −1.8 C at the ice–ocean interface. They found that brine flushing was very active. After 1 hour, the melt pond had drained almost completely and the bulk salinity had decreased by more than 2‰. Meltwater percolation into the ice sheet brings the ice sheet into an isothermal condition. Eicken et al. [2002] reported that up to 25% of the meltwater produced at the surface melt of Arctic ice is retained within the ice sheet. Due to ice permeability, some meltwater can be retained within the ice. So far, no discussions or model has been presented to explain the permeability of sea ice and its dependency on the interconnected grain and, particularly, subgrain boundaries in the bulk of sea ice. Brine Mobility Through Subgrain Boundaries The lamellar or subgrain boundaries in sea ice are in a relatively higher energy state than the ice inside the platelets (or subgrains) with their long directions in the vertical planes. Consequently, diffusivity of the boundaries is also higher than the lattice diffusion of the plates. At higher temperatures, the subgrain boundaries can act as passages allowing the brine to flow. The total surface area of the subgrain boundaries exceeds the total surface area of the grain boundaries. The network of subgrain boundaries, therefore, provides very effective passages for migration of brine. This mechanism also explains the feeding of brine channels through the connected brine pockets along the boundaries leading to the formation of tributaries of the channels. As explained earlier, the subgrain boundaries represent locations within the grain where the c axis is not matched exactly with respect to its general direction within the grain. This mismatch develops when the dendrites in the skeletal layer protrude in the

liquid and are subjected to local disturbances. Therefore, they grow at slightly different orientations, forming subgrains, and eventually merge to form a “family” of closely oriented crystalline entity known as grain. Actually, the conventional thin sectioning technique using warm plates (section 4.2.1) inadvertently shows the cellular structures of sea ice grains because of the impregnation of brine along the boundaries due to melting. This, together with smearing of brine on both surfaces of thin sections, does not assist in the clarifications of the role of individual brine pockets and the subgrain boundaries. Figure 2.34 is a photograph of a horizontal thin section of columnar-grained FYI at a depth of 2.05 m, produced using cross-polarized light (section 4.3.2). The section was prepared from an ice core, extracted from Mould Bay in March 1985, within a few hours after coring. To preserve the clarity of the surface from smeared brine, the thin section was prepared using the doublemicrotoming technique (section 4.2.2) at a low temperature of −20 C. It was later allowed to warm up inside the field laboratory to about −5 C. The lamellar structure with distinct boundaries is readily visible with the subgrain boundaries impregnated with liquid brine, forming connected passages. The same thin section was also examined at the lower temperature of thin sectioning at −20 C, but the connectivity of brine and subgrain boundaries was not as obvious as presented earlier in Figures 2.22. A few features implied in Figure 2.34 are worth pointing out though they are not related to the subject addressed in this section. These are issues related to the

Figure 2.34 Horizontal thin section of columnar-grained ice at a depth of 2.05 m in Mould Bay, March 1985, under crosspolarized light, exhibiting brine impregnated subgrain boundaries and air bubbles. Section prepared at −20 C was allowed to warm up to −5 C; double arrows indicate crosspolarizers and black color is due to c axis of this grain parallel to one of the polarizers (micrograph by N.K. Sinha).

ICE PHYSICS AND PHYSICAL PROCESSES 67

effect of temperature on the microstructure of sea ice as well as the use of the cross-polarized light in viewing thin sections of directionally solidified, columnar-grained sea ice. The orientation of the layers of brine in cross sections of columnar grain is the simplest method for ascertaining the c axis of the crystal. The approximate c axis () orientation of several grains is shown for clarification of this point. The orientations of the “polarizer” and the “analyzer” (see section 4.3.2) and hence their cross-position was marked during the photography. This is shown by the crossed double arrow positioned inside a grain with black color for maximum visibility. The pass direction of the polarizer and the analyzer are indicated, respectively by the vertical and horizontal arrows in Figure 2.34. The black grain retains this color because its average c axis was parallel to the pass direction of the polarizer. The layers of brine in this grain were nearly parallel to the pass direction of the “analyzer” and could easily be noticed by slightly rotating the specimen holder of the polariscope when the grain changes to a lighter hue. An insight into the role of subgrain boundaries for desalination can also be gained by examining the microstructure of ice surrounding the brine drainage channels in Figure 2.35. This is a cross-sectional view of a horizontal thin section of the FYI from Northstar Bay, Greenland in early March of 1994. It shows mature brine channels. Two views are presented under parallel and crosspolarized light (section 4.3.2). The parallel-polarized light photo shows clearly brine channels fed by smaller tributaries or “feed arms.” The channels, however, seem to disappear when the section was viewed under cross-polarized light. Careful examinations can still recognize their presence, but that is more of an after-effect from seeing image under parallel-polarized light. The star-shaped channels

(a)

are distributed in the ice body with their cores at distances of about 30 to 40 mm apart. The diameter of these channels or the length of the “feed arms” (or tributaries) is less than about 15 mm. Therefore, there are several subgrains with interconnected boundaries in between the core of these channels. These boundaries must have served as the paths for diffusion and hence migration of brine both laterally and vertically. As a result, the shape of the arms of the channels was linked directly to the geometry and orientation of the subgrains (this appears more clearly in Figure 2.27). Moreover, at any given stage of desalination, the brine entrapment is anticipated to increase laterally with the decrease in the distance from the core of the brine channel. This actually is the contributing factor for the visibility of the arms under parallel-polarized light in Figure 2.35. As mentioned before, this explains why sea ice researchers do not report anything about brine channels in thin sections of matured first-year sea ice. Customarily they observe thin sections through cross-polarized light; parallel-polarized light is rarely used. Even the method of making fabric diagrams, popularized by Langway [1958], involves the use of cross-polarized light (see Weeks [2010] for details of this method).

2.6. ICE DEFORMATION Except for fast ice, which forms and remains attached to a shoreline, an ice wall, a grounded iceberg, or the seafloor over shallow parts of the continental shelf, floating sea ice usually undergoes a complex perpetual motion at different scales. The mobility of the ice is triggered by one or more of the following geophysical forces: wind stress, ocean current stress, internal ice resistance, Coriolis force,

(b)

Figure 2.35 Photos of horizontal thin section showing star-shaped brine drainage channels using (a) parallelpolarized and (b) cross-polarized photographs. The section was prepared from columnar-grained S3 FYI type in Northstar Bay, Greenland, in early March 1994 at –20 C (micrograph by N.K. Sinha).

68

SEA ICE

sea surface tilt, and tidal force. Thorndike and Colony [1982] identified wind as the primary force and ocean current as the secondary. The response of the ice cover to those forcing lead to collision of individual floes (compression), shear force between the mobile ice and an adjacent stationary surface or divergence of ice sheets. Hutchings et al. [2011] indicated that deformation of ice surface is controlled by a balance between external forces and internal dissipation. Forms of deformation resulting from the compression and shear forces are shown in Figure 2.36. These forms, in addition to the formation of leads in the sea ice cover, caused by fracture of ice sheets under tensile stress, are addressed in this section. The energy expended in ice deformation determines the large-scale strength of pack ice [Weeks and Kovacs, 1970; Tuhkuri and Lensu, 2002; and Hopkins, Tuhkuri, Lensu, 1999]. Ice deformation is important because it affects key parameters that contribute to and serve as indicators to climate variability. These include ice thickness increase (when ice sheets overlap or ice blocks pile up to form ridges or rubble fields), lead opening, and (to a limited degree) the spatial variability of ice salinity since brine tends to drain faster from upturned ice blocks of ridges or rubble ice. Ice deformation also contributes to the air-water drag coefficients and the momentum transfer from the ocean currents to the bottom of the ice sheet. Thus, it is an important input to ice–ocean–atmosphere dynamic models. Sea ice deformation occurs at different temporal and spatial scales, depending on the stimulating force. Traditionally, sea ice deformation is considered at three spatial scales: small, medium, and large. Small-scale deformations range from a few hundred meters to a few kilometers and are manifested in the forms of fracturing, rafting, ridging, and rough ice surface

Compression

Rafted ice

Shear

(rubble ice). They are driven by the wind action on the ice surface as well as the interaction between individual ice floes in response to internal forces within and between them. At this scale, the deformed ice represents hazardous conditions for both marine navigation and offshore structure. Rough ice along marine navigational routes delays marine traffic, while mechanical loading of deformed ice threatens offshore structures. Medium-scale deformations are defined by a spatial scale that extends a few tens of kilometers. They are usually manifested in the form of heavy and extensive ridging as well as lead formation in the ice sheet. At this scale, the deformation is driven by mesoscale weather/or oceanic forcing. Large-scale deformations with characteristic dimensions in the order of hundred to several hundred kilometers are caused by large circulation systems, particularly in the Arctic. The circulation controls convergence and divergence of the regional ice cover and creates shear zones. A key synoptic weather pattern that causes such circulation and deformation is polar or subpolar gyre. All forms of deformation that exist at the medium scale can also be active at the large scale. There are two major circulation systems in the Arctic, which trigger large-scale ice motion that causes significant divergence and convergence (hence deformation) of the Arctic ice cover. They are the Beaufort Sea Gyre in the west and the Transpolar Drift Stream in the east (Figure 11.45). The latter was discovered in 1893 when the Norwegian explorer Fridtjof Nansen sailed to the Arctic on a specially built ship that was designed to sail through sea ice. He wanted to prove that the Arctic current flowed from Siberia westward to Greenland. To do that, Nansen boarded with his team on the ship “Fram” (meaning forward), which was purposely let caught up in the Arctic ice near New Siberian Islands off the coast of Siberia. The ship indeed moved with the pack ice but

Pressure ridge

Rubble ice

Shear ridge

Figure 2.36 Forms of sea ice deformation resulting from compression or shear forces.

ICE PHYSICS AND PHYSICAL PROCESSES 69

Rafting, along with pressure ridging, are the most common forms of ice compression at small and medium deformation scales. Rafting is more likely to occur when two thin ice floes/sheets (thickness between 20 and 60 cm) push against each other. In this case, the thinner sheet overrides the thicker one (Figure 2.36). It may also be encountered with thicker ice when relatively small-size floes collide. Obviously, rafting (and ridging) increases the ice thickness. Bailey, Feltham, Sammonds [2010] recommended that the thickness and strength of rafted ice (and of pressure ridges) has to be taken into account in the design of Arctic vessels and offshore structures. In rafting, a moving thin ice sheet/floe overrides another thin ice piece when they collide. The riding continues under compression force and against an increasing frictional force between the rafted segments. It eventually stops when the frictional force between sheets arrests motion or causes buckling in the ice sheet. Hopkins, Tuhkuri, Lensu [1999] suggested that the frictional force is proportional to the product of the difference between the unit weight and the buoyancy of either sheet multiplied by the length of the overlapping segment between the two sheets. The rafting mechanism is well documented

in literature that addresses ice jams in rivers and lakes [Michel, 1978]. Rafting is manifested in two forms: (1) simple rafting whereby one sheet simply overrides the other and (2) finger rafting whereby two sheets interlace when they meet; pushing over and under each other forming finger-like interlacing edges (Figure 2.37). The latter pattern occurs when both sheets are ≤ 15 cm thick [Weeks and Kovacs, 1970; and Vella and Wettlaufer, 2007]. In an early publication that addressed thin ice in Lake Superior, Green [1970] found that most fingers range from 0.5 to 3 m wide and 2 to 10 m long. This author observed that narrowest fingers (30–100 mm) occurred in extremely thin and fragile ice (2–3 mm thick), whereas the wider ones (up to 10 m) occurred in ice between 10 and 20 mm thick. The fingers shoved forward by traction from a light breeze at a rate ranging from about 5 to 50 mm/s. Another rafting situation, yet resulting in thicker ice, can be seen at the bottom right corner of the photograph in Figure 2.38 where the edge of the rafted floe is visibly bent. Bending stress associated with such thick floe usually results in fracturing the floe. A noticeable feature in the photograph is the piling of fractured ice from the weaker floe when collided with the stronger floe that survived the crash. Rafting has received relatively less attention in ice mechanics studies compared to ridging (particularly pressure ridging). This is because it occurs mostly in thin ice, which does not exert significant mechanical loading. As such, it has been of less concern to engineers. Rafted ice is usually composed of layers of equal thickness separated by thin layers of ocean waters. That is because when the top ice piece is dragged on the bottom piece, the brittle

(a)

(b)

rather slowly and erratically [Berens, 2010]. Nansen’s theory was proven. However, it has been proven later that the driving force for this large-scale ice motion was the surface wind, not ocean current. The wind acts through the frictional drag of the rough ice surface.

2.6.1. Rafting of Thin Ice

Figure 2.37 (a) Simple rafting shown at the bottom right corner of the ice cover in the Arctic (from CIS archive), and (b) finger rafting in the Antarctica [from Vella and Wettlaufer, 2007, American Physical Society].

70

SEA ICE Thinner ice sheet

Sail

Thicker ice sheet

Remnant of a broken floe Keel

Rafting

Figure 2.38 A variety of ice floe with different sizes and broken ice pieces in the Labrador Sea. Note the rafting of two relatively thick ice pieces at the bottom right corner of the picture and the consequent ice bending and fracturing (the photograph was taken by N.K. Sinha from the deck of an icebreaker).

dendrites will melt, leading to presence of a saline layer between the two pieces. This point is explained toward the end of section 10.1 (Figure 10.2) within the context of explaining the remarkably high backscatter from rafted ice. Modeling of thin ice rafting is not frequently presented in the literature. A one-dimensional thermal consolidation model for rafted sea ice is presented in Bailey, Feltham, Sammonds [2010] using parameters representative of ice in the Caspian Sea, the Arctic, and the Antarctic. The authors concluded that it took 15 hours for two layers of rafted sea ice to consolidate but the consolidation is sensitive to the initial thickness of the liquid layer and the fraction of salt release during freezing. More information on rafting can be found in Geiger, Hibler, Ackley [1998] and the website of the Antarctic Sea Ice Processes and Climate Program (ASPeCt). Modeling of thin ice rafting is not frequently presented in the literature. 2.6.2. Ridging of Thick Ice As a rule of thumb, if the collided ice sheets/floes are thin, then rafting is more likely to occur and if they are thick, a pressure ridge will form upon collision. Details on favorable conditions for rafting and ridging are presented in this section. In general, ridges are created by the flexural failure of the edges of the ice sheets; forming broken blocks as shown in Figure 2.39. Ridges exist in two forms depending on the mode of formation; pressure and shear ridge. Pressure (also called compression) ridges develop as a result of direct collision between ice floes. Shear ridges, on the other hand, may develop when shear force acts at the boundary between stationary (fast) ice sheet and floating ice. It may also

Figure 2.39 A sketch illustrating the formation of a ridge. The thinner ice sheet is the leading sheet that collides with the thicker sheet. Blocks are broken from the leading sheet and pushed against the thicker sheet at a constant speed to form the sail and the keel of the ridge.

develop locally between the boundary of a large floe and a highly fragmented ice zone [Kovacs, 1970]. Shear ridges are frequently short in length and straight in the plane view [Weeks, 2010]. In both cases of pressure and shear ridges the ice undergoes extensive crushing and consists of numerous upturn blocks. Most of the ice blocks go below the waterline, forming the keel of the ridge, while a smaller number of blocks are above the waterline, forming the sail of the ridge. The sail is composed of dry ice blocks with snow- and/or air-filled pores (more explanation follows). While pressure ridges contribute significantly to the ice thickness, shear ridges do not contribute as much but they impact ice drift and deformation near the shores. When the blocky edge of landfast ice at the far end from the shoreline is grounded (this is mostly a manifestation of shear ridge), the fast ice is known as Stamukhi [Barnes et al., 1987]. On mobile ice floes, pressure ridges are formed mostly parallel to the direction of the prevailing wind and/or current, but they may eventually act as sails and turn the floes to become parallel to direction of wind/or ocean current. Pressure ridges are different than the raised edges of relatively thin ice floes when they collide as shown in Figure 2.40 (finger rafting appears also at the top of the photograph). Note the curvilinear shape of the ridge, which probably followed the boundary of the ice floe. However, when ridges originate from collision of extensive ice sheets, they resume long linear shape. Raised edges are limited in width but the ridging process continues as long as the impinging process of the ice sheets continues. Piling of ice blocks at the top ice surface ends when the sail reaches its maximum height but it continues at the bottom of the ice because the maximum draft of the keel is significantly larger than the sail (usually four times larger). The brine that drains from the ice blocks at the top surface eventually freezes and fuses ice blocks together if the weather permits. Additional bonding develops if the

ICE PHYSICS AND PHYSICAL PROCESSES 71

Figure 2.40 Raised edges of thick-enough ice floes after collision. The picture was taken during the melt season when the snow cover melted and the ice blocks became visible. Edges are raised by 10–30m cm (photo courtesy of M. Johnston, NRC).

surfaces of the ice blocks melt then refreeze together. The blocks usually pile with a steep angle slope. In the case of a shallow sea, the keel of a ridge may reach the sea floor and the ridge becomes grounded. Ridges affect the ice thickness distribution; hence, a proper representation of the ridging process in large-scale sea ice models is of interest not only for marine navigation but also for climate studies in polar regions. A few studies that confirmed the role of ridging in increasing ice thickness in the Arctic include Timco and Burden [1997], Obert and Brown [2011], and Strub-Klein and Sudom [2012]. Similar studies of ridges within the Antarctic ice include Weeks, Ackley, Govoni [1989], Jeffries and Weeks [1992], and Jeffries and Adolphs [1997]. Worby et al. [1998] reported that the mean leveled FYI thickness in the Antarctic is rarely greater than 0.6 m, and that ridging rather than thermodynamic growth is the dominant mechanism for increasing the floe thickness beyond about 0.6 m. Hibler III and Ackley [2003] measured shadows from ridges in aerial photographs to determine the distribution of various heights and widths of pressure ridges. They incorporated a few concepts from the probability theory. Pressure ridges are also observed within individual floes (Figure 2.41) as well as in “ridge zones” in pack ice (Figure 2.42). Those ridges usually extend linearly or curve-linearly for a few hundred meters or a few kilometers. Ridge zones are major obstacle to the mobility of marine vehicle in the Arctic. An example of ice blocks that form a fairly small unconsolidated ridge in the Fram Strait are shown in Figure 2.43. The photo was taken during a survey of pressure ridges in March 2012. Tucker III and Govoni [1981] found that ice block size is somewhat correlated to the ridge height. They measured block

Figure 2.41 A small ridged ice floe viewed from the bridge of the Norwegian icebreaker KV Svalbard on March 13, 2012 in the Fram Strait near the eastern edge of the pack ice off the east coast of Greenland. (photo taken by D. Sudom of the National Research Council of Canada).

Figure 2.42 Pressure ridges (indicated by arrows) shown in a ridge zone in this airborne photo from the Beaufort Sea near Alaska on April 1st, 2007. Details of the ridge are shown in the inset (photographed by Dr. Pablo Clemente-Colon of the National Ice Center, Washington).

Figure 2.43 Upturned ice blocks that form a small pressure ridge in the Fram Strait off the east coast of Greenland. The ridge sail height is about 2 m and keel depth 6.5 m. (photo taken by D. Sudom of the National Research Council of Canada).

72

SEA ICE

dimensions from 30 ridges located 200 km offshore in Prudhoe Bay, Alaska. They found the largest variation of block thickness ranging from 0.7 m to 2.13 m in a single ridge and that ridge height is proportional to the square root of block thickness. The sail of a pressure ridge in the Arctic varies in height from a few meters to 20 m depending on the thickness of the original ice before deformation. The keel, on the other hand, can reach 50 m in the Arctic and 16 m in temperate regions such as the Labrador Sea [Timco, Croasdale, Wright, 2000]. Statistics of the spatial frequency and heights of ridges in different offshore areas have been presented in previous studies [e.g., Kovacs et al., 1973, Hibler III, Mock, Tucker III, 1974, and Weeks and Gow, 1980]. From measurements of mean thickness of ice blocks as well as visual observations in the southern Sea of Okhotsk (north of Japan) Toyota et al. [2007] found that ridging dominated the deformation process of ice when its thickness is above 0.4 m. The Sea of Okhotsk, however, is significantly warmer than the Arctic. Extensive field and statistical modeling on ridging of FYI and MYI were conducted by scientists in the National Research Council of Canada (NRC) in the 1990s. Timco and Burden [1997] presented a suite of geometrical parameters that characterize ridges (Figure 2.44). These include sail height Hs, keel depth Hk, sail width Ws, keel width Wk, sail area As, keel area Ak, sail porosity Ps, keel porosity Pk, sail slope αs and keel slope αk. They compared those parameters from profiles of 184 ridges reported in 22 different studies and developed exponential best fitting of data points. Table 2.3 summarizes the bestfit power regression relationships given for ridges on FYI and MYI. The number of data points (n) used in the regression and the correlation coefficient (ρ) between the two parameters in each equation are also given. The first four equations apply to FY ice ridges. Average values of ridge angles as well as simple empirical linear relationships between the ridge parameters based on analysis of

112 ridges in different sites in the Arctic are presented in Figure 2.45. Since both rafting and ridging are caused by collision of ice floes/sheets, then one of the questions that are often asked is: under what conditions does rafting or ridging occur? This question has been addressed in a number of studies including Weeks and Kovacs [1970], Parmerter [1975], Hopkins, Tuhkuri, Lensu [1999], and Tuhkuri and Lensu [2002]. The factors that determine the mode of deformation include ice thickness, modulus of elasticity, geometry of the leading edge of the sheet, and the degree of uniformity of thickness distribution within each ice floe/sheet. Generally speaking, conditions may be favorable for rafting when the two ice pieces are thin enough and have more or less the same thickness. Parmerter [1975] developed an analytical model of rafting

Ws,sail width Ps sail porosity As sail area

Sail

Hs sail height

αs

Sea level

Consolidated layer αk Keel

Hk keel depth

Pk keel porosity Ak keel area

Wk keel width

Figure 2.44 Geometrical idealization of a first-year sea ice ridge with definitions of characteristic parameters [Timco and Burden, 1997 / with permission of Elsevier].

Table 2.3 Summary of parametric relationships for ridges of FYI (first four rows) and MYI (last two rows); n is the number of data points and ρ is the Pearson correlation coefficient [adapted with modifications from Timco, Croasdale, Wright, 2000]. a Equation

n

ρ

Value

Standard deviation

Hk = aHbs

97

0.793

4.60

0.31

Wk =

aHbk

65

0.746

5.67

Wk =

aHbs

75

0.713

aAbs

b 95% confidence limit

Value

Standard deviation

95% confidence limit

4.00

0.88

0.79

0.79

1.13

3.41

0.87

0.72

1.01

30.75

1.98

16.81

0.78

0.65

0.90

33

0.896

17.46

4.82

8.73

0.82

0.70

0.94

Hkm =

aHbsm

47

0.878

3.66

0.30

3.06

0.91

0.82

1.01

Akm =

aAbsm

10

0.921

8.82

4.42

−1.42

1.00

0.76

1.24

Ak =

Note: Hkm, Hsm, Akm, and Asm are parameters pertaining to MYI, following the definitions in the text.

ICE PHYSICS AND PHYSICAL PROCESSES 73 Ws

2.6.3. Formation of Ice Rubble Field

αs = 20.7° (temperate) = 32.9° (Beaufort)

Hs

Pk = 0.14 + 0.73 Ps αk = 26.6°

Hk Wk

Hk/Hs = 4.4

Wk/Hs = 15.1

Wk/Hk = 3.9

Ak/As = 8.0

Figure 2.45 Average ratios and relations between FYI ridge parameters based on the analysis of 112 ridges. Dashed line is the sea level[after Timco and Burden, 1997 / with permission of Elsevier].

based on the geometry of the two colliding sheets. Results emphasized two findings: (1) two sheets can raft if their thickness is not significantly different and (2) the probability of rafting decreases as ice thickness or its modulus of elasticity increases. On the other hand, ridging is more likely to occur when the ice is relatively thick as mentioned earlier. Upon abrasion and crushing of the two sheets, their edges keep changing until conditions become right for ridging. This implies a transitional mode between rafting and ridging where broken ice pieces that accumulate at the surface of one sheet can lift the other sheet above it. Using a model that simulates motion of twodimensional ice blocks, Hopkins, Tuhkuri, Lensu [1999] simulated rafting and ridging between two identical sheets. They confirmed the established conclusion that rafting was the likely process when the two sheets have similar thickness even if the thickness is not small (e.g., between 50 and 90 cm). The authors also examined the effect of uneven thickness distribution on the ridging/rafting processes. The simulations showed that rafting predominates when each sheet has uniform thickness and ridging predominates when the sheets have an uneven thickness. It also showed varying mixtures of rafting and ridging behavior in a parametric space of thickness and a metric representing the uniformity of thickness distribution. In general, Hopkins, Tuhkuri, Lensu [1999] concluded that irregularity of thickness distribution of either one or the two sheets is an important parameter that determines the likelihood of rafting or ridging. Depending on the thickness of the two sheets and homogeneity of the thickness distribution, the result of the collision of the two sheets varies from (1) simple two-layer rafting, (2) multiple layering rafting, (3) ridging with some layering of rafting, or (4) pure ridging.

Following the initial impact of collision between two ice sheets, if the momentum of crushing continues, an extensive area of crushed ice (instead of just ridging) is formed. This is called rubble field. A photograph of rubble ice in Lancaster Sound, Canadian Arctic, taken from a helicopter that flew at about 100 m above the ice is shown in Figure 2.46. Rubble ice is an accumulation of randomly dispersed fragments of small ice blocks that covers a larger expanse area without any particular order. It is triggered by storm cycles that cause intensive ice compression. The broken pieces are similar to those produced when ships pass through ice. Unlike ridges, the broken ice blocks in a rubble field do not accumulate vertically; hence the thickness of rubble ice is considerably less than the thickness of ridges. Rubble ice is more observed in the Antarctic ice because ocean swells continue to break up newly formed ice and herds the broken pieces into rubble ice field. Another photo of rubble ice taken from the deck of an icebreaker sailing through the North Water Polynya in the Arctic is shown in Figure 2.47. Kovacs [1981] presented information on rubble ice formation in the Bering Sea and Norton Sound in Alaska, and quoted observation on the extent of the rubble field and the size of ice blocks. A few impacts of rubble ice have been studied. Barker and Timco [2005] studied the effect of rubble ice on the protection of offshore structures in temperate region. From the analysis of ice load data, the authors found that rubble field can be very effective in attenuating ice loads of structure. They suggested that if rubble is generated around the structure the ice load will be reduced. This is because the rubble ice dampens the direct crushing of mobile ice on the structure. Barker

Figure 2.46 A photograph of a rubble field of FYI in the Lancaster Sound, Canadian Arctic, taken in May 1991. Note the smooth floes in the middle of the rubble (photo by M. Shokr).

74

SEA ICE

Figure 2.47 Rubble ice sheet covering part of the North Water polynya, north of Baffin Bay, taken in April 1998. The dark strip at the back is an opening in the pack ice and the topography at the top of the picture is part of Ellesmere Island (photo by M. Shokr).

and Timco [2005] discussed the development and the use of the innovative technology of ice rubble generators. In an experimental study, Scourfield, Lishman, Sammonds [2022] found that the floe-floe effective friction can be raised by the presence of rubble. This has implications for assessing friction loads on ships and for modeling Arctic Ocean dynamics. 2.6.4. Fractures in Ice Cover The term “fracture” refers to any opening that exposes seawater to the atmosphere. It may vary between the narrowest form of a crack (a few centimeters to 1 m width) to the largest form of a lead, a passageway through ice which is navigable by ships. A brief description of each form is presented in the remaining of this section but more discussions on their formation identification using remote sensing observation are presented in section 10.1.4. Cracks are openings in an ice sheet, which are developed when the sheet diverges or shears in order to relieve the localized tensile stresses [Schulson and Hilber III, 1991]. They are usually observed in fast ice, consolidated ice or a single but large FYI floe. A crack may form in response to wind or tidal action which forces a split of an ice sheet or breaks fast ice. Figure 2.48 illustrates a typical crack in FYI sheet in the Resolute Passage. This crack extended for a few kilometers. Note the zig-zag path of the crack and its varying width. That is because cracks follow weak links along grain and subgrains within the ice polycrystalline structure (section 2.5.3). Figure 2.49 includes a pair of photographs of a thermally induced crack in fast ice that ran across the Frederick Hyde Fjord (83 1 N, 29 50 W), at the northern tip of Greenland in May 1994. The ice was used as an operational runway for a Boeing 727 jet (see Figure 1.16). This

Figure 2.48 A crack in 1.2 m thick FYI in the Resolute Passage, Canadian Arctic in May 1990. Note the variation of the crack width and its zig-zag path. The width is 25 cm at the widest part (photo by M. Shokr).

was the most northerly point on floating sea ice cover that a Boeing 727 jet aircraft had ever landed [Pole, 1995, Sinha, 1995]. This exercise was also used for validation of the aircraft landing manual [Sinha et al., 1996] as illustrated in Figure 1.14. The average ice thickness was 2.32 m, the average width of the crack was around 150 mm and the freeboard was almost constant at 220 mm. The crack appeared suddenly overnight on May 22, 1994 after a rapid drop in the air temperature. It widened slowly during the next three days. It became a real hazard and accordingly the “touch-down” point for the aircraft had to be adjusted. In the background of the Twin Otter aircraft in Figure 2.49, several vertically oriented snowcovered grooves on the mountain can be seen. These are sites of snow avalanches that also bring crumpled rocks at the bottom. Strangely, except for the near-shore areas, there was no snow on the entire ice cover for several kilometers. The ice looked green-turquoise from the air. The team had to haul snow from the side of the mountains to spread on the ice in order to increase the albedo of the selected 3 km long runway strip, and therefore keep the ice cooler and increase the traction [Sinha, 1995]. This process took several days.

ICE PHYSICS AND PHYSICAL PROCESSES 75

Figure 2.49 Crack in the fast FYI in Frederick Hyde Fjord, Greenland with the tail section of a Twin Otter aircraft shown on 25 May 1994 (left). From the refrozen crack depth of 200 mm (right) the freeboard of ice sheet could be measured (photo by N.K. Sinha).

Leads are the most common form of large-scale deformations caused by divergence of sea ice sheet. They are usually observed within pack ice or between pack ice and land. They can often branch or intersect, creating a complex pattern of linear features. If leads are formed between floating pack ice and fast ice, they are called flaw leads. These are commonly observed in the Eurasian Arctic region. Pärn and Haapala [2011] studied their frequency of occurrence in the Gulf of Finland during the period 1971–2007. If a lead exists between the shore and the pack ice it is called a coastal lead. In addition to providing navigational routes to marine vehicles, leads are important for Arctic sea ice ecology and have a significant effect on the heat exchange between ocean and atmosphere. While the ice sheet acts as an insulator between ocean and atmosphere, leads enhance the heat flow to the much colder atmosphere in winter in the polar regions. As a result, although leads typically represent only a few or at most several percent of the ice pack area, they may account for 50% of total heat flux from the ocean to the atmosphere [Ruffieux et al., 1995]. Lead width varies from a few meters to a few kilometers. Leads may contain brash ice and/or covered with nilas and/or YI. Figure 2.50 is a photograph of a narrow lead in the Beaufort Sea area. It shows that the lead may not constitute a homogeneous cover. In this aerial photo refrozen ice appears at one side of the lead while open water fills the rest. A few ice pieces may detach from the surrounding

Figure 2.50 A short narrow lead in FYI sheet. Note the refrozen ice at the border of the lead and the ice pieces in the middle (photo courtesy of NASA’s Operation IceBridge project).

ice sheet and drift in the lead as shown in the upper part of the figure. So, leads may contain pieces of mature ice. This complicates lead identification and statistics from remote sensing data. are

76

SEA ICE

of sea ice may respond to wind fields integrated over a distance of a few hundreds of kilometers upwind. Therefore, synoptic atmospheric pressure patterns, which create large-scale wind fields, can create divergence, convergence, shear or any combination of these stresses over a very large area of the pack ice. Using remote sensing imagery, the National Snow and Ice Data Center (NSIDC) in Colorado reported a fracture in the Arctic ice, observed in February 2013, which extended for thousands of kilometers from Ellesmere Island to Barrow (Alaska). 2.7. ICE DECAY AND AGING Figure 2.51 A lead in Fram Strait with floating ice inside, showing frost smoke generated from water evaporation at the surface and condensation above the surface (photo by D. Sudom of the NRC of Canada).

Open leads have low albedo relative to the surrounding ice. However, in the Arctic winter the enormous temperature difference between the atmosphere (typically < −30 C) and the ocean (at freezing temperature of −1.8 C) causes the lead to freeze within a few hours of formation. The temperature gradient between newly opened lead and the atmosphere can be so high that the lead “steams” with frost smoke (Figure 2.51). This is fog-like clouds formed by the contact of colder air with relatively warm water when evaporated water condensates immediately above the surface. It composes of ice particles instead of water droplets and may persist while ice is forming. This turbulent heat and mass transfer from leads to the atmosphere during winter affects the atmospheric processes hundreds of meters above and hundreds of kilometers downstream from leads. Hence, in addition to their importance for marine navigation and wildlife (polar bears often hunt near leads and seals use them to breath), leads are also important for their impact on weather and climate systems [Eisen and Kottmeier, 2000]. In concluding this section, two notes are worth presenting. The first is about the formation of cracks in the ice sheet. Although large-scale ice deformation occurs under compression forces, it is well established that tensile strength of ice is significantly lower than its compressive strength at any given rate of deformation [Hawkes and Mellor, 1972; Sinha, 1983; Sinha, 1987a; Richter-Menge and Jones, 1993]. Although compressive strength at a given temperature depends strongly on the rate of loading (strain- or stress-rate), tensile strength is significantly less rate-sensitive. However, bending of ice sheets during compressive interactions produces tensile stresses at the bottom of the ice. This results in tensile failures, thereby producing visibly large cracks. The second is about the possibility of formation of extremely long cracks. It should also be noted that a piece

It would have been more appropriate to start this section with discussions on ice aging followed by ice decay (or melting) as this is the natural sequence. However, the term “ice aging” is used in this section to refer to the transformation of FYI to MYI (not aging of FYI or MYI). On the other hand, the term “ice decay” is used to refer to the decay or partial melting of FYI before it can possibly turn into SYI. Therefore, the ice decay is discussed before moving to ice aging. It should be noted that aging and decay of ice are further discussed in section 6.2.3 in a more focused viewpoint in relation to experimental data. Similar to ice formation and accretion, the onset of ice decay depends on the latitude (a proxy of air temperature). For example, ice decay is observed during June–July around Eureka in Ellesmere Island, Canada (80 N); 2 months later than ice decay in Labrador Sea. In the Labrador Sea (latitude between 50 N and 60 N), the ice shows its first signs of disintegration in March and melts completely by the end of April. The rate of ice accretion in Eureka is more than twice the rate in the Labrador Sea. At the peripherals of the Arctic region, the overlying snow layer on FYI starts to melt in mid-June. By the end of July or August most of the FYI would have melted. 2.7.1. Ice Decay The onset of FYI decay is marked by systematic melting of the snow/ice surface which then proceeds slowly through the ice subsurface layer before it accelerates as it infiltrates the ice sheet. This process takes between two weeks to several weeks, depending on the climatic conditions and air temperature of the region and average thickness of the ice. Melting may also start at the ice bottom albeit at a much smaller rate. This decay process may accelerate in the presence of microalgae, which absorb light and convert it into heat [Zeebe et al.,1996]. The magnitude of melting at the bottom of the ice is negligible from the aspect of energy or mass balance of sea ice but may be considerable from the biological viewpoint. Ice decay causes both ice thickness and its mechanical strength to decrease [Hibler III, 1979]. For that reason,

ICE PHYSICS AND PHYSICAL PROCESSES 77

operational ice centers have incorporated information on ice decay within their products of ice strength charts. In May 2002, the CIS began the production of a prototype of ice strength charts based on an algorithm developed by the Canadian Hydraulics Centre of the NRC. The algorithm estimates the ice strength using modeled surface air temperature. The output charts display the mechanical strength of undeformed FYI during its decay relative to its mid-winter strength [Langlois et al., 2003]. The prime factors that trigger ice decay are air temperature, incoming solar radiation (ISR) and melting of snow cover. Secondary factors include tides, albedo, mechanical disruption (e.g., ice breakup due to wind), and water temperature (warm ocean current washing against the bottom of the ice in late spring causing erosion). Ice decay starts at the surface in the form of melting spots initiated by two heat sources: (1) the absorbed solar radiation and (2) the conductive heat from the surrounding air. The amount of absorbed solar radiation is determined by surface albedo, which varies with the type of surface; namely snow-covered or snow-free as well as the degree of snow wetness or ice surface dryness. Albedo values of these surfaces are presented in section 9.1. Water absorbs 90% or more of the ISR while melting snow absorbs 40–60% and dry snow absorbs 10–20%. Higher absorption of solar radiation leads to greater increase of local temperature of the surface. Because of the variation of the absorption of the ISR at the ice surface, areas that contain high concentration of dust usually act as centers for melting. Moreover, areas covered with snow with brine-rich bottom (especially when the snow is thin) become susceptible for rapid melting. This leads to the formation of a few small ponds at the locations of these spots (Figure 2.52). Once this radiationdriven process starts, the water in the pond absorbs even more radiation and the pond starts to expand and deepen slowly. The other force that comes into action is heat conduction. When warm air comes in contact with the initial pond, heat is transferred from the air to the water by conduction.

The ponds tend to form on the ice surface in depressions, or are simply retained within surviving snow pack as areas of slush. While the hydrostatic head of the surface melt provides the driving force for the water to percolate through pores and remnants of brine channels in the ice fabric, an interconnected network of pores is also formed to complete this process. Bilello [1977] described the appearance of an ice sheet while being in the state of decay as if it is covered with “giant spiders” with the “body” being the thaw hole and the “legs” being channels of meltwater draining laterally toward the hole. The ponding stage of decaying first-year (FY) level ice in Allen Bay (74.72 N and 95.25 W) near Resolute is shown in Figure 2.53. The scene is covered by melt ponds interspersed by raised areas of snow-covered ice. The figure also includes another scene of melt pond on a MYI surface depression in the same geographic site. This ice was 1.27 m thick under a hummock surface (the white areas) and 0.9 m thick under depression [Johnston, Frederking, Timco, 2003]. The drainage of water during ice decay washes away brine (see brine flushing in section 2.5.5.2). Therefore, the decayed ice (also called rotten ice) has a much lower salinity than the original ice. As the ice underneath the ponds becomes thinner and more solar radiation is absorbed, thaw holes are produced (Figure 2.52) and this enhances the melting process. A thaw hole is a large melt path through the ice at its thinnest point or at the melt pool’s deepest point. A single thaw hole is enough to drain a large area of the ice surface. Eventually meltwater drains off into the sea through thaw holes, existing cracks, /or over the side of ice floes. Some drainage may also occur in FYI covers through seal breathing holes (Figure 1.9) as reported in detail by Digby [1984]. Once the disintegration of ice sheets proceeds at the surface to the point where free water surfaces appear, the rate of ice decay significantly accelerates. In an early study, Bilello [1977] assembled weekly measurements of ice thickness over a period of 10–15 years from seven stations in Canada and Alaska to explore the effect of air temperature and ISR on ice decay.

Melting start Pond

Pond

Thaw hole

Water surface

Sea ice

3–4 weeks

Figure 2.52 Decay process in sea ice. It typically takes 3–4 weeks from the onset of surface melt for the appearance of thaw holes in Arctic ice and two more weeks until the disappearance of the ice sheet if conditions permit.

78

SEA ICE

Figure 2.53 Two scenes of decaying ice in Allen Bay in the vicinity of Resolute, Canadian eastern Arctic: FYI on June 21, 2002 (left) and MY ice on August 11, 2002 (right) [Johnston, Frederking, Timco, 2003].

–10

2

0 –5

–7

–30

–2

–1

–40

–4 –2

–4

–50

–6

–3

Ice depth (cm)

–3

–20

–2

–4

–6

–60

–6 –5

–70

–8

–80 –10

–2

–90

–3

–4

–12

–100 130

140

150

160

170

180

Julian day

Figure 2.54 Ice temperature profile before and after the onset of ice decay (JD166). Temperatures were measured from landfast ice in McDougall Sound, Canadian Arctic in 2001. Contours indicate constant temperatures. Note the increase of temperature as the ice decay progressed [Johnston, Frederking, Timco, 2002 / Government of Canada].

Properties of FYI during its decay were measured during several field programs. In the Mould Bay experiment (see section 6.2), the decay of the FYI cover from a thickness of 2.17 m on 21 June to 1.27 m on 13 July 1982 was associated with drastic change in the vertical salinity profiles from C to shape. The temperature profiles of sea ice before and during its decay are shown in Figure 2.54 from measurements conducted in McDougall Sound in the Canadian Arctic from 7 May (Julian Day 127) to 3 July C

The study used accumulated thawing degree-days (ATDDs) as indicators of air temperature. This is defined as the number of days when the daily average of air temperatures is above 0 C. It is the counterpart of the accumulating FDD, defined as the number of days when average daily temperature is below the freezing point (see section 2.4.4.1). Using air temperature data from the Resolute Bay station, the study found that the rate of fast ice decay can be defined by two linear relations (at different slopes) with air temperature. From the start of the decay until the ice reaches 0.9–1.0 m, the average ablation rate was approximately 0.009 m/ATDD. After that thickness range, the average ablation progressed at a much faster rate; approximately 0.17 m/ATDD. Two anomalies appeared in 2 years (1967 and 1971) and could be attributed to extremes of snow accumulation. Air temperature did not depart appreciably in these two seasons, but the snow cover in the beginning of the 1970–1971 winter season was double to four times as deep as it was in the beginning of 1966–1967 winter and almost twice as deep as at the beginning of the ice deterioration in May (0.55 m in 1971 and 0.28 m in 1967). Bilello [1977] concluded that the extreme variation in the rate of ice ablation can be attributed to the difference of the amount of snow on ice. Thick snow delays the ice decay and vice versa. In this case, the relation between ice and the ATDD does not apply. The ISR has the same effect as air temperature on decreasing ice thickness during ice decay. The problem, however, is the difficulty of using it as an indicator because of the unpredictable cloud cover. While many studies have addressed sea ice decay in the Arctic from the viewpoint of the ice extent decline in response to global warming [Stroeve and Notz, 2018; Liu et al., 2016], its impact on modulating the mechanical properties of ice is of great interest to marine engineers and operators. Evolution of a few physical and mechanical properties of sea ice during its decay is highlighted in the rest of this section.

ICE PHYSICS AND PHYSICAL PROCESSES 79 h_s 2000 h_s 2001 h_i 2000 h_i 2001

Thickness (m)

0.0 –0.5

Mean daily air temp. Resolute, 2001

–10

–1.5

h_s: snow depth h_i: ice thickness

–2.0 130

150

10 0

Ice surface temp., d = 10 mm, 2001

–1.0

20

170

190

Temperature (°C)

0.5

–20 –30 230

210

Julian day May

June

July

Aug

Figure 2.55 Ice and snow thickness along with air and surface temperatures of landfast ice in McDougall Sound, Canadian Arctic before and after the onset of ice decay (approximately JD166). Ice surface temperature was measured at a depth of 10 mm [Johnston, Frederking, Timco, 2002 / Government of Canada].

2000 (JD184), using thermistor string installed Johnston, Frederking, Timco, 2002]. As the season progressed, the temperature at all depths steadily increased. After 9 June (JD160) more than half of the full thickness of ice was isothermal at –3 C. By 20 June (JD170) the entire ice thickness was isothermal at –2 C during the stable morning hours. By the end of the program, 3 July (JD184), the temperature gradient had been inverted; i.e., ice in the surface layers became warmer than the bottom ice. During the same field campaign, measurements of ice thickness were related to measurements of snow depth. Figure 2.55 shows that the onset of ice ablation in mid-June coincided with the point at which the snow cover had melted completely and the ice surface reached –1.8 C. Ice ablated from an average thickness of 1.51 m to 0.83 m in about four weeks in 2000 at a rate of 22 mm/day. The 2001 field measurements, which terminated after about 1 week of ablation, showed that the ice thickness decreased from about 1.44 m in early May to 1.20 m at end of June. When the temperature of the sea ice approaches the melting point, the mechanical properties of the ice are modulated significantly, resulting in less mechanical load on marine structures. This is explained in terms of physics of ice crystalline structure as follows. As ice temperature rises, brine pockets acquire more water from the melt of the surrounding pure ice crystals, leading to expansion of the pocket. This causes an increase of brine volume associated with a decrease in brine salinity. Moreover, as the ice temperature continues to increase, the boundaries between the grains and the intragranular subgrain boundaries (Figures 2.21 and 2.22) also get impregnated with liquidlike lattice structure. The increase in temperature, therefore, affects rheological properties of polycrystalline ice [Sinha, 1978b] including the creep and failure response of

natural sea ice [Sinha, Zhan, Evgin, 1995; Zhan, Sinha, Evgin, 1996] in a profound manner. Compressive strength of ice, the most important parameter for ice-structure interactions, is extremely sensitive to strain-, stress-, or simply deformation-rate. The behavior is very complex. However, the engineering physics and micro-mechanisms of fracture and failure in both freshwater and seawater ice are well understood and the temperature and rate-sensitive strengths are predictable. Thus, even before any surface melt indication, the ice may have already lost a significant amount of its midwinter strength. This is the main drive behind the interest of the CIS and Transport Canada in ice decay. The deformation and failure mechanisms in sea ice are also controlled by the mobility of lattice dislocations. Johnston, Frederking, Timco [2001] performed their NRC-BHI tests at a fixed indentation rate (to simplify the test procedures) and used the “3mm-yield BHI strength” for quantifying the strength decay as shown in Figure 2.56. An index for strength, such as the plate pressure corresponding to the indentation depth of, say 3 mm or 5 mm (may be called as “3mm- or 5mm-yield strength”) can be used as an “index strength” for comparison purposes. The figure shows that air temperature and ice temperature influence the onset of ice decay. The mean air temperatures rose above zero for the first time on 10 June (JD162) and the snow cover rapidly melted. The first signs of decreased ice strength coincided with the decrease in snow cover and increase in air/ice temperatures. Ice desalinated rapidly and began to ablate after the insulating layer of snow melted. The ice strength continued to decrease until it reached a plateau. The stabilization in ice strength coincided with the point at which the ice had desalinated almost completely.

SEA ICE 10

Bulk ice salinity (‰)

8

10

Mean air temperature Bulk salinity

5 0

6 Top ice temperature

–5

4 –10 2

Temperature (°C)

80

–15 –20

0 0.5

16 Snow thickness

14

Thickness (m)

12 10

–0.5 Ice thickness –1.0

8 6 4

Borehole strength (MPa)

0.0

–1.5

σ 3 mm full thickness in situ borehole strength –2.0 140

150

160

170

180

190

200

2 0 210

Julian day

Figure 2.56 Relationship between physical property measurements and NRC-borehole indentation pressure, σ 3mm, corresponding to the indentation depth of 3 mm, in FYI during decay period [Johnston, Frederking, Timco, 2001 / Government of Canada]. Note that the σ 3mm is significantly less than the “ultimate compressive strength” of ice that a floating structure will face during interactions; it is only an index for comparison purposes.

2.7.2. Ice Aging If FYI survives one summer melt season, it becomes SYI on October 1 according to the nomenclature set by the World Meteorological Organization (WMO). Any ice older than SYI is referred to as MYI. Sometimes the term old ice is used to indicate SYI and older. Alternatively, the term “perennial” ice is used to differentiate the older ice from the “seasonal” ice that grows during a freezing season and completely melts by the end of the season. In this section, the term “ice aging” refers to the aging of ice from one year to the next (i.e., not aging within the same freezing season). Details on physical processes that govern this transformation are presented in section 6.2.3. Data on sea ice age are presented in section 11.6. Ice ages in the Arctic Ocean because this region is essentially land-locked, therefore floating ice remains circulating

within the basin for several years. For example, under the influence of the Beaufort Sea Gyre (BSG) in the western section, ice may remain circulating for 7–10 years before it drifts south. The drift takes the routes shown in Figure 11.45. In contrast, the Antarctic sea ice grows along the peripheral of the Antarctica; therefore it becomes free to drift into the warmer open sea northward where it melts. The ice motion in both the polar regions is driven by the ocean currents and atmospheric circulation. MYI in the Antarctic is limited compared to the Arctic. About 80% exists in the Weddell Sea as it keeps circulating this area carried by the Weddell Gyre. This is one of the two gyres that exist in the Southern Ocean, a clockwise circulating current in the eastern side of the Antarctic Peninsula. MYI ice usually accumulates at the west of the Weddell Sea (see map in Figure 13.7). Other MY ice exists in isolated embayment at other locations around the Antarctica coast.

ICE PHYSICS AND PHYSICAL PROCESSES 81

Figure 2.57 A photograph of MYI floe located in Parry Channel, Canadian central Arctic in May 1991, showing the undulating topography of the surface. The snow cover ranged between 0.1 to 0.5 m. A few elongated hummocks are marked by the arrows. The tent was put up for the research team to shelter while having hot drinks during their sampling work (photo by M. Shokr).

MYI is physically distinct from FYI by surface features, physical properties, and thickness. While FY ice exhibits level, ridged or rough (rubble-formed) surfaces, MYI surface features undulating topography in the form of alternating hummocks and depressions but both are smooth with respect to the centimeter-scale wavelength of the microwave signal. Figure 2.57 shows the surface of MYI floes in Parry Channel in the Canadian central Arctic with hummocks and depressions clearly visible. This gives the MYI surface a “hill and dale” appearance [Johnston and Timco, 2008]. Hummocks are remnants of FYI ridges, which were weathered by wind and melt actions during the melt season, hence they usually assume elongated shape (following the typical shape of ridges). Note that snow usually fills gaps between ice blocks of the ridge, so when it melts and later freezes, the entire ridge turns into a solid mass. Depressions, on the other hand, result from melt pond formation on FYI surface. During the next freezing season, the brackish pond water in the depression freezes and is usually called “melt pond

- Drainage network can be established - Bubbles are formed from remains of drainage network

ice.” This is composed mostly of superimposed ice as the meltwater or rain water freezes at the surface. From crystalline structure viewpoint, the top layer of melt pond ice is usually composed of covered with snow (granural) ice (section 5.3.3.1). MYI is much less saline and its hummocky surface contains air bubbles that could be dense with considerably larger size than the bubbles in FYI (section 5.4.2.1). However, there is still notable difference in the salinity and density between hummock and melt pond ice. The top surface of hummock ice is less saline and remarkably bubblier than the top surface of pond ice. Field measurements presented in numerous studies by several authors have confirmed that the salinity of the top 0.2 m of MYI hummocks varies between 0 and 0.5‰ while that of melt ponds is between 0.2 and 2.0‰ (section 3.3). The density varies between 0.7 and 0.85 kg/m3 for hummock ice and 0.82 and 0.92 kg/m3 for melt pond ice (sections 3.4. Photographs of thin ice sections that show the bubble contents of hummock and pond ice are presented in section 5.4.2. A scenario of bubble formation in MYI is introduced in Shokr and Sinha [1994] to explain the abundance of air bubbles in the subsurface of hummock ice and its scarcity in melt pond ice. It is conceptually introduced in Figure 2.58. As the ice temperature increases during the melt season, brine pockets/channels increase in size and tend to group into larger channels. When surface melt starts the water percolates through these channels, forming a more connected and effective drainage network. Some channels terminate at the ice surface as shown and this makes them function as a conduit to drain the surface melt as shown in the network presented in Figure 2.58. The surface melt continues to percolate through the ice volume during the summer until the freezing starts in the fall. By that time the surface melt slows down while the water in the drainage network continues to drain by gravity. As a result, channels will be partially empty. At freezing, the empty parts become air bubbles. In other words, bubbles at or near the surface are formed

- Drainage channels cannot be established - Surface melt, snow, and rain accumulate and later consolidates to form a new ice layer

Figure 2.58 A graphical representation of MYI surface during melt season. Hummock and depression (melt pond) surfaces respond differently to surface melt and that leads to the generation of more and larger air bubbles in hummock than in melt pond ice (sketch is not to scale) [Adapted from Shokr and Sinha, 1994].

82

SEA ICE

from the remains of water drainage channels established in the summer. This scenario explains also the commonly observed interconnected air bubbles near the surface of hummock ice and the decrease of bubble intensity further down the surface. Perovich and Gow [1996] confirmed the origin of air bubbles in MYI from the evolution of the brine pockets during the transition of FYI into MYI. Moreover, they pointed out the importance of understanding how the elongated brine pockets develop into the rounder air bubbles. On the other hand, the flat surface of melt pond ice (Figure 2.58) does not allow formation of drainage network. In this case, surface melt, along with possible snow and rainwater, accumulate on the top of the ice surface. On subsequent freezing season, this layer consolidates to form a “newer” superimposed ice layer that could consist of snow ice (structurally-speaking). Since the solidification develops from pure or brackish meltwater, the salinity of this ice will be minimal. Nevertheless, formation of air inclusions in melt pond ice is possible within the granular texture of snow ice through two mechanisms: (1) air entrapment between snow particles, and (2) air originally dissolved in water and later rejected during freezing. The size/geometry of air bubbles resulting from these two mechanisms is different, with much larger and intensive bubbles produce from the first mechanism (Figure 5.38). Another difference between FYI and aged MYI is manifested in the mechanical strength. Johnston, Frederking, Timco [2003] measured borehole strength of old ice floes in the Arctic. Borehole strength is three-dimensional confined, compressive strength of ice, measured using a borehole jack. The study found that FYI in summer has about 10–20% of its mid-winter strength. On the other hand, two multi-year floes of about 5 m and 6 m thick were sampled in June and August and showed that the ice strength was 56% and 47% of their maximum mid-winter strength (about 30 MPa), respectively. These results clearly highlight the importance of distinguishing between FYI and older ice for the purpose of determining the ice load on marine vessels and structures. It is commonly known that ice thickens as it gets older, whether within the same season (i.e., for seasonal ice) or as ice ages from one year to the next. In fact, data from NSIDC (see Table 13.1) have shown that average MYI ice thickness increases as it progresses from one year to the next (though these data were published in 2016 and need to be updated). However, the notion that ice thickens as it gets older is questionable for either seasonal or perennial ice. Sea ice may expose to partial surface or basal melt during the same growth season and more so during the summer melt. Since the theme of this section is the age of perennial (not seasonal) ice, then the above statement

may also raise the following two questions: (1) is MYI always thicker than FYI, and (2) given the current shrinking of Arctic ice in terms of extent and thickness (section 13.3.2), has the thickness of each ice age category declined at the same rate? In addressing the first question let us consider the transition from FYI to SYI but the same reasoning applies also to ice aging from any year to the next. The SYI develop from the partially melted FYI during summer. Since the latter has a wide range of thickness and can partially melt into any thickness during summer, the SYI can start with any thickness at the onset of the next freezing season. Therefore it is difficult to associate a thickness range to SYI. Also, it is important to consider the ensuing accretion of a new ice layer at the bottom of the SYI (or MYI) as the freezing season progresses (section 6.2.3). As a result, the thickness of the SYI or MYI, may become comparable to or even less than the typical thickness of fully grown FYI. The second question can be addressed by examining the interannual evolution of thickness of each ice age. This has not been established because it requires combining ice age and thickness, which has not been available yet from a single satellite platform. Nevertheless, data on the interannual evolution of the areal coverage (without thickness) of each ice age are offered by NSIDC. This is presented in section 11.6 (Figure 11.44) and in section 13.3.3 (Figure 13.15). The conclusion from these data is that the loss of aerial coverage has occurred at more or less the same rate for each ice age category. This reflects the fact that all ice age categories are advected out of the Arctic Basin at the same rate. Up to the mid-2000s, the maximum thickness of FYI in the Arctic was around 2.5 m while MYI thickness typically varied between 2–4 m although thinner or thicker data was reported. For example, Melling [2002] found thick MYI floes of 6–8 m thick in the Canadian Arctic Archipelago (CAA) region with hummock (hillock) of one floe reaching up to 20 m height. The locations were near-shore areas, where wind forces can thrust the ice hard against the coast of several islands. Johnston and Timco [2008] reported MYI thickness from samples obtained from a few locations in the Arctic. Figure 2.59 shows data from transects chosen arbitrary along the surfaces of two very thick MYI floes in the Nares Strait (North of Baffin Bay) and the CAA region. The CAA region is known to have the largest accumulation of MYI, blocking the North West Passage (NWP) for marine navigation. It can be seen that the draft is much rougher than the height of the surface above the freeboard. This is partly attributed to the process of the ice accretion at the bottom, which is heterogeneous as described later.

ICE PHYSICS AND PHYSICAL PROCESSES 83 5

5

Ice thickness (m)

Top surface 0

0

–5

–5

–10

–10

Top surface

Bottom surface –15

Bottom surface

–15

–20 –140

–120

–100

–80

–60

–40

–20

–20 0

Distance (m)

20

0

20

40

60

80

100

120

Distance (m)

Figure 2.59 Top and bottom surface profiles of MYI in the Nares Strait, North of Baffin Bay (left) and the Canadian Arctic Archipelago (right). The two points with thickness more than 15 m (left) represent the maximum length of the thickness auger (the ice could have been thicker). The missing points from the top surface correspond to locations of thick snow accumulation exceeded 1 m at some places so that the crew could not remove it completely [ Johnston and Timco, 2008].

While it is possible to distinguish FYI from MYI either visibly in the field or upon using radar remote sensing data, it is not easy to identify the age of old ice. However, this distinction is important for marine operators because the mechanical load is proportional to the age of the ice. In order to provide end users with information and tools to identify old ice types, the Canadian Hydraulics Centre of the NRC completed a study to develop guidelines for the identification of these types [Timco et al., 2008]. Trained ice observers in the CIS use these guidelines to distinguish SYI from MYI with sufficient confidence while boarding a ship (this obviously can happen only in summer). A few key information that supports this task includes: (1) surface appearance and history tracing of the ice floe, (2) the edge of floe as SYI floes exhibit more angular geometry than the mostly rounded edges of older ice, (3) melt ponds on SYI surface are blue or greenish blue and more elongated with preferred direction while ponds on older ice are usually blue with no preferred shape or orientation, (4) SYI has less number of hummocks with less height than older ice, (5) drainage patterns that interconnect puddles and ponds at the surface are less extensive on SYI, (6) fragments of SYI overturned by a ship or any structure are less bluish than fragments of MYI. None of the above geometric or surface features could be used to reliably distinguish between SYI and older ice in remote sensing data. However, one physical feature (though realized only through ice microstructure analtsis) that distinguishes SYI from older ice is the less intensive bubbly layer within the subsurface of SYI. This feature impacts a few radar parameters, which can be to distinguish between the two types in Shokr et al. [2022] using polarimetric SAR data.

As mentioned earlier, regardless of the age of the MYI, it always encompasses a layer of new ice that continues to grow at the ice–water interface as long as winter conditions prevail (section 6.2.3). In a study by Bjerkelund et al. [1985], the authors conducted detailed monitoring of FYI throughout its aging process into second- and third-year ice in the Mould Bay, Canadian western Arctic. The study covered the period from October 1981 to May 1983 (i.e., two successive ice formation seasons). The aging process is shown in Figure 2.60. FYI grew during the winter of 1981–82 to a maximum thickness of 2.17 m with 0.65 m of overlaid snow by mid-June 1982, when surface melt started. By end of June, the snow had melted completely and melt ponds had covered 90% of the surface. This area decreased shortly thereafter to less than 50%. By end of August 1982, the ice thickness attained a minimum of 0.7 m, by combination of surface ablation and bottom melt. Re-growth of ice at the bottom of SYI started sometime in September 1982. The growth rate was similar to that of the original FYI in the previous year. It is interesting to note that by the end of the 1982–1983 freezing season, the SYI thickness reached the same thickness of FYI observed in the 1981–1982 season. Snow also started to accumulate on the SYI surface at the same rate as in the previous year. The salinity profile of the FYI in May of the first winter (1981–1982) depicts the familiar C-shape. The remarkable observation, however, is observed in the salinity profile obtained in April 1982. There is a sharp increase of salinity across the interface between the SYI the underlying newer FYI. Salinity jumps from 0 to 3.7‰ over a short vertical distance of 0.05 m. The fabric of the columnar ice growth is preserved across the interface as shown in Figure 6.17.

SEA ICE Salinity (‰) 0 8 16

Snow

Surface 81 / 82

0 FY ice

Surface melt Snow

100 Ice thickness (cm)

100

Ice

gro

Surface 82 / 83

SYI

wt

m t to Bo elt m

"FY ice" Ice

100

gro

wth

200

300 Year 1 Growth 400

0

h

200

S

N

J

Year 2 Growth

Melt M

1981

M

J

1982

S

N

J

M

M

J

Snow depth (cm)

Salinity (‰) 0 8 16

100

Ice thickness (cm)

Snow depth (cm)

84

300

1983

Figure 2.60 Ice composition and thickness, along with snow depth in Mould Bay, Canadian western Arctic from September 1981 to mid-May 1983. Re-growth of new ice (referred to as FY) under the SY ice after September 1982 is shown in the 82/83 data [Bjerkelund et al., 1985 / Government of Canada].

2.8. SEA ICE CLASSES Ice classes presented in this section pertain mostly to sea seasonal ice (perennial ice is usually grouped into one category—old ice). This classification is relevant to the needs of operational users in ice-rich waters and to some extent the ice climatology requirements. A few existing sets of sea ice classes are based on different criteria. The most commonly-used criterion is thickness-based where each ice type is associated with a range of ice thickness. Thickness is associated with seasonal ice age and therefore this criterion is also considered to be age-based. It has been adopted by the World Meteorological Organization [WMO, 1985] and frequently used to classify ice types in operational ice charts. Other classification criteria are introduced in this section. In addition to those criteria (which are introduced in the rest of this section), sea ice can be classified using one of the three sets of dual classification; fast ice versus floating ice, level versus deformed ice, and seasonal ice versus perennial ice. It is appropriate to emphasize that the term “ice class” is designated also to ships in order to reflect their strength for navigation through sea ice. The term “ice class ship” refers to a ship with high tensile strength keel and hull material. This counters the term “ice breaker,” which refers to a ship that has heavily fortified structure. The ice classes addressed in this section pertain to the ice, not the ships.

Fast ice refers to ice attached the coastline, an ice wall, or a grounded iceberg (mostly in the Antarctic region). If the water is shallow, ice may be grounded to the sea floor. Fast ice may be formed “in situ” from water or by freezing of floating ice of any age to the shore. It can extend to a few meters or several tens of kilometers from the coast and may last for more than one year, hence second- or MYI fast ice is possible [MANICE, 2005]. Floating ice, on the other hand, is any form of ice found floating in water though sometimes is blocked by surrounding ice. When packed together in large masses, floating ice is called pack ice. Level ice has a fairly flat surface, not affected by deformation. Deformed ice has its surface affected by deformation as described in detail in section 2.6. Seasonal ice is the ice that does not survive summer melt (i.e., completely melts during summer). Perennial ice has survived at least one summer melt as mentioned earlier. The following criteria for ice classification are covered in the rest of this section. They are used in generating information in operational ice charts. 1. Ice thickness, which is related to the ice age. This criterion is known as “stage of ice development.” It includes five major ice types: new ice (NI), nilas, YI, FYI, and old ice. Sub-categories exist within each major category. 2. Ice form, which includes pancake ice, ice floes, ice breccias, brash ice, fast ice, and anchor ice. 3. Ice concentration, which is defined as the ratio of ice within a given area of the ocean. It is expressed in tenths in

ICE PHYSICS AND PHYSICAL PROCESSES 85

operational ice charts. The major categories under this criterion include consolidated ice, compact ice, very close pack/drift, close pack/drift, open drift, very open drift, and ice-free (open) water. 4. Ice surface form, which includes two major categories: level (smooth) ice and deformed ice. The deformed ice includes rafted, ridged, and hummock ice. Other subcategories are included in each category. An additional surface feature, which is described in terms of the snow cover, is called sastrugi. It is characterized by sharp, irregular wavy ridge pattern of drifted snow caused by wind erosion and deposition. The sea ice types based on the above-mentioned classification criteria are presented in Tables 2.4–2.7. Table 2.4 includes classes based on stage of development which implies ice thickness. Table 2.5 includes classes based on ice forms, while Table 2.6 includes classes based on ice concentration. The ice classes based on surface features are shown in Table 2.7. Most of the definitions of ice classes in the tables are obtained from the “Manual of Standard Procedures for Observing and Reporting Ice Conditions” or MANICE [2005], prepared under contract for the CIS.

2.9. SEA ICE REGIMES The term “sea ice regime” refers to a sea ice region that has certain features based on its geographic location or some dominant ice attributes such as ice types, forms,

thickness, mobility or degrees of deformation. Examples of sea ice regimes that are based on geographic locations include the semi-enclosed central Arctic, coastal areas, and subarctic estuaries. The regimes are described in terms of their overall ice features. For example, the central Arctic ice regime features mainly MYI all year round. The subarctic ice regime features mostly seasonal ice with medium thickness (up to 0.7–1.2 m) or thicker in winter. A specific regime such as the Gulf of St. Lawrence in eastern Canada features seasonal thin FY ice up to 70 cm thick in February/March period with a rough surface and highly dynamic nature. The Baltic Sea ice regime has unique features triggered by the brackish water from which ice is formed. This includes its low bulk salinity and porosity [Granskog et al., 2006]. The small ice regime of Storfjord around Svalbard (the Atlantic Ocean) is characterized by significant temporal variations of ice cover caused by changes of wind direction pushing the ice either toward or away from the coast. This triggers opening of coastal polynyas that are usually covered with frazil ice [Dierking, 2010]. In general, ice information that is usually used to describe these geographic-based regimes includes ice types, thickness, climatology, onset of melt, and ice duration and variability. Some ice attributes can also be used to identify an ice regime. These include fast ice, deformed ice, rough pancake ice, seasonal or perennial ice, polynyas, etc. Those regimes hold significant operational or climatic importance. Four ice regimes are discussed in this section: polynyas, pancake, marginal ice zone, and ice of glacier origin. The latter

Table 2.4 Sea ice types based on stage of ice development. Stages of Development New ice (weakly frozen crystals—have definite form only while they afloat)

Subtype Frazil Grease Slush Shuga

Nilas Young ice

First-Year ice

Old ice

Dark Nilas Light Nilas Gray Gray-White Thin FY Medium FY Thick FY Second-year (SY) Multi-year (MY)

Definition Fine spicules suspended in water. Crystals have coagulated to form a soupy layer. Reflects little light. Snow mixed with water on ice surface after heavy snowfall. Accumulation of spongy white ice lumps formed from grease ice. Very dark in color. Lighter in color. Less elastic than nilas. Breaks on swell. Usually rafts under pressure. Under pressure, it is more likely to ridge than to raft. Sea ice of developed from young ice but has survived for one freezing season only. Ice survived one summer melt season. Ice survived more than one summer melt season. Surface is smoother than SY ice.

Dimension

A few centimeters across < 5m thick > 5m thick 0.1–0.15 m thick 0.15–0.3 m thick 0.3–0.7 m thick 0.7–1.2 m thick > 1.2 m thick

86

SEA ICE

Table 2.5 Sea ice types based on ice form. Form of Ice

Subtype

Pancake

Ice floe

Definition

Dimension

Predominantly circular pieces with raised rims. It may form from nilas, shuga, or slush under a slight swell or from gray ice under severe swell or waves of gray ice.

Ice 0.3 to 3 m in diameter, up to 0.1 m in thickness

Small Medium Big Vast Giant

20–100 m across 100–500 m across 500–2000 m across 2–10 km across >10 km across

Ice Breccias

Ice pieces of different stages of development frozen together.

Brash ice

Accumulation of floating ice made up of fragments; the wreckage of other forms of ice.

Fragments < 2 m across

Ice that forms and remains fast along the coast directly or indirectly through an ice wall or front. It may be formed “in situ” from water or by freezing of floating ice of any age to shore. It may be more than one year old in which case it may be prefixed with the appropriate age category (old, secondyear or multi-year). A narrow fringe of ice attached to the coast, unmoved by tides. It remains after the fast ice has moved away. The terminal of a glacier that extends to, and floats on, the seawater but remains attached to the glacier. Its thickness can be a few hundreds of meters.

Can extend a few meters or several hundred kilometers from the coast

Fast ice

Landfast ice

Ice foot Ice shelf

Anchor ice

Submerged ice attached or anchored to the bottom, irrespective of the nature of its formation.

Table 2.6 Sea ice types based on ice concentration. Concentration Level

Definition

Consolidated ice Compact ice Very close pack/drift Close pack/drift

Floating ice in which the concentration is 10/10 and the floes are frozen together. Floating ice in which the concentration is 10/10 and no water is visible. Floating ice in which the concentration is 9/10 to less than 10/10. Floating ice in which the concentration is 7/10 to 8/10, composed of floes mostly in contact with one another. Floating ice in which the concentration is 4/10 to 6/10, with many leads and polynyas. Floes generally not in contact with one another. Ice in which the concentration is 1/10 to 3/10 and water dominates over ice. No ice present. If ice of any kind is present, this term shall not be used.

Open drift Very open drift Open water (ice-free)

implies area within sea ice that includes numerous icebergs. Before embarking on these discussions, it is worth noting that the term “ice regime” should not be confused with the same term included in the title “Arctic Ice Regime Shipping System.” The latter is a regulatory standard being used by ships in the Arctic [Timco and Johnston, 2003]. This system combines the characteristics of an ice regime based on consistent ice attributes with ice multipliers that represent the capability of a ship to navigate in ice-rich waters.

The result of the combination is an “ice numeral.” This is an index that can be used to decide if the vehicle can proceed through the given ice regime or not. 2.9.1. Polynyas A polynya (common US spelling) or polynia (common U.K. spelling) is a large area of open water with reduced ice cover that persists even when atmospheric temperature

ICE PHYSICS AND PHYSICAL PROCESSES 87 Table 2.7 Sea ice types based on surface feature. Ice Surface Type Level ice Rafted ice

Subtype/ Ice Type Rafted/ YI Finger rafted / YI

Definition Ice unaffected by deformation. Deformed ice formed by one piece of ice overriding another. Floes thrust “fingers” alternately over and under the other, common in nilas. Ridge with sharp peaks and slope of sides usually 40 degrees or more. Fragments are visible from the air at low altitude. Ridge with peaks slightly rounded and slope of sides usually 30 to 40 degrees. Individual fragments are not discernible. Ridge with tops very rounded. Slope of sides usually 20 to 30 degrees. Ridge that has undergone considerable weathering. These ridges are best described as undulations.

Ridged ice∗

New ridge /FYI

Ice piled haphazardly one piece over another in the form of ridges or walls. Usually found in FYI.

Weathered ridge /FYI

Consolidated ridge / FYI

A ridge in which the base has frozen together.

Hummock ice

Small hillock-like surface /MYI

Formed from originally piled ridge on FYI but after surviving one or more melt season(s) the ridge weathers, partially melts then refreezes and has the appearance of smooth hillocks.

Very weathered ridge /FYI Aged ridge/ FYI

Note: *A ridged ice zone is an area of many ridges with similar characteristics.

is significantly lower than the freezing point. Polynya is Russian word which means a “natural ice hole.” It was adapted by polar explorers to describe navigable routes through sea ice. This phenomenon develops in response to oceanic/or the atmosphere forcing. Polynyas exist in polar regions only and usually keep re-forming in the same geographic location every winter and last between a few weeks and a few months after formation. They impart a significant control over the energy balance of ice-covered ocean surface. They are also biologically important because they represent focal points for the intense production of herbivores that trigger the food chain; starting from planktonic and diplontic algae up to Arctic cod, seals, polar bear and eventually humans. Since polynyas are regions of high ice reproduction, they are considered to be a significant source of brine flux to the ocean. Polynyas vary in area from a few thousands of square kilometers (typical of coastal polynyas) to about 80,000 square kilometers of the largest polynya in the Arctic; known as the North Water polynya (NOW). It is usually formed by mid-winter although the formation was observed as early as late November in a few years [Shokr and Agnew, 2013]. A few studies have been published to present details on NOW [Dunbar, 1969, Steffen, 1986, Barber et al., 2001, Ingram et al., 2002, Vincent, 2020]. More information on retrieval of polynya parameters from remote sensing data is presented in section 10.3.1.

Figure 2.61 is a photo, taken on 10 April 1998, of the Canadian icebreaker Pierre Radisson stopping cruising through NOW polynya. Young ice of a variety of thickness is visible. Note how the ice breaker is pushing easily through the thin ice. In a polynya, ice is removed by melting or advection as soon as it is formed. This leaves the water surface exposed to the cold atmosphere, and therefore triggers reformation of new ice that will be removed again and the process continues. That is why polynya is viewed as an “ice factory.” Two mechanisms responsible for polynya formation are commonly known [Smith, Muench, Pease, 1990 and Martin et al., 1992]. The first is thermodynamically driven by warm upwelling water that reaches the sea surface and prevents ice formation or melts the newly formed ice. This mechanism generates what is known as sensible heat polynya. This type of polynya usually occurs away from the coast and therefore is sometimes called open-ocean polynya. The second mechanism is wind driven. Strong wind keeps removing new ice soon after formation, hence exposing the water surface to possible ice re-formation under cold air temperature (in rare occasions ice may be removed by strong ocean current). This mechanism generated what is known as latent heat polynyas. This type of polynya usually exists near coastal areas when strong offshore wind breaks the ice, which usually takes the form of landfast ice, and drifts

88

SEA ICE

Figure 2.61 Photograph of the Canadian icebreaker Pierre Radisson in the North Water Polynya in April 1998. The ice is thin (< 20 cm) but polynyas may include thicker ice (photo by M. Shokr from a helicopter).

it away. Therefore, latent heat polynyas are also called coastal polynya. Most of the polynyas in the Antarctic region are coastal polynyas driven by offshore katabatic wind. This wind carries high-density air from a higher elevation (a common feature of the Antarctica’s coast line) down a slope under the force of gravity. In general, sensible heat polynyas are less persistent in occurrence than latent heat polynyas. It is possible that the two mechanisms become active in the same polynya as in the case of NOW [Mundy and Barber, 2001]. A recurring coastal polynya in the Arctic has been observed along the west coast of the Banks Island, located in the Beaufort Sea. It is shown in the series of the advanced very-high resolution radiometer (AVHRR) thermal infrared images (channel 4) in Figure 2.62. Offshore wind breaks the fast ice along a weak boundary and pushes it away from the coastline. The breakup is

shown in the image of 3 December 1993 and the full opening of the polynya is visible in the image of 13 December 1993. It is possible in the case of coastal polynya that ice is pushed back by onshore wind to close the polynya, triggering a cycle of opening and closing the polynya. It is possible that one or both of the two conditions of polynya formation is (are) fulfilled, namely strong wind or the upwelling warm ocean water, yet the polynya may not form. This becomes possible if the flux of ice from a nearby source continues to feed into what is supposed to be the polynya area. In this case, while the abovementioned conditions are considered to be necessary for polynya formation, they may not be sufficient. To achieve the sufficiency, another condition is needed to block the ice flux into what would be a polynya area. This condition is usually manifested in the form of a natural obstacle, known as ice arch or ice bridge. This is a mechanically strong ice structure developed by nature, which blocks the incoming ice advection because it endures its massive dynamic load. An arch is usually formed between two land constriction points at the two sides of a narrow path of the ice flux. An example is shown in the optical satellite image of the NOW area in Figure 2.63. Clearly, coastal polynyas do not need arches because they are backed by land; hence no ice flux to the polynya area is possible. An arch may remain active for a few weeks or a few months or may not form at all at the designated location in some seasons. When it collapses, the associated polynya terminates. A growing instability in the duration and location of the ice arch north of the NOW was detected in Vincent [2019] using AVHRR imagery during 1979–2019. The detection was performed by visibly examining daily images. The mean duration of the arch was found to be 161±76.7 days but the average duration prior to 2007 was 177 days compared to 128 days since that time. Vincent [2019] attributed this difference to the loss of MYI in the past 25 years. Ice arch is usually composed of MYI or thick ice. More historical context about the ice arches in the NOW area can be found in Kwok et al. [2010], Ryan and Münchow [2017], and Moore and McNeil [2018]. Recently, Ren et al. [2022] used the high-resolution ASAR from the European ENVISAT satellite as well as Sentinel-1 SAR images to identify the spatio-temporal characteristics of the ice arch upstream of NOW polynya over the period 2006-2019. The interannual changes in polynya extent, heat flux, and sea ice production were estimated. A unique opportunity to monitor the ice arch evolution from onset of formation until it stabilizes emerged when daily coverage of the arch at the inlet of the Robeson Channel was available from the SAR onboard the European satellites constellation Sentinel-1A/1B.

ICE PHYSICS AND PHYSICAL PROCESSES 89

Figure 2.62 Series of AVHRR channel 4 images showing the opening of the coastal polynya of Banks Island. The first sign of the opening is observed in the December 3, 1993 image. The polynya continued to expand due to wind action. The largest area is observed in the December 13 image. In the image of December 19, the polynya was fully covered with thin ice, yet it has grown thicker as manifested by its darker appearance in the image (the figure is composed by M. Shokr).

Robeson Channel is located between Greenland and Ellesmere Island, with its northern inlet at 82 N, 60.5 W. Figure 2.64 is a geographic map showing water passages between Lincoln Sea (a southern section of the Arctic Ocean) and Baffin Bay, including Robeson Channel, Kennedy Channel, Kane Basin, and Smith Sound. The two thick lines mark locations of a possible ice arch formation at the entrance of either Robeson Channel (north) or Smith Sound (south). More reference to this figure is found in sections 10.2.1.3 and 10.3.1. The study presented in Shokr, Wang, Liu [2020] monitors the daily development of the ice arch formed at the entrance of Robeson Channel using Sentinel-1A/B. The sequence of development is revealed in the set of images in Figure 2.65. The white dashed line that appears in some panels represents the arch’s contour of the following day. For example, the dashed line in the image for 24 January

represents the arch’s contour that appears in the image for 25 January, and so on. The difference between the visible arch and the dotted line in the image of any given day identifies the ice that was detached by the wind action on that day, leading to the modulation of the arch’s shape. The first sign of the arch appeared on 24 January 2017 and the shape stabilized on 1 February 2017. The study also describes the arch formation in relation to the 10-m surface wind data obtained from atmospheric reanalysis model of the European Center for Medium Range Weather Forecast (ERA5). After the initial rapture of the ice cover that occurred on 23 January 2017 (Figure 2.65), strong southwesterly wind plays a major role in modulating the arch’s contour. It causes rupturing of more pieces of ice at weak locations of fractures. Northerly wind, on the other hand, had virtually no effect on changing the arch’s shape. This process continued

90

SEA ICE 60 °W

50 °W

78 °N

78 °N

80 °N

80 °N

82 °N

82 °N

70 °W

Figure 2.63 AVHRR image of the North Water polynya area in April 1998. Another smaller polynya, the Lancaster Sound polynya, is shown on the left side. The two arrows point to the ice arches that facilitate the formation of the polynya.

throughout the 9 days of the arch’s development by modulating the shape and location of its terminal points along the two constriction land points. It abates when the ice immediately upstream of the arch becomes too consolidated to allow more ruptures in response to wind and the arch becomes strong enough to resist the dynamic load of the incoming ice advection. More on digital methods to calculate the heat flux and ice productivity in the North Water polynya are presented in section 10.3.1. There are 23 polynyas in the Canadian Arctic region [Meltofte, 2012]. An interactive map depicting these polynyas is available through the Canadian Geographic website http://maps.canadiangeographic.ca/canadas-polynyas/. Mapping of Arctic coastal polynya is presented in Tamura and Ohshima [2011]. The number of polynyas in the Antarctic far exceeds the number in the Arctic because the strong and prevailing offshore katabatic wind keeps forming latent heat coastal polynyas. A comprehensive map of the polynyas in the Antarctic region is presented in Nihashi and Ohshima [2015]. Brief sketches showing locations and sizes of major polynyas in both polar

70 °W

Figure 2.64 Map of the Nares Strait between Greenland and Ellesmere Island including the Robeson Channel, Kennedy Channel, Kane Basin, and Smith Sound. The two thick dark lines indicate the locations of possible ice arches. When formed, a polynya is established south of the arch.

regions are presented in Martin et al. [1992], though this information might need to be updated. 2.9.2. Pancake Ice Regime As mentioned in section 2.4.1, turbulence will not allow the continuity of lateral sea ice growth and, instead, causes the frazil crystals to undergo cyclic compression following the wave action. When compressed, the herded crystals may bond to form pancake ice. When the seawater between pancakes freezes, a pancake ice regime appears. A pancake ice piece has typical diameter between 0.3 and 3 m and can grow up to 0.1 m in thickness. The pieces bond together during the early growth stage. The pancake discs may start as slush before solidification to form identifiable discs. Their mobility and collision give rise to their raised edges (Figure 2.66). This is an expression of surface roughness. The pancake surface becomes

ICE PHYSICS AND PHYSICAL PROCESSES 91

Figure 2.65 Daily Sentinel-1 images showing the development of the arch formation at the inlet of Robeson Channel, Canadian High Arctic, from 22 January 2017 until it matured on 2 February 2017. The dotted line marks the arch’s shape and location in the day following the shown day of the image.

Figure 2.66 Formation of pancake in frazil/slush ice medium. Note the raised rims of the pancakes.

even rougher when the pieces continue to thicken and overlap on top of each other as shown in Figure 2.67. The figure depicts a snow-free surface of FY thin ice (0.38 m thick) in the Labrador Sea in March 1994. Rafting of pancake ice pieces with their edges protruding upwards increases the percentage of open water available for additional frazil ice production in the water. With more frazil ice present, further pancake production occurs, followed by rafting and the cycle continues. This positive feedback cycle continues to expand the areal coverage of the pancake ice regime [Wadhams, Lange, Ackley, 1987, Lange and Eicken, 1991]. The process slows down as the ice becomes thicker and can no longer overlap (i.e., when ice thickness exceeds 0.2 m, depending on wind velocity). It should be noted that pancakes are too thin to affect marine operations. They do not have any tangible impact on climate. Pancake cycle is of interest to oceanographers because of the salt flux into the ocean, which originates

92

SEA ICE

10 cm

Figure 2.67 Rafted pancake ice surface in the Labrador Sea in March 1994. Ice was 0.38 m thick. The finger of the person in the picture points to 78 cm on the measuring tape (photo by M. Shokr).

from the release of the brine during the recurrent growth of the pancake ice. Pancake is also important within the context of interpretation of satellite radar imagery as it has remarkably high backscatter in radar image data, hence can be distinguished, yet not uniquely because MYI has comparable high backscatter (section 9.3.2). In a pancake ice field, when the initial pancakes are frozen together then ice can grow vertically to any depth. Therefore, while pancake is only a surface feature, it is used to label the entire volume of ice. Pancake ice regime means the ice surface was formed from extensive pancake cover. Since polar ice is usually covered by snow, then the ragged appearance of the pancake surface is not usually visible. It is usually identified in data from imaging radar sensors, even if it is snow-covered, because radar backscatter is sensitive to surface roughness and not affected by presence of dry snow (section 7.6.1). Formation of pancake ice field is triggered by a minimum level of surface wave action. Since intensive wave actions are found in the marginal ice zones and near ice edge (see next section), pancake ice is more likely to develop there. Pancake ice may also form by breaking nilas or even gray ice sheet under agitated conditions of ocean gravity waves or swell. Both amplitude and wavelength of the ocean waves contribute to the stresses exerted on a thin ice sheet (e.g., nilas). Long wavelengths do not break (or even bend) the ice as quickly as short wavelengths. The exerted stress is mainly a function of the rheological behavior of ice in addition to the wave characteristics, including its traveling velocity. Waves with medium amplitude (≈1 m) or higher amplitude will most likely break a sheet of gray ice (thickness between 0.1 and 0.15 m) into pancake pieces. The breakup is enhanced by waves composed of multiple frequencies

superimposed on each other. At greater distances from the ice edge into the pack ice, ocean waves subside and pancakes start to freeze together along with the frazil slush between them to form a coherent ice floe or sheet. If the pancakes are formed in open water away from the shore, they will be more or less round in shape. If they form near a shoreline, they could possibly take an elongated shape with the long axis parallel to the shore. Sometimes pancakes form at some depth at an interface between water bodies of different physical characteristics before floating to the surface. Pancake ice regimes may be linked to geographic locations where conditions are favorable for their formation. They are found in many areas of the Arctic, though not identified with certain locations, where ice is formed under turbulent atmospheric or oceanic conditions. A cyclical pancake ice regime in the east of the Greenland Sea, known as Odden Ice Tongue (Odden means headland in the Norwegian language), is frequently observed during winter months, though has become less observed since 2000. It can rapidly expand and shrink within a few days. Turbulent water prevails in this region, caused by diversion of cold polar seawater eastward away from the dominant East-Greenland Current in the vicinity of 72–74 N. Wadhams et al., [1996], estimated the salt flux into the ocean in the Odden area. Comiso et al. [2001] studied the interannual variability of the Odden ice from 1979 through 1998 in relation to the North Atlantic Oscillation. Rogers and Hung [2008] evaluated the extent of the Odden in the context of sea level pressure, surface wind, air temperature, and cloud. Pancake ice regimes are more visible in the Antarctic where severe wind and ocean waves usually prevail. One of the well-known pancake ice regimes in the Antarctic exists in the Weddle Sea. Doble, Coon, Wadhams [2003] provided a first quantitative estimate on pancake ice growth rate in the Weddle Sea. They presented photographs of six different pieces of pancake classified according to a scheme suggested in their study. Alberello et al. [2020] present in-situ measurements of pancake ice drift from pair of buoys deployed on ice floes in the Antarctic marginal ice zone where four polar cyclones affected the region. Nose et al. [2021] rationalized the thicker coagulated pancake ice in the Antarctic. 2.9.3. Marginal Ice Zone and Ice Edge Marginal ice zone is a climate-related concept. Ice edge, on the other hand, is a marine-operational-related concept. Ice edge is a conventional boundary that usually (but not always) falls within the MIZ. It is marked by a threshold on ice concentration of 10% or 15%. It should be noted that the same 15% ice concentration threshold is used to define the ice extent (section 11.3). Therefore,

ICE PHYSICS AND PHYSICAL PROCESSES 93

ice edge may coincide with the trace of the ice extent. The first part of this section addresses the definition, significance, and modeling of the MIZ. The second part addresses briefly the ice edge. 2.9.3.1. Marginal Ice Zone The marginal ice zone is a buffer between open water and the pack ice. The pack ice refers to any area of floating ice with ice concentration > 70% regardless of the ice type or form. Hence, MIZ is loosely defined as the ice regime adjacent to the open water where ice concentration is less than 70%. While this is not a rigorous definition, the following definition given by Wadhams et al. [1986] has been commonly accepted. It states that MIZ is the “part of the ice cover which is close enough to the open ocean boundary to be affected by its presence.” In other words, MIZ is defined as the area where open ocean processes, particularly ocean waves, alter the physical and dynamical properties of the sea ice cover. Ocean waves are the primary source of energy that break up the ice, determine the size of the broken floes, their concentration and mobility. The categories of ocean wave in terms of their origin, wavelength, and amplitude are presented in section 7.9.2. The two relevant categories to the present context are gravity waves (a few meters of wavelength) and swell (hundreds of meters of wavelength), which propagates from the large ocean basins. As these waves encroach on the ice cover, they leave significant impacts that determine the main features of the MIZ. According to the above definition, MIZ always exists near open ocean where wave action and interaction with ice cover is strong. It is characterized by a strong horizontal gradient of sea ice cover and vertical gradients of air– ice–sea interactions, though with seasonal dependence. Geographic regions that feature MIZ in the Arctic winter include Greenland Sea, Bering Sea, Labrador Sea, and Barents Sea. In the Southern Ocean it forms a continuous annulus belt around the Antarctica. [Wadhams, 2000]. The MIZ plays an increasingly important role in governing the evolution of sea ice and the upper layer of the ocean. Its coupled interaction with the ocean and atmosphere in terms of energy fluxes is complex. Moreover, MIZ is crucial as an ecosystem for a number of marine species from blooms of phytoplankton to seal, birds, and marine mammal. Sea ice is a platform for resting, hunting, and breeding of marine mammals [Hamilton et al. 2015]. The most visible impacts of ocean wave on ice cover include bending and potentially breaking the ice floes (or sheets) into smaller floes that can eventually be reduced to small ice fragments or a slurry. Swell can break ice floes into many smaller floes of different sizes within a few hours. Smaller floes are expected to be closest to the open water side while larger floes are usually concentrated

near the pack ice. The floe size distribution in the MIZ impacts ice strength and roughness as well as the exchanges of heat, energy, and moisture between ocean and air through ice. The penetration of the ocean waves into ice also enhances the ice floe mobility and therefore alters their shape, size, and deformation. Consequently, it reduces the resistance of the ice to the wind and ocean current stresses. In turn, this allows more wave penetration followed by more disintegration of the ice, that is, a positive feedback cycle. By breaking up the ice cover into small and scattered floes, ocean waves can accelerate lateral melting of the ice when temperature rises above the melting point. New ice is formed between ice floes when atmospheric temperature decreases. The ice cover acts as a dampener of ocean wave. These mutual effects act simultaneously to modify the spectrum of the wave traveling through the ice and determine the main properties of the MIZ. A number of field studies of the MIZ have been conducted since early 1980s. The scientific questions that were addressed in these studies involved the following issues: (1) Statistics of ice floes: size, ablation rates, melt pond, and albedo, (2) the decay of ocean waves in the sea ice cover (wave amplitude and spectrum), (3) the role of ice edge eddies in ice mass transport and melting, and (4) the verification of ice rheology models and ice edge kinematics against in-situ measurements. Two early experiments are covered in Chapter 6: MIZEX and LIMEX. In 2012, the US Office of Naval Research (ONR) had launched a 5-year program called MIZ DRI (Departmental Research Initiative). It aims at addressing a few scientific questions about the processes that govern the spatial and temporal evolution of the MIZ, the current modeling capabilities that predict the MIZ, and the roles of ocean wave, solar radiation, and heat release in governing the MIZ evolution. Summary of the program is presented in Lee et al. [2012]. The interest in studying the MIZ has gained momentum under two thrusts. The first is the incomplete understanding of the processes that govern the ocean-ice-air interactions in the MIZ. Despite the information that has been acquired about ice characteristics in the MIZ, a few questions still need to be answered. Examples are the surface energy balance, propagation of ocean wave into the ice, and the mixing of water density and salinity under the ice cover. Descriptions of key processes in ice and oceanic processes related to the MIZ are presented in Lee et al. [2012]. The second thrust is related to the recent decline of sea ice in the Arctic. As the Arctic ice has been shrinking at a fast rate during the past decade, it is possible that the extent of MIZ is declining toward locations closer to the center of the Arctic Basin above the deep ocean. Rolph, Feltham, Schröder [2020] presented the first panArctic satellite observation study to delineate the extent

94

SEA ICE

of the MIZ over the last 40 years. Interestingly, the study found no trend in the MIZ extent. This is explained based on the northward retreat of the MIZ, which is compensated by its recess southward. It is true that MIZ is recently present in areas in the Arctic that have always been covered with consolidated ice. Examples include areas in the Beaufort Sea and the Canada basin north of Alaska. New locations and ice conditions of MIZ are predicted at a variety of time and spatial scales as the trend of Arctic ice decline continues. These changes have affected the ocean-ice-atmosphere interactions significantly and caused profound impacts on the sea ice evolution. This new situation warrants further field studies of MIZ. The extent of the MIZ is determined by the distance to which waves and swell penetrates the ice cover. Swell can penetrate up to tens or even a few hundreds of kilometers into the ice cover and move at a speed of more than 50 km per day [Perrie and Hu, 1997]. Asplin et al. [2012] reported that 200–300 m wavelength swell penetrated the ice pack in the eastern Beaufort Sea in September 2009 to as much as 250 km. Statistics of ice floe size, geometry, and the spacing between floes are important parameters in modeling the dynamics, wave characteristics, energy balance, and flux exchanges in the MIZ. Lu et al. [2008] suggested that the distribution of floe sizes in the MIZ follows a zonebased pattern with small floes in the outer area, medium size floes in the interior area, and large floes near the solid ice pack. The authors also suggested a distribution of ice floe size and geometry given by the following equation: N 8 ‰ and ice temperature < −16 C. (section 3.5.3) Specific heat of snow Ice surface energy (liquid–vapor) Ice surface energy (solid–liquid) Thermal conductivity of water Thermal conductivity of air Thermal conductivity of freshwater ice at −5 C Thermal conductivity of seawater at 25 C Thermal conductivity of FYI at 6‰ and −20 C Thermal conductivity of MYI Thermal conductivity of snow for density around 917 kg m−3 at −20 C (∗) Latent heat of fusion of freshwater ice Latent heat of fusion of sea ice PAR attenuation coefficient in sea ice PAR attenuation coefficient (dry or wet snow) Thermal diffusivity of air Thermal diffusivity of water Thermal diffusivity of freshwater ice Freezing temperature of seawater at 32‰

sub-freezing temperatures (section 3.6.3). Furthermore, thermal conductivity of snow varies between 0.11–0.35 W/m.K, depending on the snow density and wetness. A few equations (mostly empirical) that describe variations of ice and snow parameters with temperature, salinity, and thickness are included in Table 3.2. More equations are included in the subsequent subsections. Physical properties of sea ice depend on ice types. Unless otherwise mentioned, the data and the equations in this chapter apply mostly to mature first-year ice (FYI). This is the case of the typical data given in Table 3.1 and the equations in Table 3.2. However, accurate representation of sea ice in weather and climate models requires distinction of ice types with their typical values or range of variation. This distinction is possible between major sea ice categories, notably perennial versus seasonal ice types. Perennial ice is characterized by its low salinity, density, surface temperature, thermal conductivity, and dielectric constant. The properties of this ice are almost independent of weather conditions though the snow cover is affected by these conditions. Seasonal ice, on the other hand, encompasses several types and

Value

Source 3

917.8 kg/m Different sources 350 ± 40 kg/m3 1025 kg/m3 1.00 kJ/K.kg 4.20 kJ/Kg 3.85 kJ/K.kg 2.03 kJ/K.kg 2.3 kJ/K.kg

Different sources Different sources Different sources Different sources Different sources Different sources Different sources Different sources Ono [1967]

2.09 J/K.kg 76 m.J/m2 33 m.J/m2 0.59 W/m.K 0.023 W/m.K 2.25 W/m.K 0.609 W/m.K 2.10 W/m.K 1.88 W/m.K 0.25 W/m.K

Different sources Boinovich and Emelyanenko [2015] Boinovich and Emelyanenko [2015] Eide and Martin [1975] Eide and Martin [1975] Fukusako [1990] Different sources Schwerdtfeger [1963] Pringle, Trodahl, Haskell [2006] Mellor [1977]

330 kJ/kg 293 kJ/kg 1.2 m−1 14.0 / 7.5 m−1 1.6x10−5 m2/s 1.4x10−7 m2/s 1.1x10−6 m2/s −1.8 C

Cheng [2002] Schwerdtfeger [1963] Grenfell and Maykut [1977] Grenfell and Maykut [1977] Different sources Different sources Eide and Martin [1975] Different sources

subtypes (e.g., new, young, and first-year) and it is difficult to assign a set of typical physical properties to each subtype. Properties of seasonal ice are not a sole function of ice thickness (age/type) but depend also on the history of ice formation, deformation, and more importantly on the snow cover and the current atmospheric temperature. Approximate values of key properties of seasonal and perennial ice types, compiled from different sources, are summarized in Table 3.3. These values can be used as coarse estimates of the given properties, bearing in mind the wide range of each parameter under different meteorological, oceanic and climatic conditions.

3.2. TEMPERATURE PROFILES IN ICE AND SNOW As mentioned before, temperature profile of sea ice cover is the driving parameter that influences the volume fractions of sea ice constituents; namely pure ice, brine, and air inclusions. Temperature also determines the brine salinity and geometric characteristics of brine pockets.

SEA ICE PROPERTIES: DATA AND DERIVATIONS 109 Table 3.2 Equations to determine a few sea ice parameters. Parameter

Equation 3

density (kg/m ) Bulk salinity (‰) versus ice thickness (m) Bulk salinity (‰) versus ice thickness (m) Porosity (%) Specific heat (kJ/kg.K) Thermal conductivity (W/m.K) Latent heat of fusion (kJ/kg) Effective heat capacity (J/K.kg) Enthalpy of sea ice (kJ/kg) Melting temperature of sea ice Energy needed to melt a unit volume of sea ice

Reference

ρsi = 917.8 − 0.14Tsi Ssi = 14.24 − 19.39hi hi ≤ 0.4 m Ssi = 7.88 − 1.59hi hi > 0.4m Ssi = 4.606 + 91.603/hi p = Ssi(0.05322 − 4.919/T) Csi = 2.11 + 17.2(Ssi/T2) ρ Ssi ρsi − ρpi ksi = si 2 11 − 0 011T + 0 09 − ρpi T si 1000 Lsi = Lpi − 2.11T − 0.114Ssi + 18.1(Ssi/Tsi) hcsi = 2113 + 17 2 × 10 − 3 Ssi T 2si T 2 esi = − 332 4 + 2 12T si + 0 008 T 2si Tm = − μSsi, where μ = 0.054 q = ρsi Lsi 1+

0 054Ssi T 2si

− ρsi Cpi 0 054Ssi + T 2si

Several sources Cox and Weeks [1974] Kovacs [1996] Frankenstein and Garner [1967] Ono [1967] Pringle et al. [2007] Weeks [2010] Untersteiner [1961] Notz and Worster [2006] Bitz and Lipscomb [1999] Bitz and Lipscomb [1999]

Note: For explanation of the subscripts, see column 1 of Table 3.1. hi is the sea ice thickness, Tsi is the sea ice temperature in C, Ssi is the bulk salinity of sea ice in ‰.

Table 3.3 Typical physical and electrical properties of four major ice types. Thickness (cm) Bulk salinity (‰) Density (kg/m3) Dielectric constant (10 GHz) Thermal conductivity (W/m.K) Brine volume fraction

New ice

Young Ice

First-Year Ice

Multi-Year Ice

< 10 14 920 5.65–j 2.25 2.14 0.20

10–30 9 900 4.0 – j 0.81 2.14 0.08

> 30 4 900 3.32 – j 0.23 2.09 0.05

> 200 0.5 800 2.77 – j 0.03 1.88 0.0

Since brine should remain at thermal equilibrium, any temperature change will adjust the salt composition inside brine pockets according to the phase diagram shown in Figure 2.3. As mentioned earlier, significant variations in the ice temperature profile cause pronounced variations in salinity, brine volume fraction, thermal conductivity, and the complex dielectric constant. Temperature profiles of Arctic sea ice have been measured extensively during numerous field studies [e.g., Untersteiner, 1964; Bilello, 1980; and Shokr and Barber, 1994]. All of these measurements were performed on an opportunity basis during field campaigns conducted on fully grown ice during short-duration field campaigns, mostly the early spring season. The wisdom of the Inuit people of the Arctic was not utilized in any of these field campaigns. Historically, the first systematic measurements on the evolution of snow and ice thickness, temperature, and salinity profiles throughout the entire ice growth seasons of 1977–78 and 1978–79 were carried out in Eclipse Sound as part of a 10-year program. An Inuit team of Pond Inlet, Baffin Island provided their assistance from the start of the freezing to the beginning

of melt the season for the entire freezing season [Sinha and Nakawo, 1981; Nakawo and Sinha, 1981]. Perovich and Elder [2001] measured the temperature profile of sea ice over an annual cycle from October 1997 through October 1998 at several sites selected to sample different ice types ranging from FYI to MYI and ice ridges. Long-term regular measurements of sea ice growth along with the record of weather data were conducted at five Canadian High Arctic weather stations: Alert, Eureka, Isachsen, Mould Bay, and Resolute Bay (see locations in Table 1.1). The long-term measurement programs undertaken during the 10-year long program (1976–1986) in Eclipse Sound was supported by Energy Mines and Resources, Canada (renamed later as Natural Resources Canada—NRCan) through the Polar Continental Shelf Project (PCSP) in association with the National Research Council of Canada (NRC) (see also section 6.2). Under steady conditions in winter when the atmosphere is considerably colder than the ocean, the temperature profile usually features a linear increase with depth from the upper cold surface to the lower ice–water interface

110

SEA ICE

where the temperature is typically −1.8 C. The linear temperature gradients Gi and Gs in ice and snow, respectively, can be written as: Tw − Ts hi

Gs =

Ts i − Ta hs

i

(3.2)

where, Ta, Ts/i and Tw are the air, snow–ice interface and seawater freezing temperatures, respectively; hi is the ice thickness and hs the snow depth. If the ice thickness increases by Δhi in a time period Δt, then the quantity of heat released during freezing is LsiρsiΔhi, where Lsi and ρsi are the latent heat of fusion and density of sea ice, respectively. This amount of heat must flow through the ice and snow to the atmosphere, assuming that the ice–water interface is at thermal equilibrium. Thus, Li ρi Δhi = ksi

Tw − Ts hi

i

Δt = k s

Ts i − Ta Δt (3.3) hs

where, ksi and ks are the average thermal conductivities of sea ice and snow, respectively. After rearrangement, the second equality in equation (3.3) gives, Ts i =

k si hs T w + k s hi T a k si hs + ks hi

(3.4)

Substituting Ts/i from equation (3.4) in equations (3.1) and (3.2), Gi =

T w k si hs T w + k s hi T a − hi hi + k si ks hs

(3.5)

Gs =

ksi hs T w + k s hi T a T a − hs k si hs + ks hi hs

(3.6)

These two expressions are equal. From this equality the ice thickness hi can be calculated, resulting in same expression as equation (2.7). The two temperature profiles in sea ice obtained on 16 February and 6 April, 1978 from fully grown FYI in Eclipse Sound [Sinha and Nakawo, 1981] are shown in Figure 3.1. The ice grew from 125 cm on 16 February to 170 cm on 6 April. Note the almost perfect linear trend shown in 6 April data and the deviation from linearity near the top of the ice surface in 16 February data because of the warm week prior to that date. The figure shows also comparison of measured data against temperature calculations using the model described above. In the model, the authors used Tw= −1.8 C, ki = 5 × 10−3 cal/cm.s.C [Schwerdtfeger, 1963], ksi = 6 × 10−4 cal/cm.s.C [Mellor, 1977], and Ta was selected to be equal to the mean air temperature. Although the results from the above model are verified for the date of 6 April, the deviation of the profile of 16 February from the model (following a week of

Mean air temperature

Water Top snow temperature surface

0

(3.1)

Snow-ice interface

Ice thickness (cm)

Gi =

50

50

Theoretical (Feb. 16)

Theoretical (April 6)

100 125 cm 150 175 cm 200

–30

–20

–10

0

Temperature (°C)

Figure 3.1 Comparison between measured and theoretical temperature distributions in an ice sheet for the two days shown in 1978. Theoretical calculations are based on mean air temperare of the two days [Sinha and Nakawo, 1981 / Canadian Science Publishing] shown.

relatively warm air temperature) indicates that the model underestimated the temperature and produced higher temperature gradient. The authors suggested that this simple model should be improved by taking into consideration three factors: (1) the thermal history in previous days, (2) the effects of wind and cloud cover on the heat transfer conditions at the exposed snow surface, and (3) the uncertainty of the thermal conductivities of snow when it acquires salinity from the ice surface. Often in ice field experiments people measure air temperature, ice thickness, and snow depth. A simple practical rule to determine roughly the temperature profiles in ice and snow as well as the snow–ice interface temperature using these three parameters was developed by the author (N. Sinha). This utterly simple method was verified and used by the Canadian Ice Service (CIS) operators, onboard the CCG icebreakers, in the early 1980s for Arctic sea ice and coined the name “Sinha Rule.” It is based on the assumption that ice is 10 times more thermally conductive than snow (in fact ki = 8.33 ks) and freezing temperature of freshwater is 0 C (−1.8 C for seawater). Since snow acts as a blanket, it is expected that the temperatures within the ice and at the snow–ice interface will be significantly higher than the air temperature during cold winter air temperature. By equating the two expressions for Gi from equations (3.1) and (3.5), the resulting equation takes the form: T si = T a hi 10hs + hi

(3.7)

SEA ICE PROPERTIES: DATA AND DERIVATIONS 111 (a)

(b) Sinha’s rule Ta

–10°C

Air temperature –10°C

Ts/i

(c) –10°C

snow

Ts/i

ice

0°C

0°C

0°C

Figure 3.2 Illustration of N.K. Sinha’s rule (coined by CIS operators) to determine roughly the snow–ice interface temperature and temperature profiles in snow and ice for freshwater ice covers. (a) snow-covered ice, (b) linear temperature distribution between the ice-water interface and the air temperature, (c) temperature profiles in snow and ice. Bottom temperature should be adjusted to −1.8 C for sea ice (courtesy of M. Johnston, NRC of Canada).

This equation can be put into practice as described below. Using a stick, one can scratch the snow surface and draw the composite of ice and snow based on the measured snow and ice depths (Figure 3.2a). A second picture can be drawn, as shown in Figure 3.2b, where thickness of the ice is kept the same but the snow depth is multiplied 10 times. This would represent the “equivalent” ice cover (i.e., snow is thermally converted to its equivalent ice thickness). A line perpendicular to the surface is then drawn to represent the reference of the water temperature at the ice–water interface. A point is marked on the top surface representing the measured air temperature Ta. Then, a straight line is connected between this point and the reference temperature at the bottom of the ice. The slope of the line gives the temperature gradient of the “equivalent” ice cover, and the point of intersection at the snow–ice surface gives the snow/ice surface temperature Ts/i. The last step (Figure 3.2c) is to connect Ts/i to Ta while using the actual snow depth (i.e., shrink back to the actual snow depth). That gives the real temperature gradient in the snow pack. Suppose, as an example, that the ice thickness is 1 m with a snow cover depth of 0.1 m. If the air temperature is −10 C and the water temperature is 0 C, then the snow–ice interface temperature will be −5 C. This shows that even thin snow can have a profound influence on the ice surface temperature and ice could be significantly warmer than the snow and air temperature. In an unpublished study conducted by M. Shokr on FYI in Mould Bay, Canadian western Arctic, temperature profiles in the ice and the overlaid snow were sampled

every 30 minutes during most of the ice growth season from 5 December 1996 until the end of May 1997. A temperature probe of 140 cm in length was inserted into the snow and ice with the top part remaining in the air. The probe had 24 thermistors, the top 16 were spaced at 4 cm and the bottom 8 at 10 cm. Only the bottom 8 thermistors were under the ice surface while the rest were covered by snow or remained exposed to the air. The temperature measurements were transmitted to the Argos satellite system and downloaded to a laboratory in Environment Canada in Toronto. Argos is a satellite data collection system dedicated to studying and protecting the environment. It receives data from any ground sensor, identifies the geographic location of the sensor and transmits the measurements to a given a specific destination so that the data can be analyzed online. Information about this system is available through the link http://www.argos-system.org/. Temperature profiles from the previously-referred study are shown in Figure 3.3 for selected days (almost weekly interval) from December 1996 to February 1997. The measured profiles end at 80 cm ice depth. Ice thickness can be estimated from the profiles by assuming that the ice–water interface is at −1.8 C and the profile continues its linear shape to the bottom of the ice. Ice thickness was also measured by the operators of the Mould Bay weather station of Environment Canada (the station was terminated in late 1997). It was 70 cm on 12 December 1996 under relatively warm temperature and grew rapidly to 100 cm on 19 December as the air temperature dropped by about −20 C between the two

112

SEA ICE 80 60 40

Depth (cm)

20 Ice/snow interface

00 –20 –40

–80

4 –4 2 –4 0 –3 8 –3 6 –3 4 –3 2 –3 0 –2 8 –2 6 –2 4 –2 2 –2 0 –1 8 –1 6 –1 4 –1 2 –1 0 –0 8 –0 6 –0 4 –0 2

–4

–4

6

–100

00

12 Dec. 96 19 Dec. 96 25 Dec. 97 2 Jan. 97 9 Jan. 97

–60

Temperature (°C) 80 60 40

Depth (cm)

20 00

Ice/snow interface

–20 –40 –60 –80

–4 6 –4 4 –4 2 –4 0 –3 8 –3 6 –3 4 –3 2 –3 0 –2 8 –2 6 –2 4 –2 2 –2 0 –1 8 –1 6 –1 4 –1 2 –1 0 –0 8 –0 6 –0 4 –0 2 00

–100

19 Jan. 97 26 Jan. 97 4 Feb. 97 11 Feb. 97 19 Feb. 97

Temperature (°C)

Figure 3.3 Temperature profiles in snow-covered FYI in Mould Bay, Northwest Territories, Canada. Ice thickness grew from 0.7 m on 12 December 1996 to 1.31 m on 26 February 1997. Snow depth was virtually constant around 50 cm. Note the linear temperature profile in sea ice and the non-linearity within the snow (data obtained and plotted by M. Shokr).

dates. It then reached 131 cm on 26 February 1997. Snow depth increased slightly from 40 cm on 12 December to about 48 cm on 19 December and remained virtually constant during the entire period. Figure 3.3 shows nearly isothermal ice temperature profile on 12 December within both the snow and ice, following a period of subzero air temperature. The figure also shows that snow temperature is more sensitive to variations of air temperature. Once air temperature remained below −30 C continuously after 9 January 2017, stable ice temperature profiles

were established. The profiles demonstrate a strong linear trend throughout the ice volume. Note the slight non-linearity of the ice temperature near the surface on 9 January when a sharp drop of air temperature (from −34 C to −44 C) occurred within a few days, which probably caused lag in thermal response of the surface. The average temperature gradient within the ice is 0.111 C/cm with a standard deviation ±0.018. Air temperature dropped from −27.52 C to −44.26 C between 25 December 1996 and 9 January 1997. This drop of

SEA ICE PROPERTIES: DATA AND DERIVATIONS 113 80 60 40

Depth (cm)

20 00

Ice/snow interface

−20 −40 −60 −80

16 Dec.96 07:32:47 16 Dec.96 19:22:36 17 Dec.96 07:09:10 17 Dec.96 19:05:44

−100 −34 −32 −30 −28 −26 −24 −22 −20 −18 −16 −14 −12 −10 −08 −06 −04 −02 −00 Temperature (°C)

Figure 3.4 Temperature profiles from the same geographic site as in Figure 3.3 showing the response of the profiles to the atmospheric temperatures cooling (−24.4 C to −32.3 C) over a short period (nearly 36 h) between 16 and 17 December 1996. Note that the top of the snow layers is more affected by variations in air temperature while ice temperature profiles remain unchanging (data obtained and plotted by M. Shokr).

17.6 C triggered a 6 C drop in ice temperature near the surface and 6 C at 80 cm depth. Temperature profiles in the snow feature a fairly linear trend of decrease from the snow base temperature upward, yet with a steeper gradient than that of ice temperature. A non-linear behavior (milder temperature decease) is observed near the top where the snow temperature becomes nearly equal to the air temperature. Figure 3.4 shows how fast the top layer of the snow cover is responsive to changes in atmospheric temperature. The figure shows the temperature profiles in snow and ice when air temperature dropped from −24.4 C to −32.3 C within 36 hours between the morning of 16 December and the evening of 17 December 1996. While the effect of air temperature is more pronounced in the upper 20–30 cm of the snow pack, the rest of the snow and certainly the entire ice depth were not affected during that short period. Variation of air temperature below –22.9 C (i.e., the precipitating temperature of sodium chloride) does not result in changes in ice properties because the ice composition remains virtually unchanged below this temperature. Petrich and Eicken [2009] presented similar profiles of snow and ice temperature from measurements using thermistor probe frozen into landfast FYI near Barrow, Alaska in 2008. They observed the transition from winter to summer conditions in terms of change in the direction of the temperature gradient. Another example of ice temperature data is presented in Perovich, Elder, Richter-Menge [1997]. This is a 15-month record of the atmosphere, ice and ocean temperatures collected using a vertical array of thermistors inserted in a

multi-year ice floe in the Beaufort Sea from late September 1993 to late November 1994. The top and the bottom of the array were exposed to the air and ocean, respectively. A data-logging system was used to collect the data autonomously. The data are presented in Figure 3.5. The annual cycle of air temperature is evident in the top graph. Contours of the ice temperature are plotted in the bottom graph. The bulk of the ice has temperatures between −6 C and −12 C during winter and −2 C during summer. Note the ice accretion at the bottom of the ice from January to May (around 40 cm) and the intensive ablation from the top and the bottom of the ice from June throughout August. About 0.5 m of ice thickness melted.

3.3. BULK SALINITY AND SALINITY PROFILE The initial salinity at the time of sea ice formation is determined by the atmospheric and oceanic conditions under which ice has formed. It varies later depending on factors such as the atmospheric temperature, precipitation, snow cover, and crystallographic structure of the ice. Salinity plays a central role in determining the thermal, mechanical, electrical, and radiometric properties of sea ice. For that reason, understanding its evolution and spatial distribution is crucial. On the average, salinity of YI is considerably higher than the mature FYI, and the latter is much higher than the virtually desalinated MYI. Sea ice salinity is required to initiate ice thermodynamics and climate models over polar regions. As a first approximation,

SEA ICE

Air temperature (°C)

114

0 –10 –20 –30 –40 50 –2.0 / 0 –4.0 / –2.0 –6.0 / –4.0 –8.0 / –6.0 –10. / –8.0 –12. / –10.

Depth (cm)

0 –50

–14. / –12. –16. / –14. –18. / –16. –20. / –15.

–100 –150 –200 –250 Sep 27

Nov 26

Jan 25

Mar 26

May 25

Jul 26

Sep 22

Nov 21

Figure 3.5 (top) Evolution of air temperature (bottom) ice temperature distribution in a multi-year ice floe in the Beaufort Sea from 27 September 1993 to 21 November 1994. The gray-shaded area denotes snow cover. Bottom contour denotes the ice–ocean interface. Each color in the color code represents a step of 2 (adapted from Perovich, Elder, Richter-Menge, 1997; Figure 1 / with permission of John Wiley & Sons).

some models adopt the crude assumption of constant bulk salinity with constant depth profile. Bulk salinity Ssi is defined as the average salinity over the sea ice thickness. Other models assume idealized time-independent salinity profiles [Hunke et al., 2015]. The best input would be time-varying salinity profiles (as ice ages brine continues to drain) that reflect also its spatial variations. As this is virtually impossible to obtain, empirical equations that relate the bulk salinity to ice thickness can be useful. In this section the bulk salinity and salinity profiles within the ice depth are presented. Both change with time as sea ice thickens. Surface salinity is the value found within the top few millimeters of the ice. It affects the reflection of solar radiation, emission of the thermal infrared (TIR) and microwave radiation, and the scattering of radar signal. Under the ice surface, the salinity profile is established. Salinity profile and bulk salinity are affected by the ice growth rate and its crystalline structure. For example, frazil and snow retain more brine (i.e., feature higher salinity) than columnar ice. Obviously, surface melt decreases both surface and bulk salinity. In general, salinity is a manifestation of the influence of air temperature. As mentioned in section 2.5.5, desalination of seasonal sea ice continues throughout its life time, depending on changes of air temperature and ice crystalline structure. Subgrains boundaries in the crystal lattice play an important role in the

desalination process (section 2.5.3). The connectivity of brine pockets and the shape of brine drainage channels also control the desalination process. In summer, surface meltwater percolates through sea ice above the freeboard, flushing away brine especially near the surface [Weeks and Ackley, 1982]. This decreases the salinity near the surface significantly and to a lesser extent throughout the entire ice thickness. Investigations on bulk salinity and salinity profiles have been conducted in numerous field experiments on different ice types; but not on YI (< 15 cm thick) as it is not safe to walk on. Unlike seasonal ice types, salinity of MYI is nearly steady with values that vary between 0 and 2‰. Most of the field studies that turned into classical publications on ice salinity were conducted during the 1970s– 1990s. Most of the data presented in the rest of this section are obtained from those publications. A relatively recent study [Gough et al. 2012] presents measurements of sea ice salinity conducted in eastern McMurdo Sound in the Antarctic during the 5 months of sea ice formation at regular intervals of 2 weeks. The common field technique used to generate these data involve coring the ice and cutting the core into sections of about 2 cm thick at known depth, then melt the specimen and measure the salinity with a refractometer. This is a rather destructive technique because the coring, extraction and the cutting processes

SEA ICE PROPERTIES: DATA AND DERIVATIONS 115

may be associated with brine drainage. Non-destructive methods were developed later by measuring the complex dielectric permittivity using a hydraprobe sensor [Backstrom and Eicken, 2006] or borehole resistance [Ingham, Pringle, Eicken, 2008].

regressed their data also into a single exponential equation as many natural phenomena tend to follow the exponential trend. Using salinity measurements from the Arctic and Antarctic sea ice, Kovacs [1996] developed another set of single monotonic equation relating bulk salinity to ice thickness:

3.3.1. Bulk Salinity

S si = 14 24 − 19 39 hi S si = 7 88 − 1 59 hi

for hi ≤ 0 4 m for

hi > 0 4 m

(3.8) (3.9)

where, Ssi is in ‰. Equation (3.9) is a fit of FYI and MYI data. The study found that the correlation coefficients for the above two regression equations are −0.78 and −0.94, respectively. It also suggested that the pronounced break in the slope at 0.4 m is due to a change in the dominant brine drainage mechanism from brine expulsion to gravity drainage (see section 2.5.5). Cox and Weeks [1974]

S si = 4 606 + 91 603 hi

for FY ice

(3.10)

S si = 1 80 + 99810 5 h2i

for MY ice

(3.11)

Applications of these equations are limited to the polar regions where cold winter conditions are nearly steady. They cannot be used for sea ice growing in temperate (subpolar) regions where freezing-thaw cycles are usually encountered in winter. The equation for the thickness of undeformed FYI [equation (3.10)] along with validation data from Arctic and Antarctic measurements are plotted in Figure 3.6. The correlation coefficient is 0.73 and the standard deviation of the measurements from the empirical relation is 1.47. Kovacs [1996] pointed out that the variations in bulk salinity data from leveled FYI grown in the Arctic and Antarctic are similar. This is in spite of the difference in seawater surface salinity (within the top 50 m) between the Southern Ocean (about 34‰) and the Arctic Ocean (about 32‰) due to lower evaporation. The author stated that “refinements and standardization of sampling techniques may one day allow us to see the effect of seawater salinity variation as well as first-year sea ice morphology and structure differences on the bulk salinity of the ice” (p.10). So far, no progress has been achieved on this point.

30 25

Bulk salinity (‰)

Bulk salinity affects the gross thermal and mechanical properties of sea ice. It is a function of ice thickness because of the continuous brine drainage throughout the lifetime of seasonal sea ice. It is remarkably high at the initial stage of sea ice formation, then decreases at a relatively fast rate during the early ice growth before the rate slows down and the salinity stabilizes. The main factors that influence the bulk salinity of sea ice Ssi are: (1) the initial growth conditions of the ice sheet (faster growth retains more salts in the ice), (2) surface melt, which accelerates brine drainage, (3) air temperature record during the lifetime of the ice sheet (higher temperatures result in more brine drainage while very low temperatures cause precipitation of salts within brine pockets, both decrease Ssi, and (4) the ice thickness. As ice thickens, the ice growth slows down and therefore less brine is retained at the growth interface (i.e., bottom of the ice). Processes that determine the bulk salinity of sea ice are discussed in Kovacs [1996] from measurements of FYI salinity data in the Arctic and the Antarctic. The study concludes that thin ice (< 5 cm thick) has high bulk salinity around 25‰ but the salinity decreases rapidly to about 6‰ as the thickness reaches 50 cm. The decrease continues though at a reduced rate and reaches about 4‰ when the ice becomes 150 cm thick. In the beginning of the melt season another tangible decrease in the ice bulk salinity to about 2‰ is observed. If the ice survives the summer to become second-year ice it will start with salinity less than 2‰ at the surface. Widely-used equations that relate bulk sea ice salinity to its thickness were developed by Cox and Weeks [1974], based on data obtained from FYI and MYI in the Beaufort Sea in winter. Two linear regression equations are applied to two ranges of ice thickness:

20 15 10 5 0

0

50

100

150

200

Ice thickness (cm) Figure 3.6 All Arctic and Antarctic first-year sea ice bulk salinity versus ice thickness data compiled from numerous sources. The solid curve is the empirical equation (3.10) [Kovacs, 1996 / U.S. Department of the Army / Public Domain].

116

SEA ICE

Ryvlin [1974] developed an equation to determine the bulk salinity at any growth stage in terms of the final bulk salinity Sbf at the end of the growth season. This is useful for retrieval of bulk salinity at any stage of ice development retroactively once the bulk salinity at the end of the season becomes known. The equation takes into consideration the seawater salinity Sw and the ice growth rate ν, S si = S w 1− S R e − a hi

05

+ SR Sw

(3.12)

where, SR is the salinity ratio (Sbf /Sw) and a is a coefficient that depends on v. For v ≥ 40 mm/day, a = 0.35, and for v ≤ 5 mm/day, a = 0.6. In between these two rates the value of a can be interpolated. This equation can only be applied retroactively because it requires the bulk salinity at the end of the season. Calculations were compared with field measurements in Ryvlin [1974]. Results show good agreement between the empirical estimate and field measurements throughout the entire ice thickness range. However, the number of measurements that were used to validate the equation was rather small. Another expression to calculate the bulk salinity in terms of the ice growth rate v (m/s) and the seawater salinity Sw is presented in Petrich and Eicken [2009]. Sb v = 0 14 S w Sw 1 35 × 10 − 7

0 33

(3.13)

The authors validated this equation using salinity profiles measured from three FYI cores obtained from Barrow, Alaska. The equation reproduced profiles of the measured bulk salinity between 0.65 m and 1.1 m fairly well. 3.3.2. Salinity Profiles Salinity profiles and their evolution during ice growth cannot be determined directly from thermodynamic models or inversely from remote sensing observations. The only source of this information has been field measurements. Measured salinity profiles from different ice types in different regions were obtained from numerous field experiments. During winter, the salinity profile of FYI usually takes a “C” shape, featuring high salinity near the top and the bottom of the ice sheet, with less salinity in the middle. This was observed in the earliest study of salinity profiles of the Arctic FYI during different seasons [Malmgren, 1927]. The high salinity at the bottom is the result of brine entrapped at the ice–water interface while freezing propagates (section 2.5.5.1). On the other hand, the high salinity at the top is a manifestation of the high brine content entrapped at the ice–water interface during the onset of freezing but did not have much opportunity to drain

because ice thickens at a fast rate during the early growth period. Deviation from the C-shape is observed in the spring and summer when the temperature gradient near the ice surface is reversed due to warmer air temperature. Sinha and Nakawo [1981] presented the first systematic salinity profile measurements and analysis on environmental conditions and growth history of Arctic FYI for two subsequent winter seasons of 1977–1978 and 1978–1979. Evolution of vertical salinity profile was recorded on weekly basis. The investigations were carried out on ice in the Eclipse Sound (latitude 72.7 N, longitude 78.0 W) near Pond Inlet, Baffin Island, Canada. The dependence of the evolution of the salinity profile on growth history, for the season of 1977–1978, was discussed in Nakawo and Sinha [1981]. For this season, measurements started shortly after the onset of ice formation in November 1977 and continued until the ice reached the stage of FYI in April 1978. Ice cores were extracted on weekly basis with the help of trained Inuit residents so that the evolution of the salinity profiles could be related to the ice growth conditions. With support from a number of organizations, especially the PCSP, Energy, Mines and Resources Canada (EMRCan) and the Arctic Research Establishment (ARE), the author (N.K. Sinha) was able to provide training to six indigenous persons. These assistants decided the time to go on the ice and take ice cores when they thought the ice cover was safe to walk on. An observation area of 100m × 100m, 0.5km from the nearest shore near Pond Inlet was established. The depth of water underneath the ice was about 150m. Most of the cores were extracted at air temperatures below −20 C and this minimizes the possibility of brine drainage during the removal of the cores from the ice sheet. The methodology used in collecting the data and the results of analysis on the ice growth, salinity profile, microstructural details and ice mechanical strength from a series of sea ice field studies near the Pond Inlet from 1977 to 1982 have been presented in Sinha and Nakawo [1981]; Sinha [1983]; and Nakawo and Sinha [1984]. Figure 3.7 shows the evolution of the salinity profiles at a 2 week interval. The salinity in the top 50 mm and the bottom of the ice sheet are consistently higher than the rest of the profile. C-shaped profiles are noticeable during the initial growth period (4 November to 2 December). After that date, the profiles are rather complex but higher salinity is still found at the surface and near the ice bottom. The average salinity decreases rapidly after initial freezing until it reaches approximately 6‰ once ice thickness exceeds approximately 0.6 m. This observation is illustrated more clearly in Figure 3.8 which shows the temporal variation of salinity for a 25 mm thick section between 0.40 and 0.425 m below the top ice surface. It can be seen that the salinity of 25‰ measured on 2 December dropped to 8.5‰ on 9 December. Thus, the major reduction in the

SEA ICE PROPERTIES: DATA AND DERIVATIONS 117 4 NOV

18 NOV

2 DEC

16 DEC

30 DEC

13 JAN

27 JAN

16 FEB

3 MAR

17 MAR

31 MAR

14 APR

28 APR

0 13.4

20

11.3

14.6 15.6 17.6

40

24.7

Ice thickness (cm)

60 80

Salinity (‰) 2 4 6 8 10

100

16.1

120 140

12.5

160 180

Figure 3.7 Evolution of vertical salinity profiles of FYI in Eclipse Sound, near Pond Inlet, Baffin Island, Canada, measured at intervals of 2 weeks during the winter of 1977–1978. A scale for salinity is shown. Vertical solid lines represent a value of 6‰ and are given as a reference [Sinha and Nakawo, 1981 / Canadian Science Publishing].

25

Salinity (‰)

20

15

10

7.4 ± 0.8 ‰

5 0 1 DEC

1 JAN

1 FEB

1 MAR

1 APR

1 MAY

Figure 3.8 Variation of salinity with time for 25 mm sections of ice at a depth between 0.40 and 0.425 m. Broken lines indicate “stable” salinity or average of all salinities except the initial high value at this depth level [Nakawo and Sinha, 1981 / Cambridge University Press].

brine content occurs within about 10 days after the formation. The figure also indicates that, as the freezing front moves downwards, the salinity at a given depth attains a “quasi-stable” value. Following the initial rapid desalination, the salinity at each depth level becomes relatively constant but has a tendency to decrease slowly during the rest of the season. This stable salinity at the level of 0.40–0.425 m is indicated in Figure 3.8 where the dashed line represents the average salinity of all the data,

excluding the initial high values. The stable salinity at a given depth is a function of depth and growth rate of ice. The values obtained from 25 mm sections for various depths at intervals of 200 mm are shown in Figure. 3.9. The profile of the stable salinity at every 25 mm segment of ice depth is shown as curve a in Figure 3.10. The variation in stable salinity at a given depth during the season was about 1‰. The figure shows also the dependence of the growth rate of ice calculated from equation (2.7) using the measured air temperature (curve b). Curve c represents the running mean of calculated growth rate for an interval ±50 mm for every 25 mm segment of ice. This is a smoother curve than the curve b and more realistic because the thermal inertia of the snow–ice system would dampen the variations in the temperature and hence the growth rate (calculated on the basis of the daily air temperature). The two curves, a and c, in the figure show an average bulk salinity of about 6‰ corresponding to a seasonal average growth rate of about 12 mm per day. A correlation between growth rate and stable salinity is apparent. The double-sided arrows point to depths where this correlation is clearly visible. Figure 3.11 shows the plot of the stable salinity of each 25 mm segment against the corresponding growth rate ν (curve a and curve b in Figure 3.10, respectively) for depths between 0.25 and 1.25 m. The number of segments available for determining each salinity value is shown in the figure. The solid line is the result of the calculations of the effective solute ratio C/C0 (i.e., Kδ) from

118

SEA ICE 1 NOV

1 DEC

1 JAN

1 FEB

1 MAR

17.5–20 cm

1 APR

1 MAY 4.5 ± 0.7 ‰

37.5–40 cm

7.2 ± 0.8 ‰

57.5–60 cm

7.3 ± 0.8 ‰

77.5–80 cm

5.6 ± 1.3 ‰

97.5–100 cm

5.7 ± 0.8 ‰

117.5–120 cm

137.5–140 cm

4.8 ± 0.7 ‰

4.4 ± 0.5 ‰

Figure 3.9 Attainment and retention of “stable” salinity at a given depth in sea ice from a site in Eclipse Sound near Pond Inlet, Baffin Island, Canada, during the winter of 1977–1978, for various depths at intervals of 0.20 m [Nakawo and Sinha, 1981 / Cambridge University Press].

Depth (cm)

0

50

c a

b

100

150 2 4 6 8 10 Stable salinity s (‰)

12

0.5 1.0 1.5 2.0 Growth rate 𝜈 (cm/day)

Figure 3.10 Stable salinity profile (curve a) within a depth interval of 25 mm, where horizontal bars represent standard deviation. Ice growth rate profile (curve b) is the calculated rate and curve c is its running mean for an interval of ±50 mm for every 25 mm segment of ice [Nakawo and Sinha, 1981 / Cambridge University Press].

equation (2.28). This is the solute concentration in the solid divided by the solute concentration of the seawater immediately under the ice–water interface. It can be seen that an ice layer can retain relatively higher brine as the growth rate increases. The same data set was used by Nakawo and Sinha [1981] to construct a composite diagram that features: (1) the stable salinity at each depth, (2) the variation of daily mean

air temperature, and (3) the growth history of the ice. The diagram is shown in Figure 3.12. Warmer periods at the end of both November and December resulted in slower growth rates, whereas colder periods in December and mid-January enhanced the rate of growth. The growth rate is determined from the slope of the ice thickness. The theoretical ice thickness (solid line) was based on equation (2.8) without considering the snow cover. It can be seen that both the observed and the calculated ice thickness (the dashed and the solid line, respectively) show the dependence of growth rate on mean air temperature (once again based on the slope of the curve). Similar to the growth rate, the stable salinity profile also exhibits a record related to the major weather changes. The low salinity of 4‰ at a depth of 10 to 30 cm seems to be related to the slow growth rate corresponding to the warm period in November. The higher salinity of 7.5‰ recorded at a depth of 40–70 cm correlates well with the higher growth rate initiated by the cold period in December. Further correspondence can be seen as the relatively lower salinities at a depth from 70 to 80 cm correlate well with the warmer periods at the end of December. The same observation applies to the low salinity between 110 and 120 cm depth (i.e., the salinity profile of 17 March or 14 April in Figure 3.7), whereas the comparatively higher salinity developed at 80–100 cm relates well with the colder period in the first three weeks of January. A non-destructive method for measuring the salinity and solid ice fraction in sea ice was developed by Notz, Wettlaufer, Worster [2005] using an instrument that measures electrical impedance. It was used in field and

SEA ICE PROPERTIES: DATA AND DERIVATIONS 119

0.8

9

1.0

Growth rate ν × 10– 5 (cm/s) 1.2 1.4 1.6

1.8

2.0

0.25

7 0.20 6

C ⁄ C0

Stable salinity s (‰)

8

5

0.15 20 < No. OF. DATA

4

15 < No. OF. DATA ≤ 20 10 < No. OF. DATA ≤ 15

3

No. OF. DATA ≤ 10

2 0.6

0.8

1.0 1.2 1.4 Growth rate ν × 10– 5 (cm/day)

1.6

0.10

1.8

Figure 3.11 Plots of stable salinity of each 25 mm segment of curve a versus corresponding growth rate of curve b in Figure 3.10 for columnar-grained ice at depths between 0.25 and 1.25 m. The number of data segment available for determining each salinity value is shown. Solid line represents equation (2.28), giving δ/D = 4.2x104s/cm and Ko = 0.12 [Nakawo and Sinha, 1981 / Cambridge University Press].

1 NOV 1 DEC 1 JAN

1 FEB 1 MAR

1 APR

1 MAY

0 –10

Stable salinity s (‰) 0

5

–20

10 –30

0

Depth (cm)

–40 50

100

150

High salinity high growth rate cold period

Mean air temperature (°C)

26 OCT

Low salinity slow growth rate warm period Theoretical

200

Figure 3.12 Composite diagram of stable salinity profile, temporal variation of daily mean air temperature and growth history of the ice (observed and calculated). High salinity is related to high growth that corresponds to colder periods. Warmer days slowed the growth of ice, resulting in lower salinity. Open and solid circles are measured and calculated ice thicknesses, respectively. The dashed line is an “eye-fit” to the observed thickness data and the solid line is based on theory. Both indicate ice growth rate [Nakawo and Sinha, 1981 / Cambridge University Press].

120

SEA ICE

laboratory measurements. It consists of a series of thin wire pairs inserted at different depths in the ice. Each wire is 14 centimeters long, so any point salinity measurement is averaged along this length. The solid ice fraction and the local bulk salinity can be determined from the measured electrical impedance between any pair of wires and the temperature at the same depth. The instrument was inserted in a 1 × 1 m2 hole into 20 cm thick fast ice cover, hence allowing salinity and temperature measurements of thin ice and mush ice layers as formed in the hole over a course of few days. This cannot otherwise be measured from ice cores because of the severe brine drainage during core extraction. In a unique data set, Notz and Worster [2008] show the rapid loss of salt during the onset of ice formation at the ice–water interface until the salinity stabilizes after a few hours or tens of hours in the case of thin ice (Figure 3.13). The history of loss of salt from a given layer is a resultant of two processes, the brine drainage to the underlying layers (or to the seawater when the ice is just formed at that layer) and brine received from the overlaid layers as the ice continued to thicken. Data in Figure 13.3 show the rapid loss of salt from the sensor at 0.5 cm depth because there is no “supply” of brine from layers above (the ice was only 0.5 cm at that time). The salinity stabilized in this layer within 24 hours. On the other hand, at 15 cm depth the freezing was encountered for the first time after about 80 hours. The salinity continued to drop at this depth but at slower rate because of the input brine received from the above 15 cm sea ice depth compensated for the loss to the underlying water or (later) ice. The salinity stabilizes at this layer after more than

60 hours. The stable salinity is about 4 ‰. Most of the salt is lost within 24 hours when the ice is ≤15 cm as the salinity drops from 35‰ to about 4‰. The rate of drop decreases as the ice thickens (i.e., as the growth rate is reduced). Notz and Worster [2008] also determined the salt flux into the underlying ocean by calculating the slope of the mean salinity during the growth of the 17 cm layer of ice. Results from two experiments conducted on different days are shown in Figure 3.14. In the second experiment the temperature was lower than that of the first experiment and the wind was much stronger (gusts at 18 m/s). This led to the formation of frazil ice in the hole, a condition more comparable to new ice formation in the open ocean. Figure 3.14 shows the temporal evolution of the mean salinity of the upper 17 cm layer before and during the freezing. The slopes indicate that in the initial 24 hours, rapid desalination of about 90 g/m2.h took place; but the desalination rate decreased to about 40 g/m2.h and remained at this rate until 80 hours when the entire length of the instrument was covered by ice. As expected, the rate of desalination decreases with time (as ice thickens). Although the authors did not confirm the presence of frazil ice by measurements in experiment 2 in Figure 3.14, the slower desalination rate in this experiment suggests presence of frazil ice crystals, which retain more salts and decelerates the brine drainage process (see sections 2.5.5.2 and 5.3.3.2). It is worth mentioning that Notz and Worster [2009] used the same instrument described in Notz, Wettlaufer,Worster

35

Smean [psu]

15.0 cm

15 10

10

5

5

0

0 20

40

60

80

100

120

140

Hours since start of the experiment

Figure 3.13 Temporal evolution of the average salinity at various depths during a field experiment on landfast ice in the Adventfjorden in Svalbard (78 N), measured in March 2005. The numbers in the legend represent location of different sensors at different depths [Notz and Worster, 2008; Figure 4 / with permission of John Wiley & Sons).

40

20

15

0

Experiment 2

–1

20

m –2 h –1

h

25

5.5 cm 8.0 cm 11.0 cm

25

g

–2

0.5 cm 1.75 cm 3.5 cm

30

40

gm

Average salinity at different depths (‰)

30

90

35

g

m –2 h –1

Experiment 1

0

20

40 60 80 100 120 Hours since start of the experiment

140

Figure 3.14 Temporal evolution of the mean salinity of the upper 17 cm of the water column as it freezes gradually. The thin lines are approximate linear fits to the data from which salinity fluxes into the ocean can be calculated. The two experiments were conducted on ice in the Adventfjorden in Svalbard (78 N, 23 E). Note that the salinity is presented in PSU, which is not very different from ‰ [Notz and Worster, 2008; Figure 5 / with permission of John Wiley & Sons].

SEA ICE PROPERTIES: DATA AND DERIVATIONS 121 (a)

(b) Hydrogen bond

H

θ

H

O

Figure 3.15 (a) Polar structure of a water molecule with the two hydrogen atoms and one oxygen atom. The angle θ is 104.5 , and (b) the hydrogen bond that attracts water molecules together.

[2005] to accurately measure the salinity in the mushy layer of the sea ice. The measurements show the continuity of the salinity across the ice–water interface. They measured a mean salinity of 26‰ when seawater was included in the sample and 18.9‰ from a drained ice sample. The continuity of the salinity across the ice–water interface was also observed by Cox and Weeks [1975]. Salinity measurements of grease ice are presented in Smedsrud and Skogseth [2006]. Only a few scattered points were measured but the work presents the first attempt to characterize salinity of grease ice, which is often found in leads and polynyas in polar winters. Grease ice has usually a brine coating on top of frazil crystals.

3.4. DENSITY OF FIRST-YEAR AND MULTI-YEAR ICE Most liquids shrink as temperature approaches their freezing point because the molecules tend to move slower. Water and ice are exceptions because of the polar nature

of their molecule as shown in Figure 3.15. Positive and negative charges are not evenly distributed across the molecule, hence the dipole nature of the water molecule. The angle θ between the two O-H bonds is 104.5 . This leads to a special force and pattern of attractions between water molecules, called hydrogen bonding. It is this bonding that gives water an unusual behavior when freezing. When water is cooled to near freezing temperature, hydrogen bonding causes molecules to rearrange into lattice structure with “open gaps” within the lattice. This causes decrease of water density below its maximum density of 999.973 kg/m3 at 4 C. Below this temperature the water becomes less dense as the molecules begin to form hexagonal lattice. When water freezes, the open solid structure of ice continues to have larger gaps, causing ice to be less dense than liquid water. The density of pure ice is 917.6 kg/m3 at 0 C; whereas water has a density of 999.87 kg/m3 at the same temperature. Density of ice increases slightly with decreasing temperature, reaching 934.0 kg/m3 at –180 C. Figure 3.16 shows: (1) the change of water density while approaching the freezing point

1010 1000 990 Density (kg/m3)

980

999.8 kg/m3

970 960 950 916.2 kg/m3

940 930 920 910 –50

–45

–40 –35

–30

–25

–20

–15

–10

–5

0

5

10

Temperature (°C)

Figure 3.16 Freshwater and ice density variation with temperature. Transition from water to ice is marked by a sharp drop in density.

122

SEA ICE

0

–4

–4

–4

–8

–8

–8

–12

–12

–12

–16 –20 –24

–16 –20 –24

–28

–28

–32

–32

FYI

–40 –20

–18

Depth (cm)

0

–36

–14

–12

–10

–8

–40

–6

–20 –24

–32

FYI

–36 –16

–16

–28

0

2

4

6

–40 0.4

8 10 12 14 16 18 20

–4

–4

–8

–8

–8

–12

–12

–12

–16

–16

–16

–24

–20 –24

–28

–28

–32

–32

MYI hummock

–18

–16

–14

–12

Depth (cm)

0

Depth (cm)

0

–20

Temperature (°C)

–8

–6

–40

0.7

0.8

0.9

1.0

0.9

1.0

–20 –24 –28 –32

MYI hummock

–36 –10

0.6

Density (g/cm3)

0

–40 –20

0.5

Salinity (‰)

–4

–36

FYI

–36

Temperature (°C)

Depth (cm)

depth to reach values around 910 kg/m3 at depth typically of 0.5 m and below. Melt pond ice of MYI, on the other hand, features nearly constant density profile with values similar to the FYI. Vertical profiles of temperature, salinity and density were obtained from 15 cores of FYI and an equal number from MY hummock and melt pond ice. All cores were obtained from the Lancaster Sound, near Resolute, Canadian central Arctic in May 1992 by the authors of this book. Cores were stored at −22 C immediately after extraction and measurements of density and salinity profiles were performed later at 4 and 2 cm depth intervals, respectively. Profiles in the top 0.4 m of two cores representing FYI and MYI hummock are shown in Figure 3.17. Note the saline free subsurface layer of the hummock ice. The salinity of the measured part of the

0

Depth (cm)

Depth (cm)

(the scale does not accentuate the peak value at 4 C), (2) the sharp drop at the freezing temperature, and (3) the increase of ice density as temperature decreases. The ice density increases slightly with decreasing temperature, reaching 925.7 kg/m3 at −100 C. Ice in nature is polycrystalline. Classification of natural ice is described in section 5.3. Sea ice density is sensitive to its polycrystalline structure, which demonstrates the porosity of the ice at any given temperature. Cox and Weeks [1986] found that the density of gas-free FYI can vary from 920 to 970 kg/m3, depending on the salinity. From numerous field measurements, the density of FYI ice is found to be within the range 890 and 920 kg/m3 and it does not usually vary much with depth. For MY hummock ice, low density is observed in the bubbly subsurface layer (typically 720 kg/m3); but increases with

0

2

4

6

8 10 12 14 16 18 20

Salinity (‰)

MYI hummock

–36 –40

0.4

0.5

0.6

0.7

0.8

Density (g/cm3)

Figure 3.17 Temperature, salinity and density profiles (top 40 cm only) of FYI and MYI hummock obtained from Lancaster Sound, Canadian central Arctic, in May 1992. These data are representative of 15 cores of FYI and 15 cores of each ice type. Note the saline free hummock ice and the remarkable decrease of density near the top surface of this ice (data obtained and plotted by Shokr and Sinha).

SEA ICE PROPERTIES: DATA AND DERIVATIONS 123

FYI core shows the beginning of C-shape near the top surface. Unlike the constant and similar density profiles of FYI and MYI melt pond (pond ice is not shown in Figure 3.17), the density of hummock ice is noticeably low (around 700 kg/m3) near the surface and increases with depth. The high porosity of the subsurface layer of hummock ice is addressed in several publications [Schwerdtfeger, 1963, Weeks, 1976, Sinha, 1984, Wadhams, 2000). Timco and Frederking [1996] presented a review of sea ice density. They reported values obtained from field measurements of FY ice between 840 and 910 kg/m3 for the ice above the waterline, and 900–940 kg/m3 for the ice below the waterline. They also reported values for MYI between 720 and 910 kg/m3. Alexandrov et al. [2010] presented a collection of earlier measurements of ice densities for YI, FYI and MYI. Timco and Frederking [1996] calculates ice density ρi for an ideal ice sheet with uniform thickness and flat snow cover, based on the isostatic equilibrium equation, where input data are ice thickness (hi), snow depth (hs), ice freeboard ( fi), which can only be measured roughly, and snow density (ρs), which increases with snow depth. ρi = ρw −

ρw f i + ρs hs hi

(3.14)

Water density ρw is set to 1025 kg/m3. 3.5. VOLUME FRACTION OF SEA ICE CONSTITUENTS In addition to the pure ice crystals, three inclusions co-exist in sea ice: brine, solid salts (precipitated inside brine pockets), and air. Volume fraction of each constituent can be represented in terms of the ratio of constituent’s volume to the total volume of the sample in parts per thousand. For example, a brine volume of 10‰ is equivalent to a brine volume fraction of 0.010. Equations to calculate the volume fractions of these inclusions in FYI were developed by Cox and Weeks [1983] based on the sea ice phase diagram developed by Assur [1958]. The equations are presented below (recall that they apply only to FYI). Table 3.4 shows the symbols used in these equations. Table 3.4 Symbols used in the equations to calculate volume fractions of sea ice constituents.

Mass Volume Density Salinity

Salinity is described in terms of mass ratio. Hence, the salinity of sea ice (Ssi) and salinity of brine (Sb) are defined as: mss + msb M si msb Sb = mb

S si =

(3.15) (3.16)

These two equations yield: msb = S b ρb V b

(3.17)

3.5.1. Brine Volume Fraction Brine volume represents the amount of liquid in the composite sea ice medium. Brine volume and brine salinity at any point within the sea ice sheet are functions of local ice temperature. Depending on the temperature, brine salinity can be determined using the sea ice phase diagram (Figure 2.2). There are two approaches to determine brine volume. The first is by deriving an expression based on equations (3.15)–(3.17) after incorporating an expression for brine salinity. This approach is presented in Cox and Weeks [1983]. The second is by using empirical equations that relate brine volume to ice temperatures and salinity. The first approach is reiterated here first. By combining equations (3.15) through (3.17) the following expression for brine volume can be obtained: Vb =

M si S si 1 ρb S b 1 + k

(3.18)

mss msb

(3.19)

where, k is defined as k=

since Mi = ρiVi, the brine volume fraction (ratio of brine volume to the bulk ice volume) can be obtained from the equation: Vb ρ S si 1 = si V si ρb S b 1 + k

(3.20)

At temperatures above −15.2 C, the presence of solid salts can be neglected. Hence k = 0 in equation (3.20) and the equation can be reduced to

Pure ice

Brine

Salt in brine

Solid salt

Air

Bulk sea ice

Vb ρ S si = si V si ρb S b

mpi Vpi ρpi -

mb Vb ρb Sb

msb -

mss Vss ρss -

ma Va ρa -

Msi Vsi ρsi Ssi

Brine density (kg/m3) is related to brine salinity (‰) through an empirical equation [Cox and Weeks, 1975]: ρb = 1000 + 0 8 S b

(3.21)

(3.22)

124

SEA ICE

The two parameters Sb and k are unique functions of temperature. The dependence of Sb on the sea ice temperature Tsi (in C) is given by the following empirical equations [Assur, 1958] based on best fit results of measured data. Since sodium sulfate starts to precipitate at −8.2 C, two equations are needed because it is difficult to develop a single equation for the full temperature range. S b = 1 725 − 18 756T si − 0 396T 2si ,

(3.23)

for − 2 0 C ≥ T si ≥ − 8 2 C S b = 57 041 − 9 29T si − 0 162T 2si ,

(3.24)

for − 8 2 C > T si ≥ − 22 9 C

Notz and Worster [2009] presented an alternative expression for brine salinity that can be used for the temperature range between −22 C and 0 C. S b = 21 4T si − 0 886T 2si − 0 017 T 3si

(3.25)

This equation is accurate at lower temperatures. It fits results from a model presented in the same reference based on data presented in Assur [1958] with a maximum deviation of less than 5%. Given the expressions of Sb [equations (3.23) and (3.24)] and ρb in equation (3.22), the denominator in equation (3.21) can be combined into one term, F1(T), and the equation can take the form: Vb ρ S si = si V si F1 T

(3.26)

The values of F1(T) are given in Cox and Weeks [1983]. Brine volume can also be expressed as a function of the average ice salinity Ssi (same as bulk salinity, but this symbol is used here to be consistent with the original reference) and temperature Tsi . A reasonable estimate can be obtained using the equation presented in Frankenstein and Garner [1967], where Ssi is in ‰. V b = 10 − 3 S si 0 532 − 49 185 T si

(3.27)

where, −0.5 C ≥ Tsi ≥ −22.9 C. A more accurate expression is presented in Cox and Weeks [1983]: V b = ρsi S si F T si

(3.28)

where,

Table 3.5 Values of the constant A versus average ice temperature Ti ( C). Tsi −2 −4 −6 −8 −10 −12 −14

A

Tsi

A

0 0.148 0.387 0.660 8.256 30.493 38.421

−16 −18 −20 −22 −24 −26 −28

44.952 50.808 56.851 63.015 217.168 537.697 842.341

3.5.2. Solid Salt Volume Fraction The mass of solid salts at a given depth is proportional to the mass of brine: mss = Amb

where, A is a function of temperature, determined from the phase equilibrium diagram, as presented in Table 3.5. This relation can be written as: ρss V ss = Aρb V b

F T si = − 4 732 − 22 45T si − 0

−0

(3.32)

from which an expression for the solid salt volume fraction can be derived V ss ρ Vb =A b V si ρss V si

(3.33)

ρb can be obtained from equation (3.22). The solid salt density ρss can be assumed to be constant =1500 kg/m3. 3.5.3. Pure Ice Volume Fraction The mass of pure ice is mpi = M si − mb − mss

(3.34)

Replacing the mass terms in the above equations by multiplication of density and volume, the above equation can be solved for the pure ice volume fraction Vpi/Vsi using V pi ρ ρ Vb = si − b 1+ A V si ρpi ρpi V si

(3.35)

The density of sea ice ρsi (kg/m3) can be calculated from the following equation, given by Pounder [1965], where the ice temperature Ti is in C: ρsi = 917 − 0 14 T si

64T 2si

(3.31)

(3.36)

01074T 3si

for − 2 ≥ T si ≥ − 22 9 (3.29) F T si = 8999 + 1309T si + 55 27T 2si + 0 71T 3si for − 22 9 ≥ T si ≥ − 30

(3.30)

3.5.4. Air Volume Fraction Air volume fraction is usually considered to be much smaller than brine volume fraction in FYI. It cannot be accurately measured. So, it is usually obtained by

SEA ICE PROPERTIES: DATA AND DERIVATIONS 125

subtracting the summation of volume fraction of the other components from the total volume fraction (i.e., 1) Va V b V pi V ss = 1− − − V si V si V si V si

(3.37)

Due to brine drainage, the air volume can be significant as air replaces some brine pockets. For this reason, some studies use the total porosity (VT), which is the summation of the brine and air volume fractions. VT = Vb + Va

(3.38)

However, air volume fraction, which can be derived from the bulk ice density, must be known accurately. Cox and Weeks [1983] developed equations to calculate the total porosity of sea ice. It should be mentioned that, unlike the case of FYI, in MYI the air volume fraction makes up a major portion of its total porosity.

temperature are plotted in Figure 3.19. As ice temperature decreases, brine volume decreases and brine salinity increases (a consequence of brine pockets shrinkage). A peak in brine salinity is noticeable at the precipitation temperature of sodium chloride (around −22.9 C). Below this temperature, a considerable mass of salt freezes and brine salinity decreases accordingly. Another slight discontinuity in the brine salinity curve occurs at about −8.2 C. This is the precipitation temperature of sodium sulfate (Na2SO4.10H2O), which constitutes only 4‰ of impurities in seawater [Assur, 1958]. Unlike brine volume, brine salinity is independent of ice salinity, as shown by

1.0 0.9

The effects of ice temperature on brine, solid ice and air volume fractions are presented in Figure 3.18 for ice density of 900 kg/m3 and salinity of 10‰. The most important feature in the figure is the non-linear variation of volume fractions of ice and brine at temperatures above −10 C. These two parameters are complementary because the air volume fraction is very small and not affected by changes in ice temperature of FYI. Below −10 C the brine volume fraction decreases linearly with decreasing temperature (the opposite is observed for the ice volume fraction). Variation of brine salinity [equations (3.23) and (3.24)] and brine volume fraction [equation (3.27)] with

0.6 0.5

Brine volume fraction Ice volume fraction Air volume fraction

0.4 0.3 0.2 0.1 0.0 –30

–25

–20

–15

–10

–5

Figure 3.18 Variation of brine, air and pure ice volume fractions versus temperature for FYI of constant density 900 kg/m3 and salinity 10‰ [Shokr and Sinha, 1994].

(b) Ice salinity = 10 ppt

0.10

Brine volume

140

Brine salinity

120 100 80

0.50 0.00 –30

–26 –22 –18 –14 –10 Temperature (°C)

–6

–2

Brine volume (%)

0.15

200

0.20 Brine salinity (ppt)

160

220

Ice salinity = 20 ppt

200 180

240

0.25

220

0.20

0

Ice temperature (°C)

240

0.25

Brine volume (%)

0.7

180 0.15

160 Brine volume

140

Brine salinity

120

0.10

100 80

60 40

0.50

20

0.00 –30 –26 –22 –18 –14 –10 Temperature (°C)

Brine salinity (ppt)

(a)

Volume fraction (%)

0.8

3.5.5. Temperature Dependence of Volume Fractions of Different Components

60 40 –6

–2

20

Figure 3.19 Variation of brine volume and brine salinity with bulk ice temperatures; calculations were made for ice salinities of (a) 10‰ and (b) 20‰ (calculations are based on equations presented in this section using data obtained by the authors from Resolute Bay in 1995).

126

SEA ICE

the nearly identical curves of brine salinity in both segments of the figure, for bulk ice salinity 10‰ and 20‰.

3.6. THERMAL PROPERTIES Since sea ice is a composite material where solid, liquid, and gas phases co-exist, its thermal behavior is also complex. Change of phase between solid and liquid takes place continually in response for variations in temperature in order to maintain the thermal equilibrium inside the sea ice composition. This is manifested in the melting of pure ice crystals at the brine pockets’ wall as the ice temperature increases (the melting point of brine decreases as brine salinity increases) and vice versa. It is also manifested in the precipitation of solid salts, which were originally dissolved in brine, when the temperature drops to reach the eutectic point of any salt constituent. The change of phase within the sea ice cover gives rise to a large change of its thermal properties. Since sea ice is essentially a solid material that hosts brine in liquid phase, heat transfer within the ice is largely accomplished by conduction and partly by convection. The key thermal processes in sea ice are: (1) the heat exchange at the ice surface (between air and ice), (2) the heat conduction through the ice and snow during winter, from the bottom up, (3) the heat exchange at the ice–water interface and (4) the latent heat gain during freezing and loss during melting and also due to precipitation of solid salts in brine pockets. Thermal parameters affected by these processes include conductivity, specific heat, and latent heat of fusion. The ratio of the thermal conductivity to specific heat is known as thermal diffusivity. These parameters depend on ice temperature, which determines the ice composition under phase equilibrium. Together with the corresponding thermal properties of the snow, the parameters can be used to estimate the heat exchange between the ocean and the air through the ice cover and therefore determine the ice growth or decay conditions. Calculations of thermal parameters of sea ice require information about seawater parameters. A comprehensive presentation of those parameters is included in Sharqawy, Leinhard, Zubair [2010]. Accurate profiles of thermal properties of sea ice and snow are needed to improve the performance of thermodynamic ice growth models. Earlier models are found in Untersteiner [1964], Maykut and Untersteiner [1971], Maykut [1978] and Semtner [1976]. Advanced models that are more suitable for climate research have been developed later by Ebert and Curry [1993], and Flato and Brown [1996]. A notable energy-conserving thermodynamic model is presented in Bitz and Lipscomb [1999] and has been used in operational ice products. The model accounts not only for the heat capacity of ice but also for the energy required for melting. As ice cools, water

in the brine pockets freezes and consequently the salinity of brine pockets increases to maintain thermal equilibrium. Inversely, as the ice warms, ice along the walls of the brine pockets melts and the salinity of brine pockets decreases. Therefore, the heat capacity of sea ice includes not only the energy required to raise the ice temperature but also to melt ice crystals along the walls of brine pockets. Winton [2000] presented a three-layer thermodynamic model of sea ice suited for climate applications, which takes into consideration the dependence of ice thermal properties on temperature and salinity (note that simple models assume the ice sheet as a single layer of fixed heat capacity). This model has been used to produce ice products from NOAA. An application of a one-dimensional thermodynamic model using different parameterizations of radiative fluxes is presented in Cheng [2002]. A more recent three-dimensional global model for sea ice dynamics and thermodynamics has been developed by Vancoppenolle et al. [2009]. It is a coupled ice–ocean General Circulation Model (GCM) where a simple representation of thermodynamics of snow and ice is incorporated (e.g., snow is one layer of constant physical properties). 3.6.1. Thermal Conductivity of Sea Ice Thermal conductivity (K) is the ratio between the heat flux (F) and the vertical temperature gradient: F = −K

∂T ∂z

(3.39)

Because of the difference in thermal conductivity between the three components of sea ice: pure ice crystals, brine and air, the overall thermal conductivity becomes a function of their relative volume fractions and geometric arrangement. Freshwater has thermal conductivity around 0.611 W/m.K, slightly higher than the conductivity of seawater with 35‰ salinity, which is around 0.6 W/m.K at 25 C. This feature is used by oceanographers to determine the salinity of seawater. The typical thermal conductivity of pure ice is 2.26 W/ m.K [Slack, 1980]. At −20 C and salinity 6‰ the thermal conductivity of sea ice is 2.10 W/m.K. That is because both the inclusions (air and brine) are poor conductors. For example, thermal conductivity of air is 0.024 W/m.K, i.e., two orders of magnitude lower than that of ice, while that of NaCl brine solution varies between 0.50 and 0.56 W/m.K for temperature between 0 C and −20 C. Levy [1982] estimated thermal conductivity of brine with 75‰ salinity at −4 C to be 0.55 W/m. K but this decreases to 0.535 W/m.K when salinity increases to 230‰. The dependence of thermal conductivity of FYI on the brine volume and salinity cannot be overemphasized. In

SEA ICE PROPERTIES: DATA AND DERIVATIONS 127

general, brine salinity is inversely related to the brine volume (expansion or contraction of brine pockets/channels decreases or increases brine salinity, respectively). Brine volume and salinity at the early phase of sea ice formation set the tone for the temporal change of thermal conductivity throughout the history of sea ice growth. These initial values are determined based on the meteorological conditions (mainly air temperature and wind) and oceanic conditions (mainly steady or turbulent) at the onset of sea ice formation. As mentioned earlier, frazil and granular sea ice with small crystals, which are usually formed under turbulent conditions, retain higher brine than the large columnar crystals, which are usually formed under quasi-steady conditions. Thermal conductivity of sea ice decreases with increasing salinity but for a given sea ice salinity, thermal conductivity decreases as temperature increases. That is because higher temperatures are associated with higher brine volume. Even a small increase of ice surface temperature in response to an increase of atmospheric temperature above −5 C will decrease the thermal conductivity [Weeks and Ackley, 1982]. While ice is warming toward the melting point and as the relative amount of liquid at the surface increases, the thermal conductivity becomes meaningless because the increased mobility of the brine triggers convective heat transfer. These issues complicate accurate estimate of thermal conductivity of sea ice. The higher thermal conductivity of ice crystals compared to the conductivity of the brine explains the vertical (cellular) growth of sea ice at the water interface (section 2.5.1). This growth pattern is caused by fact that thermal conductivity along the vertical ice cells (arms/ plates/dendrites), shown in Figure 2.19, is much higher than the conductivity across the brine between ice platelets. Therefore, the heat flows naturally along the vertical direction of those dendrites causing ice to grow in the same direction. Thermal conductivity of sea ice is a directional property. In a single ice crystal, its value parallel to the c-axis may be about 5% less than that along the normal to the c-axis [Fukusako, 1990]. This is part of the reason (though not a major reason) for the columnar growth of ice in both fresh and seawater. Several authors determined the thermal conductivity of YI and MYI from in situ measurements of temperature gradient. Other studies present models to calculate the thermal conductivity. The estimated values from measurements varied between different studies. This represents a source of uncertainty in ice thermodynamic and climate modeling. Trodahl et al. [2000] presented a set of Antarctic sea ice thermal conductivity estimates from measurements conducted in McMurdo Sound, Antarctica. They confirmed the temperature dependence of the conductivity, especially within the sub-freezing temperature range. For ice temperature range between

−10 C and −30 C the conductivity was nearly stable at 1.95 W/m.K. The authors provided evidence that change of sea ice crystal orientation does not affect the heat flow significantly unless the orientation affects the geometry and distribution of the brine channels. Pringle, Trodahl, Haskell [2006] presented thermal conductivity data from measurements on FYI and MYI cores from the same area (McMurdo Sound, Antarctic), which can be used as first approximation. They found that the thermal conductivity of FYI within the top layer (0–100 mm) and a layer at depth 450–550 mm layers were 2.14 ± 0.11 and 2.09 ± 0.11 W/m.K; respectively. For the significantly bubblier MYI they reported values around 1.88 ± 0.13 W/m.K. Empirical equations are presented in the literature to estimate thermal conductivity of sea ice (Ksi). If air volume fraction is neglected, Ksi becomes a function of the conductivities of the pure ice Kpi and brine Kb , weighed by the volume fraction of the brine. Anderson [1958] presented the following equation, which applies when the heat is transferred vertically; i.e., during ice congelation; K si = K pi 1− 0 1V b + 0 1K b V b

(3.40)

where, Kb is a function of temperature and brine salinity as will be shown later. According to equation (3.40), if brine volume is 10% (i.e. Vb = 0.1) with brine conductivity Kb = 0.2 W/m.K, Ksi drops slightly from its value of 2.25 W/m.K for pure ice 2.23 W/m.K. Air volume fraction also reduces Ksi by the ratio (1 − Va)/(1 + Va). This is a very small ratio. The effect of solid salt formation on thermal conductivity of the sea ice mixture is even smaller and often neglected. For very cold ice, when ice is largely solid with very small brine volume, another equation based on conduction in a porous media has been suggested: Kb K pi 1 + 0 1 a − 1 Vb

1 − 0 1V b 1 − a K si = K pi

(3.41)

where, a=15

Kb K pi

+ Kb

(3.42)

Expressions for Kpi and Kb can be obtained following the initial work of Schwerdtfeger [1963] that was complemented by Yen, Cheng, Fukusako [1991]. K pi = 1 16 1 91 − 8 66 × 10 − 3 T si + 2 97 × 10 − 5 T 2si (3.43) K b = 0 4184 1 25 + 0 03T si + 0 00014T 2si

(3.44)

128

SEA ICE

The CIS Community Ice–Ocean Model (CIOM) uses this conductivity parameterization. More recently, Pringle et al. [2007] proposed an alternative expression which is valid for bubbly-rich ice (i.e., MYI) as well as brine pockets rich ice (i.e., FYI):

Conductivity (Wm–1K–1)

3.0 2.5 2.0 Pure ice Brine

1.5 1.0

K si =

0.5

ρsi S si ρsi − ρpi 2 11 − 0 11T si + 0 09 − ρpi T si 1000

(3.47)

0.0 0

−5

−10 −15 Temperature (°C)

−20

−25

Figure 3.20 Calculated thermal conductivity of pure ice and brine for different temperatures from equations (3.43) and (3.44). Note the divergence of the values as temperature decreases.

where, the temperature is in C and the conductivity is in W/m.K. Plots of calculations from these two equations are shown in Figure 3.20. Conductivity of pure ice increases while conductivity of brine decreases as ice becomes colder. These two opposite trends balance their effects on the conductivity of the sea ice mixture. This implies that for the same brine volume the thermal conductivity of sea ice is a weak function of ice temperature. In other words, the brine volume is the key factor in determining the sea ice thermal conductivity. Schwerdtfeger [1963] developed a model for thermal conductivity assuming sea ice as a mix of bubbly pure ice medium enclosing a number of vertical cylindrical brine pockets (parallel to the heat flow direction). If the thermal conductivity of sea ice is considered to be equal to the summation of the conductivities of pure ice and brine weighed by their cross-sectional areas, then Ksi is given by the following equation in terms of bulk sea ice salinity Ssi, which is an indicator of brine volume: K si = K pi − K pi − K b

S si ρsi pT si ρb

(3.45)

where, ρsi and ρb are density of sea ice and brine in kg/m3, Kpi and Kb are the thermal conductivity of pure ice and brine, Tsi in C, Ssi in ‰ and p is the slope of the phase boundary in the phase diagram (1/p = −55 C). This equation is known as “bubbly-medium model” because it was derived based on the assumption of a bubbly host material with brine inclusions. Therefore, it is more valid for MY ice, especially the bubble-rich hummock ice. Another simple equation, which applies only to MYI and is being used in a few climate models, was developed by Maykut and Untersteiner [1971] based on the 1-D thermodynamic sea ice model: K si = 2 03 + 0 117

S si T si

(3.46)

The unit of each parameter is the same as used in the previous equations. Tsi is in C. The equation was verified using in situ measurements of conductivity of sea ice in the Arctic and Antarctic. Thermal conductivity can be calculated from measurements of temperature profile along the ice thickness with thermistor or thermocouple arrays conducted in the field or in a laboratory on either natural or artificial samples. Calculations are complicated because of the possibility of perturbing the local heat flow during measurements or possible temperature inversion within the ice cover. Moreover, the simultaneous effect of the latent heat required for precipitation of solid salts inside brine pockets within the temperature range between −8.2 C and −22.8 C requires careful consideration. A methodology of measurements of thermal conductivity of sea ice using this approach is described in McGuinness et al. [1998] and Pringle, Trodahl, Haskell [2006]. It assumes 1-D heat conduction and measures the temperature gradient in the direction of heat conduction z using an array of thermistors. It then uses the following form of conservation of energy equation (T is the local temperature): D

∂U ∂2 T ∂K si ∂T = K si 2 + ∂t ∂z ∂z ∂z

(3.48)

where D is the thermal diffusivity of ice and U(Sb , T) is the sea ice internal energy per unit mass found by integrating the specific heat equation [equation (3.58)]. Three thermistors are required to estimate the second derivative in the RHS in the above equation. Since the conductivity may not be constant over the depths of the three thermistors, the second term on the RHS is introduced to represent the derivatives of the conductivity. The omission of this term does not affect estimates of conductivity near the surface [Pringle, Trodahl, Haskell, 2006]. The derivatives in equations (3.48) are determined by finite differences. For example, ∂U = U t+ Δt − U t− Δt ∂t

2Δt

(3.49)

and ∂2 T = T z+ Δz − 2T z + T z− Δz ∂z2

Δz2

(3.50)

SEA ICE PROPERTIES: DATA AND DERIVATIONS 129

The last term in equation (3.48) can be expanded as follows in order to facilitate its estimation: ∂K si ∂T ∂K si ∂T ∂K si ∂S si ∂K si ∂ρsi ∂T = + + ∂z ∂z ∂T ∂z ∂S si ∂z ∂ρsi ∂z ∂z (3.51) ∂T ∂S si ∂ρ , and si can be deter∂z ∂z ∂z mined from measurements of temperature, salinity and density profiles, while the derivatives of Ksi are determined from equation (3.47). Ksi can be determined using the temperature, salinity and density measurements along with equation (3.48) as the best fit to the scatter plot of ∂U ∂K si ∂T − versus ∂ 2T/∂z2. ρsi ∂z ∂z ∂t Pringle et al. [2007] used this method to estimate the conductivity of landfast FYI in the Arctic (Chukchi Sea near Barrow, Alaska) and Antarctic (McMurdo Sound). They compared the measurements with their model [equation (3.47)] and the Maykut and Untersteiner [1971] model [equation (3.46)]. Figure 3.21 shows the measured and the modeled data. The bubbly-brine model prediction provides a better comparison with the measurements. The spatial derivatives

3.6.2. Thermal Conductivity of Snow Snow is a composite material consisting of fresh snow flakes, ice grains, air and possibly water when it is wet.

Therefore, heat transfer inside the snow pack takes one or more of the following three forms: by conduction between interconnected grains of snow ice, by convection through water contents if existing within the snow pack, and by radiation across the air inside the snow. These simultaneous mixed processes render the measurements of thermal conductivity of snow (Ks) difficult. The measured values are commonly referred to as effective conductivity [Weeks, 2010]. To further complicate the issue, the spatial variability of snow composition and properties is also high within short distances due to snow drifting and local weather conditions. The vertical profile of snow may reveal layering of different structural types and therefore properties. Sturm, Holmgren, Perovich [2002] monitored the evolution of the snow during the Surface Heat Budget of the Arctic Ocean expedition (SHEBA) in the Arctic from October 1997 to October 1998. For synopsis of this expedition see Perovich et al. [1999]. At the peak of the snow depth in April, Sturm, Holmgren, Perovich [2002] measured the stratigraphy of the snow along with profiles of a few physical parameters (density, temperature, salinity, thermal conductivity, and salinity). They identified many structurally different layers that are associated with different thermal properties. Figure 3.22 shows the different layers, which can be grouped into three major categories: recent snow, wind slab, and depth hoar. Each layer existed with a different thickness range and tended

Thermal conductivity (W.m–1.K–1)

2.6

FY McMurdo arrays FY Chukchi arrays Pringle et al. [2006] Nazintsev [1959] Nazintsev [1964]

2.4

Schwerdtfeger [1963] Lewis [1967] heat flow Lewis [1967] ice growth

2.2

Bubbly brine S=4.8 ppt M&U (1971) S=4.8 ppt

2.0

1.8 –25

–20

–15

–10

–5

0

Temperature (°C)

Figure 3.21 Measured and modeled conductivity of sea ice. Measurements were obtained from FY fast ice in the McMurdo Sound (Antarctic) and the Chukchi Sea (Arctic). Open symbols represent historical measurements as indicated in the legend. References are cited in the original paper. Solid squares (with and without the white center) are measurements presented in Pringle et al. [2007]. Model calculations are obtained from equation (3.47) (Bubbly-brine model) and equation (3.46) (M&U model). The upper and lower curves of the models’ calculations are for ice salinity 8‰ and 4‰, respectively [Pringle et al., 2007; Figure 7 / with permission of John Wiley & Sons].

130

SEA ICE

In this equation, ρs is in g/cm3 and the resulting Ks is in cal/s.cm.C. Note that 1 watt = 0.239 cal/s. This equation is most frequently cited and used to predict conductivity of the snow. Yen [1981] developed the following expression that was used later in thermodynamic model studies such as the model in Ebert and Curry [1993];

May 7 (Drifting) New & recent

Apr. 11 Apr. 7–9

recent

Jan 29 (Snowfall)

Fine-grained

Feb. 2 (Drifting) Dec. 2–8

Wind slab Depth hoar

K s = 2 2236 ρs 1000

K s = 0 023 + 0 234 ρs

Nov. 11–13

Oct. 29–30

Snow ice

(3.54)

0 156 ≤ ρs ≤ 0 6 (3.55)

The same reference suggests another logarithmic expression valid for ρs ≤ 0.6

Snow ice

Figure 3.22 Stratigraphy of snow on Arctic sea ice obtained in April 1998 during the SHEBA program. Snow depth was between 0.28 and 0.65 m [Sturm, Holmgren, Perovich, 2002; Figure 3 / with permission of John Wiley & Sons].

to be homogeneous with recognizable density, grain characteristics and hence thermal properties. These layers, however, evolve and change continually throughout the winter due to ice growth and deformation. The wide range of snow conditions and layering contributes to an equally wide range of Ks (between 0.03 and 0.80 W/m.K). Lower values are expected from dry fresh snow and higher values from wet, metamorphosed or compacted snow. Measurements of thermal conductivity of snow over land have been presented in the literature since 1886. The variability of thermal conductivity of snow on sea ice is greater than that of FYI, which varies around 2.1 W/m.K (Table 3.1). Accurate measurements of Ks that capture the spatial, temporal and stratigraphic variations of snow cover are difficult to obtain. This is partly because the snow base absorbs brine from the underlying sea ice and metamorphoses according to variations in air temperature. Simple models (mostly empirical equations) have been suggested to estimate Ks in terms of snow density, wetness, and geometrical forms of the inclusions. A few expressions for Ks as a function of snow density ρs were suggested by a few authors. Although most of these expressions were not developed for snow on sea ice, they can practically be applied to it. Bader [1962] presented a few expressions developed in studies conducted before 1945. Among them is the first equation developed in 1893, called Abel’s equation (the original paper by the same author is in German); K s = 0 0068ρ2s

ρs < 0 156

K s = 0 138 − 1 01 ρs + 3 233 ρ2s

Nov. 6

Chains of depth hoar

(3.53)

Sturm et al. [1997] suggested other expressions:

Dec. 2

Wind slab to depth hoar

1 885

(3.52)

K s = 10 0 00265 ρs − 1 652

(3.56)

In equations (3.54)–(3.56) ρs is in g/cm and Ks in W/m.K. Many methods have been developed to calculate thermal conductivity of the snow from temperature measurements. Different approaches include using steady-state heat flow across snow blocks, transient heat flow in the snow pack or its samples, and Fourier analysis of temperature cycle in the snow. The methods are discussed and assessed in Pratt [1969] in terms of their merit to predict thermal conductivity of lowconductivity materials. Sturm et al. [1997] compiled conductivity values from 27 studies conducted on snow over land before 1993. They presented the results in a table that includes the author’s name(s), location of measurements, and methods of calculating the conductivity. Figure 3.23 includes data compiled from 23 studies. Different symbols denote different data sources. References to the data sources can be found in the original figure (Figure 4 in Sturm et al. [1997]). The purpose of presenting the graph here is to show how erratic the measurements of snow thermal conductivity are. At any given density between 0.2 and 0.4 g/cm3 the snow conductivity varies largely by more than an order of magnitude. The authors attributed the scattering of the data to the inaccuracy of some measurements as well as the influence of other factors such as snow wetness, grain size, and ice layering. The snow temperature determines the water vapor contents within the snowpack and this affects the thermal conductivity. The temperature dependence of the conductivity must be taken into consideration when explaining the scattering of the data. The snow temperature from which the data in the figure were obtained varied over a wide range; between 0 and −60 C. Nevertheless, even with the significant scattering, there is an obvious trend of conductivity increase with snow density. Abel’s regression 3

SEA ICE PROPERTIES: DATA AND DERIVATIONS 131

KS (W m–1 K–1)

1

9 8 7 6 5

Center of data

4 3 2

0.19

0.5

8 7 6 5 KS (W m–1 K–1)

0.45

4 3 2

0.35 0.25

0.31 W m–1 K–1 Yen (1981)

0.15

Sturm et al. (1997) 0.05 200

0.01 0.0

0.1

250 300 350 400 Snow density (kg m–3)

0.2 0.3 0.4 0.5 Snow density (g cm–3)

0.6

450

0.7

Figure 3.23 Effective thermal conductivity measurements of snow versus density from 23 studies. The symbols refer to the different studies as listed in Sturm et al. [1997]. The regression curve is drawn using equation (3.52). The inset shows plots of equations (3.53) and (3.55). The main plot and the inset are adapted from Strum et al. [1997] and Lecomte et al. [2013], respectively.

equation [equation (3.52)] is fitted to the data in Figure 3.23 for comparison. The equation confirms the visible trend. The center of mass of the data points is at Ks = 0.251 W/ m.K and ρs = 0.273 gm/cm3. The inset in Figure 3.23 shows two plots obtained from equation (3.53) [Yen, 1981] and equation (3.55) [Sturm et al., 1997]. This inset is a modified version of a plot presented in Lecomte et al. [2013]. The snow density varies over the range between 200 and 450 kg/m3, i.e., it is part of the wider range shown in the main plot. It can be seen that the two equations depict the same trend of conductivity increase with density yet they do not coincide with Abel’s equation. The typical value of snow conductivity used in sea ice thermodynamic coarse resolution ice– ocean models (0.31 W/m.K) is based on Abel’s equation at density 330 kg/m3. This value is greater than the thermal conductivity produced by the other two equations. The above discussions point to the importance of an accurate estimate of the thermal conductivity of snow which has to be used in thermodynamic ice growth models. In general, the lower thermal conductivity of the snow protects the ice underneath it against sharp variations of atmospheric temperature (see Figures 3.3 and 3.4). Consequently, the heat flux through ice is reduced and the ice growth slows down in presence of snow.

Another factor that affects thermal properties of the snow is the solid ice inclusions. This is possible as a result of thaw-refreezing cycle. Satyawali and Singh [2008] studied the effects of a few idealized shapes of solid ice inclusions on thermal conductivity of the snow using a one-dimensional heat transfer model. Solid particles were represented as aggregations of three geometrical shapes: spherical, cylindrical, or cubical arranged in cubic packing. The authors concluded that the spherical shape of solid ice in the snow is associated with a higher conductivity than the cubical shape. Thermal conductivity from using any shape depends on the characteristic ratio between the particle size and spacing between particles. This is denoted as “particle contact ratio.” Figure 3.24 shows results from using the three particle shapes and two particle contact ratios. It also includes comparison with the logarithmic expression developed by Sturm et al. [1997] [equation (3.56)]. Once again, the results confirm the increase of snow thermal conductivity with the snow density. Equation (3.56) returns higher values than the model of Satyawali and Singh [2008] for cubical and cylindrical grains when ρs > 350 kg/m3. Note the sharper increase of conductivity as the snow becomes increasingly packed, regardless of the shape of the solid ice grain. In a more recent study, Strum et al. [2002] compiled measurements of Ks from eighty-nine points of snow on sea ice in the Beaufort Sea during the SHEBA program. Average values ranged from 0.078 W/m.K for dry fresh snow to 0.29 W/m.K for ubiquitous wind slab. The study presented a trajectory of thermal conductivity of the snow as it evolves through different phases in winter. With a steep temperature gradient within the snow depth the snow at the base metamorphoses into a depth hoar. This causes reduction of Ks. With more time, crystal size in the hoar layer increases and this metamorphism causes Ks to decrease to values approaching that of new snow. In the spring, melting and refreezing of snow produces ice with much higher values of Ks.

3.6.3. Specific Heat of Sea Ice The specific heat is defined as the amount of heat required to raise the temperature of a unit mass of a substance by one degree. This is derived from the definition of the heat capacity, which is the heat required to raise the temperature of a substance by one degree, regardless of its mass. The higher the specific heat the more difficult it is to change the temperature of the material. Freshwater has highest specific heat capacity than any liquid; around 4.187 kJ/kg.K compared to metals that have values mostly between 0.3 and 0.9 kJ/kg.K, and air that has a value around 1.0 kJ/kg.K. The specific heat of water decreases with salinity. For example, the specific heat of

SEA ICE

97 (19 al.

ca

l

et

0.6

rm

0.6

eri 0.4

0.4 her

Sp

l

ica

al 0.2

al

300

400

ica

l Cy

Cubic

200

l

r ind

dric

in Cyl

0.2

0 100

et al. (1 99 7)

0.8

Particle contact ratio = 0.2

St ur m

0.8

)

1

St u

Conductivity (Wm–1K–1)

Particle contact ratio = 0.1 1

Sp h

132

500

600

Density (kg m–3)

0 100

al

Cubic

200

300

400

500

600

Density (kg m–3)

Figure 3.24 Variation of effective thermal conductivity of snow with snow density computed from a onedimensional heat transfer model suggested by Satyawali and Singh [2008] for three shapes of ice grain inclusions in the snow (spherical, cylindrical, and cubical) and two particle contact ratios (see text for definition). Snow temperature of −5 C was used in the calculations. Empirical results from the model by Strum et al. [1997] are also included [Satyawali and Singh, 2008].

seawater with 35‰ salinity is 3.898 kJ/kg.K. Therefore, saltwater heats up faster than freshwater. During the freezing process of fresh or saline water, the specific heat becomes infinity because all the heat is consumed in the phase transformation (not raising temperature). It then decreases sharply with temperature. Ono [1967] presents tables and graphs of specific heat Csi and latent heat Lsi of saline ice within a range of temperature between −0.1 C and −8 C for discrete salinities between 0 and 11‰. At the same temperature Csi increases with salinity. The increase is tangible at subfreezing temperatures and becomes negligible as the temperature decreases. For any given salinity of water, Csi drops sharply within the first few degrees below the freezing point then the drop becomes milder until it stabilizes at temperature between −4 C and −10 C, depending on the salinity. The stable value at −16 C, for example, vary between 1.97 kJ/kg.K at zero salinity and 2.76 kJ/kg.K at 10‰ salinity. Schwerdtfeger [1963] proposed the following equation to estimate the specific heat of sea ice Csi. C si = C pi +

S si S si C w − C pi − Lsi αT i αT 2i

(3.57)

where, Ti is the ice temperature in C, Ssi is the ice salinity in ‰, Cpi is the specific heat of pure ice (0.48 cal/g. C), Cw is the specific heat of water (1.01 cal/g. C), Lsi is the latent heat of sea ice (79 69 cal/g) and α is a coefficient that reflects the linear relationship between ice temperature

and brine salinity (= −0.0182 C−1). Equation (3.57) was used to generate plots of variation of Csi with ice temperature for selected salinities (Figure 3.25). A few observations can be drawn from Figure 3.25. First, saline ice has higher specific heat than freshwater ice at the same temperature. That is because more amount of heat is required for melting or freezing the ice at the boundaries of the brine pockets or precipitating or dissoving salts inside the pockets. For example, at −8 C freshwater ice has specific heat 2.01 kJ/kg.K, but the value for saline ice with 8‰ is 4.4 kJ/kg.K. Secondly, as ice warms the heat is consumed in breaking down the structure of the solid before raising its temperature, hence the specific heat increases. Thirdly, the specific heat is very sensitive to both salinity and ice temperature at subfreezing temperatures (down to, say, −6 C). These two parameters are usually correlated within this temperature range (see Figure 6 in Shokr and Dabboor [2020]). Hence, it is difficult to parameterize the specific heat of sea ice at sub-freezing temperatures. Figure 3.25 shows that 2.30 kJ/ kg.K can be assumed as a stable value of Csi for salinities > 8‰ and temperatures below −16 C. Finally, while the specific heat of water decreases as salinity increases, the reverse is true after freezing. Another classical equation to calculate the specific heat of sea ice is presented in Ono [1967]. Here, the ice temperature does not appear explicitly in the equation but the values of the independent terms in the RHS of the equation should be used at the selected temperature. The

SEA ICE PROPERTIES: DATA AND DERIVATIONS 133 50 Salinity 0‰

Csi (KJ/Kg/K)

45

Salinity 10‰

40

Salinity 1‰

35

Salinity 2‰

30

Salinity 4‰

25

Salinity 6‰

20

Salinity 8‰

15

Salinity 10‰

10 5 0

–2

–4

–6

–8

–10

–12

–14

–16

–18

–20

–22

–24

Temperature (° C)

Figure 3.25 Specific heat of sea ice as a function of ice temperature and salinity calculated using equation (3.57).

specific heat at ice temperature T is given by the following equation, which is valid between 0 > T > − 8 C, C si − T = mi − T C i − T + mb − T C b − T + Li − T

dmb − T dT (3.58)

where, mi − T and mb − T are the mass ratio of pure ice and brine in one gram of sea ice at temperature T, respectively; Ci − T and Cb − T are specific heat of pure ice and brine at temperature T, respectively; and Li − T is the latent heat of pure ice at temperature T. Tables of specific heat of sea ice for different salinities and temperatures are presented in Malmgren [1927], Schwerdtfeger [1963], Pounder [1965] and Ono [1967]. 3.6.4. Latent Heat of Sea Ice Latent heat is defined as the heat required for changing the phase of a unit mass of material at the same temperature. Since the transformation does not change the temperature, latent heat is considered energy in hidden form. The word “latent” is derived from the Latin word “latere,” which means “to hide.” In thermodynamics, the latent heat is the counterpart of sensible heat which, by definition, causes change of temperature of the medium as a result of heat influence. For example, when colder air blows over an ice sheet, heat is released from the ice by convection (especially if the wind is strong), hence lowers the ice surface temperature. That is sensible heat. Latent heat of fusion (or melting) is associated with phase change from liquid to solid (or vice versa). On the other hand, latent heat of vaporization (or condensation) is associated with phase change from liquid to vapor (and vice versa). Latent heat of sublimation is responsible

of changing phase from solid to vapor, bypassing the liquid phase. Specific latent heat of fusion, a term which is not used for ice, is the latent heat of fusion of a substance divided by the latent heat of fusion of water. For freshwater ice, the latent heat of fusion at 0 C is about 334 kJ/kg. This is the highest value of all common material. The latent heat of vaporization of freshwater at 100 C is about 2.260 kJ/kg. Therefore, much less energy is needed for vaporization and, conversely, steam releases a great deal of thermal energy when condensates. On the other hand, the latent heat of fusion of sea ice is 290 kJ/kg, which is less than that of freshwater ice. That is because sea ice contains brine (liquid phase). Therefore, seawater freezes faster than freshwater when both reach their freezing temperature. Seawater also heats up faster than freshwater. While only 1–4 kJ of heat is needed to raise the temperature of 1 kg of sea ice by one degree (i.e., the specific heat of ice), about 300 kJ (~150 times higher) is needed to change the phase of 1 kg from sea ice to saltwater. The phase change requires much more energy than raising or lowering the temperature of ice or water. The latent heat of sublimation of ice is about 2838 kJ/kg. This is almost an order of magnitude higher than the latent heat of fusion. Several expressions for calculations of Lsi have been developed. Perhaps the original one was suggested by Malmgren [1927], which determines Lsi as a function of the ratio of sea ice salinity to brine salinity, Lsi = Lpi 1−

S si Sb

(3.59)

where Lpi is latent heat of fusion of pure (freshwater) ice. For constant brine salinity Sb , the higher the salinity of

134

SEA ICE

sea ice the less Lsi. Phase change from solid ice to liquid water is not the only process associated with latent heat. That heat is required also to change the phase of brine composition in order to maintain the thermal equilibrium with the surrounding ice. This process is not accompanied with change of temperature of brine pockets or the surrounding sea ice. Equation (3.59) takes into consideration only ice and brine salinity. Ono [1967] developed an equation to calculate the latent heat of sea ice Lsi − T at a given temperature T, where the temperature appears explicitly in the expression, Lsi − T = Li − 0 + C w − T − C i − T T

(3.60)

where Li − 0 is the latent heat of pure ice at 0 C and Cw − T and Ci − T are the specific heat of freshwater and pure ice at temperature T. The dependence of Lsi on ice temperature is justified by the fact that sea ice has no fixed temperature for phase transition because the amount of brine varies according to the record of air temperature and meteorological conditions (wind and snow) under which freezing occurs. Yen, Cheng, Fukusako [1991] developed an equation to estimate Lsi as function of ice temperature Tsi and salinity Ssi. Lsi = 79 68 − 0 505 T si − 27 3 S si + 4311 5

S si T si

(3.61)

+ 0 8 S si T si − 0 009 T 2si This equation was used to produce the graph in Figure 3.26, from which the following observations can be drawn. Similar to the specific heat, latent heat of fusion of sea ice is very sensitive to temperature in the

350

0 1

300

2 4 6

Lsi ((kJ/kg)

250

8 10

200 150

12 100 50 0

0

–1

–2

–3

–4

–5

–6

–7

–8

–9

–10 –11

Temperature (°C)

Figure 3.26 Latent heat of sea ice as a function of ice temperature and salinity, calculated from equation (3.61).

near-melting range. It is also a strong function of salinity in this temperature range. This is the range when slush ice likely develops during the freezing season or surface melt develops by the end of the freezing season. It is difficult to parameterize Lsi under this condition. However, unlike the specific heat, the latent heat of sea ice decreases as the ice temperature or salinity increases. 3.7. DIELECTRIC PROPERTIES The notion of the dielectric constant, which determines the microwave emission and scattering of sea ice, is discussed here in some details to furnish a background for more discussions on the values of those properties presented in Chapter 12. The discussions can be useful for those who use the term dielectric constant (also called complex permittivity) and need to know more about its origin and how it is related to other physical properties of sea ice as well as the crystallographic structure. A model for calculating the two components of the complex dielectric constant of sea ice and a few results are presented. Unlike conductors that have free molecular charges or insulators that have no such charges, a dielectric substance contains a number of free molecules though not appreciable. When an electrical field is applied to a conductor the molecules move steadily forming conduction current. The intensity of the current is determined by the “conductivity” of the material. On the other hand, in response to an applied electric field, molecules in a dielectric can be displaced within molecular distances. The molecular displacement can be established in either one of the following two forms. If the molecules are originally non-polar, i.e., the center of charge of a molecule does not coincide with its center of mass (as in case of the water molecules), then the displacement induces polarization. If they are originally polar but randomly oriented, then the displacement could induce a partial alignment of molecules. In either case, the effect of an externally applied electric field is to leave the interior of the dielectric material uncharged and produce, instead, a “bound” charge on surfaces normal to the field. Oscillations of these charges are manifested as an electric current, called a displacement current, to flow through the dielectric. Strictly speaking, this current is determined by the permittivity of the material. Permittivity of a dielectric is the principle electrical property as conductivity is for conductors. At this point, it would be useful to briefly introduce Coulomb’s law and its relevance to permittivity. If Q1 and Q2 are two charges of the same polarity at a distance r from each other, then the force of repulsion F12 can be written as F 12 = k Q1 Q2 r2 r

(3.62)

SEA ICE PROPERTIES: DATA AND DERIVATIONS 135

where k is called coulomb constant and r is the unit vector. The constant can be described as: k = 1 4πϵ0

(3.63)

where, ϵ0 is the permittivity of free space. A dielectric medium can be ideal or non-ideal. Ideal dielectrics possess no free charges to establish any conduction current, i.e., their conductivity is zero. They are homogeneous, isotropic, and lossless. Non-ideal dielectrics possess a very small number of free charges. Therefore, their electrical conductivity is small but not zero. Permittivity is an important property in both cases. Pure ice is almost an ideal dielectric. Saline sea ice is a non-ideal dielectric because liquid brine contains free ions. Permittivity and conductivity of dielectrics, measured in units of Farad/m, are usually combined in a single parameter called the complex dielectric constant or the complex permittivity ϵ, defined as: ϵ = ϵ − jϵ

(3.64)

The real part is the permittivity and the imaginary part, which is a function of the electrical conductivity, is called the loss factor. Qualitatively speaking, permittivity determines the portion of energy that penetrates the material (the rest will be scattered off the surface), while electrical conductivity determines the portion of the penetrated energy that is lost as heat or scattering. High permittivity means less penetration of energy, hence more reflection/ scattering at the surface. High loss means more energy dissipation inside the material. The dielectric constant of ideal dielectrics is a real number while that of non-ideal dielectrics is a complex number. The reason for combining permittivity and loss into one complex number and also for the negative sign in equation (3.63) is clarified in the following mathematical formulation. Wave propagation in an ideal (i.e. lossless) dielectric is governed by the well-known “wave” equation [Lorrain, Corson, Lorrain, 1986]. For plane-polarized waves propagating in the Z-direction, the electric field vibration in the X-direction, Ex, is determined by the equation: ∂ 2 EX ∂ 2 EX = μϵ 2 2 ∂t ∂Z

(3.65)

where, ε is the permittivity and μ is the permeability of the material (defined as the magnetic intensity caused by a magnetic field of unit strength). Assuming harmonic time dependence of the incident electric field EX, with magnitude EX0 and frequency ω: EX t = Re EX0 e jωt

(3.66)

where, Re denotes the real part, equation (3.64) can then be rewritten as: ∂ 2 EX = − ωμϵEX ∂Z2

(3.67)

The magnetic field vibration is governed by a similar equation. The electromagnetic wave is therefore affected by ϵ and μ. The former represents the capacitive property and the latter represents the inductive property of the material in response to an applied magnetic field. The permeability of most materials is close to 1, which is the value of permeability of free space. Therefore, the electromagnetic wave propagation depends only on ϵ. The solution to equation (3.67) is introduced in several text books on electromagnetic wave theory [e.g., Lorrain, Corson, Lorrain, 1986]. It shows that both electric and magnetic vectors propagate with the same velocity and phase without attenuation. The permittivity is a real number in this case, i.e., ϵ is equivalent to ϵ in equation (3.64). In nonideal (lossy) dielectrics, the energy loss is manifested in the form of conduction currents and is represented in the wave propagation equation by a “diffusion” term (the last term in the RHS of the following equation) ∂ 2 EX ∂ 2 EX ∂EX (3.68) = μ ϵ + μσ 2 2 ∂t ∂t ∂Z where, σ is the electric conductivity of the material. Again, assuming harmonic time dependence of the electric field EX, equation (3.67) takes the form: ∂ 2 EX σ = − ω 2 μ ϵ − j EX 2 ω ∂Z

(3.69)

This form can be made equivalent to the wave equation in an ideal dielectric if ϵ in equation (3.66) is replaced by [ϵ − j(σ/ω)]. This is the rationale behind the definition of the complex dielectric constant equation (3.64). By comparing equations (3.64) and (3.69), it can be seen that the real part in the definition is the permittivity and the imaginary part is proportional to the conductivity, with an explicit dependence on the frequency of the electromagnetic field. The negative sign of the imaginary part has resulted from the second derivative of the exponential term in equation (3.66). Solution of equation (3.69) in cases of low-loss, high-loss, and general-lossy media can be found in Ulaby, Moore, Fung [1986]. It suffices to mention here that the solution for the general-lossy media is a decaying harmonic function, defined as: EX Z = EX0 e jω

μω 1 − j tan δ

1

2

(3.70)

The attenuation of this function (i.e., the exponent) depends on tanδ, which is defined as (ϵ /ϵ ) and is called the loss tangent. The distance over which the amplitude of the incident radiation decreases by a factor (1/e) is called the penetration depth δ. It is related to the loss tangent by the following equation [Evans, 1965]; δ=

λ 4π

ϵ 2

1+

ϵ ϵ

−1 2

2

−1

(3.71)

136

SEA ICE

where, λ is the wavelength of the incident radiation. Derivation of this equation is presented in section 7.2.4 and calculations of microwave C-band penetration depth in FYI and MYI are presented in section 9.5.

3.7.1. Dielectric Constant of Brine The dielectric constant of brine is a function of its temperature T and normality Nb. The normality is related to the brine salinity Sb by the following relation [Klein and Swift, 1977] (Sb can be obtained from equation (3.25)), N b = S b 1 707 × 10 − 2 + 1 205 × 10 − 5 S b + 4 058 ×

10 − 9 S 2b

(3.72)

σb ϵb0 − ϵb ∞ + 2πf τb 2πf ϵ0 1 + 2πf τb

τb0 T =

1 2π

1 1109 × 10 − 10 − 3 824 × 10 − 12 T + 6 938 × 10 − 14 T 2 − 5 096 × 10 − 16 T 3 (3.79)

The rest of the functions are given by the following expressions σ b 25, N b = N b

10 39 − 2 378N b + 0 683N 2b − 0 135N 3b + 1 01 × 10 − 2 N 4b

(3.80)

a1 N b = 1 0 − 0 255N b + 5 15 × 10 − 2 N 2b − 6 89N 3b (3.81)

This equation is valid for NaCl solutions of salinities smaller than 260‰. It is therefore applicable to sea ice. Brine salinity is governed by ice temperature as presented in equations (3.23) and (3.24). Equations to calculate the complex dielectric constant of brine were developed by Stogryn [1971]. Its real and imaginary terms are given by: ϵb0 − ϵb ∞ (3.73) ϵb = ϵb ∞ + 1 + 2πf τb 2 ϵb =

polynomial to the data reported by Grant, Buchanan, Cook [1957].

2

1 0 + 0 146 × 10 − 2 TNb − 4 89 × 10 − 2 Nb − 2 97 × 102 N2b + 5 64 × 10 − 3 N3b (3.82)

c1 Δ, N b = 1 0 − 1 96 × 10 − 2 Δ + 8 08 × 10 − 5 Δ2 − N bΔ

3 02 × 10 − 5 + 3 92 × 10 − 5 Δ + N b 1 75 × 10 − 5 − 6 58 × 10 − 6 (3.83)

(3.74)

where, ϵb0 = low-frequency (static) limit of ϵb ϵb∞= high-frequency (optical) limit of ϵb σ b= ionic conductivity of brine f = electromagnetic frequency (in GHz) τb = relaxation time of brine ϵ0 = permittivity of free space (=8.854×10−2 F/m) and ϵb∞ is independent of salinity [Stogryn, 1971]. Therefore, it is equal to the optical limit of water permittivity of 4.9. The terms ϵb0, τb, σ b are functions of T and Nb as given by the following expressions: ϵb0 T, N b = ϵb0 T, 0 a1 N b

(3.75)

τ b T, N b = τ b0 T, 0 b1 T, N b

(3.76)

σ b = T, N b = σ b 25, N b c1 Δ, N b

b1 T, Nb =

3.7.2. Dielectric Mixing Models The simplest approach to determine the dielectric constant of sea ice is by using a linear model that accounts for the volume fraction of the components of sea ice: the pure ice crystals (host material) and the inclusions (brine, air, or both). The shape of the inclusions is not accounted for. If ϵh and ϵi are the complex dielectric constants of the host and the inclusion i, which has volume fractions Vi, then the dielectric constant of the sea ice mixture ϵm is defined by the following equation [Ulaby, Moore, Fung, 1986]: n

(3.77)

where, Δ = 25 − T (in C) and ϵb0(T, 0) is the static dielectric constant of brine with zero normality, i.e., of pure water. This was determined by Klein and Swift [1977] using a regression fit for measurements conducted: ϵb0 T, 0 = 88 045 − 0 4147 T + 6 295 × 10 − 4 T 2 + 1 075 × 10 − 5 T 3 (3.78) Similarly, τ b(T, 0) is the relaxation time of pure water. This was obtained by Stogryn [1971] by fitting a

ϵm = ϵh +

V i ϵi − ϵh

(3.84)

i=1

The volume fraction of each component is obtained as shown in section 3.4. The dielectric constant of air can be assumed to be 1 and the dielectric constant of brine can be calculated using the method suggested by Stogryn and Desargent [1985]. This linear model will be referred to in the rest of this section as the L-model. It is simple but does not produce accurate results. A two-phase dielectric mixing model developed by Polder and Van Santen [1946] and modified later by de Loor [1968] is pursued in the following discussions and will be referred to as the PVD model. The model is based

SEA ICE PROPERTIES: DATA AND DERIVATIONS 137

on the assumption that the dielectric is composed of a host material (of dominant volume fraction) and another material regarded as an inclusion. Inclusions are assumed to be ellipsoidal of identical shape and size but randomly spaced within the host material. They can be oriented or non-oriented. The model incorporates also an assumption that the average size of the inclusion element is an order of magnitude less than the wavelength of the incident radiation. Therefore, only the overall volume of the inclusions, rather than the size of inclusion elements, appears in the formulation. The expression for the average dielectric constant ϵm of a two-phase dielectric mixture is:

𝜖h 𝜖* = 𝜖h

EZ Fictitious host ellipsoid

EX EY

𝜖i 𝜖h

(3.85)

where, Vi is the volume fraction of the inclusions (i.e., brine pockets or air bubbles in the case of first-year and multi-year ice, respectively), ϵh and ϵi are the complex dielectric constants of the host material (i.e., pure ice in this case) and the inclusions, respectively. Both are assumed to be isotropic (scalar quantity), which is a valid assumption for sea ice. In fact, ϵh of the pure ice is constant = 3.15; ϵ∗ is the effective dielectric constant for the region immediately surrounding the inclusion element. It can be anisotropic (vector quantity) if the inclusions have an anisotropic shape or a dominant orientation. The value of ϵ∗ is determined based on an assumption concerning the mutual interaction between inclusion elements. If the volume fraction of the inclusions is less than 0.1, the mutual interaction can be neglected and ϵ∗ may be made equal to ϵh. If the volume fraction is larger than 0.1, the mutual interaction can be accounted for by assuming that each inclusion element is surrounded by a mixture of the host material and inclusions. In this case ϵ∗ = ϵm (Figure 3.27). In equation (3.85), Au is the depolarization factor of the ellipsoid inclusion particle along its u-axis (u = a, b, or c). Their sum is equal to unity [Landau and Lifshitz, 1975]. The above model assumes that inclusions have an identical ellipsoidal shape but can be randomly spaced within the host material. They can also be oriented or non-oriented (Figure 3.28). Applications of the PVD model [equation (3.85)] for three commonly observed shapes of inclusions in sea ice: sphere, oriented needle, and randomly oriented needled are introduced in Shokr [1998] and reiterated here. It should be noted that Hallikainen [1977] also evaluated the applicability of the same PVD model by using four different assumptions regarding the shape of brine pockets. Results from using the parallel-needle assumption were found to be closest to the measurements presented in Hoekstra and Capillino [1971]. Stogryn [1987] discussed the tensor properties of the dielectric constant of sea ice. The author developed a set of self-consistent

Inclusion volume fraction > 0.1 𝜖* = 𝜖m

Figure 3.27 Assumptions regarding the dielectric constant of the host material in the immediate vicinity of the inclusions [Shokr, 1998 / IEEE].

Oriented inclusions non-oriented inclusions

V i ϵi − ϵh ϵm = ϵh + ϵi 1 + Au ∗ − 1 ϵ

𝜖i Inclusion volume fraction < 0.1

EZ EX EY

Figure 3.28 Idealization of inclusion shapes, elliptic (top), needle (middle), and spherical (bottom) with oriented and random distributions used in the dielectric mixing model [Shokr, 1998 / IEEE].

equations that account for the dielectric properties of a variety of ice types over a range of temperatures. A brief review of modeling and experimental investigation of the dielectric properties of natural and simulated sea ice is presented in Hallikainen and Winebrenner [1992]. An analytical study for obtaining bounds on dielectric

138

SEA ICE

constant of composite material, with an application to sea ice, is presented in Golden [1995].

1. Spherical Inclusions These inclusions approximate the shape of brine pockets within frazil, granular, and fine-grained (transition) columnar ice structures. It also suits the shape of air bubbles in MYI. In this case, Ax = Ay = Az = 1/3. The dielectric constant is an isotropic (scalar) quantity. Equation (3.85) can be rewritten as follows: ϵm = ϵh

ϵi − ϵh 1+ 3V i ϵi + 2 ϵh

for V i ≤ 0 1

(3.86)

ϵi − ϵh ϵi + 2ϵm

for V i > 0 1

(3.87)

or ϵm = ϵh + 3V i ϵm

The unknown parameter ϵm appears on both sides of equation (3.87). Hence, the solution for ϵm can be obtained from the following quadratic equation: 2ϵ2m + ϵm ϵi − 2ϵh − 3V i ϵi − ϵh − ϵi ϵh = 0

(3.88)

a. Solution for the Case Vi ≤ 0.1 Equation (3.86) can be rationalized using the definition given in equation (3.85) to produce the following expressions: ϵi − ϵh ϵi + 2ϵh + ϵi

ϵm = ϵh 1+ 3V i

2

ϵi + 2ϵh

ϵm =

9Vi ϵi ϵh ϵi + 2ϵh

2

+ ϵi

2

2

(3.89)

2

+ ϵi

2

(3.90)

b

a1 + ja2 = 2 + jo

b1 + jb2 = ϵi − 2ϵh − 3V i ϵi − j ϵi − 3 V i ϵi c

c1 + jc2 = − ϵh ϵi + jϵh ϵi

b2 − d 2 2a

(3.97)

where, d = d1 + jd2 is the square root of the discriminant complex term (b2 − 4ac). 2. Random Needle Inclusions In this case Ax= Ay = 0.5 and Az = 0. However, due to the randomness of the needle orientation, the term involving Au in the RHS of equation (3.85) is treated as being the arithmetic average from using the three orthogonal components of Au. Hence, the dielectric constant of the mixture remains isotropic. ϵm = ϵh +

Vi ϵi − ϵh 3

1 ϵi 1 + Au ∗ − 1 ϵ

u = a, b, c

(3.98)

The second term in the RHS is the average of the three components of Au. By considering ϵ∗ = ϵh when Vi ≤ 0.1, and ϵ∗= ϵm when Vi > 0.1, equation (3.97) takes the form: ϵm = ϵh +

ϵm = ϵh +

V i ϵi − ϵh 5ϵh + ϵi 3 ϵi + ϵh

for V i ≤ 0 1

(3.99)

V i ϵi − ϵh 5ϵm + ϵi 3 ϵi + ϵm

for V i > 0 1 (3.100)

a. Solution for the Case Vi ≤ 0.1 V i B ϵh + ϵi + Aϵi 3 ϵ +ϵ 2+ ϵ 2 i i h

ϵm = ϵh + ϵm =

(3.93) (3.94)

V i A ϵh + ϵi + Bϵi 3 ϵ +ϵ 2+ ϵ 2 i i h

(3.95)

(3.101)

(3.102)

where A = 2ϵi ϵi + 4ϵh ϵi B = ϵi − ϵi

2

ϵm = − b ±

(3.96)

ϵm =

(3.92)

The two solutions of the quadratic equation (3.91) are given by: b − 4ac 2a

− b1 + d 1 2a

(3.91)

where, a, b, and C are complex coefficients. Their expressions can be obtained by comparing equations (3.91) and (3.88), taking into consideration equation (3.85): a

ϵm =

and

b. Solution for the Case Vi > 0.1 Equation (3.87) can be written as: aϵ2m + bϵm + C = 0

The solution with the positive sign was found to be acceptable in the case of brine inclusions, while the solution with the negative sign was acceptable in the case of air inclusions. The alternate solution resulted in negative values of the dielectric constant. Therefore, the expressions used to calculate the real and imaginary parts of ϵm for FYI are:

2

− 5 ϵh

2

+ 4ϵh ϵi

(3.103) (3.104)

b. Solution for the Case Vi > 0.1 The solution represented by equations (3.95)–(3.97) can be applied by using

SEA ICE PROPERTIES: DATA AND DERIVATIONS 139

ϵmz = V i ϵi

the following coefficients of the quadratic equation derived from equation (3.100): a

a1 + ja2 = 3 + jo

+ j 3 ϵh − ϵi + 5V i ϵi − ϵh c1 + jc2

c

Again, in equation (3.111) the unknown parameter ϵmx appears on both sides of the equation. The equivalent quadratic equation is:

(3.105)

b1 + jb2 = 3 ϵi − ϵh − 5V i ϵi − ϵh

b

ϵ2mx + ϵi − ϵh 1− 2V i ϵmx − ϵh ϵi = 0

(3.106)

a. Solution for the Case Vi ≤ 0.1 equation (3.114),

(3.107)

where, c1 = V i ϵi ϵh − ϵi ϵh − ϵi

2

− ϵi

2

− 3 ϵi ϵh − ϵi ϵh

ϵmx = ϵh 1+

ϵi + ϵh

(3.109)

ϵmx =

a

ϵi − ϵh ϵi + ϵh

for V i ≤ 0 1 (3.110)

ϵmx = ϵh + 2V i ϵm

ϵi − ϵh ϵi + ϵmx

for V i > 0 1 (3.111)

ϵmz = ϵh + V i ϵi − ϵh

b

(3.116)

2

(3.117)

2

ϵi + ϵh + ϵi

a1 + ja2 = 1 + j

(3.118) (3.119)

c1 + jc2 = − ϵh ϵi + jϵh ϵi

(3.120)

Results from the above model, based on inputs from the physical model of brine volume and salinity (section 3.5.1), are presented in Shokr [1998]. Figures 3.29 and 3.30 show the results from using the assumptions of random needle and oriented needle inclusions, respectively. The permittivity from using the random needle assumption is almost constant, about 3.4 for temperatures

Note that in deriving equation (3.112) the term ϵ∗ disappears, since Az= 0. Consequently, the equation is valid for any value of Vi. By rationalizing equation (3.112) (3.113)

0

0

2

Loss 6 4

8

Permittivity 2 4 6 8 10 12

4V i ϵi ϵh

2

b1 + jb2 = 1− 2V i ϵi − ϵh − jϵi 1− 2V i c

(3.112)

ϵmz = ϵh + V i ϵi − ϵh

+ ϵi

2

b. Solution for the Case Vi > 0.1 The quadratic equation (3.115) can be solved as shown in equation (3.96) and (3.97) after replacing ϵm with ϵmx and using the following coefficients of the quadratic equation derived from equation (3.111):

3. Oriented Needle Inclusion By assuming the needle axis in the Z-direction, then Ax= Ay= 0.5 and Az = 0. The dielectric constant of the mixture becomes anisotropic. Nevertheless, ϵmx = ϵmy since Ax= Ay. In this case equation (3.85) can be rewritten as follows: ϵmx = ϵh 1+ 2V i

2

(3.115)

By rationalizing

ϵi − ϵh ϵi + ϵh = ϵi

(3.108) c2 = V i 2ϵi ϵi − ϵi ϵh − ϵi ϵh + 3 ϵi ϵh − ϵi ϵh

(3.114)

12

it Salin

y

24

4 –2 ) C 6 (° –1 ture ra

8

20

–8 pe m Te

4 –2 ) C 6 (° –1 ture ra

–8 pe m Te

0

4

16 (ppt)

0

4

8

12 ity li a S n

16 p (p t)

20

Figure 3.29 Calculated permittivity (left) and loss (right) of FYI in microwave C-band versus ice temperature and salinity for a constant density of 900 kg/m3, assuming randomly oriented brine inclusions [Shokr, 1998 / IEEE].

24

SEA ICE

8 10 12 4 6

4

6

Permittivity

8 10 12

functions of temperature. Above that temperature, brine volume fractions increase almost exponentially with temperature. This observation implies that the dielectric constant of FYI becomes mainly a function of ice salinity below, say, −12 C. To simulate the permittivity of MYI, it was calculated using the assumption of spherical air bubble inclusions. Results are presented in Figure 3.31. Permittivity varies almost linearly with density, but does not vary with temperature. This is because temperature has no effect on the air volume fraction. Results also show that permittivity is almost independent of the shape of air bubbles, which means that MYI is an isotropic dielectric. Temperature dependence of the complex dielectric constant and brine permittivity are shown in Figure 3.32 from calculations using the spherical brine pocket assumption. Both permittivity and loss increase sharply at temperatures above −5 C and −11 C for salinities of 10‰ and 20‰, respectively. These are the temperatures at which the brine volume fraction reaches 0.1. Recall that, at this value, the assumption in the dielectric mixing model regarding the mutual interaction between inclusions is

0 2

2 0 pe

ur e

8

it Salin

Te m

pe

ra t

ur e

)

pt y (p

(°

C )

4

0

it Salin

24

20

18

12 8

4

4

18

24

–2

)

4

C

–2

(°

12

20

6 –1

6 –1

ra t

–8

–8

Te m

)

pt y (p

Loss

0.4

12

0.6

0

0

0.0

4

0.2

Loss

Permittivity

below −8 C or salinities less than 4‰ (for constant density). Beyond these limits, the permittivity increases non-linearly at a relatively fast rate as temperature or salinity increases. Arcone, Gow, McGrew [1986] shows similar results. The loss factor is more sensitive to temperature and salinity, as indicated by the monotonic nonlinear increase throughout the examined range. Results from using the oriented needle assumption (Figure 3.30) reflect the anisotropic character of the complex dielectric constant in this case. Both permittivity and loss components calculated for the needle axis being parallel to the electric field are higher than those calculated for the axis perpendicular to the electric field. While permittivity is higher by at most a factor of two, loss is higher by almost an order of magnitude. This demonstrates the sensitivity of the loss to the anisotropic shape of brine pockets. The rapid increase in permittivity and loss at temperatures above −8.2 C can be explained in terms of the temperature dependence of volume fractions of the sea ice composition (Figure 3.18). For an ice density of 0.9 g/cm3 and salinity 10‰ at temperatures below −8.2 C, volume fractions of brine and pure ice are weak

8

140

4 0

18

)

(ppt

Te m

pe

ra t

ur e

(°

C

12

)

8

4

)

nity Sali

24

–2

12

8

4

(° C

–2

ra ur e

20

6 –1

6 –1

pe

–8

–8

Te m

4

ity Salin

18

20

)

(ppt

0

Figure 3.30 Calculated permittivity and loss of FYI in microwave C-band versus ice temperature and salinity for a constant density of 900 kg/m3, assuming oriented needle brine inclusions. Results are from using electric field parallel (right) and perpendicular (left) to the needles [Shokr, 1998 / IEEE].

24

3 2 0

1

Permittivity

4

5

SEA ICE PROPERTIES: DATA AND DERIVATIONS 141

–8 pe

ra t

6 –1

Te m

(°C

0.6

–2

ur e

4

)

0.5

0.7 –3 m ) m·c g ( y it

0.9

0.8

s

Den

Figure 3.31 Calculated permittivity of MYI in the C-band versus ice temperature and density for a constant salinity of 2‰, assuming a spherical air bubble shape [Shokr, 1998 / IEEE].

lce salinity 10‰

6.5

1.0

Permittivity

6.0

0.8

Loss

5.5

0.6

5.0 0.4

4.5 4.0 3.5 –26

–22

–18

–14

–10

–6

Temperature (°C)

(c)

6.5

5.5

0.6

5.0 0.4

4.5 4.0

0.2

3.0 –30

–26

–22

–18

–14

–10

–6

–2

Temperature (°C) 72

Brine salinity 80‰ Brine permittivity (f/m)

0.8

Permittivity Loss

6.0

60

50

67 62

40

57 30 52

Permittivity

20

10

1.0

3.5

0.0

–2

lce salinity 20‰

7.0

0.2

3.0 –30

1.2

Brine loss (f/m)

7.0

8.0 7.5

lce permittivity (f/m)

7.5

lce permittivity (f/m)

(b)

1.2

47

Loss –30

–26

–22

–18

–14

–10

–6

–2

42

Temperature (°C)

Figure 3.32 Temperature dependence of sea ice permittivity and loss for (a) ice salinity 10‰ and (b) 20‰, and (c) brine permittivity and loss. Ice density is assumed to be 0.9 g/c3. Permittivity and loss are calculated for the microwave C-band [Shokr, 1998 / IEEE].

0.0

lce loss (f/m)

8.0

lce loss (f/m)

(a)

142

SEA ICE

adjusted. The sharp increase in permittivity near freezing temperatures is accompanied with a sharp increase in brine volume and a decrease in brine salinity (Figure 3.19). At these temperatures, charged molecules of water, not free ions in liquid brine, give rise to a high permittivity. A peak in the loss factor is noticeable at about −22.9 C, similar to a peak in brine salinity (Figure 3.19). Below this temperature, a considerable mass of salt freezes in brine pockets and brine salinity decreases accordingly. This effect reverses the trend of loss below −22.9 C; i.e., it decreases as temperature continues to decrease (a result of brine pockets shrinkage). Two observations can be concluded upon comparing results from Figure 3.32 and 3.19. The first is the similarity between the temperature dependence of sea ice permittivity and brine volume. This implies that permittivity is influenced by the overall composition of the material, of which brine volume is the most important parameter in the case of FYI. The second is the similarity between temperature dependence of the loss factor and brine salinity when the ice temperature is low enough so that the brine volume fraction is less than 0.1. This implies that the dielectric loss is influenced by detailed properties of individual inclusion elements, i.e., brine salinity in this case. The loss depends also on the shape and characteristic dimensions of brine pockets, as will be shown later. The permittivity of brine decreases linearly as temperature decreases and stabilizes at a value around 15 F/m below −22 C (precipitating temperature of sodium chloride). The dielectric loss of brine has a peak at −13 C. Both brine permittivity and loss depend only on brine salinity, which is driven by ice temperature. Applications of the above PVD-based formulations are limited to the cases when the dimensions of the inclusions are at most an order of magnitude less than the wavelength of the propagating wave. That is because the model does not take into account scattering from inclusion elements. This assumption is appropriate for (a)

the purpose of studying the interaction of the C-band (wavelength 5.4 cm) with brine pockets since their major axis is typically a few millimeters in length [Weeks and Ackley, 1982]. For air bubbles in MYI, Shokr and Sinha [1994] have found that less than 5% of bubbles have diameters larger than 0.5 cm, which is about 0.1 the wavelength of the C-band. Assessment of the model based on comparison against field measurements of the complex dielectric constant is introduced in the following section. 3.7.3. Field Measurements of Dielectric Constant More field measurements of the dielectric constant of FYI and MYI were conducted in the radio bands [e.g., O’Sadnick et al., 2016] than the C-band, which is more relevant to satellite observations. During the first three weeks of May 1993 and 1995 the dielectric constant of FYI and MYI was measured in Lancaster Sound, Canadian Central Arctic, using a portable dielectric probe (PDP) operating in the C-band and consisting of an open-ended coaxial cable whose 5-cm diameter tip is brought in contact with the ice surface. The concept of the probe is introduced in Stuchly et al. [1982], a full description of the instrument and the equations to retrieve the complex dielectric constant from the measured reflection coefficient are presented in Brundfeldt [1987]. Evaluation of the instruments’ accuracy is established in Jackson [1990] and it is concluded that permittivity could be measured at 3% accuracy, while loss was overestimated by about 15%. It should be noticed, however, that the dielectric loss is too low to be measured accurately. Using this instrument, measurements were conducted on samples from ice cores by the author of this book (M. Shokr). The instrument has two conductor cables, inner and outer. During operation, an electric field emerges from the end of the inner conductor and bridges into the end of the outer conductor, forming a non-propagating hemisphere-shaped field whose diameter is approximately 1.9 cm (Figure 3.33). The discontinuity of the impedance between the two conductors, caused by the ice dielectric,

Vertical cut from an ice core

Waveguide of (b) standing wave

Probe tip

Probe tip

Brine pockets

Horizontal cut from an ice core

Figure 3.33 Dielectric probe tip and the waveguide of the radiated electric field applied against vertical and horizontal cuts from a cylindrical ice core. The probe axis is (a) perpendicular and (b) parallel to the ice core axis. Measurements using the two configurations were averaged to obtain a single value at each depth.

SEA ICE PROPERTIES: DATA AND DERIVATIONS 143

leads to reflection of the traveling electromagnetic waves at the interface. The reflection coefficient is measured in terms of magnitude and phase, and then converted into the real and imaginary parts of the complex dielectric constant. Since brine is usually entrapped into vertically oriented-needle-shaped pockets, the dielectric constant of a sea-ice mixture is anisotropic. This anisotropy could not be measured using the above-mentioned instrument because there was no polarization definition for the standing wave originating from the probe tip. Instead, the average dielectric constant from two sets of measurements, conducted at the same depth, with the probe axis being parallel and perpendicular to the ice core axis (as shown in Figure 3.33) was obtained. Each set included 14 independent measurements distributed over the cross section of a sea ice core at the given depth. Outlier data points were rejected before averaging. Depth profiles of permittivity and loss using the abovementioned instrument are presented in Shokr [1998] (not included here). The author found that the decrease in permittivity and loss with depth replicated a similar trend shown by brine volume and salinity profiles. This was

more evident in the case of frazil ice. The permittivity values shown in the figure are in agreement with other published data in Vant, Ramseier, Makros [1978] and Arcone, Gow, McGrew [1986]. The dielectric loss values, however, are higher than other published data. For example, Vant, Ramseier, Makros [1978] reported loss values around 0.2 at temperature −14.5 C and salinity 10.5‰. Arcone, Gow, McGrew [1986] reported values around 0.1 but for less salinity, about 5‰. Nevertheless, the model results presented in Figures 3.29 and 3.30 support the possibility of higher loss values but only when brine pockets become parallel to the applied electric field. The effect of the ice crystalline structure on the measured permittivity and loss factor is revealed in Figure 3.34. The figure is a plot of measurements averaged over the top 0.15 m from 13 FYI cores obtained from Lancaster Sound, Canadian Central Arctic in May 1993. The nomenclature of the crystallographic structure is as follows: oriented coarse columnar crystals (O/C-CLMR), oriented fine columnar crystals (O/F-CLMR), random coarse columnar crystals (R/C-CLMR), snow ice (SNOW-I), oriented frazil crystals (O-FRZL), and

O/C - CLMR

O/F - CLMR

R/C - CLMR

Snow - I

O - FRZL

R - FRZL 1.3

4.3

Avg. permittivity (f.m–1)

4.2

Avg. loss (f.m–1)

1.2 1.1

4.1 4.0

1.0

3.9

0.9

3.8

0.8

Average loss (f.m–1)

Average permittivity (f.m–1)

4.4

3.7 0.7 6

7

8

9

5.9 ‰

8.0 ‰

8.2 ‰

9.6 ‰

8.9 ‰

10.5 ‰

10.9 ‰

10

11

12

13

0.6

14 cm

5

17.6 ‰

4

14.1 ‰

3

12.1 ‰

2

10.9 ‰

1

5.5 ‰

3.5

5.8 ‰

3.6

Figure 3.34 Variation of average permittivity and loss within the top 0.15 m of FYI cores with different crystallographic structures. Data were obtained from Lancaster Sound, Central Arctic in May 1993. The numbers attached to the core at the bottom are the average salinity in the top 6 cm. Nomenclature of ice crystallographic types (shown at the top) is explained in the text (data and plot are generated by M. Shokr).

144

SEA ICE

random frazil crystals (R-FRZL). The average salinity from the top 6 cm is attached to each core at the bottom of the graph. The remarkably high permittivity and loss factor measured from the random frazil cores is a manifestation of the high salinity associated with this ice type (the salinity of core #13 is 17.6‰ and the corresponding permittivity and loss are 4.37 and 1.23 F/m, respectively). Except for the random frazil, crystalline structure does not seem to be correlated with the complex dielectric constant. Permittivity and loss have small values from the oriented coarse columnar ice (cores 1 and 2). The significantly low values of loss from the oriented frazil ice core with the top 2 cm covered with snow ice (core # 11) could not be explained. Detailed measurements in the top 6 cm of this core (not shown in Figure 3.34) reveal salinity values around 14 ‰. The gas volume fraction in the top

2 cm is 4.3%, which does not explain the low permittivity and loss. Comparison of results from the PVD and the linear model (L-model) of dielectric constant against measurements obtained from the above-described field experiments are presented in Figure. 3.35 for the cases of frazil and columnar ice. For frazil ice, it appears that the PVD model results using the random needle assumption fit the measured data best. Neither the spherical shape assumption of brine pockets in the PVD model nor the L-model model is appropriate for frazil ice. This confirms the reliability of the results because the shape of brine pockets is indeed closer to the needle form. The linear model overestimates both permittivity and loss significantly. For FY columnar ice, results from using the oriented needle assumption in the PVD model appear

6.0

6.0 PVD model–spherical brine pockets PVD model–O-needle brine pockets PVD model–R-needle brine pockets Linear model

5.2

5.0 Computed loss

Computed permittivity

5.6

Frazil ice

4.8

PVD model–spherical brine pockets PVD model–O-needle brine pockets PVD model–R-needle brine pockets Linear model

4.4 4.0

4.0 Frazil ice 3.0 2.0 1.0

3.6 3.2 3.2

3.4

3.6

3.8

4.0

4.2

0.0 0.0

4.4

0.2

0.4

0.6

Measured permittivity

1.0

1.2

1.4

1.6

1.8

6.0

6.0 PVD model–spherical brine pockets PVD model–O-needle brine pockets PVD model–R-needle brine pockets Linear model

5.2

PVD model–spherical brine pockets PVD model–O-needle brine pockets PVD model–R-needle brine pockets Linear model

5.0 Computed loss

5.6 Computed permittivity

0.8

Measured loss

Columnar ice 4.8 4.4 4.0

4.0 Columnar ice 3.0 2.0 1.0

3.6

0.0

3.2 3.2

3.4

3.6

3.8

4.0

Measured permittivity

4.2

4.4

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Measured loss

Figure 3.35 Measured versus computed dielectric constant (permittivity and loss) of FYI in the C-band (for frazil and columnar types). Results from using the PVD model with three assumptions of brine pocket shape are shown along with results from using the linear model. Oriented-needle brine pockets are assumed to be vertical (M. Shokr).

1.8

SEA ICE PROPERTIES: DATA AND DERIVATIONS 145 4.0

4.0

PVD model – spherical air bubbles

3.6

PVD model – R- needle air bubbles

Computed loss

Computed permittivity

3.6

PVD model – spherical brine pockets PVD model – spherical air bubbles

PVD model – O- needle air bubbles

3.2

2.8

2.4

L-model (brine + air)

3.2 2.8

2.4 MY melt pond ice

MY hummock ice 2.0 2.0

2.2

2.4

2.6

2.8

3.0

3.2

3.4

3.6

3.8

2.0 2.0

4.0

2.2

2.4

Measured permittivity

2.6

2.8 3.0 3.2 Measured loss

3.4

3.6

3.8 4.0

Figure 3.36 Measured versus computed dielectric constant of MYI in the C-band. Results from using the PVD model with three assumptions of air bubble shape for hummock ice and two assumptions (spherical brine pockets or air bubble) for melt pond ice. Results from the linear model [equation (3.84)] are also shown for melt pond ice (M. Shokr).

data presented in Figure 3.36 show that using the brine pocket or spherical bubble inclusion assumption for this ice type in the PVD model produces unsatisfactory results. In fact, the linear model is fairly successful in reproducing the loss for melt pond ice. This model seems to be successful when brine volume fraction is sufficiently small. Variation of permittivity of MYI (combining hummock and melt pond data) versus density is shown in Figure 3.37. The data were obtained from the PVD 3.8 3.6 Measured permittivity

to be slightly better than using the random needle assumption. This applies to both permittivity and loss (for permittivity, the root mean square of the difference between measured and calculated values is 0.18 F/m and 0.29 F/m, respectively). It should be noted that accurate evaluation of using the oriented needle assumption requires measurements of the tensor components of the dielectric constant. The component parallel to the electric field direction can be twice as high as the perpendicular component, especially at temperatures above −6 C. Taking a simple arithmetic average of the two components is not the most accurate way of combining them into a single observation. The two apparent clusters in the columnar ice data in Figure 3.35 represent data points obtained from two different years (May 1992 and 1993). Obviously the ice was more saline for the cluster with higher permittivity and loss. Permittivity of multi-year hummock ice (Figure 3.36) shows that regardless of the air bubble shape, the PVD models are equally successful in approximating the measured values. This is mainly due to the fact that permittivity of air (1.0 F/m) does not significantly contrast with that of pure ice (3.15 F/m). Moreover, the composite permittivity is isotropic, since the two orthogonal components from the oriented needle assumption are found to be nearly equal. This eliminates the need for an appropriate assumption to model the highly asymmetric and variant air bubbles in hummock ice. Permittivity of hummock ice is relatively easy to model because this ice is strictly a two-phase material (pure ice and air) and the permittivity of air is independent of temperature. Melt pond ice

3.4

Hummock ice Melt pond ice Gudmandsen’s equation

3.2 3.0 2.8 2.6 2.4 2.2 2.0 0.5

0.6

0.7 0.8 Density (g.cm–3)

0.9

1.0

Figure 3.37 Variation of permittivity with density of multi-year ice. The second-order best fit is shown by the solid line and the non-linear equation (3.120) is shown by the dotted line. Melt pond ice has higher density than hummock ice [Shokr, 1998 / IEEE].

146

SEA ICE

mixing as well as measurements from the same dataset described above. While the figure does not separate the data from hummocks and melt pond ice, data points from melt pond ice are clustered around 0.9 gm/cm3 density. The solid line represents the best second-order polynomial fit: ϵ = 4 63 + 8 4 D + 7 56D2

(3.121)

where, D is ice density in gm/cm3. Results from an empirical equation developed by Gudmandsen [1971] are also shown. This equation is written as: ϵ = 1 + 0 85D

2

(3.122)

3.8. REFERENCES Alexandrov, V. et al. (2010) The relation between sea ice thickness and freeboard in the Arctic, The Cryosphere Discussions, 4, pp. 641–661. Anderson, D.L. (1958) A model for determining sea ice properties, Proceedings of Arctic Sea Ice Conference US, National Academy of Sciences, National research Council Publication 598, Washington, D.C., pp. 148–152. Arcone, S. A., Gow, A.J. and McGrew, S. (1986) Structural and dielectric properties at 4.8 and 9.5 GHz of saline ice, Journal of Geophysical Research, 91(C12), pp. 14,281–14,303. Assur, A. (1958) Composition of sea ice and its tensile strength, In: Arctic Sea Ice. US National Academy of Sciences, Washington, D.C., National Research Council, Publ. 598, pp. 106–138. Backstrom, L.G.E. and Eicken, H. (2006) Capacitance probe measurements of brine volume and bulk salinity in first-year sea ice, Cold Regions Science and Technology, 46(3), pp. 167–180. Bader, H. (1962) The physics and mechanics of snow as a material, Cold Region Science and Engineering Monograph No. IIB, US Army Cold Regions Research and Engineering Laboratory. Bilello, M.A. (1980) Decay of fast sea ice in Canada and Alaska, In: Sea Ice Processes and Models, International Association of Hydrological Sciences Publication 124, pp. 313–326. Bitz, C.M. and Lipscomb, W.H. (1999) An energy-conserving thermodynamic model of sea ice, Journal of Geophysical Research, 104, pp. 15,669–16,677. Boinovich, L.B. and Emelyanenko, A.M. (2015) Experimental determination of the surface energy of polycrystalline ice, Doklady Physical Chemistry, 459 (Part 2), pp. 198–202. Brundfeldt, D.R. (1987) Theory and design of a field dielectric measurement system, IEEE Geoscience and Remote Sensing Symposium Digest, 1, IEEE Cat.87CH2434-9, pp. 559–563. Cheng, B. (2002) On the modelling of sea ice thermodynamics and air-ice coupling in the Bohai Sea and the Baltic Sea, Finnish Institute of Marine Research, Tech. Rep. 5, Helsinki. Cox, G.F.N. and Weeks, W.F. (1974) Salinity variation in sea ice, Journal of Glaciology, 13, pp. 109–120.

Cox, G.F.N. and Weeks, W.F. (1975) Brine drainage and initial salt entrapment in sodium chloride ice, US Army Cold Region Research and Engineering Laboratory, (CRREL) Research Report #345, Hanover, NH. Cox, G.F.N. and Weeks, W.F. (1983) Equations for determining the gas and brine volumes in sea ice samples, Journal of Glaciology, 29, pp. 306–316. Cox, G.F.N. and Weeks, W.F. (1986) Changes in salinity and porosity of sea ice samples during sampling and storage, Journal of Glaciology, 32, pp. 371–375. de Loor, G.P. (1968) Dielectric properties of heterogeneous mixtures containing water, Journal of Microwave Power, 3, pp. 67–73. Ebert, E.E. and Curry, J.A. (1993) An Intermediate onedimensional thermodynamic sea ice model for investigating ice-atmosphere interaction, Journal of Geophysical Research, 98(C6), pp. 10,085–10,109. Eide L.I. and Martin, S. (1975) The formation of brine drainage features in young sea ice, Journal of Glaciology, 14(70), pp. 137–154. Evans, S. (1965) Dielectric properties of ice and snow—A review, Journal of Glaciology, 5(42), pp. 773–791. Flato, G.M. and Brown, R.D. (1996) Variability and climate sensitivity of landfast Arctic sea ice, Journal of Geophysical Research, 101(C10), pp. 25,767–25,777. Frankenstein, G. and Garner, R. (1967) Equations for determining the brine volume of sea ice from 0.5 C to −22.9 C, Journal of Glaciology, 6(48), pp. 943–944. Fukusako, S. (1990) Thermophysical properties of ice, snow and sea ice, International Journal of Thermophysics, 11(2), pp. 353–373. Golden, K. (1995) Bounds on the complex permittivity of sea ice, Journal of Geophysical Research, 100(C7), pp. 13,699–13,711. Gough, A.J. et al. (2012) Sea ice salinity and structure: A winter time series of salinity and its distribution, Journal of Geophysical Research, 117(C03008). Grant, E., Buchanan, T. and Cook, T. (1957) Dielectric behaviour of water at microwave frequencies, Journal of Chemical Physics, 26, pp. 156–161. Grenfell, T.C. and Maykut, G.A. (1977) The optical properties of ice and snow in the Arctic Basin, Journal of Glaciology, 18(80), pp. 445–463. Gudmandsen, P. (1971) Electromagnetic probing of ice, In: Wait, J.R., Ed. Electromagnetic Probing in Geophysics, Boulder, CO: Golem, pp. 321–338. Hallikainen, M.T. (1977) Dielectric properties of NaCl ice at 16 GHz, Report S-107, Helsinki University of Technology, Radio Laboratory, Espoo, Finland, 37 pp. Hallikainen, M. and Winebrenner, D.P. (1992) The physical basis for sea ice remote sensing, In: Caresy, F., ed., Microwave Remote Sensing of Sea Ice, Washington, D.C.: American Geophysical Union. Hoekstra, P. and Capillino, P. (1971) Dielectric properties of sea and sodium chloride ice at UHF and microwave frequencies, Journal of Geophysical Research, 76, pp. 4922–4931. Hunke, E.C. et al. (2015) CICE: The Los Alamos sea ice model documentation and software user’s manual version 5.1, LA-CC06-012, Los Alamos, NM, Los Alamos National Laboratory.

SEA ICE PROPERTIES: DATA AND DERIVATIONS 147 Ingham, M., Pringle, D. and Eicken, H. (2008) Cross-borehole resistivity tomography of sea ice, Cold Region Science Technology, 52(3), pp. 263–277. Jackson, T.J. (1990) Laboratory evaluation of a field-portable dielectric soil moisture probe, IEEE Transactions of Geoscience and Remote Sensing, 28, pp. 241–245. Klein, L.A. and Swift, C.T. (1977) An improved model for the dielectric constant of sea water at microwave frequencies, IEEE Transactions on Antennas and Propagation, 25, pp. 104–111. Kovacs, A. (1996) Sea ice: Part I bulk salinity versus ice floe thickness, Cold Regions Research and Engineering Laboratory (CRREL) Technology Report, 96-7. Lecomte, O. et al. (2013) On the formulation of snow thermal conductivity in large-scale sea ice models, Journal of Advances in Modeling Earth Systems, 5(3), pp. 542–557. Landau, L.D. and Lifshitz, E.M. (1975) Electrodynamics of Continuous Media, Oxford, U.K.: Pergamon. Levy, F.L. (1982) The thermal conductivity of commercial brines and seawater in the freezing range, International Journal of Refrigeration, 5(3), pp. 155–159. Lorrain, P., Corson, D.R. and Lorrain, F. (1986) Electromagnetic fields and waves, New York: W.H. Freeman and Company. Malmgren, F. (1927) On the properties of sea ice. The Norwegian North Polar Expedition with the “Maud” 1918–1925, Scientific Results, 1(5). Maykut, G.A. (1978) Energy exchange over young sea ice in the Central Arctic, Journal of Geophysical Research, 83(C7), pp. 3646–3658. Maykut, G.A. and Untersteiner, N. (1971) Some results from a time-dependent thermodynamic model of sea ice, Journal of Geophysical Research, 76, pp. 1550–1575. McGuinness, M.J. et al. (1998) Nonlinear thermal transport and brine convection in first year sea ice, Annals of Glaciology, 27, pp. 471–476. Mellor, M. (1977) Engineering properties of snow, Journal of Glaciology, 19(8), pp. 15–66. Nakawo, M. and Sinha, N.K. (1981) Growth rate and salinity profile of first-year sea ice in the High Arctic, Journal of Glaciology, 27, pp. 315–330. Nakawo, M. and Sinha, N.K. (1984) A note on the brine layer spacing of first-year ice, Atmosphere-Ocean, 22(2), pp. 193–206. Notz, D. and Worster, M.G. (2008) In situ measurements of the evolution of young sea ice, Journal of Geophysical Research, 113(C03001). Available from: doi:10.1029/2007JC004333. Notz, D. and Worster, M.G. (2009) Desalination processes of sea ice revisited, Journal of Geophysical Research, 114(C05006). Available from: doi:10.1029/2008JC004885. Notz, D. and Worster, M.G. (2006) A one dimensional enthalpy model of sea ice, Annals of Glaciology, 44, pp. 123–128. Notz, D., Wettlaufer, J.S. and Worster, M.G. (2005) A nondestructive method for measuring the salinity and solid fraction of growing sea ice in situ, Journal of Glaciology, 51(172), pp. 159–166. O’Sadnick, M. et al. (2016) In situ field measurements of the temporal evolution of low-frequency sea-ice dielectric

properties in relation to temperature, salinity, and microstructure, The Cryosphere, 10(6), pp. 2923–2940. Ono, N. (1967) Specific heat and heat of fusion of sea ice, In: Oura, H., ed. Physics of Snow and Ice, Vol. 1, Sapporo, Japan: Institute of Low Temperature Science, Hokkaido University, pp. 599–610. Perovich, D.K. and Elder, B.C. (2001) Temporal evolution of Arctic sea ice temperature, Annals of Glaciology, 33, pp. 207–211. Perovich, D.K., Elder, B.C. and Richter-Menge, J.A. (1997) Observations of the annual cycle of sea ice temperature and mass balance, Geophysical Research Letters, 24(5), pp. 555–558. Perovich, D.K. et al. (1999) Year on ice gives climate insights, Eos, Transactions American Geophysical Union, 80(41), p. 481, pp. 485–486. Petrich, C. and Eicken, H. (2009) Growth, structure and properties of sea ice, In: Thomas, D.N. and Dieckmann, G.S., eds. Sea ice, 2nd ed. Wiley-Blackwell, pp. 22–77. Polder D. and Van Santen, J.H. (1946) The effective permeability of mixtures of solids, Physics, 12, pp. 257–271. Pounder, E.R. (1965) The Physics of Ice. Oxford: Pergamon Press, 160p. Pratt, A.W. (1969) Heat transmission in low conductivity materials, In: Nye, R.P., ed. Thermal conductivity, Academic Press, pp. 301–405. Pringle, D.J., Trodahl, H.J. and Haskell, T.J. (2006) Direct measurements of sea ice conductivity: No surface reduction, Journal of Geophysical Research, 111(C05020). Available from: doi:10:1029/2005JC002990. Pringle, D.J. et al. (2007) Thermal conductivity of landfast Antarctic and Arctic sea ice, Journal of Geophysical Research, 112(C04017). Available from: doi:10.1029/2006JC003641. Ryvlin, A.I. (1974) Method of forecasting flexural strength of an ice cover, Problems of the Arctic and Antarctic, 45, pp. 79–86 (in Russian). Satyawali, P.K. and Singh, A.K. (2008) Dependence of thermal conductivity of snow on microstructure, Journal of Earth System Science, 117(4), August 2008, pp. 465–475. Schwerdtfeger, P. (1963) The thermal properties of sea ice, Journal of Glaciology, 4(36), pp. 789–807. Semtner, A.J. (1976) A model for the thermodynamic growth of sea ice in numerical investigation of climate, Journal of Physical Oceanography, 6, pp. 379–389. Sharqawy, M. H., Leinhard, J.H. and Zubair, S.M. (2010) Thermophysical properties of seawater: A review of existing correlations and data, Desalination and Water Treatment, 16, pp. 354–380. Shokr, M. (1998) Field observations and model calculations of dielectric properties of Arctic sea ice in the microwave C-band, IEEE Transactions on Geoscience and Remote Sensing, 36(2), pp. 463–478. Shokr, M. and Dabboor, M. (2020) Observations of SAR polarimetric parameters of lake and fast sea ice during the early growth phase, Remote Sensing of Environment, 247, 111910. Shokr, M. and Sinha, N.K. (1994) Arctic sea ice microstructure observations relevant to microwave scattering, Arctic, 47(3), pp. 265–279.

148

SEA ICE

Shokr, M. and Barber, D.G. (1994) Temporal evolution of physical and dielectric properties of sea ice and snow during the early melt season: observations from SIMS’90 experiment, Journal of Glaciology, 40(134), pp. 16–30. Sinha, N.K. (1983) Does the strength of ice depend on grain size at high temperatures? Scripta Metallurgica, 17, pp. 1269–1273. Sinha, N.K. (1984) Uniaxial compressive strength of first-year and multi-year sea ice, Canadian Journal of Civil Engineering, 11(1), pp. 82–91. Sinha, N.K. and Nakawo, N. (1981) Growth of first-year ice in Eclipse Sound, Baffin Island, Canada, Canadian Geotechnical Journal, 18, pp.17–23. Slack, G.A. (1980) Thermal conductivity of ice, Physical Review B, 22(6), pp. 3065–3071. Smedsrud, L.H. and Ragnheid Skogseth, R. (2006) Field measurements of Arctic grease ice properties and processes, Cold Regions Science and Technology, 44, pp. 171–183. Stogryn, A. (1971) Equations for calculating the dielectric constant of saline water, IEEE Transactions of Microwave Theory and Techniques, MIT-19, pp. 733–736. Stogryn, A. (1987) An analysis of the tensor dielectric constant of sea ice at microwave frequencies, IEEE Transactions of Geoscience and Remote Sensing, GE-25, pp. 147–158. Stogryn, A. and Desargent, G.J. (1985) The dielectric properties of brine in sea ice at microwave frequencies, IEEE Transactions on Antennas and Propagation, AP-33, pp. 523–532. Stuchly, M.A. et al. (1982) Equivalent circuit of an open-ended coaxial line in a lossy dielectric, IEEE Transactions of Instrumentation and Measurement, IM-31, pp. 116–118. Sturm, M., Holmgren, J., Perovich, D.K. (2002) Thermal conductivity and heat transfer through the snow on the ice of the Beaufort Sea, Journal of Geophysical Research, 107(C10), 8043. Available from: doi: 10.1029/2000JC000409. Sturm M. et al. (1997) The thermal conductivity of seasonal snow, Journal of Glaciology, 43(143), pp. 26–41. Timco, G.W. and Frederking, R.M.W. (1996) A review of sea ice density, Cold Regions Science and Technology, 24(1), pp. 1–6.

Trodahl, H.J. et al. (2000) Heat transport in McMurdo Sound first-year fast ice, Journal of Geophysical Research, 105(C5), pp. 11,347–11,358. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1986) Microwave remote sensing: active and passive, From theory to applications, Vol. III, Norwood, MA: Artech House Inc., p. 2058. Untersteiner, N. (1961) On the mass and heat budget of Arctic sea ice, Archiv für Meteorologie, Geophysik und Bioklimatologie, Series A(12), pp.151–182. Untersteiner, N. (1964) Calculations of temperature regimes and heat budget of sea ice in the Central Arctic, Journal of Geophysical Research, 69(22), pp. 4755–4766. Vancoppenolle, M. et al. (2009) Simulating the mass balance and salinity of Arctic and Antarctic sea ice. 1. Model description and validation, Ocean Modelling, 27(1–2), pp. 33–53. Vant, M.R., Ramseier, R.O. and Makros, V. (1978) The complex dielectric constant of sea ice at frequencies in the range 0.1–40 GHz, Journal of Applied Physics, 49(3), pp. 1264–1280. Wadhams, P. (2000) Ice in the ocean, London, UK: Gordon and Breach Science Publisher, 351 pp. Weeks, W.F. (1976) Sea ice properties and geometry, AIDJEX Bulletin, 34, pp. 137–172. Weeks, W.F. (2010) On sea ice, Fairbanks, Alaska, USA: University of Alaska Press, ISBN-13: 978-1-60223-079-8. Weeks, W.F. and Ackley, S.F. (1982) The growth, structure, and properties of sea ice, Cold Regions Research and Engineering Laboratory (CRREL), Monograph, Vol. 82–1, Hanover, NH, USA. Winton, M. (2000) A reformulated three-layered sea ice model, Journal of Atmospheric and Oceanic Technology, 7, pp. 525–531. Yen, Y.C. (1981) Review of thermal properties of snow, ice and sea ice, Cold Regions Research and Engineering Laboratory (CRREL) Report 81-10, Hanover, NH, 26 pp. Yen, Y.C., Cheng, K.C. and Fukusako, S. (1991) Review of intrinsic thermophysical properties of snow, ice, sea ice, and frost, In: Zarling, J.P. and Faussett, S.L., eds. Proceedings 3rd International Symposium on Cold Regions Heat Transfer, Fairbanks, AK, June 11–14, pp. 187–218.

4 Laboratory Techniques for Revealing the Structure of Polycrystalline Ice

4.1 Relevant Optical Properties .................................................... 151 4.1.1 Polarized Light ............................................................ 151 4.1.2 Birefringence or Double Refraction of Ordinary (Ih) Ice...................................................153 4.1.3 Optical Retardation ..................................................... 155 4.1.4 Interference Colors for White Light ............................ 157 4.2 Ice Thin Sectioning Techniques .............................................. 158 4.2.1 Hot-plate Techniques for Thin Sectioning of Ice ......... 159 4.2.2 Double-Microtoming Technique for Thin Sectioning of Ice............................................................................ 159 4.2.3 Double-Microtoming Technique for Thin Sectioning of Snow........................................................................ 161 4.2.4 Precautions for Thin Sectioning by DMT ................... 163 4.2.5 Optimum Thickness for Thin Sections of Ice and Snow ........................................................163 4.3 Viewing and Photographing Ice Thin Sections ....................... 164

4.4 Advanced Techniques for Revealing Fine Crystallographic Microstructural Features ........................................................ 173 4.4.1 Sublimation of Ice and Sublimation Pits ..................... 173 4.4.2 Etching Processes......................................................... 176 4.4.2.1 Thermal Etching of Microtomed Ice Surfaces....................................................... 179 4.4.2.2 Chemical Etching and Replicating Ice Surfaces ........................................................ 183 4.5 References............................................................................... 188

Unlike most natural materials, snow and ice exists in nature at extremely high thermal states. This is true because ice sheets float on their own melt. Their crystallographic structure and physical properties, hence structure–property relationships, are subjected to continuous changes as environmental conditions change. The word, “structure” encompasses macro- and microstructure of the material at the surface, near surface and the bulk of snow and ice bodies. In the context of this book, the structure–property relationship is narrowed down to the structure–electromagnetic wave relationship although links to thermal and mechanical properties of ice are only hinted to. Interpretation of remotely sensed images of ice require a broad understanding of the structural aspects of the observed material that affect the emission, transmission, and scattering of electromagnetic waves. Structure–property relationships are also related to engineering problems of ice. Although this is important for ice physicists and remote sensing users, it is outside the scope of this book.

Laboratory work on natural snow and ice involves measurements of a few parameters such as salinity, density, thermal and electrical properties, chemical composition, and crystallographic structure. This chapter presents techniques to reveal detailed features of the polycrystalline structure of sea ice. Techniques for in situ measurements of temperature, salinity and density profiles, or estimation of optical, thermal, electrical and radiative properties of sea ice are not covered here. The reader may refer to other sources, notably Eicken and Salganek [2010]. Information in this chapter include advanced experimental techniques to examine crystallographic structure of sea ice at macro and micro (molecular) scales in terms of preparing ice specimen, visualization and photographing the structures. This should be of interest to readers who plan to pursue this kind of experimental endeavors in future. The techniques addressed here were developed 40–50 years ago when studies of sea ice physics were flourishing as Arctic sea ice research was a priority during the

4.3.1 4.3.2 4.3.3 4.3.4

Laboratory and Hand-Held Polariscope...................... 165 Cross-Polarized versus Parallel-Polarized Light Viewing .............................................................. 168 Scattered Light and Combined Cross-Polarized/ Scattered Light Viewing............................................... 169 Circularly Polarized Light and Rapid Crystallographic Analysis ............................................ 172

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 149

150

SEA ICE

Cold War era. They were used to develop wealth of information that later became classical knowledge about sea ice crystallography. The information was initially driven by an interest in developing better understanding of the mechanical behavior of sea ice, and later by interest in using remote sensing data for better retrieval of sea ice information. While this kind of work has suffered withdrawal recently, the documentation of such advanced techniques is included here to preserve the knowledge and serve the purpose of the continuation if (or when) warranted. Readers who plan to read the material in this chapter may have to resort to information in section 5.1.3 to seek definitions of terms such as crystallographic ice type (S1, S2, etc.), grain and subgrain boundaries, basal plane, optic axis, and crystal defects and dislocations. On the other hand, readers who are interested only in developing a quick understanding of how the colorful photographs of crystalline structure of sea ice, which are presented in this and the next two chapters, are generated and interpreted may only flip through relevant material in section 4.3. Regardless of the technique used to retrieve information from sea ice field samples, precautions must be taken to preserve the integrity of the sample as existing in nature. For example, it is impossible to remove an ice core from the ice sheet without any loss of brine at the bottom of the core. Therefore, the most important precaution is to ensure that minimum of brine drainage occurs during the extraction process and prior to making thin sections for crystallographic analysis. For this reason, ice cores and blocks should be extracted at ambient temperatures as low as possible (< −20 C). This may sound very restrictive, but not impossible in the polar regions. Of course, the ambient working conditions could be far from pleasant, certainly not summer adventure tour type. It is recommended that in situ crystallographic structural properties of sea ice be examined under field conditions as soon as possible after extracting the samples from the ice field. Storing and shipping samples of sea ice cores/ blocks to laboratories far away may not be desirable in many situations because they become subjected to desalination and thereby structural changes. Field laboratories can be placed either on top of the ice covers or nearby shore closed to the sampling site or on an icebreaker if used as the platform in a scientific field expedition. Such facilities do not have to be more than a shelter protected from the elements of windy weather (example of field laboratory is presented in section 6.2.1 and Figure 6.3). In case the structural analysis of freshly sampled ice cores/blocks has to be conducted in a distant laboratory, a few precautions must be followed in transportation. The samples should be packed immediately in insulated bags and transported in insulated boxes filled with dry ice, making sure that the samples do not come in contact with

dry ice in order to avoid any thermal shocks. Upon arrival to the laboratory, the samples should then be stored in deep freezers at temperatures less than −30 C, which is below the eutectic points of most of the predominant salts in sea water. Salinity measurements should be conducted in the field as soon as possible after sampling, and then other measurements must be conducted in the laboratory to make sure that no brine drainage had occurred in the intervening period. The techniques that reveal the microstructure properties of ice involve preparation of thin sections of 1 mm or less thick. At the core of this chapter is the description of the solid-state double-microtoming technique (DMT), which preserves the integrity of the ice and snow samples. Colorful photographs of thin section using polarized light reveal microstructural and molecular processes of sea ice. Several of such photographs and micrographs are presented in this and the next two chapters. Photographs are taken using polarizers with large field of views in order to accommodate thin sections up to 300 mm in diameter, containing a large number of grains. The instruments and the techniques are also described in this chapter. It is important to mention that although snow is usually an integral part of any ice cover, the microstructure of snow on ice is never discussed. The presented DMT and the photos using polarized light can be used to reveal microstructural aspects of snow. Further treatment of thin sections, using what is known as etching techniques, can reveal a few more micro features such as boundaries between sea ice grains or subgrains can be revealed. Individual grains are defined as ice crystals with large angular difference between them; while subgrains are crystals located within a grain with small angular differences (see definitions in section 5.1.3). Thermal or chemical etching can be performed. Both require polishing of only surface of the thin section and can readily be used in the field. Replicas can be produced from these processes, preserved and transported easily for further examinations after returning from field trips. The material in this chapter proceeds to cover four subjects: (1) some properties of light relevant to viewing and photographing ice samples, with emphasis on birefringence of ice crystals and the subsequent optical retardation, (2) techniques of preparation of thin ice sections (< 1 mm thick), with emphasis on the doublemicrotoming method that preserves most of the information from the field samples of ice core or block, (3) viewing and photographing of thin sections using polarized and scattering light as well as micrographs using scanning electron microscope (SEM), and (4) advanced techniques for revealing fine crystallographic features. The latter uses simple sublimation as well as thermal and chemical etching and replicating ice surface.

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

It must be reiterated that nothing substitutes the work that has been carried out in the field if performed with utmost care and precautions. In addition to the on-site measurements of basic physical properties that are traditionally carried out, all the necessary microstructural investigations on sea ice should also be carried out in the field soon after recovering the samples from the site. Ice researchers may think that this is easier said than done, but the present authors actually adhered strictly to this principle. Instead of taking ice to distant laboratories, they took the laboratories to the ice (see sections 6.2 and 6.3). Nonetheless, various laboratory techniques and scientific principles governing the methodologies used must be learned and practiced first at home before going to the field. It is also hoped that approaches and the results described in this chapter (and elsewhere in this book) can provide some intuition to remote sensing modelers of sea ice. 4.1. RELEVANT OPTICAL PROPERTIES 4.1.1. Polarized Light Light is an electromagnetic (EM) wave in the visible range from violet to red with wavelengths (universally denoted by the Greek letter, λ) from about 400 to 800 nm (nanometer is 10−9 meter) and frequencies (denoted by the Greek letter, ν) in the range of about 750 to 380 THz (TeraHz for 1012 Hz). The frequency is the number of vibrations executed in unit time (one second). The wavelength is, therefore, defined as the ratio (c/ν), where c is the speed of the wave in a given medium. Light with one frequency is called “monochromatic.” Within the visible range, however, the sensitivity of human eye varies in a measurable way. The sensitivity of daylight-adapted “normal” healthy human eye is highest around the wavelength of about 550 nm (green). Depending on the dilation of the pupil or level of dark adaptation, the sensitivity of human eyes shifts toward the red. Speed of light or any EM wave depends on the medium through which the wave is propagating. It is (a)

also dependent on wavelength and hence frequency ν, except for vacuum. In vacuum, the speed of light c is 2.9979246 × 108 m/s. Thus, a monochromatic EM wave propagating in vacuum can indeed be described not only by a single frequency but also by a single wavelength. But inside any other medium, the wavelength of the same monochromatic wave varies depending on the medium. Due to the mediumdependency of wavelength, it is appropriate to identify EM waves only by their frequency. The wavelength can be calculated as (c/ν). Figure 4.1a shows a simple sinusoidal wave with amplitude AC (from trough to crest) traveling from the left to the right. In such a wave, the distance between two successive troughs (AA ) or crests (CC ), or a corresponding distance such as BB’, is the wavelength. Two points are in the “same phase” when they are in the same relative position and moving in the same direction, such as A and A , B and B , or C and C in Figure 4.1a. The term phase is, therefore, meant for the relative position of any two points on the wave and is one of the most important parameters of an EM wave. For successive waves, the phase angle (also called phase shift) between two waveforms of the same frequency is the angular difference between their starting points. Figure 4.1b illustrates three waves with one at phase shift ¼ λ (quarterwave or π/2 radian) and another ½ λ (half-wave or π radian) behind the first one. Two waveforms are said to be in phase when they have the same frequency and there is no phase difference between them. Two waveforms are said to be out of phase when they have the same frequency and there is some amount of phase shift between them. Polarization is another characteristic of EM wave, not only pertinent to the optical frequencies or wavelengths used in optical techniques for revealing the microstructure of polycrystalline ice, but also extremely important for microwave remote sensing purposes. It is defined as the plane of the electrical field in an EM wave. Only a cursory description will be made here as a refresher because this subject is covered in any textbook of Physics. More details are presented in section 7.2.1. Figure 4.1c (b)

λ = c/υ C

λ/2

C′

C

Monochromatic wave of frequncy ν

A

B′

λ/2

C

λ = c/υ C′

3

B

A′

151

2

B

1 B′

A

λ/4

A′

Figure 4.1 (a, b) (a) One wave and (b) three waves, numbered 1 to 3, of the same monochromatic light with wavelength, λ, but different phases: λ/4 between 1 and 2, and λ/2 between 1 and 3.

(c)

SEA ICE

Electromagnetic waves

PLANE POLARIZED LIGHT

(d) E-field

UNPOLARIZED LIGHT θiB

Optic axis

152

cte fle Re am be

Figure 4.1 (c, d) (c) Electromagnetic monochromatic wave train with linear motion from the left to the right and (d) sketch of electric field of randomly polarized (unpolarized) wave.

ICE PLATE

θrB

ed ct fra Re am be

illustrates a beam of electromagnetic monochromatic wave traveling with a linear motion from the left to the right. Figure 4.1d shows the corresponding uniform linear motion in the cross-section of the beam. The doubleheaded arrows represent the electric vector, i.e., the polarization of the waves. If electromagnetic waves are generated from a source with their electric field oriented randomly, the beam of waves is called unpolarized or “randomly polarized.” Rays of sunlight or lights from lamps and electric bulbs are randomly polarized. Solar radiation diffused through clouds or fogs are also randomly polarized, but that is not the case for sunlight scattered from the sky. Blue sky light on clear days is somewhat polarized. Lights reflected from non-metallic surfaces, such as tree leaves, water bodies, and floors are also partially polarized. Before proceeding with further discussions, it is worth noting the definition of optic axis, which will be used often in the rest of this chapter. The optic axis of a crystal is a direction in which a ray of transmitted light suffers no birefringence (this term will described in the next section). Optic axis is a term applied to crystals while optical axis is applied to an optical system. Uniaxial crystals have a single optic axis while biaxial crystals have two different optic axes. Ice crystals are uniaxial, similar to other materials such as calcite, quartz, ruby, etc. In uniaxial crystals, optic axis is the same as c axis and is perpendicular to the basal plan (plane of growth) of the crystal (definitions are presented in section 5.1.3). Since non-crystalline materials have no birefringence (in general), therefore, the optic axis definition does not hold. Figure 4.2 illustrates a special case of reflection and refraction of an unpolarized beam of monochromatic light for a critical case of the incidence angle, called Brewster angle (θiB) (section 7.2.2.1). The plate could be any transparent material, crystalline or amorphous like glass, but if the plate is made from a single hexagonal crystal of ice (Ih) with its optic axis (c axis) normal to the surface of the plate, then the Brewster angle is close to 52.6 . The specialty of this particular example is not in the characteristics of the angles of reflection or refraction. Like any flat

d

Figure 4.2 Polarization of light by reflection from a special case of a single-crystal ice plate with its optic axis (c axis) normal to the surface and the angle of incidence equals the Brewster angle, θiB, which is 52.6 for ice (sketch by N.K. Sinha).

and transparent plate, the angle of reflection in this case is also equal to the Brewster angle, θiB (Figure 4.2). If the angle of refraction that corresponds to the Brewster angle is denoted by θrB , then the refractive index of the plate (np) is given by np = sin θiB / sin θrB . For this illustrated configuration of the ice plate and for λ = 589.3 nm, np = 1.309 corresponding to θiB of about 52.6 . However, for this special case of incident angle, both the reflected and the refracted beams are plane polarized. Plane of polarization of the reflected and the refracted beams are shown by the double-headed arrows for electric vector. Note that the reflected beam is polarized in the plane parallel to the surface; whereas the refracted beam is strongly polarized in a plane 90 to the plate surface, i.e., parallel to the orientation of the optic or c axis. For this reason, the beam is traditionally called “ordinary ray” and the refractive index is usually denoted by no. In birefringent materials such as ice (presented in next section), the refractive index for a light beam with plane of polarization normal to the optic axis is called “extraordinary ray” and denoted ne. These two names, ordinary and extraordinary, have no special significances, other than the historic relevance; other names could have been more appropriate. When an unpolarized beam of light is transmitted through a linearly polarizing filter (polarizer), it becomes linearly polarized as shown in Figure 4.3. The set of double-headed arrows indicate the “pass direction” or the orientation of the electric vector. If another polarizer (called analyzer) is inserted in the way of the propagating polarized beam with its pass direction oriented at 90 to that of the polarizer, no light will be transmitted. The doubleheaded arrows on the two polarizers in Figure 4.3 indicate their orientations or pass directions. If a snowflake or a plate of single crystal of ice, with c axis parallel to the

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

153

Snow flake c - axis

No light (black)

Unpolarized light

Plane-polarized light

Polarizer

Analyzer (cross position)

Figure 4.3 Optical components of a standard large-field polariscope with its polarizer and the analyzer in cross position; a snowflake with its c axis parallel to the direction of light is inserted here to show that it will look black, irrespective of its rotation (sketch generated by N.K. Sinha)

direction of propagation of light, is placed between the polarizer and the analyzer, while the analyzer is in cross position with respect to the polarizer, it will appear as black when the viewer looks through the analyzer. This is a position of “extinction” for the snow flake or any crystal and is used extensively in determining orientation fabric diagrams for ice using polariscopes equipped with Rigsby universal stages as presented in Bader [1954] and Langway [1958] (see definition of fabric diagram in section 5.1.3). It is appropriate to mention here that similar arrangement of a thin section of quartz positioned between and parallel to crossed polarization, was defined as the “standard photometric configuration” by Martinez [1958]. The illustration in Figure 4.3 also explains why beautiful stellar flakes of snow placed flat on polariscopes exhibit no colors and may disappoint the enthusiastic explorer.

4.1.2. Birefringence or Double Refraction of Ordinary (Ih) Ice From an optical standpoint, there are essentially three divisions of crystals. They are: isotropic (isometric), uniaxial (tetragonal and hexagonal), and biaxial (orthorhombic, monoclinic, and triclinic). Ordinary ice (called Ih) belongs to the hexagonal system and consists of oxygen (O) and hydrogen (H) atoms. The stoichiometric composition of water has been investigated by numerous investigators over a very long period time and been documented thoroughly in many reports and books. The ratio of the combining volumes of oxygen (O2) and hydrogen (H2) at 0 C and a pressure of 0.76 mm Hg is O2/H2 = 1/2.00288. Essentially, two atoms of H are attached to

one atom of O in a molecule of water. In natural water, one atom of D2 (deuterium, denoted by D and is a stable isotope of hydrogen) is also present for about 6500 atoms of H2. The atomic and the molecular structure of water in all its phases (liquid and solid) have intrigued scientists for many centuries. Although the chemical formula of water is very simple, its atomic structure, depending on temperature and mechanical or electrical forces is extremely complicated and difficult to study. It was known for a long time that a molecule of liquid water consists of one oxygen (O) atom and two hydrogen (H) atoms. The H atoms are not symmetrically placed around the O atom in a molecule of water. The angle between the two O-H bonds is slightly greater than 90 , close to 104.5 as shown in Figure 3.15 [Pounder,1965, Chapter 5, p. 62]. This makes it a polar molecule with a permanent electric dipole moment. When a uniform electric field is applied to water, it becomes slightly “doubly refracting” or “birefringent.” This is known as Kerr electro-optic effect for water [see Dorsey, 1968, p. 381]. In case of a uniaxial crystal, its birefringent character means two refractive indices or doubly refracting. Birefringence is thus defined as the splitting of a light wave into two unequally transmitted or refracted waves by an optically anisotropic medium such as ice. Although water without any applied electric field is optically isotropic, ordinary ice is doubly refracting and this property is responsible for producing beautiful color in ice thin sections when observed under cross-polarized light. It is difficult to provide an unmistakable example of evidence for the duality of refraction in a single crystal of ice because ice is a weakly birefringent material with birefringence, β, of 0.0014 (to be clarified later) for

154

SEA ICE

wavelength, λ, of 546 nm (green). However, an unequivocal demonstration of double refraction can be made with a transparent crystal of calcite, popularly known as “Iceland spar” with β of 0.1738 which is more than two orders of magnitude (actually 124) greater than that for ice at the same wavelength of 546 nm. This green color (line) is predominant in mercury vapor lamp and close to the maximum sensitivity of normal human vision. Calcite is, relatively speaking, an exotic material, but chemically it is calcium carbonate (CaCO3) and, therefore, the same as marble or chalk. If an object is viewed through a clear cleavage rhombohedron of calcite, two images of the object are seen; one image is fixed while the other revolves around the fixed one if the crystal is rotated. This is illustrated in Figure 4.4. It shows two images of the word “ICE” when a clear cleaved rhombohedron shaped crystal of calcite is placed over the paper. The triangle at the bottom right corner of the crystal in Figure 4.4a is cut and polished purposely to use it as a marker, so that the rotation of the crystal can be seen without any ambiguity. Note the rotation of one image around the fixed one as the crystal is revolved clockwise in steps of 45 through a total angle of 270 . The top set of letters of the word “ICE” in Figure 4.4a is the fixed image and the bottom set is the movable image. The movable set goes above the fixed one when the crystal is rotated by 180 as can be seen in Figure 4.4d. The fixed image is produced by the

(a)

ordinary ray whereas the revolving one is produced by the extraordinary ray. A birefringent material has two velocities of light, i.e., two refractive indices nλ(o) and nλ(e). They depend on the wavelength, λ, and the direction of propagation of light. Here, nλ(o) and nλ(e) are the refractive indices of the ordinary and the extraordinary waves (to be clarified below), respectively. In general, the refractive index, nλ for a given wavelength, λ, is measured as the ratio of the speed of light in vacuum, c = 2.99792458 × 108 m/s), and its speed inside the substance. By virtue of the definition, the value of the refractive index of vacuum is 1. Because, the speed of light is slower inside a transparent body in comparison to that in vacuum, the refractive index at any wavelength is greater than 1. Obviously, the greater the value of nλ , the slower velocity of propagation of light. In isotropic (isometric) crystals the refractive index is the same for all directions. Light travels at the same velocity in all directions. This is also the case in isotropic noncrystalline materials, such as amorphous window glasses without any thermal tempering or built-in stresses. Tempered safety glass used in car windows, however, is optically anisotropic or birefringent due to the internal stresses induced by thermal tempering and behave somewhat like optically positive crystals under compressive stresses [Sinha, 1978a, 1978b]. On the other hand, a crystal of ordinary ice Ih is a naturally birefringent material and behaves like an optically positive uniaxial crystal with its

(b)

0

(d)

(c)

45

(e)

180

135

(f)

225

270

Figure 4.4 Double images of the word “ICE” as seen through a rotating clear cleavage rhombohedron of birefringent (double-refracting) calcite crystal; note the rotation of one image around the other (fixed one) as the crystal is revolved clockwise ((a) through (f)); the angle of rotation is shown at the bottom right corner (photos by N.K. Sinha).

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

optic axis parallel to its c axis. The term optically positive is used to indicate that the ordinary wave, defined as the wave that travels with the same velocity in all directions in the crystal, is faster than the extraordinary wave, defined as the wave whose velocity depends on the direction of propagation inside the crystal. Both waves are polarized and are at right angles to each other. However, they travel at the same speed in the direction of the c axis and that is the reason for calling c axis also as the optic axis. This axis can be readily visualized as the line normal to the basal plane or the flat plane of a simple snow flake as shown in Figure 4.3. NOTE ON CORNEAL BIREFRINGENCE This book is obviously written to be read by ablebodied persons capable of reading through their eyes to comprehend the images of micro-, macro- and meso-scale remotely sensed images of ice. It could be of interest to some of the readers to know that the cornea of human eyes (like those of most other animals) is also a birefringent material, like a snowflake, with the optic axis normal to the curvature of the eyes or parallel to the direction of light as it enters the eyes. The epithelium and the endothelium layers, immediately in the front and the back of the cornea, however, were shown to be isotropic. These properties of the components of the human cornea were shown (believed to be for the first time) by investigating the Rayleigh scattering of linearly polarized coherent beam of laser light, with wavelength of 632.8 nm, as it travels through the cornea [Sinha et al., 1968]. This was achieved by using the first successful He-Ne gas laser in Canada (developed at the National Research Council of Canada) which was also used to investigate stress-induced Rayleigh scattering in tempered glasses [Bateson et al., 1964, and Bateson et al., 1966]. It is appropriate to remember that the first successful laser was invented in 1963. Polarimetric studies of microwaves are strongly linked to this type of scattering of electromagnetic wave and hence remote sensing of the future generations of space-borne radar sensors.

To sum up, birefringence is defined as the splitting of a light wave into two unequally transmitted (or refracted) waves by an optically anisotropic medium. The degree of the splitting can be strong or weak depending on the degree of the geometric order of the crystalline structure. Some details of the interaction of the incident light in relation to the crystalline structure of the material can be found in Hecht [2002]. When a beam of light enters a birefringent material, the beam is refracted into two beams: ordinary (O), which

155

follows the Snell’s law, and extraordinary (E), which does not follow the same law. The polarizations of the two beams are perpendicular. The two refracted beams of which the O beam from the first one and the E beam from the second one meet at the same point at the opposite surface. Because of the different path lengths, the two will have different phase at their meeting point. The interference between these two refracted waves (constructively or destructively) determines the amplitude of the departing wave (i.e., the wave going from the opposite surface) at each wavelength. Since the amplitude will be different for different wavelengths, the light that exits the crystal will have a certain color (but not white as the incident light). The color depends on the phase shift between the ordinary and extraordinary waves at their point of meeting at the opposite side of the crystal. This phase shift is called “optical retardation.” This feature makes crystals that are oriented differently appear in different colors as can be seen in the photographs of ice thin sections that appear in this book. Many examples that show this feature are presented in the next section. The color of each crystal will be different because it is determined by three factors: (1) the difference in the velocity of the two refracted beams in the crystal, which is known as the birefringence and denoted by the Greek letter, β, (2) the orientation of the crystal’s optic axis with respect to the incident light beam, and (3) the thickness of the specimen. The relationship with the thickness of the specimen is discussed in the next section. In ordinary ice Ih, the extraordinary wave is slower than the ordinary wave. This means that the refractive index of the extraordinary wave ne(λ) is greater than that of the ordinary wave no(λ) [Dorsey, 1940, Table 120, p. 485]. The term wavelength-dependent birefringence, βλ, is defined as: βλ = ne λ − no λ (4.1) Since βλ is always positive for ordinary ice, its crystals are known to be optically positive uniaxial. However, Ih ice is also a weakly birefringent crystal and βλ depends slightly on wavelength. Its variation (also called dispersion) in the visible wavelength range can be ignored from all practical points of view. For ordinary ice with ne = 1.3118 and no = 1.3104 at –3 C, βλ = 0.0014 for λ = 546 nm. This is one of the lowest values among those of most minerals. As mentioned earlier, calcite crystal with β = 0.1738 is 124 times birefringent stronger than ice. Its no = 1.6618 and ne = 1.4880 for λ = 546 nm. With no being higher than ne, this is an optically negative uniaxial crystal. 4.1.3. Optical Retardation If a simple flat snow flake or a thin section of uniaxial ice crystal cut parallel to the basal plane (i.e., perpendicular to the c axis) is examined between crossed-polarizers,

156

SEA ICE

it remains black irrespective of the crystal rotation as pointed out in Figure 4.3. However, if a thin section (or plate) of polycrystalline ice c axis of the crystals randomly oriented) are examined between crossed-polarizers using white light as the source of illumination, then wonderful color effects would result. Each crystal in the thin section appears in different color and the section (plate) becomes a mosaic of colors (see Figures 4.13, 4.14. 4.20, and 4.21). As explained above, the special color effects are caused by the phase shift of ordinary and extraordinary refracted waves when they exit the crystal. This is known as interference of colors. The effect is very complex, but predictable. However, it requires an understanding of the physics that governs the passage of polarized light with a wide range of wavelength (such as white light) through doubly refracting crystals. This is described below, but first the effects of monochromatic light of a single frequency or wavelength, λ, will be explored. The phase shift of the ordinary and extraordinary waves when they emerge from the crystal at the same point is called optical retardation. It depends on the thickness of ice crystal and its crystallographic orientation with respect to the direction of propagation of the light beam and its plane of polarization. If the analyzer is set in the position of extinction with respect to the polarizer, i.e., in the “cross position,” as illustrated in Figure 4.3, the ice crystal, with its c axis not parallel to the direction of propagation, will appear as brighter, depending on thickness, when viewed through the analyzer. The intensity will actually vary from darkness to the maximum brightness depending on the thickness of the crystal. The phase shift does not alter the wavelength and thereby the color will remain the same as that of the chosen wavelength of monochromatic light. Only the brightness will be affected because it depends on the phase shift (or equivalently the thickness of the specimen). The maximum brightness will be for the thickness that introduces a phase shift of λ/2, for which the two waves (the ordinary and extraordinary) will combine to produce a new wave having the sum of their amplitudes. An opposite effect will occur if the phase shift is λ and the crystal will appear dark in this case. The crystal will have intermediate brightness for intermediate phase shift, for example a phase shift of λ/4. Thus, the brightness will vary from darkness for retardation of (nλ) to maximum intensity for optical retardation, R of (2n + 1)/2λ. The number n takes values 1, 2, 3, etc., and is called the order of interference. The relative optical retardation, Rλ, depends on the wavelength, λ, and is related to the birefringence, βλ , according to,

considerations have to be given to the crystallographic alignment of the crystal with respect to the direction of polarization and propagation of light, hence the equivalent optical thickness along the basal plane or normal to c axis. For the purpose of simple illustrations applicable to most minerals, for which the dispersion may not be neglected, it is suitable to calculate the dependence of optical retardation on crystal thickness using the birefringence of light with a wavelength somewhat in the middle of the visible range, 400–800 nm. A convenient wavelength for most minerals is that of the orange colored, readily available from sodium D line (589.3 nm). This is the dominant radiation from sodium lamp and the predominant color of the fluorescence bulbs or tubes of street lights used these days. As mentioned before, for this wavelength the birefringence of ice at −3 C is βλ = 0.0014 in the spectral range from 486.1 nm to 706.5 nm. Calculated results according to the linear equation 4.2 are shown in Figure 4.5. Unlike most minerals, fortuitously, the birefringence of ice Ih is very low and more importantly its dispersion within the visible range is practically negligible. Its βλ value at the lower end of the visible range, at 435.8 nm, is only marginally larger at 0.0015 [Hobbs, 1974, Chapter 3]. This renders the line depicting the dependence of retardation versus thickness in Figure 4.5 as practically universal for ice in the visible range. When a single crystal is viewed through cross-polarized light, its color will be black for all wavelengths when the thickness is zero. As the thickness increases, the crystal will appear brighter and the phenomenon that causes this is known as interference. A general case is the change in interference patterns with the increase in crystal plate thickness when viewed through plane-polarized

Rλ = ne λ − no λ t = βλ t

Figure 4.5 Dependence of optical retardation on thickness of ice crystal for linearly polarized monochromatic light transmitted through the crystal with the direction of propagation normal (90 ) to the optic or c axis.

(4.2)

where, t is the thickness of the single crystal ice plate, measured normal to the c axis. For oblique incidences,

1.2 B L A C K

Ice thickness, mm

1

V I O L E T

F O R

0.8

A L L

F I R S T

W A V E L E N G T H S

0.6 0.4 0.2

O R D E R B L A C K

G R E E N F I R S T O R D E R B L A C K

R E D F I R S T

V I O L E T

V I O L E T

O R D E R

S E C O N D

B L A C K

O R D E R

T H I R D O R D E R B L A C K

B L A C K

R E D S E C O N D O R D E R B L A C K

0 0

200

400

600

800

1000

1200 1400 1600

Retardation, nm

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

4.1.4. Interference Colors for White Light The absence of certain colors for certain thicknesses of the crystal (as shown above) complicates the situation if white light in conjunctions with cross-polarizers is used for viewing. Due to the interferences, the intensity of different parts of the spectrum of white light varies and has profound influence in determining the complimentary colors. Color of a crystal can be ascertained from the interference color chart. It starts with black and moves to dark gray which gradually becomes lighter as the thickness and hence the optical retardation increases. All the colors combine to form white light at Rλ = 250 nm as shown in Figure 4.6. At Rλ = 360 nm, combined effect is to produce bright yellow. This is followed by orange and red at 450 nm and 500 nm, respectively. Violet is the resultant

1

Ice thickness, mm

monochromatic light. The brightness will increase with the further increase in thickness up to a maximum intensity and then starts decreasing with further increase in thickness until it is dark again (black in color). The zone between the darkness at zero thickness and the first darkness defines the regime of the first order of interference. Naturally, the optical path or the amount of retardation is dependent on the wavelength. As can be seen in Figure 4.5, the first darkness for the violet color with λ = 400 nm comes when R = 400 nm. Similarly, the first darkness for green with λ = 550 nm and deep red with λ = 800 nm comes, respectively, when R is equal to 550 and 800 nm. The cycle of bright to dark repeats as the thickness increases. The zone between the first-order black and the second black is called the second order of interference. This is followed by higher orders of interferences. The range of the thickness for all the orders is the same, but it depends on wavelength. This is also illustrated graphically in Figure 4.5. In general, when viewed through polarizers using monochromatic lights with wavelengths of λ, the color of the ice crystal is brightest when Rλ = nλ/2 and black when Rλ = nλ. For the violet light with λ = 400 nm the retardation values for the black color are R = 0, 400 nm, 800 nm, 1200 nm, etc. The corresponding crystal thicknesses, according to equation 4.2 are 0.286 mm, 0.572 mm, 0.858 mm, and 1.144 mm, respectively. For the extreme end of the visible spectrum, λ = 800 nm (deep red), the crystal appears black when Rλ = 0, 800 nm, 1600 nm, etc. corresponding to thickness of 0 mm, 0.571 mm, and 1.142 mm. Therefore, the end of the first order for deep red coincides with the end of the second order for the violet, whereas the second order of interference for deep red covers four orders of interference for the violet. The color and its brightness, however, could be very different if white light is used for viewing. The absence of colors in white light leads to complimentary colors.

First order

Second order

0.8 0.6

B L A C K

W H I T E

0.4 G R A Y

0.2

157

O R A N G E

Y E L L O W

R E D

V I O L E T

I N D I G O

B L U E

G R E E N

Y E L L O W

O R A N G E

R E D

V I O L E T

0 0

200

400

600

800

1000

1200

Retardation, nm

Figure 4.6 Interference colors for various optical retardations, and the first two orders of interference colors are marked as functions of thickness if the ice crystal is viewed through cross-polarized white light transmitted through the crystal with the direction of propagation normal (90 ) to the optic or c axis; the solid line for optical retardation versus thickness is calculated for linearly polarized monochromatic light using βλ = 0.0014.

color at Rλ = 550 nm. The colors repeat with further increase in retardation. The violet color appears again at Rλ = 1100 nm. Thus, for white light the first-order color is defined as the range from the initial black to the first violet, and from this violet, up to the second violet at 1100 nm defines the “second-order colors.” Thus, both the first-order and the second-order colors are confined within a thickness of about 0.8 mm. As can be seen in Figure 4.6, the first order of interference for cross-polarized white light corresponds to a maximum thickness of about 0.4 mm. The corresponding second order of interference is limited to a thickness of around 0.8 mm. This provides an idea of the useful range of ice thickness for thin sectioning purposes. If the thickness is too thin, say less than about 0.3 mm, there may not be much differences in the color of the crystals. The same applies if the thickness is greater than about 0.8 mm. Thus, the goal for making thin sections with good colors of the crystals is to make the thicknesses somewhere between 0.4 to 0.8 mm. The above calculations are made for a single crystal of ice with its optic axis (or c axis) at right angle to the propagation of light. Pure single crystals are supposed to have no grain or crystal boundaries, but that does not apply to natural ice. This raises some interesting questions as to the choice of the thickness of ice samples for different crystalline types that allow best revealing of ice crystalline structure (i.e., the thickness of thin sections as discussed in the next section). Even the large grains of S1 type of freshwater lake ice (section 5.3.1) have subgrain structures with convoluted low-angle boundaries tilted with respect to the direction of transmitted light [Sinha, 2011]. On the

158

SEA ICE

other hand, the grains in snow ice or frazil ice could be less than 1 mm with rather complex grain boundaries and large amounts of inclusions at the boundaries. The useful thickness has to be around 0.2 mm or less in these types of fine-grained ice.

4.2. ICE THIN SECTIONING TECHNIQUES What is ice thin section and why do we need it? The answers to these two questions can be appreciated by examining the photograph of a thick section of FYI (Figure 4.7), cut horizontally from a cylindrical ice core. The ice was obtained during the AIDJEX experiment (section 6.1). The foggy appearance indicates the presence of brine, a characteristic of FYI as described in section 2.5.4. The numerous bright spots are brine pockets and brine drainage channels. The information is there, yet in a very coarse display. Thick sections can only provide crude description of sea ice composition. That was the purpose of taking the photograph shown in Figure 4.7—for just “quick-and-dirty” information. Note that the person held the specimen with bare fingers, which means that the preservation of the information was not a purpose. Thin sections (1 mm or less) are needed, along with appropriate light conditions for photography in order to reveal details of ice crystallographic structure, including brine pocket locations and distribution as described later. Taking advantage of the birefringence property of ice crystals (as explained above), thin sections are traditionally used to reveal individual crystals in

Figure 4.7 A close-up of a thick section (about 3 mm) from a horizontal cut of a circular ice core obtained from Chukchi Sea during the AIDJEX experiment in March 1975. Brine pockets and drainage channels are shown as bright objects within the foggy background (credit: Water Meier, NSIDC DAAC collection).

polycrystalline ice by different colors when observed through cross-polarized white light (section 4.3.2). A few photographs of thin sections are presented in this chapter (Figures 4.14, 4.20, and 4.21) and many more in Chapter 5. Preparation of thin sections of ice (freshwater or saline) is similar to preparation of petrographic and metallographic specimens for microstructural investigations (petrographic is description and classification of rocks by microscopic examination). This process, to a large extent, is an art. The thin sections of ice are made from thicker sections (1–2 cm thick) cut from ice cores or blocks using a bandsaw. The thinning continues to bring the thickness down to 1 mm or preferably less using melting, grinding, or microtoming. Polishing the surfaces and etching it thermally or with chemicals can be performed if required. Each of the steps requires a great deal of practices to control the quality of the finished surface. Thin sections can also be made using a very crude method that may be called as “the palm-technique.” The procedures can be extremely simple. A thick section is cut with a hacksaw to about 5 mm thick. The thickness is then reduced by successive melting from both sides using palms (if no other alternatives are available) to a thickness of about 1–2 mm. The thinning process should be progressive with periods of cooling. Reasonably good thin sections can be made with some practice, no matter how ridiculous the procedure may sound. This method can only provide crude and hasty examination. A more precise method involves use of hot glass plate to melt a thick section of a few millimeters down to a thin section of 1–2 mm. Hence, this method might be called the hot-plate technique. One of two techniques can be used. The first and certainly the most popular has been traditionally used by glaciologists. It entails using warm glass plates to melt the thick section. This technique may be subdivided further into two different procedures— soft and hard. The soft or the easiest method is to thin the section by successive melting from both the top and the bottom. The second and more painstaking, yet the most popular, is the method that employs the fixing one side of a thick section to glass plates by melting then letting it freeze, followed by the removal of materials from the top by microtoming as described in detail by Weeks [2010]. Obviously, the information at the bottom of the section is lost during melting, yet at least part of the information is survived. In general, the hot-plate technique should strictly be avoided for sea ice if the microstructural features, such as the size and shape of the air and the brine pockets or characteristics of intersubgranular boundaries and intrasubgranular crystalline defects, are to be preserved and examined. Another method, which is the most rigorous, is the DMT. It is a solid-state, cold plate process which avoids

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

using any warm glass plates. The DMT method is ideal for sea ice because it has so much structural information about inclusions, grain and subgrain boundaries and possibly very small crystals. This method was developed and presented in Sinha [1977a]. It was used for the first time to preserve the internal structure of FYI sampled from Strathcona Sound, Nunavut, Canada, during the dark polar nights of late November when the air temperature was about −30 C. 4.2.1. Hot-plate Techniques for Thin Sectioning of Ice In comparison to the steps that have to be followed in preparing petrographic specimens of rocks, ceramics and metals, the preparation of usable thin sections of ice by melting is relatively very simple. This simplicity, however, is believed to be the reason for the glaciologists to ignore the fact that ice in nature is a high-temperature material and application of any heat is liable to change the microstructure. Of course, the changes may not be noticed by unaided eyes. They may not also affect the coarse structure and texture of large-grained ice. Naturally, the practice of making specimens of freshwater or glacial ice, which have no brine inclusions, will continue in the future. In this aspect, glaciologists dealing only with large-grained ice have great advantages over geologists and metallurgists. All one needs for making a rudimentary ice section using the hot-plate technique is a sufficiently cold working area (below 0 C, say −5 C). The procedures are simple and require only a bandsaw to cut thick sections (say 5 mm in thickness), sand papers and/or wire mesh and a flat metallic plate with built-in heaters preferably using rheostats to control the temperature. Moreover, the hot-plate technique is inexpensive and requires very little efforts. It involves the removal of excess ice by melting and, with practice, good thin sections can be made. It has been proven to be very useful for examining the grain structures of freshwater lake, river, and glacier ice. The soft method the of hot-plate technique involves cutting a thick section (2–3 mm thick) using a good and stable bandsaw with sharp stainless steel blade, with tooth spacing of 1 in 6 mm, and removal of materials from both surfaces by melting. This is achieved by putting the thick section on a warm glass plate resting on a hot (warmto-touch) plate. A 15 mm thick large, 300 mm × 300 mm, metallic (preferably brass to avoid corrosion) plate, heated by insulated heating wires, provides the necessary thermal inertia for making sections. Let the bottom surface melt sufficiently to make it flat. Remove the glass plate together with the ice plate away from the hot plate, squeeze out the meltwater, flip the ice specimen and allow the composite to cool and meltwater to freeze. Once the ice section is solidly glued to the glass plate, remove the

159

excess material from the top surface gradually in small steps, making sure not to melt the glue and loose the specimen. With practice, the top surface can be made very flat. The hard hot-plate technique is more arduous and timeconsuming than the soft method. It was suggested by Langway [1958] and has been used extensively since then. Recently, a detailed description of this method is given by Weeks [2010, Appendix E] and the reader will gain an insight of not only the sectioning procedures and precautions to be taken, but also the methods to be practiced in the field for sampling ice blocks. Only a brief description is given below. 1. Saw cut the specimen. 2. Grind the surface with progressively finer sandpaper or wire mesh sand paper. 3. Hold and press firmly a warm glass plate, large enough to support the sample, on the specimen surface and let a water film form between the glass–ice interface and refreeze. 4. Using a hand/bandsaw cut the sample parallel to the glass plate leaving about a 1 mm thick section of ice adhering to the glass. 5. Reduce the section thickness from nearly 1 mm to the desired 0.4 mm to 0.8 mm thickness by: a. Sanding with sandpaper b. Using heated copper or any heated plane surface c. Using a modified standard milling machine d. Preferably a standard biological microtome with a vacuum plate attachment. Both the soft- and the hard hot-plate techniques have been used by many investigators since Rigsby [1953] and Langway [1958] while developing the methods for fabric diagrams. The hot-plate technique was also used extensively by Michel and Ramseier [1971] while developing the classifications of lake and river ice (presented in section 5.3.1). As for sea ice, it is rather very messy to melt or grind with sandpapers at any temperatures comfortable to human hands. Moreover, brine drains out of the sea ice when hot plates are pressed on the ice and during the thinning processes. Additionally, the processes of melting, especially when the final thickness is only about 0.5 mm, undoubtedly disturbs and damages the fine substructure of sea ice. Although used extensively in the past and perhaps the practice is ongoing, the hot-plate technique is not suitable for sea ice containing brine pockets, especially new, young or even FYI ice unless the goal is to quickly ascertain the type of ice required for many engineering applications. 4.2.2. Double-Microtoming Technique for Thin Sectioning of Ice The solid-state DMT [Sinha, 1977a] offers the best possible choice for making thin sections with near perfect

160

SEA ICE

parallel surfaces. The method is capable of producing surfaces without any artifacts, e.g., refrozen and mushy (brine-rich) layer of ice at the bottom as produced by hot-plate methods. The DMT is essential if investigations are to be carried out on undisturbed lattice structure of the ice surface. This includes basal and non-basal lattice dislocations in ice, internal structures of external stress/strain induced crystalline defects, such as pileup of dislocations [Sinha, 1978a], and, especially for sea ice, the subgrain structures and entrapped brine and air in the form of pockets (for definitions of the above terms see ection 5.1.3). This method has recently been extended to make sections with thicknesses down to about 0.2–0.4 mm required for fine-grained granular or snow ice or snow masses with crystal sizes of fractions of 1 mm [Satyawali, Sinha, Sethi, 2003, Klein-Paste et al., 2007). Double-microtoming refers to the method of finishing the top and the bottom surfaces of the thin sections separately without exposing the sections to any melting at any stages during the section preparation (using cold glass plates). Consequently, the entire process is solid-state and ideal for carrying out thin sectioning of ice (freshwater as well as seawater) at any temperatures without subjecting the material to any thermal and hence thermomechanical shocks. However, mechanical shocks, such as slight increase in surface defects in the form of lattice dislocations and dislocation loops, may be induced during shaving off thin layers from the surface of the specimen but this can be minimized if a trained person performs the thinning process. The procedures are described in the following. Sea ice extracted from the field in the form of cores and/ or rectangular blocks, with markings indicating their physical orientation (such as North), should be recovered only when the ambient temperatures are as cold as possible. The ice samples must be shipped and stored at −30 C or below to avoid any possibility of brine migration. Before commencing any sectioning work, the ice core or the block must be thermally stabilized by storing them inside a cold laboratory, maintained as close as possible to the working temperature. They are to be stabilized in order to minimize mechanical shocks during bandsawing and also to prevent changes of geometrical properties of brine and gas inclusions. Ideally, the working temperature of −30 C is the best for sea ice. Such a working temperature is extremely difficult for manual dexterity required for thin sectioning. A compromising temperature could be around −20 C, close to the eutectic temperature of NaCl. The thin sections are prepared from thick sections. The latter are cut using a bandsaw to approximately 5 or 10 mm in thickness, depending on the conditions of the saw and blade. Thin sections are prepared from thick sections by removing layers of ice using a sledge microtome.

This is an instrument that allows removal of very thin layers of a specimen using a steel or diamond blade. The blade shaves off a thin layer with each pass. This instrument is commonly used in medical and biological laboratories and was developed for obtaining thin layers of frozen tissue or other organic objects. The thick section is mounted on a clear glass plate. Glass plates must be kept at temperature −20 C or less. Mounting the thick ice section is accomplished by freezing a few drops of cold near-freezing water at a few points around the edge of the section; making sure that no water enters the space between the glass and the section. An eye dropper with a long nose is most appropriate for this purpose. Moreover, the water should be cooled to about 0 C by making use of an ice bath consisting of distilled water with crushed ice in it. The exposed surface of the section is then microtomed to a mirror finish in four stages using the sledge microtome machine. In the first stage, layers of 50 μm thickness each are removed in each pass of the microtome’s blade as it continues to shave the surface. The second stage starts when the thickness of the section reaches approximately 3 mm. Here, a layer of 500 μm is removed from the surface, taking only 10 μm at each pass of the blade. In the third stage the next 200 μm is removed in 5 μm layers while cleaning the microtome’s blade with a soft tissue paper once after a few passes. The final stage, which gives the final finish of the section, is achieved by removing another 50 μm from the surface in 1–2 μm layers, ensuring that the surface is clean before each pass. The quality of the finished surface has to be visually examined using reflected light from a distant source and an optical microscope. If no cracks are found, the section is removed from the glass plate by carefully cutting off the bonding ice at the points around the edge with a sharp razor blade. It should then be remounted on another clean glass plate with the finished surface facing the glass. It is preferable at this stage to build up a dam of ice completely around the edge of the thin section, making certain that no water enters beyond the outer edges of the section. This dam of extremely fine-grained freshwater is also made using an eye dropper with long nose. Its purpose is to prevent the slice from sliding during the rest of the thinning process. It also prevents moisture going and freezing between the glass surface and the bottom of the ice section. The opposite surface is then microtomed in steps, following the above procedures, until the thickness reaches the desired level and the top surface has a mirror finish. The thin section should optimally be between 0.4 and 0.8 mm as pointed out earlier; the thinner the section the better the information that can be revealed and resolved. However, a thicker specimen, around 0.8 mm, is often desirable for measurements

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

related to brine-layer spacings and related studies in FYI [Nakawo and Sinha, 1984, Sinha and Zhan, 1996]. The success of microtoming depends upon the condition of the microtome’s blade and the care with which the surface is prepared. Slight damage to the blade can cause the formation of rows of dislocations (removal of atoms) in the microtoming direction. Microtoming should produce flat and smooth surface without introducing undesirable effects at atomic scale such as modification of the density of dislocations. If thin sections are to be used later for further examination or processing by replicating or etching the surface (thermally or chemically as described in section 4.4.2), then they have to be stored in small containers at temperatures below −20 C, surrounded by crushed ice or snow in order to minimize the rate of sublimation. Thin sections of ice can be viewed and photographed using a polariscope and/or optical microscope, as described in section 4.3.

4.2.3. Double-Microtoming Technique for Thin Sectioning of Snow Almost always, ice covers have blankets of snow. Snow is also present in the environment on the grounds and the mountains. As pointed out earlier, snow cover on floating ice of lakes, rivers, and oceans has a profound influence on the growth and structure of ice, and hence on remotely sensed images acquired by using electromagnetic waves at optical or microwave frequencies. Moreover, the top surfaces of MYI may become very porous and often reach densities not far from those of consolidated snow covers. Since very little attention has been paid to the structural aspects of snow covers in earlier sections of this book, it is important to describe, albeit briefly, the characteristics of snow as a material and the extension of DMT for examining the structure of snow. This applies mostly to snow on the ground (glaciers and otherwise). Physical and mechanical properties of snow can be well understood from its microstructure and texture. Microstructure properties refer to geometry of snow crystals and pores in a sample. Texture properties refer to crystallographic orientations of snow crystals. A proper characterization of snow microstructure and texture is essential for adequate classifications of snow samples, and to understand a possible relationship between texture and material properties of snow. However, as snow is subjected to the changes in temperature, vapor pressure, etc., structural changes continue before, during, and after snowfall. Therefore, snow is thermodynamically unstable. To investigate such a material, it is important to slow down the natural snow metamorphism and minimize mechanical and thermal stresses during sample handling, transport, and analysis.

161

Snow has been traditionally classified based partly on examining thick sections. Classification of snow was first proposed by Bader et al. [1954]. These authors have developed a section preparation technique to examine snow by photomicrography and image processing. This, however, could not reach to the satisfaction level. International Commission on Snow and Ice then suggested a new classification [Colbec et al.,1990] based on the knowledge gained in few decades before 1990. The pioneering work of the Canadian researcher Perla [1982] led to the introduction of a number of possibilities for preparing plane sections for snow. One of the most useful techniques suggested by the author was the use of supercooled liquid-state dimethyl phthalate to fill the pore spaces in snow and then allowing the liquid to solidify and preserve the structure of the snow mass. Since then, those guidelines have been useful for many investigations. Good [1987] has explored it further and constructed a 3D image of faceted grain snow (depth hoar) by serial cutting of snow at different spacing. Brzoska et al. [1999] have performed serial cut on snow, but the technique is time-consuming and the ratio of the sample size to resolution is not satisfactory. Schneebeli [2000] developed a system that consisted of slicing, segmentation, and reconstruction of snow samples filled with dimethyl phthalate. The 3D structure of snow could be reconstructed and visualized on the computer within 2 h. The method is useful for structural information of snow sample but fails to get the textural information of snow. Coléou et al. [2001] used X-ray absorption tomography to build 3D high-resolution image of snow (10 μm3) and obtained the porosity and discrete local curvature of snow. No doubt, the X-ray tomography is good for analysis of single or bicrystals, but not good for polycrystalline mass. Moreover, it requires very special and expensive equipment that are not really portable for field use. Also, it requires special health-related precautions. Classification of snow samples based on thick section method is difficult in recognizing the type of snow when having similar structure as of other types. By the application of DMT together with thermal etching (presented in section 4.4.2.1), textural details of both vertical and horizontal thin sections of different snow samples can be obtained. It allows one to examine the texture of different types of snow, which was not possible by the traditional thick section technique. The DMT in conjunction with thermal etching is capable of bringing out the grain or subgrain boundaries and clearly show the type of bonds between crystals (geometric or crystalline) both by polarized light for relatively large-angle boundaries and smallangle boundaries by thermal etching. This way, one could characterize the texture of various snow types. Texture of snow, which has not been studied so far, is added

162

SEA ICE

information that can further be used to classify a snow sample in addition to the common snow class. Very little is known about morphological changes in snow on sea ice at the grain-scale level although these changes are known to affect not only the emission, absorption, and scattering of electromagnetic waves in the visible and the microwave range, but also thermomechanical properties of the top layers. To this extent, considerable progress has been made by Satyawali, Sinha, Sethi [2003] in the application of DMT to snow on mountains in an effort to develop a standardized method for investigations of snow cover and classify the snow on textural information and interpret mechanical parameters. The procedures have also been applied to study snow compacted by aircraft tire on the runways by Klein-Paste et al. [2007]. It is hoped that such scheme will be used in future for quantitatively structural aspects of snow on sea ice covers. Since snow is porous and cannot be processed by DMT, the pores have to be filled by the method suggested by Perla [1982]. This is accomplished at the sampling site and at the ambient temperature by slowly releasing a small block of sample (cube with sides of 50 mm) in a paper cup containing supercooled dimethyl phthalate. The phthalate is allowed to penetrate into the pore space. This may take a long time depending on the porosity and temperature. Once the sample is saturated with the fluid, the cup is stored below −20 C to solidify the liquid. Filling the pore space and lowering the temperature also slows down the snow metamorphism and supports the snow crystals during further handling. The paper cup can then be removed and thick sections are cut and thin sections are prepared essentially following the DMT procedures described earlier for ice. However, for the first stage of microtoming, it is better to fix the thick section on the glass plate by using a few drops of supercooled phthalate instead of water drops prescribed for ice. This is because it is often found difficult to release the sample from the glass plate after the first microtoming sequence without damaging it. Frozen phthalate is softer than ice and adheres less to the glass plate and this reduces the stresses on the sample while releasing it. The mounting step before the second microtoming sequence is performed, as prescribed for ice, with water to ensure a good sample support during microtoming to the final thickness. The final thickness for finegrained snow and snow ice has to be around 0.1–0.2mm. Sections of snow are to be cut both parallel and perpendicular to the snow sample. Since a single crystal of snow mass or a grain of snow ice is a birefringent material, it allows detection of individual crystals using polarized light. Consequently, their crystallographic orientations in a polycrystalline mass of consolidated snow and hence texture can be evaluated with no difficulty. However, thin sections of snow mass or snow

ice with extremely fine grains must be thin enough (preferably around 0.1–0.2 mm) to avoid problems created by the slanted crystal boundaries. Such thin sections can be made by DMT, but they will be white or most probably gray in color as can be figured out from Figure 4.6 and individual crystals cannot be delineated with ease. However, if a retardation plate that will introduce a relative retardation of about 400 nm is added to the light beam, for convenience between the polarizer and the thin section, then interference colors in cross-polarized mode could be within the desired zone (see also section 4.2.5) between the first and the second order of interference colors. In case of difficulty in getting a proper retardation plate, a mica sheet from an old iron can do the job. Mica sheets provide added advantages. They consist of thin layers that can be peeled for selecting the favorable thickness. This technique has been exploited to view the texture of fine-grained, faceted grain, and melt-freeze snow [Satyawali, Sinha, Sethi, 2003, Klein-Paste et al., 2007]. The extracted information cannot otherwise be obtained using other techniques. An example of Himalayan snow is shown in Figure 4.8. Another application of DMT to snow on runways is shown in Figure 4.9. Microtomed surfaces of snow samples (using DMT) can be further processed using thermal etching technique. This combination of techniques has been shown to be the best way to classify a snow sample by its texture [Satyawali, Sinha, Sethi, 2003]. Further developments in this area of snow pack characteristics can now be found in the book by Satyawali [2012]. Developments on the physics of mountain snow

Figure 4.8 Double-microtomed ultra thin horizontal section of snow subjected to equitemperature (ET) metamorphism from Himalayan Mountain under cross-polarized light using a retardation plate to show colors. Background between the colored grains is the pores filled with dimethyl phthalate and appears brownish [Satyawali, Sinha, Sethi, 2003 / Canadian Science Publishing].

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

Figure 4.9 Double-microtomed 5 mm wide horizontal thin section of aircraft tire compacted snow at a Norwegian airport under cross-polarized light with a retardation plate to bring the colors into the first order of interference (micrograph by N.K. Sinha and A. Klein-Paste).

morphology are equally applicable to those of snow covers on floating sea ice (albeit with complications caused by upward migration of brine from ice underneath). 4.2.4. Precautions for Thin Sectioning by DMT Viewing of ice thin sections through polarized light involves the passage of light through the glass plate and the slice thickness. It is an integrated method. A polarized light beam has to be transmitted through four surfaces, the two surfaces of the glass plate holding the thin section, and the top and the bottom surfaces of the section. The quality of the surfaces affects the clarity of the views in transmitted light. Scratch marks on the glass plates are the most common problems and are often difficult to avoid. For this reason, plates with scratches should be kept aside and used for the preparation of the first surfaces of thin sections while using DMT. Scratchfree glass plates should then be used when transferring the specimen following the preparation of the first surface. A common problem is also caused by the condensation of moisture from hands and breathing. Handling the glass plates during and after completion of thin sectioning should never be made with bare hands or even woolen gloves that could be very porous. Leather gloves with no furs are best for handling thin sections. One has to be careful also to avoid nose drops (common when staying in cold environment for long times) from falling on the sections or condensations coming from breaths, particularly during close examinations with magnifiers or microscopes. Another serious problem is caused by

163

the condensation of water vapor in the spaces between the bottom surface of the ice specimens and the top of the glass surfaces. The water vapor may solidify on the bottom surface of the specimen. Features created by this type of condensation may not always be recognized as artifacts and may distort the interpretations of micrographs. This is also the opportune moment to reemphasize the need for the use of DMT. It is virtually impossible to avoid undesirable artifacts at the glass–ice interface if hotplate technique is used for sectioning. However, these apparently time-consuming precautions are not necessary if thin sections are made only for classifying the type of ice and constructing the fabric diagram. They are important for the purpose of examining the detailed microstructure and contents of grains, subgrains or internal structure of brine in the form of pockets that also includes air pockets. Oblique grain boundaries also generate viewing issues. This is a serious problem for fine-grained materials. It generates technical problems in terms of usable section thickness. For ice with grain diameters less than one millimeter, the usable thickness is around 0.1–0.2 mm. To make such thin sections with parallel surfaces is by no means trivial. It is not only difficult to make such thin sections, but also the color under cross-polarized lights is gray. Enhancement of colors can be made using retardation plates as described earlier. Nevertheless, finer details such as areas between very low-angle boundaries as those between subgrains and other linear defects of deformation-induced tilt are not made visible. For sea ice, this raises some serious issues. Although grains different in their crystallographic orientations are recognizable under polarized light, the substructures inside the grains cannot be evaluated except by the examinations of the distribution of the inclusions, i.e., brine and air pockets. The low-angle boundaries, between platelets or subgrains of pure ice, with very little lattice mismatch (say less than 1 ), are not delineated as clearly as desired in polarized light. This lack of resolution is particularly critical for the oriented S3 crystallographic structure type of FYI (see definition of S3 type in section 5.3.1 and examples in section 5.3.3.5), which results in partial disappearance of distinct colors of individual crystals. It is also critical for large areas occupied by cellular structure consisting of platelets or subgrains with their smallest dimensions less than 1 mm. Such ice was named as “bottom ice” by Peyton [1966].

4.2.5. Optimum Thickness for Thin Sections of Ice and Snow It can be seen from equation 4.2 that a value of Rλ = 550 nm is obtained for a thickness, t = 0.39 mm. This amount of optical retardation, as presented graphically in Figure 4.6, defines the limit of the first-order interference or the color of violet if white light is used

164

SEA ICE

for observing thin sections between cross-polarizers. An ice section with thickness of 0.78 mm leads to the limit of the second order of interference or the appearance of the second violet at Rλ = 1100 nm. The interference colors in white light are the best if the optical retardation is within the first two orders preferably a little beyond the end of the first order. Higher orders produce faded colors. Specimens cannot be too thin either (except for snow and snow ice). Practically, no colors are produced for retardations less than about 250 nm or thickness less than about 0.2 mm. This is why thickness for thin sections of polycrystalline ice (particularly sea ice with brine inclusions) should be around 0.4 mm to 0.8 mm (as mentioned earlier) for the best interference colors in white light. Demonstration of a colorful solid-state DMT thin section of a specially fabricated polycrystalline ice is shown in Figure 4.10. The specimen thickness was about 0.4 mm and as expected from Figure 4.6, the colors are rather warm, predominantly yellow. In this case, the three letters of alphabets in the word “ICE” were initially cut from vertical thick sections of directionally solidified (DS) columnar-grained distilled-water ice, classified as S1 type (sections 5.1.3 and 5.3.3.3). In S1 ice, the crosssectional diameters of the crystals in the horizontal plane (i.e., normal to the direction of heat flow) can be 100 mm or larger. The c axis of the vertically oriented columnar grains in S1 ice tends to be also in the vertical plane. Consequently, sections made in the vertical plane tend to have the c axis in the same plane and therefore normal to the direction of propagation of light. Slight deviation of the c axis from the plane of the section, due to the formation of subgrains in S1 ice (see Sinha [2011] for details), led to differences in the colors. Each of the letters in Figure 4.10 was as large as 80 mm. Snow or granular ice forming on floating ice covers tends to have crystal sizes in the range of 1 mm or less. Spot cooling and freezing of water droplets on metallic surfaces such as in-flight icing on aircraft wings and helicopter blades may have crystals with sizes around 0.05 to 0.1 mm. Usual optimum thin sections of sea ice,

Figure 4.10 Double-microtomed thin section (about 0.4 mm thick) of S1 ice type in the shape of the word “ICE” viewed through cross-polarized white light (photo by N.K. Sinha).

discussed above, which are required for good interference colors may contain several of such grains within the depth of the sections. Moreover, their grain boundaries could be very slanted and occupy large areas if sections are in the usual range. Optimum thicknesses for such materials are about 0.1–0.2 mm. This minimizes the width of the grain boundaries that could be slanted. This thickness range, according to Figure 4.6 produces Rλ = 140 to 280 nm. This range of retardations produces gray or white color. Additional retardations can be introduced, if necessary, to the beam of light in order to bring the interference colors within the first two orders of colors. This can be done by using retardation plates such as mica sheets in between the polarizer and the analyzer either before or after the thin section of the snow is made [Satyawali, Sinha, Sethi, 2003, Klein-Paste et al., 2007]. Regardless of the thickness of thin sections, different crystals with different orientations of their c axis in a polycrystalline body appear as areas of different colors if transmitted white light is used for viewing. However, colors are never used for textural analysis of polycrystalline ice. The texture analysis is made with the help of a universal stage and positions of extinction [Langway, 1958]. Microstructural aspects involve the geometry of ice and snow crystals and pores in a sample, whereas textural aspects involve relative crystallographic orientations of ice and snow crystals in a sample.

4.3. VIEWING AND PHOTOGRAPHING ICE THIN SECTIONS As mentioned earlier, the birefringent property of ice and snow crystals is used for revealing individual crystals in different colors using plane-polarized white light. The spectacular colors are very attractive and provide the impetuous and joy for making thin sections in spite of the laborious (depending on methods used) process that has to be completed in a cold laboratory (at temperature < −10 C). The different colors (or color shades) of crystals are incidental, not essential. Moreover, black and white photographs were used traditionally for measurements, not only because the photography was cheaper, but the colors were not required for any quantitative analysis. The digital photography of today has removed the barriers of cost of producing the photographs, but, once again, the colors add only marginal advantages. In fact, the use of plane polarized monochromatic light sources provides better definitions of the intercrystalline or grain boundary regions. Once a thin section is prepared, it has to be examined visually or under microscope and then photographed. To reiterate, optical microscopes are rarely used for this purpose, though magnification can reach 200×, because

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

of their limited depth of field (the distance between the nearest and farthest objects in a scene that appears acceptably sharp in an image). Polariscopes are commonly used to view and photograph ice thin sections because they utilize the birefringence property of ice that reveals its polycrystalline structure. In fact, the optical characterizations of minerals are made by means of a special form of microscope known as the polarizing microscope in addition to standard laboratory equipment such as a refractometer, a goniometer, and an axial angle apparatus [Rogers, 1937]. However, polarizing microscopes can handle only small specimens (< 10 mm width). Therefore, they are not suitable for viewing ice thin sections which are typically made from ice cores of 10 cm in diameter or ice blocks of even larger size. Replicas of the top surfaces of thin sections (section 4.4.2.2) can also be made and viewed and micrographed using an SEM. It produces images by scanning the sample with a focused beam of electrons (instead of light) that interacts with the electrons in the sample. SEM can achieve magnification between 10× and 500,000× with an ultimate resolution better than 1 nm. The sample can be a few centimeters wide. 4.3.1. Laboratory and Hand-Held Polariscope There seems to be a general confusion as to the use of the terms “polarizing microscope” and “polariscope.” Polarizing optical microscopes have complex optical systems of condensers and objective lenses and have a very limited field of view. For this reason, the standard polarizing microscopes require small specimens (thin sections). These microscopes are used for performing petrographic analysis of minerals, among many other applications. For such applications, the specimens or slides are small, usually not more than a few millimeters in diameters. Most microscopes have specimen holding carriages with x-y movements for handling larger specimens, but only a fraction of such specimens can be examined at one time. Viewing areas (i.e., field of view) of optical microscopes, equipped with polarizers, are usually in the range of a few millimeters. The size of grains (or individual crystals) in natural ice is significantly larger (often orders of magnitude) than those of most other natural crystalline materials. The field of view necessary for petrographic studies of natural ice, therefore, has to be large and capable of handling as many grains as possible. For this reason, the polarizing microscopes are not very useful for ice unless the subgrain or the intragranular structures are to be examined. They are extremely useful for examinations of intergranular and intragranular defects in ice, as illustrated throughout this book, and the structure of snow presented earlier in section 4.2.3. The limitation of using optical microscopes for examining thin sections of sea ice has led to the development of

165

large-field optical devices or polariscopes and large-field universal stages, such as Rigsby stage, capable of handling sections of ice cores with diameters of 100 mm or specimens with dimensions up to 300 mm in diameter. Such devices also allow photographing thin sections larger than that allowed for recording with usual optical microscopes. Polariscopes are simple optical devices with large viewing areas that can handle significantly large specimens compared to those used in optical microscopes. The name polariscope is also referred to the ability of the instrument to see the “poles” or the optical axis of individual crystals in a specimen. Polariscopes are made by making use of commercially available large sheets of polarizing or Polaroid filters. It is preferred to cut the sheets as rectangular pieces with their lengths parallel to the pass direction or the direction of polarization (as illustrated in Figure 4.3). Most camera lenses can also be fitted with the commercially available polarizing filters with rotating mounts. In case of a polariscope equipped with linearly polarizing sheets, the outer rings holding the two polarizers should be marked with lines indicating their pass direction. These markings allow a very quick method of setting the two polarizers either in cross or parallel position. In the field of glaciology, the use of cross-polarizers is very popular and used almost universally for photographic illustrations of ice structure. “Large-field” polariscopes can be made relatively easily and cheaply by using commercially available plastic sheets of polarizers. Large-field of view is defined here as large viewing area so that thin sections of ice containing a large number of grains (minimum of 100) can be seen and/or photographed at the same time. Another loose definition could be a polariscope that can handle thin sections of ice cores with diameters up to, say, 300 mm. The transmitted light beam in traditional polarizing microscopes used to be polarized by making use of Nicol prisms. Since the Nicol prisms are made from quartz crystals, the available size of such crystals was the limiting factor for viewing area of microscopes. However, the situation changed drastically during the 1930s. Edwin H. Land championed the properties of dichroic crystals and patented a rather revolutionary technique in 1929 for making plastic sheets of polarizers, known as “Polaroid” filters. The H-sheet Polaroid, invented by Land in 1938 [see Land, 1951] are made from polyvinyl alcohol (PVA) polymer impregnated with iodine. The long winding chains of PVA polymer are stretched during manufacturing of sheets such that they form an array of aligned, linear molecules in the material. The aligned PVA molecules become conducting along the length of the chains when the iodine dopant attaches to them. Consequently, light polarized perpendicular to the chains is transmitted whereas light polarized parallel to the conducting chains is absorbed. Plastic polarizing sheets are

166

SEA ICE SLR Camera

Analyzer Specimen holder Goniometer

Light controller

Polarizer

Light box

Figure 4.11 National Research Council (NRC) of Canada polariscope for observations of circular thin sections of ice with diameters up to the maximum dimension of 300 mm (photo by N.K. Sinha).

now used extensively for a number of wide-ranging applications, such as sunglasses, optical microscopes, and liquid crystal displays like the screens of laptops and TVs. Commercially available PVA-iodine filter has very good polarization efficiency (99.9%) and almost neutral in color and does not distort the color significantly. Availability of large PVA panes of linearly polarizing filters made it possible to increase the field of view of polarizing microscopes with diameters of a few millimeters to large diameters. Grains (most often aggregate) of snow may be small individual crystals, possibly significantly smaller than 1 mm. However, those in natural ice are usually larger than a few millimeters and, therefore, significantly larger in comparison to those in most minerals. Consequently, for mechanical tests on ice, such as creep or strength tests, ice specimens have to be large enough to contain a sufficient number of grains. Use of rectangular specimens with dimensions of 250 mm × 100 mm or cylindrical specimens with diameters of 100 mm, made from ice cores, are common. A special polariscope system that can accommodate thin sections of ice with dimensions of 100 mm × 250 mm was designed by the second author of this book and fabricated at the National Research Council (NRC, in Ottawa). This instrument was rugged enough to be taken to the field. In point of fact, the NRC polariscope had traveled to many field laboratories in the Arctic, such as Resolute Bay, Mould Bay, Hobson’s Ice Island,

Figure 4.12 Goniometer or the universal stage (Rigsby type) to sit on the freely rotatable specimen holder of the NRC polariscope shown in Fig. 4.10 (photo by N.K. Sinha).

man-made offshore ice island, drilling platform floating on top of the North Magnetic pole, onboard the icebreaker MV Arctic, and air bases of Thule and Nord in Greenland. Actually, this polariscope was used for taking most of the polarized-light photographs illustrated in this book. Two different views of the NRC polariscope are shown in Figure 4.11. It consists of a light source composing an array of incandescent bulbs for better distribution of radiation, a light diffusing plate, two linear polarizers, a specimen holder and a holder for the universal stage (goniometer). The goniometer (Rigsby-style) stage is shown in Figure 4.12. All the components of the

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

polariscope are free to rotate 360 for the determination of fabric diagrams (see section 5.3.4). The camera mount is capable of handling 35 mm as well as large-format profession view cameras and consists of a carriage that allows rotation and adjustments of the position of the camera with respect to the object to be photographed. A builtin dimmer switch allows adjustment of the light output. To prevent infrared radiation affecting the specimen (specially sea ice thin sections), it is actually preferred to use low level of light during the adjusting periods for photography or performing measurements for fabric diagram. Narrowband or monochromatic filters can be used in the camera for increasing distinctions between individual crystals in different shades of the used color and for better resolution of the grain interfaces and inclusions inside and between crystals. Figure 4.13a shows the NRC polariscope inside the field laboratory of Mould Bay and an image of a large, 100 mm × 250 mm thin section of deformed and recrystallized sea ice is shown in Figure 4.13b. Note the cross on the left of the thin section in this photograph. The cross gives an indication of the polarizer and the analyzer and that they are in cross position. This polarizer indicator was developed by the author (N.K. Sinha) using the principle of hoops stress in circular plates. For 35 mm Single-Lens Reflex (SLR) cameras, linearly polarizing filters can be purchased from most

(a)

167

photographic stores. Since reflections from floors or table tops are elliptically polarized with strong components in the horizontal plane, a linearly polarizing filter can be distinguished readily by rotating the filter while looking at the reflected light through the filter. Make certain to buy “linearly” polarizing filters and not the “circularly” polarizing filters because the sales persons may not know the differences between these two types of polarizing filters. The auto-focus and light-metering sensors of some modern cameras may not function reliably with linearly polarizing filters. For ice, the two commonly used positions are the cross and the parallel polarizing positions. Such configurations can only be obtained with linearly polarizing filters. These filters come in the form of concentric rings. The outer ring (male) has threads on the back for mounting the filter on the camera by screwing it in front of camera lenses. The inner ring holds the polarizing filter and can rotate freely inside the male through 360 . For easy identification of the pass orientation, the axis of the polarization of the filter is marked with a single dot or a pair of dots, diametrically opposite to each other, on the front of the inner ring. While viewing through the eye piece of the camera, the inner ring can be rotated to get the cross position or any desired position with respect to the polarizer. When viewing a thin section using a polariscope, the individual crystals can be seen in different colors, with

(b)

C

100 mm Figure 4.13 (a) N.K Sinha inside Mould Bay field laboratory using NRC polariscope equipped with high-resolution 4 x 5 view-camera and (b) a 100 mm x 250 mm DMT thin section of S3 type sea ice with average c axis along the long dimension, after confined compression tests, exhibiting recrystallization; the arrow with c indicates the direction of c axis and water current in Mould Bay; the cross on the left indicates the orientation of the crosspolarizers.

168

SEA ICE

the best view revealed when using crossed-polarizers as explained earlier. Since the two polarizers are free to swing in the NRC polariscope, they can be moved out of the light train and the specimen can be illuminated from the side. The side light is scattered by the inclusions in the ice and can be used for taking scattered-light pictures. This sidelight (without the main light of the polariscope) enhances the appearance of air and brine inclusions in sea ice, but does not show the details of the grain structures. An alternative method of viewing is to combine the crossor parallel-polarized light in conjunction with the scattered light. This is described in section 4.3.3. It may not be possible to carry a large polariscope everywhere to the field. For these situations a simple method can be adapted for quick-and-easy evaluation of ice types, if necessary, using a hand-held polariscope. This polariscope consists of a 35 mm SLR digital camera and two linearly polarizing filters that can be purchased from most camera shops. The filter mounted on the camera can act as the analyzer whereas the other filter can be used as the polarizer. In case the second polarizing filter is not available, some possible options are open for consideration. The best option is to use empty white screen (LCD monitor) of a laptop, or other devices that uses such monitors, as the source of linearly polarized light. LCD monitors provide uniformly lit source of linearly polarized light with axis of polarization usually at 45 to the horizontal. The intensity of most laptop screens can also be varied to some extent. Use of manual focusing and aperture/time adjustments in case the auto-focus and light-metering system of the Digital SLR does not work. Since reflected light, as has already been discussed in section 4.1.1, is elliptically polarized with major axis in the plane of reflection, light from any source after reflection from a flat non-metallic surface such as glass or plastic can be used as a “poor polarizer.” For a horizontally oriented glass plate, an angle of incidence of about 56 from the vertical (the Brewster’s angle) is the best. Another option is to use the sky light. In the Arctic or the Antarctic, particularly during the clear polar days, the light from blue sky at 90 to the direction of the Sun can be used as a poor polarizer because scattered-light from the sky is strongly polarized. The palm-technique for making thin sections (see third paragraph in the opening of section 4.2), in conjunctions with light from a vehicle, two polarizing filters and a 35 mm Digital SLR camera was actually used by Sinha [2004] for examining the type of ice on runway at Erding Airbase in Germany during a series of friction tests under the Joint Winter Runway Friction Measurement Program (JWRFMP). This was an international collaborative project to conduct field tests using different types of instrumented aircraft and ground friction measuring vehicles. As such, microstructural analysis of the winter

contaminants on the runway was not part of the original plan and consequently necessary equipment were not taken to the field. JWRFMP was a collaborative program between Transport Canada (TC), the NRC, the Canadian Department of National Defence (DND), the US National Aeronautics and Space Administration (NASA), the US Federal Aviation Administration (FAA), the Norwegian Civil Aviation Administration and several airframe and ground friction measuring equipment manufacturers. 4.3.2. Cross-Polarized versus Parallel-Polarized Light Viewing Too much emphasis is given in the ice literature on the use of cross-polarized light to view thin sections. This is primarily due to the fact that crossed-polarizers offer the highest contrast between colors of crystals with different orientations. However, it should be mentioned that the orientation of the analyzer can be rotated to any polarization direction with respect to the polarizer. One particular direction is the parallel polarizers. This combination offers the lowest contrast in interference colors and consequently does not produce spectacular colors. For that reason, little attention has actually been paid to the use of parallel-polarized light for viewing thin sections of ice. Nonetheless, the parallel polarizers have unique advantages for the analysis of microstructures of ice with inclusions. An example of a comparative case between cross-polarized and parallel-polarized images is given in Figure 4.14 to show the usefulness of the latter viewing method in conjunctions with the former one. In cross-polarized images, crystals with their c axis parallel to either the axis of orientation of the polarizer or the analyzer, and/or parallel to the direction of light propagation appear as black. These three positions are known as positions of extinction and are, in fact, the basis of determining texture and hence fabric diagrams of ice (section 5.3.4) using universal stages. When a crystal (grain) is dark or black, the detailed intragranular structures, such as bubbles or cracks, become invisible. This can be seen if the dark crystals in the left image of Figure 4.14 are compared with those same crystals in the image on right. The relatively lighter interference colors in most of the crystals in the right (parallel-polarized) image reveal the size, shape and orientations of air/gas bubbles entrapped in the shelf ice during the consolidations of snow particles or small intragranular (confined within the crystal) and intergranular (at the grain or crystal boundaries) cracks. Healed cracks with their planes parallel to either of the two main cleavage planes of ice could be detected by the presence of row of air bubbles. Most air bubbles were found to be elongated with their long dimensions parallel to the basal plane. These

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

169

Figure 4.14 A 100 mm wide horizontal thin section of shelf ice from a block of floating ice island that was calved from Ward Hunt Ice Shelf, Ellesmere Island, Canada as observed through cross-polarized light (left) and parallelpolarized light (right) (photo by N.K. Sinha).

bubbles, with their dimensions comparable to the wavelength of microwaves, may contribute to the high backscattering observed in radar imaging system data of glaciers and ice shelves. 4.3.3. Scattered Light and Combined Cross-Polarized/ Scattered Light Viewing Saline inclusions in the form of brine pockets and gas bubbles are the primary entrapments in sea ice. Inclusions in the form of air bubbles are a commonly observed phenomenon in shelf ice and freshwater lake/river ice. The details of these inclusions are not readily observed in images of thin sections taken under cross-polarized light, except for the fact they may appear as dark spots. Often the air bubbles and brine pockets are not discernible from each other. Parallel-polarized images, as described above, are certainly useful in delineating the inclusions. An alternative method is observing thin sections using only the diffused/scattered light from a side illumination source at oblique incident angles. Another method, which is often very useful, is the use of scattered light in conjunction with polarized light images (combined method). Unpolarized light from an external source is used as shown schematically in Figure 4.15. The diffuse reflection from a surface of a translucent thin section of ice consists of light scattered from the imperfections of the surface and inclusions beneath the top surface. The scattered light may be partially polarized or not polarized at all, depending on the size of the inclusions. Therefore, only some of the light is allowed to transmit through the analyzer. Consequently, the scattering objects (e.g., inclusions) are visible beyond the analyzer. To view the thin section under scattering light only the cross-polarized light source (coming from the

Analyzer Cross position

Unpola rized light

Ice thin section

Polarizer

Unpolarized bottom light

Figure 4.15 Viewing of an ice thin section in a polariscope using transmitted cross-polarized light together with scattered light due to diffuse reflection of light illuminated from the side at oblique incidence angle (sketch by N.K. Sinha).

polariscope shown in Figure 4.11) should be switched off and the analyzer should be rotated out of the optical train. Examples of the use of scattered light and/or combinations with polarized light are given in many places within this book. A unique application of the scattered-light technique was made in viewing thin sections of ice at the ice–water interface of 1.75 m deep FYI sampled in early March, 1978 from Eclipse Sound near Pond Inlet, Baffin Island (this application is briefly covered below in this section). The goal was to explore the structure of the growing front

170

SEA ICE Sub-grains

c-axis

5 mm

W av ate er r c ag u e rre c- nt ax is

(a)

1>

00






>

Figure 4.30 Optical micrograph of thermally etched grain boundary in the middle separating two grains in S3 type sea ice and subgrain boundaries inside the two grains; orientation of c axis, parallel to the length of the elongated dislocation etch pits, is shown by < c > (micrograph by N.K. Sinha). Subgrain boundary

1mm

Grain boundary

Figure 4.29 Micrographs of a horizontal thin section of FY S3 type sea ice, taken within 20 min after completion of the DMT of thin sectioning (top) and after 96 hours of thermal etching at -10 C (bottom) (micrographs by N.K. Sinha).

Though the top surface of the section was focused, the micrograph provides three-dimensional thickness views of the brine pockets. It shows the boundary area between two adjacent grains as viewed in transmitted polarized light. The bottom image in Figure 4.29 shows thermally etched crystal boundaries and intracrystalline (inside subgrains) features after 96 hours. It illustrates several important aspects of sea ice, such as the complex network of dislocations within the subgrains and/or the orientation of the c axis of the subgrains and hence the grains. The most obvious are the elongated etch pits and the fact that grain boundary grooves are not distinguishable from the grooves along subgrain boundaries. Grain boundaries are the large angle boundary whereas subgrain boundaries are low-angle lines between platelets. No discernible differences in depths of the grooves at the boundaries between subgrains belonging to two grains with large mismatch in the orientation of their c axis, and subgrains inside a grain with little mismatch across their boundaries. Some details of thermally etched dislocation etch pits in S3 type of sea ice are shown in Figure 4.30. The photographs clearly show how the etch pits directly indicate the orientation of the c axis of the subgrains and hence the average orientation of the grains. This optical micrograph also shows a grain boundary going from the bottom

of the picture to the top in the middle of the photograph. This boundary separates the grain in the left half from that in the right half. The subgrain boundaries inside each grain indicate mismatch between the a axis and probably slight mismatch in c axis. The orientation of the c axis is indicated by the long dimension of the elongated etch pits. These pits correspond to the intersections of basal dislocations with prismatic surfaces and will be explained later in detail. One can see here, as well as in Figure 4.29, that it is difficult, if not impossible for the untrained eyes, to distinguish between grain boundaries and subgrain boundaries in sea ice. In a direct way, these images exemplify that grain boundaries in sea ice, unlike freshwater ice, are not the most important features in sea ice. Due to the propensity of subgrains, it is the subgrain size and subgrain boundaries that control the microstructure–properties relationship in sea ice. It is the subgrain boundaries where cracks are also nucleated, with contaminant generation of acoustic emission, under the influence of thermally induced strains or externally applied loads [Sinha, 1996]. Figure 4.31 shows characteristics of the same surface after 72 hours of thermal etching at −10 C. It shows triple points of subgrains with little or no entrapped brine. Note the differences in the orientations of elongated pits inside the subgrains. Note also the details of the shape of the pits. They are wedge shaped with central depression, but reverse looking in the micrograph due to the oblique transmittance of light. The proof that these pits are due to the basal dislocations intersecting the prismatic surfaces has to come from chemical etching to be presented later. It should be emphasized here that thermally etched dislocation pits like these have not been published in the sea ice literature.

182

SEA ICE



0.2 mm

Figure 4.31 Triple points of grains (actually subgrains in neighboring grains) with and without brine pockets and elongated basal dislocation pits, parallel to c axis, after 72 hours of thermal etching at −10 C. Orientation of c axis is shown by < c > (micrograph by N.K. Sinha).

< c >

0.5 mm

Figure 4.32 Thermally etched surface, normal to the long axis of a grain in columnar-grained S3 sea ice, exhibiting brine/air pockets, subgrain boundaries and stepped patterns corresponding to the network of dislocations inside one grain; orientation of c axis is shown by < c > (micrograph by N.K. Sinha).

Figure 4.32 illustrates the detailed substructures within a grain of columnar-grained S3 type of FYI with salinity of 4.3‰ at a depth of 1.83 m, extracted from the Strathcona Sound, Baffin Island, Canada. In this case a thin section, cut normal to the length of the columns, was

prepared first using DMT at −10 C. Note that this relatively high temperature was chosen purposely to have the brine pockets full with liquid. The mirror finished clean surface was then coated with a thin and uniform layer of 3% Formvar solution and allowed to dry quickly under the ambient low humidity (about 5%) conditions of the cold room. In less than an hour, once the film was dry, the section was kept inside a thermal etching box covered with a glass plate and the ice surface was monitored through the dried film under an optical microscope using transmitted light. Within one hour, the etched patterns started to develop. The large-angled grain boundaries (large mismatch in both c axis and a axis) with high-energy levels can be thermally etched within a few hours. Boundaries with lower energy levels, such as subgrain boundaries having very small lattice mismatch in the c axis across the boundaries, take longer durations. This is particularly applicable to sea ice for which there are more subgrain boundaries than grain boundaries. In fact, long-range intercrystalline boundaries with large angles of lattice mismatch, commonly seen in freshwater ice, do not exist in FYI. The physical dimensions of the large-angle boundaries are limited by the dimensions of the subgrains. Moreover, the entrapments of brine (and air) pocket along these boundaries add another level of complexities. Brine pockets are also trapped at numerous intragranular platelet or subgrain boundaries where small mismatch of lattice orientations also occur. These subgrain boundaries are also eventually thermally etched. This is why surface features can be readily seen in freshwater ice, but it takes longer for them to develop in sea ice. Moreover, relatively lower energy levels of subgrain boundaries in sea ice lead to shallower and finer etching features when thin sections of sea ice are exposed to thermal etching. In any case, while using optical microscopes care should be taken to focus on the etched features at the top surfaces of the sections because the features at the bottom surfaces close to the glass surfaces of the plates holding the specimens are also etched. If not properly focused, etched boundaries at the top as well as the bottom are visible, either overlapping each other or side by side (boundaries are rarely vertically down in case of horizontal thin sections), creating unnecessary confusions. Optical microscopy of the prepared surfaces after thermal etching reveals that the distribution of brine pockets is related to the subgrain boundaries or the bridging surfaces of the dendrites formed at the ice–water interface during freezing (see section 2.5.3 and Figure 2.22). Details of brine pockets and the precipitation pattern of the enclosed salts could be observed by replicating the microtomed surfaces and examining the replica with an SEM [Sinha, 1977a]. Most brine pockets are irregular and, when precipitation occurs, the salt crystals are loosely packed in

LABORATORY TECHNIQUES FOR REVEALING THE STRUCTURE OF POLYCRYSTALLINE ICE

183

the cavities in a random manner. Replications should, however, be performed as soon as possible after the completion of the DMT procedures in order to avoid complications due to the surface grooves formed by thermal etching during storage. For saline-free or brackish water ice, thermal etching has been found to be very effective at temperatures at or below about −10 C. For sea ice it is better to apply it, as mentioned before, for temperatures below −20 C. In polycrystalline ice, apart from defects such as grain and subgrain boundaries there are point defects such as interstitials and line defects called dislocations, where atoms are out of position in the crystal structure [Taylor, 1934] (see definitions in section 5.1.3). Dislocations are responsible for the plastic deformation of material and expected to end at subgrain boundaries. The growing of ice crystals at the ice–water interface may result in the creation of dislocations. It is possible that defects are individual (each defect is confined to a point) but in this case they are not dislocations. The controlled thermal etching techniques were applied to examine the finer aspects of the distribution pattern of dislocations inside a grain of ice, as depicted in Figure 4.33. A photomicrograph that reveals brine pockets located along mismatch subgrain boundary is shown in Figure 4.33a and another one showing an isolated inclusion, not linked to any boundary, is shown in Figure 4.33b. Aside from those inclusions, these two photos also reveal the intestine-like alimentary canals of

a complex network of basal dislocations in the ice lattice presumably generated inside the dendrites during the period when the tips of the dendritic arms protrude in the water. As the growth front progresses and bridging between the arms are created, and brine and air inclusions are trapped, the complexities in the pattern are amplified. It is worth noting that intersections of line and surface defects cannot be usually revealed by thermal etching. Chemical etching is extremely powerful and very appropriate for delineating dislocations in ice in the form of etch pits at surfaces as well as nanoscale deep holes along the cores of dislocations. Consequently, mechanically or thermally induced mobility of dislocations or network of these line defects can be detected without any ambiguities.

(a)

(b)

4.4.2.2. Chemical Etching and Replicating Ice Surfaces In case of metals and ceramics, chemical etching is achieved by the application of a solution that allows absorption of high-energy surface atoms and molecules, leaving their spots vacant in the lattice structure. As explained earlier, replicas can be made by applying replicating solutions consisting of some kind of plastics dissolved in solvents. Replication can produce imprints of the surface and the etched surface features can be examined using an optical microscope or an SEM. Due primarily to the unavoidable fact that ice exists at temperatures very close to its melting point, replicating ice

>

20 mm, respectively. Superimposed ice always forms on top of the primary ice and is caused by flooding of an ice cover from any possible water source (e.g., rain, freezing rain, or snowfall). If the snow melts or becomes wet then freezes on the ice surface, then snow ice may form. In addition of oriented needle-shaped frazil ice, type S5 is suggested for inclusion in the above table. It was not part of the original classification by Michel and Ramseier [1971].

202

SEA ICE

Primary ice is the first layer of ice or the horizontal skim layer of an ice cover. In turbulent water it may consist of frozen frazil slush, which can be very thick. If nucleation occurs by snow, the resulting congealed snow slush would also be part of the primary ice. Secondary ice forms under the primary ice and grows parallel to the direction of heat flow, which for floating ice is in the vertical plane. Its structure is mostly in the form of columnar ice, the texture of which is entirely controlled by the primary ice. It can also be in the form of frazil slush or snow slush, deposited under a primary ice layer which after some time may become an integral part of the ice cover. Superimposed ice is usually formed on top of the ice sheet (i.e., on top of the primary ice) when wet snow, rain, surface melt freezes. It is a layer of new ice that has no connection to the underlying ice and it may grow from the bottom up. 5.3.2. Extending Crystallographic Classification of Freshwater Ice to Sea Ice In the Arctic and Antarctic, the near shore water may have a salinity of about 30‰. This is close, but measurably lower than the average sea water salinity of 35‰. The lower water salinity in the polar regions is due to: (1) the less evaporation and (2) the freshwater that feeds into these areas from different land sources including rivers, snow melt, glaciers, etc. The underside of an ice cover in contact with water remains close to the freezing temperature or slightly colder depending on brine drainage, while the exposed upper surface can be close to the ambient air temperature. However, here is much wider range of environmental and oceanic conditions, above and below the sea ice covers, compared to what exists in lakes and rivers. Moreover, the growing seasons for sea ice is also significantly longer and melting periods shorter than most lake and river ice at lower latitudes. During the growing season for sea ice, depending upon the latitude especially for the Arctic Ocean and the High Arctic, there can be total absence of solar radiation or 24 h of Sun during the summer months when the ice melts. The duration and intensity of solar radiation in the polar region add another aspect of the difference between growth and decay of sea ice in those regions and lake ice in southern regions such as the Great Lakes in the North America. Of course, ice covers in rivers and lakes (even in the High Arctic) are not known to survive beyond the winter season, but that is not the case for sea ice. Physically speaking, therefore, there are differences between freshwater ice and sea ice. Compared to the types of ice that form in lakes and rivers there is more diversity in sea ice formation and its aging processes. However, the classifications of freshwater ice in Table 5.1 can be extended to ice forming in the oceans as long as the

diversities and complexities due to the entrapped salts are recognized. Differences between freshwater ice and sea ice are summarized here with emphasis on the crystallographic classification of freshwater ice in Table 5.1 that can be extended to sea ice. Although the bulk of floating freshwater ice can be divided into five basic types: S1, S2, S3, S4, and S5, the S1 type has never been reported for sea ice. S1 type of columnar-grained ice, with their c axis in the vertical plane, is expected to form under calm conditions with snow deposition activities during the initial ages of freezing. Careful examinations of the snow-free surface layers of freshly frozen sea ice in Mould Bay, Prince Patrick Island, however, revealed no signs of nucleation and growth of S1 type of sea ice at the surface level [Sinha, 1984]. It was concluded that the characteristics of S1 ice would not be maintained beyond the top 1–3 mm in sea ice. Moreover, sublimation at the surface level in conjunctions with the wind removes this thin layer as the ice cover continues to grow. Consequently, for all intents and purposes, sea ice can be assumed to have no S1 type of structure (see more on initial growth structure of sea ice in Mould Bay series of experiments in section 6.2). Type S2 is the transversally isotropic columnar-grained ice with the c axis of the grains randomly oriented in the horizontal plane or normal to the growth direction. It can form in areas not affected by any water current. Type S3 is columnar-grained with the c axis of the grains also in the horizontal plane, but tends to be oriented in a particular preferred direction. The bulk of the landfast ice usually consists of S3 type of ice. Peyton [1966] described the oriented sea ice (in sea water around Alaska) in detail and called it “Bottom Ice.” Preferred crystal orientation in bulk sea ice was well known to the Russian scientists, but their speculations as to the reasons for the formation of such ice types were never confirmed. A detailed discussion on this topic can be found in Weeks [2010]. It was shown later that the crystal orientation of this ice type is parallel to the water current under the ice sheet during the growth period. This observation was confirmed by Weeks and Gow [1978] who found that the c axis of ice in the Beaufort Sea was oriented parallel to the water current under the ice. At about the same time Sinha [1977] also observed the same in the Strathcona Sound in the Baffin Island. Type S4 ice is prevalent in many areas and often in large quantities. There are indications that this type of ice and/or mixed with unfrozen dense slush could be as thick as several meters. Icebreakers have been reported to be unable to maneuvre in sea water containing this type of conglomerate. They exist not only below the secondary ice as described in Table 5.1, but also above it. Moreover, the crystals are usually needle-like or crushed needles rather than tabular, and the structure may not

POLYCRYSTALLINE ICE STRUCTURE 203

only be equiaxed but also oriented. The structure of oriented frazil ice is already presented earlier and will be classified as S5-type ice. For general descriptions in conjunctions with the crystallographic classification, sea ice can be grouped into three major categories: granular, frazil, and columnar. Sinha [1991] suggested adaptation of names on the basis of classification for freshwater ice as deemed appropriate. Characteristics of each category of freshwater ice and sea ice are introduced in the following sections. The presentation does not cover the crystallographic classes in the same order as listed in Table 5.1. Instead, it starts with the three categories that are commonly found at the top of sea ice (T1, S4, and S5), then proceeds to the categories that make the bulk of sea ice (S1, S2, and S3). It concludes with the agglomerated type (R ice type). Several images of ice thin sections photographed between crossed or parallel polarizers are presented. More information on the thin sectioning DMT is presented in section 4.2.2, and on photographing thin sections in section 4.3.2 and 4.3.3. Before concluding this section on crystallographic classification of sea ice it is worth noting an exotic ice class associated mostly with Antarctic ice. It is called “platelet ice” by the investigators who, perhaps, were not familiar with the existence of the classification system already developed for ice. This is congealed frazil ice found at the bottom of the sea ice or ice shelves. Platelet ice was first reported (without given its name) in McMurdo Sound, Antarctica, during an expedition in 1904–1905 [Hodgson, 1907]. The numerous ice shelves around the Antarctic melt at the bottom, making the sea water slightly less saline. This water rises under buoyancy forces and spills out under the coastal sea ice. The supercooled water contains frazil ice crystals that accumulate and grow at the sea ice–water interface, forming the platelet ice layer (Figure 5.5). Therefore, a major difference between platelet ice and regular sea ice is that the former is made of less saline sea water and originates in supercooled deep water, while the latter originates at the seawater surface. A photograph of thin section depicting crystallographic structure of platelet ice is presented in section 5.3.3.8. Although platelet ice affects the crystallographic structure of sea ice, it does not have impact on its physical properties except for determining the mass balance. However, Hoppmann et al. [2020] asserts that it serves as an indicator of the extent of ice-shelf basal melting, which caused formation of frazil ice that constitutes the platelet ice. Moreover, platelet ice can host a productive ecosystem. The same reference listed other studies that highlight the importance of platelet ice from the viewpoints of glaciology, geophysics, remote sensing, and numerical modeling. Langhorne et al. [2015] used platelet ice distribution in the Antarctic water as an index for ocean-ice shelf heat flux.

Columnar sea ice growth

Consolidation of frazil ice (platelet ice)

Buoyant uprise of frazil ice

Figure 5.5 Schematic illustration showing formation of platelet ice attached to the bottom of an ice sheet. Platelet ice crystals are frazil ice formed from supercooled water resulted from basal melting of ice shelves. It is more common in the Antarctic around areas of ice shelves [adapted from Hoppmann et al., 2020].

While platelet ice is a term usually coined with Antarctic ice, a recent study [Katlein et al., 2020] presents, for the first time, in-situ observations of decametric thick platelet ice in the Arctic during the MOSAiC expedition (see section 6.10). The study used under-ice diving robot to confirm observations of platelet ice during Arctic winter in supercooled water. In general, platelet ice has not been extensively observed in the Arctic as the conditions of formation, namely from supercooled water resulting from ice shelf basal melt, are not usually encountered. The location where the icebreaker Polarstern was frozen in the Arctic Ocean during MOSAiC expedition was 85 N and 134 E is far from the closest ice shelves located along the northern section of Greenland and Ellesmere Island. Therefore, the possibility of the shelf-induced supercooled water was remote in this case. It is possible that what was considered platelet ice sticking at the bottom of the ice cover at the ship location was just frazil ice carried upward by water current. 5.3.3. Crystallographic Classes of Natural Ice 5.3.3.1. Granular or Snow Ice (T1 Ice) Below the blankets of snow, often granular ice (also called snow ice) is found at the top of ice covers in rivers, lakes, or oceans. Granular ice features randomly oriented c axis with rounded grains with diameters in the range of about 1 mm (Figure 5.6). It is equivalent to the T1 ice described in Table 5.1. It may form from a number of processes

204

SEA ICE

including freezing of wet snow or snow mixed with ice particles (e.g., fractured frazil crystals) when saturated with sea water in a slushy medium. Crushing of ice covers or deformation-induced solid-state transformation or recrystallization of secondary columnar-grained ice can also provide fine-grained particles to nucleate the granular ice type. Granular or snow ice forming in sea water traps not only air but also a measurable amount of brine (Figure 5.7). These inclusions are usually located at the grain boundaries. The density of this ice can be relatively high (could be 925 kg/m3) within a few percent of that of single crystals with density of 918 kg/m3, depending upon the amount of trapped air and brine. Macroscopically, this type of ice is isotropic (Figure 5.7) due to 3D randomness of the crystallographic orientation of constituent grains. However,

c

c c

T1 type

Figure 5.6 A three-dimensional (3D) schematic diagram of isotropic granular (snow) T1 ice (sketch drawn by N.K. Sinha).

10 mm

Fresh water snow ice

granular ice that may form from sintering of snow and crushed ice, as found in the upper sections of glaciers, ice caps, and ice shelf could exhibit marked anisotropy in shape and size of the grains and their crystallographic orientation. It should also be noted that, in contrast to the commonly observed fine-grained snow ice in freshwater ice, practically isotropic shelf ice could exhibit significantly large grains as shown in Figure 5.8. Consequently, ice islands calved from ice shelves and floating in sea mixed with new and old sea ice, can have layers of large grains and nearly equiaxed in characteristics with small anisotropy [Barrette and Sinha, 1996]. 5.3.3.2. Randomly Oriented (S4) and Vertically Oriented (S5) Frazil Ice Frazil ice consists of disk- or needle-shaped crystals suspended in the water. Unless anchored, the crystals usually form a thin, oily or translucent-looking film that tends to float to the water’s surface or suspend near the bottom surface of existing ice cover. When atmospheric temperature drops, the crystals clump together to form a dense slush saturated with water. Frazil ice develops from the consolidation and freezing of this water saturated slush containing needles of ice. Initially frazil crystals form in rapidly flowing or turbulent body of supercooled water. Frazil crystals result from the fragmentation of small and thin disk-like crystals (section 2.3.2). The disks develop as a result of the lateral growth of the nucleated crystals in the basal plane (discoid grow at a higher rate in the lateral direction). The thickness-to-width ratios of the discoid vary greatly, in the range of 1:5 to 1:20 or more [Hobbs, 1974]. Grains in frazil ice reflect the characteristics of the needles and hence the fragmented discoid. In turbulent waters, the geometric orientations of the

10 mm

Sea water snow ice

Figure 5.7 Granular T1 type of ice formed from (left) freshwater and (right) seawater as seen in thin sections produced using the DMT technique, and photographed between crossed polarizers [Sinha et al., 1996 / John Wiley & Sons, Inc.].

POLYCRYSTALLINE ICE STRUCTURE 205

AIR BUBBLES

(a)

(b)

C

C S4 type

S5 type

Figure 5.9 Three-dimensional schematic diagrams of (a) randomly oriented congealed S4 type frazil ice and (b) vertically oriented S5 type frazil ice with the long axis of the crystals in the vertical plane; arrows indicate the orientation of the c axis.

10 mm Figure 5.8 Nearly isotropic granular large-grained sintered snow ice, with elongated air bubbles trapped inside the grains, from Ward-Hunt ice shelf, Ellesmere Island, Canada (photo by N.K. Sinha).

needles could be random. If consolidation and eventual solidification occur, the frozen body maintains the randomly oriented crystal structure of the slush. This type of congealed frazil slush has been classified as S4 type by Michel and Ramseier [1971] (Table 5.1). If the mass of frazil crystals carried away by the water current is blocked by any existing structure or an ice body, the needle-shaped particles accumulate on the upstream side. Upon freezing, vertically oriented frazil ice is formed with isotropy in a plane normal to the current and the long axis of the grains tends to be vertical. This causes the c axis of the grains to be oriented in a plane normal to the current, or in the horizontal plane [Sinha, 1986]. Tidal activities may lead to the formation of oriented frazil ice in long fjords. Ocean current may also carry frazil to accumulate at the underside of ice sheets. This particular type of frazil ice may be categorized as S5 as shown in Table 5.1. Schematic diagrams of randomly oriented S4 type frazil ice and vertically oriented congealed S5 type frazil ice are shown in Figure 5.9. Since the c axis in the needle-shaped frazil crystals is normal to the long axis of the crystals, the overall orientation of c axis is random in S4 type of frazil ice. Because the crystals with their long axis tend to be vertically oriented in S5 type of frazil ice, the average orientation of the c axis is in the horizontal plane or parallel to the surface of the ice. Vertically oriented FY frazil sea ice (now classed as S5 type) was found to cover the shore-fast ice attached to

the eastern shoreline in Mould Bay, Prince Patrick Island, Canada during the early winter of 1981 (see section 6.2.1). The tide comes from the south and thereby induces northsouth water current in the bay. However, it was not the tidal current, but the westerly strong wind (recorded at the weather station) that pushed the frazil crystals toward the eastern shoreline where they congealed and eventually solidified. Microstructural studies of this ice including examination micrography images using SEM, and the rate sensitivity of compressive strength in a wide range of strain rate, were carried out. An example of the structural features seen in vertically oriented S5 type of frazil sea ice is shown in Figure 5.10. Since the discovery of this type of S5 ice in 1981, together with its recorded growth history, this type of ice has been noticed and recognized in FYI and old multi-year sea ice in many places in the Arctic. In agreement with the observations of Sinha [1986] that strong and persistent winds caused herding of oriented frazil ice, Lawson and Brockett [1990] stated that frazil deposit characteristics reflect processes and properties of their source such as depositional mechanisms and the history of ice sedimentation. Hence, their appearance in thin section photographs can also be interpreted in terms of deposit evolution and used as proxy data on the nature of subice deposition. 5.3.3.3. Columnar-Grained with c Axis Vertical (S1 Ice) As mentioned earlier, S1 type is not common in sea ice. A freely floating ice crystal grows more rapidly in any direction in the plane parallel to the basal plane or normal to the c axis. This is the situation if nucleation of ice crystals occurs at the surface of a large body of water under calm conditions with no deposition of snow. The individual crystals grow rapidly in all directions along their basal planes at the surface level until the entire surface is

206

SEA ICE Vertical section

Horizontal section

2 mm

2 mm

Figure 5.10 (Left) vertical and (right) horizontal thin section of vertically oriented S5 type of frazil sea ice, with c axis randomly oriented in the horizontal plane, photographed between crossed polarizers [Sinha et al., 1996 / John Wiley & Sons, Inc.].

c

ϕ

a Grain

c

c

a axis

a

a

xis

(b)

aa

(a)

a axis Subgrain

S1 type

c axis

c axis

Figure 5.11 Schematic diagram of columnar-grained S1 type ice with c axis in the (a) vertical plane and (b) subgrains inside grains due to small mismatches both c axis and a axis (sketch by N.K. Sinha).

covered with ice. Solidification of the entire surface layer stops the horizontal growth. Since the grains are constrained by the boundaries of the neighboring crystals, they have no choice but to grow downward. At this time, each crystal is forced to grow in the direction of maximum heat flow and in a direction parallel to their c axis. This leads to vertically oriented columnar-grained crystals with their long axis parallel to the direction of growth that coincides with the maximum heat floe (i.e., vertical direction). As the ice cover thickens, the c axis of the columnar grains remains in the vertical plane parallel to the length of the grains. The bulk of the ice sheet thus exhibits marked anisotropy with large cross-sectional dimensions. Due to the large grains, the ice also looks very clear and

transparent. This type of ice mass was classified as S1 type as illustrated schematically in Figure 5.11a. It is worth repeating that this ice type requires growth under calm meteorological and oceanic conditions, which sounds far-fetched. Purity of the water determines the quality of S1 type ice because it allows crystals to grow into relatively bigger sizes with large planar boundaries as there is no “resistance” to the growth in the horizontal plane during the initial growth. Crystals can have their cross-sectional diameters as large as 200 mm. It will be seen that the large cross-sectional areas of the grains allow the formation of substructures (subgrains) with small mismatches (mainly in their a axis) in the orientation of the ice lattice. The

POLYCRYSTALLINE ICE STRUCTURE 207

10 mm

Figure 5.12 Horizontal thin section of freshwater columnargrained S1 ice showing a single large grain with incursion of another crystal (the dark object) [Sinha et al., 1996 / John Wiley & Sons, Inc.].

characteristics of subgrains are illustrated schematically in Figure 5.11b. A photograph of a thin section, placed between cross polarizers, exhibiting a large grain of S1 ice with incursion of a second crystal is shown in Figure 5.12. Grains in S1 ice rarely compose a single crystal. As the grain grows, local disturbances develop mismatches in both the a axes and c axis within the grains. Subgrains are formed as a result. Slight variations in colors and very small changes in the shades can be noticed inside the large grains of this ice under polarized light. While presenting S1 ice type, it is informative to report some interesting results from a laboratory study on onedirectional freezing of distilled water under still conditions conducted by Perey and Pounder [1958]. They analyzed thin sections of the ice using a polariscope with universal stage (see descriptions in section 4.3.1) and determined the percentage of the area having crystals with polar angles (defined as the angle between c axis of the crystals and the vertical plane) within the range 0 –10 . The authors observed that 68% of the ice area near the surface was covered with grains with polar angles in that range. This, however, dropped to only 13% at a depth of about 130 mm. It is obvious that they did see S1 type of ice at the top, but the growth habit changed as the ice thickened. The observations led one to believe that irrespective of surface nucleation mechanism, ice would eventually grow with c axis of the crystals tending to be in the horizontal plane. This interpretation from the laboratory tests led Pounder [1965], when using his observations on seasonal sea ice, to conclude that all the ice crystals at depths below about 200 mm had an essentially horizontal c axis. In his book, Pounder also made several important remarks on natural sea ice: (1) The bulk of ice covers formed under calm conditions consists of crystals with horizontal c axis. (2) The c axis is randomly oriented in the horizontal plane (i.e., S2 type). (3) The tidal current and unusual situations like

rafting and pressure ridging occurring during growth affect the structure. Pounder [1965] also provided an explanation on how an ice crystal grows more rapidly in the basal plane than along the c axis and thereby proposed a mechanism for preferred orientation of c axis in the horizontal plane. Schematically, the authors showed the preferred growth of crystals with their c axis inclined to the vertical plane leading to the extinction of the crystals with vertically oriented c axis. This mechanism is now widely accepted. In natural freshwater lakes and rivers, the crystals in S1 ice do maintain a continuity of vertical c axis through the usually observed depths up to about 1 m. However, one clarification should be made here. If a natural ice cover is found to be clear, it is correct to assume that the growth occurred under still conditions, but that does not necessarily mean that the nucleation at the surface also occurred under clear atmospheric conditions and that the ice is of S1 type. Clear ice could also be S2 type. A light snowfall under calm conditions just before the start of freezing in lake waters can trigger nucleation of ice crystals at the surface and this could lead to S2 type ice. The type of clear ice should never be assumed without conducting a microstructural examination. Michel and Ramseier [1971] performed an extensive survey of the ice conditions in the ship channel of St. Lawrence River for the Department of Transport, Canada. This shipping channel handles all the overseas ships destined for Montreal and the Great Lakes between Canada and USA. They noticed that S1 ice type was a common occurrence in the channel. This type can, therefore, develop in rivers as well as in lakes although it is commonly seen in lakes [Ashton, 1986]. Unfortunately, clear lake ice is often assumed to be S1 type. Conversely, often S1 ice is wrongly called lake ice. The novice should remember that S1 ice and lake ice are not synonymous, by any means, and should never be used as such. 5.3.3.4. Columnar-Grained with c Axis Horizontal and Random (S2 Ice) One of the most common types of river and lake ice features vertically oriented columnar-grained material with c axis of the grains in the horizontal plane but randomly oriented within this plane. This is classified as S2 type and a schematic diagram of this type of freshwater ice is shown in Figure 5.13. The growth of S2 ice occurs after the nucleation of crystals at the water surface by falling snow under relatively calm conditions without any turbulence in the water mass and may extend across the entire water surface. Figure 5.14a shows a schematic of the nucleation of ice crystals at the water surface and how an ice cover may develop after deposition of snow on the surface with surface temperatures close to the freezing point. New crystals are nucleated and each of them starts to grow downward

208

SEA ICE

in the direction of maximum heat flow as shown in Figure 5.14a. However, not all the nucleated crystals can grow freely. The growth becomes preferable depending on their crystallographic orientations with respect to the heat flow and hence the surface. If a crystal is nucleated by a snowflake falling on the water surface with its flat surface oriented in the vertical direction so that the basal plane is parallel to the growth direction, then that crystal is in the best position to grow. This is illustrated in Figure 5.14b, which shows a layer of crystalline

c c

c c

S2 type

Figure 5.13 A schematic diagram of vertically oriented columnar-grained S2 ice with the c axis randomly oriented in the horizontal plane, parallel to ice sheet surface [Sinha, 1989].

(a)

competitions naturally developing under the snow deposit. Since the growth is preferable in the basal plane, nucleated crystals with c axis in the horizontal plane, irrespective of the orientation of the three a axes, become the preferred grains to grow in comparison to others. Crystals with c axis titled to the horizontal can still grow but their growth is interrupted by the faster growth of neighboring crystals with the preferred horizontal direction of the c axis. Thus, the neighboring crystals “compete” against each other as they grow, leading to the survival of the crystals that have their c axis horizontal or closest to horizontal. In other words, those crystals simply choke the others from growing. This process is known as “geometrical selection” of crystal growth and was critically investigated by Perey and Pounder [1958] at McGill University, Canada. They found that the competition is often concluded within about 10 to 15 mm of growth and clear ice with narrow elongated air/gas bubbles along the grain boundaries develop thereafter. The presence of transversely isotropic columnargrained S2 ice in the bulk of an ice cover is a clear indication that the freezing was initiated during a calm period when the air temperature was low yet with snowfall activities. The snow deposition must be light to heavy, so that after some initial melting, the water surface temperature was decreased to the freezing point. The cross-sectional grain size of S2 ice ranges from fine to large with usually straight boundaries. The cross-sectional area increases slightly with the increase in depth due mostly to the continuation of the selection processes. However, as the freezing front progresses, dissolved impurities including air are also pushed down. For freshwater ice, nearly transparent

(b)

Snow settling

< c

< c

< c < c

Growth competition

Ice





Elongated air bubbles

Water Figure 5.14 (a) Nucleation of crystals by snow deposition and geometric selection leading to columnar growth of S2 ice (courtesy Paul Barrette); (b) details of competition zone and domination of crystals with c axis in the horizontal plane (irrespective of dependence on a axes) (N.K. Sinha).

POLYCRYSTALLINE ICE STRUCTURE 209

ice covers (below the white snow ice, if any), composed of columnar grains start to exhibit air bubbles, often in the form of long cylinders, below a certain depth. Examples of a horizontal and a vertical section of freshwater S2 ice are shown in Figure 5.15 (Sinha et al., 1996). In this case, the pure and transparent ice was made in a laboratory using double-distilled, deaerated, and deionized water [Sinha, 1978]. Randomness in the c axis orientation of the crystals in a plane normal to the long axis of the grains can be visualized from the even distributions of “dark or black” and “bright or white” grains in crosspolarized image, similar to the example shown in Figure 5.15a. The inclusion-free clarity of the intragranular areas of the grains is particularly noticeable. It is appropriate to point out here that the clarity in the characteristics of the grain boundaries and surfaces of the absolutely freshwater ice, however, is maintained primarily because of the use of cold-state, double microtoming technique (DMT) in making the thin sections (section 4.2.2). Conventional hot-plate techniques for making thin sections are convenient, but leave residues along the grain boundaries as well as on the surfaces of the grains due to localized melting and refreezing. The distributions of dark and bright objects in horizontal sections of transversely isotropic S2 ice under crosspolarized light, such as Figure 5.15a, should change

(a)

evenly if the thin section is rotated while keeping the cross polarizers in a fixed position. The grains with their c axis parallel to the pass direction of either the polarizer or the analyzer (see section 4.3.1 for details on polariscopes) will appear dark. Grains with their c axis oriented at 45 to the pass direction of either of the two polarizing sheets appear bright. Thus, the randomness of the c axis orientation can be judged very easily by rotating horizontal thin sections under cross polarizers. The c axis of the individual grains can be determined exactly by installing the thin section on a universal stage in between the polarizer and the analyzer, and performing the analysis for textural analysis and making fabric diagrams [Langway, 1958]. The procedures are rather laborious and time consuming. However, a simple and unambiguous method for quickly ascertaining (and record-keeping purposes) the randomness of c axis orientations in S2 ice for selection of specimens for creep and strength tests was developed by Sinha [1978]. The method can be applied, if deemed necessary, to each test specimens in the form of large rectangular prisms (preferred for columnar ice) with dimensions of 100 mm × 250 mm × 50 mm before testing. It is based on chemical etching and replicating the surface, as described in section 4.4.2.2. An example is shown here in Figure 5.16. The S2 type of transversely isotropic ice can also grow in saline waters if there are no turbulences or tidal currents in the water. It has been seen to grow in protected bays of saline or brackish water. Figure 5.17 shows S2 type sea ice formed in calm water in Labrador Sea near Cartwright, Canada. The ice was sampled in March 1996 when the

20 mm

(b)

20 mm

Figure 5.15 Photographs of double microtomed (DMT) thin sections of freshwater S2 ice under cross-polarized light: (a) horizontal and (b) vertical; the scale bar in the images is 20 mm [Sinha et al., 1996 / John Wiley & Sons, Inc.].

Figure 5.16 Optical micrograph of a replica of freshwater S2 ice showing a triple point and different orientations of the c axis (< c >) indicated by the long directions of the elongated etch pits corresponding to basal dislocations intersecting the surface; note the characteristics of the nearly symmetric grain boundary in the right half of the micrograph and the two asymmetric grain boundaries in the left (specimen prepared and photographed by N.K. Sinha).

210

SEA ICE

ice thickness was 0.25 m. The differences in the intragranular characteristic in freshwater ice and sea ice are significant. While grains of freshwater S2 ice look clear and grain boundaries are linear, grains of the same type of

Figure 5.17 Horizontal thin section (100 mm diameter) at 10 mm depth of young ice grew under quiet oceanic conditions in a closed bay in the Labrador Sea, eastern Canada. The ice was 0.25 m thick. The photograph was taken between crossed polarizers. Note the random orientation of the grains, the irregular gain boundaries and the nearly the same orientation of the intragranular brine inclusions, which appear as linearly-arranged dots (photo by M.E. Shokr, unpublished).

(a)

ice from the sea are translucent and exhibit complex substructures and brine inclusions. Similar to freshwater S2 ice, the conditions for the formation of S2 type saline ice can be simulated well under laboratory conditions. Thin layers of reasonably acceptable ice up to a thickness of 0.25 m can be grown. In fact, many studies on the mechanisms of brine drainage in saline S2 ice are based on laboratory observations. The laboratory studies are, however, complicated by artifacts imposed by the limited area and depth of experimental growing tanks. Moreover, nothing can replace the natural conditions for developing columnar-grained S2 type sea ice. Natural ice provides the only opportunity for the observation of brine drainage channels in nature. An example is shown in Figure 5.18. Here the ice core was sampled from an ice cover in southern part of Tasiujaq Sound, where the ice was found to be S2 type (Tasiujaq Sound, formerly known as Eclipse Sound, is a waterway in the northern part of Baffin Island within the Qikiqtaaluk region, Nunavut, Canada). The coring was performed with the intention of capturing a vertically oriented brine drainage channel in the middle of the core. The location of the main passage of the brine channel and the orientation of its major tributaries tend to show that the interconnected subgrain boundaries must play the major roles in the desalination processes. These channels are eventually filled by water percolating from the top and refreezing inside the open spaces of the channels. This can be seen near the core of the brine channel in Figure 5.18a, with details at the center shown in Figure 5.18b. This picture and the descriptions provided here may help in understanding the brine drainage mechanism through subgrain boundaries in sea ice, discussed earlier in section 2.5.3 and illustrated in Figures 2.23 and 2.33.

(b)

5 mm

10 mm

Figure 5.18 Double-microtomed (DMT) horizontal thin section of first-year S2 type sea ice at 0.25 m depth (near the ice bottom), obtained from Eclipse Sound in the Arctic, showing a brine drainage channel in (a) at the center of the section and (b) with details in the amplified core area marked by the white box (photos by N.K. Sinha).

POLYCRYSTALLINE ICE STRUCTURE 211

5.3.3.5. Columnar-Grained with c Axis Horizontal and Oriented (S3 Ice) According to the information in Table 5.1, S3 ice type is columnar ice with preferred alignment of c axis in the horizontal plane. While working on the actions of sea ice on the newly constructed wharf at Nanisivik Naval Facility, Baffin Island, Canada in 1975, Frederking and Sinha [1977] noticed preferred c axis orientation of columnargrained ice in the horizontal plane and parallel to the direction of relatively strong tidal current in the Strathcona Sound. This alignment was also confirmed later for sea ice in Tasiujaq Sound, Baffin Island by Nakawo and Sinha [1981, 1984], who provided details of fabric diagram, texture, and subgrain microstructure, such as, the brine-layer spacing as functions of depth and growth history. However, the most comprehensive confirmation of the relationships between the preferred c axis orientation and the direction of current in the water under the ice came from the investigations of Weeks and Gow [1978] off the northern coast of Alaska. They performed an extensive survey of sea ice structure along the coasts of the Beaufort and Chukchi Seas, and concluded that the degree of orientation is more visible as the depth of the ice sheet increases. Moreover, the preferred orientation generally parallels the coastline or the narrow channel in which ice is growing. Beyond any doubt, the authors established the link between the orientation and the water current under the ice. They explained this phenomenon based on a hypothesis that ocean current flow reduces the salinity of the seawater in the direction of the current. This makes crystal growth favorable in that direction as they can extend easily farther into the melt than neighboring crystals that face “resistance” for their growth from the higher salinity solute.

θ

A schematic diagram for S3-type columnar-grained sea ice with preferred orientation of the c axis in the horizontal plane and parallel to the dominant current direction is illustrated in Figure 5.19a. At higher levels, close to the surface of ice covers, the grains are oval or elliptical in cross sections normal to the long dimension of the grains. The cross-sectional shape is usually elongated with the longer dimension normal to the direction of the water current. As usual for sea ice, grains in S3 ice consist of subgrains and inclusions in the form of brine and air pockets. Oval or elliptical shapes are also observed in the cross-sectional view of the subgrains, as sketched in Figure 5.19b. These cross-sectional views are readily noticeable in newly developed ice covers. Figure 5.20 shows examples of thin sections of 0.21 m thick “graywhite” young S3 ice sampled in March 1996 from the Labrador Sea, Canadian East Coast. The orientation of the c axis is apparent in the horizontal section. 5.3.3.6. Agglomerate Ice with Discontinuous Columnar-Grained (R Type Ice) Growth history affects the structure and texture of ice in a profound manner. In oceans, the tidal currents, wave actions, rafting, pressure ridging producing rubbles, and the subice water conditions including the mobility of frazil slush are some of the factors that change the habit of growing crystals in sea ice. The agglomerate ice, R type, is probably the most common characteristics of sea ice in highly dynamic ice regimes. It features a mix of ice types of different orientations, c axis directions, and sizes. Figure 5.21 is an example of the R type young ice, as revealed in a vertical section of the top 0.12 m of young sea ice from the Labrador Sea. The figure shows agglomeration of ice crystals of different sizes and orientations,

n ea M is x a

Sea water current c

c c

c

c

Grain boundary

c

c

r r de ate Un e w nt ic rre cu

c Air Brine a S3 type

a

Subgrain boundary

Growth direction

Figure 5.19 Three-dimensional schematic diagram of (a) columnar-grained S3 type sea ice with preferred c axis orientation; (b) locations of brine and air pockets along grain and subgrain boundaries, rows of inclusions without apparent boundaries, and orientations of a axis in two neighboring grains [Sinha, 1989 with additions of (b)].

212

SEA ICE (a)

(b)

10 mm

10 mm

Figure 5.20 Thin sections of gray-white 0.21 m thick young S3-type sea ice from the Labrador Sea in March 1996, photographed using cross-polarized light. (a) Vertical section of the entire 100-mm-diameter core and (b) horizontal section at the depth marked by the solid line in (a) (photos by M.E. Shokr).

and signifies the impact of the highly turbulent conditions that prevails during the ice formation and growth.

Figure 5.21 Vertical thin section of the top 0.12 m of R type young ice from the Labrador Sea, Canada, in March 1994, exhibiting grains of different shapes, sizes, and orientations under cross-polarized light (photo by M.E. Shokr).

5.3.3.7. Ice of Land-Based Origin Land-based ice originates from glaciers, ice sheets, and ice caps that form as a result of snow deposition, compaction, and melting and freezing over hundreds of thousands of years. A distinctive feature of land-based ice is the presence of cracks and crevices at the surface. They are often filled with melt water or rain that eventually solidifies, but this process leaves behind different families of bubbles that vary in shape and size. An example of this feature is presented in this section. Another distinctive feature of ice shelves only is sea ice formation at the bottom of the shelf. This continues in icebergs and bergy bits as they calve from ice shelves [note that the term ice islands is applied to huge icebergs calved from ice shelves in the Arctic (section 2.9.4)]. An example of this bottom-located ice (platelet ice) is presented in the next section. Figure 5.22 is an example of microstructural analysis of glaciers ice. In this case, the thin sections were prepared from an ice block sampled from the Morris Jessup Glacier, Greenland, on 5 July 1982, about 11 km east of the calving

POLYCRYSTALLINE ICE STRUCTURE 213 (a)

(b)

10 mm

(c)

10 mm

(d) N

Figure 5.22 Thin section of Greenland glacier ice: (a) cross-polarized light, (b) scattered light, (c) parallel-polarized light, and (d) the fabric diagram using only the larger grains. Note the effects of different light and imaging conditions on the apparent information in the images (N.K. Sinha).

end of the glacier. The area was heavily crevassed by streams and melt ponds. The rectangular-shaped rows of crystals lined up diagonally (particularly visible in (a)), indicate that a large crack developed at some time and the void was filled probably by the meltwater, which eventually solidified and produced two rows of large crystals. The bright objects in (b) are different families of air bubbles with different orientations as well as linear features, which are the results of healing processes of cracks. The grains shown in (a) were used to make the fabric diagram in (d). The diagram may not be significant in a statistical sense because it does not include a large number of grains. The marking N indicates the top of the micrographs and the direction parallel to the smaller dimension of the images. The orientations of the polarizers are shown by the pair of double-headed arrows in (a) (cross position)

and (b) (parallel position) with the polarizer making 30 from the direction N. 5.3.3.8. Platelet Sea Ice As mentioned in section 5.3.2 and shown in Figure 5.5, platelet ice is congealed frazil ice, which is formed in supercooled sea water under ice shelves as a result of their basal melting, and then ascend under buoyant forces toward the bottom of the ice cover where it accumulates. Figure 5.23 shows a photograph of a thin section of FYI, at approximately 0.63 m depth from the surface, where platelet (frazil crystals) ice is depicted as elongated linear (needle-shaped) features with brine pockets visible inside. This ice core (10 cm diameter) was extracted from Atka Bay, Antarctica (70.58 S, 7.85 W), which is close to the site of Germany’s Neumayer-Station III. The bay is also

214

SEA ICE

Figure 5.23 Photograph of horizontal thin section showing frazil platelet crystals (elongated features). The section was prepared from a 10-cm-diameter ice core at approximately 0.63 m, taken from Atka Bay near the German wintering station Neumayer III, Antarctica (courtesy of M. Hoppmann of Alfred-Wegener-Institute).

close to the Ekström Ice Shelf, which occupies an area of 8,700 km2 with thickness of 160 m at the edge. Sea ice does not usually grow very thick in the bay, but platelet ice caused by the presence of the ice shelf adds significantly to the thickness. 5.3.4. Stereographical Projection (Fabric Diagram) of Natural Polycrystalline Ice One of the important aspects of natural ice is the overall crystallographic orientation of c axis of the constituent crystals. They can be oriented in any direction in the 3D space. The stereographic projection is a useful tool for mapping the crystallographic planes and directions in terms of c axis of the ice grains in two dimensions: the azimuth angle and the inclination angle (direction of inclination as to left or right). It allows the visualization of the orientations of the crystallographic planes and directions in a rather straightforward manner [Reed-Hill and Abbaschian, 1992]. Two types of projections are used: the Wulff net and Schmidt net. The Wulff net is a meridional stereographic net, and is a projection of latitude and longitude lines so that the north-south axis is parallel to the plane of the paper. These lines serve essentially the same function as the corresponding lines on usual geographical maps. However, the major difference is that the measured angle of the principal axis (such as c axis for ice) is

stereographically projected instead of distances in geographical maps. The Schmidt net, on the other hand, is an equal-area net and provides another way for stereographic projection. In this projection, a unit area in any position on the net corresponds to a unit area on the spherical projection (lower hemisphere of the spherical projection is normally used) from which the net was developed. This projection has been used extensively for ice which belongs to the family of hexagonal uniaxial crystal with one major axis—the optic axis or c axis (). It is relatively simple to determine the orientation of c axis of individual crystals of ice in thin sections on the basis of extinction in cross-polarized light. For this purpose, thin sections are mounted on a universal, Rigsby type stage between two polarizers (section 4.3.1 and Figure 4.12). Measured data on the orientation of the c axis of the grains are plotted on the net as points. The projection is called “orientation fabric,” “fabric diagram,” or simply “fabric.” A minimum of 200 points are actually required for making a statistically meaningful fabric diagram, but often that is not possible for many types of glacial ice samples due to the large size of the constituent grains. Most manageable-sized thin sections, i.e., prepared from 100mm-diameter ice cores, may not contain sufficient number of grains. However, this is not considered as a major problem for highly oriented columnar-grained ice of the type S1 or S3. In case of S1 type freshwater ice, the entire thin section may contain only a fraction of one grain. This may also apply for S3 type of sea ice for most of the bulk ice away from the top surfaces. The main requirement for making the measurements for fabric diagram is the availability of thin sections, about 0.5 mm thick, with parallel, very smooth, and undistorted surfaces. The hot-plate technique or its slight variations, described in section 4.2.1, that applies surface warming and melting for mounting, thinning, and polishing thin sections is commonly used for the fabrication of usable specimens of freshwater and glacier ice. The necessary background for making relatively good thin sections of saline-free glacial ice, performing the petrographic measurements, and plotting the data for microstructural investigations of glacial ice with identifiable large crystals has been presented by Langway [1958]. This includes the basic principles developed in the field of structural petrology and metallurgy to measure the crystallographic orientations and plotting the data. It has essentially been adapted for ice to produce the fabric of columnar-grained ice with large cross-sectional grain sizes [Weeks and Gow, 1978, 1980, Nakawo and Sinha, 1981]. Recently, Weeks [2010, Appendix E] has provided an excellent description for procedures to be followed for sampling sea ice and preparing thin sections using the hot-plate technique. He has also explained the use of Rigsby universal stage and described, in detail, the

POLYCRYSTALLINE ICE STRUCTURE 215

procedures originally developed by Langway [1958] for general grain-based petrographic analysis for glacier ice. While this is highly recommended, it should be pointed out that the procedures allow one to determine an average c axis orientation for grains in case of large-grained sea ice. Unlike crystals in glacier ice or freshwater lake and river ice, grains in sea ice are actually ill-defined. Each grain consists of innumerable subgrains with slightly differing c axis orientations. Figure 5.24 shows graphical presentations of some orientation fabrics of ice found in nature. Each segment in the figure represents a horizontal circular cross section of ice. Each pair of dots inside the circles represents the two ends of c axis projected on the surface (Strictly speaking, the dots should be drawn as open or closed to represent the two ends.). If the dots are spread uniformly over the circle, it means that the projections of the c axes can have any length and may run in any direction (randomly oriented c axis). This is typically observed in isotropic, equiaxed, granular T1 type

Random orientation T1 type

Vertical orientation S1 type

of ice. If the dots are concentrated in a small area at the center, it means that the projections of the c axes on the surface are very short, implying that the optic axes of the grains are almost vertically oriented—the main characteristic of S1 type ice. If the dots are located near the peripheral of the circle, but in all directions, it means that the c axes are randomly oriented in the horizontal plane, and this characterizes the S2 type of ice. S3 ice is represented by the dots concentrated in two opposite areas close to the peripheral. Nakawo and Sinha [1984] examined the fabric diagram of fifteen cores of FYI sampled from Tasiujaq Sound, Baffin Island, Canadian Arctic, in January 1978. The samples were recovered when the ambient air temperatures were less than −20 C. The orientation of the crystallographic optic axes of crystals was measured in thin sections of the cores using a Rigsby universal stage. Figure 5.25 shows the distribution of the c axis orientation at three depth levels, plotted on a Schmidt equal-area net through the lower hemisphere. It may be seen that

T1 type superimposed on S1 type

Aligned horizontal orientation S3 type

Horizontal orientation S2 type

Figure 5.24 Fabric diagrams of a few common types of natural polycrystalline ice [Sinha, 1991b].

N

W

N

E W

S

N

E W

S

E

S

Figure 5.25 Fabric diagrams exhibiting c axis orientations at depths of 0.35 m, 0.63 m, and 0.83 m (left to right) in a first-year sea ice core, sampled in January 1978 from Eclipse Sound, Baffin Island, Canada. The two-headed arrow indicates the mean and the “bow tie” indicates the standard deviation [Nakawo and Sinha, 1984].

216

SEA ICE

the c axes of the crystals are, almost all, in the horizontal plane, but the plunges are small. Moreover, they seem to concentrate approximately in the NE-SW direction. Neglecting the slight deviation in plunge from the horizontal, i.e., assuming that the c axes are in the horizontal plane, the mean and standard deviations of the orientations of the optic axes were calculated. The mean orientation is shown by a two-headed arrow and the standard deviation by the “bow tie” in each fabric diagram. This representation was introduced by Weeks and Gow [1978] to illustrate the strong horizontal alignment in the c axis orientation of columnar-grained sea ice along the coast of the Beaufort Sea. Figure 5.26 shows an example of the depth dependence of the mean azimuth angle of the c axis. Data are obtained from the same core from which the fabric diagram data in Figure 5.25 are presented. At a depth of 0.2 m, the mean azimuth angle is about 40 (measured clockwise from the north). It remains nearly constant down to 0.6 m depth, then decreases to about 20 near the 0.8 m level, but increases to about 50 toward the bottom. The preferred orientation near the bottom, 50 to 60 from the north, agrees well with the orientation of the shorelines (60 from the north). It is probable that the direction of the water current under the ice was parallel to the shoreline. This last observation confirms an earlier observation by Weeks and Gow [1980] for ice off the Chukchi Sea coast, Barrow, Alaska.

Shoreline orientation

Depth (m)

0

0.5

1.0 Bottom 0

60 30 Azimuth angle (deg.)

90

Figure 5.26 Variation of the average azimuth angle of the c axis orientation of the same ice shown in Figure 2.25. The bar shows the standard deviation of the orientation [Nakawo and Sinha, 1984].

5.4. EXAMPLES OF CRYSTALLOGRAPHIC STRUCTURE OF NATURAL SEA ICE Examples of polycrystalline structure of seasonal sea ice, namely YI and FYI (see definition is section 2.8), are presented below to highlight main crystallographic features of each type. The focus is on some of the rudimentary areas related to the classification of sea ice based on texture, structure, and early growth processes that occur in natural bodies of sea water. The examples are of sea ice obtained from field investigations in the Canadian High Arctic (central and western) and the subarctic areas of Labrador Sea (Eastern Canada). Many persons including the staff of the Polar Continental Shelf Project (PCSP), the Canadian Coast Guard (CCG), Atmospheric Environment Services (AES) of Canada, and members of the local Inuit communities of Pond Inlet, Resolute, Grease Fjord, and Pangnirtung (all located in Nunavut, Canadian Arctic), provided assistance and passed their inherited knowledge to the authors. Some of the results presented here have never been published in the open literature. A couple of notes are needed as introduction to the following material. The first is about the columnar-grained ice, which is developed as a result of unidirectional growth of nucleated ice crystals as seeds. The growth occurs in the vertical plane along the direction of the maximum heat flow. This type of oriented growth is called directional solidification (DS) and found in high-temperature complex superalloys used in making gas-turbine engines, components for jet aircrafts and rockets [Sims, Stoloff, Hagel, 1987]. Once the DS type of growth is established in ice covers, the cross-sectional areas of the grains tend to increase with the increase in depth. This is because of the process of geometrical selection of crystal growth, described in section 5.3.3.4 and Figure 5.14. The second note is about chronic problems in the sampling and testing of natural sea ice in general, which affect the data of crystallographic structure of sea ice. They include the undesirable desalination of sea ice when samples are shipped for conducting laboratory tests, even if the ice cores or blocks are recovered at extremely cold ambient temperatures. Natural decay processes also make it impossible to take samples without disturbing the structure and its environment when the ambient temperatures are high. Moreover, small-size samples may not represent the bulk of the material. The bulk properties of any crystalline material (metals, metallic alloys, and non-metals like ceramics, rocks, and ice) can only be determined by tests that involve a large volume of the material containing many grains. Thus, for the data presented in this section, an effort was made to include microstructural information from a large number of vertical thin sections with vertical dimensions up to about 600 mm and

POLYCRYSTALLINE ICE STRUCTURE 217

horizontal thin sections obtained at different depths, in order to ensure the representativeness of the information. 5.4.1. Crystallographic Structure of Seasonal Sea Ice A common feature of seasonal sea ice is its diverse internal polycrystalline structure. This is particularly noticeable in its upper layer, which may often extend to a few centimeters in depth, and sometimes to fractions of one meter before congelation (i.e., formation of columnargrained growth). The diversity of crystalline structure is a manifestation of the wide range of oceanic conditions under which ice is formed. This is evident in both the Arctic and Antarctic data, as demonstrated in several studies on natural sea ice samples [Gow et al., 1982, Weeks and Ackley, 1982, Gow and Tucker III, 1990, Jefferies and Weeks, 1992, and Shokr and Sinha, 1994]. The microstructure of seasonal sea ice includes that of typical YI, which is usually observed in the top layer of the ice cover. This layer is considered to be an initial condition that determines the further development of the crystalline texture. The type of ice in this layer is determined mainly by the environmental conditions prevailing at the time of ice formation. It is usually composed of long needleshaped crystals–if formed in rough seas, columnar crystals–if formed under quiet oceanic conditions, or small round crystals (granular ice) –if ice is formed from snow-saturated water while freezing. A thick layer of columnar ice usually follows this initial layer. Polycrystalline ice structure is usually more diverse in subpolar areas because of the higher probability of turbulent oceanic conditions and fluctuations of atmospheric temperature around the ice freezing point during ice formation and growth. Since ice does not grow very thick in these areas (e.g., the average ice thickness in the middle of winter in the Gulf of St. Lawrence in the east coast of Canada (latitude 48.92 N) is around 0.50 m), ice congelation is less likely to occur. This contrasts the crystallographic texture of sea ice in the polar regions, where ice most often congeals into columnar structure at depths not far from the surface. Nevertheless, even in the Arctic, columnar (congealed) growth of ice may be interrupted by a frazil layer, if a herd of frazil crystals is carried away by ocean current and sticks to the bottom of the existing ice sheet. Ice may continue its growth and re-congeal at lower depths. Due to the diversity of ice formation and growth conditions, the crystalline structure of seasonal sea ice cannot be categorically characterized. All crystallographic ice classes (except S1 type) are found in seasonal ice. Perhaps the only feature that characterizes its microstructure is the presence of numerous brine inclusions, in the form of pockets and drainage channels, in addition to small air bubbles found mostly inside the brine (section 2.5.4).

5.4.1.1. Frazil Ice (S5 Type) One of the common polycrystalline categories of YI is the vertically oriented needle-shaped crystals frazil (S5). As mentioned in section 2.3.2, frazil ice develops when sea water starts to freeze under dynamic and turbulent conditions, and the small ice crystals are broken to form needles or disk-like crystals. Frazil ice is very common in the turbulent Antarctic waters and may round up at the edges of ice floes. Individual frazil crystals can act as seeds for more crystal formation. Gow et al. [1982] observed that up to 50% of the ice cover during their 1980 cruise was frazil ice, and this ice usually rounded up at the edges of ice floes. Clarke and Ackley [1982] found a predominance of frazil (70%) during their 1981 Antarctic cruise. These observations were essentially sporadic, and made on opportunity basis without any knowledge of the history of the formation of the ice cover. More systematic filed studies of Antarctic sea ice with clear objectives to study sea ice physics are still needed to make more conclusive statements about dominant ice features and crystalline structures. Structural details of the S5 type frazil ice are illustrated in Figure 5.27. Vertical and horizontal thin sections were made from a large block of young frazil ice, extracted from a fjord in the Mould Bay, on 25 October 1981, through the full depth of the ice cover. In general, the grains in the frazil layer were small and elongated as seen in the vertical section. The long axis of the grains tended to be at right angles to the ice surface. The length of the grain is between 2 and 5 mm, and the diameter is about 0.5 to 1.5 mm. Texture of this layer, therefore, was transversely isotropic with the plane of isotropy parallel to the surface. This texture in the ice sheet was developed as a result of closer packing, when the elongated frazil crystals formed in the water of the fjord were driven by wind toward the eastern shore. Records at the weather office in Mould Bay indicated heavy wind from the west before 24 September 1981, the freeze-up date for Mould Bay water. A striking observation in Figure 5.27 is the clear interface at about 0.27 m depth between the frazil ice and the underneath columnar ice (shown clearly in the insert at the bottom right of the figure). The vertically oriented frazil crystals with their c axis in the horizontal plane served well as the nucleating agents for new ice at the ice–water interface at the bottom of the frazil layer. This was indeed a classic case of nucleation from seed crystals. At this stage, the processes of columnar crystallographic growth and geometric selection of the growth were already made for the columnar grains to develop. It is not a wonder, therefore, that columnar-grained ice grew immediately at the bottom of the frazil layer. Conditions are suitable for this growth as no turbulence is encountered. Structural details of the frazil–columnar interface at 0.27 m in Figure 5.27 are shown in Figure 5.28. Since

218

SEA ICE

0.00

0.05

Ice depth, hi (m)

0.10

0.15

horizontal sections

20 mm

0.20

0.25

0.30

0.35

Figure 5.27 Thin sections of YI of S5 type viewed under crosspolarized light: (left) vertical section from top to the bottom and (right) horizontal sections at depth of 50 mm and 150 mm. At the bottom right is an inset of the frazil-columnar ice interface at about 0.27 m depth [Sinha, 1986 / John Wiley & Sons, Inc.].

the frazil ice was transversely isotropic, it must have initiated S2 type structure. However, examinations of the bottom ice reveal S3 structure. This indicates that the only frazil crystals that have a chance to grow are those with their c axis favorably oriented along the water current. For fine-grained materials like snow and frazil ice, it is not convenient to use the conventional polarized light and the universal stage method to examine the fabric of the ice. Etching and replication technique (described in section 4.4.2) was found to be ideal for frazil S5 ice. Micrographs of thermally etched surface of young frazil ice shown in Figure 5.27 are presented in Figure 5.28a and 5.28b using an optical microscope. They show grain boundaries in the horizontal and vertical sections. A micrograph of a replica of chemically etched and replicated surface (vertical section) using scanning electron micrograph is shown in Figure 5.28c. The horizontal orientation of the c axis of the vertically oriented frazil ice crystals is revealed by tiny elongated dislocation etch pits

as indicated by the doubled-headed arrows. Those etch pits correspond to the intersection of the basal dislocations with the surface under observation. The c axes of the grains, however, were found to be randomly oriented in the horizontal plane. Because of this, not all the grains in Figure 5.28c showed elongated pits, which develop only if the surface under etching coincides with one of the prismatic surfaces of the crystals. As the etching and replicating processes were carried out at −10 C, there was brine (liquid) not only in the brine pockets primarily at the grain boundaries, but also smeared on the surface prepared by microtoming. As may be seen in the figure, it was found that brine on the surface interferes with the etching processes, although not sufficiently to prevent etching of grain boundaries. 5.4.1.2. Columnar-Grained Ice (S3 Type) An example is presented from a study of YI formation and its aging, which was performed in a long fjord, with measurable tidal activities, in Mould Bay, Prince Patrick Island in 1981. Figure 5.29 includes photos of thin section of S3 type of sea ice, with its aligned c axis in the horizontal plane as appearing in the horizontal thin section in Figure 5.29b. This ice was formed under very calm conditions, with no snow activities. The vertical section clearly shows that the columnar-grained structure develops within a few mm from the top surface. The horizontal section at 15 mm depth also reveals that the horizontal preferred orientation develops soon after nucleation of the ice crystals at the surface level. This clearly refutes the idea that preferred orientation should start to develop at some depth away from the surface. Nature is quite open for all possibilities and people should keep this in mind while formulating laws of natural processes. An example of the top 140 mm layer of FYI formed under quasi-steady conditions in the Resolute Passage during 1991 winter season is presented in Figure 5.30. The ice consisted mainly of oriented columnar crystals that started right from the initial ice formation (i.e., at the top of the ice in Figure 5.30a). The striking observation in this section is the highly oriented (vertically extended) ice crystals with narrow width. It means that the c axes of those crystals were always horizontal and the entire ice depth grew under calm oceanic conditions. The big dark area in Figure 5.30a identifies grains with their c axes parallel to the direction of propagation of the polarized light (see section 4.1.1 and Figure 4.3). Note that dark areas could also mean holes in the thin section, but this is not the case here because the photo using scattered light (Figure 5.30b) shows information at the dark location in Figure 5.30a. Three sublayers exhibiting different microstructural characteristics can be identified in the figure as marked by the thick dark lines. The first layer occupies the top 7 mm, and is characterized by randomly oriented fine grains,

POLYCRYSTALLINE ICE STRUCTURE 219 (a)

(b)

1 mm 1 mm

(c)



200 μm

Figure 5.28 Optical micrograph of thermally etched ice surface showing grain boundaries in (a) vertical and (b) horizontal cross sections of frazil crystals, and (c) scanning electron micrograph of a replica of chemically etched vertical section exhibiting brine pockets (circular objects), grain boundaries (curved lines), and dislocation etch pits elongated along the c axis in the horizontal plane, as indicated by the arrow [Sinha, 1986 / John Wiley & Sons, Inc.].

typically a few millimeters in size. This layer probably constitutes slush ice formed, perhaps, during a snowfall, which indicated P4 ice class (see Table 5.1). The second layer extends from 7 mm to 25 mm depth and contains vertically oriented frazil grains, slightly larger than those in the first layer (grain size is typically 4–10 mm). The third layer features columnar-grained ice, where grains are oriented as shown in the horizontal section at 140 mm depth (Figure 5.30c). Numerous brine pockets and brine drainage channels are observed in Figure 5.30a located along the grain and subgrain boundaries. They are more apparent in Figure 5.30b, which was photographed using scattered

light. A high concentration of inclusions can be seen in two bright spots at the top surface marked with arrows. They extend horizontally indicating a brief horizontal initial growth of ice. Down from this thin layer, brine pockets and channels are arranged in a significant number (mostly thin vertical lines). Their dimensions range between a fraction of a millimeter to a maximum of 2 millimeters. A short drainage channel of approximately 1 mm width can be seen in the middle part of Figure 5.30b. Grain orientation defines the direction along which brine pockets are stretched. Information on this direction, with respect to the applied electric field, is important to determine both the dielectric constant and microwave scattering.

220

SEA ICE (a)

(b)

100 mm 1 mm

Figure 5.29 Photographs of vertical thin section, prepared using the DMT, of the top 0.21 m of freshly recovered young sea ice from Mould Bay (76 14’N, 119 20’ W) showing S3 ice type in (a) vertical section of the top 170 mm and (b) horizontal section at a depth of 15 mm, both observed with combination of scattered and cross-polarized light. The double arrow in (a) indicates the scale as well as the direction of water current and the double arrow in (b) indicates the dominant direction of water current (photograph by N.K. Sinha).

5.4.1.3. Agglomeration of Various Crystallographic Structures Young ice in highly dynamic areas in temperate latitudes can be subjected to intensive turbulent oceanic conditions during its formation, as well as melting and reformation. Growth processes are continuously disturbed by mechanical or meteorological interferences. Grains vary considerably in size and orientation, and often appear with irregular boundaries. This situation is encountered, for example, in the Labrador Sea and the Gulf of St. Lawrence, both in the eastern coast of Canada. An example of thin ice (nearly 80 mm thick) sampled from the Labrador Sea on 10 March 1994, is shown in Figure 5.31a. That was the thinnest ice that could be walked safely (but extremely carefully) on (see Figure 1.14 for an idea on the bearing capacity of sea ice covers). The general view of the ice cover featured Nilas and gray ice (as identified in the operational ice chart by the Canadian Ice Service of Environment Canada). The surface roughness of the gray ice shown in Figure 5.31b is a manifestation of the presence of pancake ice, with raised edges that has grown under intensive turbulent oceanic conditions. This ice field is another example of the agglomerate ice (R type), which was exemplified earlier in Figure 5.21. Photograph of the surface cut through 100 mm ice core (Figure 5.31c) reveals four distinctive layers separated by the broken lines. These layers can be identified clearly in the corresponding vertical thin section under polarized light (Figure 5.31d). The top layer

was probably formed from refrozen snow melt with irregular crystalline structure. This is superimposed ice (section 2.4.3). The second layer features crystals that grew horizontally (c axis vertical), and also a continuation of the irregular crystals. Columnar crystals are visible in the third layer. This is probably the first layer that was formed under relatively calm water before snow accumulated, melted, and refroze on top of the ice. The bottom layer of the ice exhibits an interruption in the columnar growth with accretion of finer frazil crystals. 5.4.1.4. Air Entrapment in Seasonal Ice Gases dissolved in sea water are rejected during the solidification processes. They become entrapped mostly inside brine pockets, but isolated air pockets are also possible (Figure 2.26). This scenario applies only to seasonal ice because air bubble formation in perennial ice follows a different mechanism illustrated in Figure 2.58. It is easier to identify isolated bubbles than the tiny submillimeter bubbles inside brine pockets. Figure 5.32 shows an example of brine and air entrapment. Three types of elongated inclusions are noticeable. They include clear brine pockets outlined by dark lines, brine pockets with small dark spherical air bubbles inside them, and dark individual larger bubbles. As explained in section 2.5.3 and Figure 2.22, brine inclusions with bubbles inside are mostly located along subgrain boundaries and associated inclined surfaces, visible in the form of diagonal bands in

POLYCRYSTALLINE ICE STRUCTURE 221 (a)

(b)

10 mm

10 mm

(c)

10 mm

Figure 5.30 Photographs of thin sections of the upper 140 mm columnar-grained ice formed under quasi-steady ocean surface: (a) photographed between crossed polarizers, (b) photographed using diffuse light, and (c) horizontal section at 140 mm depth, photographed between crossed polarizers. Arrows at the top mark spots of high concentration of inclusions within the frazil ice matrix. The arrow at the bottom marks a vertical drainage channel. The horizontal dark lines in (a), divide layers of different microstructure characteristics (photos by M. Shokr).

Figure 5.32. Note the inclusions along the boundary in the shape of an arc at the bottom left of the picture. The long direction of the inclusions tends to be parallel to the surfaces of the boundaries. In thin sections, when observed through optical microscopes under transmitted light, all air bubbles, whether isolated or entrapped within brine pockets, appear as dark objects (usually spheres) with bright spot at the center. An explanation for this appearance is given below. The dark appearance of the air bubbles can be explained on the basis of light transmission characteristics as schematically shown in Figure 5.33. The only light that can go through the objective (lens) of a transmission-type optical microscope is the central beam incident at right

angle to the surface of the bubbles. This central beam, therefore, produces the bright spot in the middle of the dark air bubbles. The dark areas of the air bubbles are caused by the scattering of the transmitted light from the source before reaching the lenses of the microscope. 5.4.2. Crystallographic Structure of Perennial Sea Ice Perennial sea ice is a common feature in the polar regions, particularly in the Arctic. More perennial ice is found in the western section of the Arctic Ocean, where the Beaufort Sea gyre causes the ice to circulate in the region for several years. Occasionally, large masses of this ice move inside the ice-free fjords and sounds during the

222

SEA ICE (b)

(a)

Gray ice Nilas

(c)

(d)

Figure 5.31 (a) A general view of an ice cover in the Labrador Sea near the shore of Cartwright, Newfoundland and Labrador province, Canada on 10 March 1994, (b) an enlargement of the edge between the Nilas and gray ice in the scene, (c) photograph of the middle plane of a vertical section of 100-mm-diameter core (cut vertically throughout the entire 80 mm thickness of the ice), and (d) the midplane vertical thin section photographed under cross-polarized light. Four distinctive layers of grain structure are delineated by white dashed lines (photographs by M. Shokr).

summers. This does not happen every year in the Low Arctic. Moreover, often grounded ridges or rubbles piled up against the land near the shorelines and survive through the summer melt season to become landfast perennial ice. This ice has not been extensively studied, and it is expected to have different crystallographic structural features than the perennial ice found in the open sea. It would likely feature less bubbles and higher salinity than the typical MYI because the mechanism of bubble formation in hummock MYI surface shown in Figure 2.58 (originated from ridging on the original FYI surface) does not quite apply to the leveled surface of landfast ice. Limited number

of studies has shown recent retreat of landfast perennial ice in a few regions in the Arctic [e.g., Pope, Copland, Mueller, 2012], but no study has focused on the crystallographic structure of this ice type. In this section, the focus is placed on crystallographic structure of perennial ice floes within the vast mobile area of the sea ice cover in the Arctic. However, a brief comparison between crystalline structures of perennial landfast ice incidentally found adjacent to FYI in the Allen Bay area near Resolute Bay, Canadian Arctic, in April 1997, is presented at the end of this section. Large-scale characteristics of MYI floes are described in section 2.7.2. As mentioned there, the surface is

POLYCRYSTALLINE ICE STRUCTURE 223

0.5 mm

Figure 5.32 Photomicrograph of brine pockets and air entrapment in FYI (N.K. Sinha).

Ice

Brine

Air

Figure 5.33 Schematic to illustrate the appearance of a dark air bubble inside a clear brine pocket, when viewed through transmission-type optical microscope. Only a narrow beam of light propagating normal to the surface of the bubble can go through the lens of the microscope, making a visible white spot in the middle of the bubble (sketch by N.K. Sinha).

characterized by hummocks and depressions (also called melt pond ice–MP ice). These are characteristically different in terms of subsurface and internal microstructure. The crystalline structure of the top layer of both types is different, and different from the top layer of the FYI from which they originate. This is due to the surface melt in the summer and refreezing in the next fall. Hummock ice features a variety of crystal shapes and sizes with convoluted boundaries. This is mainly a result of the aging processes involving partial thawing, percolation of melt water to the lower levels and subsequent refreezing. The only inclusion found in hummock ice is air bubbles, which exist with large volume fraction within the upper 20 cm or so layer (see Figure 2.15). In this case,

bubbles exist in relatively large size (order of a few millimeters in diameter) at much closer distances from each other. When hummock surface melts as a result of exposure to solar radiation during summer, meltwater accumulates in adjacent melt ponds. Albedo of the pond then decreases and this triggers positive feedback cycle that accelerates ice melting and increases the size of the melt ponds. About 1 m of ice was observed to melt on a multi-year floe in Sabine Bay, Melville Island during the summer of 1992, by the authors. The melt is usually brackish water with salinity less than 2‰. Melt pond ice features relatively large crystals at the top, which originate from freezing of brackish water. Unlike hummock ice, MP ice usually contains fewer bubbles and more brine pockets, though the salinity near the surface is generally low (1‰–3‰), depending on the flooded water salinity from which ice was formed. In the case of SYI, hummocks are usually formed at locations of ridges while MP ice is formed at leveled surface areas. The scenario of air bubble formation at the surface of old hummocks is schematically presented in Figure 2.58. It depicts the interconnectivity of the drainage network and consequently the bubbles, particularly at locations near the hummock surface. It also suggests the decrease in bubble intensity with depth as the water drainage network becomes less intensive. Other features are concluded from examining photos of crystalline structure presented later. The concave surface of MP ice, on the other hand, does not allow formation of a drainage network. In this case, water from the surface melt along with snow melt and/or rain, could accumulate on top of the ice surface. When this water freezes, it forms a new ice layer on top of the older ice (i.e., superimposed ice). The interface between the two layers can sometimes be abrupt and well defined. The crystalline structure of the new layer depends on the water quality and the atmospheric conditions. S1 type of ice with very large crystals develops for calm conditions during freezing. However, if snow falls before the initiation of freezing, then granular layer followed by S2 type of ice develop depending on the amount of snow deposition. Formation of air inclusions in this case is possible through two mechanisms. The first is air entrapment between snow particles and the second originates from the air dissolved in the water. In both cases, the air inclusions in MP ice are much smaller than air bubble in hummock ice. Examples of structural features of multi-year hummock and MP ice, as depicted in thin sections prepared from ice cores extracted from Lancaster Sound in the Canadian Eastern Arctic, are presented in the following. 5.4.2.1. Hummock Ice Figure 5.34 shows a photograph of the top 270 cm depth of a core sampled from a hummock ice. It exhibits a bubbly white layer ice extending to a depth of about

224

SEA ICE

Figure 5.34 Vertical 100-mm-diameter core exhibiting bubbly ice (white area) at the top of a hummock in an MYI floe found in Lancaster Sound, Canadian Arctic, in May 1992 (photographed by M. Shokr).

210 mm. The boundary between bubble-rich and bubblefree zones is fuzzy but well established. This termination of air bubbles can be gradual as seen in this figure, or abrupt. Based on a number of MYI hummock cores sampled from several areas in the Arctic by the authors of this book, the bubbly area usually occupies the top 50–350 mm. This porous layer may mistakenly lead one to think that the ice is composed of granular or equiaxed snow ice. Only forensic type of metallurgical investigations can reveal the true characteristics of the white top layer and the clear ice below this layer. The top 140 mm of the ice core shown in Figure 5.34 was cut vertically in the middle of the core for making the vertical thin section shown in Figure 5.35. The thin section was prepared at about −15 C inside a portable NRC cold laboratory in Resolute soon after sampling the core. The section was prepared using the DMT method with a freshly sharpened blade (section 4.2.2). The working temperature and the procedures utilized for the preparation of the section minimized the mechanical or thermal damage and protected the delicate and fragile bubbly ice. The photograph taken through transmitted cross-polarized light (Figure 5.35a) clearly brings out the characteristics of a few vertically oriented columnar grains–starting immediately from the top of the ice surface. The structure of the air bubbles, or strictly, the

(a)

(b)

2 cm

4 cm

6 cm

8 cm

10 cm

12 cm

Figure 5.35 Vertical section of the upper 140 mm depth of 100-mm diameter MY hummock core, obtained from Lancaster Sound, Canadian Arctic in May 1992, photographed (a) between crossed polarizers and (b) in scattered light [Shokr and Sinha, 1994 / Arctic Institute of North America].

POLYCRYSTALLINE ICE STRUCTURE 225

voids can be seen in Figure 5.35b, taken with scattered light by illuminating the thin section with diffuse light from the sides and switching off the transmitted polarized light (see section 4.3.3 for details). The shown layer in Figure 5.35 was part of the bulk ice that had existed at some depth below the surface in the original FYI. Due to the convex shape of hummock, it is unlikely that newer ice is formed on the surface in subsequent seasons. Most of the large black areas appearing at corresponding locations in both photographs represent large voids. The fine horizontal line is a crack formed during the microtoming and handling processes. The most important feature shown in the scattered light image is the large number of highly connected voids. In this case, air bubbles lose their individual identity and form a continuum. The measured density averaged over the 140 mm depth of the top layer was 620 kg/m3. This corresponds to

(a)

10 mm

(c)

air volume fraction of 38%. This value is too high for air to be considered as an inclusion within the background of pure ice as a host material. Under this condition, the application of a two-phase dielectric mixing model or scattering model (section 3.7.2), or the radiative transfer model to calculate the backscatter (section 12.2), both assume a dominant host (pure ice) and a minor inclusion (air) may not apply. Thin section photographs from another hummock ice core are presented in Figure 5.36. It shows a vertical section from 140 to 280 mm depth. Large columnar crystals tilted to the vertical direction are readily visible. Their major axis varies from 50 to 80 mm, although smaller dimensions around 10 mm are observed near the top. Crystals have irregular boundaries, indicating that the original ice was highly saline. The deviation of crystal orientation with respect to the vertical indicates that this part

(b)

10 mm

(d)

10 mm

Figure 5.36 Vertical section of an MY hummock core, obtained from Lancaster Sound in May 1992. Core diameter is 100 mm and depth range is 140–290 mm, photographed (a) between crossed polarizers and (b) using diffuse light. A horizontal section at 140 mm depth photographed (c) between crossed polarizers and (d) using diffuse light. White color in the photographs taken using diffuse light represents bubbly areas [Shokr and Sinha, 1994 / Arctic Institute of North America].

226

SEA ICE

2 cm

4 cm

6 cm

8 cm

10 cm

12 cm

Figure 5.37 Vertical section of the top 140 mm depth of an MY hummock ice core of 100 mm diameter, obtained from Lancaster Sound in May 1992, photographed between (left) crossed polarizers and (right) scattered light [Shokr and Sinha, 1994 / Arctic Institute of North America].

of the hummock had possibly originated from a tilted ice block in an old rubble field. The major axis of bubbles varies from 1 to 6 mm long. Once again, the boundary between the bubbly and bubble-free zones is well identified at about 220 mm below the surface. Careful examination of the air bubble orientations in the vertical and horizontal sections (the latter was made at 90 mm depth) reveals a tendency of the bubbles to stretch parallel to the dominant crystal orientation. This means that when surface melt infiltrates ice (in summer), the meltwater follows paths connecting brine pockets (which exist in the original saline FYI). Figure 5.37 shows another possible configuration of polycrystalline structure of hummock ice. It features large oriented crystals with planar boundaries with no subgrain boundaries. This is an indication that this ice grew from freshwater. Note that the long axis of the crystals is inclined at approximately 45 . Since this axis marks the crystal’s growth direction (i.e., the direction of maximum heat flow), it can then be concluded that this ice must have grown at the edge of a melt pond and, thereby, on a side of a hummock. The air bubbles in this core are vertically oriented; i.e., not parallel to the dominant crystal orientation. This is not uncommon in MYI hummock and it is in line with the configuration shown in Figure 2.58. The vertical orientation of air bubbles are indicators of the flow of meltwater following the direction of gravity, since no brine drainage channels exist in this example of freshwater ice. Irregular and interconnected bubbles, with typical dimensions between 1 to 4 mm, are observed in the top

part of the ice core (Figure 5.37). At greater depths the bubble shape becomes more elliptical with typical dimensions of the major and minor axes being about 6 mm and 3 mm, respectively. Bubbles are randomly spaced in both vertical and horizontal sections, but they tend to cluster as shown in the figure. It is worth repeating here that the distribution of air bubbles in hummock ice (or MYI for that matter) is different from the pattern of the distribution in FYI. In the latter, bubbles are usually arranged along the grain and subgrain boundaries usually inside brine pockets with their long axis parallel to the grain growth direction. As seen in the above examples, air bubbles in hummock ice are usually highly interconnected near the surface. They lose their individual identity and form a complicated network, particularly near the surface. Therefore, it is only possible to retrieve the geometrical characteristics of the bubbles, when examined at depths far enough from the surface (i.e., below the bubble-intensive area). Two micrographs of air bubbles at 0.14 m depth are shown in Figure 5.38. Bubbles appear dark and mostly located at the subgrain boundaries, which are made visible by thermal etching. A few bubbles also appear inside the grains. The colored areas inside the dark bubbles are parts of solid grains when the thin section cuts through them. 5.4.2.2. Melt Pond Ice Similar to hummock ice, MP ice exhibits a variety of microstructure patterns. Unlike hummock ice, however, two distinct types of air inclusions exist in MP ice.

POLYCRYSTALLINE ICE STRUCTURE 227

1 mm

1 mm

Sub-grain boundary

Figure 5.38 Optical micrographs, taken with polarized light, showing bubbles at a depth of 0.14 m in MY hummock ice in Resolute Passage, Nunavut, Canada in May 1992; subgrain boundary is revealed by thermal etching (N.K. Sinha).

The first, takes the form of clusters of bubbles. This is believed to originate from air entrapped between snow grains. Large clusters were consistently found in snow ice layers forming at the top layer of MP ice. This layer is formed during freezing of snow slush or waterlogged snow deposited on a pond. The second type comprises small air bubbles originally dissolved in the melt and later rejected during formation of ice in the pond. Both types differ from air bubbles in hummock ice in terms of formation mechanism and geometrical characteristics. Figure 5.39 [Shokr and Sinha (1994)] shows a vertical section of the upper 140 mm depth and a horizontal section at 90 mm depth of polycrystalline structure of MP ice in the central Arctic. The vertical section (Figures 5.39a and 5.39b) exhibits three layers of distinct structure. The top layer, about 35 mm in thickness, is comprised of numerous small ice crystals about 1 mm in diameter, which are isotropic and equiaxed (confirmed also from a horizontal section at 20 mm, which is not shown). This is granular ice most likely developed from freezing of water saturated with snow. Air entrapped between snow particles evolves into nearly spherical inclusions, depicted as large bright objects in Figure 5.39b. Their average diameter is 4 mm. These inclusions are consistently observed in, and confined only to, the top few centimeters of MP ice. Compared to commonly observed bubbles in hummock ice, they are larger, more isotropic in shape and well separated. The following 25 mm layer in the figure comprises ice crystals formed during the selection process, where crystals will survive only if their basal plane is parallel to the growth direction or the direction of maximum heat flow. Columnar-grained structure is observed in this layer. Smaller air bubbles, ranging from a fraction of a millimeter to slightly above 1 mm, are readily visible in this layer (Figure 5.39b). They are probably formed from dissolved

air in the pond water, which is rejected during solidification. This type of bubble was found consistently in MP ice, mostly around small ice crystals. While the top and middle layers contain newer ice, formed subsequent to melt season, the bottom 90 mm layer in the figure exhibits clear, nearly inclusion-free columnar crystals. The grains may appear to be of freshwater origin, but the non-planer or slightly irregular boundaries visible in the horizontal section (Figure.5.39c) suggest that the ice grew from brackish water. Measured salinity at this layer ranges between 1.1 and 1.4 ‰. The three layers contain newer ice and therefore appear to be formed subsequent to a melt season. This example shows the complexity of natural ice that defies modeling. Indeed, this is an example of MP ice, which is part of the MYI that modelers try presenting in their work. One can understand why and how many simplifications are needed to parameterize MYI in a thermodynamic or scattering model. Figure 5.40 shows a vertical section from 150 to 270 mm depth from another MP ice core. The core consisted of a 230 mm layer of non-saline granular, snow ice with grain diameters less than 2 mm. This suggested that ice was formed from water-saturated snow cover when numerous snow particles had existed as seeds. The top layer in the figure contains air inclusions of the type observed in snow ice. They are more visible against the dark background in the photo taken with the scattered light (Figure 5.40b). Their diameters range from 1.5 to 6 mm. Smaller air bubbles, typically less than 1 mm in diameter, can also be seen. Careful examination of each individual air bubble in the photograph between crossed polarizers in Figure 5.40a shows that its overall dark appearance of the large diameter objects is, in fact, cloudy with fading colors. The colors denote the presence of fine crystals within the inclusion. This supports the

228

SEA ICE (a)

(b)

10 mm

10 mm

(c)

10 mm

Figure 5.39 Vertical section of the top 140 mm of an MY melt-pond ice core, obtained from Lancaster Sound in May 1992, photographed (a) between crossed polarizers, (b) in scattered light, and (c) horizontal section at 90 mm depth photographed between crossed polarizers. The dashed lines separate areas with distinct crystalline structure [adapted from Shokr and Sinha, 1994 / Arctic Institute of North America].

premise that the inclusions were originated from larger air pockets entrapped between snow particles during the freezing process. The snow ice layer is underlain by a 30 mm transition layer of larger grains whose length varies between 1 to 4 mm. The layer marks a transition between the fine snow crystals at the top and the columnar grains that represent the original brackish ice, which has been retextured by aging. The development of columnar-grained ice also establishes the culmination of the selection process that leads to the survival of crystals with their basal planes in the vertical plane or their c axis in the horizontal plane. For such growth conditions, the impurities can be pushed down to the ice–water interface. Note the smooth transition in the grain structure. The air bubble contents in this MP ice core are similar to observations in hummock ice. The only difference, however, is the small size bubbles that are spread between large bubbles. Similar to hummock ice,

the bubbles diminish abruptly at location that marks the transition from snow ice to the re-textured ice. During a survey of sea ice cover around the area of Resolute Bay, Canadian Arctic, in May 1997, the research team found landfast ice in a confined area adjacent to the shoreline in Allen Bay, Cornwallis Island. The ERS2 SAR image of the ice cover in the Bay is shown in Figure 5.41a. Three zones can be distinguished based on the backscatter from SAR. They are labeled site 1, 2, and 3. Site 3 was confirmed by the local people to be at least SYI. Hence, there was an opportunity to examine the crystalline structure of this landfast SYI and compare it to the seasonal ice in the other two sites. It was interesting also to examine the reason for the brightest SAR backscatter from site 2 compared to site 3 as both were seasonal ice of presumably same age. Figure 5.41b shows the crystalline structures from the three sites in Figure 5.41a. Ice growth in site 1 apparently

POLYCRYSTALLINE ICE STRUCTURE 229 (a)

(b)

Snow ice

10 mm

10 mm

Figure 5.40 Vertical section of MY melt-pond ice at depths between 150 mm and 270 mm of an old floe in Lancaster Sound in May 1992, photographed (a) between crossed polarizers, (b) in scattered light illustrating bubbly snow ice and the transition from snow ice to columnar-grained ice [Shokr and Sinha, 1994 / Arctic Institute of North America]

Site 3 Site 2

Site 1

3 km

Figure 5.41a ERS-2 SAR image of sea ice cover in Allen Bay, Cornwallis Island, Canadian central Arctic, acquired in May 1997 showing three sites with different backscatter signatures. Site 1 was second-year landfast ice [ Johnston, 1998].

suffered interruption by depositing frazil ice at the locations marked by the two arrows. After the second interruption, congelation resumed without interruption as demonstrated by the continuous columnar ice growth. Site 2 has remarkably high backscatter in SAR image (Figure 5.41a) and the microstructure image is showing

fine grains, which could be instigated from frazil ice or snow ice, visible in the top 10 cm layer. The salinity profile in this layer was nearly constant around 6‰. This is a low value, which excludes the possibility of frazil ice because it typically retains considerable amount of brine. Therefore, the fine grains are likely snow ice. However, it is not possible to link this ice type to the observed high backscatter because low salinity (hence low dielectric constant) should not trigger high backscatter. The remaining possibility for the high backscatter is the surface roughness. The ice microstructure photo from site 3, which was confirmed to be second-year fast ice, shows columnar ice starting right at the top surface. The irregular top boundaries of crystals suggest that surface melt during summer had removed a layer of this ice. Two observations can be drawn from the crystalline structure of the core pertaining to this site. The first is the dominant columnar ice growth. This is expected from fast ice, which is not sensitive to atmospheric or oceanic disturbances during growth. The second is the less brine inclusions compared to those visible in the FYI cores from the other two sites. Salinity profile of this core shows relatively high value of 6‰ within the top 10 mm, then sharp drop to about 2.5‰, which remains invariant throughout the depth of the core. Therefore, it can be concluded that this landfast SYI has less salinity (and brine inclusion) than FYI. Its surface does not feature the typical geometry of alternating hummocks and MP ice. Finally, note that sites 1 and 3 have

230

SEA ICE Site 1

10 cm

Site 2

Site 3

10 cm 10 cm

20 cm 20 cm 2 cm

Figure 5.41b Cross-polarization images of vertical thin sections extracted from the three sites shown in Figure 5.41a. The differences in the microstructure are related to the ice types and their SAR backscatter signatures as explained in the text [adapted from Johnston, 1998].

nearly same backscatter. This illustrates the difficulty of linking a unique backscatter to the different ice types in these two sites. 5.5. BIOMASS ACCUMULATION AT THE BOTTOM OF THE ICE In early spring when the sun starts to shine after the end of long polar nights, algae start to grow at the bottom of sea ice. During the Arctic or Antarctic springtime, high concentrations of microalgae have been observed in the interstices of the lower margin of sea ice floes. Often algal growth also occurs in a thin layer of surface water immediately under the ice cover or associated with semi-consolidated frazil ice. For at least one to three months, ice algal blooms enhance and extend biological production in polar waters. Biomass accumulation of sea ice algal populations eventually depends largely on climatic and environmental variability. The algae are greenish brown in color, and when attached to the bottom of the ice cover are known as “ice algae.” Its growth changes the color of the ice bottom to dirty green. This biomass is important in ecosystems of the polar regions. In the Antarctica, they provide food for krill that scrape off the ice algae from the underside of the sea ice; the krill, in turn, provide food for the higher orders of lives. Since the habitat of these algae is sea ice, both marine biologists and ice scientists have studied this phenomenon. Ackley et al. [1978] used fluorescence

measurements of algal population in sea ice cores and found an algal community strongly associated with salinity maxima within the ice. It is estimated that ice algae may contribute about 57% of the total Arctic marine primary production [Gosselin et al.,1997]. In a series of three papers, Cota et al. [1991a, 1991b, 1991c] provided an in-depth analysis of in-situ measurements related to the ecology of bottom ice algae carried out in the Canadian High Arctic. The authors discussed wide-ranging aspects on ice algae, such as the variability, dynamics, distributions, productivity, and comparative physiology as well as environmental controls. Algae communities are found either attached to ice or in spaces between ice grains and subgrains. The details on these algae communities and how their presence affects the habit of sea ice crystals can only be determined by forensic type of sea ice microstructural investigations. Figure 5.42a is an illustration of the bottom section of a 1.86 m long and 100 mm diameter freshly sampled, algae-infested core of FYI with vertically oriented S3 type of columnar grain structure. The bottom of the core clearly shows the green-brown zone of algae-infestation, up to about 40 mm above the ice– water interface. A 5 mm thick and 20 mm long vertical section was cut from the core bottom for examining the extent of the algae-infestation (Fig. 5.42b). The concentration of algae population decreases rapidly with the increase in distance from the ice–water interface. Figure 5.43a is a cross-polarized light view of a horizontal thin section within or close to the skeletal layer, cut at

POLYCRYSTALLINE ICE STRUCTURE 231 (a)

(b)

Figure 5.42 (a) Bottom of an algae-infested freshly sampled 100-mm-diameter FYI core from Resolute Passage in early May 1991 and (b) 5-mm-thick vertical section of the core bottom. Spacing between dashes of the scale are in millimeters (N.K. Sinha).

(a)

(b)

Figure 5.43 (a) Cross-polarized view of a circular 100-mm-diameter horizontal thin section about 12 mm above the core bottom together with a vertical thin section of the bottom 12 mm of the core, and (b) a close-up of the boxed area in (a), but with different orientations of the polarizers (photographs by N.K. Sinha).

about 12 mm above the bottom of an ice core. A vertical thin section prepared from the central area of the bottom 12 mm of the same core (immediately below the horizontal section) is also shown in this photograph. The boxed area in Figure 5.43a is shown in Figure 5.43b after a small angular rotation of the cross-polarizers. No doubt, infestation of algae are limited along the numerous subgrain

boundaries of the grains. This was proven by looking at those boundaries through a magnifier and a microscope. When the section melted completely, one could see the algae material on the glass plate. Details of a number of grains in the horizontal section shown in Figure 5.43 are shown in Figure 5.44. Note the presence of the algae cells, with appearances

SEA ICE

c-a

xis

232

Algae cells

Figure 5.44 Cross-polarized view of a horizontal section in the skeletal zone exhibiting the algae colonies primarily along the parallel rows of subgrain boundaries and hence parallel to the basal planes of the grains or normal to the c axis, pointed out in one of the grains (micrograph by N.K. Sinha).

of round or elliptical shapes. These cells are located along the length of the parallel rows of subgrain boundaries. The microstructural information presented in this and similar figures confirms, beyond any doubt, that the presence of the brine along the boundaries between the grains and subgrains in FYI, particularly in the skeletal layer, serves as habitat for ice-specific organisms. Once again, this constitutes the base of the food web for marine species. These include not only algae but also phytoplankton, bacteria, viruses, and sufficiently small invertebrates that can traverse the narrow brine passages along the subgrain boundaries. Photomicrographs of thin sections of artificial saline ice were used by Krembs, Eicken, Deming [2011] to reveal brine pocket habitat for organisms at high magnification. Organisms that reside in brine pockets produce an organic gelatinous substance, called extracellular polymeric substances (EPS). This material is responsible for retaining salts to improve the habitability of primary production. It also changes the microstructure of the brine pockets. Krembs, Eicken, Deming [2011] found that EPS, which is observed microscopically in the brine inclusions, influences the geometry of brine inclusions. 5.6. INFORMATION CONTENTS IN POLYCRYSTALLINE ICE STRUCTURE This section includes quantitative information that can be retrieved from the crystallographic structures revealed in thin sections of sea ice. The information is provided to cover three items: (1) geometric characteristics of ice crystals, (2) brine pockets and channels and their spacing, and

(3) air bubble inclusions. Information is obtained from digital analysis of 2D photographs of thin sections. In general, the studies on the geometric characteristics of ice crystals and inclusions are limited in number and scope, with no comprehensive review published so far. Therefore, the information presented in this section, which is selected from a few studies, may not be generalized. Statistical characterization of ice crystalline structure is needed to model the reflection, radiation, and scattering of electromagnetic waves as they interact with sea ice (Chapter 12). This is also important for interpretation of remote sensing observations and the retrieval of ice parameters. Perovich and Gow [1996] suggested that a 3D characterization of the volume of ice crystals and inclusions would be ideal, but it was not possible using current technology at that time. Obviously, a large number of thin sections is needed to construct just one 3D view, which is not possible. Quantitative metallography or stereology is a comprehensive and widely accepted body of experimental and analytical methods for characterizing a 3D microstructure from 2D sections. Yet, it has not been used for sea ice thin section images. A breakthrough in 3D visualization of the sea ice composite (solid, liquid, and gaseous) was introduced by using the non-destructive technique of X-ray micro-computed tomography (μCT) to quantify the shape, size, and topology of the brine channels and air bubbles in sea ice. This technology was introduced in Ketcham and Carlson [2001]. The μCT is based on the fact that phases of different density within a composite material have varying absorption and transmission of X-ray radiation, allowing separation of different phases. That is how solid ice crystals, brine, and air in sea ice can be separated. While most of the results presented in this section were generated from digital image analysis of 2D thin section samples, brief results from using the μCT technique are shared in section 5.6.2. 5.6.1. Geometric Characteristics of Crystalline Structure A polycrystalline body of ice is an organization of crystals of different shapes, sizes, and lattice orientations into a space-filling mass. A pioneering work to generate the first geometrical characterization of ice crystalline structure and statistical characterization of the spatial variation of permittivity is presented in Perovich and Gow [1991]. The study involved digital analysis of thin sections photographs of polycrystalline saline thin ice grown in an outdoor tank at the Cold Regions Research and Engineering Laboratory (CRREL). Both geometric characterization of sea ice crystals and their permittivity are needed to model the microwave scattering [Tsang and Kong, 1981 and Winebrenner et al., 1989]. The permittivity was obtained from statistics of brine inclusions and is needed to model thermal microwave emission

POLYCRYSTALLINE ICE STRUCTURE 233

(section 12.2 and 12.3). The approach used by Perovich and Gow [1991] was simple, but rather lengthy and computationally exhaustive. Later work was presented in Johnston and Sinha [1995] that involved digital image analysis of thin sections prepared using the DMT (section 4.2.2). This was an extended version of the methodology developed in Barrette and Sinha [1996]. In polycrystalline body of ice, the grain boundaries can be used to identify individual ice crystals in photographs of thin sections. Subsequently, geometric characteristics of grains and subgrains can be obtained. Grain boundaries can be delineated either manually or using digital image processing techniques. Both horizontal and vertical sections are required for an adequate description of the grain structure in a polycrystalline mass. For directionally solidified columnar-grained S2 or S3 type of ice, horizontal thin sections are preferable for the analysis, as they provide a satisfactory description of the cross-sectional structure of the grains. Images of thin sections taken through cross-polarized and parallelpolarized light (section 4.3.2) provide complimentary information and often images taken with scattered-light are necessary for sea ice analysis. Although the grain boundaries in sea ice are most easily identified using cross-polarized light, the subgrain boundaries are distinguished more clearly using parallel-polarized light. Both types of images in full color are, therefore, necessary because the absence of details in one polarized image is usually provided by its counterpart. Moreover, images augmented by scattered lights are also found to be appropriate for the illumination of inclusions and thereby identifying the subgrain boundaries with extremely small lattice mismatch. Colored photographs are preferred for accurate identification of grain boundaries in thin section photographs. One of the most common sources of error, if gray tone images are used, is the transformation of different interference colors into the same gray value [Eicken and Lange, 1991, Eicken, 1993], thereby making the accurate identification of grain and subgrain boundaries difficult. Color, brightness, contrast, and sharpness in the digitized images of thin sections can be adjusted as required. The lack of details potentially can interfere with the ability of the image analysis to differentiate between the grain boundaries and the grains themselves. Once the digital photographic images are available, grain and subgrain boundaries in the images can be blackened with the freehand tool using any available image processing software. Utmost care is required to ensure that inter- and intragranular boundaries are outlined completely, i.e., not one pixel is absent from the darkened margins. The efficiency of digitized image analyses is demonstrated by the enhanced ability to differentiate between grains and subgrains. Since adjacent subgrains

(or cross sections of platelets) are separated by small angle boundaries with extremely small mismatch of the crystal lattice, the color gradient between subgrains is not usually well defined. Therefore, the mosaic pattern formed by intersecting rows of brine inclusions is used to outline the substructure. Close inspection of the orientation and shape of the inclusions provide the information necessary to determine the general direction of the subgrain boundary. To some extent, the accuracy of this method is compromised by the intra-subgrain brine inclusions because often rows of brine pockets do not necessarily mean any lattice mismatch on either side. An alternative method for the manual delineation of grain and subgrain boundaries in the image is to use an automated edge detection technique available in any image processing software. The software usually calculates the area of the grain or subgrain in addition to four parameters: (1) the perimeter, (2) the Feret diameter (the diameter of a fictitious circular shape that has the same area as the crystal being measured), (3) a shape factor, which is an indicator of how nearly circular the crystal is, and (4) the compactness factor (an alternative method for assessing the degree of circularity of a crystal), defined as the ratio of the square perimeter over the area. Johnston and Sinha [1995] examined the grain- and subgrain-scale characteristics of natural columnar-grained S3type FYI, sampled from Frederick Hyde Fjord (FHF) (83 11’N, 29 50’W) northern Greenland in May, 1994 using a computer assisted approach. The approach involves four steps. The first entails digitizing the colored photographs of thin sections (from cross- and parallelpolarized images) into 256 gray tone images. In the second step, the digitized images of the same thin section are input to a desktop publishing software to determine the location of grain boundaries (Figure 5.45a). The absence of details in one polarized image is usually provided by its counterpart. In case where boundaries could not be discerned from either image, they should be blackened with the freehand tool of the software. In the third step, the digitized image with defined grain and subgrain boundaries is converted to a binary image with grains presented in white and their boundaries in dark tone (Figure 5.45b). In the fourth step, image analysis software package has to be used to identify properties of grains and subgrains. The software determines the exact number of pixels in each grain/subgrain, which represents the area, as well the four parameters mentioned above. Major axis length is automatically determined by searching all of the border pixels of the object and selecting the two pixels that were separated by the furthest distance (Figure 5.45c). Once the major axis is defined, the minor axis is determined as the line drawn perpendicular to the major axis, which joins the two boundary points that are set furthest apart. The software can tabulate the exact number of grains and subgrains. The area of each object is

234

SEA ICE (a)

(b)

10 mm

0

10 mm

(c)

Figure 5.45 (a) Digital image of a horizontal thin section of columnar grained, S3 type FYI from Frederick Hyde Fjord, northern Greenland at a depth of 130 mm. (b) Delineated grains and subgrains in binary image with labeled object numbers, major and minor axes are shown, and (c) enlargement of the boxed area is shown in (b) [adapted from Johnson and Sinha, 1995 / John Wiley & Sons, Inc.].

calculated by summing the number of pixels enclosed within its perimeter. The highly oriented nature of the S3-type sea ice is made apparent in Figure 5.45c, especially with respect to the minor and major axes. Close inspection of the minor axis of subgrains provides further evidence of the geometric selection process with horizontally oriented c axis. The minor axis of the subgrains tends to be parallel to the c axis. Figure 5.46 presents a series of histogram characterizing the grains and subgrains in Figure 5.45. In this case, at a depth of 130 mm, the mean area of subgrains within a grain is approximately 60% less than the area of the overall grain. Minor and major axes lengths of the subgrains are 20–30% smaller than those of the grains. The length of the minor axis of subgrains provides an indication of the brine-layer spacing. The analysis of minor axis length for FYI from Frederick Hyde Fjord shows that the size of individual subgrains ranges from about 0.1 to 2.5 mm. The data presented in Figure 5.46 are from one depth only (130 mm below the surface). The difficulty of generating 3D picture of this information using this method can be appreciated. For thermodynamic and scattering modeling, it is better to generate this information at one or two subsurface layers, say at 10 and 20 mm, though it is still a laborious process.

There seem to be correlations between the lengths of the major and minor axes of the grains/subgrains. These are shown in Figure 5.47 for the same section in Figure 5.45. The equation most representative of the scattered data is a power law function with a constant ranging between 2.3 and 2.8 and an exponent around 0.9, as shown in the figure. The regression coefficient, R2, range near 0 indicates a poor fit while values near 1 indicate a good fit. Figure 5.47 shows better correlation between major and minor axes for the grain data (R2 = 0.74) than for the subgrains (R2 = 0.54). However, as a first approach, the regressed plots of minor versus major axes lengths for both grains and subgrains may be represented adequately by a ratio of length to width approximating 2.5:1. These results provide a statistics-based support for the observations made by Weeks and Hamilton [1962] on grains in YI at Point Barrow in Alaska. The authors reported a ratio of 2:1 for the sidewise to edgewise growth of individual grains. As the brine drains out of the FYI, many demarcation lines in the intragranular spaces seem to disappear. This applies particularly to the rows of brine pockets along the subgrain boundaries with extremely small lattice mismatch. A systematic set of results on geometric characteristics of FY columnar ice formed in Allen Bay (Canadian

POLYCRYSTALLINE ICE STRUCTURE 235 333

Area of grains

Area of subgrains 150 Frequency

Frequency

100 x = 6.05 mm2 50 0

x = 2.7 mm2

100 50 0

5

15

25

35

45

55

65

0.5

75

3

Area (mm2) Major axis of grains Frequency

Frequency

x = 4.5 mm

100 50

100

x = 3.4 mm

50 0

0 1

3

5

0.5

7 9 11 13 15 17 19 22 26 Major axis length (mm)

Frequency

Minor axis of grains

150 Frequency

13

Major axis

150

150

5.5 8 10.5 Area (mm2)

x = 1.6 mm

100 50 0 2

0.5

5 6.5 8 3.5 Minor axis length (mm)

3

13

Minor axis

100 80 60 40 20 0

9.5

8 5.5 10.5 Major axis length (mm)

x = 1.1 mm

0.1

0.7

1.3 1.9 2.5 3.1 Minor axis length (mm)

3.7

Figure 5.46 Probability distribution of macroscopic parameters for grains (left) and subgrains (right) corresponding to the objects that appear in Figure 5.45. Mean values are shown as x [adapted from Johnson and Sinha, 1995].

(a)

(b)

Grains of thin section

Subgrains of thin section 16

y = 2.6992×0.9497 R2 = 0.7419

20

Major axis length (mm)

Major axis length (mm)

25

15 10 5 0

14

y = 2.7995×0.9176 R2 = 0.5424

12 10 8 6 4 2 0

0

2

4

6

8

10

12

Minor axis length (mm)

0

2 4 Minor axis length (mm)

6

Figure 5.47 Correlation of the major and minor axes lengths of (a) grains and (b) subgrains for the images in Figure 5.45.

Central Arctic) is presented in Johnston [1998]. An example is presented in Figure 5.48 from an ice core obtained from site 3 in Allen Bay (shown in Figures 5.41a and 5.41b). Grain boundaries were delineated manually in the digitized photographs on the horizontal thin sections as shown in Figure 5.48. Selected grains in both the horizontal and vertical planes are designated g1, g2,…, g8. Their locations in the vertical sections indicate that

grains g1, g2, and g4 are eliminated before they can form well-established columns. This is also confirmed by their absence from horizontal section at depth of 100 mm. Grain g5 persists throughout the vertical growth, while other grains persist but only to certain depths. Statistics of the mean areas of grains and subgrains along with the length of their major and minor axes are presented in Figure 5.49. The area of the grains spans

236

SEA ICE (a)

(b)

g6

g4

g2

g1

5

4 cm g3

g6

5

g4 g3

g2

g1

10 mm

(c)

10 cm g8

g3

5

g7

g8

g3

5

g7

10 mm 10 mm

Figure 5.48 Photographs of (a) vertical thin section of FYI and (b) and (c) two horizontal sections obtained at the two depths marked by the dashed lines in the vertical section. The crystal boundaries in the horizontal sections are delineated manually. Grains (crystals) are designated in all sections as g1, g2,…, g8 [ Johnston, 1998 / Cambridge University Press].

560 mm2, with most of the grains having an area less than 150 mm2. In comparison, subgrains are much smaller with majority having an area less than 50 mm2. The mean area of the grains and subgrains are 86 mm2 and 15 mm2, respectively. The difference between the major and minor axes length of the grains is roughly 8 mm, while that of subgrains is only 3 mm. This means that grains tend to be elongated but subgrains are mostly isotropic in shape. 5.6.2. Geometric Characteristics of Brine Pockets in First-Year Ice When impure melt such as sea water is subjected to unidirectional solidification, microsegregation (impurities in alloy after solidification) occurs at the solid–liquid interface. The segregation processes in case of saline water, leads to the development of solid phases (i.e., crystals of pure ice) practically free of any solutes because the hexagonal ice lattice allows only traces of impurities. The impurities are pushed to the solid–liquid interfaces and the solidification front develops a cellular, dendritic, or lamellar structure filled with brine (section 2.3.2). Brine pockets and drainage channels provide pathways for the exchange

of heat, gases, salts, and are known to affect the mechanical properties of sea ice. For directionally solidified (DS) growth conditions, as described and explained in section 2.3.2, the dendritic interface of sea ice at the sea water is the initial mechanism responsible for entrapment of salts in the form of brine pockets between and within the grains. Its relatively enlarged surface area slows down the growth since the freezing temperature of brine is lower than that of seawater. For example, brine with 230‰ salinity freezes at −21.1 C or less. The dendritic structure causes the grain boundaries to be convoluted and leads to the development of intragranular substructures or subgrains with small lattice mismatches (Figure 2.19). Brine pockets are present not only at grain boundaries, but also inside grains at subgrain boundaries of sea ice. They tend to be located on planes parallel to the basal plane and hence normal to the optic or c axis of a crystal. Brine pockets observed in thin sections are the cross section of what actually exist as 3D brine layers. The spacing between adjacent planes or layers containing brine pockets is called brine-layer spacing (defined in Figure 2.19). Since brine layers are vertical in DS ice, their spacing can be represented by the spacing of the brine rows found in a horizontal section.

POLYCRYSTALLINE ICE STRUCTURE 237 0.20 Mean grain area = 86 mm2

0.30 0.20 0.10

Probability density

Probability density

0.40

0

150 50 100 Area of grains (mm2)

0.10 0.05

0

200

8 16 24 32 40 48 56 64 72 80 Area of subgrains (mm2)

0.20

0.30 Mean major axis length = 16 mm 0.20

0.10

Probability density

Probability density

Mean area of subgrains = 15 mm2

0.00

0.00

0.00

Mean major axis length = 8 mm 0.15 0.10 0.05 0.00

0

0

6 12 18 24 30 36 42 48 54 60 Major axis length of grains (mm)

6 12 18 24 30 36 Major axis length of subgrains (mm)

0.20

0.30 Mean minor axis length = 6 mm 0.20

0.10

0.00 1

4 7 10 13 16 19 22 25 28 Minor axis length of grains (mm)

Probability density

Probability density

0.15

Mean minor axis length = 3 mm 0.15 0.10 0.05 0.00 0.0 0.8 1.6 2.4 3.2 4.0 4.8 5.6 6.4 Minor axis length of subgrains (mm)

Figure 5.49 Measurements of geometry of grains and subgrains of SYI from Allen Bay, central Arctic, obtained in May 1997. Measurements are taken at 10 cm depth [Johnston, 1998].

Brine-layer spacing is one of the most important parameters in describing the structure of sea ice because of its impacts on the mechanical properties of ice and remote sensing observations. Brine-layer spacing is often called the “plate width” or “plate spacing” because it is a measure of the space between adjacent platelet ice protrusions at the ice–water interface (note how the term “platelet” is used here differently than the use of platelet ice shown in Figure 5.5, which is usually coined with Antarctic ice). It depends on growth rate and crystallographic orientation, hence on the orientation of water current under the ice. Since the growth rate is related to climatology, the vertical brine-layer spacing profile in the ice provides a record of previous weather conditions. As a rule of thumb, brinelayer spacing is inversely proportional to the growth rate of sea ice, and could also be dependent on crystallographic orientation [Tabata and Ono, 1962, Weeks and

Hamilton, 1962]. Since ice growth rate would generally decrease as ice grows thicker, it is expected that brinelayer spacing increases progressively with depth. This is confirmed in measurements on natural sea ice by Gow and Weeks [1977], Weeks and Ackley [1982] and several other classical publications on sea ice. Those publications, however, stated the difficulty to explain the fluctuations in brine spacing because of the limited data available on the growth history of sea ice. In fact, the difficulty in obtaining the complete growth history of sea ice at a given location was the main reason for the lack of meaningful data on brine-layer spacing in the pre-1984 literature. Using a record of daily mean air temperature, histories on snow and ice thickness, and appropriate physical constants for snow and ice Nakawo and Sinha [1984] estimated the growth rate of sea ice of a given thickness using equation (2.8). The calculated growth rate is plotted

238

SEA ICE 0

0

0.05 0.10 0.15 0.20

0.2

Depth (m)

0.4 c 0.6

b

0.8 a

1.0

Bottom 1.2 0

1.0 1.5 0.5 Growth rate (cm/day)

2.0 4

6 8 10 Salinity (‰)

12 0.2 0.4 0.6 0.8 1.0 1.2 Brine layer spacing (mm)

1.4

Figure 5.50 Profiles of growth rate, salinity and brine-layer spacing from an FYI core obtained in Eclipse Sound, Canadian Arctic in January 1978. The ice thickness was 1.17 m. Curve b represents the running mean of calculated growth rate (curve a), for an interval of ± 50 mm for every 25 mm. The average ice growth rate was 8.5 cm/day [Nakawo and Sinha, 1984].

1.0

0.10

0.15

0.20

0.9 Brine layer spacing (mm)

against ice depth as shown in Figure 5.50. The running average of the growth rate would be more realistic because the thermal resistance of the snow–ice system dampens the variations in daily air temperature and therefore the growth rate. The figure shows three maxima of ice growth rate at about 0.2, 0.5, and 1.0 m. The authors found that they correspond to cold periods in November, December (1977), and January (1978) near the sampling site (near Pond Inlet on Baffin Island, Canada). The small growth rate at about 0.8 m corresponds to the warm period in late December between the latter two cold periods. The similarity between the running average of the growth rate and the salinity profile (middle graph in Figure 5.50) is an indication of the correlation between the two parameters. Brine spacing was measured manually from photographs of thin sections. Figure 5.50 also shows the relation between growth rate and brine spacing; a large growth rate seems to give a small spacing, and vice versa. This general tendency agrees with previous observations as mentioned above. However, the growth rate decreased with depth near the bottom, yet the brine-layer spacing also decreased sharply. The same data from Figure 5.50 are plotted in Figure 5.51 to show the relation between the brine-layer spacing and the ice growth rate. The spacing tends to be inversely proportional to the growth rate (the broken line in the figure), in accordance with the theoretical results by Bolling and Tiller [1960] and laboratory studies [Rohatgi and Adams, 1967, Lofgren and Weeks, 1969]. The two anomaly points in Figure 5.51 reflect the anomaly shown in Figure 5.50; when the brine-layer spacing decreases while the ice growth rate also decreases near the bottom of the core. Nakawo and Sinha [1984] found that brine

0.8 0.7 0.6 0.5

Two bottom sections

0.4 0.3 0.5

1.0

1.5

2.0

Growth rate (cm/day)

Figure 5.51 Plot of average brine-layer spacing (open circles) versus the corresponding growth rate of curve b in Figure 5.50. The solid circles indicate the bottom two sections. The broken line shows a least-squares fit, assuming that the spacing is inversely proportional to the growth rate [Nakawo and Sinha, 1984].

spacing depends on crystallographic orientation, which was not taken into account in the results presented in Figure 5.50. They suggested that this might be the cause for the anomalies. In general, the correlation between brine-layer spacing and ice growth rate provides a record of the conditions, especially, the weather under which ice is formed. Jefferies et al. [1993] found similarities in the

POLYCRYSTALLINE ICE STRUCTURE 239

measured brine spacing in ice from the Arctic and Antarctic; so he concluded that fast ice growth rates are similar in the two polar regions. One of the studies on brine-layer spacing that includes a full record of the ice growth rate is presented in Sinha and Zhan [1996]. The authors conducted a study in Resolute Bay, Canadian High Arctic in May 1992, and obtained the record of the ice growth starting shortly after the ice formation in November 1991 (when it was safe to operate a light truck on it) from the Resolute weather station of Environment Canada. The growth season of 1991–92 was rather unique because the ice cover was flat and the snow cover was very uniform—a good indication of unidirectional freezing. Moreover, the mean daily air temperature for the entire ice growth season stayed well below the freezing point, which ensured almost steady growth. The ice growth record allowed the authors to develop the following relationship for the dependence of ice thickness hi on the number of freezing day N: hi = 0 0162 N − 2 71 × 10 − 5 N 2 − 1 48 × 10 − 8 N 3 (5.2)

λ1

λ1

Figure 5.52 A macrograph of a horizontal thin section at a depth of 1.59 m of unidirectionally frozen, columnar-grained sea ice in Resolute Bay (Canadian High Arctic). Brine spacing λ1 is shown. The double-headed arrows indicate the average c-axis of the grains (N.K. Sinha).

which gives a growth rate of

0.0

dhi dN = 0 0162 − 5 42 × 10 − 5 N − 4 44 × 10 − 8 N 2 (5.3) 0.5

Ice thickness (m)

With these two equations the growth rate at any depth of the ice cover can be obtained and therefore related to the estimated brine-layer spacing. Sinha and Zhan [1996] estimated the spacing through two analytical methods of thin sections. The first was in the NRC cold room in the field at Resolute, where a microscope and transmitted light were used (this is referred to as field measurements), and the second was by applying image processing techniques on digitized slides of the thin sections in a laboratory in the NRC (Ottawa). This is referred to as laboratory measurements. Figure 5.52 is an example of a macrophotograph of a horizontal thin section showing brine along grain and subgrain boundaries (white color) taken at −16 C, using combined cross-polarized and scattered light. The brine spacing λ1 is shown. Figure 5.53 shows a profile of brine spacing from one of the ice cores. Brine spacing is small near the ice surface and increases with depth. The subsurface layer represents the ice cover when it was thin, hence when the ice was growing at higher rate and subsequently smaller brine spacing. As the ice thickens, the growth rate decreases and the brine spacing increases. Large scatter is observed in both field and laboratory measurements, with the two sets differing by more than 0.1 mm at some depths. Field measurements tend to show larger fluctuations than the estimates using the digital image processing. The same reference [Sinha and Zhan, 1996] presents a plot between brine-layer spacing and the salinity (Figure 5.54). The average brine spacing λ1 tends to decrease with the increase in salt entrapment

1.0

1.5

2.0

2.5 0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

Dentritic spacing (mm) Figure 5.53 Dependence of brine-layer spacing on depth. Data obtained from columnar-grained sea ice in Resolute Bay, Canadian High Arctic. The ice core from which the data were obtained was extracted in May 1992. Open circles and squares indicate field and laboratory measurements, respectively; while solid triangles are average brine spacing [Sinha and Zhan, 1996].

(i.e., salinity) according to the following linear regression equation: λ1 = 1 11 − 0 093 S (5.4) This equation was developed from the available set of data and therefore should not be extrapolated for

240

SEA ICE

where, μ and σ are the mean and standard deviation of ln(A). The probability density function (PDF), which is the first derivative of the CDF is given by:

1.0

Brine spacing (mm)

0.9

PDF A =

0.8 λ1 = 1.11–0.093S

0.7

0.6

0.5 0

1

2

3

4

5

6

Salinity (S in ‰)

Figure 5.54 Dependence of brine-layer spacing on salinity for mature columnar-grained sea ice in Resolute Bay in May 1992 [Sinha and Zhan, 1996].

salinities higher than about 5‰ or lower than about 2‰. If extrapolated to zero salinity, the equation gives a brinelayer spacing of 1.1 mm. This is obviously erroneous and may lead to wrong deductions. For growth rates generally encountered in nature dendritic growth does not occur in freshwater. Perovich and Gow [1996] conducted a wide-ranging study to characterize brine and air inclusions in all ice types (YI, FYI, and MYI). They used an image processing system to partition images of thin sections of sea ice into constitutive components of ice, brine (in FYI), and air (in MYI). This was achieved by exploiting differences in gray shade between ice and its inclusions (inclusions are the darkest features in a black and white image). The gray shade cutoff for the ice–brine partitioning in the cases of YI and FYI was determined using the brine volume of the sample, which was determined using equation (3.26). After identifying the inclusions using standard image processing software, a suite of parameters was calculated: the number and size distribution of the inclusions, their major and minor axes, their orientation, and shape factor. A main finding from that study was the nearly perfect fit of the cross-sectional area of inclusions to the lognormal distribution. This was concluded from analysis of 40 thin sections encompassing a wide range of ice types (with granular and columnar crystalline structure), and including cases with brine pockets and air bubbles. The lognormal form of the cumulative distribution function (CDF) of the cross-sectional area A of the inclusions is given by the equation: CDF A =

1 erfc 2

– ln A − μ σ 2

(5.5)

1 − ln A − μ exp 2 A 2σ 2 2πσ 1

2

(5.6)

Perovich and Gow [1996] pointed out that the lognormal distribution is common for natural parameters when random processes contribute multiplicatively rather than additively to the parameter (the latter case leads to a normal distribution). They also indicated that the lognormal distribution has been used to describe pressure ridge keels [Davis and Wadhams, 1995] and grain coarsening in snow [Colbeck, 1987]. An example of the observed CDF and the derived PDF of the brine inclusion area from a horizontal section of columnar YI is presented in Figure 5.55. The deviation of the measured area from the lognormal distribution at the lower end of the area scale is due to the limitation of the software in measuring very small areas as indicated before. The upper area limit of 0.1 mm2 in the figure is also smaller than the findings in other studies [Cole and Shapiro, 1998 and Light, Maykut, Grenfell, 2003]. A more recent study by Light, Maykut, Grenfell [2003] used an automated imaging system for observing ice microstructure in vertical thin sections prepared from landfast ice near Point Barrow, Alaska in May 1994. The system allows detection of brine inclusions as small as 0.003 mm. This is much smaller than the minimum dimension of approximately 0.05 mm presented in Perovich and Gow [1996]. This led to different results of brine inclusion dimensions and their number density. The brine inclusion dimensions were found to range from less than 0.01 mm to nearly 10 mm, with number densities averaging around 24 brine pockets per mm3. This is an order of magnitude larger than the average brine pocket dimension of 1.0 mm, and the number densities range from 1.0 to 4.5 per mm3, as reported in Perovich and Gow [1996]. Light, Maykut, Grenfell [2003] estimated two more geometrical parameters of brine inclusions: the number density per unit volume N and the aspect ratio of the brine inclusion shape γ. They found that the data of both parameters can be well represented by the following power law in terms of the major length of the brine inclusion l: N l = 0 28 l − 1 96

(5.7)

This equation is valid for the range 0.01 mm ≤ l ≤ 8 mm. Integration of the above equation between these two limits of l yields the number density of 24 inclusions per mm3 as mentioned above. The brine inclusion shape is given by: γ l = 10 3l 0 67

(5.8)

POLYCRYSTALLINE ICE STRUCTURE 241

1.0

50

0.8

40

0.6

30

0.4

20

0.2

10

0.0 1E-3

0.01

0.1

1

Probability density

Cumulative probability

Z=19 cm

0

Area (mm2)

Figure 5.55 Horizontal thin section of columnar YI at 190 mm depth showing (left) segmentations and the distributions of the area of brine inclusions (right). The solid squares are the points on the cumulative distribution, while the best fit curve is shown as a solid line. Also shown separately is the PDF of the brine area. The scale is provided by the ruler with marking in mm at the top of the digitized photograph [Perovich and Gow, 1996, Figure 2 / with permission of John Wiley & Sons].

The equation applies for 1 ≤ γ ≤ 70 and l > 0.03 mm. Most of the data shown above on brine inclusion geometry were obtained in the 1990s, and before using images of horizontal or vertical thin sections. This restricts the retrieved information on the inclusion’s geometry to cross sections and masks the 3D nature of the microstructure. Cole and Shapiro [1998] demonstrated that additional information about ice microstructure could be obtained by complementing the data obtained from both horizontal and vertical thin sections. They found, for example, that brine inclusion shapes ranged from nearly spherical to elongated with vertical extent exceeding 15 times the diameter. Almost a decade later, researchers started to use the non-destructive μCT to detect brine pockets and channels in their 3D perspective in sea ice samples. Pringle et al. [2009] used this technique to analyze laboratory-grown sea ice and found that between −18 C and −30 C, the porosity ranged from 2% to 12% with near-parallel intracrystalline brine layers whose connectivity and morphology vary with temperature. The study, however, suggested extending the application of the μCT technique to assess the variation of the morphology of brine pockets and channels with ice growth rate. This needed to verify the empirical information obtained from using the 2D thin sections. Maus et al. [2009] discussed the potential application of the μCT technique in characterizing essential features and evolution processes of the microstructure of sea ice. Leib-Lappen, Golden, Obbard [2017] used the μCT technique to examine six ice cores extracted from

the Ross Sea in the Antarctica. An example of 3D image of sea ice samples obtained at 10 cm, 50 cm, and 150 cm depth from the same core, showing the configuration of brine channels, is presented in Figure 5.56. At 10 cm depth, the ice was composed of frazil crystals and brine channels. The crystalline structure features granular texture with a high degree of isotropy. Another high degree of anisotropy from the sample at 50 cm depth represents vertically aligned brine channel from columnar ice with spacing roughly at 0.5 mm. This is consistent with the data presented in Figure 5.54 (brine spacing increases when salinity decreases). The sample at 150 cm depth is from platelet ice at the bottom of the core, where salinity increases again. More isotropic configuration is revealed at this depth due to the higher degree of lateral branching. Leib-Lappen, Golden, Obbard [2017] also developed metrics to quantify the shape, size, and topology of the brine channels and air bubbles in sea ice from the images. Two metrics representing the diameter and the spacing of brine channels are presented in Figure 5.57 as a function of depth. The average diameter is 1 mm and the average brine spacing is 15 mm. No trend of these two parameters with depth is identified. In conclusion, the application of the X-ray micro-CT with its 3D visualization capability is a step forward to reveal more thorough perspective of crystallite structure of sea ice, in particular the brine and air bubble inclusions. This information is viewed as complementary to the information from the analysis of 2D thin sections using photography and microphotography. More applications of

242

SEA ICE

(a)

(b)

(c)

0

0

20

20

40

40

60

60

80 100 120

Depth (cm)

Depth (cm)

Figure 5.56 Images of 3D perspective of 7.5 mm3 cubes from an ice core in the Ross Sea, Antarctica, using the μCT technique with visualization software, showing brine channels at: (a) 10 cm depth, (b) 50 cm depth, and (c) 150 cm depth. Those depths include frazil, columnar, and platelet ice crystals, respectively [Leib-Lappen, Golden, Obbard, 2017, Elsevier / CC BY 4.0].

80 100 120

140

140

160

160

180

180

200 0.2 0 0.1 B. channel diameter (cm)

200 0 1 2 3 B. channel spacing (cm)

Figure 5.57 Variation of brine channel diameter (left) and brine spacing (right) with ice depth. The open square, inverted triangle, and circle are data from different sites, all FYI in the Ross Sea, Antarctica using the μCT technique [Leib-Lappen, Golden, Obbard, 2017, open source].

this technique in a variety of sea ice situations will enrich our knowledge about the inclusions in sea ice at least at a gross scale. While the traditional analysis of thin section using micrography overlooks the third dimension, it still provides unparallel details on the spatial configuration of grain, subgrain boundaries, dislocations in the crystallographic structure, and the situation of inclusions within the subgrain boundaries. This information results from molecular-scale processes, and can be revealed using micrographs of etched sections as described in section 4.4.2.

5.6.3. Geometric Characteristics of Air Bubbles Air bubbles are the key inclusions in MYI and the main scattering elements of microwave signals when wavelength matches the characteristic dimensions of the bubbles (ranging from a few millimeters to a few centimeters). Nevertheless, data on the geometrical characteristics of air inclusions in sea ice (FYI or MYI) are scarce in the literature. Studies on air inclusions in FYI were driven by the interest in the optical properties of ice. That, of course, preceded the interest in microwave interaction with ice, which identified the need for information on characteristic dimensions of bubbles in MYI as a priority to initiate scattering modeling. For FYI, Grenfell [1983] determined the bubble diameter, size, and number density from thick sections (5 mm thick) of thin sea ice in a lead in Beaufort Sea in October 1974, using a 20-power eyepiece with a 0.1 mm reticle. The study concluded that air bubbles occupied a range from 0.1 mm to 2 mm in diameter, and the size distribution could be represented by the log of the number of particles per cubic millimeter N(r) in the interval r to (r + dr). This parameter can be determined using the following power law: N r dr = N o r − 1 24 dr

(5.9)

where, No is a constant equal to 0.007. Light et al. [2003] examined the microstructure of FYI in a temperature-controlled laboratory experiment to obtain the size distribution of brine inclusions and the gas bubbles inside them. They used an imaging system of resolving inclusion sizes of less than 0.01 mm. The authors found that the dimensions of brine inclusions ranged from less than 0.01 mm to nearly 10 mm. Gas bubbles in the samples were

POLYCRYSTALLINE ICE STRUCTURE 243 Table 5.2 Second-year ice bubble characteristics measured in Mould Bay in the spring of 1983. Percentage

Depth (cm)

Number of Bubbles

Circular

Elliptical

Average Bubble Density

0.62 3.0 3.8 5.9 8.0 10.1 12.1 14.5 15.1

1048 1303 363 297 269 418 246 123 239

73.4 61.3 63.4 68.0 70.0 57.8 59.8 58.8 85.2

26.6 38.7 36.5 32.0 30.0 40.2 40.2 41.2 18.8

1.81 ± 7.9 37.2 ± 21.8 12.1 ± 7.5 9.9 ± 3.8 8.9 ± 2.8 13.9 ± 4.1 9.8 ± 3.2 4.9 ± 2.3 8.8 ± 3.4

Mean Dimension (mm) Circular 1.15 ± 0.99 ± 0.33 ± 0.69 ± 0.64 ± 0.69 ± 0.64 ± 0.52 ± 0.62 ±

0.47 0.84 0.23 0.29 0.27 0.26 0.26 0.20 0.35

Elliptical 1.07 ± 0.39 1.45 ± 1.19 1.02 ± 1.05 0.94 ± 0.44 0.81 ± 0.33 1.04 ± 0.37 0.98 ± 0.39 0.75 ± 0.27 1.26 ± 0.79

Note: Measurements were conducted on photographs of horizontal thin sections. The average bubble density is presented as area per m3 x 10−4 [Bjerkelund et al., 1985].

less than 0.2 mm. The link between the structural and optical properties of sea ice was found to be closely tied to the total cross-sectional area of the inclusions. While mean value of the major axis lengths of bubbles in FY sea ice is typically on the order of tenth of millimeters, bubbles in MY ice are measured in millimeters. A few studies on the geometrical characteristics of air bubbles in MYI were conducted in the 1990s. The first results on geometric characteristics or air bubbles in second-year sea ice appeared in Bjerkelund et al. [1985] from a study of ice in the Mould Bay, Canadian western Arctic in 1983 (see section 6.2). The authors delineated air bubbles manually in photographs of horizontal thin sections and performed the measurements also manually. They found that the bubbles were predominantly circular and located at the subgrain boundaries. Eighty-two percent of the circular bubbles had diameters of less than 1 mm. The average diameter of the circular bubbles was 0.69 ± 0.2 mm, and the average long axis of the elliptical bubbles was 1.04 ± 0.2 mm. Table 5.2 includes characteristics of bubble geometry, dimensions, and density at different depths. Shokr and Sinha [1994] identified air bubbles in multiyear hummock and MP ice from digital images of vertical thin sections. The ice was obtained from Parry Sound in the Central Arctic in May 1992. A digital image analysis technique was developed to identify bubbles and to measure the geometrical parameters of each bubble. A dataset based on observations from five hummock and five melt pond cores was generated. The smallest bubble that can be identified has a diameter of 0.1 mm. From a total of 520 bubbles in hummock ice and 1496 bubbles in MP ice, bubble statistics were derived in terms three parameters: (1) of the circle equivalent diameter “D” (equals 4A π ,where A is the area of the bubble), (2) the distance to the nearest bubble “S,” and (3) the compactness factor “C,” which is a measure of bubble shape defined as: C = A 0 0795 p2

(5.10)

Table 5.3 Characteristic dimension of air bubbles in MYI (mm). Ice type Hummock ice

Melt Pond ice

Parameter

Mean

Std. Dev.

D C S D C S

2.58 0.62 4.03 2.16 0.68 –

1.96 0.19 1.73 0.83 0.12 –

Note: C is the ratio defined in equation (5.10), D is the average diameter, and s is the standard deviation.

where, p is the bubble perimeter. The ratio A/p2 has its maximum value (1/4π = 0.0795) for a circle. Hence, C is a measure of the deviation of a bubble cross section shape from a perfect circle. As C approaches 1, bubbles become more circular, and as it approaches 0, they become very narrow relative to their length (i.e. take a needle-shape). The statistics are included in Table 5.3. Note that, while the distribution of bubble dimensions in hummock ice is nearly normal, the distribution in MP ice is binomial, as depicted in Figure 5.38. The average diameter of air bubbles in hummock ice (2.58 mm) is slightly larger than that in MP ice (2.16 mm). The difference is attributed to the presence of large bubbles in the snow–ice layer that commonly exist at the top of the MP ice. The average diameter of bubbles in this snow– ice layer in MP ice is found to be around 4.8 mm. The shape factor of bubbles in both hummock and MP ice is equal (around 0.65), indicating that the overall bubble shape is similar in both types and not far from being spherical. Nevertheless, bubbles of snow ice origin in MP ice were found to be more spherical, with an average C = 0.83. The average distance to the nearest bubble in MP ice is 4.03 mm. This distance was not calculated for bubbles in hummock ice since most bubbles were highly convoluted with a significant overlap in their projection. The probability of bubble existence at different depths

244

SEA ICE 0 20

Depth (mm)

40 60 80 100 120 140 0.00

0.02

0.04

0.06

0.08

0.10

0.12

1 cm

Probability of bubble occurrence

Mean area of bubbles in the horizontal plane = 4 mm2

0.30 0.20 0.10 0.00 0

4

8 12 16 20 24 28 32 Area of bubble (mm2)

0.40 0.30

Figure 5.59 Horizontal thin section at 100 m depth from SYI in the Allen Bay, Canadian Arctic, obtained in April 1997, used to quantify the geometry of air bubbles [Johnston, 1998 / Cambridge University Press].

Mean major axis length of bubbles in the horizontal plane = 3 mm

0.20 0.10 0.00

0 2 3 5 7 8 10 12 13 15 17 Major axis of bubble (mm)

Probability density

Probability density

0.40

Probability density

Figure 5.58 Probability of occurrence of air bubbles in MYI at discrete depths [Shokr and Sinha, 1994].

0.40 0.30

Mean minor axis length of bubbles in the horizontal plane = 2 mm

0.20 0.10 0.00 0 1 3 4 5 7 8 9 10 12 13 Minor axis of bubble (mm)

Figure 5.60 Statistics of geometry of air bubbles in SYI from Allen Bay (1997), measured in a horizontal plane at 0.10 m depth [Johnston, 1998].

is shown in Figure 5.58. Bubble density number decreases with depth. The data in the figure, however, cannot be generalized because the depth of the bubble-rich layer in hummock ice varies considerably between a few centimeters to a few tens of centimeters. In a study by Johnston [1998], air bubbles in SYI were characterized in terms of their area, major and minor axes. Data were obtained from digitized images of thin sections photographed between cross polarizer with added scattered light. An example is shown in Figure 5.59. Here the bubbles appear as dark objects against the background of colored grain. The circularity of the bubbles is visible. It can be associated with the orthotropic or transversally isotropic structure of the S2 type columnar grains found in the small bay. This ice was not affected by any preferred direction of the water current during the initial growth. The probability distributions of the area of bubbles as well as major and minor axes in the horizontal plane are shown in Figure 5.60. The mean major and minor axis length is 3 mm and 2 mm, respectively. These values agree

with the data in Table 5.2 although the probability distributions show a long tail that extends to 10 mm for the major axis. 5.7. REFERENCES Ackley, S.F. et al. (1978) Primary productivity in sea ice of the Weddell region, US Army Cold Regions Research and Engineering Laboratory, 17p. United States. CRREL Report No: CR 78-19. Ashton, G.D. (1986) River and lake ice engineering, Littleton, Colorado, USA: Water Resources Publications, 485 p. Barrette, P.D. and Sinha, N.K. (1996) Crystallographic characterization of a core from the Ward Hunt ice shelf, Canada, Proceedings of International Symposium Snow and Related Manifestations, Manali, India, September 1994, pp. 114–124. Bernal, J.D. and Flower, R.H. (1933) A theory of water and ionic solution with particular reference to hydrogen and hydroxyl ions, The Journal of Chemical Physics, 1(8), pp. 515–525. Bjerkelund, C.A. et al. (1985) The texture and fabric of the second year sea ice cover at Mould Bay, Prince Patrick Island,

POLYCRYSTALLINE ICE STRUCTURE 245 NWT, April 1983, Proceedings of International Geoscience and Remote Sensing Symposium, Vol. 1, pp. 426–431. Bolling, G.F. and Tiller, W.A. (1960) Growth from the melt: Part II cellular interface morphology, Journal of Applied Physics, 31, pp. 2040–2045. Cadek, J. (1988) Creep of metallic materials, Amsterdam: Elsevier, ISBN-10: 0-444-98916-1. Campbell, F.C. (2008) Elements of metallurgy and engineering alloys, Materials Park, OH, USA: ASM International, ISBN-13: 978-0-87179-867-0. Clarke, D.B. and Ackley, S.F. (1982) Physical, chemical, and biological properties of winter sea ice in the Weddell Sea, Antarctic Journal of the US, 17(5), pp. 107–109. Colbeck, S.C. (1987) Theory of particle coarsening with a lognormal distribution, Acta Metallurgica, 35, pp. 1583–1588. Cole, D.M. and Shapiro, L.H. (1998) Observations of brine drainage networks and microstructure of first-year sea ice, Journal of Geophysical Research, 103, pp. 21,739–21,750. Cota, G.F. et al. (1991a) Ecology of bottom ice algae: 1. Environmental controls and variability, Journal of Marine Systems, 2(3–4), pp. 257–277. Cota, G.F. et al. (1991b) Ecology of bottom ice algae: 2. Dynamics, distributions, and productivity, Journal of Marine Systems, 2(3–4), pp. 279–295. Cota, G.F. et al. (1991c) Ecology of bottom ice algae: 3. Comparative physiology, Journal of Marine Systems, 2(3–4), pp. 297–315. Davis, N.R. and Wadhams, P. (1995) A statistical analysis of Arctic pressure ridge morphology, Journal of Geophysical Research, 100, pp. 10,915–10,926. Dorsey, N.E. (1940) Properties of ordinary water-substance in all its phases: water-vapor, water and all the ices, American Chemical Society, Monograph Series, No. 81, New York: Reinhold Publishing Corporation. Dorsey, N.E. (1968) Properties of ordinary water-substance, American Chemical Society Monograph Series, New York: Hafner Publishing Company. Eicken, H. (1993) Automated image analysis of ice thin sections–Instrumentation, methods and extraction of stereological and textural parameters, Journal of Glaciology, 39, pp. 341– 352. Eicken, H. and Lange, M.A. (1991) Image analysis of sea ice thin sections: A step towards automated texture classification, Annals of Glaciology, 15, pp. 204–209. Eide L.I. and Martin, S. (1975) The formation of brine drainage features in young sea ice, Journal of Glaciology, 14(70), pp. 137–154. Frederking, R. and Sinha, N.K. (1977) Ice action on wharf at Strathcona Sound, Proceedings of 4th International Conference on Port and Ocean Engineering under Arctic Conditions (POAC 77), Memorial University of Newfoundland, St. John’s, NFL, pp. 707–717. Garofalo, F. (1965) Fundamentals of creep and creep-rupture in metals, New York: The McMillan. Gittus, J. (1975) Creep, viscoelasticity and creep fracture in solids, New York-Toronto: John Wiley & Sons. Gold, L.W. (1960) The cracking activities in ice during creep, Canadian Journal of Physics, 38(9), pp. 1137–1148.

Gosselin M. et al. (1997) New measurements of phytoplankton and ice algal production in the Arctic Ocean, Deep-Sea Research, 44, pp.1623–1644. Gow, A.J. and Tucker III, W.B. (1990) Sea ice in the polar regions, In: Smith, W.O., ed., Polar Oceanography, Part A: Physical Science, San Diego, California: Academic Press, pp. 27–122. Gow, A.J. and Weeks, W.F. (1977) The internal structure of fast ice near Narwhal Island, Beaufort Sea, Alaska, Research Report 77-29, US Army, Cold Regions Research and Engineering Laboratory, Hanover, NH, 8 pp. Gow, A.J. et al. (1982) Physical and structural characteristics of Antarctic sea ice, Annals of Glaciology, 3, pp. 113–117. Grenfell, T.C. (1983) A theoretical model of the optical properties of sea ice in the visible and near infrared, Journal of Geophysical Research, 88(C14), pp. 9723–9735. Hillig, W.B. (1958) The kinetics of freezing of ice in the direction perpendicular to the basal plane, growth and perfections of crystals, New York: John Wiley & Sons, p. 350. Hobbs, P.V. (1974) Ice Physics, Oxford: Clarendon Press. Hodgson, T.V. (1907) On collecting in Antarctic seas, British National Antarctic Expedition, 1901–1904, In: Zoology and Botany, Vol. 3, London, UK: Trustees of the British Museum, pp. 1–10. Hoppmann, M. et al. (2020) Platelet ice, the Southern Ocean’s hidden ice: A review, Annals of Glaciology, 61(83), pp. 341–368. Jeffries, M.O. and Weeks, W.F. (1992) Structural characteristics and development of sea ice in the western Ross Sea, Antarctic Science, 5(1), pp. 63–75. Jefferies, M.O. et al. (1993) Structural characteristics of congelation and platelet ice and their role in the development of Antarctic land-fast sea ice, Journal of Glaciology, 39, pp. 223–238. Johnston, M. (1998) Influence of ice microstructure on microwave scattering properties of sea ice, PhD Dissertation, Department of Civil Engineering, Laval University, Quebec, Canada, Johnston, M. and Sinha, N.K. (1995) A method for determining the macrostructural statistics of the scattering layer of first-year sea ice from Frederick Hyde Fjord, Greenland, Proceedings of 14th International Conference on Offshore Mechanics and Arctic Engineering (OMAE), Copenhagen, Denmark, Vol. 4, pp. 203–210. Katlein, C. et al. (2020) Platelet ice under Arctic pack ice in winter, Geophysical Research Letters, 47(16), e2020GL088898. Available from: http://doi.org/10.1029/2020GL088898. Ketcham, R.A. and Carlson, W.D. (2001) Acquisition, optimization and interpretation of X-ray computed tomographic imagery: Applications to the geosciences, Computer and Geosciences, 27(4), pp. 381–400. Ketcham, W.M. and Hobbs, P.V. (1967) The preferred orientation in the growth of ice from the melt, Journal of Crystal Growth, 1, p. 263–270. Krembs, C., Eicken, H. and Deming, J.W. (2011) Exopolymer alteration of physical properties of sea ice and implications for ice habitability and biogeochemistry in a warmer Arctic, Proceedings of the National Academy of Sciences (PNAS), 108(9), pp. 3653–3658.

246

SEA ICE

Langhorne, P.J. et al. (2015) Observed platelet ice distributions in Antarctic sea ice: An index for ocean-ice shelf heat flux, Geophysical Research Letters, 42, pp. 5442–5451. Available from: doi:10.1002/2015GL064508. Langway, C.C. Jr. (1958) Ice fabrics and the universal stage, US Army SIPRE Technical Report 62, p.16. Lawson, D.E. and Brockett, B.E. (1990) In-situ sampling and characterization of frazil ice deposits, Cold Region Science and Technology, 17, pp. 193–205. Leib-Lappen, R.M., Golden, E.J. and Obbard, R.W. (2017) Metrics for interpreting the microstructure of sea ice using X-ray micro-computed tomography, Cold Region Science and Technology, 138, pp. 24–35. Light, B., Maykut, G.A. and Grenfell, T.C. (2003) Effect of temperatures on the microstructure of first-year Arctic sea ice, Journal of Geophysical Research, 108(C2), p. 3051. Available from: doi: 10.1029/2001JC000887. Lofgren, G. and Weeks, W.F. (1969) Effect of growth parameters on substructure spacing in NaCl ice crystals, Journal of Glaciology, 8, pp. 153–164. Martin, S. (1970) A hydrodynamic curiosity: The salt oscillator, Geophysical and Astrophysical Fluid Dynamics, 1(1), pp. 143–160. Martinez, J.D. (1958) Photometer method for studying quartz grain orientation, Bulletin of the American Association of Petroleum Geologists, 42(3), pp. 588–605. Maus, S. et al. (2009) Synchrotron-based X-ray tomography: Insights into sea ice microstructure, In: Lepparanta, M., ed. Report Series in Geophysics 61, University of Helsinki, pp. 28–45. Michel, B. and Ramseier, R.O. (1971) Classification of river and lake ice, Canadian Geotechnical Journal, 8, pp. 36–45. Nabarro, F.R.N. and de Villiers, H.L. (1995) The Physics of creep, London: Taylor & Francis Ltd., p. 15 (Chapter 2). Nada, H. and Furukawa, Y. (2005) Anisotropy in growth kinetics at interfaces between proton-disordered hexagonal ice and water: A molecular dynamics study using the six-site model of H2O, Journal of Crystal Growth, 283, pp. 242–256. Nakawo, M. and Sinha, N.K. (1981) Growth rate and salinity profile of first-year sea ice in the High Arctic, Journal of Glaciology, 27, pp.315–330. Nakawo, M. and Sinha, N.K. (1984) A note on the brine layer spacing of first-year ice, Atmosphere-Ocean, 22(2), pp. 193–206. Perey, F.G.J. and Pounder, E.R. (1958) Crystal orientation in ice sheets, Canadian Journal of Physics, 36(4), pp. 494–502. Perovich, D.K. and Gow, A.J. (1991) A statistical description of the microstructure of young sea ice, Journal of Geophysical Research, 96(C9), pp. 16,943–16,953. Perovich, D.K. and Gow, A.J. (1996) A quantitative description of sea ice inclusions, Journal of Geophysical Research, 101(C8), pp. 18,327–18,343. Peyton, H.R. (1966) Sea ice strength, Report UAG R-182, Geophysical Institute, University of Alaska, Fairbanks, Alaska. Pope, S., Copland, L. and Mueller, D. (2012) Loss of multiyear landfast sea ice from Yelverton Bay, Ellesmere Island, Nunavut, Canada, Arctic, Antarctic and Alpine Research, 44(2), pp. 210–222.

Pounder, E.R. (1965) The Physics of ice, Oxford: Pergamon Press. Pringle, D.J. et al. (2009) Pore space percolation in sea ice single crystals, Journal of Geophysical Research, 114(C12), C12017. Available from: doi:10.1029/2008JC005145. Ramseier, R.O. (1968) Origin of preferred orientation in columnar ice, Journal of Crystal Growth, 3(4), pp. 621–624. Ramseier, R.O. (1976) Growth and mechanical properties of river and lake ice, PhD Thesis, Laval University, Quebec, Canada. Reed-Hill, R.E. and Abbaschian, R. (1992) Physical metallurgy principles, 3rd ed., Boston, MA, USA: PWS-Kent Publishing Company. Rogers, A.F. (1937) Introduction to the study of minerals, New York and London: McGraw-Hill Book Company, Inc., Part 1, 260 p. Rohatgi, P.K. and Adams, Jr., C.M. (1967) Freezing rate distributions during unidirectional solidification of solutions, Transactions of Metallurgical Society of the American Institute of Mining and Metallurgical Engineers, 239, pp. 850–857. Shaw, R.A., Durant, A.J. and Mi, Y. (2005) Heterogeneous surface crystallization observed in undercooled water, The Journal of Physical Chemistry, 109, pp. 9865–9868. Shokr, M. and Sinha, N.K. (1994) Arctic sea ice microstructure observations relevant to microwave scattering, Arctic, 47(3), pp. 265–279. Sims, C.T., Stoloff, N.S. and Hagel, W.C. (1987) Superalloys II, New York: John Wiley & Sons, Inc. Sinha, N.K. (1977) Technique for studying structure of sea ice, Journal of Glaciology, 18(79), pp. 315–323. Sinha, N.K. (1978) Rheology of columnar-grained ice, Experimental Mechanics, 18(12), December, pp. 464–470. Sinha, N.K. (1984) Uniaxial compressive strength of first-year and multi-year sea ice, Canadian Journal of Civil Engineering, 11(1), pp. 82–91. Sinha, N.K. (1986) Young Arctic frazil sea ice: field and laboratory: Field and laboratory strength tests, Journal of Material Sciences, 21, pp. 1533–1546. Sinha, N.K. (1989) Elasticity of natural types of polycrystalline ice, Cold Region Science and Technology, 17, pp. 127–135. Sinha, N.K. (1991) Microstructure and mechanical behavior of ice, In: Dodhi, D.S., ed. Proceedings of 6th International Speciality Conference, Cold Region Engineering, West Lebanon, NH, USA, American Society of Civil Engineers (ASCE), New York, pp. 519–530. Sinha, N.K. and Zhan, C. (1996) Primary dendritic spacing in land-fast polar sea ice, Journal of Materials Science Letters, 15, pp. 2118–2121. Sinha, N.K. et al. (1996) Floating ice thickness for aircraft operations, Report published by Civil Engineering Services, Engineering and Maintenance, Safety and Technical Services, Transport Canada. Tabata, T. and Ono, N. (1962) On the crystallographic study of several kinds of ice, Low Temperature Science Series, 20, pp. 199–214 (text in Japanese). Tsang, L. and Kong, J.A. (1981) Scattering of electromagnetic waves from random media with strong permittivity fluctuations, Radio Science, 16(3), pp. 303–320.

POLYCRYSTALLINE ICE STRUCTURE 247 Weeks, W.F. (2010) On sea ice, Fairbanks, Alaska, USA: University of Alaska Press, ISBN-13: 978-1-60223-079-8. Weeks, W.F. and Ackley, S.F. (1982) The growth, structure, and properties of sea ice, Cold Regions Research and Engineering Laboratory (CRREL), Monograph 82–1, Hanover, NH, USA. Weeks, W.F. and Gow, A.J. (1978) Preferred crystal orientations in the fast ice along the margins of the Arctic Ocean, Journal of Geophysical Research, 83, pp. 5105–5121. Weeks, W.F. and Gow, A.J. (1980) Crystal alignments in the fast ice of Arctic Alaska, Journal of Geophysical Research, 84(C10), pp. 1137–1146.

Weeks, W.F. and Hamilton, W.L. (1962) Petrographic characteristics of young sea ice, Point Barrow, Alaska, Research Report 101, US Army, Cold Regions Research and Engineering Laboratory, Hanover, NH, 11 pp. Whalley, E. (1969) Structural problems of ice, In: Physics of ice, pp. 19–43. Winebrenner, D.P. et al. (1989) Sea ice characterization measurements needed for testing of microwave remote sensing models, IEEE Journal of Oceanic Engineering, 14, pp. 149–158.

6 Major Field Expeditions to Study Sea Ice

6.1

The Arctic Ice Dynamic Joint Experiment (AIDJEX).......... 250

6.4

Labrador Ice Margin Experiment (LIMEX) ........................ 266

6.2

Mould Bay Experiments 1981–1984: Stories that Were Never Told ........................................................................... 252 6.2.1 Site, Resources, and Logistics.................................... 252 6.2.2 Sea Ice Conditions ..................................................... 254 6.2.3 Aging of Sea Ice: from FYI to MYI.......................... 259 6.2.4 Interface Between Old and New Ice in Second-Year Ice Profile .................................................................. 260

6.5

Sea Ice Monitoring and Modeling Site (SIMMS) Program .. 268

6.6 6.7

The Surface Heat Budget of Arctic Ocean (SHEBA) ........... 270 The Norwegian Young Sea Ice Experiment (N-ICE) ........... 272

6.8

Marginal Ice Zone (MIZ) Experiments ................................ 274

6.9 6.10

Ice Exercise by Us Navy....................................................... 277 The Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC)................................................ 278

6.11

References............................................................................. 280

6.3

High Arctic Experience with Ice of Land Origin .................. 262 6.3.1 Ward Hunt Ice Shelf and Hobson’s Choice Ice Island Experiment ................................................ 262 6.3.2 Multi-Year Rubble Field Around the Ice Island ....... 265

The initial interest in the frozen seas of the Arctic region, which motivated the early explorers in the seventeenth century onward, was not studying sea ice but rather the exploration of unknown lands and people. Ice was just a major hazardous element along the way. Apparently, exploration was also implied in an economic incentive. That interest mounted, yet in a different direction, during the post World War II era; namely in a security-related motivation in order to face potential threat by a possible attack from long-range Soviet bombers flying across the Arctic. At that time, the construction of the Distant Early Warning (DEW), jointly by the USA and Canada, proceeded along with Arctic submarine operations as a preemptive measure against any possible military showdown with the Soviet Union. Heavy construction equipment had to be shipped to the Arctic to establish the DEW system and that set in motion an influx of ships, escorted by the US icebreakers, in the summer of 1955. Obviously, that would not be possible without gathering information about sea ice. The ice information was oriented to support safe marine transportation through ice-rich routes. Apparently, the dynamic ice loads on marine vehicles and structures were the central issue and therefore the

mechanical properties of sea ice became a top research priority. That was the first aspect of scientific interest in sea ice, which was followed by a growing interest in other scientific aspects that revolve around ice climatology. This includes thermal, optical, electrical, and radiometric properties of ice with a focus on their response to climate change (see Chapter 13 on the latter aspect). Science aspects of sea ice have gained momentum when Arctic-wide ice monitoring became available using satellite sensors, especially the passive microwave in the 1970s (sections 8.1 and 8.3). The potential of sea ice parameter retrieval from satellite observations compelled funding agencies, users, and researchers to allocate resources to explore links between sea ice properties and the radiometric signal they engender. To achieve this goal, studies of physical properties of sea ice, along with modeling the satellite observations from snow-covered sea ice, had to be accelerated. This was the derive for many field and laboratory experiments as well as microwave forward modeling that flourished in the 1970s and 1980s. As is often the case with science, questions addressed have received partial answers and raised more questions. For sea ice, this has triggered more field studies, set new

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 249

250

SEA ICE

objectives and advanced measuring tools. For example, an early scientific field expedition in late 1970s (see next section) was set to answer a few emerging questions about how sea ice moves and changes in response to the influence of ocean and atmosphere. Resulting answers triggered later programs to address the marginal ice zones (MIZ) (see section 6.8). MIZ have been observed to increase in size and location as a result of the recent climate change in the Arctic. Similarly, satellite-based observations that link coincident field measurements of snow-covered sea ice to changes of its composition have contributed to the development of many field campaigns with advanced instruments to further understand key physical processes. The cumulative knowledge about sea ice physics and its impacts on improving parameter retrieval from remote sensing data and enhancing the modeling predictions will continue to stimulate more field expeditions to improve our understanding of key sea ice processes. This is particularly true as climatic changes in polar ice covers continue. There has been an apparent trend that distinguishes recent polar scientific campaigns from the campaigns conducted a few decades ago. Recent campaigns have become bigger, more interdisciplinary, with better instrumentations and defined scientific foci. This has made them more economically viable and scientifically comprehensive. The huge cost of the expedition warrants sharing the logistics between several scientific missions. Moreover, the highly interactive processes between atmosphere, ecosphere, ocean, and ice warrant such an interdisciplinary endeavor. This chapter offers a brief account of ten Arctic sea ice campaigns in a chronological order. Most of them featured international and interdisciplinary aspects. Unpublished results from experiments attended by the authors of this book are presented.

AIDJEX was the first major science field experiment conducted in the Arctic [Campbell et al., 1978, Untersteiner, 1980]. It was initiated by a consortium of a few organizations from the USA and one from Canada. The American organizations included the Division of Marine Resources, University of Washington, Seattle, the National Science Foundation (NSF), the USA Office of Naval Research (ONR), the National Aeronautics and Space Administration (NASA), the US Geological Survey (USGS), and the National Oceanic and Atmospheric Administration (NOAA). The Canadian side was represented by the Polar Continental Shelf Project (PCSP). More on this organization and its director G.D. Hobson, at the time, is revealed in section 6.2. AIDJEX was the culmination of 5 years of preparation to develop the concept, the plan, the rules of the participating agencies, and technical developments. A pilot experiment was conducted in 1972 but the actual field work was performed over 14 months on the ice north of Point Barrow, Alaska from March 1975 to April 1976. During this field work, coordinated atmospheric and oceanic measurements were collected from 4 camps on sea ice and 20 floating buoys deployed in pairs on a ring of 300 km radius around the central manned camp. Ice draft data from the USS Gurnard (SSN-662) submarine were used, from which the ice thickness was approximately estimated. Figure 6.1 shows the Jumpsuit field camp during the 1972 experiment. Satellite data available at that early time (Landsat and passive microwave) were used along with airborne imaging radar data to estimate the motion of the ice cover.

6.1. THE ARCTIC ICE DYNAMIC JOINT EXPERIMENT (AIDJEX) As the name of the experiment indicates, this experiment was about sea ice dynamics. It aimed at understanding the large-scale response of sea ice to the four driving external forces of wind, ocean currents, Coriolis force, and gradient of sea surface level (triggered by gravity). Therefore, ice physics was not at the center of the sought information (it was considered only, to a limited extent, in terms of the mechanical properties of ice). Remote sensing, though used also in a limited way, was at its early stage of development. With that in mind, the experiment had little to add to the subject of a book on ice physics and remote sensing. However, a brief synopsis is presented here to put this major experiment in the broader context of the development of Arctic sea ice knowledge and, in particular, scientific expeditions over the past 5 decades.

Figure 6.1 Aerial view showing the Jumpsuit camp of AIDJEX project in the Beaufort Sea, in 1972, on what appears to be a MYI floe (note the numerous protruding hummocks) (NSIDC courtesy Pat Martin / Flickr / CC BY 2.0).

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

As the focus of AIDJEX was on ice dynamics, the planned activities included detection of ice motion, determining the mechanisms of ice deformation, and relating the mechanisms to the external stress fields. With this information, a model of ice dynamics was pursued using a few derived principles and observable parameters. The long-term vision was to incorporate such a model into a global climate model. In the AIDJEX project the pack ice was viewed as a kinematic system composed of floes of many sizes and shapes separated by leads. Hence, unlike other field experiments introduced in the next sections, physical and radiometric properties of snowcovered sea ice were not parts of the results The scientific plan of AIDJEX addressed four questions. 1. How is large-scale ice deformation related to external stress field? 2. How does ice topography interact with large-scale stress and strain fields? 3. What are the mechanisms of ice deformation? 4. How do ice deformation and morphology affect the heat balance? Answers to these questions were assessed in Untersteiner [1980] by evaluating the accomplished work, especially about the suitability and scale of observations, the accuracy of estimating the external stresses, and the advancement of understanding the sea ice mechanics and heat balance. The assessment asserted that AIDJEX produced a quantum jump in knowledge about ice dynamics and many other aspects of the Arctic environment. The issue of scale of observation was unique to the AIDJEX project because of its set objective of ice dynamics, which takes place at different scales from a few to thousands of kilometers [McNutt and Overland, 2003]. While the scale of the Beaufort Sea Gyre (atmospheric forcing) is known to be around 1000 km, the scale of oceanic eddies was estimated to be of several tens of kilometers. Hence, arrays of measuring instruments were placed at 100 km for the camp and 1000 km for the buoys [Untersteiner, 1980]. To accommodate possible fluctuations of ice floe velocity, time scale of one day for observing ice motion was selected. Aspects of ice deformation are manifested in variable size of floes, ridging, and rubble field formation. These modulate the coupling between the ice and atmospheric boundary layer. Ice deformation was measured at an array of points within the boundaries of the ice camp from which internal stresses could be deduced. Calculations of strain field were performed using wind and ocean current forcing but did not compare well with in-situ observations. Several sources of error were identified in Pritchard, Coon, McPhee [1977], while agreement was found between observed and calculated ice velocity using different boundary layer parameters.

251

The ice dynamics model from the AIDJEX project included two sets of processes. The first was about ice motion, where parameters representing external forces and internal stresses (e.g., ice strength and drag coefficients) were obtained and linked to the geometric properties of ice. The second was about the geometric properties of ice, namely evolution of ice thickness, surface roughness, pressure ridging, and floe size. According to Untersteiner [2007], ice dynamics models at that time contained only two common vestiges of AIDJEX model development: plastic failure and a thickness distribution. The author stated that the knowledge acquired from AIDJEX about air and water drag coefficients and their seasonal variations were not incorporated into a high-end ice model. Another revisit to the assumptions used in modeling sea ice dynamics based on AIDJEX is published in Coon et al. [2007], where the authors found that the assumptions were inadequate; in particular, not enough leads are present to make the ice isotropic. With regard to using remote sensing (an aspect more relevant to the subject of this book), AIDJEX used sequential Landsat satellite images to measure displacement and deformation of sea ice surface, including lead and ridge development at the fine resolution of the sensor (a few tens of meters). That was a pioneering work. Nevertheless, only a few periods during the experiment were available to estimate the ice displacement from Landsat images because of cloud and darkness. Passive microwave images were used to estimate ice age, hence a proxy ice thickness. Calculations of the ice thickness distribution from drifting station data have been useful in heat balance studies. Progress has been made toward determining the role of open water and thin ice in the large-scale distribution of heat in the boundary layer of ocean and atmosphere. Laser ranging was used for surface topography, and airborne imaging radar for ice morphology and roughness. AIDJEX project was accomplished before the era of the space-borne synthetic aperture radar, which was used routinely in the 1990s to estimate sea ice kinematics (see section 11.7.1). Some of the highlights from the review in Untersteiner [1980] are reiterated here. In the absence of internal stress during summer, the ice drift was merely wind-driven, unlike in winter when it becomes manifested in shear discontinuity that extends for several hundreds of kilometers. Also, ice motion was found be proportional to the geostrophic wind, which means that ice thickness and ice internal stresses have minor effects. While progress has been made in defining the role of thin ice and open water on the large-scale distribution of heat within the atmospheric and oceanic boundary layers, the role of thick ice was not determined. Ice thickness information available during AIDJEX was insufficient so that in most cases the initial state of this variable had to be assumed.

252

SEA ICE

Forty digital bulletins of AIDJEX results and analyses are made available by the Polar Science Center at the University of Washington. They can be accessed through the link (psc.apl.washington.edu/nonwp_projects/aidjex/ toc.html). Final disposition of the data is given in Bulletin No. 40. Major results from the AIDJEX project are summarized in Pritchard [1980], including a review by the project manager N. Untersteiner. A retrospective of AIDJEX achievements is presented in Untersteiner et al. [2007]. For ice researchers who started their career in the 1990s, it is worth noting that AIDJEX marked the milestone of a few events in the context of sea ice field work. In March 1972, a fully loaded Lockheed C-130 Hercules (a fourengine turboprop military transport aircraft weighted 34,400 kg) made the first landing of that type of aircraft in an unprepared sea ice runway of refrozen lead. These landings became routine later. Also, the first air traffic services (ATS) 1975 satellite telephones were installed in April 1975 between the main camp in the field (76 N) and the AIDJEX office in Seattle. The routine daily contacts for one hour proved to be most beneficial. That is how AIDJEX made history. Nowadays, communication in the High Arctic region with the rest of the world during field work is achieved via internet, satellites, and telephones, though in a limited way because there is only one satellite communication system that can function north of 75 latitude.

6.2. MOULD BAY EXPERIMENTS 1981–1984: STORIES THAT WERE NEVER TOLD On 13 September 1989, the Government of Canada announced its initiative to build and operate the first Earth observation Canadian satellite, RADARSAT-1. The satellite carried a single payload sensor, a synthetic aperture radar (SAR). The design of the sensor was based on more than ten years of mission study and preparation in Canada. Monitoring sea ice was a prime objective of the mission at that time. Within this context, the series of Mould Bay experiments was launched to be an early endeavor to confirm requirements for, and specifications of, the sensor. The field experiments were conducted in the vicinities of the permanent weather station of Mould Bay (76 14 N, 119 20 W) in the Canadian western Arctic. The study period extended over four freezing/melting seasons during 1981 to 1985. Many researchers from Canada and the USA participated. One of the most important observations made during this series of experiments was the insitu observations ever carried out on the transition of FYI to MYI, and the growth of new ice under the existing MYI. Moreover, this series of experiments demonstrated

how opportunities, that might emerge during the course of the field work, could be captured to perform work which was not part of the original plan. Only limited information from these experiments has been published so far. This way, the contents of this section are about the untold story of Mould Bay experiments. The following discussions diverted sometimes from describing the recorded observations to touch upon a few physical aspects of sea ice, which can fit better in Chapter 4 or 5. This is particularly noticeable in the microstructural information of SYI presented in sections 6.3.2–6.3.4. However, the authors found it more appropriate to include it here to complement the actual source of observation. 6.2.1. Site, Resources, and Logistics Mould Bay, a 30-km long and 7–9 km wide, deep water (about 200 m) inlet is located in Prince Patrick Island of the Canadian Archipelago in Northwest Territories, Canada in the Western Arctic (Figure 6.2). The Mould Bay field experiment started in October 1981 when the ice was young. The first in the series of the experiments was during the entire month of October 1981 [Sinha, 1984, Hollinger et al., 1984, Kim et al., 1985]; the second and probably the most intensive was between 3 June and 21 July 1982 [Digby, 1984, Holt and Digby,1985, Grenfell and Lohanick, 1985]; the third was between 5 April and 10 May 1983, when a large section of the ice cover was SYI [Bjerkelund et al., 1985, Sinha, 1985]. One of the goals of the Mould Bay program was to quantify the physical characteristics of ice in relation to the microwave backscatter and emission during the annual growth and major periods of transition as the ice cover aged. The Atmospheric Environment Services (AES) of Environment Canada (renamed later to Meteorological Service of Canada—MSC) operated a permanent weather station near Mould Bay (Figure 6.2), which used to be manned year-round from 1948 to 1997. Adjacent to this weather station, an “unofficial” portable field laboratory of the National Research Council (NRC) of Canada was constructed in 1981, by staff from NRC with the help of the staff of the weather station. This field laboratory consisted of three portable trailers or mobile homes founded on huge tires. One of the trailers was then covered with sprayed foam insulation in order to use it as a cold laboratory (Figure 6.3a). A portable refrigeration/heating system was installed to control the inside temperature. Actually, these trailers were part of an old drilling camp operated by a European oil exploration company in the early 1960s. They were abandoned by this company when no oil or gas was found anywhere in the Prince Patrick Island. It proved to be very expensive for the company to move the campsite and the debris out of the island and the oil exploration company, obviously, had no

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE (a)

253

(b) –124°

–120°

–116°

–118°

Cape Leopold MClintock Satellite Bay

Arctic

N “HOBSON’S CHOICE” ICE ISLAND

–122°

Prince EUREKA

77°

Ocean

77°

Patrick

MOULD BAY

Island RESOLUTE

Lands End Mould Bay

POND INLET

Cape Manning

YELLOW KNIFE

McClure –124°

Melville

Crozier

76°

–122°

Kellett Channel Eglinton

Island

Strait

Island –116°

–118°

–120°

76°

km 0

50

100

Figure 6.2 (a) Map of Canada showing three High-Arctic long-term sea ice experimental sites—Hobson’s Choice ice island (≈ 79 23.5’N, 102 20.2’W), Mould Bay (76 14’N, 119 20’W), and Pond Inlet (72 42’N, 77 57’W) and (b) details of Prince Patrick Island viewing Mould Bay where a Canadian weather station was operational from 1948 to 1997 (N.K. Sinha and M. Shokr).

Figure 6.3a National Research Council (NRC) of Canada field Figure 6.3b The landlord visiting the field laboratory in Mould laboratory on wheels at Mould Bay weather station with Ron Bay, October 1981 (photo by N.K. Sinha). Jerome (left) and Nirmal Sinha (right); note the added foam insulation (yellow) and the deep freezer (white box) in the left above the tire (photo by AES weather station, Mould Bay, 1982).

respect for the environment. In those days, there were no environmental protection laws of the land, and mankind was free to damage the terrain of this no-man’s isolated territory. Over the years, the nature and polar bears vandalized the trailers. The test site of three trailers became the “unofficial” NRC laboratory in Mould. Soon after

the establishment of this laboratory-complex, the landlord paid a visit as can be seen in Figure 6.3b. During the project’s period, AES operated the Motorola APS-94 SLAR (side-looking airborne radar—SLAR) system onboard a four-engine Electra aircraft that used to produce real-time imagery for ice reconnaissance

254

SEA ICE

Figure 6.4 AES Electra equipped with SLAR flying over Mould Bay on 3 July 1982; note the near-absence of snow on the land and the mountains across the bay (photo by N.K. Sinha).

(Figure 6.4). This was an X-band (9.1–9.4 GHz) imaging radar with HH polarization using incidence angle of 45 . The aircraft was also equipped with two Vinten 70-mm aerial cameras and few other instruments. 6.2.2. Sea Ice Conditions A synopsis of the ice in the Mould Bay is presented since it is relevant to the subsequent discussions. During the spring of 1981, the 30-km-long Mould Bay with its length oriented in the north-south direction, as shown on the map, in Figure 6.2b, was covered with ice. Some of this ice cover was expected to remain in the bay. However, during the late summer of 1981 all of the ice in the bay moved out with the exception of a few small grounded pieces of MYI near the shore in the southern end. These MYI pieces also disappeared eventually, leaving a clear bay when the next freezing season began in October 1982. Thus, the entire bay without any ice debris became an ideal “mesoscale” laboratory for producing a sea ice cover with unidirectional flow of water under the ice due to tidal actions. The water surface conditions of the bay and the weather were monitored continuously by staff of the station after the summer of 1981. Except for a major storm a few days before the date of solidification of the entire water surface of the bay, the weather was cold, but very calm with practically no wind. The surface of the ice during the early ice growth (September–October 1981) following the shoreto-shore freezing was unique. The ice cover was as smooth as the surface of small-scale laboratory-grown ice except for major differences in the conditions below the ice that could never be reproduced in a laboratory. It is appropriate to recall the fact that in laboratory experiments, it is impossible to fulfill the thermal and chemical state of

the saline water controlled by the diurnal tidal periods and the reversible direction of the water current. Consequently, it is impossible to simulate the natural mixing of water due to tidal effects in deep waters in an ice tank, exclude lateral heat flow, maintain constant water salinity (tank water salinity increases as ice thickens), and impose unidirectional solidification for examining the crystal habit especially the alignment of c axis. During the first trip in October 1981, microstructural and mechanical properties of floating young sea ice in the bay were investigated along with those of MYI from a large old floe in Crozier Channel (this channel separates Prince Patrick Island and Eglinton Island (see Figure 6.1). During the second trip later in the same freezing season (June–July 1982), the ice was fully grown and approaching the melt season. In addition to collecting microwave data for mature FYI, one of the aims of the trip was to monitor the melting conditions of the ice cover and to examine the decaying processes in landfast ice and the mobile MYI floes in Crozier Channel. During the third trip, one year later during April–May 1983) a rare and excellent opportunity came to document SYI and the newer FYI that grew under it. The growth of ice and snow cover at various stations across the 7.2-km-wide Mould Bay was monitored every week during the major growth period for FYI (more than 300 mm in thickness) during the polar nights, from 31 October 1981 to 13 June 1982. Detailed measurements of ice thickness and snow depth collected at nine measuring stations from 24 September 1981 (start of the freezing) to 13 June 1982 are presented in Figure 6.5. As can be seen, the ice cover had reached a maximum thickness of 2.25 m and snow cover depth was 0.65 m. The staff of the weather station shared their efforts and volunteered to collect the data through the extremely cold

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE Snow, aes

2.5

Ice & snow thickness, m

255

Ice, aes Ice, 1 Snow, 1 Ice, 2

2

Snow, 2

Ice

Ice, 3 Snow, 3

1.5

Ice, 4 Large crack between station 4 & 6

1

Snow, 4 Ice, 5 Snow, 5 Ice, 6 Snow, 6

Snow

0.5

Ice, 7 Snow, 7 Ice, 8 Snow, 8

0 0

50

100

150

200

250

300

Growing days: 24 Sept. 1981 to 13 June 1982

Figure 6.5 Growth in sea ice and snow cover at various sites (identified by numbers at the right column) along the 7.2 km long experimental line across Mould Bay, from 24 September 1981 to 13 June 1982, as functions of number of days after the freezing; large crack crossing the experimental line developed a few days before 29 May 1982 (dashed vertical line). Note the virtually steady growth rate (N.K. Sinha).

0 −50

Oct 2, Core 1 Oct 2, Core 2

−100 Ice depth, mm

and dark polar nights. Traveling on the ice in the dark was by no means pleasurable. By carrying out more than their routine observations at the weather station, members of the station filled the gap in the intervening periods between field trips conducted by members of the research teams. Normally, these types of details on human aspects of science and engineering are not covered in technical books. Since snow, ice, and water play the most important roles in shaping our life and the environment we live in, scientific aspects of these materials cannot be discussed without referring to the human and environmental aspects. Vertical salinity profiles of YI and thin FYI are shown in Figure 6.6. Because of the high latitude of Mould Bay, there was practically no solar radiation to complicate the growth conditions. The surface temperature probably kept in phase with the ambient air temperature. Moreover, since there was no snow deposition during the first eight days of ice growth in October 1981, the ice surface remained exposed to the atmosphere with clear sky. Due to the persistent low air temperatures, the ice surface was obviously subjected to sublimation. Due to the favorable thermal gradient, brine also stayed in the upper layer and probably migrated toward the colder top. This dual affects satisfactorily explains the relatively high surface salinities shown in Figure 6.6. While the surface salinity of the C-profile can be physically and experimentally justified, the lower portion of the profile is subjected to errors, mainly due to problems inherent in the experimental procedures followed, namely brine drainage during the removal of the

Oct 2, Core 3 Oct 2, Core 4

−150 −200 −250

Young sea ice

−300 −350

October 18,

−400 0

4

12 8 Salinity, ppt

16

20

Figure 6.6 Vertical salinity profile of 8-day old Young columnar-grained S3 type sea ice from four ice cores in Mould Bay on 2 October 1981, ice was grown at an average rate of 24.4 mm/day; data on 24-day old FYI, obtained on 18 October, is shown for comparison (N.K. Sinha).

cores from the ice sheet. Major part of the drainage near the bottom occurs as the core auger is lifted up the hole. Further drainage occurs sideways when the core is placed in a container invariably placed horizontally on the ice surface. The ambient air temperature also plays a big role. Further desalination may occur during storage, shipping, and cutting into segments; but this depends entirely on the ambient temperature. For those reasons, the non-destructive method for measuring salinity profile of YI developed by Notz, Wettlaufer,

256

SEA ICE

Worster [2005] (see section 3.3.2) by avoiding ice core extraction is considered to be a more appropriate tool for measuring the salinity of YI. As expected, the ice in Mould Bay during the experiment was found to be columnar-grained S3 type with the c axis of the grains in the horizontal plane and oriented predominantly along the north-south direction, parallel to the tidal water current. The preferred crystallographic orientation of the ice was found to start right from the top of the ice surface. For this purpose, thin sections were prepared using DMT (section 4.2.2) inside the aforementioned rudimentary field laboratory, during the cold periods at temperatures between –15 C and –25 C offered by nature. The sections were made from freshly recovered large rectangular blocks through the full depth of the ice cover obtained with a gasoline-powered chain saw. Before cutting the ice sheet, the orientation of the ice blocks was marked to show the north-south direction, i.e., the longitudinal axis of the fjord, and hence the “expected” direction of the tidal current. It is better to sample rectangular blocks because it is very difficult to mark the orientation of the ice at the top before coring, although special inks were tried with limited success. Moreover, the core may break and may not be possible to match fragmented sections. Of course, it would be difficult to take blocks deeper than about 1.5 m. In October 1982, parts of the matured FYI cover that survived through the summer melting, entered the life of its second year. With the advent of the winter in the fall of 1982, the shore leads started to freeze and “locked” the remnants of the winter of 1981–1982 ice in the bay. New ice also started to grow below this survived ice. Unlike the previous growth season, mainly due to lack of travel funds, no ice data were collected during the fall of 1982 and winter of 1983 until the early spring of 1983 (31 March to 10 May 1983), when the AES/NRC research team revisited the site long before the end of the third and last experiment of this series. A photograph of the surface of the SYI cover on 31 March 1983 is shown in Figure 6.7. It shows the expected hummocky surface features, yet not well identified as appearing in the MYI surface in Figure 2.57. Hummocks are not as well developed in SYI as they would be in MYI. In fact, this feature is used to discriminate between the two types on the ground (but hardly in radar remote sensing images though a first suggestion for this discrimination from using polarimetric SAR data is presented in Shokr et al. [2022]). The relief around the melt ponds and drainage channels must, therefore, continue to influence subsequent ablation of ice at surface levels and later stimulate the formation of snow drifts, winddriven compaction of snow particles and morphological changes in the snow mass during the growth of the winter season of 1982–83.

Figure 6.7 View of predominantly second-year cover in Mould Bay on 31 March 1983, showing surface undulations and a track made by a snow plow for establishing the experimental line across the bay (photo by N.K. Sinha, unpublished)

0.50 Snow cover

0.00 10

9

8

7

6

5

4

3

2

1

Station

Old ice of 1981–82

0.50 1.00

Old ice – new ice interface

New ice of 1982–83 winter on 5 April 1983

1.50 2.00 Ice – water interface

2.50 0

1.2

2.4

3.6

4.8

6.0

7.2

Figure 6.8 Profile of snow and ice along the experimental line with 10 stations (marked as 1–10) in Mould Bay on 5 April 1983 [Adapted from Bjerkelund et al., 1985].

Ten equally spaced sampling stations were established along the experimental line close to the path created by a snow plow for ease of traveling with snowmobiles. The snow depth varied between 0.1 to 0.6 m and, incidentally, very similar to that observed in June 1982. The ice thickness profiles across the bay along the experimental line, shown in Figure 6.8, indicate that the old ice during the past freezing season was located between stations 2 and 9 with a refrozen lead near station 7. Undoubtedly, this lead was linked to the large crack mentioned earlier. The thickness of the old (second-year) ice that had gone through two complete seasons of growth varied between a minimum of 0.17 m near station 2, and a maximum of about 0.9 m between stations 4 and 6. This was also the region of maximum snow accumulation. Consequently, the growth of the new ice in this region was

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

expected to be lowest. The thickness of the underlying new ice under the existing SYI (to be considered as FYI) was found to be lowest of 1.35 m at station 5. The shore leads consisted of top-to-bottom FYI with the maximum thickness of about 2.3 m around station 10 close to the western shore, which also coincided with near-absence of snow. The interface between the old ice and the underlying new ice is also shown in Figure 6.8. It corresponds to approximate line of abrupt changes in the salt content through the ice thickness. Therefore, this line of demarcation provides important information on the transmission characteristics of EM waves through the ice cover, especially for long radar waves. If incident microwave is transmitted through the full thickness of the virtually salt-free SYI, it will be absorbed rapidly soon after penetration through this interface. In order to examine the ice surface roughness and the vertical variation of the voids (or air inclusions), 40-mm-thick rectangular vertical sections (100 mm wide and 200 mm deep) were cut from large ice blocks using a bandsaw inside the field laboratory and kept at about −15 C, making certain not to damage the original top surfaces. The saw-cut surfaces were then sanded lightly, and after brushing off the loose ice particles, the grounded surfaces were wiped lightly using soft fabric moistened (not soaked) with chemically pure ethyl alcohol. This cleaning treatment allowed inclusions to be viewed clearly as shown in Figure 6.9. The vertical profile of the SYI in Figure 6.9a illustrates the general characteristics of entrapped air bubbles and their variations in shape and size with depth. The degree of surface roughness and the two foam-like white layers near the top can be seen clearly in Figure 6.9b. These two layers seemed to be created by the coalescence of spherical bubbles with diameters up to about 8 mm. They

(a)

257

seem to originate from superimposed snow ice but this can be checked only in photos of thin sections prepared using the DMT technique. Rows of this type of large and isolated spherical bubbles with irregular surfaces are visible down to a depth of about 40 mm. All the layers in Figure 6.9b were above the measured freeboard of around 200 mm and above the region of buoyancy forces. The central plane of the third layer features a row of large spherical bubbles with their surfaces touching each other. These voids were obviously created when the old ice melted and the meltwater percolated downward through the ice matrix. The shape of those voids changed rather abruptly to long and narrow cylindrical shape. The relatively transparent bands of ice and layers of elongated air bubbles may “hastily and erroneously” be interpreted as refrozen layers of meltwater ponds with S1 and/or S2 ice commonly observed at the surfaces of MYI floes, as well as in flat areas of shelf ice and glaciers. Forensic type of microstructural investigations were used to reveal the true nature of the ice in this region, which is important for microwave backscatter. Parallel- and cross-polarized views of a thin section at a depth of about 3 mm (marked by the top arrow in Figure 6.9) are shown in Figure 6.10. This thin section corresponds to the mid-plane of the top porous and bubbly white layer visible in Figure 6.9b. The photographs show, without any doubts, that the ice was (a)

(b)

(b) 10 mm

20 mm

20 mm

100 mm

Figure 6.9 Vertical thick section (40 x 100 x 220 mm) of SYI from Mould Bay in April 1983 showing (a) concentration and layering of air inclusions and (b) enlarged profile of the top surface and shape of air bubble in the bulk. Arrows mark different layers. (photo by N.K. Sinha, unpublished).

Figure 6.10 Double-microtomed, 100 mm x 200 mm, horizontal thin section at about 3 mm below the SYI surface (corresponding to the mid-plane of the topmost bubbly white layer in Figure 6.9b) viewed through (a) parallel- and (b) cross-polarized light; the arrow indicates the long axis of Mould Bay and hence the tidal current (photo by N.K. Sinha).

258

SEA ICE

(a)

20 mm

(a)

(b)

20 mm

Figure 6.11 Double-microtomed, 100 mm x 200 mm, horizontal thin section at about 13 mm below the SYI surface shown in Figure 6.9b, viewed through (a) parallel- and (b) cross-polarized light; the arrow indicates the long axis of Mould Bay (photo by N.K. Sinha).

oriented columnar-grained S3 type with preferred c axis orientation parallel to the water current along the length of Mould Bay. The photo using parallel-polarized light shows rows of dark areas corresponding to the crosssectional views of the aligned air pockets. The surface of the thin section was also examined with a hand-held magnifier and an optical microscope to confirm that the darker spots were indeed produced by the air bubbles. The S3 type vertically oriented columnar nature of the ice grains (actually family of subgrains) can be seen clearly when the photographs in Figure 6.10 are compared with those in Figure 6.11 for a thin section at a depth of 13 mm (marked by the middle arrow in Figure 6.9). This section was prepared for examining the structure of the narrow, partially transparent layer sandwiched between the top two thick porous layers visible in Figure 6.9b. Expectedly, it exhibits lower density of entrapped air bubbles, but shows cross sections of several isolated large air bubbles with irregular edges. The cross section of a particularly large bubble with rather rough periphery can be seen in the middle of the micrograph. The continuity of the columnar S3 structural ice type, but with drastic reduction in the size of air inclusions, can be seen in Figure 6.12 for a horizontal thin section at a distance of 200 mm from the top surface (marked by the bottom arrow in Figure 6.9). The ice at this depth was practically

(b)

20 m m

20 m

Figure 6.12 Double-microtomed, 100 mm x 200 mm, horizontal thin section at a depth of 368 mm from the SYI surface shown in Figure 6.9b, viewed through (a) paralleland (b) cross-polarized light; the arrow indicates the long axis of Mould Bay (N.K. Sinha).

transparent with density of about 915.9 kg/m3, which is very close to that of absolutely clear transparent pure ice with density of 917.8 kg/m3. The parallel- and crosspolarized views of the thin section show complex substructure, which is similar to that of mature FYI except for the fact that the usual rows of brine pockets are not present at all. Usually, SYI maintains the microstructure of the FYI that it has grown from, except that most of the brine pockets have disappeared. The polarized light, however, is incapable of bringing out the details of the substructure. This can be better seen by the application of thermal etching as will be shown next. In all the macrographs of the SYI in Figures 6.10, 6.11, and 6.12, the substructures are somewhat ill-defined and often “fuzzy” in appearance unless, of course, examined with microscopes. These illustrations bring forward the difficulties in defining a grain and its boundaries, let alone determine representative grain sizes in sea ice. In reality, microstructure of directionally solidified (S2 and S3 type) sea ice is a conglomeration of subgrains with slight mismatches (say, less than 1 ) in the orientation of their lattice (c- and a axis). Consequently, it is not possible to use the conventional polarized light method to examine detailed fabric of the material. The presence of brine pockets along the boundaries of the substructures in FYI helps, somewhat, in delineating the structural details. However, as brine drains almost completely in SYI, brine pocket arrangement can no longer be used to identify subgrain boundaries. Nevertheless, the small mismatches in the

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

259

The SYI grains and subgrains with their basal planes oriented in the vertical planes and parallel to the growth direction, and their c axis favorably oriented along the direction of water current, act as the source for nucleation of the new growth.

6.2.3. Aging of Sea Ice: from FYI to MYI

C 1 mm

Figure 6.13 Optical micrograph of the surface of a microtomed horizontal section showing thermally etched subgrain boundaries and cross sections of irregularly shaped air bubbles inside a grain (strictly speaking, a family of subgrains) of SYI from Mould Bay in April 1983; the arrow indicates approximate c axis orientation of the family of subgrains [Sinha, 1985 / John Wiley & Sons, Inc.].

lattice orientations of the subgrains and air pockets remain locked in the ice. An example is shown in Figure 6.13. It shows etched subgrain boundaries and the structural details of entrapped air pockets and their locations with respect to these boundaries. Thermal etching technique (section 4.2.2.1) can easily be applied in the field laboratory and indeed add a new dimension to the possibility of extending forensic investigations to the study of ice substructure. It is obvious that almost complete desalination of sea ice under isothermal conditions during the final stages of melting occurs predominantly by gravity drainage. Unlike the early growth period when brine channels play important roles, drainage during the summer melt must occur primarily through grain and subgrain boundaries because of the absence of open brine channels in mature FYI. Brine channels formed near the growth front are eventually filled as the ice–water interface progresses downward during the growing season. The flushing of meltwater does not affect the fabric of the ice. The completion of the desalination processes leads to the strengthening of the grain and subgrain boundaries and hence closure of these boundaries for subsequent upward migration of brine (due to buoyancy forces, for example) when the drainage comes to a virtual end and the new growth starts to occur with the advent of the winter season.

Aging from FYI to older ice occurs when an FYI sheet survives the complete melt during summer. In most areas above the Arctic Circle, sea ice is known to survive through one or more melt seasons. Melting occurs at the top and the bottom surfaces of the ice cover. Floating sea ice floes or land-fast ice may go through a continuous aging process over a number of years. As a result, the drifting pack ice in the polar region of the north always contains a mixture of sea ice with different ages. Total disappearance of the annual sea ice in the oceans surrounding the continent of Antarctica is not unheard of. But nearly all the oceanic areas at latitudes below about 60 South, mainly in west Antarctic region, remain ice rich in terms of disintegrated fragments of old land-based ice (icebergs) and fragments of landhugging shelf ice. When matured FYI cover goes through its first melt season, normally during the middle of July to middle of September in the northern hemisphere, the reduction in thickness occurs due to melting from both the top and bottom. As the remaining sheet of ice enters a second season of winter freezing, new growth occurs under the remnant FYI (now SYI). This new growth continues until the end of the second growth season. At the end of the second growth season, the composite ice cover is called SYI; but this does not imply that the entire through-thickness ice sheet is two years old. An SYI floe consists of a layer of older ice on top of another layer (perhaps thicker) of new ice. When this composite ice sheet goes through a second melting season, three distinct possibilities can develop by the end of the second summer: (1) the entire floe may melt and disappear, (2) the top layer of older ice dissolves completely leaving only a reduced layer of FYI, (3) the top layer of older ice melts partially near the top surface and the newer ice at the bottom also melts partially. The composite ice cover, in case of scenario (3) with a layer(s) of newer ice at the bottom is still called old ice. This ice continues to face a new freezing period and a number of melting scenarios during its lifetime. The processes of partial melting and freezing may continue for many seasons. That is why even though we give age to an MYI floe, the floe may comprise layers of different ages lying on top of each other. No need to go through this complication. A fourth-year ice floe, for example, means that at least part of the floe has survived three meting seasons

260

SEA ICE

with no concern for the age(s) of the layers that compose it. The usual practice of identifying the age of an ice floe in floating pack ice is by no means simple, either visibly in nature or in remote sensing images. In most cases, it is not more than a wild guess work and the quality of results depends on the experience of the observer but often the results are erroneous. Examination of ice microstructure from sample cores may provide the right information, but this is a tedious and extremely time-consuming process. For all practical purposes, and to avoid complications and conflicts, sea ice floes older than FYI are recognized as MYI. However, with this rather gloomy view, sea ice age has been estimated based on tracking individual sea ice parcels, where each parcel is considered as a Lagrangian particle. Results are available in the form of age maps from a few agencies (see section 11.6). The product has been used in many studies to monitor the ice age across the Arctic. Given the coarse spatial and temporal resolution of the product, it is still good enough for practical applications.

6.2.4. Interface Between Old and New Ice in Second-Year Ice Profile

To the surprise of the team, the pit clearly showed a line of demarcation between the SYI and the underlying new FYI. The clarity and uniformity in the thickness of the SYI layer, as can be seen in Figure 6.15, were beyond the imaginations of the team members. The ice freeboard was estimated from the average densities, and the thicknesses of SYI (including the bottom layer of the new FYI) was around 200 mm. A hole was drilled with an ice auger in the middle of the open pit to the bottom of the ice sheet for flooding. The water indeed came to the level of estimated free board. The installation of a sled-mounted microwave antenna (University of Kansus scatterometer) is shown in Figure 6.16. Backscatter of sea ice formed in the pit, with related chemical and physical observations including the formation and growth of frost flowers, was recorded until the ice thickened to the gray-white stage (between 150 and 300 mm). It should be noted that ice formation under very calm conditions in the pit would be different than formation in open sea area. From one of the large blocks of ice obtained from the pit, a 100-mm-wide and 8-mm-thick vertical section containing the SYI and a part of the new ice at the bottom was cut with a bandsaw, and both surfaces were then sanded and polished manually. This section is shown in

As mentioned above, while old ice is labeled as such, its bottom part always contains new ice formed during the current freezing season. This is readily known but an opportunity emerged during the 1983 Mould Bay experiment to capture the interface between the two ice ages while studying the ice thickness profile. A 3 m × 3 m pit was cut with chain saws to a depth of about 0.7 m in the SYI sheet. The open-cut pit is shown in Figure 6.14.

Figure 6.14 AES-NRC team beside a 3 m x 3 m (about 0.7 m deep) pit they prepared near station 9 in Mould Bay (Figure 6.8), in April 1983; note the sharp demarcation line near the bottom between lighter SYI and darker new (FY) ice (photo by N. K. Sinha).

Figure 6.15 Vertical surface of a section of the pit in SYI smoothed to show layered structure of 0.45 m deep old ice and the darker new ice that grew below the old ice in September 1982; the horizontal lines marked on the ice are 0.05 m apart (photo by N.K. Sinha).

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE (a)

261

(b)

SY Ice

FY Ice

Interface

Figure 6.17 A vertical thick section (100 mm wide and 5 mm thick) of ice showing (a) the entire depth of the SYI and the SYI–FYI interface, and (b) details around the interface line (photos by N.K. Sinha).

Figure 6.16 A sled-mounted antenna was installed for monitoring microwave backscatter from the surface as the ice formed and grew in the 3 m x 3 m pond (photo by N.K. Sinha).

Figure 6.17a. It shows the lower density layer near the top surface in white and the layering in the SYI. Some of the layers were almost transparent and contained very little air bubbles (compare to the bubbly layer from MYI hummock ice in Figure 2.15). The layered structure indicates clearly the variability in the rate of growth when this section of the ice grew during the 1981–1982 winter season. The long vertical section in Figure 6.17a also shows the interface between the old ice and the new growth. A DMT thick section (about 5mm in thickness) was processed and its photograph showed the interface better (Figure 6.17b). This thick-section photograph was taken by using scattered light with two sources of illumination from the sides. It clearly shows the differences in the optical characteristics of the two types of ice. The relatively clear, brine-free SYI at the top appears to be dark except for the trapped elongated air bubbles in it. Note particularly the differences in the characteristics of the scattering of light immediately above and below the interface. The interface is remarkably sharp. The new ice immediately below the interface, for depth of about 5 to 10 mm, is very

cloudy due to the entrapment of extremely small, but numerous brine and air pockets. About 10 mm below the old/new ice interface, the new growth starts to exhibit some of the macroscopic features usually noticed in seasonal sea ice. The top end of the cloudy zone clearly shows a demarcation line between the old and the new ice, but it will be seen that there was continuity in the crystalline characteristics of the two types of ice across the border. Figure 6.18 shows thermally etched surface features in a solid-state DMT thin section of the ice above and below the SYI–FYI interface. Thin sectioning was performed as soon as possible after recovery of the ice blocks, and etching/photographing was performed in the field laboratory in Mould Bay. In this case, the thin section was kept inside the thermal etching box (see Figure 4.27) and allowed to etch thermally. The continuity of subgrain boundaries across the interface are clear indications that the dendrites at the bottom of the ice acted as the seed crystals and that each of the vertically oriented subgrains (or platelets) with their c axis in the horizontal plane continued to grow in the favorite direction of the vertical plane. As can be seen in the micrograph of Figure 6.18, the subgrains and hence the columnar gains simply continue to grow down when the winter arrived due to their favorable crystallographic orientations for growth. However, the rapid growth rate obviously allowed microscopic inclusions of air and brine to be trapped not only along the subgrain boundaries but even inside the subgrains.

262

SEA ICE

Subgrain Boundary

Interface

1 mm

Figure 6.18 Thermally etched surface of a vertical thin section near SYI–FYI interface in Mould Bay on 5 April 1983, exhibiting continuity of subgrain boundaries (thin lines) from SYI (top half ) to the cloudy new ice (bottom half ) (micrograph by N.K. Sinha).

What can happen when an ice floe composed of S3 type layers of SYI and FYI survives through the third of summer melts and starts to grow again? Will the S3 structure continue to grow? With anticipations, visual observations were continued at the weather station of Mould Bay on the SYI–FYI sheet during the summer melt of 1983. Unfortunately, the entire composite ice cover in Mould Bay did not survive the melt season of 1983. 6.3. HIGH ARCTIC EXPERIENCE WITH ICE OF LAND ORIGIN 6.3.1. Ward Hunt Ice Shelf and Hobson’s Choice Ice Island Experiment An ice shelf generally forms where one or more land glaciers spread out onto sea to produce a laterally widespread, relatively flat floating body of ice [Thomas, 1979]. In the Arctic, the Ward Hunt Ice Shelf (WHIS) is probably the most known feature. It is located in the northern part of Ellesmere Island at latitude 83.2 and between meridians of longitudes 72.0 and 78.5 (Figure 6.19). Currently, it covers a triangular area of about 440 km2. The bottom of ice shelves consists of layers of sea ice. The layer may also contain frazil ice if generated from supercooled water under and around the shelf. In glaciers, the deposited snow goes through morphological changes and the grains tend to grow fairly rapidly during cycles of thaw-refreezing or under the weight of the overlaid snow. They may start off as characteristically millimeter in size but they tend to grow bigger at greater depths near the bottom of the glaciers, where shear deformation is usually observed and strain-induced re-

crystallization changes the habit of the crystals. However, as the tongue of the glacier (i.e., the ice shelf ) spreads on top of the sea water, saline ice starts to develop at the bottom of the shelf, and frazil crystals may form in the water and float back up to attach to the bottom of the sheet. Although both glaciers and ice shelves consist of superimposed ice, one difference is worth noting. Glaciers are exposed to shearing forces as the ice body glides on mountain slopes. Therefore, glacier ice is subjected to complex thermal and mechanical histories. Microstructural analysis can assist greatly in identifying the texture and structural characteristics of ice and hence the source. Glacier ice tends to exhibit shearing features in the grains and also different families of healed cracks. Natural ice islands in the Arctic are massive tabular bodies of ice which periodically break off the east side of the WHIS. As mentioned in section 2.9.4, if the broken ice mass of land origin in the Arctic is huge, it is called ice island (as opposed to the smaller size icebergs). Any size of calved chunk from an ice shelf in the Antarctic region (no matter how huge it could be) is called an iceberg. Ice islands drift and remain circulating in the Arctic basin for several years, particularly in the Beaufort Sea region where motion is driven by the Beaufort Sea Gyre. They become hazards to offshore petroleum development in the coastal waters of the Beaufort and Chukchi seas. While irrelevant to the subject of this section, it is interesting to note that smaller “man-made sea ice islands” were also made inside the Canadian Archipelago during the years 1975–1985, by artificially thickening the sea ice cover. They were made for use as offshore drilling platforms for oil and gas exploration [Sinha, Strandberg, Vij, 1986]. These man-made ice islands, of course, never survived more than one or two years of summer melting, and drifted within the region of the Canadian Archipelago. The hazards of natural ice islands (tabular icebergs) in the waters of Arctic Ocean, Beaufort Sea and the Northwest passage were assessed extensively during the late 1980s to assist in the safe operation of ships and offshore structures (particularly offshore drilling platforms). Research activities had focused on the number of floating ice islands, their morphology and detection using remote sensing [Jeffries, Sackinger, Shoemaker, 1988, Jeffries and Sackinger, 1989], dynamics, motion, and recurrence intervals [Lu, 1988], physical-structural characteristics and stratigraphy [Jeffries, Sackinger, Shoemaker, 1988], strength-related properties of shelf ice [Frederking and Sinha, 1987, Jeffries and Sackinger, 2012], and mediumscale, macro-scale in-situ strength and laboratory strengths of multi-year ridge ice [Gagnon and Sinha, 1991, Meaney, Kenny, Sinha,1992]. When the leading edge of the WHIS, west of the northernmost weather and upper air station in Alert (Figure 6.19) broke away in 1982, the Canadian

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

263

Figure 6.19 Drift of the tabular iceberg (ice island), named “Hobson’s Choice” during 1983 to 1990, after calving from the WHIS in Ellesmere Island and final fragmentation during summer of 1990 between Ellef Ringnes Island and Axel Heiberg Island (from L. Jackson article in Canadian Geographic Magazine, Dec’88/Jan’89; with changes, courtesy of Polar Continental Shelf Project).

Government, through its PCSP, decided to occupy the largest chunk of 45 m thick tabular body of ice measuring 8 km × 3 km, shown in the aerial photo in Figure 6.20. This tabular platform was nicknamed by the scientists as the “Hobson’s Choice” ice island. This name may imply the meaning “no choice.” However, there was another deeper meaning for the name. It coincided with the contemporary Director of PCSP, Mr. George Hobson, who played the pivotal role in the success story of this organization and was much loved and respected by the scientists. Recognizing its possibilities as a moving research base, the PCSP erected a base camp in 1984, on Hobson’s Choice and constructed several shelters to accommodate research teams. The teams conducted several research studies of different geophysical aspects between 1984 and 1990. Following the separation from the East WHIS in 1982, Hobson’s Choice drifted south-west in the Canada Basin along the shores of the islands for about six years, as shown in Figure 6.19. This drift pattern was expected to

Figure 6.20 Aerial photo showing corrugated shelf-ice section of the “Hobson’s Choice” ice island surrounded by crushed sea ice (courtesy of Polar Continental Shelf Project).

continue, but the floating island stopped drifting after the winter of 1988–1989, and remained in the vicinity of the Meighen Ice Cap, between Axel Heiberg Island and Ellef Ringnes Island, until fragmented during the summer

264

SEA ICE

of 1990, when it moved inside the Sverdrup basin. Massive rescue operations were then performed to recover all the facilities and heavy test equipments. Since a wide sea-ice rubble field developed at the floating edge of the East WHIS before the calving events occurred in 1982, the ice island had a wall of old ice, to be called “shelf-fast rubble field” attached to it. During its floating life as an ice island, it developed a new wall of sea ice rubble field around the fractured surface created in 1982–1983. Eventually, therefore, the island had developed two sea ice components. Like a sandwich, it consisted of a 40 to 45 m thick strip of shelf ice with 5 to 10 m thick ridged MYI on either side, as illustrated in Figure 6.21. Its total area was almost 34 km2 and the mass exceeded 7 × 108 tons [Jeffries, Sackinger, Shoemaker, 1988, Jeffries and Sackinger, 1989]. The surface of the shelf ice component was distinguished by an undulating topography of nearly parallel hummocks and depressions, spaced about 200 m apart and extending for thousands of meters as can be seen in the photograph (Figure 6.20). PCSP built a campsite in 1984 on one of the hummocks schematically illustrated in Figure 6.21. In November 1988, thirteen hummock and eleven depression specimens from the shelf ice of Hobson’s Choice were sampled from areas around the PCSP site and shipped to the NRC laboratory in Ottawa by air in insulated boxes with dry ice. Once in Ottawa, the specimens were allowed to equilibrate to the cold-room temperature (−10 C). These pieces were then mounted on a lathe and machined to final test specimens with diameters of 67.7 to 74.8 mm, lengths of 142.4 to 169.5 mm and length–diameter ratios of 2.0 to 2.3. Horizontal thin sections were prepared by the method of DMT and photographed in cross-polarized light. Mean grain diameter was then determined according to d (mm) = (4 / πNA)−1/2, where NA is the number of grains per unit area

in mm2. The bulk density and ice crystal diameter data for one hummock and one depression cores are plotted in Figure 6.22. Although there is significant scatter in the data, a visible trend of increasing bulk density and crystal size with depth is apparent. The ice in both cores was superimposed ice, resulted from melting and refreezing of snow packs on the WHIS. The ice structure and stratigraphy of the hummock and depression cores were studied in detail and found to

Test site

PCSP camp

Shelf-fast rubble my ice

My-pack ice Air strip

Ward hunt shelf ice

5 km

Figure 6.21 Schematic of ice island (Hobson’s Choice) showing corrugated shelf ice in the middle, older pre-1982 shelf-fast MYI rubble field, newer post-1982 MYI rubble produced by pack ice, the PCSP camp, the air strip and the site for 65-m-long trenchbased medium-scale indentation tests of 1989 and 1990 experiments (sketch by N.K. Sinha).

0

0

–10

–10

Hummock

Depth, m

Depth, m

Hummock

–20 Depression –30

Depression

–20

–30 East ward hunt ice shelf (ice island) 1986 ice cores

–40 860

East ward hunt ice shelf (ice island) 1986 ice cores

–40

870

880

890

Density, kg·m–3

900

910

920

0

5

10 Grain size, mm

Figure 6.22 Depth dependence of density (left) and grain size (right) for the cores from hummock and depression of Hobson’s Choice ice island (fragment of Ward Hunt Shelf Ice) in the High Arctic (N.K. Sinha).

15

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

50 mm

contain equiaxed isotropic ice as well as transversely isotropic ice. The latter is indicative of refrozen melt ponds, and this was more common in the depression core where ice specimens are, on average, slightly denser and have larger crystals than hummock specimens. It is fair to assume that all the test specimens comprise mostly equiaxed isotropic ice, with a minor transversely isotropic ice component most likely in the depression specimens. Figure 6.23 shows the assemblage of the grain based on

Figure 6.23 Delineation of crystals in a vertical thin section of an ice core from the WHIS, Canada. The diagram at the right shows the configuration of the grain assemblage [National Research Council of Canada].

Depth, m

0.0

boundary morphology at 5.1 m depth of the hummock ice near a site close to the main camp site. 6.3.2. Multi-Year Rubble Field Around the Ice Island As shown in the schematic diagram of the Hobson’s Choice ice island, in Figure 6.21, the ice island consisted of three sections: corrugated shelf-ice, pre-1982 (i.e., older) shelf-fast MYI rubble field, and post-1982 (newer) MYI rubble produced by pack ice during the drift of the ice island. This is essentially a common feature of all the ice islands in the Arctic Ocean because sea ice rubble field develops adjacent to the freshly created fracture surface of the shelf ice after any calving incident, and as the fragmented section floats away another rubble field forms on the other exposed surface. Both the rubble fields grow and age as the islands drift year after year, but naturally one is always older than the other. Macrostructure of the ice could be seen clearly after grinding and polishing sections of the vertical sea ice walls required for the indentation tests. The polished surfaces revealed the outlines of ice blocks that could be seen at many places in the walls. They indicated that the ridging occurred when the thickness of the sea ice cover was about 0.5 m thick. The ice showed characteristics typical for old, consolidated ridged rubble field of sea ice examined in Mould Bay [Sinha, 1987]. Although the density varied only slightly between 875 and 886 kg/m3 (Figure 6.26), the salinity varied significantly with depth and location [Sinha, 1991]. Core 1 was situated approximately 20 m away from core 2 (no hummock or depression surfaces can be identified in such rubble MYI field). Ln general, the ice salinity increased from zero at the surface level to about 3‰ at a depth of 4 m. It is understood that the relatively high salinity, shown in Figure 6.24, is a manifestation of the

Core - 1 Core - 2

Core - 1

0.5

Oriented frazil

1.0

Columnar

1.5

Oriented frazil

2.0 2.5 3.0

Granular 3.5 4.0

Columnar 0

1

2

3

Salinity, ‰

4

5

850

265

900

Density, kg·m–3

950

100 mm Core-1

Figure 6.24 Salinity, density, and structure of rubble MYI that surrounded Hobson’s Choice ice island [Sinha, 1991 / American Society of Civil Engineers].

266

SEA ICE

formation of the MYI rubble, which is different than the typical mechanism of MYI floes. The rubble formation does not feature the mechanism of water percolation through brine channels before the channels become partially empty and enclose air bubbles (section 2.7.2). An interesting and unique experiment was conducted on the rubble MYI attached to the Hobson’s Choice ice island. Medium-scale indentation strength tests were performed in long trenches prepared in the newer MYI rubble area (Figure 6.25). Never in the history of sea ice research such long trenches were prepared in old rubble or ridged MYI ice fields. Tests were carried out in 1989 [Frederking, Jordaan, McCallum, 1990] and in 1990 [Jordaan et al., 1992]. In both the test years, the 65 m long, 3 m deep and 3 m wide trenches prepared for the mediumscale indentation tests showed that the ice was fully consolidated and free from any large-scale voids. The trench

walls were deep blue in color, as illustrated in Figure 6.26. This certainly was not expected in a ridged rubble field (white was expected due to the presence of numerous air inclusions of different sizes). Obviously, the aging process involved melting, filling the empty spaces between the fragmented pieces of ice, and refreezing, which caused the disappearance of voids. Thus, the consolidation processes in the rubble that started to form after calving in 1982 were complete by 1989 and 1990 when the trenches were prepared. 6.4. LABRADOR ICE MARGIN EXPERIMENT (LIMEX) The Labrador Ice Margin Experiment (LIMEX) was a series of two field programs. The first was LIMEX’87, conducted in the Grand Banks region off the east coast of Newfoundland, during the last two weeks of March [McNutt et al., 1988, Carsey et al., 1989]. The second was LIMEX’89, also conducted off the east coast of Newfoundland between 4 March and 4 April [Raney, Digby-Argus, McNutt, 1989]. Both experiments were timed to coincide with the period of maximum ice extent and the onset of ice retreat. The location of LIMEX’89 is shown in Figure 6.27. LIMEX was modeled on MIZEX

52°

Figure 6.25 Medium-scale indentation test site showing the 65 m long, 3 m wide, and 3 m deep trench in MYI rubble field of Hobson’s Choice ice island.

51°

Phase 1 Phase 2

50°

Gander

49°

(Airborne operations base)

Grand Banks

48°

St. John’s 47°

Ship-based observations 46° 55°

Figure 6.26 Trench wall in MYI rubble field attached to Hobson’s Choice ice island exhibiting a few ground and polished test surfaces.

Scale

50° 60 nm

Figure 6.27 Map of the LIMEX’89 site showing areas of shipbased activities.

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

activities (section 6.8). The LIMEX study area was chosen to be within the range of excursion from the most important oil exploration site on the Canadian east coast, namely Hibernia, which is not far from the ice extent. Moreover, the site is part of a highly operational marine route for ships traveling between Canada and Europe. LIMEX’87 and LIMEX’89 experiments sought to guide the development of the (then) future SAR systems onboard the European ERS-1 and the Canadian RADARSAT satellites. A polarimetric SAR system onboard aircraft Convair 580, operated by Canada Center for Remote Sensing at that time, in addition to an SLAR system onboard Lockheed Electra aircraft operated by the Canadian Atmospheric Environment Service were used. Participants in the experiments were from Canada, USA, UK, and Germany. The overall objectives of LIMEX’87 included: (1) studying the dynamics of the Labrador Sea ice cover, (2) surveillance of ice and oceanic conditions with airborne microwave instruments, (3) exploring the roles of oceanic and atmospheric processes in influencing ice conditions, and (4) development of air-sea-ice models to predict the behavior of the ice extent, compactness, and motion [McNutt et al., 1988]. In LIMEX’89, two ships served as research platforms, the Canadian MV Terra Nordica and the Canadian Coast Guard icebreaker Sir John Franklin. Figure 6.28 is a photo of Terra Nordica during its operation. Unlike the wind conditions that prevailed during LIMEX’87 and caused the compaction of ice against the shore, the wind

Figure 6.28 The Canadian MV Terra Nordica sailing through thin FYI in the MIZ in Grand Banks, east coast of Newfoundland, Canada in March 1989 during LIMEX’89 field expedition (courtesy of Canadian Meteorological and Oceanographic Society archive).

267

during LIMEX’89 prevailed from the northwest, causing the ice to be dispersed into the open ocean. The focus in LIMEX’89 was more on relating key ice and ocean properties to remotely sensed data using a relatively large suite of sensors. That included, airborne X- and C-band SAR, as well as the X-band SLAR, aerial photos microwave radiometer, a laser altimeter, a stereographic camera, and a search radar. Data were also acquired from ship-mounted scatterometer and sled-mounted radiometer. The focus on remote sensing made this experiment notable. In-situ measurements included snow wetness, albedo, floe sizes, rubble ice heights, and ice salinity, thickness and strength [Raney, Digby-Argus, McNutt, 1989]. Prinsenberg and Peterson [1992] presented sea ice properties in LIMEX’89. For LIMEX’89, a few physics-related and SAR-related objectives were set and pursued. The physics-related objectives included: (1) sea ice edge processes and structure (e.g., wind-driven motion, ice formation and decay, lateral and vertical heat fluxes, etc.), (2) ice–wave interaction in terms of wave penetration into the pack ice and the effect of swell on floe size, and (3) verification of environmental simulation models in the MIZ. The SAR-related objectives included: (1) acquisition of backscatter of ice under different conditions from a collection of C-band and X-band SLAR systems and comparison with backscatter modeling, (2) examining the physical properties of sea ice and snow in relation to the observed backscatter, (3) resolving the minimum floe size and lead dimensions that can be detected by the radar systems, (4) comparison of information obtained from the employed SAR and SLAR systems against information obtained from passive microwave data, and (5) simulation of C-band ERS-1 and RADARSAT images in preparation for data products to monitor offshore ice conditions. These objectives laid a powerful background for using microwave remote sensing in sea ice applications. Looking at the outcome form LIMEX, one can realize the priorities of remote sensing sea ice research that were developed beyond 1980s. For one thing, the condition of ice in the study area during LIMEX’87 was peculiar; hence more ice and remote sensing data had to be acquired under different conditions. For example, the relatively warm weather during LIMEX’87 caused more moisture in the snow and brine drainage within the sea ice. Therefore, the radar backscatter data were affected by these conditions. Due to dominant onshore winds, the ice regime during this experiment was compacted against the shore and became landfast. However, the penetration of swell broke it and produced pancake and brash ice covering an extensive area. Moreover, rafting dominated the offshore region. All those factors limited the application of the observations to areas of similar ice conditions, which usually exist in other southern

268

SEA ICE

latitude locations in the northern hemisphere, in the Southern Ocean [Carsey et al., 1986], or perhaps in recently formed MIZ regimes in the Beaufort Sea north of Alaska, where Arctic ice is being declined at a fast rate (see section 2.9.3.1) A few observations during LIMEX’87 have become classical knowledge about SAR–ice interactions. For example, at the ice edge, bands of bright signature appeared in the image of 23 March with width less than 1 km. On 26 March, the width increased to a few kilometers [Carsey et al., 1989]. Observations from the ship and helicopter indicated that those bands represented a composition of granulated brash ice between pancake ice pieces. Pancake and brash ice were caused by the wave action and both should trigger high radar backscatter in the C-band. When brash increases, surface roughness increases and the backscatter is expected to increase. Inversely, the decrease in the amount of brash should lead to decrease in backscatter. However, less backscatter was observed as the quantity of brash ice was increased. More close-up photos of the ice field showed that the brash increase reduced the exposure of pancake floe edges and consequently the backscatter. This striking observation was accompanied by an opposite observation, namely increasing radar backscatter when the amount of brash decreased due to melt. In this case, the more exposure of the pancake edges led to the increased the backscatter. LIMEX’87 was one of the early experiments, where modeled backscatter was calculated based on field observations. Detailed observations of snow and ice surface properties were used in a simple model of radar backscatter in the C-band [Drinkwater, 1989]. For example, roughness of snow and ice surfaces was obtained, for the first time, using an apparatus that enables quantification of roughness in terms of surface height, and correlation length. A few conclusions were drawn from results of the backscatter modeling, which helped to establish a framework for understanding sea ice radar backscatter later. For example, when a damp snow cover is present, the C-band signature becomes largely responsive to snow rather than the underlying sea ice parameters. When melt ponds are formed, a specular reflection of the backscatter results in the steepest angular variation within an incidence angle between 40 and 80 . In the absence of snow, the backscatter is controlled by the ice salinity and surface roughness. Also, backscatter has been proven to be a suitable parameter to identify areas of deformed and undeformed ice. It is interesting to note that ice motion was tracked in LIMEX’87 using analysis of sequential SAR images in 4 days in a row. In areas where individual floes could not be identified, ice tracking was made possible using feature tracking in successive images (e.g., bright ridges in SAR images). When floes can be identified, tracking of

individual ice floes has proven to produce most accurate motion vector field (see section 11.7.1). LIMEX also developed information on oceanic processes. One of oceanographic features that was revealed for the first time by SAR data was the mesoscale eddies at the ice edge [Gascard and Clarke, 1983]. Tang and Ikeda [1989] showed evidence and presented explanation of upwelling water near the ice edge as indicated by increase of surface water density. This was observed in previous field studies. Data from both LIMEX experiments were most useful in identifying gross ice features, such as ice edge location with its composition of floe types and sizes, ice drift through preliminary ice tracking algorithms, ice melt rate, ice thickness, and wind-driven ice divergence. The limited ice and snow properties obtained were not enough to cover the suite of parameters needed to understand SAR images. That was partly because knowledge about SAR–ice interaction was not matured at that time. This, perhaps, was one of the reasons for the underuse of remote sensing data from LIMEX. In this context, it should be mentioned that problems in calibrating the polarimetric SAR from the airborne sensor onboard the Canadian Convair-80 aircraft hindered the use of these data. Results from LIMEX’87 were published in the special issue of the IEEE Transactions on Geosciences and Remote Sensing (TGRS) in 1989 (volume 27 number 5) and results from LIMEX’89 were published in the special issue of Atmospheric-Ocean in 1992 (volume 30 number 2).

6.5. SEA ICE MONITORING AND MODELING SITE (SIMMS) PROGRAM The SIMMS program was a Canadian initiative conducted in the Arctic spring period (April–June) over 8 years from 1990 to 1997 to study snow-covered sea ice processes interacting with atmospheric parameters and remote sensing observations. This program was contemplated and founded by the Late Dr. David Barber, when he was a graduate student at University of Waterloo and later a professor at University of Manitoba (both in Canada). The base site was in the Resolute Bay, Nunavut, Canada, but several manned observation camps were established in different areas around Resolute to study ice and snow at sites of different FYI and MYI. SIMMS field work attracted several Canadian and US partners. The field work was supported through the Institute for Space and Terrestrial Science at the University of Waterloo during the first 5 years, and the Centre for Earth Observation Science, University of Manitoba for the rest of the program period. A remarkable feature of this program was the participation of many graduate students; some of them chose sea ice and cryosphere as research or operational career later. Resources for this program

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

Figure 6.29 The warehouse of the Polar Continental Shelf Program, Government of Canada, in Resolute Bay (photo by M. Shokr).

included instrumentations to measure snow, ice, and atmospheric parameters (including radiative parameters), and data from airborne and satellite-borne SAR systems, notably ERS-1 (launched in 1991) and RADARSAT-1 (launched 1995). A significant resource was the accommodation and logistic facilities of the PCSP (see section 6.2.1). A photograph of the warehouse of PCSP in Resolute is shown in Figure 6.29. In addition to the site camps on ice, a few researchers stayed in the fancy (by Arctic standard) PCSP accommodation facility and commuted to the ice by snowmobiles. Sites for conducting measurements were selected based on satellite and airborne SAR surveys. The overall objective of SIMMS program was to characterize the interannual variability of the snow-covered sea ice in the experimental site and understand the physical mechanisms that give rise to the major processes operating across the ocean–sea ice–atmosphere (OSA) system. Those processes were used to improve models of ice– atmosphere dynamics. Five specific objectives were set [LeDrew and Barber, 1994]: (1) to measure physical parameters of snow and sea ice over a continuum of space and time scales, (2) to understand how the major fluxes of energy and mass are partitioned within the OSA interfaces, (3) to retrieve geophysical properties and energy fluxes from remote sensing data, (4) to study and model processes within OSA, and (5) to monitor the OSA change and interannual variability. Objectives 1 to 3 were directly addressed within the SIMMS field programs, and objectives 4 and 5 were achieved by employing observations into more regional and hemispheric studies. Within SIMMS program, numerous research projects were developed by many investigators to address science investigations within the set objectives. The following paragraphs highlight results relevant to the remote sensing observations linked to physical parameters of snow-covered ice. Snow parameters included

269

grain morphological parameters (size and statistical moments), snow wetness, density, thickness, and hoar depth. These were performed at numerous snow pits on FYI and MYI. Data were sampled at intervals along the snow depth to generate profiles of the parameters. Special focus was placed on the hoar layer at the base of the snow cover on FYI. This is a highly saline snow layer with salinity exceeding a range of 10‰ and may reach 50‰ [Barber et al., 1992]. Snow grain morphology was measured using the crystallography method and found to vary along the depth with coarser gains usually found at the snow base or within layers of the snow column, possibly resulting from thaw-freezing cycles. Sea ice physical parameters were measured in detail and numerous ice cores were extracted and examined using the DMT technique and etching methods of thin sections. This resulted in wealth of information, some is included in this book. SIMMS data were used in a microwave scattering modeling study [Barber and Nghiem, 1999] to investigate the dependence of radar backscatter on snow thickness over smooth FYI. The study showed that the thermodynamics of the snow cover affect wave propagation, attenuation, and scattering through the brine volume distribution. However, the effect was subtle and specific to certain ranges of salinity, surface roughness, and thickness of sea ice. In general, findings from SIMMS about snow wetness and metamorphism stimulated several studies to retrieve snow parameters from microwave observations. For example, a theoretical framework developed during SIMMS program was used later to develop methods for retrieving snow water equivalent (SWE) and snow thickness. One conclusion was that, a thin snow cover will respond to changes in air temperature but the change is more pronounced in the snow–ice interface temperature. This will result in significant change in complex permittivity and radar scattering. A thick snow cover, on the other hand, produces negligible change in the temperature at the snow–ice interface. One of the salient features of the SIMMS program was the use of a large suite of remote sensing instruments. The space-borne sensors included the SAR systems from the European ERS-1 and the Canadian RADARSAT-1, Landsat TM (Thermal Mapper), SPOT (Systéme pour d’observation de la terre), the US AVHRR (Advanced Very High- Resolution Radiometer), and the passive microwave radiometer SSM/I (Special Sensor Microwave/Imager]). In addition, two aerial SAR systems were used; the X-/C-band polarimetric radar system of Canada Center for Remote Sensing (CCRS) onboard the CV-580 and the X-band airborne system of Intera Technologies. Some of these observations came to a successful end but others did not. SAR images were the prime tool for selecting camp sites to conduct in-situ measurements on different ice types. SAR data, both from airborne system and

270

SEA ICE

satellite, were used to develop seasonal evolution of backscatter from FYI and MYI (see section 9.3.1). Observations provided first indications that discrimination between the two types, which is possible during winter, is not possible once the snow cover becomes wet and later when the ice surface becomes flooded. Later studies avoided using SAR observations for sea ice classification in the summer. Another example on the use of remote sensing data was presented in Key et al. [1994] to validate estimation of skin surface temperature (section 11.5.1) using thermal infrared AVHRR observations obtained in May–June 1992, to validate skin temperature derived from a split-window technique presented in Key and Haefliger [1992]. The observed skin temperature ranged from 0.1 C to 3 C. The advantage of using a field opportunity such as SIMMS was to test a number of parameter retrieval methods. In Key et al. [1994], the skin temperature (actually the radiating-layer temperature) was obtained using three methods: (1) derivation from measurements of upwelling longwave (broadband) radiation, (2) hand-held IR thermometer to measure spatially distributed temperatures, and (3) a thermocouple placed approximately 1 cm below the snow–air interface. Comparison of the AVHRR-driven temperature against data from each method indicated that the temperature from using methods 1 and 2 were higher than the AVHRR temperature by a small fraction of a degree, while method 3 produced less temperature deviation by the same fraction, illustrating the insulating quality of the overlying snow. The intensive work on linking remote sensing data to geophysical parameters during SIMMS program led to identifying a link between the backscatter coefficient σ o from SAR observations and shortwave albedo. Another link was observed, yet with some ambiguity, between σ o and the transmitted Photosynthetically Active Radiation (PAR) for a snow-covered FYI [Barber and LeDrew, 1994]. Modeling results indicate that σ o from both 5.3 GHz (C-band) and 9.25 GHz (X-band) frequencies, at HH polarization and incidence angles of 20 , 30 , and 40 can be used to estimate the daily averaged integrated climatological albedo (α). The models of the C-band at any incidence angle produced equally precise estimate of α. In general, the 8-year SIMMS program achieved some of its set objectives, but the opportunity attracted many partners to launch their own investigations, and that enriched the overall outcome from the program. The fact that the program relied on the Canadian Government’s facilities established by the PCSP (accommodation and logistics), with no need for a ship platform, made it relatively inexpensive, hence more attractive to foreign participants. To the US partners, it was interesting to conduct such a study in an area different than the traditional coastal area of Barrow at the northern tip of Alaska. SIMMS study area during the field expedition in the

spring (April–June) was always covered by thick FYI with some scattered MYI.

6.6. THE SURFACE HEAT BUDGET OF ARCTIC OCEAN (SHEBA) SHEBA was a three-phase scientific mission. The first was conducted in 1995–1996, which involved reviews of modeling studies, including general circulation models (GCM), and previous data sets to understand interactions within the ocean–atmosphere–sea ice system and the impact of climate change on the Arctic environment. The second phase, which is the subject of this section, was a year-long (2 October 1997 to 12 October 1998) field experiment, where the Canadian Coast Guard Ship Des Groseilliers was frozen into the ice about 570 km north and east of Prudhoe Bay, Alaska, and an ice camp was established around it (Figure 6.30). Throughout the period of this phase the ship drifted over 2800 km, starting at (75 N,142 W) and ended at (80 N,162 W). The third phase (2000–2002) included data analyses and continuation of modeling activities using the observations acquired in phase 2. The SHEBA winter was slightly colder than average, but the melt season was longer, which resulted in overall thinning of ice [Perovich et al., 2003]. A summary of the SHEBA project with a focus on the field experiment program is presented in Uttal et al. [2002]. A comprehensive presentation of the atmospheric and sea ice results is presented in Serreze, Maslanik, Key [2002]. A list of SHEBA publications in the Journal of Geophysical Research: Oceans is available through the link https://agupubs.onlinelibrary.wiley.com/doi/toc/ 10.1002/(ISSN)2169-9291.SHEBA1. SHEBA reconnaissance imagery dataset is publicly available through https://nsidc.org/data/G02180/versions/1.

Figure 6.30 The Canadian Coast Guard Ship Des Groseilliers frozen in the ice cover during SHEBA experiment and the ice camp around it. The location was 75 N, 142 W (credit NOAA/ESRL/Daniel Wolfe).

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

The program was based on the premise that more data were needed to understand the variability of Arctic sea ice cover, which was predicted differently using different versions of the GCM. The program was timely because the trend of sea ice continued to show decrease in the extent and thickness after the SHEBA’s period. The ice extent set an all-time low record in September 2012. The idea of a ship-based experiment was most suitable because it provided logistics to accommodate a large number of researchers, power to operate their instruments and laboratories, and a machine shop to support the projects. This idea was implemented for the first time in recent history during the SHEBA program and stimulated more implementations later in other Arctic scientific expeditions (e.g., MOSAiC program, as described later). It is worth noting that in 1893, the Norwegian explorer Fridtjof Nansen used the same idea by freezing his ship Fram into the ice pack near the New Siberian Islands, hoping the ice drift under the transpolar drift stream would take it to the North Pole point. It did not happen, but the drift routed the ship toward the Atlantic Ocean. The overall objectives of SHEBA were to: (1) identify the feedback mechanisms of surface albedo and cloud radiation into the ocean–ice–atmosphere processes that control the Arctic ice, and (2) improve the simulation of ocean–atmosphere–ice interactive processes in climate models. Specific objectives were set to guide the research proposals, which included: (1) relating ice surface processes to the atmospheric and oceanic forcing, (2) developing relations between snow-covered ice surface and albedo measured using any given shortwave radiation, (3) identifying the scale of ice and atmospheric processes (from local to aggregates) that suits climate models, (4) constructing a dataset for developing and testing climate models, (5) improving the interpretation of satellite remote sensing data in order to be able to interpret the Arctic climate system better, and (6) documenting and understanding the mechanisms that extend the influence of the Arctic system to the global environmental system. The observational phase included coordinated field measurements by nearly 200 scientists to address numerous questions using in-situ observations and satellite remote sensing data. A few results relevant to sea ice properties and its remote sensing observations are highlighted in the following. Perovich et al. [2003] measured ice thickness at 135 sites of different ice types to estimate the sea ice mass balance. The thickness ranged between 0.3 and 8 m with snow cover between a few centimeters to more than a meter. The average winter ice growth was 0.51 m and the summer melt was 1.26 m, which consisted of 0.64 m of surface melt and 0.62 m of bottom melt. Accretion ranged from zero for thick ridged ice to more than a meter for YI during the ice growth period. Melt ponds tended to have a large amount of surface melting. Maximum surface melting

271

was observed in July, while bottom ablation peaked in August. Such a wealth of in-situ measurements constitutes the best truth data that can be used to validate information retrieved from satellite observations. Unfortunately, advanced satellite remote sensing tools and their retrieval methods were not available during SHEBA. That could have utilized the detailed surface measurements to advance ice parameter retrieval at higher accuracy. For example, to advance the applications of SAR polarimetry (section 7.6.3) in retrieving snow and ice parameters, an in-situ data obtained from a field program similar to SHEBA is needed. The future carries the possibility for this option. Sea ice extent and concentration were compiled from products of the passive microwave observations during SHEBA. The purpose was not to verify existing operational products, for that cannot be achieved by using the fine-resolution in-situ measurements (a few meters) against the coarse-resolution products (a few tens of kilometers). The use of the ice concentration data was to monitor the ice drift of the SHEBA camp to ensure that it would not reach an MIZ with low ice concentration. Another component of field/satellite study was ice drift, which was summarized in Serreze, Maslanik, Key [2002] and highlighted here. Once again, the drift data were obtained from sequences of passive microwave satellite observations using SSM/I satellite data with application of the familiar maximum cross-correlation technique (section 11.7.1.1). In addition, the drift was also estimated from a set of buoys provided by the international Arctic buoy program (IABP). The results at 62.5-km grid showed agreement between SSM/I and IABP estimates with difference root mean square around 6 km/day. When both sources were merged via optimal interpolation, the daily average data were used to produce monthly and annual means for the 1988–1994 period, and the data were made available via AVHRR pathfinder. Ice drift was examined in relation to wind and the already established relation between large-scale drift pattern and geostrophic wind was confirmed. As mentioned in the beginning of this section, phase 1 of SHEBA involved assembling existing datasets of the mean and variability of the Arctic atmospheric and sea ice conditions. Those datasets were generated inhouse and in previously published results. The analysis concluded that since late 1980s, an increase in cyclone activity caused a shift in atmospheric circulation favoring ice advection away from the Siberian coast. Results of ice drift as well as ice divergence and convergence from the IABP system were not direct output from in-situ observations, yet they fulfilled a purpose of the project. The multi-sensor data obtained during SHEBA observational phase along with in-situ coincident data served to cross-validate ice parameters. The use of airborne sensors revealed heterogeneity of surface parameters. Haggerty, Maslanik, Curry [2003] found a large increase in the

272

SEA ICE 1.0

0.8

Dry snow

0.2

0.0 Apr 1

May 1

Pond formation

0.4

Melting snow

Albedo

0.6

Jun 1

Pond evolution

Jul 1 Date

Aug 1

Fall freezeup

Sep 1

Figure 6.31 Evolution of wavelength-integrated albedo from 1 October 1998 to 27 September 1998, during SHEBA field experiment when the sky was clear. The standard deviation of the albedo measurements from all points along the transect line is shown as open circles [adapted from Perovich et al., 2003].

surface temperature of thin ice during the transition month of May and much smaller increase of thick ice temperature with limited spatial variability of 0.5 K. Generally speaking, major field studies such as SHEBA are excellent opportunities for quantification of the spatial heterogeneity and temporal evolution of ice surface parameters during the transition seasons (freezing and melting) using an extensive sampling scheme or data from fine-resolution airborne sensors. This is manifested in a study by Sturm, Holmgren, Perovich [2002] to estimate the spatial variability and temporal evolution of the snow in winter during SHEBA. Snow depth and ice thickness were measured at transect lines ranging from 200 to 500 m in length. Measurements were performed 14 to 16 times during the year. The study found that the snow started in October, reached maximum depth in midDecember, and then remained relatively unchanged. The mean snow depth was 33.7 cm with a bulk density of 0.34 g/cm3 and average SWE of 11.6 cm. As expected, the greatest depth and the greatest variability were found in ridged and rubble ice areas. Sturm, Holmgren, Perovich [2002] also showed that snow depth and stratigraphy varied significantly with the underlying ice type. This kind of information can be used to support the snow depth retrieved from remote sensing observations. Another advantage of field measurement during SHEBA was monitoring the seasonal evolution of the albedo from MYI [Perovich et al., 2003]. Though this was performed in the previous and later field experiments, SHEBA offered a long-term monitoring of this parameter during the daylight season (April to October). Spectral and wavelength-integrated albedo were measured along a

transect of 200 m, every 2.5 m using field-portable spectroradiometers and shortwave radiometers. The short spatial interval of the measurements allowed linking the variability of the albedo to the variability of snow conditions; namely thickness, metamorphism, melt, and weather events. As a result, the study identified five distinct snow phases linked to the evolution of the albedo: dry snow, melting snow, pond formation, pond evolution, and fall freeze-up. These categories were also season-driven as shown in Figure 6.31. High albedo (0.8–0.9) was associated with dry snow. In summer, albedo from bare ice surface and dark melt pond was around 0.65 and 0.1, respectively. Note the gradual decrease of albedo as spring conditions advanced to summer, and the sharp increase in the fall during refreezing and snow accumulation (Figure 6.31). Perovich et al. [2003] reported good agreement between albedo from ground measurements and aircraft measurements using two identical Eppley radiometers, one pointing up and the other pointing down, even though the footprint of the radiometers were larger than the point measurements on the ground. Characterizing surface condition with its albedo can be achieved more accurately using in-situ surface ground measurements.

6.7. THE NORWEGIAN YOUNG SEA ICE EXPERIMENT (N-ICE) The Norwegian Young Sea Ice Experiment took place between 12 January and 24 June 2015, along a route north of Svalbard. The expedition was carried out using the Norwegian Polar Institute’s research vessel Lance as a

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

Figure 6.32 The Norwegian research vessel Lance, on 17 February 2015 during the N-ICE expedition north of Svalbard (courtesy: Mats Ganskog, Norwegian Polar Institute).

research base (Figure 6.32). A few camps were setup on the floes around the ship. The ship was drifted with the ice through a route bounded by 78.5 N–81 N and 8 E– 28 E. The total distance covered by the ship was about 7420 km in 111 days, with 69 researchers and 27 support staff. The drift was so fast at times that caused the ice floes carrying the research camp to break up a few times, and a new camp had to be established every time. The ice pack consisted of FYI and SYI, and ice thickness on the camp floes ranged between 0.81 and 2.7 m with mean 1.49 m, while snow depth varied between 0.13 and 1.12 m with mean 0.48 m [King et al., 2018]. While this ice thickness did not fit the YI category as specified in the title (YI is defined as ice of thickness less than 0.30 m), it was considered by the research team thinner than the thickness of the pack ice in the central Arctic (hence the word “young” in the title). There was probably YI along the ship’s route when ice broke but was not sampled because of safety issues. Indeed, YI is the most challenging ice type to study not only because of its unsafe thickness but also due to its high spatial and temporal variability. This is reflected in its wide range of radiometric and scattering measurements [Shokr, Asmus, Agnew, 2009]. Nevertheless, N-ICE2015 was one of the pioneering field studies of sea ice in such marginal Arctic region, which is characterized by its highly dynamic ice and less ice concentration. As Arctic sea ice cover is shifting to younger and thinner pack, it was thought that Arctic sea ice data available prior to the time of N-ICE’2015 did not reflect this new reality. Therefore, the objective of the program was to understand the mutual effect between Arctic sea ice and other components the Arctic system (atmosphere and

273

ocean) under the new reality, and how these effects impact the ecological and biological components of the Arctic [King et al., 2018]. Some research questions specific to the study region were: (1) What causes ice melt: warm Atlantic water or solar heat? (2) How does thinner ice cover affect ice dynamics and drift modeling? (3) How does ice thinning affect the ice-associated ecosystem? and (4) How does thinner ice respond to extreme weather systems? The program was a multi-disciplinary endeavor (as usual with most Arctic scientific expeditions) with participation from 5 Norwegian and 14 international collaborators. Measurements included energy fluxes, sea ice dynamics, as well as atmospheric, oceanic, and sea ice parameters. The biological and atmospheric components of the program were more intensive than the sea ice component. Results from the experiment are published in special issues in three sections of the Journal of Geophysical Research: Oceans, Atmospheres and Biogeosciences. More sea ice papers, with key points from each paper, are found in the Journal of Geophysical Research: Oceans, special issue, which is available through the link https:// agupubs.onlinelibrary.wiley.com/doi/toc/10.1002/(ISSN) 2169-9291.NICE1. Datasets from the experiment, covering several atmospheric, oceanic, and sea ice aspects have been made available through the link https://www.npolar.no/en/projects/ n-ice2015/#toggle-id-2. Examples of datasets relevant to the subject of this book include snow depth, ice thickness, sea ice biogeochemistry, surface topography of a selected floe using photogrammetric measurements, sea ice and snow surface elevation from laser leveling exercises, ship radar images, snow bit data, buoy data, and ice core physics of temperature, salinity, and density. Highlights of the results are presented in a summary of the experiment, available through https://www.npolar.no/ en/projects/n-ice2015/. A few highlights of the results are presented here. (1) Snow depth was larger than what was predicted by the models. (2) Many storms occurred in winter, bringing moist and warm air, which slows down the ice growth and affects ocean mixing. (3) Thinner ice was easy to break and deform, hence more ridging and lead formation were observed. (4) Leads allowed enough heat to reach the water and therefore maintain an algae bloom under thick ice cover in the vicinity of the leads. Results relevant to sea ice physics and remote sensing are presented in the following. Johansson et al. [2017] investigated the backscatter from three satellite SAR systems in X-, C-, and L-band in relation to ice thickness, snow depth, and surface roughness measured from a helicopter-borne thickness sensor. The study showed how backscatters from the three systems complement each other, and raised awareness of the limitation on using the X- and L-band data. It has

274

SEA ICE

drawn the all too familiar conclusion that the full range of ice thickness cannot be directly retrieved from SAR data. Surface roughness and snow pack structure have stronger influence on sea ice classification than ice thickness. In another study [Provost et al., 2017], snow and ice instruments were deployed on FYI and MYI floes to document temperature and thermal diffusivity profiles at 2 cm vertical resolution. From the profiles, snow and ice were distinguished and so was the temporal evolution of each medium. Surface flooding was detectable from the evolution of the thermal diffusivity, temperature, and heat propagation through the underlying ice. Slush was also detected and upon freezing was defined as snow-ice (yet with no examination of the crystalline structure). Recall that snow-ice type with its fine- to medium-size equiaxed grains is formed when water saturates the snow deposited on ice surface then freezes (this is T1 ice in Table 5.1). Provost et al. [2017] noted that surface flooding resulted from two processes: (1) storm-induced breakup of snowcovered floe and (2) loss of buoyancy due to basal ice melt. The study offered explanation of how heat flux from ocean to ice increases after the storm, causing the ice to be warm and permeable, hence surface flooding would be associated with vertical intrusion of brine. It also asserted that superimposed ice (see definition in section 5.1.3) with frozen slush may become more frequently observed on the surface of the Arctic sea ice in future. Since N-ICE’2015 was conducted in a highly dynamic sea ice regime, it was expected to generate information on ice motion and deformation. Deformation is described and quantified in a study by Oikkonen et al. [2017] that addressed small-scale ice deformation within compact ice pack and MIZ. Measurements were performed using ship radar images covering 15 × 15 km2 with a resolution of 12.5 m to monitor deformation of four ice floes while drifting during the experiment period. Image analysis methods were used to identify deformation features in a sequence of images. Mean deformation rate was estimated for each floe with different length and time scales. High deformation rates occurred within the ice pack only under high wind and drift speed, while deformation was observed within the MIZ during calm conditions. The deformation rates followed power law with respect to length and time scale. Tracking the development of the deformation reveals information on the mechanical structure of the floe and the wind history because deformation events are initialized along lines of previous damage [Oikkonen et al., 2017]. Since an objective of N-ICE’2015 was to capture the scope of the recent ice thinning in the Arctic, Rösel et al. [2018] compared ice thickness and snow depth measured during the experiment using helicopter-borne electromagnetic induction (EM) instrument and in-situ

measurements. The observed mean snow depth was 0.53 m, which was 73% above the average from historical observations covering the years 1955–2017. The total ice and snow thickness from the EM data was 1.7 m (measured in April–June 2015). This was below the values ranging between 1.8 and 2.7 m, reported in historical observations in the same region and time of the year. The ratio of snow depth to ice thickness was exceptionally high just because of ice thinning. While the ice/snow thickness study by Rösel et al. [2018] offered an insight into the declining trend of the ice cover, it is important to integrate the information with ice type concentration, which implies the ratio of thick to thin ice (thin according to the standard definition of less than 0.30 m thick). Data from field studies has continued to be useful for many years after acquisition as new opportunities of applications arose. Based on measurements of snow data during N-ICE’2015, Rostosky et al. [2020] evaluated uncertainty of snow depth retrieval from passive microwave data using the Microwave Emission Model for Layered Snowpacks (MEMLS) (section 12.2). Depending on the sensor’s frequency, the uncertainty of depth retrieval varied between 11% and 19%. Snow properties were found to be a major source of uncertainty when using the 6.9 GHz frequency, while snow properties, ice properties, and clouds were equally strong sources at the higher frequency of 36.5 GHz. In a study by Rösel et al. [2017], the authors attempt to address the hanging question of how representative sample measurements in the field are of a wider footprint of satellite scale? They used extensive ground-based and airborne observation and compared them with classified SAR images from ALOS-2 Palsar. The idea was to verify the ice class using the relevant ground and airborne measurements. Because of the difficulty of point-to-point comparison of in-situ measurements to coincident observations of drifting ice, a statistical comparison was used. Three ice classes were used: thin, level, and deformed. These classes fit SAR observations more than the traditional age-based classes (section 7.6.3.4). Findings from the study confirmed the utility of the satellite data in retrieving key information such as fraction of thin ice, heavily ridged ice, and other features with SAR surface expression. While the results did not provide exclusive answer of the posed question, the study provided important information to expand the operational use of SAR for sea ice.

6.8. MARGINAL ICE ZONE (MIZ) EXPERIMENTS As a result of the significant retreat of Arctic sea ice since the first minimum ice extent record in 2007, a few coastal areas and continental shelves became ice free.

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

At the same time the ice edge started to shift northward, forming MIZs in areas that were traditionally covered with compact ice. This is particularly true in the Pacific Arctic region, where greatest sea ice retreat and thinning has been observed. Interest in the MIZs started earlier than 2007 but increased after that date. A number of MIZ field studies have been conducted since early 1980s. MIZ field studies usually carry strong oceanographic and biological components than sea ice component and this is, in general, typical for ship-based experiments in the Arctic even within the pack ice. McPhee [1983] provided an interesting expression for the status of sea ice work within the entire scheme of a scientific polar sea expedition, “the joker-in-the-deck.” The author suggested that MIZ modifies momentum transfer from the atmosphere, alters surface albedo, serves as an efficient thermal insulator, damps surface wave motion, and may substantially change both temperature and salinity structure in the upper ocean by melting or freezing. A comprehensive description of the science and processes in MIZ is given in Lee et al. [2012]. The study pointed out the need for work on more specific questions to further our understanding of such processes and feedbacks, hence constrain model representation of sea ice distribution and thickness. Two MIZ experiments followed these recommendations. In the rest of this section, brief information about four major MIZ experiments are presented: the Marginal Ice Zone Experiment (MIZEX), the Co-ordinated Eastern Arctic Experiment (CEAREX), the marginal ice zone observations and processes experiment (MIZOPEX), and the MIZEX-DRI (Departmental Research Initiative) of ONR. A special issue of Journal of Geophysical Research: Oceans on MIZ was published on 20 September 2012. MIZEX was conducted between 1983 and 1987, over two winters and two summers in the two Arctic regions, the Bering Sea and the Greenland Sea. MIZEX was sponsored by the ONR with participation of several American and international institutions. It was composed of two parts, MIZEX west, conducted in the Bering Sea between 5–27 February 1983 [Cavalieri et al., 1983], and MIZEX east, conducted in the Fram Strait region in 1983 (June– August), 1984 (June–July), and 1987 (March–April) [MIZEX Group, 1986, Johannessen, 1987]. The scientific questions relevant to sea ice that were addressed in the MIZEX program revolved around a few issues that included: (1) sea surface temperature, (2) statistics of ice floes including floe size, ablation rates, melt pond occurrence, and albedo influence, (3) ocean waves attenuation in the sea ice cover (measuring wave decay, change in spectrum and direction while penetrating sea ice, and estimating the effect of wave penetration on floe size distribution), (4) the role of ice edge eddies in ice mass transport

275

and melting (measured using radar transponders and buoys equipped with weather stations and current meters), and (5) the verification of ice rheology models and ice edge kinematics against in-situ measurements. The remote sensing component of the MIZEX east program in 1983/1984 was one of the most comprehensive, ever used in a sea ice field experiment. Two typical objectives were set for the remote sensing use [Johannessen, 1987]: (1) to obtain a synoptic picture of the ice conditions in order to put local measurements in a regional context, and (2) to develop understanding of electromagnetic wave interaction with ice and ocean surfaces during summer. A suite of airborne sensors included SAR, scatterometer, microwave radiometer, and aerial camera, all were onboard the CCRS (at that time) CV-580 aircraft. Radar altimeter, passive microwave imager, and aerial camera, all onboard NASA’s CV-990 aircraft, were also used in addition to an SLAR system, laser profilometer, and stepped frequency radiometer onboard NOAA P-3 platform; and another SLAR system onboard CNES B-17 platform. Satellite data were obtained from visible and infrared sensors of AVHRR, the night-time light operational line scan (OLS), Landsat MSS, and the microwave sensor SMMR. More ship- and surface-based remote sensing instruments were used, including passive microwave radiometers, scatterometer systems, and an instrument to measure dielectric constant. With such large and diversified suite of sensors, the location of the ice edge was monitored at high temporal frequency, thermal structure of the ocean off the ice edge was examined along with ice/ocean eddies at 1 km resolution, maps of mesoscale ice concentration and ice edge structure were produced, and floe size distribution was estimated. Imaging radar and helicopter-based scatterometer data show a distinct change in backscatter across the front of the ice edge caused by the damping of capillary waves in the meltwater area. All these were novel remote sensing observations at that time, which laid foundations for more observations to follow using advanced remote sensing tools. MIZEX’87 conducted in the Fram Strait and Greenland Sea in March–April timeframe included surfacebased active and passive microwave measurements made in conjunction with ice properties measurements from distinct ice types. More ships from Norway and Germany participated. Activities and initial results from MIZEX were published in a series of bulletins, which can be accessed through the Old Region Research and Engineering Laboratory (CRREL) archival system. Results of the MIZEX field studies were published in three special issues of the Journal of Geophysical Research: Ocean in 1983, 1987, and 1991. The CEAREX, funded by ONR, was carried out in the East Greenland Sea west of Svalbard from September

276

SEA ICE

1988 to June 1989 (freezing season). In addition to Norway, USA, Canada, Denmark, and France participated in the experiment. The primary objectives included understanding the processes regulating exchange of momentum, heat and biomass in the recurrent MIZ in the eastern Arctic at mesoscale (order 10 km) and submesoscale. One component was about sea ice behavior and associated acoustic ambient noise caused by the movement of the ice. The measurements contributed to understanding the relationship between stresses in individual ice floes and the geophysical driving forces and to identify noise generated by different processes. Another component was about ice drift, which focused on the dark winter period to observe atmosphere, ice, and ocean behavior simultaneously [CEAREX Drift Group,1990]. Data from CEAREX are available for download via the National Snow and Ice Data Center (NSIDC) site upon registration. Data include sections on sea ice, meteorology, oceanography, hydrology, biophysics, bathymetry, and position. This coordinated dataset became a source for understanding the interactions among ice, ocean, and atmosphere. However, results were not presented in many journal publications. A program called MIZOPEX was carried out between 21 July and 9 August 2013, in a recurrent MIZ area north of Alaska. It was conceived in response to NASA’s request for research effort to fill gaps of information about parameters/processes in MIZ, and to advance applications of their unmanned airborne systems (UASs) [Maslanik et al., 2016]. It took advantage of the capabilities of multiple classes of NASA’s UASs along with satellite observations and limited in-situ measurements. Other sensors used included infrared pyrometers, broadband and hyperspectral sensors, SAR, lidar, and electro-optical cameras. Scientific research issues addressed in the MIZOPEX included: (1) relationships between ocean skin temperatures and subsurface temperatures, and how these evolve over time in an Arctic environment during summer, (2) variability in sea ice conditions such as thickness, age, and albedo, (3) interactions between sea surface temperature (SST), salinity, and ice conditions during the melt cycle, and (4) validation of satellite-derived SST and ice concentration using field data. The program was exclusively American with participation from NASA, US Department of Energy, US Air Force, the NOAA, and the Federal Aviation Administration. Measurements during the program included skin temperature, spectral reflectance and albedo, sea ice freeboard and roughness, ice melt pond characteristics, atmospheric state variables, and wave height. While the program was mainly about data collection and testing UASs (no funding for data analysis), some work was done with the data though mostly on oceanic and atmospheric radiation parameters

(not so much on sea ice). UASs data confirmed the complex relationship between ice floes, meltwater and disruption of the mixed layer by drifting floes. Some applications of the data were developed for validation of satellite-retrieved and model-derived parameters. In response to the retreating MIZ in the Beaufort Sea during the past few decades, the ONR had launched a 5-year program, which started in late 2012 called “Emerging Dynamics of the Marginal Ice Zone,” and designated MIZ-DRI (Departmental Research Initiative). It built upon the information obtained from previous MIZ experiments. The program aimed at addressing a few scientific questions that included: (1) What are the processes that govern the spatial and temporal evolution of the MIZ? (2) What are the roles of ocean wave, solar radiation and extra heat release in governing the MIZ evolution and how do these processes couple? (3) What are the current modeling capabilities that predict the MIZ and how can it be improved? (4) What is the ice floe response to ocean wave; and what are the short-scale flexural variations across the floe? The planning phase of the experiment is described, in detail, in Lee et al. [2012]. Based on the above-mentioned scientific questions, the overall objectives of MIZ-DRI, as defined in Lee et al. [2012] were: (1) studying the key processes that control the evolution of the MIZ with sufficient spatial and temporal scope to the range of environmental influences; (2) identifying key interactions and feedbacks in the ice– ocean–atmosphere system, and predicting their changes with the increased seasonality on the MIZ, (3) evaluating the existing models that predict the seasonal evolution of the MIZ, and improving the parameterizations of key processes needed to improve the models. Some science objectives related to sea ice included: (1) understanding the physics of sea ice breakup, (2) characterizing changes in physics of the MIZ associated with changes in open water concentrations, (3) exploring feedbacks in the ice– ocean–atmosphere system that contribute to further decline of sea ice in the MIZ, and (4) collect a benchmark dataset for refining and testing models. Complete information about the MIZ-DRI, along with annual reports, is available at the University of Washington Applied Physics Laboratory project website. The main field experiment occurred between March and October 2014, in the Beaufort Sea north of Alaska. The observational component from the research cruise on the USA RV Sikuliaq during the fall of 2015 is presented in Thomson [2015]. Using data from the same cruise, Wadhams et al. [2018] found vast fields of pancake ice at the advancing ice edge as summer ended. Using SAR images from Sentinel-1 and COSMO-SkyMed (CSK), the authors applied an SAR-waves technique to estimate the thickness of frazil-pancake ice fields.

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

277

6.9. ICE EXERCISE BY US NAVY Ice Exercise (ICEX) is a biennial exercise, conducted north of Prudhoe Bay, Alaska. It is a United States Navy mission in the Arctic Ocean. The lead organization is the Navy’s San Diego-based Arctic Submarine Laboratory (ASL). A core component of each exercise is an ice camp, which accommodates research and logistic-support teams. The camp is usually established on a drifting MYI floe to ensure its safety against possible ice cracking, but it should also be adjacent to leveled FYI to allow supply aircraft landing and takeoff. Selection of the site is performed using satellite and airborne data. In-situ measurements are carried out by members from research institutions such as the Cold Region Research and Engineering Laboratory (CRREL) and University of Alaska. Measurements are usually conducted on a variety of surrounding ice types including MYI, deformed and leveled FYI and new ice. In some years the ICEX site was surveyed by NASA’s Operation IceBridge (OIB) mission. ICEX aims to offer the US Navy submarines a platform to assess their operational readiness and provide training to other services from partner nations to increase their experience in ice-covered regions. With that goal, ICEX promotes interoperability between allies to maintain Arctic operational readiness in terms of testing combat and weapon systems, communication and navigation systems, and sonar systems. Obviously, it is a defense-related rather than science-related mission. However, a science component is provided to develop support to the military exercise and produce limited scientific data about Arctic ice, e.g., thickness, mechanical behavior, and other icerelated environmental information. Arctic under-ice submarine operations have been conducted since 1947 in support of inter-fleet transit, training, cooperative allied engagements and routine operations. At that time submarines were diesel-operated and had to surface from time to time for air. In 1958, the first nuclear-powered US submarine Nautilus was introduced. This kind of submarines can stay under water for as long as there is enough food for the crew. Nautilus crossed the Arctic Ocean under the pack ice for the first time. With such a long history of operations in the Arctic ice-covered region, surfacing through the ice is still a tricky maneuver. It requires going up (vertically) at zero lateral speed. Sea ice is concealment for submarines, but they have to punch through it to communicate, launch missiles or support surface activities. Figure 6.33 is an aerial photograph of the USS submarine Hampton punching through Arctic sea ice during ICEX 2014 (March). A synopsis on the three recent ICEX campaigns is provided here. The 2016 ICEX was conducted over a fiveweek period during March/April with 200 participants

Figure 6.33 The Los Angeles-class attack submarine USS Hampton breaks thin ice to surface in the Arctic Ocean during the 2014 ICEX campaign (credit: US Navy, courtesy Hamilton Ingalls Industries, Chris Oxley).

from the United States, Canada, United Kingdom, and Norway. The ASL acted as a liaison between the civilian science community and submarine operations. An ice camp (called Sargo) was established and two submarines conducted Arctic transits training exercises including surfacing the ice. ICEX 2018 was also a five-week exercise with participation of two US submarines and one British submarine, both surfaced north of Alaska. The submarines executed 14-day tactical development exercises. The ice camp Skate was established and 50 persons participated. More quantitative data were collected using instruments from the University of Alaska and images from RADARSAT to track ice floes. CRREL conducted research to characterize ice–ocean properties and deployed an autonomous ice mass balance buoy and field ice stress sensors. Japan Agency for Marine Earth and Technology deployed buoys to obtain real-time information about sea ice drift, ocean temperature, and salinity as well as under-ice images during sea ice melt season. The US submarines collected water samples to contribute to the scientific data collection. This included salinity, nutrients, Oxygen-18 (18O), ice draft, bottom sounder data of bathymetry, and profiles of temperature and conductivity. ICEX 2020 was a 3-week exercise, which involved five nations, two submarines, and more than 100 participants over the three weeks of operations. An ice camp, called Seadragon, was established to house 45 personnel on MYI floes. In ICEX 2020, the International Arctic Research Center of the University of Alaska offered scientific support to the mission by providing information on wind, temperature, ice breaking, blowing snow, and other indicators on how the ice responded to weather and ocean conditions. Ice conditions were monitored using satellite imagery, GPS trackers and weather simulations. Some research work was conducted, jointly with

278

SEA ICE

CRREL team, to measure mechanical properties of sea ice and assess the potential for ice breakup. Radar instruments were installed to determine ice motion on scales that varied from a few centimeters to several kilometers. Strain in the ice was measured using a laser strain observer. Interferometric Ku-band radar was used to create 3D elevation maps of ice in the area and identify small fractions down to the wavelength of the radar. Since ICEX is mainly an operation-driven program and the accompanied scientific tasks aim basically to support the operational goal as mentioned above, there have been very few journal publications and white papers on the program, mostly published in the Journal of Geophysical Research. Readers can explore several websites containing information on individual ICEX campaigns. 6.10. THE MULTIDISCIPLINARY DRIFTING OBSERVATORY FOR THE STUDY OF ARCTIC CLIMATE (MOSAiC) MOSAiC was the largest scientific research expedition in the Arctic. The German icebreaker RV Polarstern was frozen into the Arctic Ocean near the North Pole and drifted with the ice for a year from September 2019 to September 2020. The RV Polarstern is a 12,000-ton ship with 20,000 horsepower engine, and can push through up to 2 m thick ice. It has nine laboratories, focused on everything from chemistry to ocean acoustics to fish biology. Figure 6.34 shows the ship during the polar night. Onboard the ship, and spread out in camps around the ship, were nearly 600 researchers collecting scientific data. They were affiliated with more than 80 institutions from 20 nations and representing dozens of scientific disciplines. The icebreaker started its voyage in September, the month of minimum ice extent, in order to cut as far

Figure 6.34 German RV Polarstern frozen in Arctic ice during MOSAiC research expedition (courtesy of Gunnar Spreen, University of Bremen).

north as possible to areas near the North Pole where thick ice was expected. However, thin ice prevailed the central Arctic at that time, and the selection of a thick ice floe to moor the ship and set up the research camp was a challenging task. Finally, a 2 × 3 km2 FYI floe was found with variable thickness that reached 3.8 m at some ridged areas. More description of the MOSAiC expedition can be found in Shupe et al. [2020]. MOSAiC was a scientific research mission dedicated to studying the Arctic climate system with a focus on ocean– ice–atmosphere interaction. Unlike previous expedition, which focused on specific science questions, MOSAiC tackled the most comprehensive question of Arctic amplification in relation to the Arctic system; namely the cause of the higher rate of Arctic warming which is more than double the global average. To address this question, all components of the system were studied simultaneously; ocean, ice, atmosphere, radiation, clouds, ecosystem, chemistry, and marine biology. In addition, the full year mission linked the four seasons and their transitions in the central Arctic, an area never visited in any scientific research expedition before. The study also aimed at improving the representation of all Arctic processes in global climate models. MOSAiC costed about US $150million, half of the budget was contributed by the German Federal Ministry of Education and Research. Aside from the unprecedented size of this field program in terms of personnel, diversity of study discipline, allocated resources, and the duration of the field study, what made it also unique is the number and modernity of the measurement instruments. Several new/or improved instruments were used. Progress of our understanding of physical processes hinges upon accurate measurements of interacting parameters. This progress later feeds into improved modeling. A stipulation of the MOSAiC project is that all the garnered data will be free and open to the public for generations. The rest of this section presents highlight of the planned work on sea ice and remote sensing as described in the expedition program published by the Alfred-Wegener Institute in October 2019 (https://epic.awi.de/id/eprint/ 50082/1/Expeditionsprogramm_PS122_leg2.pdf). The objectives of the sea ice component included: (1) characterization of ice and snow properties with their spatial and temporal variability, (2) determination of mass and energy balances of snow and ice, (3) description of variability of ice thermodynamics and dynamics on regional scale, (4) integration of sea ice measurements with other components at multiple scales. The objectives of the remote sensing component included: (1) ground truthing of existing satellite data with in-situ observations, (2) collecting in-situ data to support development of ice and snow parameters retrieval methods, (3) collecting and archiving a comprehensive satellite remote sensing data set covering the entire experiment period,

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE

(4) supporting operations in the field (e.g., navigation and selection of measurement sites. Another goal was to use satellite remote sensing to extend observations to larger scale and relate them to other environmental conditions. Airborne remote sensing was used to bridge gaps between satellite observations and in-situ measurements. Remote sensing instruments encompassed ship- and surface-based, satellite-borne and airborne instruments. The ship- and surface-based instruments included a Ku-/Ka-band radar; L-, C-, X-, and Ku-band scatterometer, five MW radiometers, and three cameras (infrared, video, and hyperspectral). Figure 6.35 shows a suite of those instruments on the ground. The satellite instruments included three SAR sensors (Sentinel-1, TerraSAR-X, and Cosmo-SkyMED), two optical sensors (AVHRR and MODIS), two passive microwave sensors (AMSR2 and SSMIS), a scatterometer (ASCAT), and an altimeter (CryoSat-2). Airborne photography and laser scanning were used to collect surface data along transects in order to detect surface conditions such as ice types, deformation and melt ponds. Data were also used to up-scale in-situ measurements from floe scales to regional scales. Helicopter-borne electromagnetic system (EM-Bird) was used to survey ice thickness along certain transects. So far, results from MOSAiC have appeared in a few publications [Krumpen et al., 2020, Stroeve et al., 2020, Munoz-Martin et al., 2020, Katlein et al., 2020]. More are expected to appear in the future. A comprehensive

IR Camera

summary of snow and ice work of MOSAiC is presented in Nicolaus et al. [2022]. Highlight of some relevant results are presented here. Krumpen et al. [2020] investigated the initial conditions of the ice at the start of the experiment and found that it was exceptionally thin compared to the conditions in the past 26 years. The authors argued that this may be the new normal of the Arctic conditions and that would make future follow-up campaigns of this scale increasingly difficult. However, the important issue, if this projected scenario is realized, would be the impact of thinner, perennial-ice-free cover on Arctic and global climate. Observations like this should be confirmed at larger scale in order to guide modeling studies. It has been long recognized by the remote sensing community that snow influences observations and hinders retrieval of ice parameters from all types of remote sensing instruments; optical, infrared, and microwave. Stroeve et al. [2020] investigated the potential for combining Ku- and Kaband scatterometer data to estimate both snow depth and ice thickness (Ku and Ka wavelengths are 7.5–11.1 mm, and 1.65–2.5 cm, respectively). The two bands were combined in a surface-based scatterometer/ altimeter instrument, referred to as KuKa radar. Determining snow depth should improve ice thickness retrieval from altimeters (section 8.6). The instrument was used to collect observations from the start of freeze-up throughout the spring, and from FYI and SYI floes as well as new ice in leads. Stroeve et al. [2020] describe the

L-Scat Radar

ELBARA Radiometer SSMI Radiometer

279

Ku/Ka Radar

HUTRAD Radiometer

GNSS-R

Snow Sampling

Figure 6.35 Ground-based remote sensing instruments used in MOSAiC experiment (courtesy of Gunnar Spreen, University of Bremen).

280

SEA ICE

operation of the instrument to estimate snow depth and a few snow parameters triggered by weather conditions. This will assist in characterizing radar backscatter response to changes in atmospheric and ice geophysical conditions. Munoz-Martin et al. [2020] used the Global Navigation Satellite System Reflectometer (GNSS-R) combined with an L-band radiometer to estimate the ice thickness. The purpose was to validate results from the same instrument currently onboard ESA’s Sentinel Small Satellite (S^3). Data collected by the surface-based GNSS-R for different frequencies showed that sea ice thickness and snow depth can be retrieved using multi-angular and multi-frequency observations. The preliminary results from this instrument are yet to be validated in different sites and in combination with other instruments. Katlein et al. [2020] observed and studied platelet ice in the Arctic during the midwinter leg of the MOSAiC expedition. As mentioned in section 5.3.3.8, platelet ice is a term coined with Antarctic sea ice and it consists of frazil crystals originated in supercooled sea water under-ice shelves. While ascending toward the surface, driven by buoyant forces, the crystals accumulate at the bottom of the ice and become part of the ice column. Platelet ice has not been extensively observed in the Arctic perhaps because of the limited number of ice shelves compared to the Antarctic. Katlein et al. [2020] used remotely operated underwater vehicle to observe the growth of the platelet ice layers, and concluded that its formation was widespread in Arctic winter though it has been overlooked. More studies are needed to confirm this observation and clarify the difference between platelet ice layer and frazil when both stick to the bottom of the sea ice cover. Platelet ice crystals are, in fact, frazil crystals with their typical needle-shape. They both interrupt the crystalline growth habit (e.g., columnar growth) but frazil crystals may act as seeds for subsequent ice growth. 6.11. REFERENCES Barber, D.G. and LeDrew, E.F. (1994) On the links between microwave and solar wavelength interactions with snowcovered first-year sea ice, Arctic, 47(3), pp. 298–309. Barber, D.G. and Nghiem, S.V. (1999) The role of snow on the thermal dependence of microwave backscatter of sea ice, Journal of Geophysical Research, 104(C11), pp. 25,789–25,803. Available from: doi:10.1029/1999JC900181. Barber, D.G. et al. (1992) Spatial and temporal variation in sea ice geophysical properties and microwave remote sensing data: The SIMS’90 experiment, Arctic, 45(3), pp. 233–251. Barrette, P.D. and Sinha, N.K. (1996) Crystallographic characterization of a core from the Ward Hunt ice shelf, Canada, Proceedings of International Symposium on Snow and Related Manifestations, Manali, India, September 1994, pp. 114–124. Bjerkelund, C.A. et al. (1985) The texture and fabric of the second year sea ice cover at Mould Bay, Prince Patrick Island,

NWT, April 1983, Proceedings of International Geoscience and Remote Sensing Symposium, 1, pp. 426–431. Campbell, W.J. et al. (1978) Microwave remote sensing of sea ice in the AIDJEX main experiment, Boundary Layer Meteorology, 13, pp. 309–337. Carsey, F.D. et al. (1986) Weddell-Scotia Sea marginal ice zone observations from space, October, 1984, Journal of Geophysical Research, 91, pp. 3920–3924. Carsey, F.D. et al. (1989) Overview of LIMEX’87 ice observations, IEEE Transactions of Geoscience and Remote Sensing, 27(5), pp. 468–482. Cavalieri, D.J. et al. (1983) MIZEX West: Bering Sea marginal ice zone experiment, Eos, Transactions, American Geophysical Union, 64(40), pp. 578–579. CEAREX Drift Group (1990) CEAREX drift experiment, Eos, Transactions, American Geophysical Union, 71(40), pp. 1115–1118. Coon, M. et al. (2007) Arctic Ice Dynamics Joint Experiment (AIDJEX) assumptions revisited and found inadequate, Journal of Geophysical Research, 112(C11S90). Available from: doi:10.1029/2005JC003393. Digby, S.A. (1984) Remote sensing of drained ice areas around the breathing holes of ice-inhabiting seals, Canadian Journal of Zoology, 62, pp.1011–1014. Drinkwater, M.R. (1989) LIMEX ‘87 ice surface characteristics: Implications for C-Band SAR backscatter signatures, IEEE Transactions of Geoscience and Remote Sensing, 27(5), pp. 501–513. Frederking, R.M.W. and Sinha, N.K. (1987) Technical properties of shelf ice, Proceedings of Workshop on Extreme Ice Features, Banff, Alberta, Canada, 3–6 November 1986, National Research Council of Canada, Technical Memorandum No. 141, pp. 67–78. Frederking, R., Jordaan, I.J. and McCallum, J.S. (1990) Field tests of ice indentation at medium scale Hobson’s Choice Ice Island, 1989, Proceedings of IAHR International Symposium on Ice, Espoo, Finland, Vol. 2, pp. 931–944. Gagnon, R.E. and Sinha, N.K. (1991) Energy dissipation through melting in large scale indentation experiments on multi-year sea ice, In: Ayorinde, O.A. et al., eds. Proceedings of the 10th International Conference on Offshore Mechanics and Arctic Engineering, Stavanger, Norway, Vol. 4. Arctic Polar Technology. New York, American Society of Mechanical Engineers, pp. 157–161. Gascard, J.C. and Clarke, R.A. (1983) The formation of Labrador sea water: Mesoscale and smaller-scale processes, Journal of Physical Oceanography, 13(10), pp. 1779–1797. Grenfell, T.C. and Lohanick, A.L. (1985) Temporal variations of the microwave signatures of sea ice during the late spring and early summer near Mould Bay, Northwest Territories, Journal of Geophysical Research, 90(C3), pp. 5063–5074. Haggerty, J.A., Maslanik, J.A. and Curry, J.A. (2003) Heterogeneity of sea ice surface temperature at SHEBA from aircraft measurements, Journal of Geophysical Research, 108(C10), p. 8052. Available from: doi:10.1029/2000JC000560. Holt, B. and Digby, S.A. (1985) Processes and imagery of firstyear fast sea ice during the melt season, Journal of Geophysical Research, 90(C3), pp. 5045–5062. Hollinger, J.P. et al. (1984) Microwave emission from High Arctic sea ice during freeze-up, Journal of Geophysical Research, 89(C5), pp. 8104–8122.

MAJOR FIELD EXPEDITIONS TO STUDY SEA ICE Jeffries, M.O. and Sackinger, W.M. (1989) Analysis and interpretation of an airborne synthetic aperture radar image of Hobson’s Choice Ice Island, In: Axelsson, K.B.E. and Fransson, L.A., eds. Port and Ocean Engineering under Arctic Conditions (POAC), 12–16 June 1989, Lulel, Sweden, Vol.2, pp.1032–1041. Jeffries, M.O. and Sackinger, W.M. (2012) Ice island detection and characterization with airborne synthetic aperture radar, Journal of Geophysical Research, 95(C4), pp. 5371–5377. Jeffries, M.O., Sackinger, W.M. and Shoemaker, H.D. (1988) Geometry and physical properties of ice islands, In: Sackinger, W.M. and Jeffries, M.O., eds., Port and Ocean Engineering under Arctic Conditions (POAC), Geophysical Institute, University of Alaska Fairbanks, USA, Vol.1, pp.69–83. Johannessen, O.M. (1987) Introduction: Summer marginal ice zone experiment during 1983 and 1984 in Fram Strait and Greenland, Journal of Geophysical Research, 92(C7), pp. 6716–6718. Johansson, A.M. et al. (2017) Combined observations of Arctic sea ice with near-coincident colocated X-band, C-band, and L-band SAR satellite remote sensing and helicopter-borne measurements, Journal of Geophysical Research: Oceans, 122, pp. 669–691. Jordaan, I.J. et al. (1992) Analysis of ice failure processes, Final Report to Natural Sciences and Engineering Research Council (NSERC) on Collaborative Research and Development Grant, No. 661-078/89, Faculty of Engineering and Applied Science, Memorial University of Newfoundland, St. John’s, Canada. Katlein, C., et al. (2020) Platelet ice under Arctic pack ice in winter, Geophysical Research Letters, 47(16), e2020GL088898. Available from: http://doi.org/10.1029/2020GL088898. Key, J.R. and Haefliger, M. (1992) Arctic ice surface temperature retrieval from AVHRR thermal channels, Journal of Geophysical Research, 97(D5), pp. 5885–5893. Key, J. et al. (1994) On the validation of satellite-derived sea ice surface temperature, Arctic, 47(3), pp. 280–287. Kim, Y.S. et al. (1985) Towards identification of optimum radar parameters for sea ice monitoring, Journal of Glaciology, 31(109), pp. 214–219. King, J. et al. (2018) Comparison of freeboard retrieval and ice thickness calculation from ALS, ASIRAS, and CryoSat-2 in the Norwegian Arctic to field measurements made during the N-ICE2015 Expedition, Journal of Geophysical Research: Ocean, 123. Available from: https://doi.org/10.1002/ 2017JC013233 (open access). Krumpen, T. et al. (2020) The MOSAiC ice floe: Sediment-laden survivor from the Siberian shelf, The Cryosphere, 14, pp. 2173–2187. LeDrew, E.F. and Barber, D.G. (1994) The SIMMS program: A study of change and variability within the marine cryosphere, Arctic, 47(3), pp. 256–264. Lee, G.M. et al. (2012) Marginal Ice Zone (MIZ) program: Science and experimental plan, Technical Report APL-UW 1201, Applied Physics Laboratory, University of Washington, Seattle, Washington. Lu, M. (1988) Analysis of ice island movement, M.Sc. thesis, University of Alaska, Fairbanks. Maslanik, J.A. et al. (2016) Investigations of spatial and temporal variability of ocean and ice conditions in and near the

281

marginal ice zone: The Marginal Ice Zone Observations and Processes Experiment (MIZOPEX) final campaign summary, Technical Report, US Department of Energy, DOE/SCARM-15-046. McNutt, S.L. and Overland, J.E. (2003) Spatial hierarchy in Arctic sea ice dynamics, Tellus A, 55, pp. 181–191. McNutt, L. et al. (1988) LIMEX’87: The Labrador Ice Margin Experiment, March 1987; A pilot experiment in anticipation of RADARSAT and ERS-1 data, Eos, Transactions, American Geophysical Union, 69(23), pp. 634–643. McPhee, M.G. (1983) Greenland Sea ice/ocean margin, Eos, Transactions, American Geophysical Union, 64(9), pp. 82–83. Available from: doi:10.1029/EO064i009p00082. Meaney, R., Kenny, S. and Sinha, N.K. (1992) Medium-scale ice-structure interaction: Failure zone characterization, Musk-Ox, 39, pp. 24–30. MIZEX Group (1986) MIZEX East 83/84: The summer marginal ice zone program in the Fram Strait/Greenland Sea, Eos, Transactions, American Geophysical Union, 67(23), pp. 513–517. Munoz-Martin, J.F. et al. (2020) Snow and ice thickness retrievals using GNSS-R: Preliminary results of the MOSAiC experiment, Remote Sensing, 12(24), p. 4038. Available from: https://doi.org/10.3390/rs12244038. Nicolaus, M. et al. (2022) Overview of MOSAiC expedition snow and sea ice, Elementa Science of Anthropocene, 10(1). Available from: https://doi.org/10.1525/elementa.2021.000045. Notz, D., Wettlaufer, J.S. and Worster, M.G. (2005) A nondestructive method for measuring the salinity and solid fraction of growing sea ice in situ, Journal of Glaciology, 51(172), pp. 159–166. Oikkonen, A. et al. (2017) Small-scale sea ice deformation during N-IVE2015: from compact ice to marginal ice zone, Journal Geophysical Research: Oceans, 122, pp. 5105–5120. Perovich, D.K. et al. (2002) Seasonal evolution of the albedo of multiyear Arctic sea ice, Journal of Geophysical Research: Oceans, 107(C10), 8044. Available from: doi:10.1029/ 2000JC000438. Perovich, D.K. et al. (2003) Thin and thinner: Sea ice mass balance measurements during SHEBA, Journal of Geophysical Research: Oceans, 108(C3), p. 8050. Available from: doi:10.1029/2001JC001079. Prinsenberg, S.J., Peterson, I.K. (1992) Sea-ice properties off Labrador and Newfoundland during LIMEX ‘89, AtmosphereOcean, 30(2), pp. 207–222. Pritchard, R., ed. (1980) Sea ice processes and models, Proceedings of the Arctic Ice Dynamics Joint Experiment, International Commission on Snow and Ice Symposium, University of Washington Press, Seattle. Pritchard, R.S., Coon, M.D. and McPhee, M.G. (1977) Simulation of sea ice dynamics during AIDJEX, Journal of Pressure Vessel Technology, 99(3), pp. 491–497. Provost, C. et al. (2017) Observations of flooding and snow-ice formation in a thinner Arctic sea-ice regime during the NICE2015 campaign: Influence of basal ice melt and storms, Journal of Geophysical Research: Oceans, 122, pp. 7115–7134. Raney, R.K., Digby-Argus, S. and McNutt, L. (1989) Labrador Ice Margin Experiment LIMEX’89—An overview, Proceedings of International Geoscience and Remote Sensing Symposium (IGARSS’89), Vancouver, Canada, pp. 1517–1519.

282

SEA ICE

Rösel, A. et al. (2017) Can we extend local sea-ice measurements to satellite scale? An example from the N-ICE2015 expedition, Annals of Glaciology, 59(76pt2), pp. 1–10. Rösel, A. et al. (2018) Thin sea ice, thick snow and widespread negative freeboard observed during N-ICE2015 north of Svalbard, Journal of Geophysical Research: Oceans, 123, pp. 1156–1176. Rostosky, P. et al. (2020) Modeling of the microwave emission of snow on Arctic sea ice for estimating the uncertainty of satellite retrievals, Journal of Geophysical Research: Oceans, 125(3), e2019jc015465. Serreze, M.C., Maslanik, J.A. and Key, J.R. (2002) Atmospheric and sea ice characteristics of the Arctic Ocean and the SHEBA field region in the Beaufort Sea, National Snow and Ice Data Center (NSIDC), Special Report 4. Available from: https://nsidc.org/pubs/special/4/index.html. Shokr, M., Asmus, K. and Agnew, T.A. (2009) Microwave emission observations from artificial thin sea ice: The ice-tank experiment, IEEE Transactions of Geoscience and Remote Sensing, 47(1), pp. 325–338. Shokr, M. et al. (2022) Observations from C-band SAR fullypolarimetric parameters of mobile sea ice based on radar scattering mechanisms to support operational sea ice monitoring, Canadian Journal of Remote Sensing, 48(2), pp. 197–213. Shupe, M.D. et al. (2020) The MOSAiC expedition: A year drifting with the Arctic sea ice, Arctic Report Card 2020, National Oceanic and Atmospheric, Administration. Available from: https://arctic.noaa.gov/Report-Card/Report-Card-2020. Sinha, N.K. (1984) Uniaxial compressive strength of first-year and multi-year sea ice, Canadian Journal of Civil Engineering, 11(1), pp. 82–91. Sinha, N.K. (1985) Confined strength and deformation of second-year columnar-grained ice in Mould Bay, Proceedings of 4th International Symposium Offshore Mechanics and Arctic Engineering (OMAE/ASME), Dallas, Texas, USA, 10– 15 February, 1985, Vol. 2, pp. 209–219. Sinha, N.K. (1987) The borehole jack—Is it a useful Arctic tool?, Journal of Offshore Mechanics and Arctic Engineering, Transactions of the ASME, 109(4), pp. 391–397. Sinha, N.K. (1991) Microstructure and mechanical behavior of ice, In: Dodhi, D.S., ed., Proceedings of 6th International

Speciality Conference, Cold Region Engineering, West Lebanon, NH, USA, American Society of Civil Engineers (ASCE), New York, pp. 519–530. Sinha N.K., Strandberg, A. and Vij, K.K. (1986) In situ assessment of drilling platform sea ice strength using a borehole jack, Proceedings of In-situ Testing and Field Behaviour, 39th Canadian Geotechnical Conference, 27–29 August 1986, Ottawa, Ontario, Canada, Canadian Geotechnical Society, pp. 153–157. Shupe, M.D. et al. (2020) The MOSAiC expedition: A year drifting with the Arctic sea ice, Arctic Report Card 2020, National Oceanic and Atmospheric Administration. Available from: https://arctic.noaa.gov/Report-Card/Report-Card-2020. Stroeve, J., et al. (2020) Surface-based Ku- and Ka-band polarimetric radar for sea ice studies, The Cryosphere, pp. 4405–4426. Sturm, M., Holmgren, J. and Perovich, D.K. (2002) Winter snow cover on the sea ice of the Arctic Ocean at the Surface Heat Budget of the Arctic Ocean (SHEBA): Temporal evolution and spatial variability, Journal of Geophysical Research, 107(10), p. 8047. Available from: doi:10.1029/2000JC000400. Tang, C.L. and Ikeda, M. (1989) Ice edge upwelling off the Newfoundland coast during LIMEX, Atmosphere-Ocean, 27(4), pp. 658–681. Thomson, J. (2015) ONR Sea State DRI cruise report: R/V Sikuliaq, Fall 2015 865 (SKQ201512S) (Technical report). Seattle, WA: University of Washington. Thomas, R.H. (1979) The dynamics of marine ice sheet, Journal of Glaciology, 24, pp. 273–286. Untersteiner, N. (1980) AIDJEX review, In: Pritchard, ed. Sea ice processes and models, Seattle, London: University of Washington Press. Untersteiner, N. et al. (2007) AIDJEX revisited: A look back at the US-Canadian Arctic Ice Dynamics Joint Experiment 1970–78, Arctic, 60(3), pp. 327–336. Uttal, T., et al. (2002) Surface heat budget of Arctic ocean, Bulletin of American Meteorological Society, 83(2), pp. 255–276. Wadhams, P. et al. (2018) Pancake ice thickness mapping in the Beaufort Sea from wave dispersion observed in SR imagery, Journal of Geophysical Research: Oceans, 123(C3), pp. 2213–2237.

7 Remote Sensing Fundamentals Relevant to Sea Ice

7.1

General Principles of Satellite Remote Sensing .................... 284

7.2

Electromagnetic Wave Properties and Processes .................. 289 7.2.1 Polarization of EM Wave.......................................... 290 7.2.2 Reflection, Transmission, Absorption, Scattering, and Emission .................................................................... 292 7.2.2.1 Reflection and Fresnel Model......................293 7.2.2.2 Transmission................................................295 7.2.2.3 Absorption and Scattering Losses................296 7.2.2.4 Emitted Radiation (Re-radiation)................296

7.6.3.2

7.7

Polarimetric Parameters Derived from the FP SAR Data ..............................................317 7.6.3.3 Linking Radar Scattering Mechanisms to Ice Features .................................................320 7.6.3.4 Age-Based versus SAR-Based and ScatteringBased Sea Ice Classification........................... 321 Scatterometer Systems .......................................................... 322

7.8

Altimeter Systems ................................................................. 323

7.9

Radiative Processes in Relevant Media ................................ 325 7.9.1 Atmospheric Influences.............................................. 325 7.9.1.1 Influences of Atmosphere on Optical and Infrared Observations ..................................325 7.9.1.2 Atmospheric Correction for Passive Microwave Observations .............................328 7.9.2 Seawater .................................................................... 330 7.9.2.1 Seawater in the Optical and Thermal Infrared Data...............................................330 7.9.2.2 Seawater in the Microwave Data.................331

7.2.3 Brightness Temperature and Emissivity..................... 297 7.2.4 Penetration Depth...................................................... 299 7.3

Optical Sensing ..................................................................... 300

7.4 7.5

Thermal Infrared Sensing ..................................................... 303 Microwave Remote Sensing.................................................. 305

7.6

Imaging Radar Sensing......................................................... 308 7.6.1 Imaging Radar Principles .......................................... 308 7.6.1.1 Radar Equations and Spatial Resolutions of RAR and SAR ........................................309 7.6.1.2 Coherency and Polarization of Radar Signals .........................................................311 7.6.1.3 Radar Scattering Mechanisms .....................312 7.6.2 Multichannel SAR..................................................... 313 7.6.3 SAR Polarimetry: Formulation and Derived Parameters.................................................................... 315 7.6.3.1 Formulation of Polarimetric Measurements ..316

Space-borne remote sensing refers to the recording of reflected, scattered, or emitted radiation by the earth’s surface and atmosphere using sensors onboard satellite platforms. The observations are usually presented in the form of imagery data representing measurements from individual pixels called sensor’s footprint within the imaged area. A few sensors, however, provide measurements along the satellite track on the ground only. Processing the data leads to retrieval of geophysical and geometrical properties of the imaged area. Remote sensing is different from remote sounding, which measures the properties of the atmospheric column between the sensor and the ground. Space-borne remote sensing is commonly referred to in the literature as “earth

7.10

7.9.3 Snow on Sea Ice: Physical and Radiative Processes .................................................................... 333 7.9.3.1 Snow in Optical and Thermal Infrared Data.............................................................335 7.9.3.2 Snow in the Microwave Data ......................336 References............................................................................. 341

observations from space,” or simply “earth observation” (EO). More than 100 EO satellites, developed by many national and commercial space agencies are orbiting the earth at this time. Sea ice applications use data from a few satellites, especially those carrying microwave sensors. The spatial resolution for operational ice monitoring range between a few meters or tens of meters (fine resolution), hundreds of meters (medium resolution) for tactical navigation, or a few kilometers (coarse resolution) for mesoscale monitoring. Operational sea ice monitoring agencies usually assess the trade-off between spatial resolution and the swath in order to preserve the information at the finest possible scale while minimizing the data volume to be processed within the allowable turnaround time.

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 283

284

SEA ICE

Remote sensing is the primary tool for monitoring and retrieving information about sea ice because most of it is located in the polar regions. Therefore, it was not unexpected to envision sea ice as a leading application of EO data, since its inception in the 1970s. In fact, one of the driving forces behind the development of the spaceborne synthetic aperture radar (SAR) and scatterometer systems was to monitor polar sea ice and seawater. Initially, SAR was important to support marine navigation and safety of marine structures for gas and mineral exploration. Since the late 1990s, Arctic sea ice monitoring has become more important from climatic viewpoint as it continues to shrink in extent and decrease in thickness, which are alarming signs of global warming. Sea ice data from satellite microwave sensors, especially passive microwave (PM), represent one of the longest records of EO. Data from optical sensors have also been used during daylight seasons but under cloud-free skies. They are particularly useful to identify the onset of ice melting. Scatterometer data are used to monitor the polar-wide ice conditions. Altimeter data from both radar and laser sensors have become essential tools to monitor polar ice thickness at synoptic scales. The prime sensor for mapping sea ice types, concentration, and surface features is the space-borne SAR. Today, it is impossible to imagine an operational ice monitoring and analysis program that does not depend heavily on satellite remote sensing. Sea ice can be easily discriminated from the surrounding open water in remote sensing observations by utilizing their contrasting radiometric signatures, which are triggered by their different physical properties. These properties include surface albedo (for optical sensors), physical temperature (for thermal infrared sensors), salinity, albedo, emissivity, dielectric constant, and surface roughness (for microwave sensors). Discriminating age-based (or equivalent thickness-based) ice type in data from any single sensor is challenging because of the overlap of observations from different ice types as explained later in Chapter 9. Remote sensing observations are triggered by the optical, thermal, and electrical properties of the earth’s cover within the penetration depth of incident or emitted radiation. In sea ice, this depth is usually limited to a few millimeters or centimeters, depending on the wavelength of the signal, and can be extended to tens of centimeters in dry snow or saline-free ice (e.g., freshwater ice or MYI). Information from areas below the penetration depth is not implied in the remote sensing observations. The only exception is the ice thickness estimate of thick ice, which is available from using altimeter systems because the analysis of the data from this system employs laws of buoyancy, not just reflected radiation. Because of this limitation of remote sensing data (except for the altimeter systems), ancillary data such as meteorological conditions, regional

climate, and recent history of the ice cover are usually used in operational ice monitoring centers to complement information from remote sensing imagery data and support retrieval of sea ice information. The core information in this chapter addresses a few fundaments of remote sensing relevant to sea ice monitoring and parameter retrieval. In order to appeal to readers seeking quick acquaintance with scientific issues that are outside their domain of experiences, the presentation favors the breadth rather than the depth of the information. This also conforms to the overall flavor of the book—to work across disciplinary boundaries in order to realize the promise of the integrated information. The reader will find a bias in the presentation toward microwave remote sensing although details of optical and thermal infrared (TIR) sensing are also presented. From climatic viewpoint, space-borne observations from PM and radar systems have been the most common sources for monitoring sea ice in the polar region on a large scale. From operational viewpoint, SAR systems are the prime data source used in mapping sea ice to support marine operations in ice-rich waters. The chapter starts by introducing basic definitions and principles of satellite remote sensing (section 7.1). An introduction to a few theoretical concepts of electromagnetic (EM) wave and its interaction with matter (namely, reflection, transmission, absorption, scattering, and emission) is included in section 7.2. Also presented in the same section is a brief explanation of the concepts of brightness temperature, emissivity, and penetration depth. Sections 7.3, 7.4, 7.5, and 7.6 present basic principles of the four categories of remote sensing: optical, thermal, PM, and radar, respectively. More emphasis is placed on radar polarimetry since its operational applications are emerging and its theoretical background has not yet been fully comprehended within the operational sea ice community. Section 7.7 and 7.8 cover theoretical background of scatterometer and altimetry systems, respectively. Remote sensing observations of sea ice are affected by the presence of three other media; atmosphere, seawater, and snow on ice. Section 7.9 addresses basic physical and radiative processes pertaining to the EM propagation in these media. One subsection is dedicated to each medium. This information furnishes necessary background to augment the material on the ice parameter retrieval presented in Chapters 10 and 11.

7.1. GENERAL PRINCIPLES OF SATELLITE REMOTE SENSING The three EM spectral regions that are used in the EO data applications are optical, TIR, and microwave. The spectral range of each region and its divisions into

Optical spectral range

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE • Visible region VIS (0.4–0.7 µm) – Blue (0.4–0.5 µm) – Green (0.5–0.6 µm) – Red (0.6–0.7 µm)

285

Visible spectrum

0.4

0.5

0.6

0.7

Wavelength, µm

• Near-infrared region (NIR)

0.7–1.3 µm

• Short-wave IR (SWIR)

1.3–3 µm

• Middle IR (Mid-lR)

3–8 µm

• Thermal infrared region (TIR)

8–15 µm

• Far infrared (FIR)

15–1000 µm

• Microwave region (MW)

> 1 mm

Reflected IR

Emitted radiation (modeled by black body radiation)

Figure 7.1 Categories of remote sensing based on the EM spectral bands.

subregions are shown in Figure 7.1. The optical region covers the three spectral bands: the visible (VIS), near IR (NIR), and shortwave IR (SWIR). Those are the bands found in the solar radiation before atmospheric absorption. Optical sensors are commonly used in sea ice applications, but they are constrained by the availability of sunlight and a cloud-free sky in order to be able to record the reflection from the surface. They detect solar reflection using radiometers, photographic, or digital cameras. Sensors operating in the NIR and SWIR bands detect the reflected signal using films with IR-sensitive emulsion spectroradiometers. Radiometers are also used in TIR sensors (8–15 μm wavelength) to detect thermal emission. Under cloud-free sky, TIR sensors can be used to determine the surface (skin) temperature of sea and ice surfaces. This may offer discrimination between the two surfaces. They can also be used to identify MYI because it is usually colder than the surrounding ice. Observations in the far IR region (FIR) are not used in sea ice applications. The microwave region occupies a wide range of wavelengths from 1 millimeter to 1 meter (300–0.3 GHz frequency). The longer wavelength range (1 m to 100 km), which is not shown in the figure, is known as radio wave. Microwave sensors in the millimeter and centimeter wavelength range are the prime sensors for sea ice monitoring. Microwave emission from all ground cover surfaces is weak, so it has to be integrated over a large area to make reasonably high signal strength compared to instrument’s noise. This means coarser resolution of PM observations. Microwave emission from seawater is considerably lower than sea ice, hence water is considered to be radiometrically colder than ice. This is the essential property that allows discrimination of seawater and ice in PM data, hence estimation of ice concentration.

The total amount of radiative flux emitted from the sun or the earth is determined from the Planck’s equation (7.21), assuming that the sun is a blackbody (definition of blackbody is presented in section 7.2.3). Blackbody radiation is a function of the physical temperature of the emitting body and the frequency of the radiation. The top graph of Figure 7.2 is a sketch depicting the spectral energy of emitted radiation from the sun (with temperature of 6000 K) and the earth (288 K). This is following Planck’s equation and assuming that both are blackbodies. Solar radiation peaks at wavelength of approximately 0.5 μm (blue-green light) in the VIS spectrum. This band offers the highest radiation level available for remote sensing detectors. It is also close to the center of the EM wavelengths that can be detected by human eyesight. On the other hand, the emitted radiation from the earth’s surface peaks near 10.0 μm wavelength, i.e., within the TIR range. Solar radiation starts to diminish in the TIR and becomes negligible in the microwave region. A portion of the solar energy that strikes the earth’s surface is reflected, and another portion is absorbed. The absorbed energy is re-emitted since all objects radiate energy at temperatures above the absolute zero (−273 C) by virtue of their atomic and molecular oscillations. After reflection or emission, part of the surface radiation is absorbed and scattered by the atmospheric constituents (gases and aerosols) in certain spectral bands. The atmosphere allows transmission in other bands called atmospheric transmission windows. They are located in the ranges 0.4–1.3 μm, 1.5–1.8 μm, 3.6–4.0 μm, 10.5– 12.3 μm, and throughout the wide range of the microwave spectrum from 10 mm to 10 cm (bottom of Figure 7.2). The spectral bands used in satellite remote sensing for surface observations fall into these transmission windows. Microwave sensors are successful in sea ice applications

TIR

Microwave

SEA ICE UV VISIBLE

286

FIR

Energy

Infrared solar radiation earth emission

0.3 μm

1.0 μm

10 μm

30 μm

1 mm

1m

1 mm

1m

Wavelength

Transmittance (%)

Atmospheric windows 100

Absorbed

0

0.3 μm

1.0 μm

10 μm

30 μm

Wavelength

Figure 7.2 Spectral emitted radiation by the sun and the earth (top) and the atmospheric windows for spectral radiation in the range 0.3 μm to 1 m (bottom). Outside the atmospheric windows, the radiation is partly or fully absorbed by various atmospheric constituents. The earth’s surface emits energy that peaks around 10 μm. Most of the radiation received by satellite sensors around this band is emitted from the surface.

because of their very low absorption by the atmosphere and clouds. The summation of energy incident upon a point on the earth’s surface from all directions above the surface (i.e., the hemisphere) is called irradiance. The total reflection or emission leaving any point at the surface is known as radiance. The unit of irradiance is power per unit area, while the unit of radiance is power per unit area per unit solid angle. If a point source (e.g., the sun) radiates uniformly in all directions through a non-absorptive medium, the irradiance decreases inversely with the square of the distance from the object. Remote sensing instruments operating in the optical bands measure the solar reflection off a surface or atmospheric constituent. The green band (wavelength around 0.51 μm) is particularly sensitive to sea ice regardless of its age or surface conditions. In general, optical remote sensing data are useful for sea ice applications because of their high sensitivity to water–ice boundaries and flooded water on an ice surface. Their major drawback, however, is the dependence on sunlight and the extinction by clouds and other atmospheric constituents. Thermal infrared sensors measure the emitted heat radiation from the surface or atmospheric constituents. Emitted energy in this spectral region is also absorbed (and re-emitted) by clouds and fog. Therefore, it is useful to combine VIS and TIR imagery data to identify clouds and remove their influences. Microwave radiation is emitted by the Earth’s surface in very small amounts. Therefore, the observed

signal should be integrated over a large footprint in order to increase the signal-to-noise ratio. This means the spatial resolution has to be coarse; a few kilometers or tens of kilometers. Microwave signals traveling in the atmosphere are not affected by clouds although the shorter wavelengths (< 20 mm) can be absorbed by raindrops in severe storms. Active microwave (radar) sensors generate their own illumination, hence the observations are independent of solar radiation. They are also not affected by presence of clouds of precipitation. For that reason, they are called all-time, all-weather sensors. The spatial resolution from radar sensors can be as fine as a few meters. However, for operational ice monitoring 100 m is a practically sufficient resolution. Based on the orbit type, EO satellites are grouped into two categories: polar orbiting and geostationary. Polar orbiting satellites orbit the earth at altitudes between 500 and 1,400 km. The orbit is typically inclined by 5 to 15 degrees to the vertical (Figure 7.3a), so they actually follow a near-polar orbit. At altitudes between 800 and 900 km, a satellite completes a full orbit in about 100 minutes. During that time the earth will rotate around its polar axis by 25 degrees. This represents the shift of the ground track of the satellite between successive orbits. Due to the curvature of the earth, a polar orbit satellite can cover the polar region much more frequently than the equatorial. For example, a satellite at 800 km orbit passes over the Arctic or the Antarctic 11–12 times per day. That is enough to cover the entire region if the swath

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE (a)

(b) N

tor Equa

S

Near 90° orbit

Or

bit

pla

ne

The orbit plane rotates at The same rate of earth’s Rotation around the sun

Figure 7.3 Configuration of a near-polar orbit, sun-synchronous satellite, (a) the near-polar orbit with inclination to the vertical plane, and (b) the sun-synchronous mode where the orbital plane must rotate by approximately one degree per day eastward to keep pace with the earth’s revolution around the sun.

of the satellite is wide enough (e.g., around 2000 km). The repeat cycle of a polar orbit satellite is defined as the number of days between two successive identical orbits. It usually varies between 6 and 35 days depending on the altitude of the orbit. The altitude and inclination of the orbit can be combined to produce what is known as a sun-synchronous orbit. In this trajectory, the angle between the sun–earth line and the plane of the satellite remains constant throughout the complete rotation of the earth around the sun (Figure 7.3b). This ensures that the satellite ascends (heading north) or descends (heading south) over any given point along the equator at the same mean solar time. For example, a satellite might ascend across the equator twelve times a day, each time at approximately 3:00 pm mean local time. This property is important in the case of optical sensors because the surface illumination angle will be nearly the same every time the satellite crosses the given latitude. Most of the EO satellites are sun-synchronous. For example, the Canadian RADARSAT-1 is placed in a sun-synchronous orbit at 798 km altitude but inclined by 98.4 to the equator. With these parameters it crosses the equator at 6 pm and 6 am in the ascending and descending modes, respectively. It should be noted that at locations away from the equator, particularly in the polar regions, the local time of observation of a given swath can vary quite considerably. The other category of satellite orbit is the geostationary. In this mode, the orbit is a circular path directly above the equator at an altitude of 35,786 km. The orbital speed at this altitude is 3.07 km/s, which equals the linear speed of the earth’s rotation. Therefore, the satellite views the same part of the earth all the time (uninterrupted). To an observer on the earth, the satellite appears stationary in

287

the sky, hence the name geostationary. The satellite usually covers one-third of the earth’s surface. Therefore, the entire surface (except the extreme polar regions) can be covered with three geostationary satellites positioned over the equator at longitudes separated by 120 degrees. Geostationary satellites are the preferred choice for tracking weather systems; hence, they are also referred to as meteorological satellites. A network of operational geostationary satellites is used to provide images of the earth’s surface in the VIS and IR bands. They include the United States Geostationary Operational Environmental Satellite (GOES), the European Space Agency’s EUMETSAT operational satellite series, the meteorological mission of the Japanese Multi-functional Transport Satellite (MTSAT), and India’s weather satellite series Indian National Satellite System (INSAT). Since most of the ice in nature exists at or near the polar regions, data from polar orbit satellites are used more often than geostationary satellites in ice applications. Depending on the source of the illumination, spaceborne sensors are grouped into two categories: passive and active (Figure 7.4.). Passive sensors have the solar radiation as the source of illumination. They measure the reflected radiation in the VIS/NIR bands or the emitted radiation in the TIR or microwave bands. Active sensors, on the other hand, transmit their own illumination signal (radar or laser pulses). They include imaging radar, scatterometer systems, and altimeter systems (radar and laser). The transmitted signal is reflected/ scattered off the surface or within the volume, and the receiving antenna records the part of the signal that scatters back to the antenna (called backscatter). In monostatic radar systems, the same antenna is used for transmitting and receiving the signal. Active sensors can image the earth’s surface during daytime or at night. Measurement of radiation by space-borne instrument can be performed by a non-scanning (profile) or scanning (imaging) sensor. Active non-scanning sensors include altimeters that look down at the surface at nadir direction, measuring the precise time between transmitted and received signals, and extract the height of the observed surface with respect to a given reference. Wind scatterometer is another example of an active non-scanning sensor that observes the same part of the ocean surface from different (at least three) view angles. It is mainly used to estimate winds over ocean surface. Active scanning sensors, on the other hand, include SAR and its predecessor real aperture radar (RAR). Fundamentals of these systems are presented in section 7.6. Sensors classified as passive non-scanning include space cameras (although these are imaging systems) such as the one onboard the Russian COSMOS satellite. Passive scanning (imaging) systems are classified into space TV cameras, multispectral optical-mechanical scanner, and scanning microwave

288

SEA ICE

Active sensor

Passive sensor

Figure 7.4 Illumination source for active and passive sensors.

(a)

(b)

Ro

Sate

llite 6.56 motion km/s

tat

45°

mi

833 km

ing

rro

F

r

Sc

gle an 2° 10

53°

θ

85 GHz 37 GHz

e

n ca

an

lin

19 GHz

S

.5

12 km

Figure 7.5 Configuration of (a) cross-track scanning radiometer and (b) a conical scanning radiometer (composed by M. Shokr).

radiometers. The latter has been frequently used in monitoring sea ice in the polar regions since early 1970s. The two most commonly used types of scanning radiometers are cross-track and conical scanners (Figure 7.5). In cross-track scanners (Figure 7.5a), the radiometer scans the swath width (F) in a series of lines using a rotating mirror (A). The lines are oriented perpendicular to the direction of motion of the satellite platform (i.e., across the swath). Obviously, the incidence angle varies across the swath. Image is constructed from sequential scanned lines. The scanner can record the radiation while moving in one direction only (sidewise) or in both directions.

When moving in both directions, the successive scan lines make a zigzag. This problem can be rectified by moving the lenses by infinitesimally small distance along the satellite track line in order to offset its forward orbital motion. This compensation can work only in the case of optical systems such as multispectral scanners. It does not work in the case of microwave radiometers because of the large aperture of the antenna. As a result, the microwave antenna of a cross-track scanner usually scans successive lines in one direction only. Conical scanners are commonly used in space-borne TIR and PM sensors. The scanner views the surface while

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

looking forward or backward at a constant incidence angle. An example of a backward conical scatter is shown in Figure 7.5b. Each line is scanned separately, and a single observation represents radiation integrated over a specific “footprint” of the sensor, which is also called the integrated field of view (IFOV). Its dimensions are determined by the angular resolution based on 3 dB beamwidth of the antenna pattern projected on the earth’s surface. The advantage of a conical scanning system is the fixed incidence angle. This eliminates the need for correction of the measured radiation due to varying incidence angle across the scan line. However, conical scanning causes the successive IFOV to overlap more toward the end of the swath. This point will be explained in section 7.5 (Figure 7.17). The IFOV determines the spatial resolution of the sensor on the ground, but the pixel size in the imagery data is determined by the sampling rate of the observed radiation. If the sampling takes place at an interval that corresponds to a ground distance shorter than the characteristic dimension of the IFOV, the “pixel spacing” becomes smaller than the IFOV and successive samples from adjacent pixels overlap (oversampling). This is different from the overlapping of the IFOV near the end of the swath due to conical scanning as mentioned before. This situation also arises when the data are gridded with grid spacing less than the IFOV dimensions. In this case information of emitted radiation from each IFOV is distributed over an area of a few grid cells. The most authentic radiometric information is found in two cases: when the sampling rate matches the footprint of the sensor or when the data is provided in non-gridded format. However, data are usually offered in gridded projection so that multisensor data can be integrated and geographic information can be overlapped. Quantitative retrieval of information from remote sensing data requires radiometric calibration of the data. This entails converting the pixel value provided by the vendor to the actual radiometric or scattering value measured by the sensor. Absolute calibration accounts for sensor parameters such as gain, offset, viewing geometries, antenna pattern, and power spread of the illuminated signal (in case of the radar sensor). The absorption and scattering of the signal by the atmosphere is usually accounted for in a subsequent process (not as part of the calibration). This requires mathematical forward modeling to simulate the observations. Remote sensing imagery data provided to users are usually not absolutely calibrated. However, absolute calibration can be performed using software available in most major remote sensing image processing packages. Relative calibration, on the other hand, is a simpler task that aims at normalizing data from multiple passes of the same satellite or

289

from multiple sensors on different satellites. It involves tuning the observation to match the radiation from a ground target of known reflectance, emission or scattering, which is considered to be constant over time. Satellite operators usually provide coefficients to tune the calibration if a drift in sensor measurements occurs over time. Relative calibration should be satisfactory in sea ice applications such as delineation of sea ice boundaries, identification of ice surface features that contrast the background ice, or tracing trajectory of ice features in sequential images. Most of the image classification techniques employ procedures that require relative calibration only. If, on the other hand, the purpose is to retrieve geophysical parameters such as surface temperature, roughness, ice concentration, or any parameter that depends on the actual value of the observed radiometric/scattering, then absolute calibration becomes necessary. Examples include quantitative retrieval of salinity, moisture, surface temperature, and roughness. It is also important for approaches that involve combining data from different sensors or different dates. Producing operational ice charts based on visual analysis of remote sensing images does not require calibrated data. 7.2. ELECTROMAGNETIC WAVE PROPERTIES AND PROCESSES Information in this section highlights a few definitions and describes processes of EM wave propagation relevant to remote sensing of sea ice. Since microwave remote sensing is at the center of this application, the section starts with discussions on polarization of EM wave, a particularly important wave property for microwave sensing. This includes definitions and mathematical formulations of polarized signal, which furnishes a necessary background to understand the full potential of microwave data. Mechanisms of depolarization of the received signal in the case of radar data are also presented in order to support the discussions on the use of radar polarimetry in section 7.6.2.3. This is followed by presenting the five basic processes of EM wave interactions with matter, namely reflection, transmission, absorption, scattering, and emission. These processes are frequency-dependent. The discussions highlight a few points relevant to understanding the data from optical and TIR sensors. The section addresses also the concepts of brightness temperature and the emissivity, and concludes with information on penetration depth of the EM signal, which determines the depth from which the recorded observations are engendered. For more detailed and comprehensive discussions on these fundamental concepts the reader may check books on fundamentals of remote sensing [e.g., Reese,

290

SEA ICE

2013, Joseph, 2017], satellite remote sensing [e.g., Chuvieco, 2020], and microwave remote sensing [Ulaby, Moore, Fung, 1981, 1982, Elachi and van Zyl, 2006, and Ulaby and Long, 2014].

Polarization of EM wave refers to the alignment of the electric field vector in the wave. If the vector is aligned to a certain plane or varies according to a predictable alignment while the EM wave is propagating, the wave is called polarized. In the first scenario the wave is called linearly polarized and in the second it is called elliptically polarized (circular polarization is a special case of elliptical polarization). On the other hand, if the wave has random time-varying electric field vector it is called unpolarized. Between these two ends the signal can be partially polarized. In section 4.1.1 rudimentary aspects of polarization of EM waves is introduced to highlight the use of polarized light, particularly plane-polarized light, for examining the grain structure and texture of polycrystalline ice in thin ice section samples. In this section, more account on the polarization of EM waves is presented and discussed in the context of remote sensing observations. While the electric field of the incident wave can be in any plane, it is customarily resolved into two orthogonal components. One being the plane defined as containing the propagation vector of the EM wave, and the normal to the surface that the wave strikes. If the electric field vector lies in this plane, it is said to be vertically polarized and referred to as a traverse magnetic (TM wave). If it lies in the perpendicular plane, the wave is said to be horizontally polarized and referred to as traverse electric (TE) wave. Most sources of light are classified as unpolarized or at least partially polarized. This makes the reflected observations from optical and the emitted observations from TIR sensors also unpolarized. Therefore, the polarization of

the observations in these cases becomes irrelevant, i.e., and does not carry information about the observed object. In PM remote sensing, the emitted radiation can have a varying degree of polarization. For example, the emission from ocean surface has considerably higher component in vertical than horizontal polarization. Hence, the difference between emission measured using the two polarizations is an important parameter used in the discrimination between seawater and sea ice. In radar remote sensing the polarization of the received signal is even more important because the transmitted radar signal is already polarized. Hence, if the observed surface depolarizes (i.e., change the polarization) of the incident signal, the received signal will carry information about the surface. The depolarization of the radar signal by the observed surface is discussed later, but it suffices here to confirm that recording the polarization of the scattered radar signal can provide important clue to identify the observed surface. The transmitted radar signal can be linearly polarized (vertically or horizontally according to the aforementioned definition) or elliptically (or circularly in a special case) polarized. The linearly polarized wave can also be generated at any plane other than vertical or horizontal. Figure 7.6 shows the mechanisms of generating the two types of signal polarization: linear and circular. Two orthogonal polarizations waves are generated by the antenna and the resulting polarization is determined by the vector addition of the two orthogonal waves. If the two waves are in-phase, the resulting plane-polarized wave will oscillate within a fixed plane and the projection of its tip on a plane perpendicular to the wave propagation is a line. Therefore, this is called linearly polarized wave. The slope of the line is determined by the relative amplitude of the two orthogonal transmitted waves. If the amplitudes are equal, the slope will be 45 , as shown in the figure. In the current operational SAR systems the linear polarization is either 0 (horizontal) or 90 (vertical).

(a)

(b)

7.2.1. Polarization of EM Wave

Figure 7.6 Illustration of two antenna-transmitted EM waves whose electric components are orthogonal but the phase is different. In (a), the phase difference is zero and in (b) it is π/2. On a plane perpendicular to the wave propagation the tip of the resultant vector traces a line in (a) and a circle in (b).

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE (a)

(b)

y

(c)

y

y z

z

z

x

291

x

x

Figure 7.7 Propagation of the electric field in an EM wave, (a) linear polarization where the waveform remains in the same plane, (b) elliptic polarization, and (c) circular polarization.

“polarization ellipse.” Its geometry is defined in terms of the orientation ψ and ellipticity χ as shown in the figure. The orientation is the angle of the semimajor axis, measured counter-clockwise from the positive horizontal axis. Ey It takes on values between 0 and 180 . The ellipticity a (also called eccentricity) measures the degree to which E the ellipse is oval [χ = arctan (b/a)]. It takes values between χ° ψ° b −45 and +45 . The zero value corresponds to linear x polarization and the ± 45 values correspond to circular Ex polarization. Table 7.1 shows the phase difference between the two orthogonal transmitted waveforms needed to generate different shapes of the polarization ellipse. If the two components of the transmitted signal are in-phase (ΔΦ = 0 or π/2), the polarization of the resultFigure 7.8 Polarization ellipse of polarized EM wave showing its ant signal becomes linear (χ=0). As a special case, to two characteristic parameters: the orientation ψ and ellipticity χ. generate a linear polarization in the H plane, the V component must be zero and vice versa (not shown in the table). When the phase shift between the two equally The alternative scenario involves two simultaneously transmitted waves increases to π/2 radians, the ellipticity transmitted waves out-of-phase. This case is illustrated in increases to 45 and the polarization becomes circular, Figure 7.6b, where the phase shift is quarter cycle (i.e., regardless of the orientation ψ. Only two values of the oriπ/2). In this case, the tip of the propagating wave traces entation angle are shown in Table 7.1: 45 or 135 . The a circle on the perpendicular plane (assuming the amplirecently developed SAR compact polarimetry (CP) mode tudes of the two transmitting waves are equal). This is (see section 7.6.3) onboard RADARSAT Constellation known as circularly polarized wave. If the phase shift is Mission (RCM), which was launched in June 2019, transbetween zero and π/2, the tip of the wave traces an ellipse, mits a right-circular polarized wave and receives linear and hence known as an elliptically polarized signal. The polarization: horizontal (H) and vertical (V). Hence, the progression of the time-varying direction of the electric received signal is designated RH and RV, respectively. field vector for linear, elliptical, and circular polarizations In the classical literature, the polarization state of a is shown in a 3D perspective in Figure 7.7. As can be seen, plane wave is expressed by a vector, known as the Stokes the linear and circular polarizations are two limiting cases vector. Its elements are functions of the orientation and of the more general configuration of elliptical polarization. ellipticity of the signal and can be written as: Polarization state of an EM wave can be defined in terms of two parameters; orientation and ellipticity. Ev 2 + Eh 2 Io I0 They are explained in Figure 7.8 using the most general Q I 0 cos2ψcos2χ Ev 2 − Eh 2 = (7.1) = case of elliptical polarization. In this case an ellipse is ∗ U I 0 sin 2ψcos2χ 2 Re E vEh traced by the tip of the electric field vector in the plane perV I 0 sin 2χ 2Im E v E ∗h pendicular to the wave propagation. This is known as y

292

SEA ICE Table 7.1 Phase shift (Δϕ) between the electric field of the two orthogonal transmitted EM components and the shape and parameters of polarization ellipse associated with the resulting wave. Δϕ

0

π/8

π/4

3π/8

π/2

5π/2

3π/4

7π/8

π

ψ

45°

45°

45°

45°

any

135°

135°

135°

135°

χ

0°

11°

23°

34°

45°

34°

23°

11°

0°

shape

where, I0, Q, U and V are known as Stokes parameters. The asterisk refers to the complex conjugate. For fully polarized wave, the Stokes parameters are related by the equation: I o 2 = Q2 + U 2 + V 2

(7.2)

Therefore, only three parameters of the Stokes vector elements are independent. The parameter Io2 represents the total power in a completely polarized signal. For a partially polarized wave, the total power is greater than the summation of the squares of the other three terms. The degree of polarization dp of a signal is determined according to this ratio:

dp =

Q2 + U 2 + V 2 I 20

(7.3)

This parameter has been successfully applied for target scattering characterization [Touzi, 1992] from polarimetric radar observations (section 7.5.2.3). In radar terminology the term “depolarization” refers to the situation when the dominant polarization of the scattered signal is different than the polarization of the transmitted signal. This term does not apply to the received signal in PM because there is no transmitted signal. Nevertheless, the emitted signal can be recorded in terms of its horizontal or vertical polarization components or both even if no preferred polarization exists (i.e., randomly polarized emitted signal). The depolarization of the scattered radar signal depends on the surface roughness and the subsurface composition. As a rule of thumb, depolarization of radar signal is caused by multi-scattering mechanism, i.e., when the scattered signal keeps bouncing off different facets of the surface or scattering elements (i.e., inclusions) within the volume. This is manifested in the scattering from surfaces covered with random structure such as

forest trees or various urban structures. For sea ice, the multiple scattering is possible from ice blocks of a ridge or from a volume that has inclusions in the form of scattering elements (e.g., air bubbles in MY hummock ice). In order to trigger volume scattering, the characteristic dimensions of those elements must be comparable to the wavelength of the incident signal. The depolarization of the scattered microwave signal from a rough surface is discussed in Fung [1966]. The study shows that both quasispecular scattering and the multiple scattering are responsible for the polarized return, while multiple scattering alone is responsible for depolarization. In a follow-up study, Fung [1967] calculated backscatter from rough surfaces and inhomogeneous (i.e., anisotropic) volumes in microwave frequencies and concluded that crossdepolarized backscatter measurements are much less dependent on incidence angle than co-polarized backscatter. More information on depolarization of radar scattering and the partial polarization of PM emission is included in Fung and Chen [2009]. 7.2.2. Reflection, Transmission, Absorption, Scattering, and Emission The EM wave interaction with any medium is manifested in five radiometric processes: reflection, transmission, absorption, scattering, and emission. Absorption and scattering represent losses. The summation of the transmission, absorption, and reflection power components is equal to the power of the incident radiation. The ratio of the amplitude of the reflected wave to the amplitude of the incident wave is known as reflection coefficient. Similar ratios are known as transmission coefficient and absorption coefficient (though these terms are loosely used). If power ratio is used instead of amplitude ratio, the first three processes are defined in terms of reflectivity, transmissivity, and absorptivity. Transmittance and absorptance are also used though not

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

universally adopted. The scattering may be measured in terms of radar cross section (RCS). This is a hypothetical area which scatters the signal isotropically with the same power intensity as received by the antenna of the radar system. However, a more common observation parameter is radar scattering coefficient. This is the average radar cross section value per unit area of a large target. More on these two concepts is presented in section 7.6.1.1. The emitted radiation results when part of the incident energy is absorbed and re-radiated later to maintain thermal equilibrium with the surroundings. Objects of the earth system emit energy in the TIR and microwave spectral regions. Every object emits energy determined mainly by its physical temperature (emission ceased if the temperature reaches above absolute zero or −273.18 C). In addition to the physical temperature, emission is also determined by the emissivity of the material (see definition in the next section). All of the above-mentioned coefficients are wavelength-dependent. It should be mentioned that emitted radiation is relevant only to the TIR and microwave regions. In optical remote sensing only reflection applies because no component of the earth system emits light (sun is the only natural source).

7.2.2.1. Reflection and Fresnel Model Reflection takes place at the interface between two media of different refractive indices (in the optical region) or different dielectric constants (in the microwave region). Part of the energy that penetrates the medium may undergo scattering and a portion of the scattered signal may reach back to the reflective surface and get refracted across it. That part is known as internally reflected portion and it adds to the surface reflection. This explains the high reflectivity of the snow in the VIS region. It is not only caused by the white smooth surface but also by the added reflection from within the composite snow flake and air under the surface, which is refracted back to the air across the snow surface. Since the earth does not emit radiation in the visible spectrum, the reflectivity of the sunlight in the VIS range determines the visibility of the surface. Spectral (mirror) or scattered (diffuse) reflection depends on the surface roughness with respect to the wavelength of the incident signal. Spectral reflection features reflection in a single direction away from that of the incident wave. It occurs if the surface is smooth (its root mean square of height is small compared to the wavelength of the incident radiation). On the other hand, scattering occurs if the surface is rough and it increases with roughness. It should be noted, however, that a smooth surface may lead to diffuse reflection if scattering is caused by crystallites, impurities, or inclusions within a layer immediately under the surface. Specular reflection is governed by the laws of geometrical optics.

293

There is lack of standardization regarding the use of the term “reflection” in the field of remote sensing. Erroneous or ambiguous nomenclature regarding this term has led to misunderstanding of data and inability of comparing measured or modeled quantities. Schaepman-Strub et al. [2006] presented recommendations to standardize the use of the term. The definitions adopted in this book entail that reflectance and reflectivity are expressions of the power fraction of the reflected to the incident wave. The difference between the two terms is introduced in the next paragraph but both are scalar quantities. On the other hand, reflection coefficient is a complex number determined by Fresnel equations (Fresnel is pronounced fray-nel) for spectral reflection off a single layer reflector. The equations are introduced later. There is a difference between reflectivity and reflectance (though both are power ratios and assume a smooth surface to satisfy the condition for specular reflection). Reflectivity is restricted to the reflection off the upper (or first) surface. Reflectance takes into consideration the part of the scattered energy within the volume and the reflected energy off the lower (opposite) surface. Both are known as internal reflection and may go back to the upper (first) surface and get refracted through. This adds another component to the reflectivity. Obviously, the contribution of the internal reflection varies with the thickness and composition of the medium. With this in mind, reflectivity is considered to be the limiting value of reflectance as the medium becomes very thick with no scattering elements. Reflectivity is a property of the material, while reflectance is the fraction of EM power reflected from a specific sample; hence it implies information on the thickness of the sample. The term reflectance is usually used in a non-rigorous sense when addressing the reflection off any surface, including sea ice sheet. Specular reflection at an interface of mismatched media can be modeled using Fresnel equations. The interface must feature a quasi-steady smooth surface. For optical signal, which is always unpolarized; the plane of electric field changes randomly while the direction of the wave propagation is constant. Therefore, the wave can be identified in terms of its two components: one on the plane of incidence and the other perpendicular to the plane of incidence. The plane of incidence is defined as the plane combining the wave propagation and the perpendicular to the interface. Fresnel equations define the reflection and transmission (refraction) coefficients of the reflected and the refracted waves, each with their two components parallel and perpendicular to the plane of incidence (Figure 7.9). These components are denoted Rh or the parallel component (||) and Rv or the perpendicular component (⊥). When an optical EM beam is propagating in a medium with refractive index ni and strikes another medium with a

294

SEA ICE

Inc

ide

Plane nt

be

am

of inc ‖

idenc

e r‖

ted

Reflec

⊥ r⊥

Interf

ace

ted rac Ref

t⊥

t‖

Figure 7.9 Illustration of an incident beam with its reflected and refracted beams at an interface of mismatched media. The two orthogonal components of the wave are shown by red when in plane of incidence and blue when perpendicular to the plane of incidence (see text for definition of the plane of incidence).

different refractive index nt at an incidence angle θi, the two components of the reflection coefficient are given by (note that the refractive index is a complex number);

Rh =

ni cos θi − nt cos θt = ni cos θi + nt cos θt

1−

ni sin θi nt

2

ni cos θi − nt

1−

ni sin θi nt

2

ni cos θi + nt

(7.4)

Rv =

ni cos θt − nt cos θi = ni cos θt + nt cos θi

1−

ni sin θi nt

2

ni

1−

ni sin θi nt

2

ni

− nt cos θi + nt cos θi (7.5)

where, θt is the local angle of refraction at the interface. The second form in each one of the above two equations is derived from the first by eliminating θt using Snell’s law (also known as the law of refraction), which has the following expression: sin θi nt = (7.6) sin θt ni Equations (7.5) and (7.6) show that the reflection increases as the contrast in the refractive index between the two media increases. Fresnel equation is directional, i.e., it determines only the specular reflection of the two wave components at any given incidence angle and the results vary with the angle. It is applicable whether the incident EM wave is polarized or unpolarized. The situation is different for the linearly polarized radar signals because the wave propagates in either the

horizontal or the vertical plane. So, there is no need to decompose it into two orthogonal components in the incidence plane and the perpendicular. Moreover, since Fresnel equation is applied to smooth surface only, and this surface does not depolarize the reflected signal, Fresnel equation gives the reflection in the same plane, i.e., in the co-polarized observation of transmit–receive configuration, namely Rhh or Rvv. The complex dielectric constant of the incident and transmitted waves ϵi and ϵt, respectively, are used instead of the refractive indices ni and nt (definitions of the complex permittivity are introduced in section 3.7.1). The refractive index (a complex number) is related to the complex permittivity ϵ [also called dielectric constant and defined in equation (3.64)] and the permeability μ of the medium by the equation: n=

ϵμ

(7.7)

For most natural materials, μ is very close to 1 at optical frequencies, therefore: n=

ϵ

(7.8)

For an EM wave traveling in air (ϵi = ni = 1) and striking a surface with complex permittivity ϵt at incidence angle θ, the two Fresnel equations can be obtained by replacing nt in equations (7.4) and (7.5) with n and substituting it with ϵ from equation (7.8). Assuming ni = 1, the resulting equations take the form: Rhh ϵt , θ = Rvv ϵt , θ =

cos θ −

ϵt − sin 2 θ

cos θ +

ϵt − sin 2 θ

ϵt − sin 2 θ − ϵt cos θ ϵt − sin 2 θ + ϵt cos θ

(7.9)

(7.10)

These are complex quantities since ϵt is a complex number. Their magnitude varies between 0 and 1 (with 1 indicating total reflection). The reflectivity in the power domain can be obtained from the calculated reflection coefficients using the equations: Γ hh = Rhh R∗hh = Rhh

2

(7.11)

Γ vv = Rvv R∗vv = Rvv

2

(7.12)

Equations (7.11) and (7.12) are plotted in Figure 7.10 for a smooth surface of a lossless medium with dielectric constant (3.2 − j0). The equations simulate the change of radar reflection from smooth lake ice surface with incidence angle. Similar data for sea ice can be obtained by using the appropriate dielectric constant. This information can be used in modeling the backscattering as presented in Chapter 12. In general, the horizontally polarized waves are reflected more readily than vertically polarized waves. It increases monotonically with increasing incidence angle and reaches a peak of 1 at the grazing

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE Unpolarized incident light

V re e rti fle ca ct lly e d -; lig pola h t ri ze d

1.0

295

0.8

mismatch interface Reflectivity

6.0

Γh

tally izon d r o H rize ght pola cted li a refr

0.4 Brestwer’s angle Γυ

0.2

Figure 7.11 Illustration of unpolarized light incident at an interface at Brewster angle showing the reflected and refracted signals polarized in vertical and horizontal polarization, respectively. 0.0 0.0

20

40

60

80

Incidence angle

Figure 7.10 Reflectivity calculated from Fresnel equations for the radar signal as a function of incidence angle for a smooth surface of lossless medium with a dielectric constant ϵ = 3.0 − j0 (representing freshwater ice).

angle (close to 90 ). On the other hand, reflection of the vertically polarized waves decreases as the incidence angle increases until it reaches zero at an angle known as the Brewster angle [named after the Scottish physicist Sir David Brewster (1781–1868)]. Beyond this angle, the reflection increases as shown in Figure 7.10 and reaches its peak also at the grazing angle. In using equations (7.11) and (7.12) for sea ice it should be noted that its dielectric constant in the microwave region varies over a relatively wide range depending on the ice type and frequency of the EM wave. The typical value for FYI is 3.5 − j2.0. Ellison et al. [2003] measured the permittivity of seawater in the microwave spectral range 30–105 GHz with varying temperature. At −2 C, they found that complex permittivity ϵ was 10.66 − j19.46 and 7.43 − j9.38 for 37 and 89 GHz, respectively. It is obvious that the loss component of seawater is significantly higher than that of sea ice. It is appropriate to point out that the concept of Brewster angle (also known as the polarization angle) applies to the unpolarized EM (e.g., optical signal). What happens when light strikes an interface at an angle equal to the Brewster angle? The signal that contains equal amounts of vertical and horizontal polarization (Figure 7.11) is

reflected with only horizontally polarized component, and refracted with only vertically polarized component. Of course, for an incoming vertically polarized EM wave at this angle there is no reflection. This confirms what is already mentioned in section 4.1.1, namely when an unpolarized monochromatic light beam strikes a transparent plate at incidence angle equal to the Brewster angle for the given material and the wavelength, the reflected beam becomes plane-polarized.

7.2.2.2. Transmission The part of the EM wave that is not reflected off the surface penetrates the medium. Depending on the manner of the wave interaction with the atoms and electrons, part of the penetrated signal can be scattered or absorbed (next section). If no losses occur, the signal can be fully transmitted and the medium is considered fully transparent. This is the case, for example, of the atmosphere, which is transparent for longwave of microwave (> 10 cm). The other extreme end is a medium where the entire power of the incident radiation is lost within, i.e., fully absorbed. This medium is considered to be opaque. Anything between these two limits is a partially transmitting medium. The EM wave transmission is a function of the optical and the electrical properties of the material as well as the wavelength λ of the penetrated signal. The transmittance of a signal Tλ is defined as the ratio of the intensity (power) of the transmitted to the incident radiation. Tλ = I I0

(7.13)

296

SEA ICE

7.2.2.3. Absorption and Scattering Losses Absorption and scattering are two mechanisms of loss of an EM wave while propagating in a medium. When the radiation interacts with an atom, the latter may be excited to a higher energy level. This energy may transform into kinetic energy when the atom collides with other atoms, raising the temperature of the medium. This loss of energy is known as absorption. On the other hand, if the excited atom emits the extra energy immediately (i.e., within nanoseconds) as a photon (should be at the same frequency as the frequency of the propagating EM wave), the process is known as atomic scattering (or just scattering). The scattering may also be caused by reflection of the EM wave off tiny elements such as dust particles in the atmosphere. Brief information about absorption and scattering processes in the atmosphere as well as the atmospheric agents that trigger them is presented in section 7.9.1. Absorption and scattering losses have direct impact on remote sensing observations because the recorded reflected or emitted radiation from the surface has to be corrected to account for both forms of losses while traveling in the atmosphere. Absorption and scattering are described by their relevant coefficients. The absorption coefficient α (also called the attenuation coefficient) is defined as the fractional decrease in the intensity of the radiation dI over a distance dx. It has dimension L-1. dI = α x dx I

(7.14)

If α < 1 0, part of the radiation is attenuated within the depth of the material. If α = 1 0 all radiation is attenuated within the unit width. In general, both I and α depend on frequency, and if equation (7.14) is integrated over a finite distance x in the atmosphere between x = 0 where I = I o and x = X , then: X

I = I o exp

α x dx

(7.15)

0

Similar equations can be written for the scattering loss coefficient σ (x) if α in equations (7.14) and (7.15) is replaced with σ. If both absorption and scattering coexist then they can be combined into what is known as the extinction coefficient κ: κ=α+σ

I = I oe − τ

(7.17)

This term is used in a simplified version of radiative transfer equation throughout this book [e.g., equation (7.56)] and also to retrieve ice surface temperature using microwave observations (section 11.5). Conservation of energy implies that the amount of incident energy is equal to the sum of the reflected, transmitted, and absorbed energy. Therefore, in a simplified view, reflectivity, transmissivity, and absorptivity should add up to 1. Since most materials, including ice, are opaque to thermal radiation, it is safe to assume that the sum of reflectivity and absorptivity equals one in this case. This means that good reflectors are poor absorbers and vice versa. This principle of conservation of energy is known as Kirchhoff’s law. 7.2.2.4. Emitted Radiation (Re-radiation) Part of the incident solar radiation in the VIS and NIR bands is absorbed and later re-radiated (emitted) at longer wavelengths in the TIR, FIR, or microwave spectral region. Most of the energy emitted by the earth’s surface, with its temperature ranging roughly between −50 C to 50 C, is in the TIR regions. Unlike other radiometric processes, incident radiation is not directly involved in the emission process. When incident radiation is absorbed, an atomic process is triggered and results in emitting radiation at frequencies different than that of the incident radiation. Emission is determined by the physical temperature of the medium and its emissivity. The latter is a function of its physical composition of the medium. Surface roughness determines the angular distribution of the emitted radiation. In the TIR region, emission is a measure of how an object radiates heat after its temperature rises as a result of energy absorption. In the microwave region emission is more of a function of the molecular properties of the medium. The concept of emissivity is introduced in the next section, but it is appropriate to introduce its connection to the reflectivity to support further discussions of the radiometric processes in this section and in Chapter 12. Under thermal equilibrium, the medium must emit the same amount of energy that has been absorbed (i.e., emissivity and absorptivity are equal). This principle of conservation of energy is a corollary of the well-known Kirchhoff’s law of thermal radiation. It takes the form:

(7.16)

εv θ = 1 − Γ v θ

(7.18)

The integral of the loss coefficient in equation (7.15), whether it is absorption, scattering, or extinction, with respect to the distance x is called the optical thickness τ, which is dimensionless (also known as coefficient of attenuation). The intensity of the radiation decreases over the distance τ according to the equation:

εh θ = 1 − Γ h θ

(7.19)

The RHS in the above equation is the absorptivity (the complementary of the reflectivity Γ). The reflectivity can be obtained from equations (7.11) and (7.12). Qualitatively speaking, the equations indicate that, for optical

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE (a)

297

(b) 1.0

1.0

εv 0.8

εh Emissivity

Emissivity

0.8

6.0

0.4

0.2

0.0 0.0

εv

6.0

0.4

εh

0.2

20

40

60

80

Incidence angle

0.0 0.0

20

40

60

80

Incidence angle

Figure 7.12 Plots of microwave emissivity using Kirchhoff’s law calculated from Fresnel equations using dielectric constant that represents (a) freshwater ice and (b) seawater.

and thermal radiation, strong absorbers (i.e., poor reflectors) at a given wavelength are also strong emitters at or around the same wavelength. In the TIR region, open water is a good absorber and therefore a good emitter with emissivity values around 0.98. Similarly, fresh snow is a good absorber and good emitter with its emissivity very close to one. In the microwave range the Fresnel reflection and consequently the emissivity depend on the dielectric constant through the reflectivity Γ in equations (7.18) and (7.19). Figure 7.12 includes plots of microwave emissivity calculated using equations (7.18) and (7.19) for two values of dielectric constant: (3.0 − j0) and (18.5 − j31.3), which represent freshwater and seawater, respectively (these values are valid for microwave frequency 35 GHz and temperature 20 C). Due to its higher dielectric constant, the seawater has considerably less microwave emissivity than sea ice, especially in the horizontal polarization. This is the basis of sea ice and water discrimination in microwave retrieval algorithms. The peak emissivity in the vertical polarization from water occurs near 80 and from ice near 60 (i.e., ice has smaller Brewster angle than water). It should be mentioned that the wind-driven surface roughness of seawater will increase the emissivity by expanding the sea surface emission area. Jin et al. [2019] proved this point by studying performances of different sea surface emission models in combination with in-situ measurements. 7.2.3. Brightness Temperature and Emissivity Emission in the mid-IR, FIR, TIR, and microwave regions (Figure 7.1) is expressed in terms of brightness temperature Tb , which is defined as the temperature of a blackbody that emits same amount of radiation

observed by the sensor. The justification for this parameter is rooted in the fact that the emitted radiation from a blackbody is governed by a deterministic equation. A blackbody is a hypothetical material which absorbs all incident radiation at all frequencies or just a certain band of frequencies and therefore reflects none. In other words, a blackbody is a perfect absorbent and perfect emitter of energy. At low temperature, a blackbody appears extremely black to the eye—hence the origin of its name. However, at high temperature, it may emit a spectrum of photon energies in the visible EM bands, hence it may appear white. The behavior of a given material with regard to being a blackbody varies depending on the frequency of the incident EM wave. In the optical band, many surfaces approach the ideal blackbody behavior in terms of their ability to absorb the solar radiation. Examples include soot, carbide, silicon, and platinum. In the TIR band, water and ice absorb radiation well and therefore can be considered nearly blackbodies. In the microwave (and radio wave) bands the absorptive properties of water and ice differ from each other and none of them is considered a blackbody. The total radiation flux per unit area RB (in W/m2) across all wavelengths emitted by a blackbody at temperature T (in Kelvin) varies as the fourth power of the physical temperature of the body. The relationship is known as the Stefan-Boltzmann law: RB = σ b T 4

(7.20)

where, σb is the Stefan-Boltzmann constant (5.671 x 10-8 W/m2.K4). Using a quantum mechanics model, Max Planck developed an expression for the radiation flux density RB(f) from a blackbody at a given frequency f and

298

SEA ICE

physical temperature T (in Kelvin), which is known as Planck’s equation: RB f =

2hf 2 1 c2 ehf kT − 1

(7.21)

where, h is Planck’s constant, k is Boltzmann’s constant, and c is the speed of light. The emitted energy reaches a peak at a certain wavelength λmax (in microns) which is proportional to the physical temperature of the object. This is known as Wien’s displacement law (the hotter the object the shorter the wavelength of the maximum emitted radiation). λmax = a T

(7.22)

where, a = 2.879x10 m/K, and T is in Kelvin. Figure 7.13 is a plot of the radiation flux using equation (7.21) for different physical temperatures. The approximate average temperature of the sun’s surface is 6000 K. Although the sun does not fit the definition of blackbody because it does not absorb energy, its radiation follows the blackbody radiation and peaks at approximately 0.5 μm wavelength. The terrestrial radiation resembles that of blackbody at 300 K (average temperature of the earth). The earth system (surface and atmosphere) emits radiation, known as outgoing longwave radiation, in the wavelength range from 3 μm to 100 mm. It peaks around 10 μm, which corresponds to an atmospheric window in the TIR region (Figure 7.2). Earth surface radiation in the microwave range is much weaker than that in the TIR. Therefore, in order to achieve a reasonable signal-to-noise ratio, microwave sensors should integrate the emitted radiation over a large IFOV (typically a few kilometers width).

Spectral radiance (W/m–2. Sr –1. μm–1)

1010 10000 K

108

6000 K (sun)

106 3000 K (Tungsten filament)

104

1000 K 500 K

100

300 K (Earth)

1 100 K

0.01 0.1

1 10 Wavelength (μm)

100

Figure 7.13 Radiation from blackbody (Planck’s equation) against wavelength for different physical temperatures. Shaded area marks the visible spectrum.

In the microwave range of 1 to 300 GHz, the term hf /kT in equation (7.21) becomes very small because of the small frequency of the microwave signal. In the limit, the exponential term in this equation becomes small and can be approximated with the first-order term in the Taylor polynomial. Planck’s equation can then be reduced to a form known as Rayleigh-Jean’s equation: RB f =

2hkTf 2 c2

(7.23)

The interesting feature of this equation is the linear relationship between the radiation and the physical temperature of the object. This is not the case in the TIR region, where the emitted energy is proportional to T4 as indicated by equation (7.20). The relations between the emitted energy and the physical temperature [equations (7.22) and (7.23)] allow the conversion of energy measured by a radiometer to temperature. This temperature would be equivalent to the temperature of a blackbody that radiates the same amount of the measured energy. It is called brightness temperature Tb. Measurements of radiance RB by any radiometer are usually provided as Tb. Radiation from a TIR channel can be converted to brightness temperature by the inversion of Planck’s equation [Gumley, Hubanks, Masuoka, 1994]: T b = c2

λ ln c1 λ5 RB λ, T b 106

+1

(7.24)

where RB is Planck radiance in W/m2.sr.mm−1, λ is wavelength in m, c1 = 1.1910439 × 10−16 W/m2, and c2 = 1.4387686 × 10−2 m.K. Most objects in nature radiate less energy compared to a blackbody at the same temperature. Such an object is known as “gray” body and its radiation is denoted (RG). Emissivity ε defined as the ratio of the emitted radiation from a given body to the radiation from a blackbody at the same physical temperature, (RB). It is the physical property of a material that describes its ability to emit radiation: RG (7.25) ε= RB The emissivity is always less than one but it approaches one as the object features more resemblance to blackbody behavior. In terms of temperature ratio, emissivity takes two different forms; one applies to the TIR and other to the microwave emission. Since the earth is usually considered a blackbody with emitted radiation that peaks in the TIR (Figure 7.13), the expression for the emissivity in the TIR region can follow from equation (7.25) using equation (7.20) as an expression for RB (which mainly implies energy flux through the TIR region in this case): ε=

RG σT b 4 Tb = = 4 RB T σT

4

(7.26)

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

or T b = ε14 T

(7.27)

In the microwave region the expression of emissivity results by combining equations (7.25) and (7.23). The emissivity becomes simply the ratio between the brightness temperature and the physical temperature of the medium (more precisely the physical temperature of the radiating layer of the medium): ε=

Tb T

(7.28)

Therefore, while the constant of proportionality between the brightness temperature and physical temperature of a gray body is ε in the case of the microwave region, it is ε14 in the case of the TIR region. This makes brightness temperature more sensitive to changes in emissivity in the microwave than the TIR region. In fact, in the TIR region the brightness temperature becomes much more sensitive to variations in the physical temperature of the observed object, especially for surfaces of nearly equal emissivity close to 1.0. This is in the cases of water, ice, and snow (TIR emissivity around 0.96). That is, how the brightness temperature in the TIR observations is approximately equal to the physical temperature and it differs between ice and water (in winter) because water is much warmer than ice. On the contrary, in the microwave region the emissivity of water is much smaller than that of ice and snow (water is radiometrically colder in the microwave region). Furthermore, the emissivity of ice or snow varies substantially in response to changes in atmospheric temperature or precipitation (especially when temperature changes around the melting point). Therefore, the measured microwave brightness temperature cannot be used to estimate the physical temperature unless the emissivity becomes known or assumed. More on the range of microwave emissivity from sea ice and snow is presented in section 9.4. In a concluding note, it should be mentioned that brightness temperature depends on the physical temperature of the medium [equations (7.27) and (7.28)]. Hence, unlike the emissivity, it is not an intrinsic radiative property of the material. For that reason, the polarization and gradient ratios (both derived from brightness temperature) are usually used instead of brightness temperature because they are independent of the physical temperature (see definitions and equations in section 9.2). 7.2.4. Penetration Depth Penetration depth is defined as the depth over which the radiative power is attenuated by a factor of 1/e (i.e., about 63% of the radiation is absorbed). Remote sensing

299

observations from any surface are affected by the composition of the observed surface within the penetration depth of the incident wave. For optical and TIR radiation, the penetration depth is usually negligible as it measures in sub-millimeter. Therefore, the theoretical background presented in the following applies to the microwave region only. As mentioned earlier, EM wave propagation undergoes extinction that takes the form of absorption or scattering. The wave extinction per unit length is known as the extinction coefficient. It has units of dB/length. According to the Beer-Lambert law [Beer, 1852], the intensity of an EM wave inside a given substance decreases exponentially with depth. Following this concept, if the transmitted power at a point just below the surface is P(0+), then the power at a depth z is given by the following equation [Ulaby, Moore, Fung, 1982]: P z = P 0+ exp −

z

ke z dz

(7.29)

0

where, ke is the extinction coefficient. From equation (7.29), the definition of δp can be implied in the following equation: δp

ke z dz = 1

(7.30)

0

If ke is assumed to be approximately constant with depth, then δp is given by: δp =

1 ke

(7.31)

For large values of ke, the signal penetration is very limited, hence any contribution to the observed signal in remote sensing data is practically triggered by the surface. According to the two extinction processes mentioned above, ke can be written as the summation of two coefficients; one for the absorption loss ka and the other for scattering loss ks: ke = ka + ks

(7.32)

The ratio k s ke is known as the scattering albedo, which varies between zero and one. Higher values imply that most of the extinction is due to scattering. In practice, the scattering loss in the medium is difficult to compute and is therefore usually ignored. In this case, the penetration depth can be calculated using equation (7.31) with k a substituted instead of ke . Penetration depth can be derived in terms of the complex permittivity ϵ [equation (3.64)] from the solution of the simplified equation of the EM propagation in a lossy homogeneous medium when the wave is assumed to be traveling along the positive z-axis [Ulaby, Moore, Fung, 1981, page 65]:

300

SEA ICE ϵ

E = E 0 e jkz

(7.33)

where, k is the wave number in free space (= 2π/λ), λ is the wavelength in meters. In the case of a lossless material such as freshwater ice or MYI ϵ can be much smaller than the real part ϵ , then: ϵ=

ϵ +j

ϵ

(7.34)

2 ϵ

Substituting equation (7.34) into equation (7.33) yields:

E = E0e

jkz ϵ

e

− kz

ϵ 2 ϵ

(7.35)

The exponents in the above equation define the attenuation (absorption) constant α and the phase constants β: α=

kϵ πϵ = 2 ϵ λ ϵ

(7.36)

β=k ϵ

(7.37)

and

In the case of a generally lossy medium, which is the case of saline FYI, equation (7.34) can be written as: ϵ=

ϵ 1 − j tan δ

1 2

(7.38)

where tan δ is called the loss tangent: tan δ =

ϵ ϵ

(7.39)

It is defined as the tangent of the angle in a plane between the lossy and the lossless components of an EM field. Using equations (7.36), (7.38), and (7.39), the attenuation coefficient α can be redefined as [Ulaby, Moore, Fung, 1981, page 225]: 2π α= λ

μϵ 2

1+

ϵ ϵ

2

1 2

1 2

−1

(7.40)

where, μ is the permeability, which is almost equal to 1 in the microwave region for most materials. This is the reciprocal of equation (3.71). The penetration depth is defined as half of the attenuation coefficient: δp =

1 2α

(7.41)

At a wavelength of 1.55 cm (near the operational PM frequency of 18 GHz), the penetration depth is on the order of millimeters for OW and FYI, decimeters for MYI, and meters for dry snow. More data on penetration depth of microwave frequencies are presented in section 9.5. Penetration depth in the snow is important for interpretation of observations from snow-covered sea ice. Snow on sea ice usually stratifies as the snow ages

and becomes exposed to different weather conditions. In general, the microwave emission tends to decrease almost exponentially with the snow depth [Parkinson et al., 1987].

7.3. OPTICAL SENSING About 37% of the solar energy received at the earth’s surface (called irradiance) is reflected back to the atmosphere (called radiance). Most of the irradiance (consequently the radiance) is in the visible band (Figure 7.2). Optical sensor measures reflectance at the top of atmosphere (TOA). The radiance (surface reflectance) can be retrieved from the TOA reflectance measurement after accounting for atmospheric effects, which include absorption, scattering, and emitted radiation. The use of the term “reflectance” in the literature is somehow ambiguous, but a standard reference of reflectance nomenclature is presented in Nicodemus et al. [1977]. Reflectance depends on the local geometry of the surface and the viewing angle of the sensor; therefore, it is a directional function. The surface that exhibits an isotropic scattering pattern is called Lambertian surface. In nature, the rougher the surface the closer its behavior would be to Lambertian scattering, but the reverse is not true. Sea ice surface can be distinctly considered as non-Lambertian. This is mainly because the combined effects of ice mobility and snow metamorphosis make the surface non-uniformly rough with respect to the wavelengths of many satellite sensors, particularly optical sensors. The angular variation of reflectance over the surface area of the hemisphere is known as the bidirectional reflectance distribution function (BRDF). It is the ratio of the reflectance off a surface, in a certain direction, to the would-be uniform (isotropic) reflectance of the same incoming irradiance. BRDF is a frequency-dependent function that describes the angular behavior of surface scattering for a given viewing angle of the sensor. If R and E are the reflected and incident energy, respectively, then BRDF

λ

θ i , ϕi , θ o , ϕo =

R θi , ϕi , θo , ϕo , λ E θ i , ϕi , λ

(7.42)

where, λ is the wavelength of the observed radiation; θi and ϕi are angles of the incoming radiation in spherical coordinates (zenith and azimuth), θo and ϕo are viewing angles of the outgoing (reflected) radiation in the same coordinates. The BRDF characterizes the scattering pattern of the surface. It simply describes the familiar observation when an object looks different when illuminated and viewed from different directions. It assumes that the light striking the surface at some point will be reflected from that same point. In other words, it ignores subsurface scattering and the contribution of the neighboring

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

pixels (i.e., leakage) to the illumination of the given pixel. An idealization of the BRDF is represented by equal distribution of scattering in all directions, i.e., from a Lambertian surface. To estimate the BRDF, reflectance from all angles must be measured for each incidence angle. This can be accomplished using either ground measurements or satellite observations. When using satellite, measurements have to be obtained from multi-date and multi-orbit satellite observations. They can then be used as input into a numerical model to produce the BRDF. Various investigators have proposed theoretical and empirical models [e.g., Wanner et al., 1997]. Rösel [2013] verified the assumption (against a common notion) that the ice surface can be a Lambertian reflector and therefore the BRDF is not crucial for calculations of sea ice reflectance or albedo from optical sensors. Only a few studies of BRDF for sea ice have been accomplished. Li and Zhou [2002] determined the BRDF of FYI and MYI from 512 spectra channels of wavelength between 350 and 1050 nm. Data were obtained during two austral summer cruises in 1999 and 2000 in the Southern Ocean. They found that FYI exhibited stronger surface anisotropy than the MYI because the latter had smoother surface. Moreover, the BRDF pattern of the FYI had a minimum in the backward direction where the pattern for MYI had a peak. The authors attributed the peak to the reflectance from the fore slopes of surface undulation. Heygster et al. [2012] asserted the importance of determining the BRDF for snow-covered sea ice since it is an essential input to the retrieval of the reflectance. The BRDF has limited use in sea ice, but it is important to clarify the notion of the albedo. Albedo is defined as the weighted angular integration of the BRDF over the hemisphere. It varies between zero and one; zero means fully absorbing medium and one means fully reflective. While the above-mentioned definition is widely recognized, the term is loosely used in the literature. Its usage is confused with the reflectance. Obviously, unlike reflectance, albedo is not an angledependent parameter. Therefore, it is a property of the medium. Unless the surface is Lambertian, albedo depends only on the direction of the incoming radiation, which is a major factor in determining the directional distribution of the BRDF. It is the most commonly used surface property derived from space-borne optical sensors. Surface albedo is a regular product from MODIS, calculated accurately on a 16-day cycle basis [MODIS, 1999]. The method is summarized in Luo et al. [2005]. Its main application in sea ice is in discrimination of ice from the surrounding open water. It is also used in the energy balance at snow-covered ice surface and therefore in climate models. Data on albedo from sea ice are presented in section (9.1).

301

Albedo can be presented as an integration over a narrow band of the EM wave spectrum or over the entire optical spectrum (0.4–3.0 μm wavelength). The former is called “spectral albedo” and the latter is called “broadband albedo” or “climatological albedo.” Sometimes broadband albedo is separated into two parts, one is integrated over the visible spectrum (0.4–0.7 μm) and the other integrated over the NIR range (0.7–3.0 μm). When quoted unqualified, it usually refers to the integrated reflection across the visible spectrum. In the absence of an accurate estimate of the BRDF and if the surface is not Lambertian, the estimated albedo from satellite measurements may not be accurate. Lindsay and Rothrock [1994] estimated the total narrow band albedo from reflectance measured by AVHRR augmented by the BRDF for sea ice. Albedo may also be presented as summation of two components: (1) black-sky albedo (also called direct albedo); resulting from directional hemispheric reflectance at a given solar zenith angle, and (2) white-sky albedo (also called diffuse albedo); resulting from bi-hemispherical reflectance under conditions of isotropic illumination). The former represents the angular reflection of the direct radiation that strikes the surface computed at solar noon, while the latter represents the reflection of the diffuse radiation generated across the hemisphere and has the angular dependence removed. Black-sky albedo is calculated by integration of the BRDF for particular illumination geometry. White-sky albedo is calculated by integration of the BRDF for the entire viewing and solar hemisphere. MODIS-derived albedo products include both black-sky and white-sky albedo, each one is provided in the form of spectral and broadband [Strahler et al., 2003]. According to MODIS [1999] document on albedo product, the black-sky albedo αbs and white-sky albedo αws are given by the equations: αbs θ, Λ =

f k k

Λ hk θ

(7.43)

αws Λ =

f k k

Λ Hk

(7.44)

where, θ is the zenith solar angle, Λ is the waveband, fk is the BRDF kernel k model parameter, Hk is the integral of hk(θ) over θ, and hk is the integral of BRDF model kernel k over the view zenith angle ϑ and the view-sun relative azimuth angle ϕ. hk(θ) and Hk are given by the equations: hk θ =

2π π 2 0

K k θ, ϑ, ϕ sin ϑ cos ϑ dϑdϕ

(7.45)

0

and π 2

Hk = 2

hk θ sin θ cos θ dθ

(7.46)

0

where, Kk(θ, ϑ, ϕ) in the BRDF model kernel k and Hk is the integral of hk(θ) over θ. Kernel functions of sea ice are presented in Zhou [2002].

302

SEA ICE

Estimates of snow-covered sea ice albedo in polar regions has been presented in many studies. Model results presented in Maykut and Untersteiner [1971] indicated that a 20% decrease of albedo (i.e., more absorption of sunlight) would cause the disappearance of perennial ice. An important factor that determines surface albedo of polar ice regimes is snow cover. Fresh snow reflects most of the radiation in the optical spectrum [Wiscombe and Warren, 1980, Perovich, 2001]. Therefore, even a few millimeters of snow cover will decrease the penetration of solar radiation to the underlying ice significantly [Allison, Brandt, Warren, 1993]. Losing the snow cover, on the other hand, will decrease the surface albedo and contribute to higher temperatures of the ice. As pointed out before, if only a limited number of reflectance measurements are available (for a few solar incident and satellite viewing angles), some knowledge (or assumptions) about the anisotropy of the surface become necessary in order to estimate the albedo. Some authors employ anisotropic reflectance factors (ARFs) to convert a reflectance measurement to a channel albedo. There are ARFs for various surface types and solar viewing geometries–some derived empirically and some from BRDF models. Simple linear models also exist to further extend channel albedo acquired via narrowband reflectance measurements from sensors such as AVHRR to broadband climatological albedo. De Abreu [1996] used both approaches to derive albedo of Arctic sea ice from AVHRR measurements. In retrieving surface albedo from optical remote sensing observations, the data should be pre-processed to account for atmospheric and cloud influences. Methods for atmospheric correction are mentioned in section 7.8.1. An analytical expression to retrieve the ice surface albedo from radiance measured at the TOA over a Lambertian surface was developed by Vermote et al. [1997] and used in De Abreu [1996]. The radiance LTOA recorded by a satellite optical sensor can be decomposed into two components: radiation reflected from the surface (Ls) and the perturbation contributed by atmosphere due to scattering (La) (known as path radiance): LTOA θi , ϕi , θo , ϕo = Ls θi , ϕi , θo , ϕo + La θi , ϕi , θo , ϕo (7.47) By assuming the surface to be a uniformly Lambertian reflector, the surface reflection can be written as: Ls =

ρs θi , ϕi , θo , ϕo E TOA T cos θ π 1 − ρs θi , ϕi , θo , ϕo s

(7.48)

where, ETOA is the solar spectral radiance incident at the TOA, ρs is the surface reflectance integrated over the sensor’s footprint, T is the total transmittance of the

atmosphere due to scattering and absorption, and s is the reflectance of isotropic light entering from the base of the atmosphere (also called atmospheric spherical albedo). Combining equations (7.47) and (7.48) and by normalizing the radiance received by the satellite with respect to the incident solar radiation (ETOAT cos θ), the reflectance received at the TOA (ρTOA) can be written as: ρTOA θi , ϕi , θo , ϕo = T g ρa θi , ϕi , θo , ϕo +

T t ρs θi , ϕi , θo , ϕo 1 − ρs θi , ϕi , θo , ϕo s (7.49)

where, ρa is the contribution of the atmospheric reflectance measured at TOA, and Tt and Tg are two-way transmittance due to scattering and gaseous absorption effects, respectively. These parameters are examined in detail in Tanré, Holben, Kaufmann [1992]. The unknown in equation (7.49) is the surface reflectance ρs. This can be obtained by rewriting the equation to be in the form: ρs = y 1+ y s

(7.50)

ρTOA θi , ϕi , θo , ϕo − T g ρa θi , ϕi , θo , ϕo T gT a

(7.51)

where, y=

If the reflectance from the atmosphere and the two-way atmospheric transmittance of the radiation are given, then equation (7.51) can be used to retrieve the surface reflectance from the measured TOA reflectance. The transmittance is determined mainly from the scattering and absorption properties of the aerosol, ozone, and water vapor. For a successful atmospheric correction, the ozone amount, water vapor content, and aerosol optical depth should be accurately determined. The albedo of seawater typically falls between 0.04 and 0.08, depending on the incidence angle of the incoming solar radiation [Ivanov, 1979]. This is much lower than the albedo from sea ice or snow, which falls between 0.4 and 0.8 (although small values around 0.08 are observed for dark nilas). Similarly, the reflectivity in the NIR spectrum shows the same contrast between open seawater and sea ice. The NIR reflectivity of OW is near zero, although it increases when the number of suspended particles becomes significant. This usually happens during events of phytoplankton blooms, upwelling bottom sediments, and particle transportation from rivers [Doron et al., 2011]. In March 2002, Winther et al. [2004] measured the spectral reflectance in the VIS and NIR bands from fast ice in Kongsfjorden, Svalbard. They present ice surface albedo as well as the normalized irradiance from seawater at two locations: directly below the ice sheet and one meter below the sheet (Figure 7.14). The surface reflectance from

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

Spectral surface reflection/ Normalized under-ice irradiance

1.0

0.8

Surface reflectance

0.6

Normalized irradiance at 0 m

0.4

0.2

Normalized irradiance at 1 m

0.0 300

400

500

600

700

800

900

1000 1100

Wavelength (nm)

Figure 7.14 Measurements of surface reflectance from sea ice surface, normalized under-ice irradiance taken directly below the ice, and normalized under-ice irradiance taken 1 m below the underside of the ice. Measurements were conducted on fast ice in Kongfjorden, Svalbard, in March 2002 [Winther, et al., 2004 / Taylor & Francis].

sea ice does not change much between the VIS and NIR spectral region (ice was covered by 1–2 cm snow layer; hence albedo values in the figure are the relatively high). However, results show nearly total absorption of the NIR radiation (wavelength >700 nm) by water. One purpose of measuring the under-ice irradiance was to demonstrate the intensities of the Photosynthetically Active Radiation (PAR) (400–700 nm). Figure 7.14 demonstrates that the PAR at 1 m depth is about 50% of the intensities measured directly under the interface of sea ice with water. 7.4. THERMAL INFRARED SENSING There is no universally accepted range of infrared radiation, but it is nominally ranges between 0.7 and 1000 μm. Within this range, the longwave radiation occupies the wavelength range between 4 and 30 μm. The TIR is part of the longwave radiation with typical wavelengths between 8 and 15 μm. It is popularly known as heat energy although absorbed EM in any wavelength can heat the surface. Since TIR sensors can acquire data independent of daylight, they become particularly useful in polar regions during the dark winter seasons. However, similar to the optical sensors, they cannot acquire data about the surface under cloudy sky. Although clouds reflect insignificant amount of TIR originating from the surface, they emit significant amount of TIR. Most of the operating TIR sensors measure emitted radiation in the spectral band 10–12 μm wavelength, which marks the maximum emission from the earth’s surface. It is also part of the atmospheric window (Figure 7.2). For example, the

303

radiometers onboard AVHRR record observations in two bands: 10.3–11.3 μm (channel 4) and 11.5–12.5 μm (channel 5). Similar channels exist on MODIS: 10.78–11.28 μm (band 31), and 11.77–12.27 μm (band 32). The Japanese meteorological counterpart, the MTSAT, carries two channels with 10.3–11.3 μm and 11.5–12.5 μm wavelengths. As mentioned earlier, TIR observations are suitable for discriminating sea ice from the surrounding water because the observed radiation is a strong function of physical temperature. Since the emitted radiation in the TIR is proportional to T4 (see explanation in section 7.2.3), a small change in surface temperature causes a large change in the emitted radiation. For example, as the ice surface temperature in the temperate regions is usually lower than the surrounding OW temperature by only 5% (on the Kelvin scale), the measured Tb from the ice will be lower by 21% [Stringer, Barnett, Godin, 1984]. The difference between surface temperature of ice and water in polar winter can be higher than 20 C, which leads to much higher difference in Tb. Snow cover hardly affects the capability of ice/water discrimination based on TIR data because it has a slightly lower emissivity than ice and its physical temperature is usually not very far from that of ice. This discrimination capability is lost in the summer when the surface temperature of ice and OW becomes almost equal. The most commonly retrieved parameter from the TIR observations is surface temperature (also referred to as skin temperature). Sea surface temperature (SST) can be retrieved fairy accurately from TIR because the seawater emissivity in this spectral region is very stable and close to 1.0 (around 0.98). On the other hand, retrieval of ice surface temperature (IST) can be problematic because of possible large variability of surface emissivity in presence of metamorphosed snow. In both cases, the surface temperature actually means the mean temperature of the radiating layer (i.e., the penetration depth). For TIR, this layer is usually less than one millimeter deep. Interest in IST is not limited to discrimination between ice and water. A climatic record of IST in polar regions is an important indicator for the decade-scale changes in sea ice extent and volume. In order to obtain accurate estimates of IST from TIR data, the observations have to be corrected to account for atmospheric influences. The simplest equation that describes the received radiance L(Tb) in the TIR region can be written following [Yu, Rothrock, Lindsay, 1995]: L T b = ε Lsfc T s τ + Latm + 1− ε τ Latm

(7.52)

where, Lsfc is the radiance emitted from the radiating layer, Latm is the radiance emitted by the atmosphere, Ts is surface temperature, ε is the surface emissivity, and τ is the atmospheric attenuation [equation (7.17)].

304

SEA ICE

The third term in the RHS represents the reflected downwelling of atmospheric radiation off the surface. It is reasonable to assume that in the cold season under clear sky conditions, the polar atmosphere is nearly transparent to the 11.0 μm radiation [Yu, Rothrock, Lindsay, 1995] (i.e., τ = 1) and Latm = 0.Therefore, in this spectral band, equation (7.52) can be reduced to: L T b = εLsfc T s

(7.53)

Since the emissivity of snow-free sea ice or seawater is virtually constant and very close to 1, the estimation of ice surface temperature Ts should be nearly equal to the brightness temperature Tb at 11.0 μm. This approximation is confirmed in Key et al. (1997). Equation (7.53) is the simplest, first-cut estimate of SST and IST when the surface is snow-free or covered with dry snow (dry snow is nearly a blackbody in the TIR region with emissivity around 0.98). Yu, Rothrock, Lindsay [1995] estimated the IST from equation (7.53) using data from the 11 μm AVHRR channel assuming appropriate values of the snow cover emissivity that varied with the scan angle of the sensor. Assumptions include cold and dry ice surface under clear sky. The authors also developed an empirical equation to calculate the IST directly from the AVHRR TIR channel 4: T s = − 4 349 + 0 863T 4

(7.54)

The authors verified the estimates from this equation against measurements from drifting buoys and ice stations in the central Arctic throughout the year of 1989 except May (Figure 7.15). The difference between the two sets –10

Buoy lce station

Ts (°C)

–20

–30 Correlation coeff. = 0.96

T s = a + bT 11 + C T 11− T 12 + d T 11− T 12 sec θ − 1 (7.55)

Mean (Ts–Tair) = 0.0

–40

Stand. dev. (Ts–Tair) = 1.68 (°C) –40

–30

–20

varies between −4.6 C and 3.2 C, and the correlation coefficient is 0.96. The accuracy from using equation (7.54) or equation (7.53) was found to be the same. These initial results confirmed the possibility of using the 11 μm TIR channel for estimating IST using a simple equation provided that no atmospheric influence is “contaminating” the observed T b Comparison between brightness temperature observations from 11 μm and 12 μm channels or the derived surface temperature from each channel using any method provides an acceptable assessment of the effects of the atmosphere. If the difference is smaller than 0.39 C, then clear atmosphere can be assumed [Riggs, Hall, Ackerman, 1999]. In the presence of atmospheric influences, a more accurate estimate of the IST requires correction of the satellite observations for attenuation of the observed radiation. This is caused mainly by water vapor and partly by several atmospheric constituents (e.g., carbon dioxide, nitrogen oxide, methane, ozone). A commonly used technique that corrects for atmospheric influences in calculating surface temperature is called the “split window.” It was originally suggested by Prabhakara, Dalu, Kunde [1974] and first used by McMillin [1975] to obtain accurate estimates of land surface temperature. The technique estimates surface temperature using the 11μm and 12μm channels because the 12 μm is more sensitive to water vapor than the 11 μm channel. Therefore, the difference between them is a function of the absorptive response of the water vapor content in the atmosphere. The more water vapor contents, the higher the difference (or the ratio) between observations. The split window technique incorporates a linear or a nonlinear equation that combines the observations from the two channels to determine the surface temperature Ts. Several studies conducted in the early 1990s employed different forms of the split-window equation using TIR data from AVHRR and other instruments [Key and Haefliger, 1992, Lindsay and Rothrock, 1993, Massom and Comiso, 1994, Yu, Rothrock, Lindsay, 1995]. The following form was used in Key et al. (1997). It provides better accuracy because it accounts for the atmospheric effects as well as the sensor’s scan angle.

–10

Tair (°C)

Figure 7.15 Plot of surface temperature derived from AVHRR using equation (7.54) (solid line) and the measured surface air temperature from buoys and ice stations. Data obtained in the central Arctic throughout 1989 (except May) [Yu, Rothrock, Lindsay, 1995, Figure 3 / with permission from AGU].

where, T11 and T12 are the measured brightness temperatures in the 11 and 12 μm channels, respectively, θ is the sensor scan angle (at each pixel) that accounts for the path length of the radiation in the atmosphere, and a, b, c, and d are regression coefficients. Accuracy of surface temperature estimation from the above equation ranges between 0.5 and 1.5 K, depending on the accuracy of determining the coefficients [Key et al., 1997].

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

The coefficients in equation (7.55) are usually determined using measurements of Ts regressed against estimates from using the RHS of the equation. T11 and T12 can be obtained from satellite measurements after corrections for atmospheric influences. Alternatively, they can be obtained using modeled brightness temperature (in this case there is no need for the atmospheric correction). A commonly used radiative transfer model for this purpose is LOWTRAN (LOW-resolution atmospheric TRANsmission). This is a computer code for predicting atmospheric radiance and transmittance for a given surface condition, including composition and temperature, as well as the characteristics of local atmosphere [Kneizys et al., 1989]. The input atmospheric parameters are usually obtained from measurements using a radiosonde or a drifting ice station. The model also needs a value for local surface emissivity at the point where brightness temperature is to be estimated. The emissivity is often assumed, but it can be a source of error if the surface includes a highly compositional mix of ice and metamorphosed snow. Climatological emissivity data can be used though the data are less commonly available in TIR than in microwave region. 7.5. MICROWAVE REMOTE SENSING In general, passive and active microwave remote sensing are the most commonly used tools to monitor sea ice in the polar regions because of their ability to image the surface in the absence of sunlight. They are also unaffected, to a large extent, by atmospheric conditions especially when the wavelength of the radiation exceeds 10 mm. Table 7.2 includes the frequency and wavelength of the currently operational microwave sensors (active and passive). PM sensors measure emitted radiation while active sensors measure the backscattered power of the radar signal after radar signal strikes the surface. Table 7.2 Frequency and wavelength of operational microwave sensors (active and passive).

Active

Passive

Band symbol

Frequency (GHz)

Wavelength (cm)

P L S C X Ku K Ka Ka Q W

0.3–1.0 1.0–2.0 2.0–4.0 4.0–8.0 8.0–12.5 12.5–18.0 18.0–26.5 26.5–40.0 18.0–19.0 36.0–38.0 85.0–90.0

100–30 30–15 15–7.5 7.5–3.75 3.75–2.4 2.4–1.67 1.67–1.13 1.13–0.75 1.66–1.58 0.83–0.79 0.35–0.33

305

Unlike TIR radiation, which is influenced by the physical temperature of the object, PM radiation is affected by the emissivity of the radiating layer. The emissivity in the microwave region is influenced by physical composition and properties of the medium such as salinity, moisture content, surface roughness, and crystalline structure. As for sea ice, some of these parameters such as salinity and geometrical characteristics of brine pockets are primarily triggered by variations in temperature. Therefore, in addition to being an explicit parameter in determining the emitted radiation, ice temperature influences most of the parameters that affect the emissivity and consequently the emitted radiation. Within the radiating layer, the radiation may be extinct through scattering by inclusions that have different dielectric constants than the host material (i.e., ice crystals). It can also be scattered at the surface if it has appropriate roughness scale. Figure 7.16 shows emitted radiation scattered when refracted across a rough boundary between two dielectrically mismatched media. It also shows the coherent summation of scattered signals that meet at a given point at the surface before getting emitted in the direction of the satellite observation θr. In the microwave region, ice crystals emit higher energy than water molecules; i.e., the emissivity of ice is significantly higher than that of water. As mentioned before, this is the premise for using PM sensing for discriminating sea ice from surrounding seawater. Moreover, the gradual evolution of emissivity from the low values of OW to the higher values of ice can also be utilized to identify thin ice and estimate its thickness. Stringer, Barnett, Godin [1984] compared microwave brightness temperature from OW and sea ice using the following numerical example. For wavelength 1.5 cm (nearly 19 GHz) the brightness temperature from OW at 273 K with emissivity 0.5 is 136 K, and for FYI at 250 K with emissivity 0.92 it is 230 K [from using equation (7.28)]. Therefore, although ice is physically colder than the surrounding OW, it emits more energy in the microwave band because of its higher emissivity. Hence, ice is considered to be radiometrically warmer than water. Operational frequencies of PM sensors (Table 7.2) are selected within the range 4.9–94 GHz such that spectral regions of high atmospheric opacity (20–24 GHz and 40–80 GHz) are avoided. Those frequencies are also suitable for sea ice applications in the polar regions because they are associated with low radiation emission from clouds, especially at the lower frequencies. As mentioned earlier, the weak radiation emitted from the earth’s surface in the microwave region results in coarse spatial resolution from PM sensors (a few kilometers or tens of kilometers), and this does not allow resolving information on a small scale to identify narrow features such as leads with a few hundred meters in width. However, the wide swath of the data is useful for regional and global ice

SEA ICE (a)

(b) Scattered radiation triggered by the surface roughness θr air

Si Tb by ngle vo be lum am e s tr ca igg tte ere rin d g

306

ice θi

Figure 7.16 (a) Emitted radiation across a rough boundary between two media of different dielectric constants. (b) Volume emitted radiation contributing to a single brightness temperature observation at the viewing angle θr of the receiving antenna.

Satellite position

Al

on

g

sc

an

an

c ss

os Cr

2 1

Track line

3

Along track

monitoring. PM remote sensing has provided the most comprehensive long-term daily observations of sea ice in all geographic regions. Recent operational microwave sensors that have been used widely in sea ice applications include SSM/I, AMSR/E, and AMSR2. Details about their radiometric and geometric characteristics can be found in Hollinger, Lo, Poe [1987], Kawanishi et al. [2003] and Kachi et al. [2008]; respectively (more sensors are listed in section 8.3). The radiometers on those sensors are conical, which look backward and scan the swath at a constant angle (53 in case of SSM/I). The scan lines draw non-concentric arcs which tend to converge more toward the edge of the swath as the distance from the ground track of the sub-satellite point increases (Figure 7.17). This viewing geometry produces footprints in the form of ellipses with minor and major axes in the direction of scan lines and across scan lines, respectively. More overlap of the footprints (also called IFOV) is visible toward the end of each scan line. The footprint dimensions from commonly used channels of PM sensors are included in Table 8.3. However, data are usually provided to users in gridded format such as the equal area projection of EASE grid, which is suitable for polar regions [Brodzik et al., 2012]. This avoids the issues of footprint overlap although it takes from the integrity of the data due to data sampling involved in this process. The sampling of the measured radiation can be achieved at a rate finer than that of the footprint of the radiometer. This is common when generating gridded data. For example, the National Snow and Ice Data Center (NSIDC) resample the SSM/I 19 and 37 GHz data at 25 km X 25 km grid spacing. This is smaller than the IFOV dimensions of the same channels as shown in (Table 8.3). Therefore, the observation at each grid point is produced from about 7 to 9 overlapped footprints in the case of the 19 GHz channel and 4 to 5 footprints in the

Scan angle 3 2 1 IFOV

Grid point

Figure 7.17 Scanning geometry of SSM/I showing footprints of a given radiometric channel (not-to-scale). Data are sampled over 104.2 from end-to-end per scan line.

case of the 37 GHz channel (more pixels are averaged near the end of the swath than at nadir). This is how the integrity of the data is compromised due to averaging measurements from several footprints, let alone with the possibility of overlapping. On the other hand, data from the 85 GHz channels were sampled at 12.5 km pixel spacing. This is very close to the dimensions of IFOV of that channel (13 km X 15 km as shown in Table 8.3). Therefore, only a slight overlap exists between adjacent pixels. The observed radiation by a PM sensor (presented in the form of Tb) is the summation of four sources (Figure 7.18): surface radiation, upwelling atmospheric

307

Space

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

Atmosphere

2

Emission

1 3 4

Absorption Reflection

Figure 7.18 Sketch showing the components of the observed radiation by a PM sensor. The component numbers 1, 2, 3, and 4 represent the emission from the surface, the upwelling radiation from the atmosphere (Tup), the reflected downwelling radiation (Tdown − ref) and the reflected space radiation (Tsp); respectively.

radiation, reflected downwelling atmospheric radiation, and reflected space radiation. The algebraic form of the simplest radiative transfer equation (RTE) that describes this summation is the basis for many algorithms of sea ice geophysical parameter retrieval from PM remote sensing. T b = εRL T RL e −τ + T up + T down−ref 1−ε e −τ + 1−ε T sp e −2τ

Assuming constant temperature and atmospheric attenuation profiles in equations (7.57) and (7.58), and substituting the results of the integration back into equation (7.56), a simplified form of the RT equation (7.56) can be re-written as: T b = T RL ϵRL e − τ + T ∗air 1− e − τ + T ∗air 1− ϵRL 1− e − τ e − τ + T sp 1− ϵRL e − 2τ

(7.56)

(7.59)

where, Tb is, as usual, the observed radiation, εRL and TRL are the emissivity and the average physical temperature of the radiating layer, τ is the total (integrated) atmospheric opacity, which is wavelength-dependent, Tup is the upwelling atmospheric radiation, Tdown − ref is the reflected downwelling radiation, and Tsp is the average temperature of free space. Tup and Tdown − ref are given by the equations:

Here, T ∗air is the effective physical temperature of the atmospheric radiating layer (theoretically it should be the entire atmospheric column). This parameter is usually defined as a weighted average atmospheric temperature in the lower troposphere, Tair [Gloersen et al., 1978]. Svendsen et al. [1983] used the following approximation, which is considered to be sufficiently good, given the uncertainties in the value of the atmospheric opacity.

τ

T up =

T z e−τ + τ

z

dτ z

(7.57)

0 τ

T down − ref = k T z e − τ

z

dτ z

(7.58)

0

where T(z) is the temperature of the atmosphere at height z and τ (z) is the atmospheric opacity from the surface to a height z. The constant k in equation (7.58) is introduced to account for the approximation of the diffuse reflection from rough surfaces as being along the line-of-sight. Its value depends on the distribution of τ and the degree of isotropy of the diffuse reflection at the surface [Gloersen et al., 1978]. Strictly speaking, the integrals in equations (7.57) and (7.58) should be carried out over the entire hemisphere.

T ∗air = 0 9 T air

(7.60)

Since Tsp is extremely small (approximately 2.7 K), the fourth term in equation (7.59) is usually neglected. In the case of thick ice (say, ice thicker than 30 cm), TRL can be approximated as 0.9 Tair, but this is not always a valid assumption. Given an observation of Tb, equation (7.59) can be solved for TRL or ϵRL. If TRL and Tair can be parameterized or related by an equation, then equation (7.59) can be solved for ϵRL. Svendsen et al. [1983] suggested the following relation: T RL = αT air + 272 1− α

(7.61)

The authors obtained Tair using drifting buoys in an area north of Ellesmere Island in the Arctic. They estimated α to be roughly 0.4 from using scanning

308

SEA ICE

multi-channel microwave radiometer (SMMR) measurements. Zwally et al. [1983] used different equation: T RL = T air + f T w− T air

(7.62)

where, Tw is the freezing point of seawater and f is an empirical parameter determined using Tb. Zwally et al. [1983] examined ice concentration values derived from ESMR data and adjusted f’ until the correct concentration value was reached. They found that 0.25 is a reasonable value for use in both Arctic and Antarctic regions. In reality, the magnitude of f varies with the thickness of the ice and snow. Due to their coarse resolution (a few kilometers or tens of kilometers), footprints of space-borne PM radiometers are almost always heterogeneous. A basic linear model that decomposes the observed brightness from a heterogeneous footprint into its components from each surface is commonly used. If a footprint is composed, for example, of OW, FYI and MYI with concentrations COW, CFY, and CMY, respectively, then the observed Tb can be decomposed into three components from the three surface types as follows: T b = C OW T b,OW + C FY T b,FY + C MY T b,MY

(7.63)

where, Tb,OW Tb,FY, and Tb,MY are typical brightness temperatures (called tie points) from the relevant surface type. The above equation is used in many algorithms for retrieval of ice concentration. The estimation of the tie points is usually achieved by sampling brightness temperature from areas of uniform surface cover (ice type or OW). However, some ice types such as YI exhibit a wide range of Tb under different atmospheric temperatures or other meteorological conditions (see Figure 9.11). Emissivity from OW is a weak function of water salinity and temperature though it is affected by ocean foam, winddriven sea surface roughness, and rainfall. This broadens the range of Tb from sea water. Representing such surface by a single tie point may not be appropriate in this case. An appropriate approach would be to consider the variability of all possible Tb caused by meteorological, snow and ice deformation factors, within each surface type (see the method in section 11.2.2.4). Over consolidated ice fields in the Arctic, the atmosphere is usually dry with prevailing stratus clouds that contain mostly ice crystals. These clouds are almost fully transparent in the microwave region. However, Gloersen et al. [1973] mention that clouds increase the PM satellite observation from Arctic ice by about 7 K.

7.6. IMAGING RADAR SENSING While PM sensors, with their wide swath and frequent temporal coverage, are suitable for synoptic-scale polar

sea ice monitoring, active microwave sensors (i.e., radar systems) are more suitable for tactical scale monitoring at finer spatial resolution (tens or hundreds of meters), which is required to facilitate marine navigation. The term radar was coined in 1941 by the US Navy as an acronym for radio detection and ranging. Radar systems are traditionally used for object detection (i.e., aircrafts, ships, guided missiles, heavy rain, etc.). In remote sensing, there are three types of radar systems, all used in sea ice applications: imaging radar, non-imaging radar (scatterometer), and radar altimetry. Obviously, imaging radar generates imagery data, but these data can also be generated from the scatterometer systems. This section addresses the imaging radar technology while scatterometer and altimeter systems are addressed in sections 7.7 and 7.8, respectively. Imaging radar systems measure the part of the radiation scattering from a ground target/surface that travels back to the receiving antenna, i.e., the backscattering. If the same antenna is used to transmit and receive the data, the system is called monostatic. Generally speaking, strong backscatter is engendered by one or more of the following three sources: (1) a surface that features strong dielectric mismatch with the overlying medium (e.g., saline sea ice underlying air), (2) rough surfaces, and (3) volume with numerous scattering elements (e.g., air bubbles in the crystalline lattice of MYI). Surface scattering is usually associated with a mechanism called single-bounce of the incident signal while volume scattering is associated with multiple-bounces. More details on this point are presented in section 7.6.1.3. Information on the commonly used imaging radar systems with their operational modes is presented in section 8.4. Data from those systems are the prime source of information for sea ice monitoring. 7.6.1. Imaging Radar Principles An imaging radar system transmits pulses from its antenna to illuminate strips of the surface, covering the entire swath, at one side of the orbit direction, i.e., left or right. For example, RADARSAT-1 antenna was right-looking and RADARSAT-2 has the two options of either right- or left-looking. Figure 7.19 shows a single pulse scanning a single strip on the ground scene. This configuration applies more of the RAR as explained later. The pulse repetition frequency is designed to give an appropriate ground resolution of the imaged area in the cross-track direction. When the transmitted pulse scans the scene across the swath, it interacts with the ground cover causing a fraction of the incident energy to be scattered off the surface back to the antenna. This is the backscatter received by the same antenna and its power (Pr) is related to the transmitted power (Pt) through what is known as radar equation [Ulaby, Moore, Fung, 1982 and Elachi, 1988].

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE An

309

ten

na Side lobes

e

ls

ed itt

pu

sm

a th Sw

an Tr

Main lobe

Rr Rt

an

ine

l

Sc

Figure 7.19 Transmitted radar pulse from a right-side-looking airborne or space-borne antenna. Each pulse scans a line across the swath.

σ

Radiating pattern

Scattering surface

The early version of imaging radar systems was the airborne RAR. In this system the backscatter signal from each transmitted pulse is processed separately. The spatial resolution across the scan line (range direction) is usually fine (a few meters), but the resolution along the azimuth direction (normal to the scan line) is very coarse. The next generation of imaging radar, SAR, was developed to overcome the drawback of the coarse azimuth resolution of RAR. The following section discusses the derivation of the radar equation, which applies to both radar systems. Equations of the spatial resolution of each system are also given. SAR has been the prime tool for sea ice monitoring, since the beginning of 1990s. 7.6.1.1. Radar Equations and Spatial Resolutions of RAR and SAR Figure 7.20 shows the geometry of transmitted and received radar signals and a few parameters involved in the radar equation. For convenience of interpretation, the transmitted and received points are shown at different locations on the graph, but in all operating SAR systems they are located at the same point (i.e., the same antenna). This type is referred to as monostatic radar system. For “point target,” i.e., a strongly scattering target located against a non-scattering background, the radar equation for the received power Pr usually takes the forms Pr =

Pt 1 Gt σ Ar 4πR2t 4πR2r

(7.64)

The RHS of the equation is composed of three terms. The first represents the power distribution of the transmitted power Pt at distance Rt from the transmitting antenna following the antenna pattern Gt. As shown in the figure,

Figure 7.20 Illustration of a radar transmitted pulse, its interaction with a surface and the backscatter that travels back to the receiving antenna. Parameters used in the “radar equation” are also shown.

the pattern consists of a main lobe as well as side lobes. The power intensity is inversely proportional to the square of the distance and is determined by the antenna gain within Gt. The third term represents the power of the received backscattering. The term Ar is the effective aperture of the receiving antenna and Rr is the distance from the scattering element back to the antenna. The middle term, σ, represents the effect of the ground target on the balance of power between the transmitted and received signals. This term has units of area because the first term in equation (7.64) is power per unit area and the third term is dimensionless. Hence, σ is called radar cross section. Conceptually speaking, the transmitted power is intercepted by the “effective receiving area” of the ground target Ars. When this happens, the target absorbs a fraction of it, denoted pa. This excites electric currents on the scattering elements within the target, which then becomes an antenna re-radiating with its own antenna pattern Grs. The re-radiating power is measured by the radar cross section. Therefore, σ takes the form: σ = Ars 1− pa G rs

(7.65)

This equation presents σ as being the area that characterizes the target in terms of its ability to scatter radar signal. As the equation implies that the scattering from the target is isotropic, σ is considered to represent the cross

310

SEA ICE

section of an equivalent isotropic scatterer that would be required to generate the same power density as observed by the receiver. One of the problems of using the concept of radar cross section is its dependence on the surface area of the imaged footprint (i.e., the ground resolution of the sensor). Therefore, when radar sensors of different resolutions view the same surface then σ may be different. A more useful quantity is obtained by dividing σ by the area of the observed terrain. This parameter is called backscatter coefficient or “sigma naught” σ o. It is a dimensionless quantity, which can be treated as an intrinsic property of the surface. Moreover, if the ground target is composed of a large number of distributed scattering elements as opposed to a single element, then the integration of scattered differential power from all scattering elements within the resolution cell can be written as: pr =

Pt G t o Ar σ ds 2 4πR 4πR2r Ao r

(7.66)

This is a model for distributed target. This equation shows that the backscatter coefficient is a statistical measurement that represents the average power returned from the distributed scattering elements. It can be conceived as the ratio of the statistically averaged scattered power density to the average incident power density. Backscatter coefficient is determined by the surface roughness, the physical and electrical properties of the scattering elements as well as the radar parameters; namely the wavelength, polarization, and the local incidence angle of the incident signal. Since values of σ o are usually very small, they are presented in decibels (dB) rather than linear scale. When averaged over a number of pixels (sampling area), the average must be calculated from the values in the linear power scale and then converted to dB using the equation: σ 0 = 10 log 10 σ 0linear

(7.67)

The standard deviation of σ oover the sampling area is expressed in dB using the following equation: SDσ 0db = 10 log 10

Mean σ 0linear − SD σ 0linear Mean σ 0linear

LSA = θa R = λR L (7.68)

The resolution of the RAR system in the range direction Xr and azimuth direction Xa are given by the following equations. The derivations can be found in several text books [e.g., Elachi, 1988]. cτ Xr = sin θr (7.69) 2 X a = λh L cosθ

where, τ is the pulse width, c is the spepd of light, θr is beam width of the transmitted signal in the range direction, λ is the wavelength of the transmitted signal, h is the altitude of the platform, L is the antenna length, and θ is the incidence angle at the middle of the swath. According to equation (7.69), for a pulse width of 5 × 10-8 second and θr= 20 , the resolution in the range direction is Xr= 22 m. This is reasonable. On the other hand, Xa calculated from equation (7.70) can be very large. For h = 800 km, λ = 23 cm (L-band), L=12 m and θ = 20 , Xa= 16.3 km. This is an extremely coarse resolution. In order to have Xa = 100 m, an antenna length of 1958 m is required. Of course, this is not practical and that is what prohibits the use of RAR when fine resolution in the azimuth direction is required. SAR is based on the concept of combining signals from all radar beams that view the same ground resolution cell (or same single target) as long as it remains “visible” by successive beams. The collection of these beams starts with the first beam that “sees” the ground cell and ends with the last beam that sees it. An important fact that allows the utilization of this concept is that radar pulses travel between the radar antenna and the target at the speed of light, which is five orders of magnitude faster than the speed of the satellite. Therefore, several thousand echoes can see the same ground cell/target as the satellite passes over it. Obviously, this requires a relatively high pulse repetition frequency such as 1400 Hz used in the SAR systems onboard the European Remote Sensing Satellites ERS-1 and ERS-2. Figure 7.21 shows the string of dots representing positions along the SAR satellite path at which radar pulses are transmitted. The first and last beams that illuminate the marked ground target are also shown. A few thousand transmitted pulses usually detect (or simply “see”) the same ground cell/target along the satellite path. As the platform continues to move forward, all echoes from those successive pulses are recorded. The distance between the position of the first and last pulse that detects the target is the length of the synthetic (virtual) antenna. This is denoted LSA in Figure 7.21 and is given by the following equation:

(7.70)

(7.71)

where, L is the physical length of the antenna and θa is the angular bandwidth corresponding to the synthetic aperture length. The exposure time Te for a given target to remain within the successive illuminating beams is given by the following expression, where V is the speed of the satellite: T e = LSA V = λR LV

(7.72)

In SAR systems, both the target’s position and ground resolution in the range direction are determined from the

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE th

re er tu

p

ca heti

t Syn

Aperture azimuth beamwidth (θa)

L SA

leng

311

V e ang et r g r Ta (R)

h θa

k trac

L SA

Sub

ath Sw

L SA

Ground target resolution element

Figure 7.21 Concept of SAR is demonstrated by nine positions of antenna transmitting beams (the dots along the satellite track). The distance over which the beams illuminate the ground target is the length of the synthetic aperture LSA. It can be considered as a virtual antenna.

return time of the echo; i.e., similar to RAR. Hence, Xr is obtained from equation (7.69). On the other hand, the resolution in the azimuth direction is determined using one of the following two approaches, which lead to the same result. The first approach, called synthetic array, assumes a fictious antenna length (LSA in Figure 7.21), which amounts to a very small resolution. The second approach is based on the concept of Doppler frequency shift. Here, the backscatter signal from each transmitted pulse within the synthetic aperture is recorded in terms of intensity and phase shift with respect to the transmitted signal (the successive transmitted pulses are coherent as will be pointed out shortly). The range of the Doppler shift from the first and last pulses that “see” the ground resolution is recorded and used to determine the ground resolution in the azimuth direction. Both approaches yield the following expression for the azimuth resolution: Xa = L 2

(7.73)

The striking feature about this expression is that the azimuth resolution equals half of the physical antenna length and is independent of the altitude of the platform. More information about SAR and the derivations of the expressions of spatial resolutions in the range and azimuth directions can be found in a few reference books [Ulaby, Moore, Fung, 1982, Elachi, 1988, Culander and McDonough, 1991, Jansing, 2021]. It is worth mentioning that while backscatter is usually recorded in terms of the power ratio of the received to the transmitted signal (i.e., σ o), the recording of the amplitude and the phase shift is achieved through a particular mode called single look complex (SLC). Here, the pixel value is presented in terms of a complex number that includes both the amplitude and the phase of the received signal.

The phase carries information on the imaged surface because the transmitted radar pulses are coherent as explained below (the phase from the reflected or emitted signal from any passive sensors is not useful because the incident or the emitted signal is incoherent). In the SLC product the position of the pixel is defined along the slant range (i.e., the line connecting the SAR platform and ground pixel), not the ground range. In the case of RADARSAT-1 and RADARSAT-2, SLC has been available from all beam modes except the ScanSAR (Figure 8.2), which is the most suitable mode for ice monitoring operational requirements. 7.6.1.2. Coherency and Polarization of Radar Signals Radar pulses generated by SAR are coherent, meaning that they have the same phase. This feature makes measuring the phase of the return signal with respect to the “reference” phase of the transmitted signal meaningful. The phase can be part of the information that identifies the type, geometric characteristics and elevation of the observed surface. It is the essential parameter used in estimating the elevation of the imaged surface, a technology known as SAR interferometry technology. Since no applications of this technology exist in the field of sea ice, it will not be pursued in the following discussions. The phase measurement is retained only in the SLC product. In sea ice applications, SLC data are used only within the polarimetric SAR acquisitions as will be shown later in this section and the phase (part of the complex number identification of the pixel) is used in deriving all polarimetric parameters as explained in section 7.6.3.1. In addition to their coherency, the transmitted radar pulses are also polarized. This is far more useful feature for ice application than the coherency. As mentioned in

312

SEA ICE

V

H

Vertically- oriented target

section 7.2.1, a polarized EM wave has a predictable timevarying alignment of its electric field vector (i.e., it may prescribe oscillations in a fixed plane or predictable time-varying plane). Upon scattering off the observed surface, the radar signal often exhibits a change of polarization. Therefore, the received signal may be partially polarized and the degree of polarization can be determined from elements of the Stokes vector as shown in equation (7.3). The degree of polarization constitutes part of the information that can be used to identify the surface. This can be achieved by recording the received backscatter in two polarization components, usually (but not always) in the same polarization plane as that of the transmitted pulses and the orthogonal polarization plane. Therefore, the received signal is denoted by a pair of symbols, the first refers to the polarization of the transmitted and the second to the polarization of the received signal. Hence, HH means that the detected signal is from horizontally polarized transmitted wave and horizontally polarized scattered wave. HH and VV are called copolarization signals and HV and VH are called crosspolarization signal. The reciprocity theorem dictates that the backscatter in HV and VH are equal. Interpretation of the radar images in terms of their polarization requires understanding of two processes: (1) how the vertical and horizontal polarizations differ in their interactions with the surface and (2) what causes the depolarization of the scattered signal. Vertical and horizontal polarization signals interact differently with vertically oriented surface structures. Figure 7.22 illustrates these interactions. As vertical scattering elements act as dipoles, vertically polarized signals will interact strongly with vertical structures and produce relatively high backscatter in the vertical co-polarization (σ ovv ).

V

H

Figure 7.22 Interaction of vertical and horizontal polarization signals with vertically oriented objects. The vertical polarization signal interacts strongly with these objects but the horizontal polarization does not; instead, it interacts better with the underlying ground surface.

On the other hand, horizontally polarized signals do not interact with vertical structures as much as they do with the underlying surface, resulting in low backscatter in the horizontal co-polarization (σ ohh ). The difference between σ ohh and σ ovv is, therefore, used to identify objects that stand vertically. This is particularly useful in agricultural and forestry applications. Information about the canopy is usually embedded in σ ovv , while information about the underlying soil can be retrieved from σ ohh . In sea ice applications, the difference between the two copolarization backscattering coefficients has been observed in the return from rough seawater surface. Here, σ ovv from a rough water surface is significantly higher than σ ohh and may overlap with the backscatter from sea ice. This observation was made by SAR image analysts at the Canadian Ice Service in early 1990s, when they started to use the σ ovv from ERS-1 with its high backscatter from rough water surface. It confused them as they were used to the much lower σ ohh backscatter from the same water surface, which was available from the prime operational airborne SAR system since 1987. This was an X-band HH polarization called Sea Ice and Terrain Assessment (STAR), operated by Intera Technologies Limited of Calgary, Alberta, Canada. 7.6.1.3. Radar Scattering Mechanisms When a radar pulse strikes a surface, scattering is triggered through one or more of the following four mechanisms: (1) single-bounce (SB), (2) double-bounce (DB), (3) multiple-bounce (MB), and (4) helix scattering. Sea ice may trigger one or more of the first three mechanisms, while helix scattering is relevant to complicated manmade structures in urban areas, which is not relevant to the sea ice applications. Figure 7.23 illustrates the first three mechanisms and their associated phase shift ϕ between the incident and scattered signals. It uses general geometries but geometries relevant to sea ice are presented in Figure 7.26 (section 7.6.3.3). The surface associated with the single-bounce scattering mechanism should be smooth with respect to the wavelength of the incident signal. The surface associated with multiple-bounce scattering should be very rough (deformed) where the signal keeps bouncing off several facets. Multiple bounces can also occur within the volume where the signal keeps bouncing off numerous scattering elements. Double-bounce mechanism is triggered by dihedral structure. Different names of those mechanisms may be used in the literature. For example, odd-bounce can be used instead of single-bounce and random scattering can be used instead of multiple-bounce. In the rest of this text the names single-bounce, double-bounce, and multiplebounce are used, with acronyms SB, DB, and MB, respectively. However, the terms surface scattering and volume

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE ϕ ≅ 180°

Single-bounce scattering (surface scattering)

313

0° < ϕ < 180°

ϕ ≅ 0°

Double-bounce scattering

Multiple-bounce scattering (volume scattering)

Figure 7.23 Illustration of the three scattering mechanisms identified with radar polarimetric data.

scattering are occasionally used to further illustrate a point in the text. The SB scattering is usually triggered by a smooth surface such as level sea ice, quasi-steady ocean surface, a sphere or reflector of trihedral type. Hence, it is expected to be in the form of quasi-specular reflection. The term SB mechanism is more accurate than surface scattering because highly deformed surfaces may engender SB and/or MB mechanism(s). While SB mechanism usually features a single-bounce, it can feature a small odd number of bounces. This mechanism maintains the same polarization as that of the incident signal. Hence, high co-polarized backscatter should be expected. The DB scattering is triggered by a dihedral surface such as vertical trees or buildings. For sea ice, this structure is manifested in ridging or (to a lesser extent) highly deformed surface. Similar to the SB, this mechanism does not depolarize the signal but it reverses its orientation. The MB scattering mechanism is produced by a surface with structure (e.g., forest, urban areas, ridged or heavily deformed ice) or heterogeneous volume with numerous scattering elements. This can be found in the bulk volume of a forest or vegetation fields. In the case of sea ice, it is usually generated at a surface that features a large number upturned ice blocks (heavy rubble field), or from within a volume that numerous scattering inclusions with size compatible with the wavelength of the incident radar signal. An example is the air bubbles within the subsurface of MYI hummocks. The MB mechanism depolarizes the incident signal, resulting in high cross-polarized backscattering. The degree of depolarization depends on the degree of the intensity and randomness of the scattering elements. Power from the three mechanisms can be calculated using a set of parameters called Yamaguchi decomposition parameters (section 7.6.3.3) It is possible that more than one scattering mechanisms may emerge from the same ice type, depending on the scale of surface roughness or scattering inclusion elements with respect to the wavelength of the incident radar signal. For example, a rough surface that generates MB scattering from an X-band signal ( 3 cm wavelength) might be

considerably smooth for an L-band signal ( 23 cm wavelength) and hence generates dominant SB scattering. Less MB (volume scattering) from MYI is observed from the L-band than C-band because the wavelength of the former is considerably larger than the typical size of the air bubbles (the main scattering element in MYI). Since the incident radar signal is always coherent and fully polarized, depolarization of the radar backscatter from any surface is always positive, i.e., showing a lesser degree of polarization with respect to that of the incident signal. The odd- and double-bounce mechanisms are usually called deterministic as opposed to the random scattering. Some ice types such as new ice and MYI features a dominant scattering mechanism (namely MB in this case). More on this point is presented in section 9.3.3 and Table 9.8. The scattering power from the SB, DB, and MB mechanisms are denoted as Ps , Pd , and Pv , respectively in Table 9.8. 7.6.2. Multichannel SAR The legacy of the first space-borne SAR onboard the short-lived satellite Seasat (June–October, 1978) had created an atmosphere of hope within the sea ice community. Seeing images of the extensive coverage of sea ice in the polar regions for the first time was fascinating. It was hoped, at that time, that operational ice parameters such as ice type and concentration could be retrieved from SAR data. As the research studies progressed, it turned out that what was possible to estimate from visual interpretation of the images might not be retrievable using quantitative analysis of a single-channel SAR data. Except for ice tracking algorithm, no other ice parameter could be retrieved using single-channel SAR data in a way to meet the accuracy and robustness requirements of operational ice monitoring programs. The main obstacle was the overlap of backscatter signatures between different types [Mäkynen and Hallikainen, 2004] and the wide range of possible backscatter from certain ice types, particularly YI. The latter is caused by a few factors that include: (1) fluctuations of atmospheric temperatures around the

314

SEA ICE

freezing point during the early stage of ice formation, (2) different forms of precipitation falling at the surface, (3) metamorphism of snow cover, and (4) dynamic ocean state in marginal ice zones. Under certain conditions of metamorphosed snow, the backscatter is generated mainly from within the snow pack, thereby masking the underlying ice. Moreover, ambiguity between backscatter from ice surface and wind-roughened OW surface complicates the discrimination between these two surfaces. The only ice type that can be identified using a single-channel SAR is MYI. This type gives relatively high backscatter in both co- and cross-polarization triggered by the intensive volume scattering within the bubbly layer near the surface of hummock ice. MYI can be unambiguously identified using C-band and to a less extent from X-band SAR observations. The discrimination is more successful under the dry and cold Arctic winter conditions [Smith, Barrett, Scott, 1995, Shokr, Ramsay, Falkingham, 1996, and van der Sanden, 2004]. Using the L-band has proven to be less successful in identifying MYI because its wavelength (15–30 cm) is an order of magnitude larger than the dimension of the scattering air bubbles. In addition, single-channel SAR can be useful in discriminating sea ice from open water when the water surface is quiescent and the ice is cold and dry. In this case the backscatter from water surface becomes significantly lower than that from sea ice. By late 1990s there was a consensus among the sea ice remote sensing community that identification of ice types and accurate retrieval of ice parameters from SAR required multichannel data. Three different approaches have been suggested to increase the dimensionality of the data using multi-frequency, multi-polarization, or adding texture to account for the spatial context of the backscatter. There is a fourth approach calling for use of multi-temporal observations to identify surfaces that change with time, but this is not suitable for retrieving sea ice parameters since ice is mobile. However, aside from the issue of increasing the dimensionality of the data, sequential SAR images of sea ice have been used successfully to identify ice floe displacement, which became wellestablished application and was used to generate ice motion maps (section 11.7.1). Increasing the dimensionality of SAR data using multi-frequency, multi-polarization, and texture approaches is discussed in the following. Multi-frequency SAR observations have the potential to identify ice types. Among the first experimental SAR multi-frequency systems was the NASA’s Jet Propulsion Laboratory Airborne SAR (AIRSAR) operating in the C-, L-, and P-band frequency. The potential of this approach for sea ice classification was presented in Drinkwater et al. [1991]. Another airborne multi-frequency SAR system was developed in the late 1980s by Canada

Centre for Remote Sensing and installed onboard the Convair 580 airplane [Livingstone, 1996]. It operated in the X- and C-bands. These two systems never intended for operational use though studies showed the potential of multi-frequency data for discrimination between FYI and MYI. A few studies have been conducted lately to compare data from space-borne SAR data from different frequencies and integrate them to enhance information retrieval. Dierking and Busche [2006] explored how the C- and L-band could complement each other and concluded that their combination resulted in a more detailed view of the state of sea ice cover. It is not likely, however, that a space-borne multi-frequency SAR system will be developed in the near future. Perhaps the most viable approach to generate multichannel SAR data is through using multi-polarization channels. While multi-frequency SAR data requires more than one antenna, multi-polarization requires one antenna. The underlying theme behind the use of multipolarization to enhance the sea ice parameter retrieval is the possibility of the depolarization of the scattered signal, which holds clues about the surface type and composition. Multi-polarization SAR systems have been developed recently into dual polarization and quadpolarization (also called fully polarimetric) systems. Information about fully polarimetric SAR is introduced in the next section. A dual polarization SAR mode was introduced for the first time in the Advanced SAR (ASAR) system onboard the European satellite Envisat (2002–2012). It recorded backscatter in one of three alternating polarization combinations (HH and HV), (VV and HV), or (HH and VV). Later, the Japanese ALOS-1 PALSAR (January 2006–April 2011) and the Canadian RADARSAT-2 (December 2007–recent) developed SAR with a dual polarization mode operating with either one of the two selections (HH and HV) or (VV and VH). Increasing the dimensionality of SAR data by using texture has been examined in several research papers but has not been proven viable for operational use. A number of textural measures can be generated from a single-channel observation; many of them are uncorrelated. This approach has been tested extensively since the early years of SAR sea ice applications [Shokr, 1991, Soh and Tsatsoulis, 1999, and Clausi and Yue, 2004, Chen et al., 2020). A key question regarding this approach is how to capture the texture of different sea ice surfaces at the appropriate scale for each surface. The texture of natural ice surface exists at different scales ranging from smallscale roughness all the way to large-scale blocks that form ridges and deformed ice surface. A texture technique usually encompasses convolution of the image with a kernel operator of a given size. The window captures texture of scales that “resonate” with its size. Therefore, a texture parameter calculated using a given window size may fail

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

to capture the texture which is effective at a scale different from that size. More on this subject is presented in section 11.1.3.1. Finally, phase difference between the received and transmitted radar signal can be used as an additional dimension in SAR observations. This issue has been addressed in a recent publication by Ding et al. [2020] but not generally pursued in the literature. The reasons are twofold. First, ice types do not have unique surface geometry that warrants generation of unique phase information associated with each type. Second, it is not clear how the phase information can be useful in resolving some of the snow issues that hamper sea ice classification using SAR data. Recall also that the phase information is accessible only through the SLC SAR acquisition mode. This product is not available from the ScanSAR wide swath mode, which is frequently used in operational sea ice applications.

7.6.3. SAR Polarimetry: Formulation and Derived Parameters Unlike the single- and dual-channel SAR, fully polarimetric (FP) SAR (also called quad-pol SAR) was developed to acquire full polarimetric set of parameters that describe the imaged surface/target better. The suite of polarimetric parameters is the subject of this section. However, it is worth starting the discussions by noting the original paper on SAR polarimetry by Kennaugh [1951]. Research on evaluation of SAR FP data for sea ice applications started in late 1980s and early 1990s after the NASA/JPL airborne multi-frequency SAR FP system (AIRSAR) became operational in 1987. Sample data from this mode are presented in section 9.3.3. Readers who wish to acquire more information on the theory and applications of the FP system may refer to a few books on the subject, e.g., Ulaby and Elachi [1990], Lee and Pottier [2009], and van Zyl and Kim [2011]. An FP system transmits alternating pulses in two orthogonal polarizations, e.g., vertical and horizontal. For each pulse, the system records the backscatter (intensity and phase) in the two orthogonal polarization components. This means that the system records four observations from each ground cell (all complex numbers to combine the intensity and the phase) representing HH, HV, VV, and VH backscattering signals. With this information, the user can derive many parameters that fully describe the scattering behavior of the imaged surface/target. That is why FP is considered to be the ultimate polarization-based technology of SAR. The derived parameters can be grouped into three sets, which are described in section 7.6.3.2. FP SAR mode has been available from ALOS-1 PALSAR and RARADSAT-2 systems.

315

While the FP mode allows generation of the largest amount of polarimetric information, the shortcoming of this mode is the narrow swath (10–40 kilometers in the current systems). This makes it unfavorable for operational sea ice monitoring. Moreover, the linear polarization suffers from range ambiguity, which limits the FP mode to a maximum of 40 incidence angle [Freeman and Raney, 2008]. For that reason, the linear FP mode from RADARSAT-2, for example, is limited to the standard and fine beam acquisition modes (Figure 8.2). A more recent mode that approximates the FP but produces images with hundreds of kilometers swath is called hybrid compact polarimetry (HCP) or CP for short. The concept of SAR CP was suggested by Green [1968] and introduced in Raney [2007]. In this technology a single linear (vertical or horizontal) or circular polarization (right or left) is used in the transmitted pulses and the system records the two linear backscatter returns. As the CP mode transmits only one polarization signal, it is considered to be a variation of the dual polarization and that is why it is known as “partial polarimetry.” However, unlike the dual polarization, it retains the phase information. The CP is an efficient step up from the single and dual polarization toward FP systems to realize many (but not all) of its benefits without the attendant coasts of doubling the average transmitted power [Raney, 2007]. Brisco, Mahdianpari, Mohammadimanesh [2020] stated that the CP overcomes the drawback of conventional dual systems, namely the lacking of fully polarimetric information and the shortcoming the FP systems with their small swath width. The CP mode of the RCM (launched in June 2019) transmits a right-circular polarization and receives two mutually coherent orthogonal polarizations (σ 0RH and σ 0RV ). The resolution is 50 m covering a swath width of 350 km. This constellation system, which is composed of three satellites following each other in the same orbit, covers the Arctic region daily. In summary, the CP mode is considered to be a dual polarization system, but the set of polarimetric parameters from this system is comparable to those derived from the FP mode [Bouzerar et al., 2020]. Dabbor and Geldsetzer [2014] calculated 23 parameters from simulated CP SAR data and evaluated their potential for sea ice type discrimination. The study identified the three parameters for the optimal classificaation of sea ice and open water. However, Touzi [2009] provided a critical comparison between fully polarimetric SAR (with its linear polarization) from RADARSAT-2 and the simulated CP SAR. The study concluded that the FP should remain the chioce if the high quality of crosspolarization return is the priority. That is because FP SAR provides cross-polarization measurements of much

316

SEA ICE

Table 7.3 Characteristics of SAR dual, fully polarization, and compact polarization modes. Dual polarization Transmission Reception Phase between channels Output polarization combinations

Fully polarimetric (FP)

Compact polarimetry (CP)

Single linear polarization H or V Non-coherently H and V Not known

2 linear polarizations (H, V) 2 circular polarizations (R, L) Coherently H and V for any transmitted polarization Known

Single linear polarization (H or V) or single circular Right (R) or Left (L) Coherently H and V or R and L for any transmitted polarization Known

(HH, HV) or (VV, VH)

HH, HV, VH, VV RH, RV, LH, LV

(L45L45, L45L−45) or (L−45L−45, L−45L45) or (RH, RV) or (LH, LV): same as dual polarization

Note: L45 and L−45 are linearly polarized signals at 45 and −45 orientation, respectively.

higher signal-to-noise ratio than the CP SAR. Nevertheless, CP mode provides meaningful phase information that might be worth exploiting. Table 7.3 summarizes the polarization characteristics from the three SAR modes: dual, FP, and CP.

7.6.3.1. Formulation of Polarimetric Measurements The four complex scattering measurements that completely describe the scattering properties of the target in polarimetric SAR are usually expressed in a vector or a matrix form. A few forms have been developed from which several parameters can be extracted and used in qualitative and quantitative analysis of the data. Among these forms are Mueller matrix and scattering matrix (also called Sinclair matrix). Elements of the latter are used to compose two vectors from which two second-order matrices called covariance and coherency are generated (see details below). Details on derivation of these description matrices can be found in a number of references including Zebker and van Zyl [1991], Cloude and Pottier [1997], Cloude [1997], Freeman and Durden [1998, Touzi et al. [2004], and Touzi [2007]. Here, a brief summary of the scattering matrix as well as the covariance and coherency matrices is introduced. The scattering matrix is the basic expression that describes the transformation of the electric field of the incident wave to the scattered wave: E sh E sv

= S

E ih E iv

=

S hh S vh

S hv S vv

E ih E iv

(7.74)

where, E is the electric field intensity, superscripts i and s denote the incident signal (transmitted) and the scattered (received) signal; respectively, and subscripts h and v denote the vertical and horizontal polarizations, respectively. The elements of the matrix in the RHS of equation (7.74) are complex quantities. Each element combines intensity and the phase of the measured signal. The effective backscattering coefficients σ 0 in different polarizations are related to the elements of the scattering matrix by the following equations:

σ 0hh = 10 × log 10

S hh

2

σ 0hv = 10 × log 10

S hv

2

σ 0vv = 10 × log 10

S vv

2

(7.75)

where .. denotes the averaging over a neighborhood of N pixel as the equation (7.75) represents the multi-looked intensity images for the three shown polarizations. The conversion of scattering elements in equation (7.74) to backscatter coefficient is implemented in commercial software packages such as ESA’s SNAP or ASF’s MapReady. Second-order matrices have elements representing second order statistics based on vectorization of the scattering matrix. With the reciprocity assumption of the crosspolarization elements (Shv = Svh), two commonly used scattering vector expressions have been derived using elements of the scattering matrix. The rationale behind their development is presented in Hellman [1999]. The first vector is known as the scattering vector: T

K C = S hh , 2S hv , S vv

(7.76)

A well-known matrix that can be derived from this vector is called covariance matrix [C]. It is constructed in the power domain by multiplying the vector by its conjugate transpose. This an example of second-order matrix: S hh C = K C K ∗C T =

2

√2S hv S ∗hh S vv S ∗hh

√2S hh S ∗hv 2 S hv

2

√2S vv S ∗hv

S hh S ∗vv √2S hv S ∗vv S vv

2

(7.77) where, superscripts ∗ and T denote the complex number conjugate and vector transpose, respectively. Decomposition of this matrix is used to generate parameters that represent the power from each scattering mechanisms. This is known as Yamaguchi decomposition (presented in the next section). The other vector expression, known as Pauli target vector, is also composed using combinations of elements from the scattering matrix: 1 S hh + S vv , S hh− S vv , 2 S hv T (7.78) KP = 2

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

Roughly speaking, the square of the first and second elements represent contributions from the SB and DB scattering mechanisms, respectively. In general, crosspolarization return (the third element) is associated with MB (random scattering) although under some special geometrical arrangement it can be caused by DB scattering mechanisms. A color composite image, known as Pauli image, can be formed from the square of the three elements of Pauli vector. Assignment of the elements to the three colors (RGB) is implemented such that the color distribution in the

1 T = 2

T = K P K ∗P T

S hh 2 − 2 Im S hh S ∗vv − S vv

2

2S hh S ∗hv + 2S vv S ∗hv

+ 2 Im S hh S ∗vv − S vv

2

S hh 2 − 2 Re S hh S ∗vv + S vv

2

2S hh S ∗hv − 2S vv S ∗hv

2

2S hv S ∗hh

+

2S hv S ∗vv

2S hv S ∗hh

= N < C > N

T

where N =

1 2

1 1 0

0 0 2

− 2S hv S ∗vv

1 −1 0

(7.81) Both covariance and coherency matrices contain the same information about the scattering measurements (amplitude, phase, and correlation) as well as scattering mechanisms. The trace of each matrix represents the total power of the scattered wave. The coherency matrix is closely related to the physical and geometrical properties of the imaged surface and therefore allows better physical interpretation. Both matrices should be averaged over a window of neighboring pixels before deriving the polarimetric decomposition parameters as described below. This process should further reduce the speckle in the data though it degrades the spatial resolution. The scattering mechanisms from a given surface (Figure 7.23) are expressed in terms of parameters derived from the decomposition of the covariance or the coherency matrix. It is worth noting that decomposition of a first-order matrix (e.g., scattering matrix) leads to parameters that can characterize a coherent target within the resolution cell, namely a sole target responsible for the measured backscatter. On the other hand, decomposition of a second-order matrix leads to parameters that characterize incoherent targets; namely, distributed scatterers

(7.79)

This is a 3 x 3 Hermitian positive semidefinite matrix which can be written as:

2

The three diagonal elements of the coherence matrix are the squares of the elements in the Pauli target vector [equation (7.84)]. Therefore, they represent the three scattering mechanisms as mentioned above. The covariance and coherency matrices are linearly related. The covariance matrix can be converted into coherence matrix using the following linear transform. Note that an average (i.e., multi-look) of [T] and [C]should be used: T

image is close to the surface color in nature. For example, if the blue color is assigned to the square of the first element (Shh + Svv), the sea surface in the image will appear blue because it triggers surface scattering (odd-bounce) mechanism. Another second-order matrix, known as coherency matrix [T], can be constructed from Pauli target vector:

S hh 2 + 2 Re S hh S ∗vv + S vv S hh

317

4 S hv

7 80

2

across the resolution cell. Sea ice surface usually belongs to this category. 7.6.3.2. Polarimetric Parameters Derived from the FP SAR Data Three sets of polarimetric parameters can be derived from four backscatter elements of the scattering matrix at each pixel. The parameters can be used to reveal information that cannot otherwise be revealed from using any other SAR mode. The first is called polarization signature plot. It is based on the idea that the backscatter from any linearly polarized transmit–receive observation can be calculated from the combination of the four basic received orthogonal polarizations σ ohh, σ ohv, σ ovv, and σ ovh. This allows generation of a 3D plot of the scattering power (in dB) as a function of ellipticity χ and orientation ψ of the transmitted signal (Figure 7.8). In fact, the scattering power can be determined as a function of the four wave polarization variables, a pair of ψ and χ of the incident signal, and another pair of ψ and χ of the scattered signal, however, this would be extremely tedious and inconvenient task. Therefore, to simplify the process, two plots are usually generated: one with the polarization of the backscatter is the same as that of the incident wave (co-polarization) and the other for the orthogonal polarization with respect to that of the incident wave (cross-polarization). This can be portrayed in two surface plots called the co-pol and cross-pol signatures. An example of polarization signature of Brag scattering from ocean surface is presented in Figure 7.24. Obviously, these two signatures do not represent every possible transmit–receive polarization combination, yet they provide a useful visualization configuration that can hopefully be unique to facilitate identification of a few ice types. Examples of polarization signatures of selected ocean and urban surfaces are shown in Evans et al. [1988].

318

SEA ICE

Power

Co-polarized signature

Power

Co-polarized signature

0 0

0 0

Or

45

ien

90 135

tat

ion

–15 0 –45 –30

180

15

30

45

g)

ity (de

(de

elliptic

g)

Or

45

ien

90 135

tat

ion

180

(de

–15 0 –45 –30

g)

15

30

45

g)

ity (de

elliptic

Figure 7.24 Polarization signature in co- and cross-polarization for Brag scattering from ocean water surface (courtesy Canada Center for Remote Sensing).

Experimentation with this analysis tool flourished in 1980s using data from NASA’s multi-frequency fully polarimetric AIRSAR. The research has declined later as the FP mode from SAR was not readily available for almost a decade until it became available from the Japanese PALSARALOS and the Canadian RADARSAT-2 systems (section 8.4). However, to the best knowledge of the authors of this book, the polarization signature has not been used to explore its potential for ice type classification. Part of the reason is the difficulty of quantifying the 3D geometric shape associated with different ice types. The example in Figure 7.24 is easy because it features a well-defined shape. Complex shapes with too many perturbations become difficult to geometrically characterize. Nevertheless, it is recommended to explore the possibility of using this tool for ice type identification and perhaps snow cover characteristics. Generation of polarization signature is yet to be included in SAR polarimetry analysis toolkits. The second set, known as polarimetric parameters, describes the polarimetric behavior of the ground cell/target in terms of the intensity and phase of each polarized component of the backscatter. It is derived directly from the four backscatter components included in the scattering matrix [equation (7.74)]. They include the total power (SPAN), the co-polarization and cross-polarization ratios (Rhh/vv, Rhh/hv, and Rvv/hv), the depolarization correlation coefficient Rdepol, the co-polarization phase difference ϕhh − vv, and the co-polarization correlation coefficient rhh − vv. Mathematical expressions of those parameters are as follows: SPAN = S hh S ∗hh + S vv S ∗vv + 2 S hv S ∗hv Rhh Rhh Rdepol =

(7.82)

vv

=

S hh S ∗hh S vv S ∗vv

(7.83)

hv

=

S hh S ∗hh S hv S ∗hv

(7.84)

S vh S ∗hv S vv S ∗vv S hh S ∗hh

(7.85)

ϕhh − vv = tan − 1

rhh − vv =

Im S hh S ∗vv Re S hh S ∗vv S hh S ∗vv

S hh S ∗hh S vv S ∗vv

(7.86)

(7.87)

where Re and Im are the real and imaginary components of the complex number, ■ denotes the average over a number of neighboring pixels, ∗ denotes complex conjugate, and |■| is modulus of complex number. The crosspolarization ratio and the depolarization correlation coefficients are indicators of the multiple (random) scattering mechanism, which depolarizes the scattered signal. ϕhh − vv of 180 indicates ideal SB scattering from a smooth surface and 0 indicates ideal DB scattering from a dihedral-like landscape structure. Data from this parameter set are presented in section 9.3.3. The third set, known as polarimetric decomposition parameters, comprises parameters that express, directly or indirectly, the scattering mechanisms from the observed medium either. The parameters are derived from the decomposition of the second-order forms of the scattering matrix. This set is divided into two subsets. The first is obtained from a purely mathematical-based decomposition such as the well-known Cloude-Pottier eigen decomposition of the coherency matrix. The second set is obtained from a model-based decomposition such as the Yamaguchi decomposition of the covariance matrix. Both types of decompositions and their derived parameters are introduced in the following paragraphs. Ground cover (e.g., sea ice types) can be uniquely identified by parameters from this set if associated with a certain dominant scattering mechanism. More on the radar scattering mechanisms from different types of sea ice is presented in section 7.6.3.3. It is worth noting that decomposition of the first-order scattering matrix leads to generation of parameters that can characterize a coherent target within the resolution cell, namely a sole target that largely contributes to the

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

measured backscatter. On the other hand, decomposition of a second-order matrix leads to parameters that characterize incoherent targets; namely, distributed scatterers across the resolution cell. Sea ice surface usually belongs to this category. Two commonly used decompositions (for coherency and covariance matrices) with the relevant polarimetric decomposition parameter sets from each one are presented in the rest of this section. A commonly used decomposition approach is the wellknown Eigen decomposition of the coherency matrix [T]. It decomposes this matrix into three unitary orthogonal eigenvectors related to the three scattering mechanisms is presented in Cloude and Pottier [1997]: T

= V Λ V

−1

(7.88)

where [V] is the matrix of the eigenvectors, and [Λ] is the 3 x 3 diagonal matrix of the eigenvalues λ1, λ2, and λ3. This equation can be re-written in the form: T = λ1 e1 eT1 + λ2 e2 eT2 + λ3 e3 eT3

(7.89)

where ei are the three orthogonal eigenvectors, with e1 being the largest and e3 the smallest. The relative values of λi are indicators of the dominant scattering mechanisms. For example, if the primary eigenvalue λ1 is much larger than the secondary ones (λ2 and λ3) it means that one scattering mechanism is dominant. An eigenvector ei can be written as: ei = eiϕi cos αi λj i , sin αi cos βi ejδi , sin αi sin βi ejγ i

T

(7.90) where αi is an angle related to the scattering processes, βi is the orientation angle of the target about the radar line-ofsight, and δi and γ i are phase angles related to the target material. A few parameters derived from this decomposition are commonly used in applications of polarimetric data [Cloude and Pottier, 1997]. They include entropy H, which takes values between 0 and 1; anisotropy A, which also takes values between 0 and 1; and the weighted average alpha angle α with values between 0 and 90 . The first two parameters are derived from the eigenvalues and

the third from the eigenvectors. Denoting the probability obtained from the eigenvalues λi as Pi: P i = λi

3

λ; j=1 j

A≅0 λ2 ≅ λ3 Two secondary mechanisms effective

(7.91)

The definitions of H, A, and α can be written as: H = − P1 log 3 P1 − P2 log 3 P2 − P3 log 3 P3 A = λ2 − λ3

λ2 + λ3

α = P1 α1 + P2 α2 + P3 α3

(7.92) (7.93) (7.94)

where, αi is the arccos of first element of the eigenvector ei as indicated in equation (7.90). Equation (7.94) represents a weighted average of the αi values in the three eigenvectors. The above three parameters are related to the number and the type of the scattering mechanisms. Yet, they are not direct measurements of the power contained in each scattering mechanism as shown later when introducing the Yamaguchi decomposition. When calculating the parameters, they should be averaged over a specified neighborhood window (e.g., 5 x 5 pixels) in order to furnish a multi-look (speckle-filtered) data. Without this step some parameters such as entropy [equation (7.92)] cannot be interpreted correctly because the randomness it represents will be attributed to the speckle noise rather than authentic randomness of the signal scattered off the surface. Entropy is an indicator of the multi-scattering mechanisms (namely single- or double-bounce) or the randomness of the scattered signal (namely, multiple scattering). Small values signify the presence of a deterministic (odd- or double-bounce) scattering mechanism. In practice, 0.3 is considered a small value. Large values signify multiple scattering. In practice, 0.6 is considered a large value. Values between these two limits mean limited contribution by the random scattering and/or comparable contributions of the single- and double-scattering. But it does not inform much about the relative weights of the two comparable mechanisms. In this case the anisotropy can be useful. Figure 7.25 illustrates the meanings of H and

No dominant scattering; “A’’ is considered for this range of “H’’

H≅0 λ1>>λ2 and λ3; λ2 and λ3 ≅ 0

319

A≅1 H≅1 λ2 >> λ3 λ1 ≅ λ2 ≅ λ3; One secondary mechanism effective

Figure 7.25 Illustration of the physical meanings of the entropy and anisotropy parameters in terms of the number of the effective scattering mechanisms.

320

SEA ICE

A (which are complementary) in relation to the three eigenvalues of the coherence matrix. According to equation (7.93), low values of A mean that the two secondary mechanisms associated with λ2 and λ3 (recall that λ2 > λ3) have nearly equal values; i.e., both mechanisms are equally effective. High values mean that only one secondary mechanism is effective (namely, the one associated with λ2). However, there is caveat here. It is possible that a surface features a dominant scattering mechanism while the power from each one of the other two mechanisms is very low and nearly equal. In this case, λ1 becomes very high while both λ2 and λ3 become very low, which simply means no significant secondary scattering mechanisms. According to equation (7.93), the anisotropy in this case can be very high or very low because it implies a small quantity divided by another small quantity. In other words, the anisotropy value means nothing but noise. This is observed in the case of frazil or grease ice with their few millimeters thickness (see definition in section 2.3.2), from which the scattering is predominantly single-bounce while negligible traces of the other two mechanisms exist. One should not bother with the interpretation of the anisotropy in this case. The case of MYI floes offers another interesting interpretation of the anisotropy. This parameter has remarkably lower values from MYI compared to other ice types. Hence, it can be used to identify this ice type (Figure 9.28 and Table 9.8). So, why the low anisotropy from MYI? Both surface and volume scattering mechanisms take place in MYI with surface being the dominant and the volume as a significant secondary mechanism. In this case, A is expected to assume high value because λ2 is expected to be much larger than λ3. However, this is not the case. Low values of A mean that λ2 > λ3, but not λ2 >> λ3. The alternative reason for the low A from MYI can be explained by the fact that the secondary scattering mechanism is MB. Since this is not a deterministic scattering mechanism, its randomness may bring λ2 to lower values close to λ3, which leads to low values of A. This is possible because MB is active only from hummock surface while most of the MYI floe area is depression (melt pond) surface. The last parameter derived from the eigenvector decomposition of the coherency matrix is alpha angle (α) [equation (7.94)]. It varies between 0 and 90 and is related to the type of scattering mechanism. When α = 0 , it means that scattering features SB mechanism. When α = 90 the scattering is generated through DB mechanism (e.g., scattered high-rise buildings in flat area). For α = 45 the scattering is generated from the volume, through MB mechanism. As mentioned before, Yamaguchi decomposition [Yamaguchi et al., 2005] is a model-based decomposition. It models the covariance matrix [C] into four scattering matrices corresponding to the SB, DB, MB, and helix

scattering mechanism. It is an extension of the threecomponent Freeman-Durden decomposition [Freeman and Durden, 1998], which is more suitable for agricultural applications. The four components are denoted by the subscripts s, d, v, and c (respectively) as shown in the following equation: C = fs C s + fd C

d

+ fv C v + fc C

c

(7.95)

where, f is the coefficient pertaining to each scattering mechanism. From fs , fd , fv , and fc , the power of the SB, DB, MB, and helix scattering mechanisms, respectively, can be calculated as shown in Yamaguchi et al. [2005]. In this book, the power of SB, DB, and MB are denoted as Ps , Pd , and Pv , respectively. Sample data are presented in section 9.3.3) from several ice types. The ratio Pv /Ps was found to be useful in identifying ice types associated with strong Pv power, namely MYI and new ice. An example that addresses rafted ice is presented in section 10.1 and another example that address possible discrimination between hummock and melt pond ice using scattering power is presented in section 11.1.3.3 (Figure 11.12).

7.6.3.3. Linking Radar Scattering Mechanisms to Ice Features Figure 7.26 is an illustration of the three radar scattering mechanisms in relation to a few key features of sea ice. SB is triggered by smooth or rough surface, defined as surface with its root mean square (RMS) of height being of the same order of magnitude as that of the incident radar wavelength. The DB mechanism is activated when the signal is bounced twice between two orthogonal surfaces; hence it is likely triggered by a ridge on the ice surface. The MB occurs when the incident wave undergoes many random bounces between numerous scattering elements within the ice volume (e.g., the bubbly sublayer of MYI hummock) or at the surface (heavily deformed ice blocks). Since MB mechanism depolarizes the scattered signal while SB does not, the power ratio of MB/SB mechanisms (Pv /Ps) can be used as a proxy indicator of the degree of the depolarization of the backscatter. This parameter can be derived from FP or CP data according to equation (7.3). As mentioned above, it can be used to identify two cases of ice types where MB scattering power is comparable or exceeds the SB scattering power. These are the cases of MYI and new ice with thickness less than the penetration of incident radar signal. They both depolarize the scattering signal because of the dominant MB mechanism. This is shown in cases (c) and (d) in Figure 7.26 for the cases of MYI and new ice, respectively. Another example that demonstrates the use of this ratio (Pv /Ps) to identify these two ice types is presented in Figure 9.28 (section 9.3.3). Seawater surface that features no wave or capillary waves (with wavelength < 17 mm) is considered smooth

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE (a)

(c)

321

(b)

(d)

sea ice

h

sea water

Figure 7.26 Radar scattering mechanisms from a few sea ice features: (a) single-bounce scattering from leveled ice, (b) DB scattering from ridged ice, (c) MB scattering from air bubbles within MYI hummock and (d) MB scattering from the dendritic ice–water interface of very thin ice layer. h is the thin ice thickness [Shokr et al., 2022 / Taylor & Francis].

with respect to the much larger wavelength of the incident C-band (5.3 cm). Hence, the scattering mechanism will be strictly SB with mostly specular scattering. When the surface features gravity waves with wavelength of a few centimeters as the wind speed exceeds 5 km/h, Bragg scattering will be activated and the backscatter increases [LeBlond and Mysak, 1981] (section 7.9.2.2). Nevertheless, the scattering mechanism continues to feature dominant SB. Only large ocean swell that exceeds 1 m may trigger DB mechanisms. Yet, this may be observed in open ocean away from the ice, hence not relevant to sea ice observations. A few roles on using the radar scattering mechanisms, estimated from polarimetric decomposition parameters described in the previous section, have been suggested in Shokr et al. [2022] and summarized in the following. 1. The scattering from the OW surface is strictly SB, whether the surface is smooth or wind-roughened (assuming C-band radar signal). 2. Thin new ice, up to a few centimeters thick, triggers dominant MB scattering if the incident signal penetrates the ice thickness and interacts with the dendritic ice–water interface (shown in Figure 2.19). However, the MB scattering vanishes gradually as ice thickens, allowing for more SB scattering. 3. Level and relatively rough ice surfaces trigger mainly SB scattering with low and high backscatter, respectively. 4. Rafted ice surface may return very high backscatter with relatively high component of DB power (though it is always smaller than the power from the SB and MB mechanisms). This is explained in section 10.1 and Figure 10.2. 5. Deformed ice surface triggers high backscatter with SB, MB, and perhaps DB in the presence of large upturned blocks.

6. The dihedral-shaped ridges may trigger DB scattering. 7. MYI triggers MB scattering power comparable to the SB. However, different scattering mechanisms are revealed upon examining the scattering from hummock and melt pond surfaces [Shokr et al., 2022] (see also Figure 11.12 in this book). 8. The ratio of the scattering powers MB/SB (Pv /Ps) can be used to identify thin new ice and MYI. These two types are associated with significant component of MB. (See Table 9.7. 9. While the total power (SPAN) has nearly equal, high value from rough and deformed ice, the MB/SB ratio is much higher from deformed ice. This parameter can be used to discriminate between these two types. 7.6.3.4. Age-Based versus SAR-Based and ScatteringBased Sea Ice Classification Ice classification for operational or scientific research applications is based on the categories set by the World Meteorology Organization (WMO). These are age-based classes (see section 2.8). When imaging radar was introduced from airborne sensors in the 1980s and later from space-borne sensors in 1990s, data were used to identify the same age-based ice types, although the backscatter intensity (σ o) is not directly sensitive to the age or the thickness of the ice. That is why computer-assisted ice classification schemes have proven to be of limited operational use, especially from single and dual SAR systems. This is simply because the backscatter does not uniquely identify any age-based category except for MYI and only if the right wavelength is used (e.g., the C-band). This fact is demonstrated in several studies that address methods of sea ice classification from SAR (see a few references quoted in section 11.1.3). The pioneering study on

322

SEA ICE

developing knowledge-based rules for ice classification [Soh et al., 2004] used information beyond SAR backscatter to avoid the failure of SAR to detect the continuous development of ice age/thickness. Nevertheless, the operational success of the study was limited. When the FP data became available, studies continued to explore the utility of using the polarimetric parameters (section 7.6.3.2) in classifying ice types, but once again, based on the WMO categories [Isleifson et al., 2010, Gill and Yackel, 2012, Moen et al., 2015]. Aside from the issue of the snow cover that may complicate the sea ice classification when snow metamorphoses under varying meteorological conditions, a new concept for using SAR in sea ice classification is worth developing. This entails pursuing one of the following two approaches: using new categorization of ice classification based on what SAR actually “see” (may be referred to as SAR-based ice categorization) or using the power from each one of the three scattering mechanisms to identify ice types that feature certain dominant mechanism (may be called scattering-based ice categorization). Radar backscatter is triggered by factors that include surface roughness/deformation, ice dielectric constant, the composition of the ice cover within the penetration depth of the incident radar signal, and last but not least, the snow cover when metamorphosed to include wetness or ice layers/crystals. These factors are not particularly relevant to ice age or thickness. That is why in operational SAR analysis ancillary data about weather, climatology, recent history of the ice cover, and observations from other sensors have become (and will continue to be) indispensable to support mapping ice cover in terms of thickness and age. Switching from age-based to SAR-based or scatteringbased ice classification warrants better use of SAR data. This can be achieved using new categories based on backscatter intensity or scattering mechanism power. These categories are identified with ice features that trigger the observed backscatter intensity or scattering mechanisms. Potential new categories for SAR-based ice types (based on SAR intensity) include: (1) level versus rough or deformed ice surface, (2) annual versus perennial ice, and (3) bare versus snow-covered ice with metamorphosed snow. Categories for scattering-based ice types are already presented in section 7.6.3.3. There is potential to add subcategories on ice thickness within the thin (new) ice category. Using SAR to its full potential for sea ice classification and information retrieval requires avoiding forcing SAR backscatter to identify unwarranted age-based categories such as thin versus thick FYI, gray versus gray-white YI, and SYI versus MYI. Attempts to automate age-based sea ice classification have been undertaken by numerous research studies during the past four decades. Approaches

range from establishing empirical relations between backscatter and ice thickness, knowledge-based rules, image segmentation, pattern recognition methods, and recently adopting deep learning techniques. A new SAR-driven ice categorization, as outlined above, does not mean abandoning the knowledge and methodologies that have accumulated over decades to support the age-based categorization. Admittedly, this knowledge has been used successfully to support operational sea ice classification, though with employing ancillary data and rule-based framework. What a new SAR-based (or scatter-based) categorization would add is identification of additional categories. Examples include identification of different scales of surface roughness, thickness-based subcategories within the YI, and possible discrimination between SYI and MYI. More accurate identification of OW, new ice, ridged ice, and MYI are other examples of using scatter-based categorization. More information on this point can be found in Shokr et al. [2022] and in Figures (9.29) and (11.12) in this book.

7.7. SCATTEROMETER SYSTEMS Scatterometer is a side-looking radar sensor designed mainly to estimate the speed and direction of wind across the ocean surface. However, scatterometer systems have proven very valuable in monitoring sea ice in the polar regions and relating the interannual variability to climate change. Technical information on scatterometer systems is included in section 8.5. The system usually transmits C-band or Ku-band pulses. At these frequencies, microwaves penetrate only a few millimeters into seawater, so the backscatter is significantly affected by surface roughness. This is how it becomes effective in measuring the surface wind speed as it modulates the water surface roughness. Scatterometer systems have coarse spatial resolution measured in tens of kilometers but with large swath measured in a few thousands of kilometers. The sensor measures backscatter coefficient with a precision within a tenth of a decibel. Sea ice applications using scatterometer data have grown. Today, these data are essential tools used in generating sea ice products such as ice types, extent and age covering the polar regions. There are two types of scatterometer viewing configurations, fixed fan-beam and rotating fan-beam (also called pencil beam) (Figure 7.27). The fixed fan-beam usually consists of three pairs of antennae. The pairs transmit radar pulses at 45 forward, backward, and orthogonal to the satellite’s ground track on both sides of the flight path. Each scanning swath has a certain width with a gap between the two swaths. No data are collected within this gap, which is a disadvantage of the system as the full coverage of a certain region (e.g., a polar region) takes

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE (a)

323

(b) Pencil (rotating)-beam configuration

tra ck Sa tell

ite

Forward left

13

Middle left

Backward left

Left swath

336 km

Right swatch

7 km

Middle right

Backward right

/s

m

5k

5°

° ° 45 34 90° 25° 53°

ack

it tr

Incidence angle Forward right 64°

4 12 18

rpm

Inner beam

m

0k

0 11

Orb

802 km

Fixed fan -beam configuration

ack

it tr

Orb

700 km

Outer beam

900 km Cro ss t ra

ck

550 km

Figure 7.27 Viewing configuration of (a) the fixed fan-beam scatterometer onboard ESA’s scatterometer systems carried on ERS and ASCAT satellites, and (b) rotating pencil-beam scatterometer (values shown belong to SeaWinds scatterometer on NASA’s QuikSCAT satellite).

more orbits to accomplish. Moreover, since the incidence angle varies across the swath, the measured backscatter becomes incidence-angle-dependent. This can be viewed as another disadvantage if the backscatter varies sharply with the incidence angle as in the case of open water. Another feature of the fixed fan-beam system is the recording of co-polarization backscatter only. The systems onboard ESA’s satellite platforms ERS and ASCAT record the backscatter in VV polarization. The rotating-beam scatterometer, on the other hand, features a transmitted pulse that rotates around the satellite nadir axis at a fixed angular speed. The system transmits two beams as shown in Figure 7.27b, inner and outer. The three operational scatterometer instruments onboard NASA’s QuikSCAT, the Indian OSCAT and the Chinese HY-2 satellites (see section 8.5) have pencil-beam configuration with slightly different viewing parameters. For example, the SeaWinds scatterometer on QuikSCAT has the inner beam measuring HH polarization at incidence angle 46 and the outer beam measuring VV polarization at 54 . The scanning rotation is 18 rpm. On the other hand, OSCAT has the inner beam measuring HH polarization at incidence angle of 48.9 , while the outer beam measures VV polarization at incidence angle 57.6 , both with scanning rotation of 20.5 rpm. Two advantages of the rotating-beam configuration should be pointed out. The first is the elimination of the nadir gap (unlike the gap that exists in the data from the fixed fan-beam system). This allows more coverage per orbit and makes signal processing easier. The second is the fixed incidence angle, which requires no correction to account for angular variation of the measured backscatter. Moreover, though not relevant to the sea ice applications, with

the two-beam system, four looks become possible at any ground cell, thereby allowing directional ambiguities of the ocean winds to be resolved and the wind direction to be determined. The co-polarization ratio available from the rotating-beam system facilitates discrimination between water and ice and to some extent between ice types.

7.8. ALTIMETER SYSTEMS Space-borne altimeter sensors have been developed primarily to measure wave height over open ocean and thickness of ice sheets. However, their applications have been extended to sea ice and snow cover. Altimeter data have proven to be cutting edge technology for estimation of polar sea ice thickness at synoptic scales. Unlike sea ice thickness retrieved from microwave and TIR sensors, which is limited to a maximum of 20–50 cm (see section 11.4), altimeter system can estimate the full range of thickness, though with less accuracy, for thin ice of a few centimeters thickness. There are two types of altimeters, defined according to the transmitted signal, radar and laser. They both operate in a very similar way. A radar altimeter transmits radar pulses, which can map the surface at a resolution of a few kilometers. A laser altimeter, which is also called laser image detection and ranging (LIDAR), transmits laser pulses, which can map the surface with much finer resolution, typically 70 m. Therefore, the laser altimeter can detect steep changes in the surface elevation with higher accuracy. In both types, the system measures the response time of the return signal as well as the shape of the returned pulse. The response time determines how far

324

SEA ICE

h

Radar b e

am

Laser beam

the satellite is from the reflective surface (this can be snow or ice surface as shown later). Ice thickness is estimated from this measurement using buoyancy law as explained later. This allows monitoring the energy and mass balance of polar ice sheets, particularly in the Antarctic. On the other hand, the shape of the returned pulse carries information about the surface roughness and stratification of ice grain size of snow pack. This information can be retrieved directly based on the waveforms or by interpretation of derived quantities such as the backscatter coefficient. The latter method does not require further signal processing. The change in the received echo strength and shape from the ice and water surfaces allows mapping of the ocean-ice boundary and providing other potentially useful information on the sea ice thickness and snow depth. Radar altimeter transmits pulses that penetrate the snow and reflect off the ice surface. On the other hand, pulses from a laser altimeter reflect off the first surface they hit, be of open water, bare sea ice, or accumulated snow (Figure 7.28). The distance from the reflective surface to the satellite platform (determined from the time travel by reflected signal) is used to calculate sea ice thickness. However, this requires estimation of the freeboard as a reference. It can be estimated when the satellite flies over a gap between the ice floes or over leads within the ice cover. The freeboard is roughly equal to one-eighth of the ice thickness. Kwok et al. [2007] determined the mean

Snow fradar

Reference level

flaser

H Sea ice

Ocean

Figure 7.28 Schematic diagram showing the principles of observations of the laser and radar altimeter beams with definitions of the penetration paths flaser and fradar, the snow depth (h) and ice thickness (H).

freeboard of FYI and MYI in the Arctic to be 14 and 35 cm in the fall, and 27 and 43 cm in the winter. The increases of 13 cm of FYI and 8 cm of MYI freeboard during winter are caused by the 4 months of ice growth and snow accumulation. However, these approximate estimates can only produce rough estimates of ice thickness. Better and timely freeboard data have been used. The accuracy of the calculated ice thickness is tied to the uncertainties in the estimation of the freeboard. The latter is estimated with increasing uncertainty as ice becomes thinner (less than a few tens of centimeters). By measuring the distance from the reflective surface to the satellite platform, and knowing the freeboard of the snow-covered sea ice, the path of the laser and radar beams, flaser and fradar can be calculated (Figure 7.28). The ratio of this measurement to the total thickness of the ice sheet is proportional to the ratio of the ice density (in the case of radar altimeter) or the density of the ice/ snow composition (in the case of laser altimeter) to the water density. Therefore, this ratio can be used to estimate the ice thickness through application of the buoyancy law. In this process, the snow and sea ice cover are assumed to be in a state of hydrostatic balance. Both snow depth and the densities of seawater, sea ice, and snow must be specified. The equations used to estimate ice thickness H from flaser or fradar are presented here following Kwok and Markus [2017]. Other forms using different input parameters are presented in Kim et al. [2020]. More details on the principles of radar altimetry are presented in Laxon et al. [2013] and Ricker [2015]. H=

ρw ρw − ρs f − h ρw − ρi laser ρw − ρi

(7.96)

H=

ρw ρs f − h ρw − ρi radar ρw − ρi

(7.97)

where, the subscripts ρw , ρi , and ρs are the density of water, snow ice, and snow, respectively, and h is the snow depth. The accuracy of the measured freeboard used in equation (7.96) and (7.97) is crucial because freeboard represents only a small fraction of the ice thickness (10%– 15%, depending on the snow cover). Therefore, even a small error in its measurement will be magnified when estimating ice thickness. Several methods were developed to estimate the freeboard. Products of sea ice freeboard are available from ESA’s Climate Change Initiative (CCI), the Alfred Wegener Institute (AWI), and the Operation IceBridge (OIB) products. In a recent study, Zhang et al. [2021] estimated the Arctic sea ice freeboard and its variations for the period from 2002 to 2012 by reprocessing Envisat satellite altimetry data using a new model to remove geoid undulations.

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

Although estimates of ice thickness from the laser altimeter are produced at a resolution acceptable for ship navigation, the satellite overpasses provide measurements only along the ground track of the satellite. This means that filling gaps to produce a regional map of ice thickness requires data from numerous passes that can only be acquired over a long period. This is operationally unacceptable. The real value of ice thickness data from laser altimetry resides in providing climate-related information on ice thickness and volume that reveals trends of interannual variability of polar sea ice at a synoptic scale. 7.9. RADIATIVE PROCESSES IN RELEVANT MEDIA Remote sensing observations of sea ice are influenced by radiative processes from three relevant media: atmosphere, seawater, and snow over ice. In order to retrieve ice parameters from satellite observations, the influences of these media should be accounted for. This section addresses a few key issues of the EM wave interactions with these media. The information is rather basic and descriptive in order to serve the purpose of the readers who seek brief knowledge about the impacts of the three media on the remote sensing observations of sea ice. Nonetheless, the text includes references to sources with more in-depth information on the subjects. 7.9.1. Atmospheric Influences In addition to the radiance that originates from the surface, the radiation received at the TOA has contributions from the atmosphere. The gaseous and particulate matter (aerosols) components may affect the spectral information of the received radiation. The effects depend on the wavelength of the radiation, its path length through the atmosphere, the viewing geometry of the sensor and most importantly the atmospheric constituents. In the case of Visible 0.4–0.7

optical sensors (where the source of illumination is the sun), the atmosphere intervenes with the path of both incident and reflected radiations. On the other hand, only the path of the emitted radiation affects observations for TIR and PM sensors. Lack of understanding of the EM wave interactions with the atmosphere will impact negatively on the interpretation of the data. It is possible to use human intuition to guide interpretation of remote sensing imagery data in the VIS spectral region, but this is not as readily available and reliable when it comes to the TIR or microwave region. In short, for accurate retrieval of surface parameters, the influence of the atmosphere should be removed from the measured signal. 7.9.1.1. Influences of Atmosphere on Optical and Infrared Observations The atmosphere influences the observations through one or more of the following three processes: absorption (attenuation), scattering, and emission of the radiation. The processes are triggered by atmospheric gases, aerosols, and clouds. These processes are described in detail with their governing equations in several books [e.g., Elachi, 1988, Stephens,1994, Marzano and Visconti, 2003]. The task of accounting for these processes in the remote sensing observations is called atmospheric correction. This is perhaps the most important pre-processing task of optical and infrared data especially in the cases where multi-temporal images are to be compared and analyzed. The correction is not crucial in the case of microwave sensing since the atmosphere is mostly transparent with respect to microwave signal. The exception is the effect of water vapor and cloud liquid water contents, on emitted radiation at high microwave frequencies (> 37 GHz). Absorption is the process by which radiation is absorbed and converted into other forms of energy. In general, the atmosphere absorbs about 25% of solar radiation. Atmospheric gases absorb radiation only at certain wavelengths as shown roughly in Figure 7.29. Oxygen (O2) absorbs

Mid-infrared 0.7–1.3

Thermal infrared 3–14

Transmission

Photographic infrared Mid-infrared 0.7–1.3 1.3–1.3

100%

O2

O2

325

8–14 3–5

H2O

10.5–12.5

CO2

O3

H2O

CO2

CO2

Absorption 0 0.1

0.2

0.3

0.4

0.6

0.8

1

1.5

2

3

4

5

Wavelength (μm)

Figure 7.29 Atmospheric transmission windows by different gases (shown in white).

6

8

10 12

15

20

30

SEA ICE

In Mie scattering, the size of the scattering elements is comparable to the wavelength of the radiation. For optical radiation, Mie scattering is usually caused by water vapor, dust, and other aerosol particles ranging from a few tenths of a micrometer to several micrometers in diameter. Compared to Rayleigh scattering, Mie scattering is greater in intensity and tends to affect longer wavelengths. While Rayleigh scattering is nearly isotropic, Mie scattering has a dominant forward direction. This property makes Mie scattering responsible for the beautiful sunrise and sunset colors of the sky, when the atmosphere holds a great amount of smoke and dust. These elements scatter more of the shorter wavelength violet and blue color in different directions (Rayleigh scattering) and allow only the longer wavelength orange and red colors to reach our eyes through the forward Mie scattering. Lastly, the non-selective scattering occurs when there are particles in the atmosphere several times the diameter of the radiation being transmitted. In this type of scattering all wavelengths of light are scattered, thus the name non-selective. The water drops in clouds satisfy this condition and that is why clouds appear white. In general, atmospheric scattering is minimal for the relatively long wavelength of TIR radiation. The radiative processes that contribute to the observed radiation in the optical region are summarized in Figure 7.31. The figure shows the radiation received by an optical sensor (LS ) along with its three components: the terrestrial reflection/radiation (LT ), the cloud reflection (LC ), and the path radiation (LP ). For surface applications of remote sensing data, terrestrial radiation is the desired contribution to be received. According to figure (7.31), The solar radiation received at the IFOV encompasses direct solar radiation (1), indirect radiation

8 lar

Lp 5

LT

6 4

3

θ0

θr

diation

1

0λ

Smoke, dust, etc. (size ~ λ)

eE

Diameter

Mie scattering

nc

Gas molecules (size > λ)

Figure 7.30 Three types of scattering in the atmosphere, caused by different sizes of the scattering elements.

Neighboring pixel

IFOV pixel

Figure 7.31 Interaction of solar radiation with atmosphere and surface. The components of radiation that reaches the satellite point of observation are shown. Number and symbols are indicators of radiative processes and components of the received signal as explained in the text.

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

through scattering by the atmosphere (3), and the “leakage” contribution from the neighboring areas (7), which is called adjacent effect. Some solar and scattered radiation that reaches neighboring pixels (4) is scattered back to the atmosphere (6). This, along with the scattered and emitted radiation by atmospheric elements (5) constitutes the path radiance (LP ). Clouds receive sun irradiance (8) and reflect approximately 19% of it, forming the LC component. In general, in the optical and infrared regions, reflection and emission occur from cloud top only and, therefore, become weakly related to cloud contents. Temperature of cloud top can be inferred from TIR observations. Based on the above qualitative description, the radiance received by an optical sensor can be formulated as the summation of the three components: LS = Hαt + LP + LC

(7.98)

where, H is the total downwelling radiance, α is the surface reflectance and t is the atmospheric transmittance (the fraction of radiation that survives the trip over the distance x). The first term in the above equation represents LT. However, this representation does not include the leakage contribution from neighboring pixels, but this effect is usually small and can be neglected, yet it is always better to be accounted for. In order for the optical observations to represent the intrinsic directional reflectance characteristics of the surface, the observations must be corrected to remove the influences of the path radiation (atmospheric influences) and the cloud reflection. Several algorithms have been developed to perform the atmospheric correction. In an original study, Lindsay and Rothrock [1994] estimated the seasonal cycle of clear sky hemispherically integrated surface albedo of sea ice in the Arctic from measurements made with the AVHRR onboard the NOAA-10 and NOAA-11 satellites. Details of the method include calculations of the TOA reflectance of a Lambertian surface, an account for the non-isotropic reflectance of the ice and atmosphere, and a correction for atmospheric interference. De Abreu et al. [1994] presented a correction scheme for the retrieval of albedo over Arctic ice using data collected during the Sea Ice Monitoring and Modeling Site (SIMMS) experiment in Resolute Passage, Canadian Central Arctic in the spring of 1992 (SIMMS is presented in section 6.5). The scheme accounts for the intervening atmosphere, viewing geometry, and sensor spectral response in the VIS range. The authors found that the atmospheric correction increases the TOA albedo by 27% to 32% (meaning that atmospheric constituents absorb a significant amount of radiation in the VIS range). After removing the effects of viewing geometry, the variability of surface albedo between orbits decreases. When corrected for viewing geometry, the satellite-

327

derived surface albedo over snow-covered sea ice ranged from 0.68 to 0.82. Other atmospheric correction schemes that target a wide range of applications are presented in Fraser et al. [1989], Rahman and Dedieu [1994], and Karpouzli and Malthus [2003]. A comparison between different schemes is presented in Norjamäki and Tokola [2007]. Atmospheric correction that accounts for the scattering, attenuation, emission, and adjacent effects is addressed in several books [e.g., Schowengerdt, 2006, Wang, 2010, and Kondratyev, Kozoderov, Smokty, 2012]. Commercial software packages for atmospheric corrections are also available. Two commonly used packages are the Quick Atmospheric Correction (QUAC) [Bernstein et al., 2006] and the Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) [Anderson et. al., 2002], both developed by Spectral Sciences in United States. The software packages retrieve spectral reflectance from multispectral and hyperspectral radiance images. QUAC conducts a more approximate atmospheric correction (accuracy within ±15%), but at a faster speed than FLAASH or other physics-based first-principle methods. FLAASH, on the other hand, is a first-principle atmospheric correction tool that corrects radiation in the VIS through NIR and shortwave infrared regions, up to 3μm. Details on the principles and operation of these two packages of software are included in ITT Visual Information Solutions [2009]. Cloud filtering or masking is another issue that should be addressed at the pre-processing level of the data. Due to the higher reflectance of clouds in the visible range, cloud cover has been found to obstruct the sea ice scenes during most of the daylight season in the polar regions [Chen et al., 2002]. On an average, 80% of the Arctic Ocean is covered with clouds. For clouds to be removed and sea ice information to be recovered, the area should be observed for a long enough period. Daily observations would probably be sufficient. Johannessen et al. [2007] recommended receiving images from four to five orbits per day in order to recover surface information at least once in any 3–5 days period. It is worth mentioning that undetected cloud pixels also impede the retrieval of atmospheric information (e.g., aerosol parameters) and render correction for atmospheric influences difficult. Numerous techniques have been developed to filter the cloudy segments from the optical and TIR imagery data, minimize the cloud effects on surface observations, or just identify the cloud situation at the pixel level as being fully cloudy, partially cloudy, or clear. Traditional techniques of cloud masking for low and medium resolution data (e.g., AVHRR and MODIS) are usually based on empirically tuned thresholds from VIS and IR channels [Ackerman et al., 1998, Dybbroe, Karlsson, Thoss, 2005]. More sophisticated approaches that were developed later included the use of neural networks [Jang et al., 2006],

328

SEA ICE

linear unmixing techniques, time-series analysis [Gómez-Chova et al., 2007], and parallel Markovian segmentation [Le Hégarat-Mascle and André, 2009]. Cloud masks are part of geophysical parameter products from operational satellites. The MODIS Cloud Mask aims to minimize the effect of cloud contamination [Ackerman et al., 1998]. It also classifies each pixel as either confidently cloud, probably cloud, confidently clear or uncertain. Yu and Lindsay [2003] stated that while various cloud masking techniques are available, no reliable algorithm has been developed so far for the Arctic dark season. Visual inspection of the TIR images is still a usable method for this case. Schweiger et al. [2008] explored connections between cloud cover and sea ice variability in the marginal ice zone in the Arctic during autumn using 40-year weather re-analysis from the European Center for Medium-Range Weather Forecast (ECMWF) and the Television and Infrared Observation Satellite (TIROS), Operational Vertical Sounder (TOVS), and Polar Pathfinder satellite datasets. They found that cloud cover variability is strongly linked to sea ice variability. Changes in cloud cover can be explained in terms of the increase of near-surface temperatures resulting from the removal of sea ice. 7.9.1.2. Atmospheric Correction for Passive Microwave Observations Due to their long wavelengths relative to the dimensions of the atmospheric scattering elements, observations from microwave sensors are not generally affected by gases or aerosols in the atmosphere, especially at longer wavelengths. Usually, observations from channels lower than 19 GHz (higher than 1.55 cm wavelength), which includes frequencies of operational radar remote sensing, can “see” through the atmosphere and clouds provided that clouds are not precipitating. For PM frequencies of 37 GHz channel and higher (wavelength < 1.0 cm), observations can be affected by atmospheric constituents to some degree, depending on the type and size of suspended elements. Therefore, a correction must be applied to recover the microwave emitted radiation from the surface. The correction should be applied particularly to the SSM/I 85.5 GHz. AMSR-W 89.0 GHz, SSMIS 91.1 GHz or equivalent high-frequency channels from future PM sensors. The atmospheric influences are triggered by three parameters: integrated water vapor along a vertical column of the atmosphere (IWV), cloud liquid water contents (CLW), and the surface wind speed over open ocean (V). Cloud thickness is also a factor but it cannot be corrected for. Therefore, pixels that include clouds thicker than a certain threshold should be excluded from any surface parameter retrieval. The water vapor is concentrated within the first 5 km of the troposphere. The cloud liquid water is defined as the integration of all forms of water including ice crystals in a cloud volume.

Data of IWV and CLW and V, nearly coincident with the satellite overpass, are needed to conduct the correction. There are two approaches to obtain these data. The first is by using the data from operational weather models. The second is to calculate these parameters using PM observations from the water vapor channel (around 22 GHz) in combination with channels that are not strongly sensitive to these parameters. The second approach has traditionally been used to obtain the parameters over open ocean surface only. However recent studies estimated the integrated water vapor from the strong absorption lines spanning the highly opaque 183 GHz line. The lines are employed in spectral channels183 ± 1 GHz, 183 ± 3 GHz, and 183 ± 7 GHz of the Special Sensor Microwave/Temperature 2 (SSM/ T2) on board the Defense Meteorological Satellite Program (DMSP) series F-11 to F-15 as well as the Advanced Microwave Sounding Unit-B (AMSU-B) on board NOAA-15, 16, and 17. These channels measure radiation originating from a number of different layers in the troposphere. One of the methods that have been proposed to retrieve IWV from SSM/T-2 over sea ice in the Arctic is developed by Miao [1998] and in the upper troposphere by Sohn et al. [2003]. Qiao and Miao [2003] present maps of monthly averages IWV in the Arctic using AMSU-B channels. One of the operational weather models that produce IWV and CLW among a large suite of other atmospheric parameters is the Global Environmental Multi-scale (GEM). This is the model used at the Canadian Meteorological Centre (CMC) for short-range regional forecasting and medium-range global forecasting [Bélair et al., 2009]. The model outputs the results at time steps of 7.5 minutes but updates and archives the analysis after assimilating the model’s results with remote sensing observations and other ground measurements at synoptic times 00, 06, 12, and 18 GMT. GEM produces the meteorological parameters at a standard operational geographic grid of 33 km. However, coincident and co-located data with satellite overpasses can be generated when the model runs in a hindsight mode (to coincide with the satellite overpass) and by interpolating the model’s gridded results to be co-located with the location and geometry of the observed footprints. This approach was used in Shokr and Markus [2006] and Shokr and Agnew [2013]. There are two difficulties regarding the use of data from weather models in a correction scheme that accounts for the atmospheric influences on the PM observations. The first is the coarse spatial resolution of the model’s grid with respect to the spatial resolution of the observations to be corrected (the current grid spacing from GEM is 10 km or coarser while the footprint of the AMSR-E 89 GHz is 4 x 6 km2). The second is the unreliable estimate (or even the unavailability) of the CLW parameter.

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

A review on using PM observations to determine the water vapor profile and the integrated water vapor is presented in Urban [2013]. The IWV is usually estimated over open ocean from the weak absorption line of water vapor near 23.0 GHz (e.g., the 23.8 GHz channel of AMSR-E). The retrieval equation usually combines observations from this channel with other low frequency channels. For example, Boukabara [1997] proposed the following equation: IWV = 23 66 − 1 44 log 280 − T b19V − 2 47 log 280 − T b19H − 2 70 log 280 − T b229V − 0 811 log 280 − T b37V + 2 43 log 280 − T b36H (7.99) where, IWV is in kg/m2, Tb is the brightness temperature in K and the subscripts denote the channel frequency and polarization. Other equations are suggested by Alishouse et al. [1990] and Petty [1993]. A few methods to retrieve CLW over open water are also available. Water in clouds emits microwave radiation, and therefore CLW can be retrieved upon comparing the observation against a radiometrically cold background. This is the case of clouds over ocean, where Tb above clouds is significantly high compared to the values from the radiometrically cold ocean. It contrasts the case of clouds over land, where cloud bottom reflects the surface emission back, resulting in Tb above clouds being lower than Tb from the surface. O’Dell, Wentz, Nennartz [2008] presents a review on some of the methods of CLW retrieval from PM observations. Bobylev et al. [2010] present a neural network approach to retrieve IWV and CLW over open water in the Arctic using microwave radiometer data. They estimated the error in the retrieval and compared it to other algorithms. The method of Karstens, Simmer, Purpecht [1994] provided reliable results over open water. It is based on the following equation where CLW is in kg/m2: CLW = 3232 16 + 2 28 T b19v + 492 98 ln 280 − T b22v − 1850 55 ln 280 − T b37v + 433 686 ln 280 − T b37h (7.100) In the polar regions the precipitation was observed to occur at CLW > 0.35 kg/m2 [Kern, 2001]. However, due to scattering of microwave signal off precipitation particles, estimations of CLW tend to be biased toward values above 0.35 kg/m2. Kern [2001] presents a table of monthly averages of IWV, CLW, and V, in the Arctic and Antarctic. The data are based on retrievals of these parameters from SSM/I. In general, scattering by precipitation particles (in liquid or solid form) decreases Tb at the higher frequencies of PM, while absorption and emission increases it [Fuhrhop and Simmer, 1996]. The net effect is an increase

329

of Tb with IWV or CLW. Likewise, an increase in the surface wind V over ocean causes roughening of the water surface, and this increases the brightness temperature from both horizontal and vertical polarizations. However, the increase in Tb from horizontal polarization occurs at a higher rate, which causes the polarization difference (Tbv − Tbh) from open water to decrease significantly with increasing wind speed [Fuhrhop and Simmer, 1998]. This leads to overestimation of ice concentration in presence of rough open water surface (section 11.2.2). Ice, on the other hand, has remarkably less polarization difference than the surrounding open water. Upon estimating the required meteorological parameters, a correction scheme can be applied to account for the effects of those parameters on the observations. A commonly used scheme involves using an empirical equation to subtract the effects of IWV, CLW, and V from the observed brightness temperature. The equation can be developed by compiling a database of simulated observations using a radiative transfer model with different inputs of IWV, CLW, and V. This approach was adopted in Kern [2004] where the MicroWave MODel (MWMOD) [Fuhrhop and Simmer, 1998] was used to calculate simulated brightness temperature for the operational PM frequencies. From the database, a linear polynomial regression was developed and used to determine corrected values of the brightness temperature. The correction for IWV and CLW is performed using a single equation by subtracting the contributions of these two parameters from the observed Tb, p for a given polarization p. If the footprint contains open water and ice, an initial estimate of ice (or equivalently water) concentration C is needed. The correction, denoted by Tb, p − corr − WL, is given by the following equation (WL denotes correction for IWV and CLW): T b,p − corr − WL = T b,p −

n k=1

Ck

4 i,j = 0

aij εp V

IWVi CLW i

k

(7.101) Here the subscript k refers to each one of the two surface types (i.e., ice and open water), n is the number of surfaces (=2). The term between the square brackets in the RHS is a fourth-order polynomial (i = j = 4), where the coefficients aij depend on the surface emissivity εp. The latter varies with the wind speed if the correction is made over open water. A lookup-table of the coefficients, along with the open water emissivity variation with wind speed εp(V), is given in [Kern, 2001]. Different sets of coefficients are obtained for the ice and water surfaces. The correction for the ice and OW in a heterogeneous footprint is weighed by their concentrations Ck in the above equation. Correction for the influence of wind speed over ocean

330

SEA ICE

surface is carried out using the following equation. The parameter at the LHS is the brightness temperature corrected for the three factors of IWV, CLW, and V; denoted by the letters WLV: T b,p − corr − WLV = T b,p − corr − WL − C OW

4

bV i=0 i

i

(7.102) where, the coefficients bi are also generated from regression of data obtained using MWMOD and provided in Kern [2001]. COW is the concentration of OW. An initial concentration should be assumed. The increase of Tb caused by IWV and CLW becomes more significant at higher frequencies and over open water areas (since they have more clouds and water vapor). Therefore, it is important to correct the high-frequency observations over marginal ice zone. Other atmospheric factors that affect Tb include scattering by drizzle, frozen hydrometeors, and raindrops. They become increasingly important at frequencies greater than 30 GHz [Gasiewski, 1993]. The first two factors cause a decrease of brightness temperature, but, on the other hand, emission from raindrops also increases. The net effect of these two opposing mechanisms depends not only on the frequency of the recorded microwave signal but also on its polarization. Although the effect has not been modeled and fully described, Shokr, Asmus, Agnew [2009] found experimentally that rain over open water causes more increase of Tb from the horizontal than vertical polarization, particularly from higher frequency channels. Their data were obtained from measurements of microwave emission from laboratory-grown sea ice in an outdoor tank using a surface-based radiometer (Figure 9.13). This depolarization of the wave causes open water to be misidentified as sea ice by most ice concentration algorithms. 7.9.2. Seawater Understanding radiometric properties and processes of seawater is crucial for sea ice parameter retrieval from PM data because of two reasons: (1) the radiometric signatures of sea ice and seawater overlap under certain conditions (e.g., wet sea ice surface or rough seawater surface) and (2) the possibility of coexistence of ice and water in the same footprint of the observation. Moreover, the radiometric properties (as well as radar backscatter) of newly formed ice (< 1 cm), mark a significant transition from those of open water. The differences in seawater properties, compared to sea ice, between VIS, TIR, and microwave regions should be understood in order to interpret the data correctly. For example, while seawater is almost a perfect blackbody in the TIR region, it is a gray body in the microwave frequencies with a wide range of frequency-dependent emissivity. Understanding radiative

processes of open water is particularly important for quantitative analysis of sea ice data in the marginal ice zones.

7.9.2.1. Seawater in the Optical and Thermal Infrared Data Seawater is a highly absorbent medium in the optical region of the EM wave. The reflection coefficient varies between 0.02 and 0.1 in the VIS, and about 0.008 in the NIR regions, respectively. This property sustains the average temperature of the earth at 15 C, and hence sustains life. If the entire 71% seawater cover (of which 13% is covered with sea ice) freezes, the energy absorption would drop significantly and the average temperature of the earth will drop drastically from 15 C to −40 C. The reflection, absorption, and emission of seawater in the optical and IR regions are determined by five water surface parameters: salinity, turbidity, temperature, emissivity, and roughness. Influence of these parameters on the radiation in the optical spectrum is presented in several text books [e.g., Wozniak and Dera, 2007]. A brief account of these influences is discussed in the following. The salinity and turbidity are particularly important for the VIS and NIR observations. However, the turbidity of the water is not relevant to the remote sensing of sea ice because it exists in significant qualities only near shore, i.e., far from the open ocean where most of the mobile sea ice exists. The SST and emissivity are the two factors that determine the TIR observations. Since water is almost a blackbody at TIR wavelength, the emissivity of saline water varies over a narrow range between 0.987 and 0.993 [Konda et al., 1994]. It is a weak function of the salinity. Part of the solar radiation that strikes the seawater surface is refracted and travels at a slower speed in water compared to air. The refractive index that determines the amount of refraction is a function of the water salinity and temperature [Austin and Halikas, 1976]. It increases with increasing salinity and decreasing temperature. This relationship is used to determine the seawater salinity by measuring the refractive index of a sample at a constant temperature. The radiation that penetrates the ocean surface is attenuated by absorption. The average percentage attenuation of the blue and red wavelengths in one meter is 6.5% and 46%, respectively. The non-absorbed radiation is scattered and part of it may eventually reach the surface and get refracted back to the air. This and the reflected part determine the ocean color. The color is detected by satellite sensors such as the NASA’s Seaviewing Wide Field-of-view Sensor (SeaWiFS). For clear water, regardless or surface roughness, the spectral albedo is essentially independent of the wavelength in the optical range [Perovich, 1991].

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

Calm sea Sun glint

Figure 7.32 MERIS image showing sun glint in a scene in the South Pacific [Kay et al., 2009 / Reproduced from MDPI / CC BY 3.0].

Roughness of the ocean is manifested in the form of surface waves, which are mainly triggered by surface wind. It is the key factor in determining the scattering of EM waves in general and particularly in the microwave region. In optical remote sensing, if the angle of the reflected beam is nearly equal to the viewing angle, reflectance will be high. This phenomenon is called sun glint (Figure 7.32). Its occurrence depends on the sun’s position, the viewing angle of the sensor, and the local tilt of the ocean surface. The orientation and the extent of sun glint have been used to identify regions of calm sea surface (sun glint has a better chance to occur in calm seawater). This phenomenon has no bearing on sea ice but it has to be corrected for if oceanic information such as the level of chlorophyll is sought. A review of the correction methods is presented in Kay, Hedley, Lavender [2009]. Ocean foam (or whitecaps) is another phenomenon that affects the reflection of the solar radiation from seawater. It is generated during active breaking of the ocean waves when the ratio of the wave height to its wavelength is greater than 0.1. At a wind speed of 20 m/s the foam covers one-third of the open ocean [Ross and Cardone, 1974]. Ocean foam (with its heavily bubbly contents) changes the optical and microwave observations of the surface. Since bubbles scatter light, reflection from foam is usually high, and therefore the foam-covered surfaces appear white (hence, the name whitecap). The reflection is isotropic and independent of the wavelength in the VIS region but decreases with wavelength in the NIR region. Since this phenomenon usually occurs in the open ocean, it does not affect the interpretation of sea ice imagery data. It only impacts a few air–sea processes at the ocean surface, including gas fluxes and turbulent mixing. Spectral reflectance of whitecaps is reviewed in Kokhanovsky [2004]. 7.9.2.2. Seawater in the Microwave Data Compared to the optical and infrared observations, knowledge about microwave scattering mechanisms from

331

ocean surface is more critical for the interpretation of sea ice imagery. The ocean generates waves with a wide range of wavelength in response to the wind forcing. Some wavelengths and heights resonate with the incident microwave signal (unlike the optical radiation with its extremely small wavelength). In the microwave region, different scales of wavelength of ocean surface activate different scattering mechanisms. A particularly important mechanism is the well-known Bragg scattering, which is responsible for high power of the received backscatter. Bragg scattering represents the constructive addition of the radar signal when scattered off successive ocean waves. To fulfill the condition for Bragg scattering, the wavelength of the incident radar wave should be twice the wavelength of the ocean wave (measured in the slant range). This relationship is presented later in this section, but first a brief presentation on categories of ocean waves in relation to the microwave scattering mechanism is introduced. Detailed information on this subject can be found in a few text books on microwave remote sensing of ocean [e.g., LeBlond and Mysak, 1981, Robinson, 2004, Comiso, 2010]. Ocean waves are categorized into four classes based on their wavelengths: (1) capillary waves, (2) gravity waves, (3) swell, and (4) tsunamis. The first three are generated by wind in open ocean, and they generally coexist as their driving forces act simultaneously. Tsunamis, which are not relevant to this discussion, are caused by geological effects such as earthquakes in deep water. Capillary and gravity waves are particularly relevant to the identification of the microwave scattering mechanisms. Capillary waves, also known as ripples, are small waves with a height of about 20 mm and a typical wavelength of a few millimeters though it can reach 17 mm. They are generated by fresh wind (light breezes) blowing on a smooth water surface. Their dynamics are dominated by surface tension. The frication caused by the wind stretches the water surface, causing waves to be generated as surface tension tries to restore it to smoothness. Capillary waves are usually found where a boundary is placed around the water body such that it restricts the growth of the waves. This is frequently observed in narrow leads and polynyas. Beyond the wavelength of 17.3 mm the gravity supersedes capillary action at the ocean surface as the dominant restoring force of the wave (i.e., to restore the displaced water element toward equilibrium). The oscillation about the equilibrium state in this case (i.e., on the air–sea interface) is known as surface gravity waves. The wavelength is typically a few meters, but it can reach tens of meters. There is a range of overlap between capillary and gravity wave called capillary-gravity wave. As gravity waves build up, their wavelength tends to lengthen and the speed increases until it matches the speed of the wind. At this point they can no longer extract energy from the wind. Gravity waves inside the water body, i.e., between layers

332

SEA ICE

s ck Ba rin tte ca

ds

g

θ

inθ

r tte

θ

ing

It is clear that the wavelength of the ocean wave has to be almost equal to the radar wavelength for Bragg scattering to occur. The wavelength range of the operational SAR systems is between 2 and 30 cm (for the x- and Lband, respectively as shown in Table 7.2). This falls into the wavelength range of the capillary-gravity wave. The minimum wind speed that triggers a dominant (first-order) Bragg scattering in the C-band ( 5.4 cm wavelength) is estimated to be about 3.25 m/s at 10 meters above the surface. It should be noted, however, that the

ca

(7.103)

s ck

λ = 2d sin θ

(a)

Ba

of different densities, are called internal waves (unlike the surface gravity waves, they are not generated by wind). The next category of wind-generated ocean surface waves is the swell. It is not generated by the local wind but rather by a series of gravity waves generated elsewhere or sometime ago by the wind action. Swell waves have a wide range of wavelengths measured in hundreds of meters and depend on the size of the water body and the wind event that originates them. In most severe storms, the wavelength of swells is occasionally longer than 700 m. Gravity waves and swells penetrate sea ice at its edges, leaving their imprint as a wavy pattern in radar images of the ice sheet. This is usually observed when the ice is thin enough. The last category of tsunamis features huge waves measuring a few hundreds of kilometers in wavelength (100–500 km) and 100–200 m in height. They are not visible in the ocean but they grow to devastating proportions near the coast due to reduced water depth. Tsunami is a Japanese word meaning “harbor wave.” They are not relevant to remote sensing of ocean and certainly not of ice. If the roughness scale of the ocean surface is much smaller than the wavelength of the incident microwave signal, the surface is considered to be smooth and the scattering follows a specular pattern, producing near-zero backscatter. This is particularly manifested at large incidence angles. Backscatter increases as the surface roughness scale increases with respect to the wavelength of the microwave signal until it reaches its peak. The peak is triggered by Bragg scattering. In this mechanism, the radar signal resonates with components of the surface wave spectrum. The radar signal is scattered from successive ocean waves and added coherently as shown in Figure 7.33a. The scattering from different waves may or may not be in-phase because of the different path lengths as shown in the figure. If scattering from successive waves is in-phase the waves may then interfere constructively, producing high backscatter. This is the essence of Bragg scattering. The Bragg equation relates the ocean wavelength d required to trigger the Bragg scattering to the radar wavelength (λ) and the incidence angle θ measured from the datum of the ocean surface:

d

(b) ble Dou

Wind direction

g

ncin

bou

ent

Incid

e wav

Ocean curren

t

Figure 7.33 Two configurations of radar scattering mechanisms from ocean wave: (a) Bragg scattering and (b) DB scattering mechanism generated by huge gravity waves or swell. Capillary waves that trigger Bragg scattering can be superimposed on gravity wave that generates the DB.

received signal from Bragg scattering is Doppler shifted with a specific shift determined by the velocity of the gravity wave that fulfills the condition for Bragg scattering. Naturally, the ocean wave is composed of many waves of different wavelengths. In this case, the radar backscatter may be triggered by Bragg scattering and modulated by the larger surface gravity waves. These waves as well as the swell may cause DB scattering mechanism of the radar signal as shown in Figure 7.33b. Compared to radar signal, the emitted radiation from PM is less affected by surface roughness of the ocean. The wavelength of the operational PM sensors occupies a range between 0.33 and 1.67 mm (Table 7.3), which is smaller than the typical wavelength of the capillary waves of the ocean surface (2–17 mm). However, water surface is highly polarizing in the PM frequencies. It is also polarized in radar data near the Brewster angle (between 83.5 and 66.5 for microwave frequencies between 1 GHz and 100 GHz). Operational PM sensors have typical viewing angle of approximately 53 . This ensures a large polarization difference between the horizontal and vertical emission. That is why PM data are used to discriminate between open water and ice, which has much less polarization difference. The advantage of the polarization difference, which is triggered by calm seawater, is lost gradually as the sea roughness increases. Ulaby, Moore, Fung [1986] present

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

results of seawater emissivity for different wave roughness scales using a facet-model, which is a modified Fresnel reflection model. They used probabilities of sea surface slopes as a function of the wind speed. The authors concluded that the seawater emissivity at 19 GHz horizontally polarized EM wave at incidence angle close to that of operational PM sensors increases with surface roughness. Physically speaking, this is explained by the increasing surface area of the rough surface. At the same time the vertically polarized emission is independent of the windinduced roughness. The difference between the emissivity from the vertical and horizontal polarizations for rough water surface is less than that of calm water by 0.2–0.25 K [Comiso et al., 1992]. This confuses the discrimination between ice and open water in PM observations when seawater becomes rough. However, it can be used to infer the wind speed over ocean [Swift, 1980, Goodberlet, Swift, Wilkersen, 1989]. It should be noted that the observed PM signal is integrated over large FOV areas that measure in tens or hundreds of square kilometers. This is too large to have a uniform sea roughness scale across the footprint. Ocean foam increases the microwave emissivity in both polarizations over the entire range of incidence angle. This has been shown in a study by Camps et al. [2005] on the emissivity of foam-covered water surface at the microwave L-band. They found that the foam-induced emissivity increases at a rate of approximately 0.007 per millimeter of foam thickness when extrapolated to nadir direction. They also found that the foam cover increases the polarization difference relative to the foam-free surface. This is contrary to a finding by Stogryn [1972], which confirmed that foam causes a significant decrease in polarization difference. Ulaby, Moore, Fung [1986] present data showing emissivity increase due to presence of foam at 20 GHz frequency as a function of the ratio of foam layer thickness to the wavelength of the PM emission. At this frequency, the emissivity increases from 0.42 over foam-free open water to 0.92 over a foam layer with the above-mentioned ratio equaling 1. The emissivity saturates at this level for thicker foam layer. Foam is highly bubbly medium, which scatters EM with almost all wavelengths. 7.9.3. Snow on Sea Ice: Physical and Radiative Processes Snow is one of the most complex and changeable substances on Earth. While ice exists in nature near its melting point, snow exists near its “triple point,” meaning that the transition between its three phases of solid, liquid, and vapor can take place very fast. Dramatic and rapid changes of snow occur from the instant snow hits the ground. This starts a long process of metamorphism.

333

The process is triggered by changes in atmospheric temperature, pressure, humidity, and other forms of precipitation that fall at the snow surface. Even a subtle change in any of these parameters can have a significant effect on the snow form and properties. That is how the presence of snow modifies satellite observations from the underlying sea ice and complicates the ice parameter retrievals. Unless the snowpack on sea ice is fresh and dry, its presence can impact the remote sensing observations significantly. Langlois and Barber [2007] present a creditable review on the current state of knowledge pertaining to the geophysical, thermal, and dielectric properties of snow on sea ice as well as the different microwave emission and scattering mechanisms associated with it. The factors that affect the reflection, emission, and microwave scattering include snow metamorphism (which implies grain size and snow density), depth, salinity, and wetness. Impacts of these factors on the observations are discussed in the following. Snow metamorphism takes one or more of the following forms: snow transformation to ice crystals (called snow grains), ice lenses, surface crust layer, and surface glaze. Except for the surface crust, which is wind-driven, the other forms are thermodynamically driven. Snow grains are the most common form of metamorphism. Under its own weight or when exposed to midwinter melt-refreeze cycles, snowflakes are transformed into ice crystals (grains) that may bond together and squeeze the air between them. The snow density increases accordingly. Snow grains may also be formed as a result of a high-temperature gradient within the snowpack. Hightemperature gradient (in excess of 0.3 /cm) generates vapor diffusion (snow sublimation). The vapor transport from the snow base upward into colder temperatures, resulting in fast growth rate of grains with dendritic structure and faceted shapes. This kind of growth takes place more often near the snow base and is responsible for forming what is known as hoar layer. In areas where temperature gradient is low, usually near the top of the snow, grains tend to form into well rounded tiny grains with fine texture. Another form of snow metamorphism is ice lenses. These are frozen layers of water that accumulate at the snow surface or within the snowpack through drainage. They tend to reduce the bulk temperature of the snow and create impermeable ice barrier layers. Snow crust is wind-generated snow packing formed at the surface. Wind-driven compacted crusts may form a wavy pattern known as sastrugi. Glaze is a smooth impermeable surface, which is developed at the snow surface when rainwater freezes. All forms of snow metamorphism affect the microwave scattering and emission properties, in some cases significantly. Empirical data and modeling of these phenomena are needed to better understand and interpret remote sensing observations.

334

SEA ICE

Snow depth (cm)

The most significant snow metamorphism takes place within the hoar layer at the snow base. All physical parameters in that layer are significantly affected. An example is shown in Figure 7.34 from a study conducted by Barber and Nghiem [1999] that aimed at exploring the effect of snow thermodynamic processes on radar backscatter in the C-band. The data in the figure are representative of the shown parameters in the presence of a metamorphosed saline snow within the hoar layer near the ice–snow interface. This layer, shown within the bottom 4 cm in Figure 7.34, features higher salinity due to

12

May 6

12

10

May 9

10

8

8

6

6

4

4

2

2

0

0 0

5

12.5

20

2

50

150

250

350

0

2

4 6

8

10 12

Average grain size (mm2)

Density (kg/m–3)

Salinity (%°)

Snow depth (cm)

brine wicking mechanism by the snow from the highly saline ice surface. It is also characterized by relatively large snow grains (size > 4 mm2). This is usually associated with relatively low density since the large grains cannot be packed as tightly as the fine grains. Snow density varies significantly with age and compactness as shown in Table 7.4. Note that sea ice density occupies the range 830–917 kg/m3. The two components of the complex dielectric constant at the snow base increase following the increase of brine volume at the ice–snow interface.

12

12

10

10

8

8

6

6

4

4

2

2

0

0 0 .005

.015

.025

Brine volume fraction

2

2.5

3

3.5

Modeled permittivity (ϵ′)

.05

.15

.25

Modeled loss (ϵ″)

Figure 7.34 Profiles of physical parameters in snow pack overlaying FYI in Barrow Strait, central Arctic. The salinity, density, and average snow grain size were measured from samples extracted at approximately solar noon on 6 and 9 May 1993, while brine volume fraction, permittivity, and loss factor were calculated using models presented in Barber and Nghiem [1999] [Barber and Nghiem (1999), Figure 2 / John Wiley & Sons].

Table 7.4 Density of different snow types. Snow type Density (kg.m-3)

New snow

Damp new snow

Settled snow

Depth hoar

Packed snow

Firn snow

50–70

100–200

200–300

100–300

350–400

700–800

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

7.9.3.1. Snow in Optical and Thermal Infrared Data In principle, discrimination between ice and water is possible from optical data because there is enough contrast between the albedo from the two surfaces. Albedo of seawater is usually below 0.1; whereas albedo of bare ice surface varies between 0.5 and 0.7. Albedo of snow on sea ice features a wide range of values; from 0.9 for fresh snow to 0.5 or lower values as the snow ages [Grenfell and Maykut, 1977]. When the snow starts to melt the albedo decreases to 0.4 or lower. Because of this wide range of albedo, the status of the snow on ice becomes important for determining the onset and the rate of ice melt (melting accelerates when the albedo of the overlain snow is low). For that reason, many studies were launched to explore the effect of physical properties of the snow on its optical behavior in relation to the wavelength of the incident radiation. A comprehensive review on the subject is given in Warren [2012]. It demonstrates the dependence of albedo on the snow grain size, wetness, impurity contents, and wavelength and viewing angles of the radiation. Other details on optical remote sensing of snow-covered sea ice can be found in Warren [1982], Zhou [2002], and Pedersen [2007]. Kashiwase et al. [2017] have shown that the increasing rate of breaking and divergence of the Arctic ice cover during early melt season triggers more spatial contrast of albedo, which leads to more divergence. This explains, at least partly, the recent acceleration of the Arctic sea ice loss during summer. The NSIDC reported that the air temperature across much of the Arctic increased by nearly 4 C during the period 1960–2019 (https://nsidc.org/cryosphere/arcticmeteorology/climate_change.html). Based on this information, more temperature fluctuations around the melting point of the snow are expected with more snow metamorphism. This affects albedo as well as TIR and microwave emission. Albedo decreases slowly with increasing wavelength in the VIS region and more rapidly in the NIR region. It can reach low values below 0.1 for NIR wavelengths between 1.5 and 2.0 μm. For that reason, NIR data are used to discriminate between snow and ice on one hand (with their low albedo) and clouds on the other hand (with their relatively high albedo) [Sandven and Johannessen, 2006]. The albedo of a thin snow layer depends on the albedo of the underlying ice surface. When the snow becomes optically thick, the effects of the underlying surface can be ignored. In addition to the wavelength of the incident radiation, the optical depth of the snow depends primarily on the snow grain size. A snow layer of 10 cm depth would make it impossible to identify the underlying ice type in the visible or NIR wavelengths [Perovich, 1991]. Snow density indirectly affects the optical properties of snow through the grain size. As described above, the grain size increases as the snow pack ages from about 50 μm for

335

new snow to about 1 mm for old snow that has gone through wetness and re-freezing. This results in an increase of density and a decrease in albedo. The grain size of the snow also affects the scattering pattern of the optical and infrared radiation. In an early study on the optical properties of snowpack, Bohren and Barkstrom [1974] found that most of the scattering of light beam in the snow is the result of change in direction upon transmission through snow grains, rather than reflection. Wiscombe and Warren [1980] presented plots showing the effect of snow grain size and wetness on albedo (see section 9.1, Figure 9.5). They also presented a model for calculating the spectral albedo of snow in the VIS/IR spectrum. It accounts for the direct or the diffuse incident radiation at any zenith angle. Zhou [2002] found that snow albedo in the NIR bands is more sensitive to the grain size at the top layer (< 5cm), while the albedo of the visible bands is sensitive to the grain size of a much thicker snow layer. Therefore, the NIR solar irradiance plays a major role in the energy balance at a snow surface and consequently snowmelt. Liquid water in snow does not affect the albedo as much as it affects the absorption of the VIS/NIR radiation, hence the emission of TIR radiation. It decreases the albedo slightly, but when the snow starts to melt the albedo starts to decrease sharply and that leads to more snowmelt (i.e., a positive feedback). Wet snow is encountered more in the snow cover on YI during fall or on thicker ice in the spring. Liquid water in snow is a catalyst for snow grain clustering. Upon re-freezing, larger effective grain sizes of the snow cause more decrease of albedo. This was verified using field measurements obtained in the Antarctic during the austral summer of 1999 [Zhou, Li, Morris, 2001]. It is worth mentioning that the albedo of the snow on glaciers and ice sheets can be reduced due to the presence of powdery wind-blown dust made of a combination of small rock particles and soot, called cryoconite. The long periods of low sun elevation in the polar areas limit the possibility to estimate the surface albedo using optical data. On this account, Yackel et al. [2007] explored the correlation between albedo and the backscatter from snow-covered sea ice, hoping to develop a capability to retrieve albedo from SAR backscatter measurements. The study used an extensive data set collected during the Seasonal SIMMS field program from 1992 to 2001 except the year 1996, and the CollaborativeInterdisciplinary Cryospheric Experiments (C-ICE) from 1999 to 2001. All data were obtained from Lancaster Sound near Resolute Bay in the Canadian Central Arctic. The authors modeled the stable winter ice backscatter σ 0w as a function of the radar incidence angle and measured the seasonally evolving backscatter coefficient σ 0. They found an inverse correlation, valid during the early melt season, between the optical albedo and the deviation of

SEA ICE 10

0.9

8

0.8 Site 1 FY ice

6

0.7

o

ΔσW

Site 2 FY ice

4

Instantaneous albedo

0.6

2 0.5

0 –2 –4 135

0.4

June 7 140

145

150 155 160 Day of the year

165

Instantaneous albedo

336

170

0.3 175

Figure 7.35 Time-series of the deviation of backscatter coefficient of snow-covered, smooth FYI in the Arctic from the typical winter values (Δσ0W). Backscatter was measured using RADARSAT-1. The coincident measured albedo is also shown. A least-squares cubic polynomial is fitted to each dataset. Inverse correlation between Δσ0W and the albedo is visible after 7 June [Yackel et al., 2007, Figure 9 / with permission from Wiley].

the backscatter from its typical winter values. That deviation is expressed as Δσ 0W = σ 0 θ − σ 0w θ , where θ is the radar incidence angle. Figure 7.35 displays this relationship. It shows time evolution of Δσ 0W measured from RADARSAT-1 images over two test sites of snowcovered smooth FYI and the “instantaneous” in-situ measurements from the snow-covered sea ice within 15 minutes of the satellite overpass. The satellite measurements were obtained from a variety of beam modes using ascending passes only. The incidence angle was accounted for. The inverse correlation between Δσ 0W and the albedo measurements is visible in the data of the early melt season (June 7 to June 21). σ 0 increases during this period by up to 8 dB, and the albedo decreases from about 0.8 to 0.3. Yackel et al. [2007] attributed the increase in backscatter to two factors: the growth of snow grains at the snow basal plane, which causes large volume scattering, and the increase of dielectric constant near surface following an increase in snow wetness. This leads to an increase in surface scattering although the wetness attenuates the volume scattering. 7.9.3.2. Snow in the Microwave Data The factors that influence the microwave interaction with the snowpack over sea ice (passive or active) include snow thickness, density, wetness, salinity, grain size, and snow layering at the surface or within the snowpack (e.g., ice crust, lenses, and glaze). These factors interact, so it is not possible to isolate each factor in an experimental or modeling study. That is why study of microwave emission or backscatter from snow is challenging. However, established effects of a few of the

above-mentioned factors are addressed below but first, a brief account of a few key characteristics/processes of snow on sea ice is presented. Details on microwave emission and radar backscatter from snow on sea ice are presented in Markus and Cavalieri [1998] and Langlois and Barber [2007]. Sturm et al. [2006] presented detailed characterization of the physical properties of snow on sea ice in Barrow, Alaska, based on measurements from 118 snow pits. A snow microwave emission model, known as Helsinki University of Technology (HUT) is presented in Pulliainen, Grandell, Hallikainen [1999]. It describes the emission behavior of a homogeneous snowpack as a function of snow-water equivalent, grain size, and density. So far, this model has not been applied to snow on sea ice, but Gunn et al. [2011] applied a modified version to model microwave emission from snow over lake ice. Snow properties are affected by properties of the underlying sea ice. This includes surface roughness, salinity, and the presence of a slush layer at the snow–ice interface. These properties determine the emitted radiation from the snow base. Brine at the subsurface of FYI can be wicked into the snow. Usually, the lower 4 cm layer of snow on FYI contains significant amount of brine (Figure 7.34). A salinity of 4‰ at the snow base would double the value of the dielectric loss (recall that the loss is a strong function of the brine salinity and brine pocket shape, as explained in the last part of section 3.7.2). On the other hand, snow on MYI is almost saline free. That is why it is easier to identify the effect of snow wetness on emission/scattering over MYI than FYI. Depending on the wavelength, the microwave signal in the snowpack may undergo scattering and absorption. Scattering occurs when the signal encounters elements

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

with dielectric constant different than that of the host medium. This includes air inclusions and ice grains of millimeter or sub-millimeter scales in the case of PM, or centimeter or sub-centimeter scales in the case of radar. Absorption is caused mainly by the moisture within the snow and the highly saline hoar layer at the snow base. For all operational PM frequencies, the penetration depth of microwave radiation in the snow decreases as the snow wetness or salinity increase (see section 9.5). Beyond a certain upper limit of snow wetness or grain size the snowpack masks the information about the underlying ice to a great extent. This affects the retrieval of ice types, concentration, thickness, and surface temperature. An example that shows the masking of air bubbles (main radar scattering elements) in MYI is presented at the end of this section. Cycles of melting and re-freezing cause changes in snow grain size, shape, and may add ice layering within the snowpack or ice crust at the snow top. These features cause significant scattering, hence decrease Tb especially at lower frequencies (≤ 37 GHz) [Comiso et al., 1997]. This effect is more pronounced for the horizontal polarization at high incidence angles due to the lower penetration depth and stronger snow layering effects [Hallikainen, 1989]. Tonboe, Andersen, Toudal [2003] also found that the re-freezing of liquid water in the snow decreases attenuation and increases scattering of microwave emission, leading to an increase in Tb. To account for the snow contribution in microwave observations, snow depth and its gross features must be known. This is the greatest challenge in interpreting microwave observations. Daily maps of snow depth in the Arctic have become available from NASA’s Goddard Earth Science Center, but snow properties are always unknown. While snow wetness has direct impact on modulating the emitted or scattered microwave signal from the snow cover, it has an indirect impact on the information from the underlying sea ice. For example, it partially or fully masks volume scattering from the bubbly layer at the subsurface of MYI. For this reason, it is important to distinguish between what the snow wetness does to the signal from the snow, and what it does to the signal from the combination of the snow and the underlying ice. The two effects are discussed in the following. Effect of Dry Snow Depth In principle, the effect of snow depth on microwave emission/scattering cannot be correctly characterized unless the snow is dry. That is why attempts to correlate snow depth with satellite observations have produced conflicting information as they are not usually accompanied with full description of the snow conditions (e.g., snow wetness, stratification, grain size, etc.). For example, Lohanick and Grenfell [1986] found no correlation between microwave emission and snow

337

depth from measurements of Tb at 37 GHz conducted over cold FYI of thickness up to 30 cm near Tuktoyaktuk, Northwest Territories. On the other hand, Markus and Cavalieri [1998] presented correlation between Tb from 37GHz, obtained from SSM/I, with in-situ snow depth measurements from the Bellingshausen-Amundsen Sea in the Antarctic. The correlation coefficient over a depth up to 80 cm was −0.64 and −0.66 for observations from vertical and horizontal polarization, respectively. These correlations were confirmed in later studies and therefore have led to the general acceptance, though with some reservations, of the inverse relationship between microwave brightness temperatures and dry snow depth. This is caused by the increased scattering within the snow pack given the small wavelength of the PM emission. The relationship applies to both horizontal and vertical polarizations, and the rate of decrease of brightness temperature is higher at higher frequencies as radiation at those frequencies are scattered more strongly off the air inclusions in the snow. This means that brightness temperatures at 37 GHz are reduced more than brightness temperatures at 19 GHz with increasing snow depth. The implication is that the gradient ratio between 37 and 19 GHz channels GR37V19V (defined in equation (9.2) can be used as an indicator of snow depth. This ratio is always close to zero for bare ice surface. Scatter plots that depict this relationship are presented in Markus and Cavalieri [1998]. Retrieval of snow depth using this feature is presented in section 10.3.2. Effect of snow density Fresh snow becomes compacted under its own weight. Densification of snowpack increases its permittivity, hence suppresses microwave emission from the underlying sea ice [Pullianen and Hallikainen, 2001]. The increase of permittivity will also reduce the penetration of the radar signal. This increases the backscatter from snow and hinders its interaction with the underlying ice, hence suppressing information from the ice. Effect of Snow Grain Size and Ice Layering Snow grains are developed as a result of two processes. (1) When vapor freezes within the snow pack, and this is observed more often near the snow base. (2) After a thaw/freezing cycle when snow develops moisture as the temperature rises then freezes when the cold temperature returns. This is observed along the entire snow depth. The mean grain size increases if existing grains bond together through enhanced packing. In addition, water that percolates through the snow and later refreezes can form ice lenses and ice glands. Re-freezing of wet snow leads to the formation of ice layering within the snow pack and ice crust at the surface, both cause significant scatter and drop in Tb, especially at lower frequencies (≤ 37 GHz) [Comiso

338

SEA ICE

et al., 1997]. Among the several field measurements of snow grain size and shape was the work presented in LeDrew and Barber [1994] based on snow measurements conducted in the 1990s during the Seasonal Ice Monitoring and Modeling (SIMMS) program. Sampling snow from pits that are far apart and not covering wide area may lead to wrong conclusions because of the high spatial and temporal variability of snow properties. Colbeck [1983] and Langlois et al. [2008] suggest that the snow morphology is extremely variable and can change within a few hours. In an experimental study of snow on sea ice in Barrow, Alaska, Sturm et al. [2006] found the depth hoar grain size distribution consisted of both large grains (2–10 mm) and small grains (0.1–0.3 mm), and both within small scales. Snow grains, ice lenses, and thick crust cause the microwave emissivity to drop. This is because they enhance the scattering of the emitted radiation through its upward travel to the surface. Yan, Weng, Meng [2008] estimated the drop to be from 0.94 to 0.74 at 31.4 GHz. As a rule of thumb, for fine-grained snow crystals (relative to the wavelength of the incident signal), absorption and scattering losses are typically small, while for coarse-grained, volume scattering becomes significant. Depending on the incidence angle of the receiver, scattering of emitted radiation from FYI may or may not cause a decrease in Tb but it almost certainly increases the radar backscatter. The effect of snow grains within the snow pack on increasing radar backscattering from FYI is demonstrated in results presented in Du, Shi, Rott [2010] and

shown in Figure 7.36. The study uses a multiple layer scattering model. Backscatter from the volume scattering increases as the radius of snow grain increases. The difference between results from the multiple scattering models and the other two models (which are shown in the figure) also increases with radius of the snow grain. The copolarized snow volume backscattering simulated from the three models has comparable values from each model. The cross-polarized backscattering coefficients, on the other hand, are underestimated by the first-order scattering model. Regardless of the model used to produce these results, the underlying theme is the increase of both coand cross-polarization return with the radius of snow grains before they saturate around a radius of 1.5 mm. The increase in the cross-polarization scattering is a manifestation of multiple scattering which triggers depolarization of the radar signal (section 7.6.1.3). An implication of these results is that as snow on FYI acquires wetness and then refreezes, snow grains grow bigger and therefore the backscatter from this medium increases. If this happens to the snow on FYI, it will likely make the backscatter signature as high as that from MYI and thus confuses the ice classification. This scenario can be used to explain anomalies in the Arctic FYI signature in the spring as presented later. Due to lack of field measurements of snow grain size, most of the information about the effect of this parameter on microwave signal (passive and active) has been obtained using a radiative transfer model of snow (section 12.2). Studies of sensitivity of Tb to grain size

5 Multiple scattering model

0

Multiple scattering model First-order scattering model Second-order scattering model

0

First-order scattering model Second-order scattering model

–5

σVV (dB)

σVV (dB)

–10

–10

–20

–30

–15

–40 VV polarization

VH polarization –50

–20 0.0

0.5

1.0 Radius (mm)

1.5

2.0

0.0

0.5

1.0

1.5

2.0

Radius (mm)

Figure 7.36 Simulated volume backscattering coefficient calculated from the first- and second-order scattering models as well as a multiple scattering model presented in Du, Shi, Rott [2010] as a function of grain size in snowpack. Calculations are performed for snow-water equivalent of 100 mm at frequency 17 GHz and incidence angle 40 [Du, Shi, Rott, 2010, Figure 4 / with permission from Elsevier].

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

are presented in Li, Tan, Tsang [2006], Durand, Kim, Margulis [2009], Kontu and Pulliainen [2010], and Derksen et al. [2012]. A main conclusion from these studies is that most of the differences between observed and calculated brightness temperatures from snow can be explained in terms of snow grain size. Effects of Snow Wetness In principle, microwave emissivity increases as the snow becomes wet. This is caused by more absorption of solar radiation by water. However, an increase in the emissivity does not necessarily mean increase in brightness temperature because the latter is also determined by the physical temperature of the surface. On the other hand, radar backscatter decreases as the snow wetness increases since wet snow is a lossy medium. However, wet snow on FYI triggers another property that contributes to the microwave emission/scattering. As mentioned earlier, the bottom few centimeters at the snow base usually acquire brine through vertical expulsion from the underlying saline sea ice surface. Therefore, snow wetness causes brine volume in this layer to increase. This lossy medium may partially or fully mask the microwave emission and radar scattering from the underlying ice. On the other hand, when the saline-free snow on MYI is exposed to near melting temperatures, only snow wetness will determine the emission and scattering from the snow. Thus, modeling these quantities from low-salinity MYI is easier than high-salinity FYI. However, in the case of MYI, snow wetness may mask the radar signal interaction with subsurface bubbly scattering layer, hence rendering this ice dark (low backscatter) similar to the FYI. More discussions on the effect of snow wetness on the emissivity and dielectric constant follow. Contrary to the basic principle stated above, Markus, Stroeve, Miller [2009] found the emissivity from wet snow to be less than dry snow. Snow wetness lowers the emissivity in the microwave region and eventually brings it to values close to those of OW. For Arctic sea ice, Mathew, Heygster, Melsheimer [2009] reported a similar drop of FYI emissivity in the presence of wet snow during transition season (April and May). This may decrease the brightness temperature from FYI. Lohanick [1993] reported a dramatic decrease in Tb at 10 GHz due to the presence of a slush layer at the snow–ice interface immediately after the snowfall. Ye et al. [2016] showed that brightness temperature from FYI decreased when the snow became wet during the spring transition season. On the other hand, an early study by Comiso [1985] showed that emissivity of snow-covered MYI increased with increasing snow wetness (data were available up to 5% volumetric water content), causing the brightness temperature of MYI to approach values of FYI. This information is confirmed in Mathew, Heygster, Melsheimer [2009] during the spring season in the Arctic when snow

339

on MYI becomes wet. The subsequent change of microwave emission from snow-covered MYI is not fully understood but it is likely to increase. Since wet snow is a lossy medium at microwave frequencies, its dielectric constant increases as wetness increases. The increase in the real part causes an increase in backscattering off the surface but accompanied with a decrease of volume scattering. The net effect is an increase in backscattering as snow wetness increases. This is confirmed in a study on saline-free snow presented in Koskinen, Pulliainen, Hallikainen [2000], where the authors show a linear increase of backscatter from −22 dB to −16 dB as the snow wetness varies from 0% to 10%. Barber and Nghiem [1999] reached the same conclusion from a modeling study of backscatter from snow over FYI. The model estimated an increase of 3.5 dB associated with physical change in snow induced by temperature increase. In cold temperature, the σ ovv at the C-band was −22.1 dB at a look angle of 23 , but this value increased to −18.6 dB in warm snow cover. Ye et al. [2016] demonstrated an increase of backscatter from snow-covered FYI using Ku-band scatterometer data as the snow became wet. The high backscatter reaches values typical of MYI, which confuses the discrimination between these two types. At the same time the backscatter from MYI decreases because wet snow suppresses the volume scattering from the bubbly underlying MYI. Shokr and Dabboor [2020] also show decrease of backscatter from bubbly lake ice when the overlaid snow became wet. Effects of snow wetness on suppressing the backscatter and enhancing the PM emission from MYI are presented in Shokr and Agnew [2013] using data from the QuikSCAT Ku-band scatterometer and AMSR-E 36.5 GHz radiometer (Figure 7.37). The top two panels show mosaics of Tb, 36h from AMSR-E and backscatter σ 0hh from QuikSCAT, constructed from all available satellite orbits on 18 September 2007. The figure shows the effect of snow wetness on the observed signal from MYI (the bright area in the QuikSCAT image). An anomalous signature of the MYI is marked by the solid arrows. It features higher-than-normal Tb,36h in the AMSR-E image and lower-than-normal σ 0hh in the QuikSCAT image. These anomalies are linked to a northeast heat wave caused by a large cyclone just north of Greenland that brought warm southerly wind (temperature around 0 C) into the region as shown in the near-surface temperature map (12 m above the surface). The map is produced using the Canadian GEM weather model. The warm spill coincided with the location of the signature anomalies in the image. Warm air raises the physical temperature of the snow, and this increases the snow wetness and leads to increase in brightness temperature (for sure due to increase in physical temperature and, perhaps, an increase

340

SEA ICE Am5h–o–Arco7–261–261

AMSR HIRZ

QuickScot

quoh–o–Arc07–261–261

2007–9–18

QuikSCAT σohh

AMSR-E Tb36h

2007–9–18

–5 dB

300 K

–10 dB 250 K –15 dB 200 K AMSR-E 36.5 GHz–H 18/9/2007

–20 dB

Quitsoot L1B hh 18/9/2007

150 K –25 dB

100 K

–30 dB

Surface temperature 256 K 258K

2007–9–18 75 N 80 N

260 K 262 K

85 N

264 K 266K 268 K 270 K 272 K

80 N

274 K 276 K

75 N

278 K 70 N

Figure 7.37 Daily average mosaics of AMSR-E Tb, 36h and QuikSCAT σ0hh over the Arctic basin for 18 September 2007 (top panels), and the coincident near-surface air temperature map obtained using the weather model GEM used by the Canadian Meteorological Center (bottom). The bright area in the QuikSCAT image is MYI. Anomalies in the remote sensing signatures of the MYI image are marked by the solid arrows and they correspond to higher air temperature in the temperature map. The dotted arrow indicates an area of very cold temperature (below −20 C) [Shokr and Agnew, 2013 / from Elsevier].

in emissivity, though this is not confirmed as shown by the conflicting results presented above). The snow wetness also attenuates the radar backscatter. Three days later (on 21 September 2007) the atmospheric temperature at

the same location of the anomaly decreased to values below −10 C and the satellite signatures returned to their typical values (data not shown). The short duration of the anomaly confirms that snow on MYI restores its

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE

properties as soon as the cold temperature is restored. Snow on MYI is saline-free so there will be no loss of brine during the warm temperature. The same study by Shokr and Agnew [2013] presents another anomaly of the microwave signature of snow on FYI caused by warm air temperatures during the spring. Unlike snow on MYI, the snow on FYI is saline at the base as mentioned earlier and presented in Drinkwater and Crocker [1988]. Warm atmospheric temperature will not only increase the snow wetness but will also cause brine drainage from the snow on FYI. Shokr and Agnew [2013] found that this scenario leads to a decrease of brightness temperature and an increase of backscatter from QuikSCAT. They also noted that these changes are not reversed after the return of the cold temperature. This is unlike the changes of the same parameters from the snow on MYI in response to warm temperature. Warm temperature causes irreversible changes of snow properties on FYI because it removes some brine contents. It would be appropriate to conclude this section with the following two notes. The first is about the interaction of the microwave signal with the existing snow and its ongoing physical processes. Findings from many studies of snow on sea ice continue to raise questions more than providing answers. For example, the re-freezing of liquid water in the snow decreases the attenuation and increases the scattering of microwave emission, leading to an increase in Tb [Tonboe, Andersen, Toudal, 2003]. Yet, re-freezing of liquid water may also lead to the formation of ice layering within the snowpack and ice crust near the surface of the snow base. This causes significant scattering and consequently a decrease of Tb especially at lower frequencies (≤ 37 GHz) [Comiso et al., 1997]. This effect is more pronounced for the horizontal polarization at high incidence angles due to the lower penetration depth and stronger snow layering effects [Hallikainen, 1989, Langlois and Barber, 2007]. Montpetit et al. [2013] studied the effect of ice lenses within a snowpack on microwave emission. These results would raise the question: what is really going on? Is there any conclusive information on effects of snow metamorphism on the microwave emitted radiation? The second point is about microwave observations during snow fall events. In a laboratory study of microwave emission from artificial thin ice grown in an outdoor tank, Shokr, Asmus, Agnew [2009] observed a sharp drop in Tb while fresh snow was falling on thin ice (< 5 cm thick). The drop was sharper from the lower frequency channels (19 GHz). It was followed by a gradual increase as the snow settled until the brightness temperature regained its typical value for thin ice. Unlike the observations by Grenfell and Comiso [1986], the sharp drop in brightness temperature found in Shokr, Asmus, Agnew [2009] was associated with an increase in the polarization difference. This

341

implies that the retrieval of any parameter of thin ice (< 5 cm thick) based on the polarization difference (e.g., ice concentration) will be in error if the data are obtained during an event of fresh snowfall. The large polarization difference during this event will make thin ice mistakably identified as open water. 7.10. REFERENCES Ackerman, S.A. et al. (1998) Discriminating clear sky from clouds with MODIS, Journal of Geophysical Research, 103 (D24), pp. 32,141–32,157. Alishouse, J. et al. (1990) Determination of total precipitable water from the SSM/I, IEEE Transactions on Geoscience and Remote Sensing, 28, pp. 811–816. Allison, I., Brandt, R.E. and Warren, S.G. (1993) East Antarctic sea ice: Albedo, thickness distribution, and snow cover, Journal of Geophysical Research, 98(C7), pp. 12,417–12,429. Anderson, G. et al. (2002) MODTRAN4-based atmospheric correction algorithm: FLAASH fast line-of-sight atmospheric analysis of spectral hypercubes, SPIE Proceedings, 4725. Austin, R.W. and Halikas, G. (1976) The index of refraction of seawater, University of California, San Diego Visibility Laboratory of Scripps Institute of Oceanography, Technical Report No. 76-1. Barber, D.G. and Nghiem, S.V. (1999) The role of snow on the thermal dependence of microwave backscatter of sea ice, Journal of Geophysical Research, 104(C11), pp. 25,789–25,803. Available from: doi:10.1029/1999JC900181. Beer, A. (1852) Bestimmung der Absorption des rothen Lichts in farbigen Flüssigkeiten (Determination of the absorption of red light in colored liquids), Annalen der Physik und Chemie, 86, pp. 78–88. Bélair, S. et al. (2009) Medium-range quantitative precipitation forecast from Canada’s new 33-Im deterministic global operational system, Weather and Forecasting, 24(3), pp. 690–708. Bernstein, L.S. et al. (2006) Improved reflectance retrieval from hyper- and multispectral imagery without prior scene or sensor information, SPIE Proceedings, Remote Sensing of Clouds and Atmosphere XI, 6362. Bobylev, L.P. et al. (2010) Atmospheric water vapor and cloud liquid water retrieval over the Arctic Ocean using satellite passive microwave sensing, IEEE Transactions on Geoscience and Remote Sensing, 48(1), pp. 283–294. Bohren, C.F. and Barkstrom B.R. (1974) Theory of the optical properties of snow, Journal of Geophysical Research, 79(30), pp. 4527–4535. Boukabara, S. A., (1997) Couplage des mesures hyperfrequences actives et passives, PhD thesis, University of Paris VII, Paris. Bouzerar, H. et al. (2020) Compact and full polarimetric SAR imaging for target characterization, Proceedings of Mediterranean and Middle-East Geoscience and Remote Sensing Symposium (M2GARSS), pp. 61–64. Available from: doi: 10.1109/ M2GARSS47143.2020.9105325. Brisco, B., Mahdianpari, M. and Mohammadimanesh, F. (2020) Hybrid Compact Polarimetric SAR for Environmental Monitoring with the RADARSAT Constellation Mission, Remote Sensing, 12, 3283. Available from: doi:10.3390/ rs12203283.

342

SEA ICE

Brodzik, M.J. et al. (2012) EASE-Grid 2.0: Incremental but significant improvements for Earth-gridded data sets, ISPRS International Journal of Geo-Information, 1(1), pp. 32–45. Available from: http://www.mdpi.com/2220-9964/1/1/32. Camps, A. et al. (2005) The emissivity of foam-covered water surface at L-band: Theoretical modeling and experimental results from the FROG 2003 field experiment, IEEE Transactions on Geoscience and Remote Sensing, 43(5), pp. 925–937. Chen, P.Y. et al. (2002) An automated cloud detection method for daily NOAA-14 AVHRR data for Texas, USA, International Journal of Remote Sensing, 23, pp. 2939–2950. Chen, S. et al. (2020) MYI floes identification based on the texture and shape feature from dual-polarized Sentinel-1 imagery, Remote Sensing, 12, 3221. Available from: doi:10.3390/rs12193221. Chuvieco, E., ed. (2020) Fundamentals of satellite remote sensing: an environmental approach—Third edition, London, New York: CRC Press, ISBN-10:1-13858-383-9. Clausi, D.A. and Yue, B. (2004) Comparing co-occurrence probabilities and Markov random fields for texture analysis of SAR sea ice imagery, IEEE Transactions on Geoscience and Remote Sensing, 42(1), pp. 215–228. Cloude, S.R. (1997) Wide band polarimetric radar inversion studies using the entropy-alpha decomposition, In: Mott, H. and Boerner, W.M., eds. Wideband interferometric sensing and image polarimetry, Proceedings of SPIE, Bellingham, Washington: International Society for Optical Engineering, Vol. 3120, pp. 118–129. Cloude, S.R. and Pottier, E. (1997) An entropy based classification scheme for land applications of polarimetric SAR, IEEE Transactions on Geoscience and Remote Sensing, 35(1), pp. 68–78. Available from: doi: 10.1109/36.551935. Colbeck, S.C. (1983) Theory of metamorphism of dry snow, Journal of Geophysical Research, 88, pp. 5475–5482. Comiso, J. (1985) Remote sensing of sea ice using multispectral microwave satellite data, In: Deepak, A., ed. Advances in Remote Sensing Retrieval Methods, Hampton, VA, USA, pp. 349–369. Comiso, J. (2010) Polar oceans from space, Atmospheric and Oceanographic Sciences Library, 41, New York: SpringerVerlag, 522 p. Comiso, J. et al. (1992) Microwave remote sensing of Southern Ocean cover, In: Carsey, F.D., ed. Microwave Remote Sensing of Sea Ice, Washington, D. C.: American Geophysical Union Monograph, Vol. 68, pp. 243–259. Comiso, J. et al. (1997) Passive microwave algorithms for sea ice concentration—A comparison of two techniques, Remote Sensing of Environment, 60, pp. 357–384. Culander, J.C. and McDonough, R.N. (1991) Synthetic aperture radar: system and signal processing, Wiley series in Remote Sensing and Image Processing, ISBN-13: 978-0-47185770-0, 672p. Dabbor, M. and Geldsetzer, T. (2014) On the classification of sea ice types using simulated Radarsat Constellation Mission (RCM) compact polarimetric SAR parameters, Proceedings of American Society of Photogrammetry and Remote Sensing (ASPRS) Conference, Louisville, Kentucky, March 23–28. De Abreu, R. (1996) In situ and satellite observations of the visible and infrared albedo of sea ice during spring melt, PhD thesis,

Department of Geography, University of Waterloo, Ontario, Canada. De Abreu, R. et al. (1994) Comparison of in situ and AVHRRderived broadband albedo over Arctic sea ice, Arctic, 47(3), pp. 288–297. Derksen, C. et al. (2012) Evaluation of passive microwave brightness temperature simulations and snow water equivalent retrievals through a winter season, Remote Sensing of Environment, 117, pp. 236–248. Dierking, W. and Busche, T. (2006) Sea ice monitoring by L-band SAR: An assessment based on literature and comparisons of JERS-1 and ERS-1 imagery, IEEE Transactions of Geoscience and Remote Sensing, 44(2), pp. 957–970. Ding, F. et al. (2020) Is radar phase information useful for sea ice detection in the marginal ice zone?, Remote Sensing, 12(11), pp. 1847. Available from: doi:10.3390/rs12111847. Doron, M. et al. (2011) Spectral variations in the near-infrared ocean reflectance, Remote Sensing of Environment. Available from: doi:10.1016/j.rse.2011.01.015. Drinkwater, M.R. and Crocker, G.B. (1988) Modelling changes in the dielectric and scattering properties of young snowcovered sea ice at GHz frequencies, Journal of Glaciology, 34(118), pp. 274–282. Drinkwater, M.R. et al. (1991) Multifrequency polarimetric synthetic aperture radar observations of sea ice, Journal of Geophysical Research, 96(C11), pp. 20,679–20,698. Du, J., Shi, J. and Rott, H. (2010) Comparison between a multiscattering and multi-layer snow scattering model and its characterized snow backscattering model, Remote Sensing of Environment, 114(5), pp. 1089–1098. Durand, M., Kim, E.J. and Margulis, S.A. (2009) Radiance assimilation shows promise for snowpack characterization, Geophysical Research Letters, 36(L02503). Available from: doi:10.1029/2008GL035214. Dybbroe, A., Karlsson, K.G. and Thoss, A. (2005) NWCSAF AVHRR cloud detection and analysis using dynamic thresholds and radiative transfer modeling. Part I: Algorithm description, Journal of Applied Meteorology and Climatology, 44, pp. 39–54. Available from: doi:10.1175/JAM-2188.1. Elachi, C. (1988) Spaceborne radar remote sensing: applications and techniques, New York, USA: IEEE Press. Elachi, C. and van Zyl, J.J. (2006) Introduction to the physics and techniques of remote sensing, 2nd ed., New Jersey: John Wiley & Sons Inc., 552 p. Ellison, W.J. et al. (2003) A comparison of ocean emissivity model using the Advanced Microwave Sounding Unit, the Special Sensor Microwave Imager, the TRMM Microwave Imager, and airborne radiometer observations, Journal of Geophysical Research, 108(D21), p. 4663. Available from: doi:10.1029/2002JD003213. Evans, D.L.T. et al. (1988) Radar polarimetry: Analysis tools and applications, IEEE Transactions on Geoscience and Remote Sensing, 26(6), pp.774–789. Fraser, R.S. et al. (1989) Algorithm for atmospheric corrections of aircraft and satellite imagery, NASA Technical Memorandum No. 100751, Goddard Space Flight Centre, Greenbelt, MD, USA. Freeman, A. and Raney, R.K. (2008) Improving range ambiguity performance in Quad-Pol SAR, Proceedings of the

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE International Geoscience and Remote Sensing Symposium (IGARSS’08), Boston, USA, pp. 349–356. Freeman, T. and Durden, S.L. (1998) A three-component scattering model for polarimetric SAR data, IEEE Transactions on Geoscience and Remote Sensing, 36(3), pp. 963–973. Available from: doi: 10.1109/36.673687. Fuhrhop, R. and Simmer, C. (1996) SSM/I brightness temperature corrections for incidence angle variations, Journal Atmospheric and Oceanic Technology, 13(11), pp. 246–254. Fuhrhop, R. and Simmer, C. (1998) MWMOD User Manual, Version 1.12, Kiel, Germany: Institut für Meereskunde. Fung, A.K. (1966) On depolarization of electromagnetic waves backscattered from a rough surface, Planetary and Space Science, 14(7), pp. 563–568. Fung, A.K. (1967) Character of wave depolarization by a perfectly conducting rough surface and its application to Earth and Moon experiments, Planetary and Space Science, 15(9), pp. 1337–1347. Fung, A.K. and Chen, K.S. (2009) Microwave scattering and emission models for users, Norwood, MA, USA: Artech House, 430 p. Gasiewski, A.J. (1993) Microwave radiative transfer in hydrometeors, In: Janssen, M.A., ed. Atmospheric remote sensing by microwave radiometry, John Wiley & Sons, ISBN-13: 978-0-47162-891-0, 572 p. Gill, J.P.S. and Yackel, J.J. (2012) Evaluation of C-band SAR polarimetric parameters for discrimination of first-year sea ice types, Canadian Journal of Remote Sensing, 38(3), pp. 306–323. Gloersen, P. et al. (1973) Microwave signatures of first-year and multi-year sea ice, Journal of Geophysical Research, 78, pp. 3564–3572. Gloersen, P. et al. (1978) Time-dependence of sea ice concentration and multiyear ice fraction in the Arctic Basin, Boundary Layer Meteorology, 13, pp. 339–359. Gómez-Chova, L. et al. (2007) Cloud screening and multitemporal unmixing of MERIS FR data, European Space Agency (Special Publication), ESA SP, (SP-636). Goodberlet, M.A., Swift, T.C. and Wilkersen, J.C. (1989) Remote sensing of ocean surface winds with the Special Sensor Microwave Imager, Journal of Geophysical Research, 94, pp. 14,547–14,555. Green, P.E. (1968) Radar measurements of target scattering properties, In: Evan, J.V. and Hagfors, T., ed. New York: McGraw-Hill. Grenfell, T.C. and Comiso, J.C. (1986) Multifrequency passive microwave observations of first-year sea ice grown in a tank, IEEE Transactions of Geoscience and Remote Sensing, GE-24 (6), pp.826–831. Grenfell, T.C. and Maykut, G.A. (1977) The optical properties of ice and snow in the Arctic Basin, Journal of Glaciology, 18 (80), pp. 445–463. Gunn, G.E. et al. (2011) Evaluation of the HUT modified snow emission model over lake ice using airborne passive microwave measurements, Remote Sensing of Environment, 115 (1), pp. 233–244. Gumley, L.E., Hubanks, P.A. and Masuoka, E.J. (1994) MODIS Technical Report Series: MODIS Airborne

343

Simulator Level 1B Data User’s Guide, NASA Technical Memorandum 104594, Vol. 3, 37. Hallikainen, M.T. (1989) Microwave radiometry of snow, Advances in Space Research, 9, pp. 267–275. Hellman, M. (1999) Classification of full polarimetric SAR data for cartographic applications, PhD thesis, Dresden University of Technology, Dresden, Germany. Heygster, G. et al. (2012) Remote sensing of sea ice: Advances during the DAMOCLES project, The Cryosphere, 6, pp. 1411–1434. Available from: doi:10.5194/tc-6-1411-2012. Hollinger, J.P., Lo, R. and Poe, G. (1987) Special Sensor Microwave/Imager User’s Guide, Washington, D.C.: Naval Research Laboratory. Isleifson, D. et al. (2010) C-band polarimetric backscattering signatures of newly formed sea ice during fall freeze-up, IEEE Transactions on Geoscience and Remote Sensing, 48(8), pp. 3256–3267. ITT Visual Information Solutions (2009) Atmospheric correction module: QUAC and FLAASH User’s Guide, ITT Visual Information Solutions Report 20AC47/DOC. Ivanov, A. (1979) Absorption of solar energy in the ocean, In: Krauss, E.V., ed. Modelling and prediction of upper ocean layers (in Russian), Leningrad: Gidrometeoizdat, pp. 64–77. Jang. J.D. et al. (2006) Neural network application for cloud detection in SPOT vegetation images, International Journal of Remote Sensing, 27, pp. 719–736. Jansing, E.D. (2021) Introduction to synthetic aperture radar: concepts and practice, McGraw-Hill Education, ISBN-13: 978-1-26045-896-1, 208 p. Jin, X. et al. (2019) Assessment and improvement of sea surface microwave emission models for salinity retrieval in the East China Sea, Remote Sensing, 11, 2486. Available from: doi:10.3390/rs11212486. Johannessen, O. et al. (2007) Remote sensing of sea ice in the Northern Sea route: Studies and applications, Springer-Praxis Books in Geophysical Sciences, ISBN 10: 3-540-24448-4. Joseph, G. (2017) Fundamentals of remote sensing, Universities Press, ISBN-10: 9-386-23546-6, 624 p. Kachi, M. et al. (2008) Status of GCOM-W1/AMSR2 development and science activities, In: Meynart et al., ed. Sensors, systems, and next generation satellites XII, SPIE Proceedings, Vol. 7106. Available from: doi:10.1117/12.801228. Karpouzli, E. and Malthus, T. (2003) The empirical line method for the atmospheric correction of IKONOS imagery, International Journal of Remote Sensing, 24(5), pp. 1143–1150. Karstens, U., Simmer, C. and Purpecht, E. (1994) Remote sensing of cloud liquid water, Meteorology and Atmosphere Physics, 54, pp. 157–171. Kashiwase, H. et al. (2017) Evidence for ice-ocean albedo feedback in the Arctic Ocean shifting to a seasonal ice zone, Scientific Reports, 7, p. 8170. Available from: https://doi.org/ 10.1038/s41598-017-08467-z. Kawanishi, T. et al. (2003) The Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E), NASDA’s contribution to the EOS for global energy and water cycle studies, IEEE Transactions of Geoscience Remote Sensing, 41(2), pp.184–194.

344

SEA ICE

Kay, S., Hedley, J.D. and Lavender, S. (2009) Sun glint correction of high and low spatial resolution images of aquatic scenes: A review of methods for visible and near-infrared wavelengths, Remote Sensing (Open Access Journal), 1, pp. 697–730. Available from: doi:10.3390/rs1040697. Kern, S. (2001) A new algorithm to retrieve the sea ice concentration using weather-corrected 85 GHz SSM/I measurements, PhD dissertation, University of Bremen, Germany. Kern, S.A. (2004) A new method for medium-resolution sea ice analysis using weather-influence corrected Spatial Sensor Microwave/Imager 85 GHz data, International Journal of Remote Sensing, 25(21), pp. 4555–4582. Kennaugh, K. (1951) Effect of type of polarization on echo characteristics, Technical Report 389-4, The Ohio State University, Antenna Laboratory, Columbus, OH, 35 p. Key, J.R. and Haefliger, M. (1992) Arctic ice surface temperature retrieval from AVHRR thermal channels, Journal of Geophysical Research, 97(D5), pp. 5885–5893. Key, J.R. et al. (1997) High-latitude surface temperature estimates from thermal satellite data, Remote Sensing of Environment, 61, pp. 302–309. Kim, J.M. et al. (2020) Differences between ICESat and CryoSat-2 sea ice thicknesses over the Arctic: Consequences for analyzing the ice volume trend, Journal of Geophysical Research: Atmosphere, 125, e2020JD033103. Kneizys, F.X. et al. (1989) Users Guide to LOWTRAN, AFGLTR-0177, US Air Force Geophysical Laboratory, Bedford, MA. Kokhanovsky, A.A. (2004) Spectral reflectance of whitecaps, Journal of Geophysical Research, 109(C05021). Available from: doi:10.1029/2003JC002177. Konda, M. et al. (1994) Measurement of the sea surface emissivity, Journal of Oceanography, 50, pp. 17–30. Kondratyev, K.Y., Kozoderov, V.V. and Smokty, O.I. (2012) Remote sensing of the earth from space: Atmospheric correction, Springer, ISBN-13: 978-3-64276-749-4, 489 p. Kontu, A. and Pulliainen, J. (2010) Simulation of spaceborne microwave radiometer measurements of snow cover using in situ data and brightness temperature modeling, IEEE Transactions on Geoscience and Remote Sensing, 48(3), pp. 1031–1044. Koskinen, J., Pulliainen, J. and Hallikainen, M. (2000) Effect of snow wetness to C-band backscatter—A modeling approach, Proceedings of International Geoscience and Remote Sensing Symposium, Honolulu, 4, pp. 1754–1756. Kwok, R. and Markus, T. (2017) Potential basin-scale of Arctic snow depth with sea ice freeboards from CryoSat-2 and ICESat-2: An exploratory analysis, Advances in Space Research, 62(6), pp. 1243–1250. Kwok, R. et al. (2007) Ice, Cloud and land Elevation (ICESat) over Arctic sea ice: Retrieval of freeboard, Journal of Geophysical Research, 112(C12013). Available from: doi: 10.1029/2006JC003978. Langlois, A. and Barber, D.G. (2007) Passive microwave remote sensing of seasonal snow-covered sea ice, Progress in Physical Geography, 31(6), pp. 539–573. Langlois, A. et al. (2008) The response of snow thermophysical processes to the passage of a polar low pressure system and its

impact on in-situ passive microwave data: A case study, Journal of Geophysical Research, 113(C03S04). Available from: doi:10.1029/2007JC004197. Laxon, S.W. et al. (2013) CryoSat-2 estimates of Arctic sea ice thickness and volume, Geophysical Research Letters, 40, pp. 732–737. Le Hégarat-Mascle, S. and André, C. (2009) Use of Markov random fields for automatic cloud/shadow detection on high resolution optical images, ISPRS Journal of Photogrammetry and Remote Sensing, 64(4), pp. 351–366. LeBlond, P.H. and Mysak, L.A. (1981) Waves in the ocean, Elsevier Oceanography Series, Amsterdam, The Netherlands: Elsevier Scientific Publishing Company, ISBN-10: 0-44441602-1. LeDrew, E.F. and Barber, D.G. (1994) The SIMMS program: A study of change and variability within the marine cryosphere, Arctic, 47(3), pp. 256–264. Lee, J.S. and Pottier, E. (2009) Polarimetric radar imaging: From basics to applications, CRC Press, ISBN-13: 978-1-42005-4972, 422 p. Li, S. and Zhou, X. (2002) Derivation and comparison of bidirectional reflectance functions of first-year and multi-ear sea ice types in the Southern Oceans, Proceedings of IEEE International Geoscience and Remote Sensing Symposium IGRASS’02, Vol. 3, pp. 1944–1946. Li, Z., Tan, Y. and Tsang, L. (2006) Modelling the passive microwave remote sensing of snow using dense media radiative transfer theory with NMM3D rough surface boundary conditions, Microwave and Optical Technology Letters, 48 (3), pp. 557–562. Lindsay, R.W. and Rothrock, D.A. (1993) The calculation of surface temperature and albedo of Arctic sea ice from AVHRR, Annals of Glaciology, 17, pp.174–183. Lindsay, R.W. and Rothrock, D.A. (1994) Arctic sea ice albedo from AVHRR, Journal of Climate, 7, pp. 1737–1749. Livingstone, C.E. et al. (1996) The CCRS airborne SAR systems: Radar for remote sensing research, Canadian Journal of Remote Sensing, 21(4), pp. 468–491. Lohanick, A.L. (1993) Microwave brightness temperatures of laboratory-grown undeformed first-year ice with an evolving snow cover, Journal of Geophysical Research, 98, pp. 4667–4674. Lohanick, A.W. and Grenfell, T.C. (1986) Variations in brightness temperature over cold first-year ice near Tuktoyaktuk, Northwest Territories, Journal of Geophysical Research, 91 (C4), pp. 5133–5144. Luo, Y. et al. (2005) Surface bidirectional reflectance and albedo properties derived using a land cover-based approach with Moderate Resolution Imaging Spectroradiometer observations, Journal of Geophysical Research, 110(D01106). Available from: doi:10.1029/2004JD004741. Mäkynen, M. and Hallikainen, M. (2004) Investigation of CBand and X-Band backscattering signatures of the Baltic Sea ice, International Journal of Remote Sensing, 25(11), pp. 2061–2086. Markus, T. and Cavalieri, D.J. (1998) Snow depth distribution over sea ice in the Southern Ocean from satellite passive microwave data, In: Jeffries, M.O., ed., Antarctic sea ice:

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE physical processes, interactions, and variability. Antarctic Research Series, Vol. 74, American Geophysical Union, pp. 19–39. Markus, T., Stroeve, J.C. and Miller, J. (2009) Recent changes in Arctic sea ice melt onset, freeze-up, and melt season length, Journal of Geophysical Research, 114(C12024). Available from: doi:10.1029/2009JC005436. Marzano, F.S. and Visconti, G., eds. (2003) Remote sensing of atmosphere and ocean from space: Models, instruments and techniques, New York, Boston, London, Moscow: Kluwer Academic Publishers. Massom, R. and Comiso, J.C. (1994) The classification of Arctic Sea ice types and the determination of surface temperature using advanced very high resolution radiometer data, Journal of Geophysical Research, 99(C3), pp. 5201–5218. Mathew, N., Heygster, G. and Melsheimer, C. (2009) Surface emissivity of the Arctic sea ice at AMSR-E frequencies, IEEE Transactions on Geoscience and Remote Sensing, 47(12), pp. 4115–4124. Maykut, G.A. and Untersteiner, N. (1971) Some results from a time-dependent thermodynamic model of sea ice, Journal of Geophysical Research, 76, pp. 1550–1575. McMillin, L.M. (1975) Estimation of sea surface temperature from two infrared window measurements with different absorption, Journal of Geophysical Research, 80, pp. 5113–5117. Miao, J. (1998) Retrieval of atmospheric water vapor content in polar regions using spaceborne radiometry, PhD thesis, Reports on Polar Research, 289, Alfred-Wegner Institute for Polar and Marine Research (AWI). MODIS (1999) MODIS BRDF/Albedo product: algorithm theoretical basis document—Version 5, MODIS Product ID: MOD43, Available from: http://modis.gsfc.nasa.gov/data/ atbd/atbd_mod09.pdf. Moen, M.A.N. et al. (2015) Assessing polarimetric SAR sea-ice classifications using consecutive day images, Annals of Glaciology, 56(69). Available from: doi: 10.3189/2015AoG69A80. Montpetit, B. et al. (2013) Snow microwave emission of ice lenses within a snowpack using the microwave emission model for layered snowpacks, IEEE Transactions on Geoscience and Remote Sensing, 51(9), pp. 4705–4717. Nicodemus, F.E. et al. (1977) Geometrical considerations and nomenclature for reflectance, US Department of Commerce, National Bureau of Standards (US) Monograph, Washington, D.C., 67 p. Norjamäki, I. and Tokola, T. (2007) Comparison of atmospheric correction methods in mapping timber volume with multitemporal Landsat images in Kainuu, Finland, Photogrammetric Engineering and Remote Sensing, 73(2), pp. 155–163. O’Dell, C.W., Wentz, F.J. and Nennartz, R. (2008) Cloud liquid water path from satellite-based passive microwave observations: A new climatology over the global oceans, Journal of Climate, 21, pp. 1721–1739. Parkinson, C.L. et al. (1987) Arctic sea ice: 1973–1976, NASA Special Report SP-489, Washington D.C., 295 p. Pedersen, C.A. (2007) Optical properties of snow and sea ice field measurements, parameterization schemes and validation, PhD Dissertation, University of Tromso, Norway.

345

Perovich, D. (1991) Seasonal changes in sea ice optical properties during fall freeze-up, Cold Region Science and Technology, 19, pp. 261–271. Perovich, D.K. (2001) UV radiation and the optical properties of sea ice and snow, In: Hessen, D., ed. UV radiation and Arctic ecosystems, Berlin: Springer Verlag, pp. 73–89. Petty, G.W. (1993) SSMI/algorithms for the estimation of column water vapor, Proceedings of the Shared Processing Network DMSP Special Microwave Imager Symposium, The Fleet Numerical Meteorology and Oceanography Center, Monterey, California, 1993. Prabhakara, C.P., Dalu, G. and Kunde, V.G. (1974) Estimation of sea surface temperature from remote sensing in the 11–13m window region, Journal of Geophysical Research, 79(33), pp. 5039–5044. Pullianen, J. and Hallikainen, M. (2001) Retrieval of regional snow water equivalent from space-borne passive microwave observations, Remote Sensing of Environment, 75, pp. 76–85. Pulliainen, J.T., Grandell, J. and Hallikainen, M.T. (1999) HUT snow emission model and its applicability to snow water equivalent retrieval, IEEE Transactions on Geoscience and Remote Sensing, 37(3), pp. 1378–1390. Qiao, M. and Miao, J. (2003) Retrieving integrated water vapor using AMSU-B channels over Arctic region, International Journal of Energy and Environment, 2(3), pp. 149–156. Rahman, H. and Dedieu, G. (1994) SMAC: A simplified method for atmospheric correction of satellite measurements in solar spectrum, International Journal of Remote Sensing, 15, pp. 123–143. Raney, R.K. (2007) Hybrid-polarity SAR architecture, IEEE Transactions on Geoscience and Remote Sensing, 45(11), pp. 3397–3404. Reese, W.G. (2013) Physical principles of remote sensing, 3rd ed., New York: Cambridge University Press, 442 p. Ricker, R. (2015) Sea ice thickness derived from Cryosat-2: Validation and uncertainties, PhD thesis, Jacobs University Bremen, Germany. Riggs, G.A., Hall, D.K. and Ackerman, S.A. (1999) Sea ice extent and classification mapping with the moderate resolution imaging spectroradiometer airborne simulator, Remote Sensing of Environment, 68(2), pp. 152–168. Robinson, I.S. (2004) Measuring the oceans from space: the principle and methods of satellite oceanography, New York: Springer-Verlag, pp. 504–561. Rösel, A. (2013) Detection of melt ponds on Arctic sea ice with optical satellite data, New York, Dordrecht, London: Springer Heidelberg. Ross, D.B. and Cardone, V. (1974) Observations of oceanic whitecaps and their relation to remote measurements of surface wind speed, Journal of Geophysical Research: Ocean, 79(C3), pp. 444–452. Sandven, S. and Johannessen, O.M. (2006) Sea ice monitoring by remote sensing, In: Gower, J., ed. Remote Sensing of the Marine Environment, Manual of Remote Sensing, 3rd ed., Vol. 6, Chapter 8, Maryland, USA: American Society for Photogrammetry and Remote Sensing, pp. 241–283. Schaepman-Strub, G. et al. (2006) Reflectance quantities in optical remote sensing–Definitions and case studies, Remote Sensing of Environment, 103, pp. 27–42.

346

SEA ICE

Schowengerdt, R.A. (2006) Remote sensing: Models and methods for image processing, 3rd ed., Academic Press, ISBN-13: 9780-12369-407-2. Schweiger, A.J. et al. (2008) Relationships between Arctic sea ice and clouds during autumn, Journal of Climate, 21, pp. 4799–4810. Shokr, M. (1991) Evaluation of second-order texture parameters for sea ice classification from radar images, Journal of Geophysical Research, 96(C6), pp. 10,625–10,640. Shokr, M.E. and Agnew, T.A. (2013) Validation and potential applications of Environment Canada Ice Concentration Extractor (ECICE) algorithm to Arctic ice by combining AMSR-E and QuikSCAT observations, Remote Sensing of Environment, 128, pp. 315–332. Shokr, M., Asmus, K. and Agnew, T.A. (2009) Microwave emission observations from artificial thin sea ice: The ice-tank experiment, IEEE Transactions on Geoscience and Remote Sensing, 47(1), pp. 325–338. Shokr, M.E. and Markus, T. (2006) Comparison of NASA Team2 and AES-York ice concentration algorithms against operational ice charts from the Canadian Ice Service, IEEE Transactions on Geoscience and Remote Sensing, 44(8), pp. 2164–2175. Shokr, M.E., Ramsay, B. and Falkingham, J.C. (1996) Operational use of ERS-1 SAR images in the Canadian ice monitoring program, International Journal of Remote Sensing, 17(4), pp. 667–682. Shokr, M. and Dabboor, M. (2020) Observations of SAR polarimetric parameters of lake and fast sea ice during the early growth phase, Remote Sensing of Environment, 247, p. 111910. Shokr, M. et al. (2022) Observations from C-band SAR fullypolarimetric parameters of mobile sea ice based on radar scattering mechanisms to support operational sea ice monitoring, Canadian Journal of Remote Sensing, 48(2), pp. 197–213. Smith, D.M., Barrett, E.C. and Scott, J.C. (1995) Sea-ice type classification from ERS-1 SAR data based on grey level and texture information, Polar Record, 31(177), pp. 135–146. Soh, L. and Tsatsoulis, C. (1999) Texture analysis of SAR sea ice imagery using gray level co-occurrence matrices, IEEE Transactions on Geoscience and Remote Sensing, 37(2), pp. 780–795. Soh, L. et al. (2004) ARKTOS: An intelligent system for SAR sea ice image classification, IEEE Transactions on Geoscience and Remote Sensing, 42(1), pp. 229–248. Available from: doi: 10.1109/TGRS.2003.817819. Sohn, B. et al. (2003) Estimating upper-tropospheric water vapour from SSM/T-2 satellite measurements, Journal of Applied Meteorology, 42, pp. 488–504. Stephens, G.L. (1994) Remote sensing of the lower atmosphere: an introduction, Oxford University Press, ISBN-13: 978-019508-188-6, 544 p. Stogryn, A. (1972) The emissivity of sea foam at microwave frequencies, Journal of Geophysical Research, 77, pp. 1658–1666. Strahler, A. et al. (2003) MODIS BRDF/Albedo Product (MOD43B) User’s Guide, NASA, Washington, D.C. Available from: http://geography.bu.edu/brdf/userguide/index.html. Stringer, W.J., Barnett, D.G. and Godin, R.H. (1984) Handbook for sea ice analysis, Naval Environmental Prediction Research Facility, Rep. No. CR 84-03, Monterey, California.

Sturm, M. et al. (2006) Snow depth and ice thickness measurements from the Beaufort and Chukchi Seas collected during the AMSR-Ice03 campaign, IEEE Transactions on Geoscience and Remote Sensing, 44(11), pp. 3009–3020. Svendsen, E. et al. (1983) Norwegian remote sensing experiment: Evaluation of the Nimbus 7 scanning multichannel microwave radiometer for sea ice research, Journal of Geophysical Research, 88(C5), pp. 2781–2791. Swift, C.T. (1980) Passive microwave remote sensing of the ocean—A review, Boundary Layer Meteorology, 18, pp. 25–40. Tanré, D., Holben, B.N. and Kaufmann, Y.J. (1992) Atmospheric correction algorithms for NOAA-AVHRR products: Theory and application, IEEE Transactions on Geoscience and Remote Sensing, 30(2), pp. 231–248. Tonboe, R., Andersen, A. and Toudal, L. (2003) Anomalous winter sea ice backscatter and brightness temperature, Danish Meteorological Institute, Scientific Report 03-13, Copenhagen, ISBN-10: 8-77478-489-7. Touzi, R. (1992) Extraction of point target response characteristics from complex SAR data, IEEE Transactions on Geoscience and Remote Sensing, 30, pp.1158–1161. Touzi, R. (2007) Target scattering decomposition in terms of roll-invariant target parameters, IEEE Transactions on Geoscience and Remote Sensing, 45(1), pp. 7–84. Available from: doi: 10.1109/TGRS.2006.886176. Touzi, R., (2009) Compact-hybrid versus linear dual and fully polarimetric SAR, ESA Publication SP-668, Proceedings of 4th International Workshop on Science and Applications of SAR Polarimetry and Polarimetric Interferometry (POLInSAR workshop), Frascati, Italy. Touzi, R. et al. (2004) A review of polarimetry in the context of synthetic aperture radar: Concepts and information. Ulaby, F. and Elachi, C. (1990) Radar polarimetry for geosciences applications, Norwood, MA, USA: Artech House, ISBN-13: 978-0-89006-406-1, 364 p. Ulaby, F.T. and Long, D.G. (2014) Microwave radar and radiometric remote sensing, University of Michigan Press, ISBN13: 978-0-47211-935-6. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1981) Microwave remote sensing: active and passive, In: Vol. I: Microwave remote sensing fundamentals and radiometry, Norwood, MA: Artech House Inc., p. 225. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1982) Microwave remote sensing: active and passive, In: Vol. II: Radar remote sensing and surface scattering and emission theory, Norwood, MA: Artech House Inc., p.847. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1986) Microwave remote sensing: active and passive, Vol. III: From theory to applications, Norwood, MA: Artech House Inc., p. 2058. Urban, J. (2013) Satellite sensors measuring atmospheric water vapor, In: Kämpfer, N., ed. Monitoring atmospheric water vapor: Ground-based remote sensing and in-situ methods, ISSI Scientific Report 10, Chapter 9, New York, London: Springer. van der Sanden, J.J. (2004) Anticipated potential of Radarsat-2 data, Canadian Journal of Remote Sensing, 30(3), pp. 369–379. van Zyl, J.J. and Kim, Y. (2011) Synthetic Aperture Radar polarimetry, Hoboken, New Jersey: John Wiley & Sons Inc., 312 p, ISBN-13: 978-1-11811-511-4.

REMOTE SENSING FUNDAMENTALS RELEVANT TO SEA ICE Vermote, E.F. et al. (1997) Second simulation of the satellite signal in the solar spectrum, 6S, IEEE Transactions on Geoscience and Remote Sensing, 35(3), pp. 675–688. Wang, M. (2010) Atmospheric correction for remotely-sensed ocean colour products, International Ocean Color Coordinating Group (IOCCG) Report Number 10, 2010, 78 p. Wanner, W. et al. (1997) Global retrieval of bidirectional reflectance and albedo over land from EOS MODIS and MISR data: Theory and algorithm, Journal of Geophysical Research, 102(D14), pp. 17,143–17,161. Warren, S.G. (1982) Optical properties of snow, Reviews of Geophysics and Space Physics, 20(1), pp.67–89. Warren, S.G. (2012) Optical properties of snow, Reviews of Geophysics, 20(1), pp. 67–89. Winther, J. et al. (2004) Surface reflectance of sea ice and underice irradiance in Kongsfjorden, Svalbard, Polar Research, 23 (1), pp. 115–118. Wiscombe, W.J. and Warren, S.G. (1980) A model for spectral albedo of snow, I: Pure snow, Journal of the Atmospheric Sciences, 37(12), pp. 2712–2733. Wozniak, B. and Dera, J. (2007) Light absorption in sea water, Atmospheric and Oceanographic Library, 33, 452 p. Yackel, J.J. et al. (2007) First-year sea ice spring melt transitions in the Canadian Arctic Archipelago from time-series synthetic aperture radar data, 1992–2002, Hydrological Processes, 21, pp. 253–265. Yamaguchi, Y. et al. (2005) Four-component scattering model for polarimetric SAR image decomposition, IEEE Transactions on Geoscience and Remote Sensing, 43(8), pp. 1699–1706. Ye, Y. et al. (2016) Improving multiyear ice concentration estimates with sea ice drift, Remote Sensing, 8, p. 397. Available from: doi:10.3390/rs8050397.

347

Yan, B., Weng, F. and Meng, H. (2008) Retrieval of snow surface microwave emissivity from the advanced microwave sounding unit, Journal of Geophysical Research, 113 (D19206). Available from: doi:10.1029/2007JD009559. Yu, Y. and Lindsay, R.W. (2003) Comparison of thin ice thickness distributions derived from RADARSAT Geophysical Processor System and advanced very high resolution radiometer data sets, Journal of Geophysical Research, 108(C12), p. 3378. Available from: doi: 10.1029/ 2002JC001319. Yu, Y., Rothrock, D.A. and Lindsay, R.W. (1995) Accuracy of sea ice temperature derived from the advanced very high resolution radiometer, Journal of Geophysical Research, 100(C3), pp. 4525–4532. Zebker, F.A. and van Zyl, J.J. (1991) Imaging radar polarimetry: A review, Proceedings of the IEEE, 79(11), pp. 1583–1606. Zhang, S. et al. (2021) Arctic sea ice freeboard retrieval from Envisat altimetry data, Remote Sensing, 13(8), p. 1414. Available from: http://doi.org/10.3390/rs13081414. Zhou, X. (2002) Optical remote sensing of snow on sea ice: Ground measurements, satellite data analysis, and radiative transfer modeling, PhD thesis, University of Alaska Fairbanks, 219 p. Zhou, X., Li, S. and Morris, K. (2001) Spatial variation of, and correlations among, snow surface albedo and physical parameters of summer snow cover on sea ice in the Ross Sea, Antarctica, AGU 2001 Fall Meeting, Eos, suppl. 82(47), p. F550. Zwally, H.J. et al. (1983) Antarctic Sea Ice, 1973-1976: Satellite Passive-Microwave Observations, Washington, D.C.: NASA SP-459, National Aeronautics and Space Administration, 206 pp.

8 Satellite Sensors for Sea Ice Monitoring

8.1 Historical Synopsis of Remote Sensing Satellites for Sea Ice.. 349

8.5 Scatterometer Sensors ............................................................. 358

8.2 Optical and Thermal Infrared Sensors.................................... 352 8.3 Modern Passive Microwave Sensors....................................... 353

8.6 Altimeter Sensors.................................................................... 359 8.7 References............................................................................... 360

8.4 Modern Imaging Radar Sensors............................................. 355

This chapter presents a brief account on satellite remote sensors used to study and monitor sea ice, particularly in the polar regions. The material covers sensors that provide information at three spatial scales: synoptic scale covering the entire polar region, regional scale (hundreds or thousands of kilometers coverage), and tactical scale (at fine resolution of tens or hundreds of meters), which is mainly offered by synthetic aperture radar (SAR). The chapter covers all categories of optical, thermal infrared, passive microwave (PM), SAR, scatterometer and altimeter systems. It presents technical details about the sensors and, in some cases, examples of the products from the data. The data are used to retrieve key sea ice parameters. Retrieval methods are presented in Chapters 10 and 11 for surface information and ice geophysical parameters, respectively.

8.1. HISTORICAL SYNOPSIS OF REMOTE SENSING SATELLITES FOR SEA ICE While sea ice observations from coastal stations and ships have a history of more than 100 years, observations from satellite sensors are relatively new. A short historical synopsis of satellite remote sensing of sea ice is presented in this section; while modern sensors, methods and products (developed after mid-1990s) are presented later. The earliest satellite programs that collected data on sea ice included NASA’s Television and Infrared Observations

Satellite (TIROS), Nimbus (a meteorological research satellite), and the Earth Resources Technology Satellite (ERTS) programs. ERTS was renamed later to Landsat. The first satellite image of sea ice was captured by a television camera onboard TIROS-1 on its second day of operation, 2 April 1960. It was an image of the Gulf of St. Lawrence in Canada (Figure 8.1). It showed areas of dark tone, which could be interpreted as cloud cover. Upon comparison against aerial photographs from the Canadian Meteorological Service at that time, it was confirmed that the observed gray tone represented sea ice. That, indeed, was the first indication that satellite ice reconnaissance would be a valuable tool to add to traditional ice observation wards at that time; namely ships, aircrafts, and meteorological stations. Interpretation of this image can be found in Wark and Popham [1960]. The Nimbus program was a series of seven satellites launched over 30-year period from 1964 to 1994. These were Sun-synchronous near-polar orbit satellites aimed at collecting meteorological data. The name Nimbus is the Latin word for “rain cloud.” The program effectively marked the beginning of the Earth observations era. Each satellite within this series carried different payload of sensors. The first four satellites carried visible and infrared sensors. They collected orbital data on the ice extent in the polar regions during the period from the mid-1960s to 1980. However, by late 1960s, it became clear that frequent synoptic observations of polar ice required microwave sensors because of the limitations of the VIS and

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 349

350

SEA ICE

ce

t

fS

en wr La

o ulf

G

c

e eb

Qu

Figure 8.1 The first satellite image of sea ice captured by the camera onboard TIROS-1 satellite on 2 April 1960, showing sea ice cover in the Gulf of St. Lawrence (the gray area in the middle) (photo from the archive of US National Oceanic and Atmospheric Administration).

IR sensors during the long dark winter months and the frequent cloudy conditions in the remaining months of the year. Nimbus 5 (1972–1983) carried a passive microwave (PM) sensor called Electrically Scanning Microwave Radiometer (ESMR) and Nimbus 7 (1978–1994) carried another PM sensor, the Scanning Multichannel Microwave Radiometer (SMMR). This sensor was operational until 1987 and provided the first consistent long-term climatology of polar sea ice. Nevertheless, the first PM satellite sensors were launched on the Russian Kosmos-243 and Losmos-384 on 23 September 1968 and 10 December 1970, respectively. ESMR was a cross-track scanning instrument (see the configuration in Figure 7.5), which had one horizontally polarized radiometer operating at a frequency of 19.35 GHz (1.55 cm wavelength). This channel discriminates well between sea ice and open water; but a single observation has to be integrated over large footprints. The spatial resolution of ESMR was 25 km × 25 km near nadir, degrading to 160 km × 45 km at the end of the scan. This sensor allowed the first estimates of ice concentration during lifetime of the satellite from 1973 to 1976. Daily and monthly averaged sea ice concentration maps were calculated later from re-processed data for the Arctic and the Antarctic from 12 December 1972 through 31 December 1976. The maps are available in 25 km gridded resolution from the National Snow and Ice Center (NSIDC) via the link http://nsidc.org/data/ nsidc-0009.html. The data were re-processed to reduce weather and coastal contamination effects and presented in the Hierarchical Data Format (HDF). Zwally et al. [1983] produced the first atlas of sea ice in the Antarctica from ESMR data. It was interesting to see for the first time the sea ice field that encases the continent. One of the most unforeseen discoveries from this data set was a large “hole” of open water in the middle of the sea ice

cover during the winter of 1974. That was the first observation of the phenomenon that became later known as polynya [section 2.9.1]. ESMR was the precursor to the more widely used microwave sensor, the SMMR, which was launched on Nimbus 7 and operated for nine years from 1978 to 1987. The sensor recorded dual-polarized microwave emission at 6.63, 10.69, 18.0, 21.0, and 37.0 GHz. This allowed mapping of sea ice concentration and distinction between FYI and MYI types. The first algorithms to retrieve sea ice concentration used SMMR data. Among them were NORSEX [Svendsen et al., 1983], NASA team [Cavalieri, Gloersen, Campbell, 1984], and Bootstrap [Comiso, 1986]. Due to power limitations, SMMR collected data every other day. The radiometers on SMMR and all of the subsequent PM systems have been conical (see the configuration in Figure 7.5). Following SMMR, the Special Sensor Microwave Imager (SSM/I) was launched aboard by the Defense Meteorological Satellite Program (DMSP). The sensor was onboard a series of satellite platforms that started in 1978: F8, F10, F11, F13, F14, and F15. The last platform F15 operated until August 2009 but measurements from SSM/I have troubles since 2006. In addition to a single vertical polarization channel operating at 22.2 GHz, the SSM/I carried radiometers that operated in dual-polarization mode at 19.3, 37.0, and 85.5 GHz. Data are sampled over 104.2 from end to end per scan line while the scanner is looking backward with incidence angle 53.1 . The spatial sampling rate for the first three channels was 25 km and for the 85 GHz channel 12.5 km. More details about the sensor are presented in Hollinger, Lo, Poe [1987], and an evaluation of its performance (stability of the gain, radiometric calibration, co-registration, electronic noise and sensitivity, etc.) is presented in Hollinger, Pierce, Poe [1990]. SSM/I has been used to create the longest record of sea ice extent and concentrations for the polar regions during its lifespan of 31 years. It was decommissioned in February 2009, leaving a legacy that stimulated more PM sensors to monitor polar ice. By inter-calibrating data from different PM sensors on different satellites, researchers could compile an even longer record of sea ice in the polar regions to study the variability and identify trend of ice distribution and extent. Cavalieri, Parkinson, Vinnikov [2003] present an analysis of 30 years of Arctic sea ice from PM data. They used data from the ESMR (December 1972–March 1977), SMMR and SSM/I (October 1978–June 1988), and SSM/I alone (June 1987–December 2002). Operational ice charts from the US National Ice Center (NIC) were also used to fill gaps. Other records of PM data that revealed seasonal, regional, and interannual variability of Arctic ice are presented in Parkinson et al. [1999], Parkinson and Cavalieri [2002], Zwally et al. [2002a], and LeDrew et al.

SATELLITE SENSORS FOR SEA ICE MONITORING 351

[1992]. The National Snow and Ice Data Center - Distributed Active Archive Center (NSDIC-DAAC) distributes Level-3 of sea ice concentration record, which covers a period from October 1978 to present. The record is offered in EASE-Grid and polar stereographic projections. It uses data from SMMR, SSM/I and Special Sensor Microwave Imager Sounder (SSMIS). The SSMIS is described in section 8.3; since it is not part of the historical sensors addressed in this section). All these sensors are part of US programs. Other modern PM sensors generated by Japan and China are also described in section 8.3. Passive microwave observations have fairly coarse resolution of a few tens of kilometers from the 19 GHz and 37 GHz channels and a few kilometers from the highfrequency end of the 85 GHz. Therefore, they are more suitable for synoptic observations that serve climatic applications. For tactical and mesoscale ice information, which is required for planning the navigation routes of marine vehicles through ice-covered water, imaging radar systems are more suitable. Therefore, small-scale ice features such as ridges, small leads, and coastal polynyas can be identified better with the radar systems. The first space-borne imaging radar sensor, namely the SAR, was launched onboard the Seasat satellite on 28 June1978 carrying an L-band HH polarization antenna (see section 7.6.1.2 for definitions). The mission ended abruptly on 9 October 1978 due to a power failure. In spite of its short period of operation (105 days), Seasat provided data of great scientific value and proved the successful use of SAR for oceanic and sea ice applications. It sets the stage for more space-borne SAR missions that were developed later. The 25 m fine-resolution SAR imagery data from Seasat were used to produce the first detailed sea ice motion maps. Individual ice floes could be identified in sequential images and that enabled the monitoring of ice field and ice tracking [Hall and Rothrock, 1981]. An ice deformation grid was also produced for the first time from Seasat [Fily and Rothrock, 1987]. This allowed the identification of openings in the ice cover and determining heat exchange from the ocean to the air. Carsey and Holt [1987] measured the impact of the wind on ice drift at the marginal ice zone from Seasat data (see more about this zone section 2.9.3.1). Based on these successful initial applications, NASA supported the development of the Geophysical Processor System (GPS) as a subsystem of the Alaska Satellite Facility (ASF) at the University of Alaska, Fairbanks. This system has been dedicated to producing sea ice and ocean products from SAR imagery [Kwok, et al., 1990, Weeks, Weller, Carsey, 1991, Kwok and Baltzer, 1995]. In 1999 it was upgraded to process SAR Arctic ice images from other satellites, namely the European ERS and the Canadian RADARSAT into ice data products.

The first space agency that captured on the legacy of Seasat to develop the next space-borne SAR system was the European Space Agency (ESA). They launched their first SAR sensor onboard the European Remote Sensing (ERS-1) satellite on 17 July 1991. That was followed by its replica onboard ERS-2 on 21 April 1995. Both sensors operated in the C-band (around 5.6 cm wavelength) with a single polarization channel (VV). The viewing geometry was limited to one mode featuring a 100 km swath over an incidence angle ranging from 20 to 26 at a resolution of approximately 30 m. ERS-2 operated for more than 16 years by the time it was decommissioned in July 2011. Data from ERS-1 were used in operational demonstration programs to explore their utility in sea ice mapping in terms of ice classification and extent [e.g., Hakansson, Moberg, Thompson, 1995, Shokr, Ramsay, Falkingham, 1996, and Johannessen et al., 1996]. Some success had been achieved in terms of identifications of ice features in the image but no method had proven to satisfy operational requirements as explained in section 7.6.2. Data from ERS-2 were heavily used in the 1990s to develop methods of sea ice classification but an overall conclusion highlighted the limitation of using a single-channel SAR data for this purpose because of the overlap of typical backscatter from different ice types (section 9.3.1). Although efforts for developing automated ice classification from these early SAR systems had never matured, retrieval of ice motion was developed at ASF. This was achieved with limited success because of the narrow swath of the ERS data (100 km) and the infrequent satellite revisits over the same site (the satellite repeat cycle was 35 days). Dokken [2000] compiled a number of research summaries of ERS SAR projects conducted under the auspices of ESA, based on documentations provided by ESA and the Canadian Space Agency (CSA). In February 1992 the National Space Development Agency of Japan (NASDA), which was renamed to Japan Aerospace Exploration Agency (JAXA) in 2003, launched its first space-borne SAR sensor onboard the Japanese Earth Resources Satellite (JERS-1). It featured an L-band (23.5 cm wavelength) operating at HH polarization. This relatively long wavelength allowed more penetration through the ice. Therefore, the received backscatter carried more information about the ice subsurface properties. In essence, the longer wavelength of the L-band is more appropriate than the C-band in detecting ridges, rubble fields, and brash ice. Moreover, the relatively shallow incidence angle of JERS-1 (about 35 at the middle of the swath) made it more capable of detecting other ice surface deformation. Dierking and Busche [2006] conducted a useful comparison between the L-band JERS-1 and the C-band ERS-1 and demonstrated that the images of both sensors complement each other, resulting in a more detailed view of the sea ice cover state.

352

SEA ICE

8.2. OPTICAL AND THERMAL INFRARED SENSORS The first satellite with an optical sensor used in sea ice and snow applications was the Advanced Very HighResolution Radiometer (AVHRR). This is a cross-track scanning radiometer system, which was available in a series of 19 sensors. Depending on the model, it encompasses four or five channels in the VIS, NIR, and TIR spectral bands. Data from the VIS and NIR channels can be converted to albedo while data from TIR channel can be converted into brightness temperature. Calibration information appended with the file allows for the conversion. AVHRR series was carried on NOAA’s PolarOrbiting Environmental Satellites (POES) series beginning with TIROS-N in 1978. It has provided global coverage since June 1979 with resolution 1.1 km at nadir but degrades toward the far range. Each pass of the satellite covers a wide swath of about 2400 km. The Europe’s contribution to the Initial Joint Polar System shared with NOAA produced the European Meteorological Operational satellites MetOp, which continues to new versions of AVHRR. These satellites orbit the Earth at altitudes between 830 and 870 km, circling the Earth approximately 14 times per day with each orbit taking approximately 102 minutes. The orbits are timed to achieve global coverage twice per day, normally daytime and night time. Table 8.1 includes information on the platforms that carry AVHRR. Note the long lifetime of NOAA-15. Data can be obtained from the USGC and NOAA POES websites. Three data products are available from the POES AVHRR: the High-Resolution Picture Transmission (HRPT), which is real-time downlink data; the Local Area Coverage (LAC), recorded onboard and later transmitted to the ground receiver; and the Global Area Coverage (GAC), which is a reduced resolution image processed onboard the satellite. In addition to several meteorological and climatological applications, AVHRR data have been used in discriminating sea ice from open

water (hence determining sea ice extent), estimating sea and ice surface temperature, melt pond on ice, among other applications. There are three generations of AVHRR. Table 8.2 includes information on the spectral bands of the successive generations AVHRR/1, AVHRR/ 2 and AVHRR/3. Another optical/TIR sensor from which data have been used in sea ice applications is NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS). This was a follow-on AVHRR series. Two identical MODIS sensors were launched on NASA’s satellites Terra (launched on 18 December 1999) and Aqua (launched on 4 May 2002). Both are operating until this time (2022). Both satellites orbit the Earth at 705 km latitude. The sensor has 36 channels, of which 19 operate in the VIS/NIR range [Salomonson, Barnes, Masuoka, 2006]. MODIS has a viewing swath width of 2,330 km and views the entire surface of the Earth every one to two days. Table 8.1 Satellite platforms that carried AVHRR sensors, NOAA and MetOp series. Satellite name TIROS-N NOAA-6 NOAA-7 NOAA-8 NOAA-9 NOAA-10 NOAA-11 NOAA-12 NOAA-14 NOAA-15 NOAA-16 NOAA-17 NOAA-18 NOAA-19 MetOp-A MetOp-B MetOp-C

Launch date

End date

13 October 1978 27 June 1979 23 June 1981 28 March 1983 12 December 1984 17 September 1986 24 September 1988 13 May 1991 30 December 1994 13 May 1998 21 September 2000 24 June 2002 20 May 2005 6 February 2009 19 October 2006 17 September 2012 7 November 2018

30 January 1980 16 November 1986 7 June 1986 31 October 1985 11 May 1994 17 September 1991 13 September 1994 15 December 1994 23 May 2007 present 9 June 2014 10 April 2013 present present present present present

Table 8.2 Spectral bands of the channels of AVHRR modules and their NOAA or MetOp satellite platforms.

Channel

AVHRR/1 on NOAA-6,8,10

AVHRR/2 on NOAA-7,9,11,12,14

AVHRR-3 on NOAA-15,16,17,18,19 and MetOp-A,B

1 2 3A 3B 4 5

0.58–0.68 0.725–1.1 – 3.55–3.93 10.5–11.5 Band 4 repeated

0.58–0.68 0.725–1.1 – 3.55–3.93 10.3–11.3 11.5–12.5

0.58–0.68 0.725–1.0 1.58–1.64 3.55–3.93 10.3–11.3 11.5–12.5

SATELLITE SENSORS FOR SEA ICE MONITORING 353

This is the longest-living sensor in the history of satellite remote sensing. The sensor has 36 spectral bands ranging in wavelength from 0.4 μm to 14.4 μm. Several products are available from this sensor free of charge (upon registration), which were developed (and continue to be improved) using advanced algorithms with inputs from combinations of the available channels. MODIS data are available in three levels, starting at a swath level and progressing through higher level products. Level 1B is swath raw data geolocated to latitude and longitude centers of 1 km resolution. The data should be converted to radiance, then surface reflectance (for the VIS channels) or brightness temperature (for the TIR channels) before performing any analysis or parameter retrieval. Level 2 is a geophysical product usually provided in the sinusoidal projection. Level 3 data are gridded into map projection format. Sea ice products are provided in the three levels. A list of the products, data format, algorithm description and quality assessment are included in Riggs and Hall [2015]. Sea ice products from MODIS start at level 2 with a swath coverage of 1354 km (cross-track) by 2030 km (along-track) and nominal resolution of 1 km. They progress in sequence. Each product in the sequence is made using the previous sea ice product as input. The first product is swath sea ice extent and ice surface temperature (IST). These two parameters are included in the second, third, and fourth products. The second product is intermediate multidimensional created by mapping pixels to their Earth locations on the Lambert Azimuthal equal area EASE-Grid projection. The third product is daily daytime ice cover and IST at 1 km resolution. The fourth product is a daily global map at 4 km resolution. The Visible Infrared Imaging Radiometer Suite (VIIRS) is a follow-on development of AVHRR and its MODIS successor. It is developed by NASA and flies aboard two satellites: Suomi National Polar-orbiting Partnership (Suomi NPP). Data are distributed by NASA’s Land Processes Distribution Active Archive Center (LP DAAC). While MODIS and VIIRS have similar orbits, the spatial resolution of the “thermal bands” on each sensor differs. The VIIRS instrument offers 22 spectral bands ranging between 412 nanometers and 12.5 micrometers at two spatial resolutions of 375 m and 750 m. Several VIIRS bands are spectrally broader than corresponding MODIS bands. For sea ice applications, data from VIIRS are available since 2012 to users through the NASA’s data products after resampling to 500 m, 1 km, and 0.05 degrees. This is to promote consistency with the MODIS data. The 3000 km swath of VIIRS is larger than MODIS swath. VIIRS L3 products of ice thickness, ice concentration, and IST are also available through NOAA’s CoastWatch dataset. Snow and ice products from VIIRS are summarized in Key et al. [2013].

The next-generation of NASA-NOAA Joint Polar Satellite System (JPSS) include four missions planned to last through 2031. Each of them will host a VIIRS instrument as part of their payload. Note that JPSS-1 (NOAA20) was launched on 18 November 2017, and JPSS-2 was launched on 10 November 2022.

8.3. MODERN PASSIVE MICROWAVE SENSORS The most commonly used PM sensors in sea ice applications are listed in Table 8.3 (in the past and present). The shown resolutions are the dimensions of the integrated field of view (IFOV). However, data are usually provided to users in gridded format such as the equal area projection of EASE grid at finer pixel spacing (casually called resolutions). It varies between 25 × 25 km2 and 12 × 12 km2, depending on the sensor and the spectral channels. All sensors listed in Table 8.3 have radiometers operating in the horizontal and vertical polarizations. At the time of writing this chapter, only the Advanced Microwave Scanning Radiometer 2 (AMSR2) has been widely used for near-real-time ice information. The Chinese Fengyun FY-3 PM radiometer is also operational yet with limited applications of sea ice as the data accuracy and robustness are being established. However, the Advanced Microwave Scanning Radiometer— Earth Observing System AMSR-E and, to a limited extent, SSMIS sensors (both are successors of SSM/I) are still used in non-real-time applications. AMSR-E was developed by JAXA in collaboration with NASA. It was launched on NASA’s Aqua platform in 2001. The sensor was used extensively in sea ice and snow research and monitoring until it failed in October 2011. In addition to sea ice concentration, AMSR-E was used to determine sea ice thickness (section 11.4.2) and surface temperature (section 11.5.2). SSMIS operated onboard the satellite DMSP platform series F-16, F-17, F-18, and F-19 between October 2003 and April 2016. It was the successor of SSM/I and had 26 channels ranging from 19.53 GHz to 183.31 GHz. With availability of highfrequency sounding channels, data were used to retrieve a variety of atmospheric parameters (e.g., temperature, moisture) and land parameters. However, data were also used to estimate ice and snow parameters, which are provided in global maps at 25 km × 25 km resolution (https://nsidc.oeg/data/amsre). AMSR2, developed by JAXA, was launched on 18 May 2012 onboard the Global Change Observation Mission (GCOM-C1) satellite and still operational at the time of writing this book. It has provided continuation of observations after the failure of its predecessor AMSR-E but with slightly enhanced resolution (Table 8.3). The first composite image of the Arctic ice was acquired on 3 July 2012.

354

SEA ICE

Table 8.3 Specification of commonly used PM sensors with the channels used in sea ice applications.

Sensor

Developer

Frequency (GHz)

Resolution (IFOV) (km)

Swath (km)

Start date

End date

Along scan

Cross scan

SSM/I

NASA/NOAA

1400

July 1987

Feb. 2009

19 37 85

43 28 13

69 37 15

AMSR-E

JAXA

1445

May 2002

Oct. 2011

18.7 36.5 89

16 8 6

27 14 4

SSMIS

NASA/NOAA

1700

Oct. 2003

Feb. 2016

19.35 37.0 91.66

73 41 14

47 31 13

AMSR2

JAXA

1450

May 2012

–

18.7 36.5 89

14 7 3

22 12 5

SMOS

ESA

1000

Nov. 2009

–

1.4

35

50

MWR onboard FY-3B/C/D

CMA/NSMC China

2250

Sept. 2013

–

18.70 36.50 89.00

30 18 9

50 30 15

The Soil Moisture and Ocean Salinity (SMOS) satellite mission is known as “water mission.” It was developed by ESA under its Living Planet Program in collaboration with the National Centre for Space Studies (CNES in French) in France and the Center for Technological and Industrial Development (CDTI) in Spain. This is the only PM sensor that operates in a single low-frequency band in both vertical and horizontal polarizations, the L-band (1.4 GHz). While the mission did not aim at retrieval of sea ice parameters, ice thickness (up to 50 cm thick) has been retrieved and commonly used in research studies (see section 11.4.2). The Level 3 daily maps of sea ice thickness are provided by ESA’s Earth observation gateway in NetCDF format. The maps are available in areas at the edge of the Arctic Ocean during winter (October–April) from year 2010 onwards. Daily maps are also generated by the Alfred Wegener Institute (AWI), projected on polar stereographic grid of 12.5 km, with uncertainty of retrieval attached to each pixel. This product is complemented with sea ice thickness measurements from ESA’s CryoSat and the Copernicus Sentinel-3 missions (section 11.4.3). The Fengyun satellite series FY-3 has been developed by China Meteorological Administration/ National Satellite Meteorological Center (CMA/NSMC). The satellite carries a microwave radiometer imager (MWRI), which has been used in limited sea ice applications so far. Operational missions of this satellite series include FY-3C (launched in September 2013), FY-3D (launched in November 2017), FY-3E (launched in July 2021) and FY-3F (launch schedule in 2023). Recent publications on using FY-3C/MWRI for sea ice include Shi et al. [2021] and Zhao et al. [2021] on estimation of sea ice

concentration as well as Li, Chen, Guan [2021] on retrieval of snow depth on Arctic sea ice. The sea ice operational community was concerned about a possible discontinuity in the availability of space-borne PM data after the decommissioning of AMSR2. In response to this concern, ESA is currently developing a mission for Copernicus Imaging Microwave Radiometry (CIMR). Copernicus is a European system for monitoring the Earth to support of European Union’s Earth observation program and its integrated policy for the Arctic. Planning of the CIMR started in 2018 and the satellite carrying the radiometer is scheduled for launch in 2025 or beyond. This will be a global wideswath, conically scanning multi-frequency imaging radiometry with a focus on polar regions. More information on this mission is found in Kilic et al. [2018]. By inter-calibrating data from the NASA’s SSM/I and SSMIS, a long record of sea ice concentration known as NOAA/NSIDC Climate Data Record (CDR) was established [Meier et al., 2017]. The algorithms blend two well-known methods: the NASA Team (NT) [Cavalieri, Gloersen, Campbell, 1984] and NASA Bootstrap (BT) [Comiso, 1986]. The CDR record begins in 1987 with DMSP SSM/I PM data, rather than in 1978 with NASA Nimbus-7, SMMR data because the latter did not meet the transparency requirements of brightness temperature. Cavalieri, Parkinson, Vinnikov [2003] presents an analysis of 30 years of Arctic sea ice from PM data. They used data from the ESMR (December 1972–March 1977), SMMR and SSM/I (October 1978–June 1988) and SSM/I alone (June 1987–December 2002). Operational ice charts from the US NIC were also used to fill gaps.

SATELLITE SENSORS FOR SEA ICE MONITORING 355

8.4. MODERN IMAGING RADAR SENSORS Theoretical and operational information about SAR is introduced in section 7.6.1. Satellite SAR systems that have been used in sea ice applications (in past and present) are included in Table 8.4. Data from a few systems have been available and used offline after the end of the satellite mission. It is important to understand the difference between various systems, especially between their frequency and polarization modes, and to consider how to homogenize the data sets to optimize their combined use. The RADARSAT program was borne out of the need for effective monitoring of ice in Canada’s sea and lake waters, especially in the Arctic region. The CSA launched its first SAR system onboard the RADARSAT-1 satellite in November 1995. The satellite was decommissioned in March 2013 surpassing its originally-designed 5-year lifetime by 13 years. It carried a C-band SAR with single HH polarization but had seven imaging modes that offer different viewing geometry and spatial resolutions. The standard mode had seven beams (referred to as S1 to S7), each one covered approximately 100 km swath at different viewing angles with 25 m spatial resolution and 12.5 m pixel spacing. The most frequently used mode for operational ice monitoring was the ScanSAR-Wide, which had nominal resolution of 100 m and nominal swath of 500 km or 450 km. CSA and the Canada Center for Mapping and Earth Observations (CCMEO), a branch on Natural Resources Canada, is making available 36,500 historical RADARSAT-1 images available for public at no cost. This is part of Canada’s Open Government efforts to encourage novel Big Data Analytic and Data Mining (BDADM) activities by users. Data can be accessed through the data management system of CCMEO.

ENVISAT was launched on 1st March 2002 and lost suddenly on 8 April 2012. It carried the C-band Advanced SAR (ASAR), which was the first system that featured multi-polarization data acquisition through its dualpolarization mode. In addition to its single polarization (VV), which was similar to its predecessor ERS-1/2 systems, ASAR operated in one of five selective polarization modes: two single co-polarization HH or VV and three alternating co- and cross-polarization modes: (HH and HV), (VV and VH), or (HH and VV). The system had 12 viewing modes including the most usable in sea ice applications, namely the Image Mode Medium and Alternating Polarization Medium modes (both have 150 m nominal resolution). ENVISAT data can be accessed via the On-The-Fly (OTF) service established by ESA Earth Online Dissemination Service. This long record has been utilized in numerous research studies to establish interannual variability of sea ice and snow parameters in the polar regions. Animation of ASAR mosaics of the Arctic Ocean between June and September 2008 was generated. The entire record of ASAR can be used to generate a longer record of Arctic ice variability during the lifetime of the satellite. Dierking and Pedersen [2012] presented an overview on the achievements made in sea ice monitoring by using ASAR during ENVISAT’s lifetime. The Phased Array type L-band Synthetic Aperture Radar (PALSAR) was launched onboard JAXA’s Advanced Land Observing Satellite (ALOS-1) in 2006. It was decommissioned in 2011. This was followed by launching PALSAR-2 onboard ALOS-2 in 2014 (and still operational). Data from these two satellites were available from single, dual and, for the first time, full polarimetric polarization. Each sensor has seven viewing modes, including ScanSAR Normal and Wide. The latter has a swath of 490 km with 60 m resolution. However, data

Table 8.4 Space-borne SAR systems used in sea ice applications. The S, D, and F in the polarization columns refer to single, dual, and fully polarimetric mode, respectively. Satellite/Sensor Seasat ERS-1/ERS-2 JERS-1 RADARSAT-1 ENVISAT ASAR ALOS- PALSAR series RADARSAT-2 TerraSAR-X & TanDEM-X COSMO-SkyMed series RISAT-1 Sentinel-1 GF-3 RADARSAT Constellation Mission (RCM)

Agency/Country

Band

Pol.

Lifespan

Resol. (m)

Swath (km)

NASA, USA ESA JAXA, Japan CSA, Canada ESA JAXA, Japan CSA, Canada DLR, Germany ASI, Italy ISRO, India ESA CNSA Canada

L C L C C L C X X C C C C

S (HH) S (VV) S (HH) S (HH) S, D S, D, F S, D, F S, D, F S, D S, D, F, CP S, D S, D, F S, D, F, CP

1978–1978 1991–2011 1992–1998 1995–2013 2002–2012 2006– 2007– 2007– 2007–2021 2012–2017 2014– 2016– 2019–

25 30 18 × 24 10–100 30–500 7–100 3–100 1–18 15 1–50 5 × 20 1–500 variety

100 100 75 50–500 5–406 20–350 10–500 100 × 150 30–40 10–225 250 10–650 variety

SEA ICE

from this mode were not available for operational sea ice monitoring. It should be mentioned also that ALOS-2 includes, for the first time also, a compact polarimetric (CP) mode but only from the High Sensitive Mode (6 m resolution) and Fine Mode (10 m resolution) (see section 7.6.3 for description of the CP mode). As far as the authors of this book know, the CP mode on PALSAR systems was never tested for sea ice applications. Several studies were performed to evaluate the use of other PALSAR modes for sea ice and compare results against the use of other C-band systems. A recent study by Karvonen et al. [2020] evaluated the use of dual-polarization PALSAR-2 ScanSAR data for their utility in estimating sea ice concentration, type, thickness and drift. PALSAR data can be accessed through NASA Earth Data website or ASF DAAC website https://earthdata.nasa.gov/eosdis/daacs/asf. RADARSAT-2 was launched in December 2007, carrying also a C-band SAR and it is still operational at the time of writing this chapter. It is a commercially operational system. Therefore, data availability for scientific research has been limited within a few initiatives managed by the CSA. The government of Canada financed the development of this satellite, which was developed, owned, and operated (along with the ground segment) by MacDonald Dettwiler Associates (MDA) Ltd. In return, data have been available to Canadian Government agencies for free. RADARSAT-2 has the capability of acquiring data from single polarization, dualpolarization in two modes: (HH and HV) or (VV and VH), as well as fully polarimetric (FP) mode (see

section 7.6.3). However, only the single polarization (HH) has been used in the sea ice monitoring program by the CIS using visual analysis of the images. This has been the prime data source in their program. Compared to RADARSAT-1, more modes have been added to RADARSAT-2 as shown in Figure 8.2. The ScanSAR wide and narrow modes remain to be the commonly used options for sea ice monitoring. Some RADARSAT-2 capabilities that benefit sea ice applications are the multipolarization options that improve ice edge detection, ice type discrimination, and ice topography and structure information. The Ultra-Fine beam mode (3 m resolution) offers the potential for ship classification. Data from the FP mode are not used operationally because of their limited swath (25 km). Weekly year-round maps of sea ice in three Arctic regions: western, eastern Arctic and Hudson Bay are generated by CIS using RADARSAT-2 complemented with Sentinel-1 SAR. The maps can be accessed free of charge through the CIS website. Details of RADARSAT-2 modes and technical specifications are available in Fox, Luscombe, Thompson [2004] and MacDonald Dettwiler and Associates [2009]. TerraSAR-X is a German SAR mission developed through a joint venture between the German Aerospace Center (abbreviated DLR in English) and the European Aeronautic Defence and Space Company (EADS). It was launched in June 2007 and its successor TanDEM-X was launched in early 2010. Both sensors are still operational and have single, dual, and fully polarimetric options. Though not relevant to the subject of this book, it is worth mentioning that the two satellites fly at distance of a few

Satellite velocity vector All beam modes available In Right-and Left-Looking

Spotlight

50

0k

m

0k m

0k

25

25

Ultra-Fine

49°

m

50

0k

Extended High

Extended Low

ScanSAR Wide/Narrow

Multi-look Fine

Quad-pol standard/Fine

m

20°

Standard Beam

Sub-satellite ground track

Wide Beam

356

Figure 8.2 RADARSAT-2 imaging modes. Nine modes are available. (ScanSAR has two modes: Wide and Narrow) (Adapted from MacDonald Dettwiler and Associates, 2009).

SATELLITE SENSORS FOR SEA ICE MONITORING 357

hundred meters from each other. This unique constellation allows generation of digital elevation models of the Earth’s land surface with an unprecedented vertical accuracy of 2 or 10 meters. The ScanSAR mode of TerraSAR-X covers up to 100 km × 150 km at a spatial resolution of up to 18.5 km, while the Wide ScanSAR (WS) mode covers up to 270 km × 200 km at a spatial resolution of 40 m. Data from TerraSAR-X are not often used in sea ice applications. Most of the studies using this sensor address data comparison against other SAR data. A notable study by Eriksson et al. [2010] compared multi-polarization SAR data from the L-band ALOS, the C-band ASAR, the Cband RADARSAT-2, and the X-band TerraSAR-X satellites for sea ice monitoring in the Baltic Sea. A main conclusion from the study is that the information content in the X-band and C-band is largely equivalent, whereas L-band data provide complementary information. The Constellation of Small Satellites for Mediterranean basin Observation (COSMO-SkyMed, also referred to as CSK) series is composed of four spacecrafts, conceived by the Agenzia Spaziale Italiana (ASI), and funded by the Italian Ministry of Research (MUR) and the Italian Ministry of Defense (MoD). This is an X-band SAR system with resolution up to 40 m depending on the operation mode with swath width up to 200 km from the ScanSAR mode. The overall objective of this program was set to be a contribution to global Earth observations. While the frequency and duration of the coverage allows access of details on the evolution of sea ice conditions, data have not been used often in sea ice applications. Data are archived and can be accessed through ESA’s Earth observation gateway. The series of the Radar Imaging Satellite (RISAT) was developed by the Indian Space Research Organization (ISRO). This is another multi-mode C-band system. Data from RISAT-1 have been used in experimental studies to explore their applications in sea ice. SAR onboard this satellite featured the three polarization options of single (S), dual (D) and FP (Table 8.4) in addition to the CP option. After user adoption, RISAT-1 became the first EO system to include a CP SAR mode. The potential utility of this option in retrieving sea ice parameters was explored for the first time in a few studies [e.g., Singha and Ressel, 2017]. Espeseth, Brekke, Johansson [2017] compared RISAT-1 CP data against RADARSAT-2 FP data measurements and showed that results were comparable. However, to the best knowledge of the authors of this book the satellite was decommissioned in 2017 without developing serious data applications for sea ice. Sentinel-1 is a constellation of two satellites, Sentinel1A (launched in April 2014) and Sentinel-1B (launched in April 2016), both carrying identical C-band SAR systems and following the same orbit. This is the first constellation within the Copernicus Programme, developed and

operated by ESA. The repeat cycle of each satellite is 12 days, so it is reduced to 6 days when data from both satellites are used. SAR on this satellite has four modes of radar operation Strip Map (SM), Interferometric Wide swath (IW), Extra-Wide swath (EW), and Wave (WV). The IW and EW modes are the most usable in sea ice applications. The IW mode features 5 m × 20 m spatial resolution with 250 km swath. The EW mode features 20 m × 40 m spatial resolution with 400 km swath. Both modes offer single polarization (HH or VV) or dual polarization [(HH + HV) or (VV + VH)]. SAR data from these two satellites are used frequently in sea ice research and operational applications because ESA and the European Commission’s policies have made data access free of charge for registered users. Data can be accessed through the Copernicus Open Access Hub and ASF DAAC. Sentinel-1C and 1D are in development with Sentinel-1C set for launch in 2022. Gaofen-3 (GF-3) is the first civilian C-band polarimetric SAR system developed by the China National Space Administration (CNSA). The satellite was launched on 10 August 2016 but data were made available to Chinese users about 3 years later. One of its main purposes is monitoring ocean and coastal areas. This SAR has multiple working modes such as two side view imaging modes (left and right). The nominal resolution ranges from 1 m to 500 m, and the nominal swath varies from 10 km to 650 km. There are 12 viewing modes, of which three modes are of FP capability. More technical specifications of the sensor are included in Zhang [2017]. Although GF-3 SAR has been operational for a period of time, the data have not been shared internationally. This has limited the exploration of the data in different applications. However, thousands of images have been distributed to Chinese institutions and used in a few studies to assess their quality and potential applications. Studies that address sea ice classification have been published in Li et al. [2017], Zheng, Li, Ren [2018], and Zhang et al. [2021]. Utility of the FP mode from GF-3 still needs to be assessed as the data are becoming more available. The RCM, launched in June 2019, is a trio Earth observation satellite platform. The satellites are evenly spaced on the same orbital plane, separated by 32 minutes or approximately 1,46,000 km. They were declared operational in November of same year and expected to operate for 7 years. RCM builds on the legacy of RADARSAT-1 and RADARSAT-2, and on Canada’s expertise and leadership in Earth observation from space. It has offered a significant advancement in the field of SAR architecture by including a hybrid polarity system known as compact polarimetry [Raney, 2007]. This system transmits right circular polarization pulses and receives two linear orthogonal signals, i.e., offering two polarization configurations: RH and RV. It can provide more-or-less same

358

SEA ICE

polarimetric parameters as offered by the FP mode of SAR systems. In addition to the CPO mode, RCM has 16 more imaging modes. The potential of this mode was summarized in a preliminary review on the future applications of the data [Charbonneau et al., 2010]. A few investigations on potential applications of CP data for sea ice used simulated data based on RADARSAT-2 FP mode [Dabboor and Geldsetzer, 2014, Geldsetzer et al., 2015, Dabboor, Montpetit, Howell, 2018; Dabboor and Shokr, 2020]. More studies using actual data from the RCM are emerging since data have been released to users first and foremost in the Canadian Government. Data may be available to users outside the Government of Canada subject to exceptions in terms of security, privacy and confidentiality. Data from the three satellites that constitute the RCM system, will allow imaging the same site more often, i.e., improve temporal resolution. RCM has the capacity to view any point on 90% of the world’s surface every 24 hours (except around the South Pole). It covers areas in the Arctic up to four times a day. This capability will facilitate the generation of composite images that monitor changes in sea ice daily. The CIS is planning to use images from the RCM CP mode as the prime source in the sea ice operational monitoring program.

8.5. SCATTEROMETER SENSORS Table 8.5 includes a list of operational scatterometer systems used for sea ice applications since 1991. Before this date, the short-lived Seasat satellite carried its Ku-band (13.8 GHz) fan-beam scatterometer in 1978. That was followed by the C-band scatterometer systems onboard ERS-1 and ERS-2 (1991–2011), which were called Active Microwave Instrument (AMI). These scatterometers operated in the C-band (5.3 GHz) VV polarization. Data from these sensors were primarily dedicated to retrieval of sea surface wind measurements; but they were used in limited ways for sea ice mapping in the Arctic. The system featured

fixed-fan-beam configuration with three antenna beams (section 7.7, Figure 7.27). It samples the sea surface at azimuth angles of 45 , 90 , and 135 with respect to the satellite flight direction. The scatterometer illuminates a 500-km-wide swath at incidence angles ranging from 18 to 57 . Backscatter measurements are binned over 50 km wide overlapping cells, and offered in 25 km × 25 km grid cell. In 1996, NASA launched a Ku-band scatterometer known as NSCAT but it failed shortly after nine months. Since the data from this satellite were not used often for research or operational purposes, it is not included in Table 8.5. As a quick replacement, NASA launched the SeaWinds Ku-band (13.4 GHz) scatterometer onboard the QuikSCAT satellite on 19 June 1999. This sensor provided valubale sea ice data over the polar regions for more than 10 years before its rotating antenna stalled on 23 November 2009. Sea ice products available from this sensor include Arctic and Antarctic ice exent (available from NSIDC website), first-year and multi-year ice classification (availabe from Young Brigham Young University FTP server), ice age (NASA JPL website) and calibrated backscatter coefficient. In 2006 ESA launched another C-band scatterometer, called Advanced Scatterometer (ASCAT) onboard the European polar-orbiting satellites MetOp-A and MetOp-B. Similar to the ERS scatterometer systems, it measures backscatter in VV polarization only. Using ASCAT and the PM SSMIS sensors, sea ice age maps have been produced for the Arctic for the period 2014–2019, available through the BYU website. Sea ice extent, approximate ice age and normalized backscatter maps from the long records of ERS, QuikSCAT and ASCAT scatterometer are generated by the EUMETSAT’s Ocean and Sea Ice Satellite Application Facility (OSI-SAF) in 12.5 km polar stereographic grid. The Ku-band scatterometer that followed QuikSCAT is called Ocean Scatterometer (OSCAT), built by the ISRO. This sensor was launched aboard the Oceansat-2 satellite on 23 September 2009. Data from the SeaWinds

Table 8.5 Space-borne scatterometer systems used in sea ice applications. The letters (i) and (o) denote the inner and outer beams, respectively. Satellite/Sensor ERS-1/-2 / Wind Scatterometer QuikSCAT / SeaWinds ASCAT / MetOp OceanSat-2 / OSCAT SCATSAT-1 / OSCAT-2 HY-2A HY-2B Scatterometer

Country/Agency ESA USA ESA India India CASTC

Band

Pol.

Lifespan

Resol. (m)

Swath (km)

Ku Ku C Ku Ku Ku

VV HH, VV VV HH, VV HH, VV VV

1991–2011 1999–2009 2006– 2009– 2016–2021 2011–2021 2018–

25/50 4.5 (resamp.) 12.5 km 4.5 (resamp.) 4.5 (resamp.) 33 × 23 (i) 37 × 26 (o)

500 1400 (i), 1836 (o) 1100 1400 (i), 1836 (o) 1400 >1400

SATELLITE SENSORS FOR SEA ICE MONITORING 359

and OSCAT have been reconstructed onto a polar stereographic grid using knowledge of the scan geometry and antenna pattern. China Aerospace Sceince and Technology Corporation (CASTC) launched a Ku-band pencil beam conically scanning scatterometer onboard the satellite series Hai Yang (HY-2A/B). The outer beam measure backscatter in VV polarizationat 49 incidence angle (footprint 37 × 26 km) and the inner beam in HH polarization at 41 incidence angle (footprint 33 × 32 km). A few studies were conducted to use data from this sensor in sea ice applications [Zhao et al., 2021, Shi et al., 2017]. Enhanced resolution data from QuikSCAT, ASCAT, and OSCAT over the polar regions are made available in an open-source database generated by the Center of Remote Sensing, Brigham Young University [Early and Long, 2001]. The image reconstruction of these data was possible because the scatterometer system makes multiple overlapping paths over each polar region. Reconstructed data can be downloaded through the website ftp.scp.byu.edu. Products are available as daily averaged maps from the rotating-beam Ku-band scatterometer at the enhanced resolution of 4.45 km and every second day from the fixed fan-beam C-band scatterometer.

8.6. ALTIMETER SENSORS As mentioned in section 7.8, satellite altimetry is a powerful tool for observing sea ice and ice sheets. They are mainly designed to measure the variation of global mean sea level but data found applications in monitoring the sea ice thickness at synoptic scale (section 11.4.3). The system measures sea ice and ice sheets topography at high latitudes. There are two types of altimeters, radar and laser.

Table 8.6 includes a list of operational space-borne altimeters used in sea ice applicarions. Radar altimeters were part of the payload of ERS-1, ERS-2, and ENVISAT. They all operated in the Ku microwave band (13.8 GHz). However, as those satetellites were Sun-synchronous, the orbit inclination was 98.52 (far enough from the 90 ). This did not allow data collection over an area centered around the geographic polar points. Sea ice applications of those sensors included ice thickness, lead identification, ice freeboard retrieval, and ice/water classification. The ENVISAT altimetry data are available to the user community via direct download from the ESA online archives. The Joint Altimetry Satellite Oceanography Network (JASON) series is a US–European satellite altimeter mission dedicated to measure the ocean surface topography, hence calculate the speed and direction of ocean currents and monitor global ocean circulation. The program is also named after the Greek mythological hero Jason. Once again, data found applications in sea ice. Another altimeter system, called Altika was launched onboard the Satellite with ARgos and ALtiKa (SARAL). It is a cooperative altimetry technology mission developed by ISRO and CNES. A prime payload of SARAL is the altimeter Altika. It is the first space-borne altimeter to operate at Ka band. It is set to take over ocean monitoring applications from ENVISAT. A few research studies on sea ice have been accomplished using data from this sensor. For example, Zakharova et al. [2015] used the maximum power of waveform to identify leads during the freezing season in the Arctic. Leads as narrow as 200 m in width could be identified. All SARAL data products are distributed in the Sentinel Standard Archive Format for Europe (SAFE) format.

Table 8.6 Space-borne altimeter systems used in sea ice applications. Satellite/Sensor

Agency

Type

Lifespan

Resol. (km)

ERS-1/2 ENVISAT

ESA ESA

Radar-Ku Radar-Ku

1991–2011 2002–2012

Jason-1/ -2

NASA/CNES

Radar- Ku

2001–13/2008–

Altika / SARAL ICESat / GLAS CryoSat-2 SIRAL ICESat-2 / ATLAS Jason-3 HY-2A/B Sentinel-3 / SRAL

CNES/ISRO NASA ESA

Radar- Ka laser Radar-Ku

2013– 2003–2009 2010–

16–20 0.35 (along-track) 80 (across-track) 11.2 (along-track) x 5.1 (across-track) 2 0.07 0.25

NASA NASA, CNES, NOAA, CASTC ESA

laser Radar-Ku Radar-Ku, C Radar-Ku

2018– 2016– 2011–/ 2018– 2016–

0.03–0.10 2–10 depending on sea wave height 16 0.3

360

SEA ICE

The first laser altimeter, Geoscience Laser Altimeter System (GLAS), was launched onboard the NASA’s Ice Cloud and Elevation Satellite (ICESat) on 13 January 2003 and mission ended on 14 August 2010 [Zwally et al., 2002b]. The system operated at two wavelengths of 1064 nm and 532 nm and the mission provided data on ice thickness across the polar regions. Measurements were affected by optically thick cloud cover and weather patterns in the polar regions. Technical difficulties allowed the sensor to operate over periodic intervals, so data were collected only through a patchy time series. GLAS is a profile sensor that acquires data from narrow stripes of 70 m width (cross-track) with footprints 3 km apart. Images can be constructed from sequential satellite overpasses. A summary of ICESat measurements of Arctic sea ice thickness is presented in Kwok and Rothrock [2009]. The second generation of the orbiting laser altimeter, the Advanced Topographic Laser Altimeter System (ATLAS), was the sole instrument on ICESat-2, which was launched in September 2018. The sensor operates at same wavelengths of ICESat. During the 8 years period between the end of ICESat and the launch of ICESat-2, the operation IceBridge program, an air-borne laser mission, was launched by NASA. ICESat-2 launched with advanced laser technology, shooting 10,000 laser pulses per second as compared to the 40 pulses per second on the first ICESat. Data were intended to measure changes in elevation of polar ice sheets as well as sea ice thickness. Using snow depth and density data from the NASA Eulerian Snow on Sea Ice Model, Petty et al. [2020] presented the first map of winter sea ice thickness (February– March 2019) from ICESat-2 data. The authors show that ice has lost about 37 m of thickness (about 20%) during the period from 2008 to 2019. ICESat-2 data are available through the NSIDC DAAC ICESat/GLAS collection. The NSIDC DAAC also provides data from ICESat and IceBridge missions. ESA’s CryoSat-2 mission was launched on 8 April 2010, almost five years after its predecessor CryoSat-1 was lost in a launch failure. CryoSat-2 is still operational, carrying the single payload of SAR Interferometric Radar Altimeter (SIRAL). The sensor operates in the Ku-band (13.4 GHz) (similar to the altimeters onboard ERS-1, ERS-2 and ENVISAT). The satellite has been designed to cover as much of the polar regions as possible. Consequently, the orbit of the satellite has a high inclination of 92 , taking it just 2 short of the poles, to acquire images in the polar areas up to 88 latitude. In order to achieve this goal, the orbit has to be a non-Sun-synchronous. As such, CryoSat-2 mission has been considered an “ice explorer.” SIRAL has three modes of operation. The SAR mode provides the finest ground resolution of 250 m in the along-track and about 15 km in the cross-track directions.

The repeat cycle of CryoSat-2 is 369 days with a track spacing of over 4 km. This means that mapping ice thickness over the entire polar region requires numerous passes that have to be acquired over several weeks. CryoSat data are distributed free of charge through ESA’s Earth Explorer database. Regular ice freeboard data and the derived ice thickness maps over the Arctic are available through NSIDC. China National Space Administration included a radar altimeter system on its HY-2 satellite series. Data were mainly used to estimate surface height and compare it against results from other systems such as Jason-2. Data were also tested for their potential applications in sea ice classification. So far, the data have not been made available to public. ESA’s SENTINEL-3 satellite carries an improved along-track resolution (approximately 300 m) altimeter called Synthetic aperture Radar ALtimeter (SRAL) and dual frequency PM radiometer (MWR). SRAL is a dual frequency system operating in Ku and C-bands. It provides continuity with ERS, ENVISAT and CryoSat altimeter missions but offers estimated ice thickness at higher accuracy and resolution. The sensor has two modes of operation, a low-resolution mode (LRM), which is a conventional pulse-limited mode, and a SAR mode with high along-track resolution. In the SAR mode, the sensor facilitates measurement over sea ice and land ice. The Level-2 SRAL/MWR products have three dedicated outputs for ice sheets, ice and sea ice thickness. Data can be accessed through ESA’s Copernicus Programme. 8.7. REFERENCES Carsey, F.D. and Holt, B. (1987) Beaufort-Chukchi ice margin data from Seasat ice motion, Journal of Geophysical Research, 92(C7), pp. 7163–7172. Cavalieri, D.J., Gloersen, P. and Campbell, W.J. (1984) Determination of sea ice parameters with the NIMBUS 7 scanning multichannel microwave radiometer, Journal of Geophysical Research, 89, pp. 5355–5369. Cavalieri, D.J., Parkinson, C.L. and Vinnikov, K.Y. (2003) 30-Year satellite record reveals contrasting Arctic and Antarctic decadal sea ice variability, Geophysical Research Letters, 30(18). Available from: doi:10.1029/2003GL018031. Charbonneau, F.J. et al. (2010) Compact polarimetry overview and applications assessment, Canadian Journal of Remote Sensing, 36(Suppl. 2), pp. S298–S315. Comiso, J.C. (1986) Characteristics of Arctic winter sea ice from satellite multispectral microwave observations, Journal of Geophysical Research, 91(C1), pp. 975–994. Dabboor, M. and Geldsetzer, T. (2014) On the classification of sea ice types using simulated Radarsat Constellation Mission (RCM) compact polarimetric SAR parameters, Proceedings of American Society of Photogrammetry and Remote Sensing (ASPRS) Conference, Louisville, Kentucky, March 23–28.

SATELLITE SENSORS FOR SEA ICE MONITORING 361 Dabboor, M. and Shokr, M. (2020) Compact polarimetry response to modeled fast sea ice thickness, Remote Sensing, 12, p. 3240. Available from: doi:10.3390/rs12193240. Dabboor, M., Montpetit, B. and Howell, S. (2018) Assessment of the high resolution SAR mode of the RADARSAT constellation mission for first year ice and multiyear ice characterization, Remote Sensing, 10, p. 594. Dierking, W. and Busche, T. (2006) Sea ice monitoring by L-band SAR: An assessment based on literature and comparisons of JERS-1 and ERS-1 imagery, IEEE Transactions on Geoscience and Remote Sensing, 44(2), pp. 957–970. Dierking, W. and Pedersen, L.T. (2012) Monitoring sea ice using ENVISAT ASAR—A new era starting 10 years ago, IEEE International Geoscience and Remote Sensing Symposium 2012, Munich, Germany, pp. 1852–1855. Dokken, S.T. (2000) Sea ice and ocean environmental applications of spaceborne SAR, PhD thesis, Chalmers University of Technology, Gothenburg, Sweden. Early, D.S. and Long D.G. (2001) Image reconstruction and enhanced resolution imaging from irregular samples, IEEE Transactions on Geoscience and Remote Sensing, 39(2), pp. 291–302. Eriksson, L.E.B. et al. (2010) Evaluation of new spaceborne SAR sensors for sea-ice monitoring in the Baltic Sea, Canadian Journal of Remote Sensing, 36(Suppl. 1), pp. S56–S73. Espeseth, M.M., Brekke, C. and Johansson, A.M. (2017) Assessment of RISAT-1 and RADARSAT-2 for sea ice observations from a hybrid-polarity perspective, Remote Sensing, 9 (11), p. 1088. Fox, P.A., Luscombe, A.P. and Thompson, A.A. (2004) Radarsat-2 SAR modes development and utilization, Canadian Journal of Remote Sensing, 30(3), pp. 258–264. Fily, M. and Rothrock, D.A. (1987) Sea ice tracking by nested correlations, IEEE Transactions on Geoscience and Remote Sensing, GE-25(5), pp. 570–580. Geldsetzer, T. et al. (2015) All-season compact-polarimetry C-band SAR observations of sea ice, Canadian Journal of Remote Sensing, 41(5), pp. 485–504. Hakansson, B., Moberg, M. and Thompson, T. (1995) Real time use of ERS-1 SAR imagery for ice service and ice breaking operations in Baltic Sea, International Journal of Remote Sensing, 16(17), pp. 3441–3458. Hall, R.T. and Rothrock, D.A. (1981) Sea ice displacement from Seasat synthetic aperture radar, Journal of Geophysical Research, 86(C11), pp. 11,078–11,082. Hollinger, J.P., Pierce, J.L. and Poe, G.A. (1990) SSM/I instrument evaluation, IEEE Transactions on Geoscience and Remote Sensing, 28(5), pp. 781–790. Hollinger, J.P., Lo, R. and Poe, G. (1987) Special Sensor Microwave/Imager User’s Guide, Washington, D.C.: Naval Research Laboratory. Johannessen, O.M. et al. (1996) Near-real-time sea ice monitoring in the Northern Sea route using ERS-1 SAR and DMSP SSM/I microwave data, Acta Astronautica, 38(4-8), pp. 457–465. Karvonen, J. et al. (2020) On suitability of ALOS-2/PALSAR-2 dual-polarized SAR data for Arctic sea ice parameter estimation, IEEE Transactions on Geoscience and Remote Sensing, 58(11), pp. 7969–7981.

Key, J.R. et al. (2013) Snow and ice products from Suomi NPP VIIRS, Journal of Geophysical Research: Atmosphere, 118, pp. 1–15. Kilic, L. et al. (2018) Expected performances of the Copernicus Imaging Microwave Radiometer (CIMR) for an all-weather and high spatial resolution estimation of ocean and sea ice parameters, Journal of Geophysical Research: Oceans, 123 (10), pp. 7564–7580. Kwok, R. and Baltzer, T. (1995) The geophysical processor system at the Alaska SAR facility, Photogrammetric Engineering and Remote Sensing, 61(12), pp. 1445–1453. Kwok, R. and Rothrock, D.A. (2009) Decline in Arctic sea ice thickness from submarine and ICESat records: 1958–2008, Geophysical Research Letters, 36(L15501), 5 pp. Available from: doi:10.1029/2009GL039035. Kwok, R. et al. (1990) An ice-motion tracking system at the Alaska SAR Facility, IEEE Journal of Oceanic Engineering, 15(1), pp. 44–54. LeDrew, E. et al. (1992) Canadian ice atlas from microwave remote sensing imagery: July 1987 to June 1990, Ottawa, Canada: Canadian Communication Group Publishing, ISBN-10: 0-660-57966-9, pp. 50–70. Li, L., Chen H. and Guan, L. (2021) Retrieval of snow depth on Arctic sea ice from the FY3B/MWRI, Remote Sensing, 13(8), p. 1457. Li, J. et al. (2017) Gaofen-3 sea ice detection based on deep learning, IEEE Progress in Electromagnetic Research Symposium-Fall (PIERS-FALL), IEEE, pp. 933–939. MacDonald Dettwiler and Associates (2009) RADARSAT-2 product description, Technical report No. RN-SP-52-1238, Richmond, British Columbia, Canada: MacDonald Dettwiler and Associates Ltd. Meier, W.N. et al. (2017) NOAA/NSIDC Climate Data Record of passive microwave sea ice concentration, Version 3, Technical Report, Boulder, Colorado USA: National Snow and Ice Data Center. Available from: https://doi.org/10.7265/ N59P2ZTG. Parkinson, C.L. and Cavalieri, D.J. (2002) A 21-year record of arctic sea-ice extents and their regional, seasonal, and monthly variability and trends, Annals of Glaciology, 34, pp. 441–446. Parkinson, C.L. et al. (1999) Arctic sea ice extents, areas, and trends, 1978–1996, Journal of Geophysical Research, 104 (C9), pp. 20,837–20,856. Petty, A.A. et al. (2020) Winter sea ice thickness from ICESat-2 freeboard, Journal of Geophysical Research, 125(5), e2019JC015764. Raney, R.K. (2007) Hybrid-polarity SAR architecture, IEEE Transactions on Geoscience and Remote Sensing, 45(11), pp. 3397–3404. Riggs, G.A. and Hall, D.K. (2015) MODIS sea ice products users guide to Collection 6. Available from: https://modissnow-ice.gsfc.nasa.gov/uploads/siug_c5.pdf Shi, L. et al. (2017) Sea ice extent retrieval with HY-2A scatterometer data and its assessment, Acta Oceanologica Sinica (English Edition), 36(7), pp. 1–8. Shi, L. et al. (2021) Sea ice concentration products over polar regions with Chinese FY3C/MWRI data, Remote Sensing, 13(11), p. 2174. Available from: https://doi.org/10.3390/ rs13112174.

362

SEA ICE

Shokr, M.E., Ramsay, B. and Falkingham, J.C. (1996) Operational use of ERS-1 SAR images in the Canadian ice monitoring program, International Journal of Remote Sensing, 17(4), pp. 667–682. Singha, S. and Ressel, R. (2017) Arctic sea ice characterization using RISAT-1 Compact-Pol SAR imagery and feature evaluation: A case study over northeast Greenland, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 10(8), pp. 3504–3514. Salomonson, V.V., Barnes, W. and Masuoka, E.J. (2006) Introduction to MODIS and an Overview of Associated Activities. In: Qu J.J. et al., eds. Earth Science Satellite Remote Sensing, Berlin, Heidelberg: Springer. Available from: https://doi.org/ 10.1007/978-3-540-37293-6_2. Svendsen, E. et al. (1983) Norwegian remote sensing experiment: Evaluation of the Nimbus 7 scanning multichannel microwave radiometer for sea ice research, Journal of Geophysical Research, 88(C5), pp. 2781–2791. Wark, D.Q. and Popham, R.W. (1960) TIROS-I observations of ice in the Gulf of St. Lawrence, Monthly Weather Review, 88 (5), pp.182–186. Weeks, W.F., Weller, G. and Carsey, F D. (1991) The polar oceans program of the Alaska SAR Facility, Arctic, 44(1), pp. 1–10. Zakharova, E.A. et al. (2015) Sea ice leads detection using SARAL/AltiKa Altimeter, Marine Geodesy, 33(Suppl. 1).

Available from: http://dx.doi.org/10.1080/01490419.2015. 1019655. Zhang, Q. (2017) System design and key technologies of the GF3 satellite, Acta Geodaetica et Cartographica Sinica, 46, pp. 269–277. Zhang, T. et al. (2021) Deep learning-based sea ice classification with Gaofen-3 fully polarimetric SAR data, Remote Sensing, 13, p. 1452. Available from: doi.org/10.3390/rs13081452. Zhao X. et al. (2021) Sea ice concentration derived from FY-3D MWRI and its accuracy assessment, IEEE Transactions on Geoscience and Remote Sensing. Available from: doi: 10.1109/TGRS.2021.3063272. Zheng, M., Li, X., Ren, Y. (2018) The method study on automatic sea ice detection with GaoFen-3 synthetic aperture radar data in polar regions, Haiyang Xuebao, 40(9), pp. 113–124. Zwally, H.J. et al. (2002a) Variability of Antarctic sea ice 1979–1998, Journal of Geophysical Research, 107. Available from: doi:10.1029/2000JC000733. Zwally, H.J. et al. (2002b) ICESat’s laser measurements of polar ice, atmosphere, ocean, and land, Journal of Geodynamics, 34, pp. 405–445. Zwally, H.J. et al. (1983) Antarctic sea ice, 1973–1976: Satellite Passive-Microwave Observations, NASA SP-459, Washington, D.C.: National Aeronautics and Space Administration, 206 pp.

9 Radiometric and Scattering Observations from Sea Ice, Water, and Snow

9.1 Optical Reflectance and Albedo Data .................................... 364 9.2 Microwave Brightness Temperature Data .............................. 370

9.3.3 Fully Polarimetric Data............................................... 387 9.4 Emissivity Data in the Microwave Bands............................... 395

9.3 Radar Backscatter .................................................................. 376 9.3.1 Backscatter Databases from Single-Channel SAR....... 378 9.3.2 Dual Polarization Data ............................................... 384

9.5 Microwave Penetration Depth................................................ 403

Typical radiative and scattering observations from sea ice, snow, and sea water are needed in many uses in the realm of remote sensing of sea ice. They are needed to support interpretation of remote sensing data and quantitative retrieval of ice parameters from the observations. The retrieval can be achieved using inverse models to recover prognostic ice/snow parameters or a search scheme to find the closest point (in a multi-dimensional space of physical parameters) to the given observation. The typical radiative/ scattering parameters can also be used to support development of forward models that simulate the observed reflected/emitted radiation or scattered signal from snow-covered sea ice. Typical values of ice albedo and emissivity find uses as input to climate and weather models. Typical backscatter from different ice types is needed to train ice classifiers, and typical microwave penetration depth is used to estimate ice thickness. In general, radiometric and scattering data from sea ice, sea water, and snow serve the purpose of developing ballpark values related to different snow and ice conditions. This may lead to developing an insight into the processes that trigger those values. Researchers who work on ice parameter retrieval algorithms have always hoped to identify observations or parameterizations that can unambiguously discriminate between ice types. However, more often than not, values of the parameters from different ice types or surface conditions overlap and sometimes seriously. This is

demonstrated in some data sets presented in this chapter. In certain situations of the observed data, the presented information is restricted to the snow cover because of the limited penetration of the incident or the emitted signal (i.e., the signal remains within the snow cover). Therefore, snow parameters should be carefully considered when analyzing the observations because the contribution from the ice may not be part of the observations. The chapter includes collections of radiometric and scattering data, or derived parameters, from different ice types, snow, and open water under different meteorological and surface conditions. They include direct observations obtained from satellite sensors, airborne sensors, and ground measurements from field experiments or laboratory studies of sea ice. No attempt has been made to explain discrepancies between data sets of the same parameter from different sources for any given ice type. Reasons for discrepancies may include lack of characterization of ice surface or snow cover during measurements, or inaccuracy in the absolute calibration of the instruments. Occasionally, brief outlines of techniques to derive or compile parameters are presented (as shown in the case of microwave emissivity in section 9.4). Furthermore, physical explanation is provided to clarify differences of certain parameters (e.g., albedo) between certain surfaces (e.g., ice and water). Five sets of parameters are presented. They include optical, passive microwave (PM), and radar backscatter

9.6 References............................................................................... 407

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 363

364

SEA ICE

observations as well as emissivity in microwave bands, and microwave penetration depth. The data are mostly grouped following the major age-based categories of sea ice; namely young ice (YI), first-year ice (FYI) and multi-year ice (MYI) in addition to open water (OW). One approach to resolve the overlap of values between different ice types is by increasing the dimensionality of the observations or the derived parameters (e.g., adding texture). A multisensor approach of the data (multispectral and multi-polarization) is also useful, hence included in some parts of the presentation. 9.1. OPTICAL REFLECTANCE AND ALBEDO DATA As defined in section 7.2.2.1, reflectance is the fraction of electromagnetic power reflected off a specific interface compared to the power of the incident radiation. Albedo is the hemispheric integration of reflectance in the optical spectrum (wavelength 0.30 to 3.0 μm). When albedo is determined from successive narrow bands it is called spectral albedo. When spectrally integrated over all solarreflective bands (VIS and NIR), it is called broadband albedo. The conversion from spectral to broadband albedo is accomplished via measurements from available spectral bands of a multispectral sensor and simulation of unavailable bands. Albedo is a crucial component in the radiative schemes of climate and sea ice models. Seong et al. [2022] present a credible study on the roles of Arctic sea ice surface albedo and skin temperature on radiative forcing. It also contributes to the self-perpetuating nature of sea ice as it triggers positive feedback that enhances the growth of sea ice. A classical review of optical properties of sea ice is presented in Perovich [1996]. Other comprehensive reviews are presented in Pedersen [2007] and Warren [2019]. A review of snow optical properties is presented in Mobley et al. [1998]. Useful discussions on optical properties of snow and ice, in general, and albedo in particular are presented in Lubin and Massom [2006]. Assessment of the decrease of albedo as a result of the decrease of Arctic sea ice is presented in Pistone, Eisenman, Ramanathan [2014]. In principle, albedo of water surface is considerably lower than that of ice, and albedo from ice is lower than that of snow. Albedo from ocean water surface is 0.07 and from FYI it is between 0.5 and 0.7, depending on the spectral band, with possible higher values around 0.9 when covered by snow [Allison, Brandt, Warren, 1993]. However, the refractive index of ice and snow varies around 1.31, depending on the wavelength. This is slightly less than the 1.39 from seawater. Since reflectivity is inversely related to the refractive index, these values mean that ice is only slightly more reflective than water, which

does not comply with observations in nature, where significantly higher reflection is observed from ice than water. A physical explanation is warranted at this point. Water has smooth surface with fairly homogenous (inclusion-free) subsurface layer. Therefore, the refracted light propagates down into the water without scattering. On the other hand, snow overlaid on sea ice has numerous micrometer-scale air bubbles and perhaps other scattering inclusions. Sea ice also has brine and air inclusions within the submillimeter depth under the surface. When the incident light refracted at the surface, it scatters off those inclusions and part of the scattered light reaches back to the surface and is refracted back to the air through the ice/air interface. This is how albedo of sea ice becomes higher than that of sea water. If no inclusions exist, the result will be a transparent ice sheet with very small albedo. The point to be emphasized here is that albedo is triggered not only by the refractive index at the surface but also by the composition of the material immediately under the surface. That is why snow has higher albedo because it is made up of flakes or small ice crystals with numerous air gaps between them. Light that reaches this composition will scatter causing higher albedo. Similarly, ocean water surface will feature high albedo only when foam is formed as a result of wave action. Foam is created by agitation of sea water when it contains higher concentration of dissolved air and organic matters. Following the above discussions, low albedo values result from conditions that enhance the absorption immediately under the surface. This includes the presence of liquid water within the snow volume or at the bare ice surface. Brine pocket in FYI may also contribute to changing albedo. As discussed in section 2.5.4, brine pockets within the polycrystalline sea ice matrix expand or shrink as ice temperature increases or decreases, respectively. Upon expansion, more absorption of optical radiation takes place (meaning, less albedo) and vice versa. Below the eutectic temperature of the sodium chloride (−22.8 C), the salt in the brine solution precipitates. This turns brine pockets into scattering elements instead of being absorbents. As a result, the albedo of sea ice below this temperature may increase as reported in Grenfell [1983]. Shortly after the onset of freezing, albedo from grease ice and nilas increases to values between 0.1 and 0.2, and it continues to increase as sea ice thickens. This results in continuous decrease of absorbed radiation and therefore less surface temperature (given constant air temperature). The opposite is true when sea ice starts to melt, i.e., surface albedo drops to values around 0.2, which means more absorption of solar energy and enhanced melt process. This is how the ice-albedo feedback loop acts positively to enhance booth ice growth and decay processes once they start. For this reason, sea ice albedo has been identified as an important climatic parameter, not only

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 365

because it modifies the surface radiation balance but also due to its possible role in perpetuating global temperature change. Climatologists are interested in the parameterization of sea ice albedo in order to strengthen the sea ice components in climate models. The purpose is to develop methods for estimating the global radiative forcing caused by sea ice-albedo feedback in the polar regions [Hudson, 2011]. Inaccurate estimate of this forcing leads to inaccurate prediction of climate change and its impacts. Climate models have underestimated the currently observed thinning of Arctic sea ice by a factor of 4 on the average and failed to predict the decrease of its extent [Rampal et al., 2011]. While this can partly be blamed on misrepresentation of the albedo, it can also be caused due to poor representation of other factors such as the sea ice southward drift, especially through Fram Strait. Nevertheless, the sea ice-albedo feedback is a key driver of the Arctic climate change [Thackeray and Hall, 2019]. Currently, models usually assume a constant albedo for sea ice or use climatological albedo compiled from numerous measurements of satellite remote sensing. Data of sea ice albedo are obtained from field measurements, physical-based models, or satellite observations. Each approach has its advantages and limitations. Field measurements data were obtained on opportunity basis, hence sporadic. This has led to an inconsistency of the recorded values between different investigations, which preclude a straightforward comparison. De Abreu [1996] addressed this point. The work presented physically based parameterization for albedo from snow-covered sea ice. It addressed the sea ice and snow parameters that affect the albedo and identified those that need to be included in modeling the albedo. Satellite observations have been used to map albedo at regional and global scale. Sea ice albedo measurements on a grid scale suitable for climate models can be obtained from the operational gridded albedo products of satellite data. This is currently available through NASA from MODIS data and from the European Ocean and Sea Ice Satellite Applications Facility (OSI SAF) products using AVHRR data. From the physical processes point of view, differences in the magnitude of sea ice broadband albedo are mainly attributed to differences in scattering, while variations in spectral albedo are mainly attributed to different absorption at different wavelengths. In the rest of this section, albedo data of sea ice and seawater, obtained from selected publications, are presented with links to factors that influence this parameter. Broadband albedo data are presented first followed by spectral albedo. One of the earliest compilations of surface-based broadband albedo measurements is presented in Perovich, Maykut, Grenfell [1986] and summarized in Table 9.1 with additional data from Grenfell and Maykut [1977].

Table 9.1 Broadband albedo under clear sky measured from Arctic sea ice and its snow cover. Surface type Sea water Bare FYI Frozen white ice Melting blue ice Melting white ice Ponded FYI Refrozen melt pond New snow Dry snow Wind packed snow Melting old snow

Albedo (P) 0.06 0.52 0.70 0.32 0.68 0.21 0.40 0.87 0.81 0.81 0.77

Albedo (G)

0.25 / 0.39 0.56 / 0.76 0.22 / 0.29

0.63 / 0.77

Note: Letters P and G stand for data obtained from Perovich, Maykut, Grenfell [1986] and Grenfell and Maykut [1977], respectively. The value after the “/” is albedo under cloudy sky.

The “white” and “blue” ice types are defined based on the presence or absence of elements that scatter light. Pure ice and seasonal sea ice with negligible air volume appear blue because they absorb all wavelengths of the visible spectrum, but the blue. In this ice, the absorption of light at the red end of the spectrum is six times greater than at the blue end as a result of an overtone of an oxygen– hydrogen (O−H) bond stretch [Braun and Smirnov, 1993]. On the other hand, when ice contains bubbles that scatter light at all visible wavelengths, as in the case of glacier ice and to some extent MYI, the reflected light increases and the perceived color changes from blue to white. Table 9.1 shows also that bare FYI has albedo around 0.52, which is significantly higher than that of the ponded ice (typical albedo of 0.21). It is interesting to note the difference between albedo from dry fresh snow (0.87) and melting old snow (0.77) in the table. Lower albedo can be reached as snow wetness increases. The water contents in the melted snow absorb more radiation in the NIR than the VIS regions. The low albedo of OW (0.06) is usually associated with a sun elevation of 30 or higher. Albedo increases at lower sun elevation and in the presence of capillary ocean wave. For example, Johannessen et al. [2007] found that the albedo of OW could reach 0.025, when sun elevation reached 10 over calm sea, but this decreased to 0.02 in the presence of moderate wave or lower with larger waves. In-situ measurements of broadband albedo from several age-based ice types in the Antarctic were compiled in Warren, Roesler, Brandt, [1997] from measurements obtained during East Antarctic cruises in the austral spring. Results are presented in Table 9.2 for six ice types. OW and grease ice have nearly equal low albedo. As mentioned in section 2.3.2, grease ice is not solid material but rather a “soupy” layer that does not reflect much light

366

SEA ICE

Table 9.2 Broadband albedo (0.3–2.8 μm) of standard sea ice types measured during East Antarctic cruises in the spring [Adapted from Warren et al.,1997].

Open water Grease ice Nilas: < 10 cm Gray ice: 10–15 cm Gray-white ice: 15–30 cm FY (thin) ice: 30–70 cm FY (thick) ice: > 70 cm

Snowfree

Thick snow ≥ 30 mm

0.07 0.09 0.16 0.25 0.35 0.42 0.49

— — 0.42 0.52 0.62 0.72 0.81

— — — 0.70 0.74 0.77 0.85

and cannot hold snow. After this early phase of ice formation, albedo increases rapidly as ice develops into nilas and gray ice types. The main point that can be drawn from Table 9.2 is the significant increase of surface albedo by adding even a thin layer of snow (< 3 cm) (further discussion on this point can be found in Allison, Brandt, Warren [1993] and Warren, Roesler, Brandt [1997]). Albedo from thick ice with more than 3 cm of snow increases to 0.85 up from 0.49 for the same ice without snow. The high albedo of the snow is mainly caused by irradiance at the shortwave [Brandt, Roessler, Warren, 1999]. Clouds increase the broadband albedo from sea ice. Grenfell and Perovich [1994] reported values 0.05 higher than the corresponding values under clear skies. Grenfell and Maykut [1977] measured clear-sky broadband albedo values of 0.63 and 0.22 for melting old Arctic ice with snow cover and with mature melt ponds, respectively. Their measurements of the equivalent values under cloudy skies were 0.77 and 0.29, respectively. Spectral albedo from sea ice demonstrates similar behavior to that from snow. Both start with high values in the VIS range, which decrease rapidly in the NIR range until it levels off to a minimum value throughout the rest of the spectrum. Spectral albedo of bare sea ice in the VIS range has been shown to vary significantly during early stages of ice formation. Snow-covered albedo also varies considerably during cycles of cold and warm spells of atmospheric temperatures as physical changes become linked to the fluctuating temperature. On the other hand, as the attenuation of incident radiation increases rapidly with wavelength in infrared region, albedo in the NIR spectral range becomes less sensitive to temperature fluctuations. Variation of spectral albedo with sea ice thickness can be examined using a modeling approach. In an early study, Grenfell [1983] used a multi-stream single layer model to generate thickness-based variation of spectral albedo of bare sea ice surface as shown in Figure 9.1. The ice thickness range (1–300 cm) in the figure represents a transition from the gray/black surface of a nilas ice sheet

1m

3m 0.8 Albedo

Ice type

Thin snow < 30 mm

1.0 50 cm

0.6 25 cm 0.4 10 cm 0.2 0.0

1 cm

400

600

800

1000 1200

1500

2000

2500

Wavelength (nm)

Figure 9.1 Modeled spectral albedo from bare sea ice surface with density 0.88 gm/cm3, salinity 11‰, and temperature −2.8 C, for different ice thickness [Grenfell, 1983, Figure 4a / John Wiley & Sons].

to the brighter surface of fully grown thick FYI. Albedo attains a maximum value between 400 to 550 μm wavelength, followed by a sharp decrease. In the NIR around 1 μm wavelength and beyond, albedo levels off to a constant low value ( 0.06) and therefore becomes largely insensitive to ice thickness [Allison, Brandt, Warren, 1993, Perovich, 1996]. The albedo of thick ice (≥ 1 m thickness) is about eight folds its value for 1 cm ice thickness in the VIS range, but it has the same low value in the NIR range. The rate of increase of albedo in the VIS range is higher during the early growth stage (for thickness 100% or < 0%). Data in Figure 9.10 show that the central Arctic is fully covered with FYI and MYI, while the distribution of the data from the Bering Sea is rather cumbersome as more sporadic distribution is observed along the line (OW-FY). According to Cavalieri [1994], it is not clear whether this distribution (from the Bering Sea) is due to the presence of thin ice types or an actual mixture of thicker ice types (FYI) with OW. It could also be due to water vapor contents in the atmosphere, generated by the large area of OW in that region. The wide spread of the data outside the triangle of the tie points as in Figure 9.10 highlights the difficulty of using typical values (tie points) of radiometric observations as representative of an ice type. Thin ice is particularly problematic because radiometric (or scattering) measurements spread over a large range. This is because air temperature fluctuations around freezing point during the lifetime of this ice type lead to significant changes in brine volume fraction and the subsurface composition. Same happens to the overlain snow. From a laboratory experiment on simulated sea ice grown in an

(b)

(c)

0.10 OW 0%

OW

OW

0%

0%

0.05

MY

n

co

nc

en

tra

tio

n tra

nc co

ce 80%

Central Arctic

MY

100%

Se

ai

ce ai 80%

Entire Arctic

100%

60%

Se

80%

40%

en

en nc co ce

60%

ai

–.05

FY

tio

tio

40%

FY

tra

40%

60%

MY

20%

n

FY

0.0

20%

Se

GR37V19V

20%

Bering Sea

100%

–0.1 0

0.2

0.1 PR19

0.3 0

0.1

0.2 PR19

0.3 0

0.1

0.2 PR19

Figure 9.10 Polarization versus gradient ratios derived from SSM/I observations over (a) entire Arctic, (b) the central Arctic, and (c) Bering Sea, all obtained from observations of on 4 April 1988 [adapted from Cavalieri, 1994, Figure 2 / with permission from John Wiley & Sons].

0.3

374

SEA ICE

outdoor ice tank, Grenfell and Comiso [1986] showed significant changes in the polarization ratio as ice grows up to 5 cm thick. This is a transition from the high polarization difference of water to the much less difference of sea ice. Grenfell et al. [1992] showed temporal trajectories of undisturbed YI in the GR19V37V and PR19 space. They confirmed that changes in both parameters occurred quickly during the early phase of ice growth but slowed down later on (i.e., thickness-dependent change). One of the most detailed studies on the behavior of thin ice in the GR19V37V and PR19 space is presented in Wensnahan, Maykut, Grenfell [1993]. They established a physical basis for the time evolution of these two parameters during the early ice formation stage. PM data from YI was studied, in detail, in Shokr, Asmus, Agnew [2009] through a series of simulated sea ice experiments grown in an outdoor tank in the premises of the National Research Council of Canada in Ottawa. The ice was grown up to 22 cm thick under natural cold atmospheric temperature in winter, while the surface was exposed to a variety of weather conditions and precipitations. Scatter plots of PR37 and GR37v19v from specific days with different ice thickness and surface conditions are shown in Figure 9.11. Points in the graphs represent data sampled every 15 minutes using a ground-based radiometer from the selected days shown by the labels in each panel. The vertices of the curvilinear triangle are tie points of OW, thin ice (< 10 cm thick), and thicker ice (> 15 cm thick), averaged from the entire data set under clear sky and dry ice/snow surface. The scattered points in the figure are connected to mimic their temporal evolution. Therefore, in addition to showing the spread of the data, the figure shows also how far the actual radiometric measurements are from the set of tie points. The temporal variability of the measurement is not caused by variability of thickness but rather of meteorological conditions (temperature, wind, and precipitation), It shows how sensitive emitted radiation from thin ice cover is to environmental conditions. This is the reason behind the difficulty of representing thin ice using a single tie point as clearly demonstrated in this figure. It is the major issue in estimating any thin ice parameter, thickness, concentration, or surface temperature. Figure 9.11 also shows that on 8 December 2005, the ice was 4 cm thick with a snow-free surface. The corresponding trajectory of the points is spread mainly outside the triangle. This means that the thin ice used in the calculations of the tie point of this ice did not represent the thin ice on that day. During the snowfall on 9 December 2005, the ice remained thin, and PR19 took values significantly below the tie point value. However, when the temperature rose to ice melting point after the snowfall, the values of both parameters, especially PR19, spread over a very wide range. This is a situation that shows how quickly the radiometric measurements from thin ice vary in response to

varying weather conditions. On the morning of 16 December 2005 (Figure 9.11d), heavy snow started to fall in the morning on a snow-free ice surface and by late afternoon it reached 16 cm depth. Once again, the trajectory of the two parameters spread over a wide range. This marks a period of unsettled snow, which extended for the rest of the day. Both PR19 and GR19V37V are clearly influenced by processes that take place during the snow settling. These include snow compactness, possible grain formation, and brine wicking from the ice surface. Apparently, these processes take place at a very short temporal scale. Bindschadler et al. [2005] provide more information on these processes. Data from wet snow surface obtained on 18 December 2005 (Figure 9.11e), also show a trajectory whose points fall outside the triangle (recall that the tie points were established without taking into account the wet snow condition). Toward the end of the experiment on 7 January 2006 (Figure 9.11f ), when the ice was 24 cm thick with 20 cm layer of dry stable (yet metamorphosed) snow, the trajectory forms a very narrow cluster centered near the tie point of the “thicker ice” as expected. That is the perfect condition for accurate parameter retrieval using the concept of tie points. In order to account for the variability of radiometric or scattering measurements from any ice type, especially thin ice, the full range of possible values should be considered. This can be achieved by employing the probability distribution of all possible values compiled under all surface and weather conditions. However, regardless of using a single representative value or the probability distribution of all possible values, the data should account for seasonal and regional variability as well as sensor drift. Algorithms implemented in the Ocean and Sea Ice Satellite Application Facility (OSI-SAF) use monthly sets of tie points to account for sensor drift and, therefore, allow for intersensor comparison. To examine the differences in microwave emission from different surface conditions, Shokr, Asmus, Agnew [2009] used the same simulated sea ice to examine Tb and different polarization and gradient ratios from seven surfaces developed in response to different weather conditions: saline water, wet slush, wet bare ice, refrozen slush, dry sow, wet snow, and dry ice surface (Figure 9.12). A trend of increase of Tb as the surface progresses from wetter to drier surface can be detected but the overlap of Tb from different surfaces is not small. Discrimination between water and slush surfaces is possible especially from 37 and 19 GHz. The transition between water and slush is associated with an increase of Tb by about 20 K (or more) from all channels except 85 V, where Tb is increased by only 10 K. Water is also separable from all ice surfaces in both gradient and polarization ratios. As for the gradient ratio, the narrowest and largest dynamic range is of GR19V37V and GR85H19H, respectively. Values of

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 375 (a)

(b)

.10

.10

Dec. 8, 2005: 4 cm ice, no snow Tie-point triangle: Thin ice, Thick ice, and OW

.08

.08

artifical sea ice: outdoor tank OW 20%

.06

GR 37v19v

GR 37v19v

.06 40%

.04 60% .02

80%

Thin ice < 10 cm

100%

.00

.04

.02

.00

Thicker ice > 10 cm –.02 .02

–.02 .04

.06

.08

.10

.12

.14

.16

.18

.20

.22

.24

.26

.02 .04 .06 .08

.10

PR19

(c)

(d)

.06

.06

GR 37v19v

GR 37v19v

.08

.04

Snow fall .02

–.02 .08

.10

.12

.14

.16

.18

.20

.22

.24

After snow fall

.04

.02

Heavy snow fall

–.02

.26

.02

.04

.06

.08

.10

PR19

(e)

(f)

.14

.16

.18

.20

.22

.24

.26

.10

Dec. 18, 2005: 12 cm ice + 6 cm slush + 5 cm snow .08

.08

.06

.06

GR 37v19v

GR 37v19v

.12

PR19

.10

surface flooded

.04

.20 .22 .24 .26

.00

Temperature approached melting; surface wet .06

.18

Dec. 16, 2005: 12 cm ice, snow fall (17 cm)

.08

.04

.16

.10

Dec. 9, 2005: 5 cm ice, during snow fall: 5 cm snow

.02

.14

PR19

.10

.00

.12

.02

Jan. 7, 2006: 23 cm ice + 4 cm frozen slush + 13 cm snow

.04

.02

.00

.00

–.02 .02

.04

.06

.08

.10

.12

.14

PR19

.16

.18

.20

.22

.24

.26

–.02 .02 .04 .06 .08

.10

.12

.14

.16

.18

.20 .22 .24 .26

PR19

Figure 9.11 (a) Tie points triangle in the PR19 and GR37v19v space for thin ice, thick ice, and OW obtained from simulated sea ice grown in an outdoor tank. The other panels (b–f ) show the temporal evolution of the two measurements, sampled every 15 minutes, during the selected day as marked in the panel. Different days are associated with different ice thickness, surface, and meteorological conditions as shown in the labels. Trajectories show wide range of variability except under dry surface conditions on 7 January when ice became thick (23 cm) [Shokr et al., 2009 / IEEE].

SEA ICE

300 280 260 240 220 200 180 160 140 120 100 80 60

0.35 0.30 0.25 Gradient ratio

Brightness temperature (K)

376

Water

Wet slush

Wet Refrozen ice srfc. slush

0.20 0.15 0.10 0.05

19V

19H

0.00

37V 85V

37H 85H

–0.05

Wet snow

Dry ice/ snow.

Water Wet slush Wet bare ice Refrozen slush Wet snow Dry snow Dry ice srfc.

–0.10 GR19V /37V

GR19H GR85V GR85H GR85V GR85H /37H /37V /37H /19V /19H

0.30 Water Wet slush Wet bare ice Refrozen slush Wet snow Dry snow Dry ice srfc.

Polarization ratio

0.25 0.20 0.15 0.10 0.05 0.00 –0.05 19 GHz

37 GHz

85 GHz

Figure 9.12 Plots of brightness temperature measurements from different surfaces of thin ice grown in an outdoor tank (from onset of freezing to 22 cm thick) and the corresponding polarization and gradient ratios. Error bars are standard deviation (M. Shokr).

GR85H19H from different surfaces overlap heavily. The ratio GR85H19H has its lowest value for the cases of dry ice surface and dry snow. It increases for all other surfaces indicated in the figure. This is in line with a remark made in Markus and Cavalieri [2000] that when “surface effects” come into play the ratio GR85H19H increases significantly. The seven surfaces shown in Figure 9.12 cannot be reliably distinguished using any polarization ratio, particularly from the 19 GHz channel. Microwave brightness temperature from ocean surface is fairly stable. However, it takes different values during rainfall. Since physical temperature of the water surface does not change during rain events, then the change in Tb must be caused by change of surface emissivity as a result of rain. As rain does not change the composition of the surface, then the change of emissivity must be caused by emission from the hydrometeors. In fact, the modulation of Tb during rainfall events over the ocean has been used to estimate the rate of rainfall over ocean [Lin and Rossow, 1997]. Using surface-based radiometer measurements from simulated seawater Shokr and Kaleschke [2012] found that the increase of Tb during

rainfall events is particularly pronounced at higher frequencies, and the increase from the horizontally polarized emission is significantly larger than that from the vertically polarized. This is shown in Figure 9.13, which depicts the evolution of Tb during a rainfall event. It peaked when the rainfall peaked. The maximum values of Tb-19h, Tb-37h and Tb-85h, observed during the peak of the rain were 122.0 K, 161.1 K, and 239.9 K, respectively, with the corresponding PR values of 0.190, 0.155, and 0.043. These values are significantly higher than those obtained from water surface with no rain (0.27, 0.25, and 0.19, respectively). It is worth mentioning that rainfall over land results in less observed Tb because it lowers the surface temperature and it is not associated with hydrometeors.

9.3. RADAR BACKSCATTER After the remarkable success of the first space-borne SAR onboard the short-lived Seasat in 1978, it was rightly felt that radar was well suited as an operational tool for

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 377 300

Tb-19V Tb-37V Tb-85V

Rain on OW

280

Tb-85V

240 220 200

Tb-37V

180

Tb-19V

160 Tb-85H

140

Tb-37H

120

337.01:25

336.23:25

336.19:23

336.17:23

336.15:22

336.11:20

336.13:21

336.06:57

336.04:57

336.02:56

336.00:55

335.22:54

335.18:53

335.20:54

335.16:52

335.14:51

335.08:46

335.06:45

335.04:44

335.02:44

335.00:43

334.22:42

336.08:58

Tb-19H

100

336.21:24

Brightness temperature (K)

260

Tb-19H Tb-37H Tb-85H

Figure 9.13 Evolution of Tb during an event of rainfall on simulated sea water in an outdoor tank. The rain started on 1 December at 8:46 am (335:08:46 denotes ddd:hh:mm). As soon as rain ended at 336:08:58 Tb resumed its typical value from OW [Shokr and Kaleschke, 2012 / with permission of Elsevier].

sea ice monitoring in polar oceans. As a result, many research programs were dedicated to explore the backscatter signature of different ice types with different surface conditions in different seasons using different radar frequencies and a variety of viewing geometries. A few programs were conducted throughout the 1980s using surface-based scatterometer systems to measure radar backscatter from natural Arctic sea ice [e.g., Onstott, 1979, Onstott and Gogineni, 1985]. A list of surface-based scatterometer investigations on natural and laboratory-grown sea ice conducted in the 1980s is presented in Onstott [1992]. A continuation list that covers the period from 1989 to 2006 is presented in Geldsetzer et al. [2007]. Radar backscatter measurements from thin ice types were also conducted on laboratory-grown ice in outdoor facilities. Of particular importance were the two series of experiments conducted in the Cold Regions Research and Engineering Laboratory (CRREL), located in Hanover, New Hampshire. The first, was conducted in the mid-1980s under the title CRREL Experiment (CRRELEX) [Swift et al., 1992], and the second was initiated by the US Office of Naval Research (ONR) in 1992, under the title Accelerated Research Initiative (ARI) to study the electromagnetic properties of sea ice [Jezek et al., 1998]. Airborne SAR systems such as NASA’s AirSAR and the CCRS’s Convair-580 were used to acquire backscatter data from ice types mainly in the western Arctic around Alaska and the east coast of Canada, respectively.

After the launch of SAR onboard the European satellite ERS-1 in 1991 and its successor on ERS-2 in 1995 (almost coincident with the Canadian RADARSAT-1 SAR launch), the number of SAR ice backscatter studies peaked. A commonly adopted approach involved linking the acquired backscatter data to ground observations from snow-covered ice types. In addition to constructing backscatter database of sea ice types, the aim was to develop empirical equations to identify ice types and surface conditions from the observed backscatter. Ice classification based on ice age/thickness was the main drive behind these activities in order to support national operational ice monitoring programs. In relation to this task, the pressing questions were about the information contents of the observed backscatter signal, namely about the factors that trigger the backscatter and their inks to the snow and ice conditions. This triggers significant backscatter modeling studies from snow and ice. The rest of this section includes samples of, and references to, backscatter data from selected sources that span a long period starting from 1980s until recently. The backscatter data are grouped according to the traditional thickness-based ice types, but some data are grouped according to surface conditions. In section 9.3.1 the presentation focuses on single-channel SAR then moves to databases from multi-channel SAR in section 9.3.2. Data from the recently developed fully polarimetric and compact polarization SAR modes (section 7.6.3) are presented to show what these technologies can offer to improve the

378

SEA ICE

information retrieval from snow-covered sea ice. This is covered in section 9.3.3. 9.3.1. Backscatter Databases from Single-Channel SAR The first in-situ measurements of radar backscattering from Arctic sea ice were conducted in May 1977 near Point Barrow, Alaska using a surface-based FM-CW scatterometer within the Transportable Microwave Active Spectrometer (TRAMAS) [Onstott, 1979]. Measurements were conducted over an incidence angle range from 10 –70 at radar frequencies ranging between 1 and18 GHz with co- and cross-polarizations. Although many elements regarding radar/ice interactions have been unraveled, the complexity of the composition of snowcovered sea ice process still poses challenges to understand the backscatter measurements and mechanisms. Radar pulses interact with surface roughness, salinity, crystalline structure, volume scattering elements, ice–water interface and snow metamorphism. This may result in a wide range of backscatter intensity from same ice types especially when covered by metamorphosed snow, hence overlap between backscatter from different ice types may be significant. More information of these interactions has been revealed using backscatter modeling with physical and radiative transfer components [Tonboe, Andersen, Toudal, 2003] as presented in Chapter 12. Many field and laboratory experiments followed the pioneering TRAMAS program. Onstott [1992] listed 17 investigations, conducted between 1977 and 1989, using ground-based scatterometer systems and airborne radars. The author presented sets of graphs of backscatter coefficients from the major ice types at L-, C-, X-, and Kubands and different incidence angles for winter and summer ice conditions. The first three frequencies became operational on satellite SAR platforms and the Ku-band became operational on several scatterometer platforms. Measurements aimed at offering clues to decide on the optimum frequency of future SAR systems for sea ice monitoring. A few trends of backscatter behavior from sea ice were identified and confirmed later in many studies. For example, backscatter from sea ice was found to decrease with increasing incidence angle, MYI returned highest backscatter under winter conditions, L-band returned highest backscatter from pressure ridges, and the backscatter from the Ku-band was the lowest among the tested frequencies. Figure 9.14 is a classical illustration of the angular behavior of radar backscatter from sea ice in winter, presented in Onstott [1992]. It shows the decrease of backscatter with the incidence angle and the capability of each frequency to discriminate between ice and OW. The high angular sensitivity of the Ku-band measurements from calm water is noticeable. However, Ku-band was shown later to increase with water surface

roughness (not shown in the figure). The trends of backscatter shown in Figure 9.14 are not greatly affected by the seasonal variations of ice cover. In a more recent study Eriksson et al. [2010] showed that the information content in the X- and C-band data is largely equivalent, whereas L-band data provide complementary information. This band is better for identifying ridges and heavily deformed ice surface. It seems to be less sensitive to wet snow cover on the ice. Cross-polarization data are not shown in Figure 9.14 but they become available from later studies as shown later in this section. Another dataset of radar backscatter was acquired during the Winter Weddell Gyre Study (WWGS’92) in 1992, in the Antarctic using a ship-borne C-band microwave scatterometer [Drinkwater, Hosseinmostafa, Gogineni, 1995]. Data were obtained at a variety of incidence angles between 15 and 70 . Graphs showing results from this experiment are compiled in Drinkwater [1998] and reproduced in Figure 9.15. The figure shows the angular variation of backscatter coefficient for four types of ice: white ice, smooth FYI, rough FYI, and MYI. White ice in this figure refers to thin FYI of thickness between 30–70 cm (a definition unique to this data set). Rough FYI refers to heavily deformed, ridged, or rubble surface. The data in the figure were obtained from ice that exceeded 60 cm in thickness. The cross-polarization return is significantly lower than the co-polarization return, but the latter drops considerably at higher incidence angles for white ice and smooth ice so that it becomes similar to the copolarization return. A remarkable observation from this figure is the indistinguishable backscatter from rough ice and MYI in the ERS-1 incidence angle range though the scattering mechanisms are different—surface scattering from rough ice and volume scattering from MYI. A sketch that summarizes the angular variation of the backscatter from OW, FYI, MYI, and marginal ice zone (MIZ) ice is presented in ESA (1998) and reproduced in Figure 9.16. The steep rate of the decrease of σ 0 from OW is noticeable. The overlap of the backscatter from the ice types with water highlights the importance of accounting for incidence angle variation within the wide swath of ScanSAR mode. More on the angular variation of polarimetric radar parameters follows in this section. The backscatter range from a few ice types with different surface conditions and different seasons was obtained from ERS-1 SAR (C-band VV polarization) during the Seasonal Ice Zone Experiment (SIZEX 92) in the Barents Sea [Sandven et al., 1999]. Results are presented in Figure 9.17. Note the wide range of backscatter coefficients from thin ice types including new, nilas, gray-white, and frost flower covered surfaces (−25 to −4.5 dB). Overlap of backscatter from different surfaces is also observed. In the summer, backscatter from FYI and MYI overlap heavily and occupy a relatively narrow range. This is

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 379 1.3 GHz, VV polarization 4 L-band

–12

C-band –10

σ° (dB)

σ° (dB)

–4

–20

5.2 GHz, HH polarization

0

Calm water Multiyear ice Thick first-year ice Pressure ridge Freshwater lake

–28

–20 –30 Calm water First-year ice Multiyear ice

–40

–36 15

–50 30

45

0

60

0

40

20

60

80

Incidence angle

Incidence angle 9.6 GHz, HH polarization

13.6 GHz, HH polarization

20 X-band

Ku-band

10 –10 σ° (dB)

σ° (dB)

0 –20

–10 –20

–30

–40 10

Multiyear ice First-year ice Calm water

20

Multiyear ice First-year ice Calm water

–30

30

40

50

60

Incidence angle

–40 0

10

20

30

40

50

60

70

Incidence angle

Figure 9.14 Co-polarization radar backscatter coefficients of winter Arctic sea ice at L-, C-, X-, and Ku- bands as a function of incidence angle [Onstott, 1992, Figure 5.19 / with permission from John Wiley & Sons].

because all ice surfaces, regardless of the ice type, become flooded. Backscatter from OW occupies a wide range from less than −20 dB for a quasi-steady ocean surface to −4.5 dB when the surface wind is 8 m/s or stronger. It is interesting also to note the wide range of backscatter from an ice category labeled “unknown transition.” This ice exists within the ice edge region in this dataset. It is the most complicated region of ice cover (section 2.9.3.2) in terms of physical and radiometric properties. This is not a stand-alone ice type, but the figure points to the difficulty of information retrieval at the ice edge. Data in Figure 9.17 highlight the need for using multi-channel SAR as a potential tool to reveal ice conditions more accurately. Variation of backscatter from sea ice during its growth from a few millimeters until it reaches maturity at around 50 cm was studied in late 1980s and early 1990s, in a series of outdoor laboratory experiments in CRREL in Hanover, New Hampshire. In Grenfell et al. [1992], the authors

concluded that C-band backscatter from very thin ice (~ 2 mm thick) increased with ice thickness until it reached a peak (around 30 cm thick) before declining gradually at a very mild rate to reach a minimum (at about 150 cm thick). Gradual increase of backscatter from the phase of grease/frazil ice throughout the phase of light nilas (i.e., within the category of new ice < 10 cm thick) was also observed in Beaven, Gogineni, Shanableh [1994]. Other radar backscatter databases are presented in Kwok et al. [1992] for the major ice types from C-band VV polarization scatterometer data. Drinkwater [1998] compiled a regional and seasonal database of backscatter from Antarctic sea ice using the C-band scatterometer (20 km resolution) onboard both ERS satellites. Measurements showed a large variability of backscatter from the same ice type under variable surface conditions. A first attempt to explore the possibility of mapping Arctic sea ice using backscatter from a scatterometer data is presented in Ezraty and Cavanie [1999]. The

380

SEA ICE

10 GHz as shown in Figure 9.35. If thick snow covers both hummock and melt pond ice, the difference in their emissivity is reduced because snow is also a scattering medium. This complicates the estimation of emissivity from MYI. Figure 9.36 [also from Grenfell, 1992] confirms the separability between hummock and melt pond ice in the 2D space of the 18 GHz and 37 GHz emissivity. The emissions from the two frequencies are highly correlated. This correlation is generally observed between frequency channels at different polarization or polarization channels at different frequencies, regardless of the type of ice cover. Once again, the lower values correspond to data from hummock ice. Surface roughness is another factor that influences the emissivity. As a rule of thumb, when the roughness scale matches the wavelength of the radiation more scattering takes place, therefore less radiation will be received by the radiometer. Grenfell et al. [1988] studied this effect using simulated sea ice of 15 cm thickness and resembled a rubble ice field by covering the surface with a layer of semicircular ice disks of about 25 mm thickness and 50 mm diameter. The influence of this layer on the 1.0

Emissivity 18 GHz

0.9

0.8

0.7

H polarization V polarization Open water (theory)

0.6

0.5 0.6

0.7

0.8

0.9

1.0

Emissivity 37 GHz

Figure 9.36 Correlation between emissivity of MYI at 18 and 37 GHz, measured at an incidence angle of 50 over more than 200 sites in the Arctic. Two clusters can be seen at the two sides of the vertical line. The clusters separate measurements from hummock ice (left) and melt pond ice (right) [Grenfell, 1992, Figure 5 / with permission from John Wiley & Sons].

emissivity is presented in Figure 9.37, which shows results before and immediately after the layer was added as well as the results obtained 12 hours later when the rubble pieces had frozen to the ice surface. Surface roughness triggers more emission at vertical polarization than horizontal polarization. The undisturbed ice surface also has higher emissivity in the vertical polarization. At 6 GHz, the emissivity from vertical polarization is the same after adding the roughness disks. This shows how the 25 mm thick disks appear smooth to the 6 GHz emission (about 5 cm wavelength). The presence of surface roughness decreases the emissivity at all higher frequencies as more scattering takes place. Freezing of the rubble has virtually no effect on the spectrum of the emissivity. The situation of the horizontal polarization is more complex. The undisturbed ice surface (no rubble) shows a distinct minimum near 6 GHz followed by a gradual increase to a maximum near 18 GHz, then it decreases at a slight rate. Grenfell et al. [1988] explained the minimum emissivity as being the effect of a first-order fringe pairs of the surface that were captured at lower frequencies. On the other hand, the higher order fringes that should have been captured at higher frequencies were not observed due to attenuation and scattering in the frost layer that covered the surface at the time when the rubble layer was introduced. When the rubble ice layer was added, the emissivity at low frequencies rose by about 0.15 at low frequencies then decreased monotonically at higher frequencies. As mentioned in section (7.9.3.2), presence of dry snow reduces the emissivity of sea ice due to increased scattering within the snow volume. Results in Figure 9.38 include measurements before and after a snowfall of about 45 mm on ice with average thickness of 120 mm. Once again, this was conducted on laboratory-grown snowcovered sea ice in CRREL [Grenfell and Comiso, 1986]. The fresh snow reduces the emission but not by the same amount for different frequencies and polarizations. Significant reduction can be seen in the 37 GHz data with horizontal polarization. Almost no reduction can be seen in 10 GHz data with vertical polarization. The angular decrease of the emissivity is in the horizontal polarization data. Looking at this graph, one can easily identify a pattern of angular variation of the emissivity for different microwave frequencies and polarization for bare ice surface or surface with dry snow. The real challenge of the snow, however, is connected to its metamorphism, which instigated by cycles of air temperature and wind changes during the lifetime of sea ice. Moisture in the snow causes higher absorption of solar radiation. Therefore, wet snow becomes a good emitter in microwave bands (i.e., has higher emissivity). Emissivity continues to increase with snow wetness until saturation is reached at emissivity value around 0.95. Moreover, as

402

SEA ICE 1.0

0.9

Emissivity

0.8

0.7 H polarization

V polarization

0.6

0.5

No rubble

No rubble

25 mm rubble layer

25 mm rubble layer

Rubble frozen to ice

Rubble frozen to ice

0.4 4

10

40

100

Frequency (GHz)

4

10

40

100

Frequency (GHz)

Figure 9.37 Emissivity spectra of 150-mm-thick ice before and after addition of a 25 mm layer of ice rubble to roughen the surface [Grenfell et al., 1988, Figure 3 / with permission from IEEE].

1 0.9

Before snow fall After snow fall

V pol.

V pol. H pol.

0.8

Emissivity

V pol.

V pol. H pol.

0.7

H pol.

Before snow

H pol.

0.6

After snow 0.5 10 GHz 0.4 30

18 GHz

37 GHz

Angle 60 30 Angle 60 30 Angle

90 GHz 60 30 Angle

60

Figure 9.38 Comparative spectral emissivity before and after a 45 mm of snowfall. Data from vertical and horizontal polarizations are shown [Grenfell and Comiso, 1986, Figure 5 / with permission from IEEE].

snow acquires wetness, emission from the underlying sea ice decreases. The above data demonstrate the effects of ice and snow parameters on the microwave emissivity. This information is required in forward modeling of remote sensing observations or for retrieval algorithms of sea ice parameters (Chapter 12). The information also advances our understanding of emissivity in relation to ice physics. Weather and climate models need only monthly or seasonal averaged emissivity in given regions. A complete set of monthly average emissivity of FYI and MYI in the Arctic was generated by Mathew et al. [2008] for the

microwave frequencies of the Advanced Microwave Sounding Unit (AMSU) radiometer onboard NOAA polar-orbiting satellites. AMSU-A has 15 channels in the frequency range 23–89 GHz, and AMSU-B has five channels in the frequency range 89–183 GHz. The observations scan the scene in a range of local zenith angles that vary between ±57 . Another study (Mathew, Heygster, Melsheimer, 2009) established same data for the frequencies of AMSR-E. In addition to the satellite observations, simulated brightness temperatures were used to allow calculation of emissivity using equation (9.13). The simulation was based on atmospheric model profiles of temperature and humidity from the ECMWF. The above two references present tables and graphs of monthly averaged emissivity of Arctic sea ice. For example, Mathew et al. [2008] included graphs of angular variation of the monthly averaged emissivity in 2005 for the MYI and FYI in the north of Greenland. The variation of the emissivity (represented by the standard deviation) is high during the summer months when the ice surface melts. For all frequencies and months, the variation of emissivity with local zenith angle is negligible up to 45 and then decreases as the angle increases. Another study [Willmes, Nicolais, Haas, 2013] used the microwave emission model MEMLS to study the seasonal and regional patterns of the emissivity in winter and spring in the two polar regions between 2000 and 2009. The authors found that the emissivity increases from winter to early summer and the steepest increase was estimated in the 85 GHz data during the month of June, in the Arctic. The corresponding increase in the 37 and

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 403

19 GHz were rather milder. No increase was estimated in the austral spring of the Antarctic (i.e., in December). The stable emissivity was lowest from the 85 GHz (around 0.7) and highest from the 19 GHz (around 0.95).

9.5. MICROWAVE PENETRATION DEPTH This section includes data on penetration depth (δp) in sea ice or snow obtained for the microwave bands of existing space-borne operational radiometers and SAR systems. This parameter depends on the frequency of the emitted or incident radar signal, its local incidence angle, and most importantly the composition and the degree of heterogeneity of the medium. The factors that affect δp in the case of snow-covered sea ice include the salinity and conductivity of the snow cover [Vant, Ramseier, Makros, 1978 and Hallikainen and Winebrenner, 1992), brine volume [Arcone, Gow, McGrew, 1986, and Ostott and Shuchman, 2004], brine salinity [Stogryn and Desargant, 1985, and Shokr and Barber, 1994], and ice porosity or air bubble contents of MYI [Fung and Eom, 1982]. As δp increases, the probability for EM wave of encountering absorbing inclusions, discrete scatterers or layers with dielectric mismatch within the radiating layer is enhanced. Those elements modulate the signal. Usually, the radiating layer of the emitted microwave is assumed to be equal to the penetration depth. The effective temperature within this layer determines the observed brightness temperature. Classical data sets of penetration depth within major sea ice types are reported in the literature with lower penetration expected from highly saline ice (lossy medium) and vice versa. However, within a given ice type, variability is encountered depending on the composition of the sea ice and the snow cover. It will be shown later how δp varies with ice thickness, salinity, crystalline structure, snow wetness, and snow grain size. Data in Table 9.10, obtained from Ulaby, Moore, Fung, [1982], show the ranges of δp for FYI, MYI, and snow with two different liquid water contents. Note the wide range associated with the gross categories of FYI and MYI. It is interesting to mention that δp from dry snow at 36.5 GHz is estimated to be around 92 cm [Ulaby, Moore, Fung, 1981). This means

that dry snow is almost transparent to the microwave at this frequency. This value of δp is significantly higher when compared to the values from 1% and 2% of liquid water content. Calculations of the penetration depth in the C-band microwave frequency using equation (3.71) with measurements of the two components of the dielectric constant ϵ and ϵ from FYI and MYI ice cores were obtained by M. Shokr (unpublished data) during field experiments in the Resolute Passage, Canadian Arctic between 1992 and 1995. Sea ice cores were used to measure the dielectric constant. The average FYI and MYI ice thickness were 140 and 210 cm, respectively, but measurements of the dielectric constant and the calculation of the penetration depth were performed on the top 40 cm. Results are presented in Table 9.11 for melt pond and hummock of MYI and three crystallographic structures of FYI: randomly oriented frazil, vertically oriented frazil, and columnar (see definitions and illustrations in section 5.3.3). The penetration depth in hummock ice (saline-free but bubbly ice) is larger than that in melt pond ice (slightly saline but almost bubble free). This means that the absorption within a lossy saline medium causes more signal extinction than that caused by the scattering off the bubble inclusions. Penetration depth within FYI with random frazil structure is noticeably less than that from oriented frazil, and the latter is less than that from columnar grained. Note also the remarkably smaller penetration depth in FYI compared to MYI, which is a manifestation of the lossy medium of saline FYI. Spectral variation of δp in the microwave range for FYI, MYI, pure ice is presented in Ulaby, Moore, Fung [1986] Table 9.10 Penetration depth (cm) for FYI, MYI, and snow with liquid water contents of 1% and 2% for the AMSR-E frequencies of 6.9, 18.7, and 36.5 GHz (data are compiled from Ulaby, Moore, Fung, 1982). Surface type First-year ice Multi-year ice Snow (1%) Snow (2%)

6.9 GHz

18.7 GHz

36.5 GHz

5–20 20–40 30 10

2.7 7–15 11 4

1–4 4–8 6 2

Table 9.11 Penetration depth in sea ice (cm) calculated from equation (3.71) using measurements of the complex dielectric constant in the microwave C-band for MYI (melt pond and hummock) and FYI (random frazil, oriented frazil, and columnar) (M. Shokr, unpublished data) Multi-year ice Melt pond 18.07

First-year ice Hummock

Random frazil

Oriented frazil

Columnar ice

35.58

3.61

4.55

7.49

404

SEA ICE

and reproduced here with addition of snow data (Figure 9.39). The calculations were performed using average values of complex permittivity for each surface type; all valid at −10 C. The upper and lower frequencies in the range 1–100 GHz are used. The average δp for FYI at 6 GHz and 19 GHz are 10 cm and 3 cm, respectively, while the corresponding values for MYI are 60 cm and 11 cm. For the L-band ( 1.4 GHz), δp is 2.5 m for MYI and 0.5 m for FYI. These agree with values presented in Heygster et al. [2009]. Freshwater ice has a much larger penetration depth; near 10 m at 6 GHz and 1 m at 102

Dry Mult

snow

i-yea

100

r ice

Pu re i

ce

First -yea ri

ce

10–1

100

85

37

6

10–3

19

10–2 1.4

Penetration depth δ (m)

101

101

102

Frequency (GHz)

Figure 9.39 Spectral variation of penetration depth in the microwave range for dry snow, pure ice, MYI, and FYI. The double-headed arrows indicate the range of variability of the penetration depth [adapted from Ulaby, Moore, Fung, 1986].

19 GHz. The penetration depth of the L-band (1.4 GHz) in dry snow is an order of magnitude higher than δp at 85 GHz. That is why the L-band radiometer on the SMOS satellite could be used to estimate ice thickness up to 50 cm (section 11.4.2). For OW (not shown in the figure) the penetration depth is only a few millimeters over the entire frequency range. Another dataset of δp covering nilas, dry snow, FYI with 13 cm overlaid snow, and MYI with 20 cm overlaid snow is presented in Table 9.12 for frequency range from 1.4 GHz to 150 GHz. Data were compiled from two sources: Wigneron et al. [2006] and Mathew et al. [2008]. Note the very large penetration in the snow compared to the ice and the larger penetration in MYI compared to FYI. There is a significant increase in the penetration of the L-band (1.4 GHz) in both ice types compared to the next frequency of 6 GHz. A more detailed graph of spectral variation of the penetration depth δp for FYI and MYI types, calculated from equation (3.71) using input from experimental values of the absorption coefficient (dB/m), is presented in Hallikainen and Winebrenner [1992]. The frequency range is limited between 1 to 10 GHz (Figure 9.40). This is the range that covers the three operational SAR systems (X-, C-, and L-bands), which is below the range of PM radiometers. Hence, data in the figure are more useful in the context of radar signal penetration. Data are compiled from studies by Vant et al. [1974], Vant, Ramseier, Makros [1978]; Hallikainen [1983], and Hallikainen, Toikka, Hyyppä [1988]. The logarithmic scale of δp in the figure is selected to reveal the wide range of values as they change with ice temperature and salinity. For FYI, δp increases as temperature decreases or as salinity

Table 9.12 Penetration depth (cm) at different microwave frequencies for the shown ice with different snow cover. Frequency (GHz) 1.4 6 10 19 23

5 cm Nilas

Dry snow

FYI /13 cm snow

MYI /20 cm snow

> 5.0 5.0 4.0 2.0 1.0

134.0 130.0 96.0 28.0 13.0

49.0 7.0 4.0 2.2 2.0 1.52 1.0 1.45 1.0 1.28 1.0 0.94 0.0∗ 0.75

160.0 32.0 22.0 13.0 11.5 7.76 9.0 7.32 3.5 6.23 0.90 4.0 0.0∗ 2.9

37

1.0

50

1.0

89

1.0

150

1.0

Note: The values shown in regular font are from a graph on page 390 in Wigneron et al. (2006) and in bold font are from Mathew et al. (2008). The 0.0∗ denotes penetration depth limited to 11 cm in the dry snow (i.e., does not reach the ice surface).

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 405 101

Fy 1.0‰

Penetration depth, δp (m)

5

My 1 .3‰, –9°C –20°C My 1.3‰ , –4 –5°C

2

Salinity by weight

Vant et al. [1978] Vant et al. [1974] Hallikainen [1983] Hallikainen et al. [1988]

My 1.3‰

°C

–20°C –2°C Fy 5.1‰, – 20°C –20°C Fy/columnar Fy 5 Fy 7.5 .1‰ –10°C –0.5°C F ‰, –2 4.6‰ , –5° y 10 –5°C 0°C –5°C C Fy 5.5‰ .5‰ , –20°C , –2 Fy 1 0°C 0.5% –10°C Fy 5. , –5 5‰, –5°C °C –5°C Fy 7.5 –2°C ‰, –5 –20°C °C –10°C –1°C –5°C

100 5 2 10–1 5

Fy/frazil 4.4‰

2 10–2 0

1

2

3

4

5 6 7 Frequency GHz

8

9

Fy0.8‰ 10

11

Figure 9.40 Variation of penetration depth in FYI and MYI with microwave frequencies obtained for different ice temperatures and salinity. Data at 10 GHz are shown for MYI, columnar FYI, and frazil FYI. Data are compiled from the shown studies [Hallikainen and Winebrenner, 1992, Figure 3-7 / with permission from John Wiley & Sons].

decreases for a given temperature. For the C-band (around 5 GHz), the penetration depth in FYI with salinity 7.5‰ is 0.14 m at ice temperature −5 C and 0.38 at temperature −20 C. For the L-band (frequency between 1 and 2 GHz) δp for FYI at −20 C is 0.45 m and 0.9 m for salinity 10.5‰ and 5.1‰, respectively. On the other hand, δp for MYI with salinity 1.3‰ and temperature −9 C is 3.4 m. Salinity of MYI varies between 0 and 2‰ (nearly negligible), hence δp becomes a function of temperature only. The data in Figure 9.40 show that δp from MYI decreases with increasing temperature. This cannot be explained since MYI is almost saline free, hence change of temperature should impact its composition. The only explanation for the decrease of δp in MYI is the possibility of developing wet snow cover at −4 C, with its dielectric constant incorporated in the model used to generate the data. At temperature −10 C and salinity 1.3‰, the penetration depth in MYI is 0.4 m. Figure 9.40 also includes comparison of penetration depth between frazil ice and columnar FYI at 10 GHz. At the temperature −10 C, δp takes values around 0.07 m for frazil ice with 4.4‰ salinity, and 0.16 m for columnar ice with 4.6‰ salinity. Although salinity is nearly the same between frazil and columnar ice in these data, the penetration depth in columnar ice is almost double its value in frazil ice. Frazil ice crystals are much smaller in size than the columnar ice and therefore have more brine pockets. This causes the bulk salinity of frazil ice to be distributed into more brine pockets within interstitial spaces of the smaller ice crystals in frazil ice. Brine is a lossy medium, so more brine pockets result in more loss of the radiation. Moreover, more pockets mean more

scattering at their boundaries, which is another form of losses that further reduce the penetration depth. Due to rapid desalination processes during the early growth phase of sea ice (the first few centimeters thick), δp varies significantly with ice thickness. This was captured in calculations presented in Onstott [1991] using a model developed by Kovacs et al. [1986] to calculate the dielectric constant profile in sea ice as a function of brine volume. Results are presented in Table 9.13 for a few microwave frequencies. For ice less than 2 cm, the penetration depth of the C-band signal exceeds the given thickness. At 16.6 GHz, δp is limited to the top few millimeters. This is the frequency used in the Ku-band of scatterometer systems, and it is not very far from the operational 18–19 GHz used in PM sensors. Snow over sea ice contributes to the penetration depth. While fresh dry snow is transparent at microwave frequencies, wet or metamorphosed snow causes signal loss through absorption and scattering, and hence decreases the penetration depth significantly. Theoretical Table 9.13 Variation of penetration depth of microwave signal in sea ice with thickness in the range from 1 to 8 cm, obtained from model results [Onstott, 1991 / Environmental Research Institute of Michigan]. Ice thickness (cm) 1 2 4 8

Penetration depth in sea ice (cm) 5.25 GHz

9.6 GHz

13.6 GHz

16.6 GHz

>1 2.0 3.3 3.7

0.8 1.0 1.6 1.6

0.5 0.7 1.0 1.0

0.4 0.5 0.8 0.8

406

SEA ICE

calculations presented in Hallikainen and Winebrenner [1992] show that the penetration depth in dry snow varies between 0.6 m at 100 GHz to 6 m at the longer wave of 1 GHz. However, once the snow temperature rises to values near the melting point, the volumetric water contents (VWC) increase and the dielectric loss becomes considerably higher [Hallikainen and Winebrenner, 1992]. This causes significant decrease of δp. Variation of δp with VWC in the snow for a range of microwave frequencies is presented in Hallikainen and Winebrenner [1992] and reproduced here in Figure 9.41. The calculations were based on experimental data presented in Hallikainen, Ulaby, Abdeirazik [1984], Mätzler [1985], and Tiuri et al. [1984]. δp decreases by one order of magnitude as the VWC increases from 1% to 12%. This is mainly caused by increase in the dielectric loss. A larger rate of decrease is found at the transition from dry to slightly wet snow. For example, at 37 GHz the dry snow has penetration depth around 1.1 m (not shown in the figure), but this value decreases to approximately 3.5 cm when VWC reaches 2%. At 18 GHz the penetration depth is roughly 10 cm at 1% VWC and about 0.8 cm at 8% VWC. For VWC > 10%, δp stabilizes at a value around 35 cm for 1 GHz and 40 cm for 37 GHz. Onstott et al. [1987] measured VWC in 45 cm snow cover in the Fram Strait region and found it to be around 5%–6%. The authors also found the penetration depth to be about 100 cm and 5 cm for the L-band (1–2 GHz) and C-band (5–6 GHz), respectively. These values are not very far from the values shown in Figure 9.41. In general, wet snow may cause microwave penetration depth to be confined within the snow depth (depending on the frequency). This is a serious issue that

hampers retrieval of sea ice information from remote sensing observations. It also affects the input of the correct emissivity of snow-covered sea ice in a radiative transfer modeling scheme. If there is 5 cm saline-free snow on top of sea ice, the C-band signal will not “see” the ice if VWC exceeds 6%. Similarly, the 37 GHz emitted radiation will not have contribution from the ice if VWC exceeds 2%. Penetration depth is also affected by the size of snow grain if formed within the snow cover under compactness or thaw-refreeze processes. Bingham and Drinkwater [2000] studied the seasonal and interannual variability of the Antarctic ice sheet using SSM/I radiometers and ERS scatterometer systems. They presented a graph showing variation of penetration depth in dry snow within microwave frequency range between 1 and 40 GHz. The data were produced using a Rayleigh scattering model. Results of spectral variation of δp for three radii of grain size are shown in Figure 9.42. Although the data are for snow on ice sheet, the trend should be applicable to the snow on sea ice. At low frequencies (< 2 GHz), δp is measured in hundreds of meters, regardless of the snow grain radius. This means that observations from the L-band PM radiometer or SAR system are not affected by the snow grain size. However, δp decreases sharply with increasing frequency. The decrease becomes steeper as the grain size increases. Fine-grained snow (recall that all the data in Figure 9.42 are from dry snow) is characterized with small absorption and scattering losses. In this 103

102

102

100 1 GHz

2

4

6

8

10

12

m

0

10–1

m

10–3

m

5 10 18 37

mm

m

10–2

100

0.5

2

2

r=

5

10–1

101

r=

Penetration depth (m)

101

r=

Penetration depth δp, m

Wet snow experimental data

10–2 14

Volumetric water content, %

Figure 9.41 Variation of penetration depth with the percentage of VWC in the snow. Snow density was assumed to be 0.385 g/cm3 [Hallikainen and Winebrenner, 1992, Figure 3–14 / with permission from John Wiley & Sons].

0

5

10

15

20

25

30

35

40

Frequency (GHz)

Figure 9.42 Calculated microwave penetration depth for snow with a density of 0.3 g/cm3 at a temperature of 0 C with different snow grain radii (r) [Bingham and Drinkwater, 2000, Figure 1 / with permission from IEEE].

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 407

case δp can be a few hundred times the wavelength. For coarse-grained snow, the scattering losses become increasingly significant and therefore the penetration depth is significantly reduced. The figure shows the frequencies that are associated with near zero δp for each grain size. Bingham and Drinkwater [2000, p. 1811] stated that “in terms of microwave signature, regions consisting of coarse-grained firn have a higher backscatter coefficient and lower brightness temperature than regions of finegrained firn” ( firn is a German word referred to compacted granular snow. It is the intermediate stage between snow and glacial ice).

9.6. REFERENCES Aldenhoff, W. et al. (2020) First-year and multiyear sea ice incidence angle normalization of dual-polarized Sentinel-1 SAR images in the Beaufort Sea, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 13, pp. 1540–1550. Allison, I., Brandt, R.E. and Warren, S.G. (1993) East Antarctic sea ice: Albedo, thickness distribution, and snow cover, Journal of Geophysical Research, 98(C7), pp. 12,417–12,429. Arcone, S.A., Gow, A.J., McGrew, S. (1986) Structural and dielectric properties at 4.8 and 9.5 GHz of saline ice, Journal of Geophysical Research, 91(C12), pp. 14,281–14,303. Arrigo, K.R., Sullivan, C.W. and Kremer, J.N. (1991) A biooptical model of Antarctic sea ice, Journal of Geophysical Research, 96(C6), pp. 10,581–10,592. Barber, D. (2005) Microwave remote sensing, sea ice and Arctic climate, Physics in Canada, 8, pp. 105–111. Beaven, S.G., Gogineni, S.P. and Shanableh, M. (1994) Radarbackscatter signature of sea ice in the Central Arctic, International Journal of Remote Sensing, 15 (5), pp. 1149–1154. Bindschadler, R. et al. (2005) Detecting and measuring new snow accumulation on ice sheets by satellite remote sensing, Remote Sensing of Environment, 98(4), pp. 388–402. Bingham A.W. and Drinkwater, M.R. (2000) Recent changes in the microwave scattering properties of the Antarctic ice sheet, IEEE Transactions on Geoscience and Remote Sensing, 38(4), Part 2, pp. 1810–1820. Brandt, R.E., Roessler, C.S. and Warren, S.G. (1999) Spectral albedo, absorptance, and transmittance in Antarctic sea ice, Proceedings of 5th Conference on Polar Meteorology and Oceanography, American Meteorological Society, Boston, MA, pp. 456–459. Brandt, R. et al. (2005) Surface albedo of Antarctic sea ice zone, Journal of Climate, 18, pp. 3606–3622. Braun, C.L. and Smirnov, S.N. (1993) Why is water blue?, Journal of Chemical Education, 70(8), p. 612. Cavalieri, D.J. (1994) A microwave technique for mapping thin sea ice, Journal of Geophysical Research, 99(C6), pp. 12,562–12,572. Cavalieri, D.J. et al. (1991) Aircraft active and passive microwave validation of sea ice concentration from the DMSP SSM/I, Journal of Geophysical Research, 96, pp. 21,989–22,008.

Comiso, J.C. et al. (1989) Passive microwave in-situ observations of winter Weddle sea ice, Journal of Geophysical Research, 84(C8), pp. 10,891–10,905. De Abreu, R. (1996) In situ and satellite observations of the visible and infrared albedo of sea ice during spring melt, PhD thesis, Department of Geography, University of Waterloo, Ontario, Canada. Dierking, W. and Pedersen, L.T. (2012) Monitoring sea ice using Envisat ASAR—A new era starting 10 years ago, IEEE International Geoscience and Remote Sensing Symposium 2012, Munich, Germany, pp. 1852–1855. Drinkwater, M.R. (1998) Active microwave remote sensing observations of Weddell Sea ice, In: Jefferies, M.O., ed., Antarctic Sea Ice: Physical Processes, Interactions and Variability, Antarctic Research Series, Vol. 74, American Geophysical Union, pp. 187–212. Drinkwater, M.R., Hosseinmostafa, R., Gogineni, S.P. (1995) C-band backscatter measurements of winter sea ice in the Weddell Sea, Antarctica, International Journal of Remote Sensing, 16, pp. 3365–3389. Eppler, D.T. et al. (1992) Passive microwave signatures of sea ice, In: Carsey, F., ed. Microwave remote sensing of sea ice, Geophysical Monograph Series, Vol. 68, Washington, D.C.: American Geophysical Union, pp. 47–71. Eriksson, L.E.B. et al. (2010) Evaluation of new spaceborne SAR sensors for sea-ice monitoring in the Baltic Sea, Canadian Journal of Remote Sensing, 36(Suppl. 1), pp. S56–S73. ESA (1998) Further achievements of the ERS missions, The Netherlands: ESA Publication SP-1228, European Space Agency Publication Division. Ezraty, R. and Cavanié, A. (1999) Intercomparison of backscatter maps over Arctic sea ice from NSCAT and the ERS scatterometer, Journal of Geophysical Research, 104(C5), pp. 11, 471–11,483. Fung, A.K. and Eom, H.J. (1982) Application of a combined rough surface and volume scattering theory to sea ice and snow backscatter, IEEE Transactions on Geoscience and Remote Sensing, GE-20(4), pp. 528–536. Fuhrhop, R. et al. (1998) A combined radiative transfer model for sea ice, open ocean and atmosphere, Radio Science, 33(2), pp. 303–316. Geldsetzer, T. and Yackel, J.J. (2009) Sea ice type and open water discrimination using dual co-polarized C-band SAR, Canadian Journal of Remote Sensing, 35(1), pp. 73–84. Geldsetzer, T. et al. (2007) Surface-based polarimetric C-band scatterometer for field measurements of sea ice, IEEE Transactions on Geoscience and Remote Sensing, 45(11), pp. 3405–3416. Gill, J.P.S. and Yackel, J.J. (2012) Evaluation of C-band SAR polarimetric parameters for discrimination of first-year sea ice types, Canadian Journal of Remote Sensing, 38(3), pp. 306–323. Grenfell, T.C. (1983) A theoretical model of the optical properties of sea ice in the visible and near infrared, Journal of Geophysical Research, 88(C14), pp. 9723–9735. Grenfell, T.C. (1992) Surface-based passive microwave studies of multiyear ice, Journal of Geophysical Research, 97(C3), pp. 3485–3501. Grenfell, T.C. and Comiso, J.C. (1986) Multifrequency passive microwave observations of first-year sea ice grown in a tank, IEEE Transactions on Geoscience and Remote Sensing, GE-24(6), pp. 826–831.

408

SEA ICE

Grenfell, T.C. and Lohanick, A.L. (1985) Temporal variations of the microwave signatures of sea ice during the late spring and early summer near Mould Bay, Northwest Territories, Journal of Geophysical Research, 90(C3), pp. 5063–5074. Grenfell, T.C. and Maykut, G.A. (1977) The optical properties of ice and snow in the Arctic Basin, Journal of Glaciology, 18(80), pp. 445–463. Grenfell, T.C. and Perovich, D.K. (1994) Analysis of surface based passive microwave observations during LEADEX 1992, Proceedings of Geoscience and Remote Sensing Symposium (IGARSS’94), 2, pp. 1005–1007. Grenfell, T.C. et al. (1988) Multifrequency passive microwave observations of saline ice grown in a tank, Proceedings of the IGARSS’88 Symposium, Edinburgh, UK, pp. 1687–1690. Grenfell, T.C. et al. (1992) Considerations of microwave signature of thin ice, In: Carsey, F., ed. Microwave Remote Sensing of Sea Ice, Geophysical Monograph, Vol. 68, Washington, D.C.: American Geophysical Union. Grenfell, T.C. et al. (1998) Evolution of electromagnetic signatures of sea ice from initial formation to the establishment of thick first-year ice, IEEE Transactions on Geoscience and Remote Sensing, 36(5), pp.1642–1654. Haggerty, J.A. and Curry, J.A. (2001) Variability of sea ice emissivity estimated from airborne passive microwave measurements during FIRE SHEBA, Journal of Geophysical Research, 106(D14), pp. 15,265–15,277. Hallikainen, M.T. (1983) A new low-salinity sea ice model for UHF radiometry, International Journal of Remote Sensing, 4, pp. 655–681. Hallikainen, M. and Winebrenner, D.P. (1992) The physical basis for sea ice remote sensing, In: Carsey, F., ed. Microwave remote sensing of sea ice, Washington, D.C.: American Geophysical Union. Hallikainen, M.T., Toikka, M., Hyyppä, J. (1988) Dielectric properties of low-salinity sea ice, Proceedings of IGARSS’88, pp. 419–420. Hallikainen, M., Ulaby, F.T., Abdeirazik, M. (1984) The dielectric behaviour of snow in the 3 to 37 GHz range, Proceedings of International Geoscience and Remote Sensing Symposium (IGARSS’84, Strasbourg, ESA Scientific & Technical Publications Branch, pp. 169–174. Harlow, R.C. (2011) Sea ice emissivities and effective temperatures at MHS frequencies: An analysis of airborne microwave data measured during two Arctic campaigns, IEEE Transactions on Geoscience and Remote Sensing, 49(4), pp. 1223–1237. Haas, C. (2001) The seasonal cycle of ERS scatterometer signatures over perennial Antarctic sea ice and associated surface ice properties and processes, Annals of Glaciology, 33, pp. 69–73. Heygster, G. et al. (2009) L-band radiometry for sea-ice applications, ESA ESTEC Contract Report 21130/08/NL/EK, Institute of Environmental Physics, University of Bremen, 219 p. Hudson, S.R. (2011) Estimating the global radiative impact of the sea ice–Albedo feedback in the Arctic, Journal of Geophysical Research, 116(D16192), 7 pp. Available from: doi:10.1029/2011JD015804. Huck, P. et al. (2007) Mapping sediment-laden sea ice in the Arctic using AVHRR remote-sensing data: Atmospheric correction and determination of reflectances as a function of ice type

and sediment load, Remote Sensing of Environment, 107, pp. 484–495. Isleifson, D. et al. (2010) C-band polarimetric backscattering signatures of newly formed sea ice during fall freeze-up, IEEE Transactions on Geoscience and Remote Sensing, 48(8), pp. 3256–3267. Jezek, K.C. et al. (1998) A broad spectral, interdisciplinary investigation of the electromagnetic properties of sea ice, IEEE Transactions on Geoscience and Remote Sensing, 36(5), pp. 1633–1642. Johannessen, O. et al. (2007) Remote sensing of sea ice in the Northern Sea route: Studies and applications, Springer-Praxis Books in Geophysical Sciences, ISBN-10: 3-540-24448-4 Johansson, A.M. et al. (2017) X-, C-, and L-band SAR signatures of newly formed sea ice in Arctic leads during winter and spring, Remote Sensing of Environment, 204, pp. 162–180. Kern, S. (2001) A new algorithm to retrieve the sea ice concentration using weather-corrected 85 GHz SSM/I measurements, PhD dissertation, University of Bremen, Germany. Key, J.R. et al. (2001) Estimating the cloudy-sky albedo of sea ice and snow from space, Journal of Geophysical Research, 106(D12), pp. 12,489–12,497. Kovacs, A. et al. (1986) Modelling the electromagnetic trends in sea ice and example impulse radar and frequency domain electromagnetic ice thickness sounding results, Technical Report, CRREL, Hanover, New Hampshire. Kwok, R. et al. (1992) Identification of sea ice types in spaceborne Synthetic Aperture Radar data, Journal of Geophysical Research, 97(C2), pp. 2391–2402. Langlois, A. et al. (2012) Coupling the snow thermodynamic model SNOW-PACK with microwave emission model of layered snowpacks for sub-Arctic and Arctic snow water equivalent retrievals, Water Resources Research, 48 (W12524). Available from: doi:10.1029/2012WR012133. Light, B. et al. (1998) The effect of included particulates in the optical properties of sea ice, Journal of Geophysical Research, 103, pp. 27,739−27,752. Lin, B. and Rossow, W.B. (1997) Precipitation water path and rainfall rate estimates over oceans using special sensor microwave imager and International Satellite Cloud Climatology Project data, Journal of Geophysical Research: Atmospheres, 102(D8), pp. 9359–9374. Available from: doi: 10.1029/ 96JD03987. Lindsay, R.W. and Rothrock, D.A. (1995) Arctic sea ice leads from advanced very high resolution radiometer images, Journal of Geophysical Research, 100, pp. 4533–4544. Liu, H., Guo, H., Zhang, L. (2014) Sea ice classification using dual polarization SAR data, 35th International Symposium on Remote Sensing of Environment, Beijing, China, IPO Conference Series: Earth and Environment Science, Vol. 17, 012115. Lo, R.C. (1983) A comprehensive description of the emission microwave imager SSM/I environmental parameter extraction algorithm, Naval Research Laboratory Memo. Rep. 5199, Washington. Lu, J., Heygster, G., Spreen, G. (2018) Atmospheric correction of sea ice concentration retrieval for 89 GHz AMSR-E observations, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 11(5), pp. 1–16.

RADIOMETRIC AND SCATTERING OBSERVATIONS FROM SEA ICE, WATER, AND SNOW 409 Lubin, D. and Massom, R. (2006) Polar remote sensing: Atmosphere and ocean, Chichester, UK: Praxis Publishing Ltd. Markus, T. and Cavalieri, D.J. (2000) An enhancement of NASA Team sea ice algorithm, IEEE Transactions on Geoscience and Remote Sensing, 38(3), pp. 1387–1396. Markus T. and Dokken, S.T. (2002) Evaluation of late summer passive microwave Arctic sea ice retrievals, IEEE Transactions on Geoscience and Remote Sensing, 40(2), pp. 348–356. Mathew, N., Heygster, G., Melsheimer, C. (2009) Surface emissivity of the Arctic sea ice at AMSR-E frequencies, IEEE Transactions on Geoscience and Remote Sensing, 47(12), pp. 4115–4124. Mathew, N. et al. (2008) Surface emissivity of Arctic sea ice at AMSU window frequencies, IEEE Transactions on Geoscience and Remote Sensing, 46(8), pp. 2298–2306. Mätzler, C. (1985) Interaction of microwaves with natural snow cover, Switzerland: Institute of Applied Physics, University of Bern, p. 148. Mätzler, C. (2005) On the determination of surface emissivity from satellite observations, IEEE Geoscience and Remote Sensing Letters, 2(2), pp. 160–163. Mills, P. and Heygster, G. (2011) Sea ice emissivity modeling at L-band and application to 2007 Pol-Ice campaign field data, IEEE Transactions on Geoscience and Remote Sensing, 49(2), pp. 612–627. Mobley, C.D. et al. (1998) Modeling light propagation in sea ice, IEEE Transactions on Geoscience and Remote Sensing, 36(5), pp. 1743–1749. Mouche, A., Hauser, D. and Kudryavtsev, V. (2006) Radar scattering of the ocean surface and sea-roughness properties: A combined analysis from dual-polarizations airborne radar observations and models in C band, Journal of Geophysical Research, 11(C09004), 18 pp. Available from: doi:10.1029/ 2005JC003166. Nghiem, S.V. et al. (1995) Polarimetric signatures of sea ice 2. Experimental observations, Journal of Geophysical Research, 100(C7), pp. 13,681–13,698. NORSEX Group (1983) Norwegian remote sensing experiment in a marginal ice zone, Science, 220(4599), pp. 781–787. Nürnberg, D. et al. (1994) Sediments in Arctic sea ice— Implications for entrainment, transport and release, Marine Geology, 119, pp. 185−214. Onstott, R.G. (1979) Surface-based scatterometer results of Arctic sea ice, IEEE Transactions on Geoscience and Remote Sensing, GE-17(3), pp. 78–85. Onstott, R.G. (1991) Active microwave observations of the formation process of simulated sea ice, Technical Report 2395002-T, Ann Arbor, Michigan: Environmental Research Institute of Michigan. Onstott, R.G. (1992) SAR and scatterometer signatures of sea ice, In: Carsey, F., ed. Microwave remote sensing of sea ice, Geophysical Monograph, Vol. 68, Washington, D.C.: American Geophysical Union, pp. 73–104. Onstott, R.G. and Gogineni, S.P. (1985) Active microwave measurements of Arctic sea ice under summer conditions, Journal of Geophysical Research, 90(C3), pp. 5035–5044. Onstott R.G. and Shuchman, R.A. (2004) SAR measurements of sea ice, In: Jackson, C.R. and Apel, J.R., eds., Synthetic Aperture Radar. Marine User’s Manual, National Oceanic

and Atmospheric Administration, Washington D.C.: US Department of Commerce, pp. 81–115. Onstott, R.G. et al. (1987) Evolution of microwave sea ice signatures during early summer and midsummer in the marginal ice zone, Journal of Geophysical Research, 92(C7), pp. 6825–6835. Parkinson, C.L. et al. (1987) Arctic sea ice: 1973–1976, NASA Special Report SP-489, Washington D.C., 295 p. Pedersen, C.A. (2007) Optical properties of snow and sea ice field measurements, parameterization schemes and validation, PhD dissertation, University of Tromso, Norway. Perovich, D.K. (1996) The optical properties of sea ice, Technical Report 96-1. US Army Cold Regions Research and Engineering Laboratory (CRREL), Hanover, NH. Perovich, D.K., Maykut, G.A. and Grenfell, T.C. (1986) Optical properties of ice and snow in the polar regions. 1: Observations, Proceedings of SPIE Ocean Optics, VIII(637), pp. 232–241. Pistone, K., Eisenman, I. and Ramanathan, V. (2014) Observational determination of albedo decrease caused by vanishing Arctic sea ice, PNAS, 111(9), pp. 3322–3326. Pohl, C. et al. (2020) Broadband albedo of Arctic sea ice from MERIS optical data, The Cryosphere, 14, pp. 165–182. Rampal, R. et al. (2011) IPCC climate models do not capture Arctic sea ice drift acceleration: Consequences in terms of projected sea ice thinning and decline, Journal of Geophysical Research, 116, 17 pp. Available from: doi:10.1029/ 2011JC007110. Rubinstein, I.G., Nazarenko, D.M. and Tam, S. (1994) Passive microwave systems, In: Haykin, S. et al., ed. Remote sensing of sea ice and icebergs, New York: Wiley, p. 177. Sandven, S. et al. (1999) Barents Sea seasonal ice zone features and processes from ERS 1 synthetic aperture radar: Seasonal Ice Zone Experiment 1992, Journal of Geophysical Research, 104(C7), pp. 15,843–15,857. Sandven, S. et al. (2012) Sea ice classification using RADARSAT-2 dual polarisation data, SEASAR 2012, The Fourth International Workshop on Advances in SAR Oceanography, June 2012, Tromso, Norway. Seong, N. et al. (2022) Evaluation of sea ice radiative forcing according to surface albedo and skin temperature over Arctic from 1982–2015, Remote Sensing, 14(11), p. 2512. Available from: https://doi.org/10.3390/rs14112512. Shokr, M. (2009) Compilation of radar backscatter database of sea ice types and open water using operational analysis of heterogeneous ice regimes, Canadian Journal of Remote Sensing, 35(4), pp. 369–384. Shokr, M. and Barber, D.G. (1994) Temporal evolution of physical and dielectric properties of sea ice and snow during the early melt season: Observations from SIMS’90 experiment, Journal of Glaciology, 40(134), pp. 16–30. Shokr, M. and Dabboor, M. (2020) Observations of SAR polarimetric parameters of lake and fast sea ice during the early growth phase, Remote Sensing of Environment, 247, 111910. Shokr, M. and Kaleschke, L. (2012) Impact of surface conditions on thin sea ice concentration estimate from passive microwave observations, Remote Sensing of Environment, 121, pp. 36–50.

410

SEA ICE

Shokr, M., Asmus, K. and Agnew, T.A. (2009) Microwave emission observations from artificial thin sea ice: The ice-tank experiment, IEEE Transactions on Geoscience and Remote Sensing, 47(1), pp. 325–338. Shokr, M.E., Lambe, A., Agnew, T. (2008) A new algorithm (ECICE) to estimate ice concentration from remote sensing observations: An application to 85 GHz passive microwave data, IEEE Transactions on Geoscience and Remote Sensing, 46(12), pp. 4104–4121. Shokr, M. et al. (2022) Observations from C-band SAR fullypolarimetric parameters of mobile sea ice based on radar scattering mechanisms to support operational sea ice monitoring, Canadian Journal of Remote Sensing, 48(2), pp. 197–213. Spreen, G., Kaleschke, L. and Heygster, G.H. (2008) Sea ice remote sensing using AMSAR-E 89 GHz channels, Journal of Geophysical Research, 113(C02S03). Available from: doi:10.1029/2005JC003384. Swift, C.T. et al. (1992) Laboratory investigations of the electromagnetic properties of artificial sea ice, In: Carsey, F., ed. Microwave remote sensing of sea ice, Geophysical Monograph, Vol. 68, Chapter 9, Washington, D.C.: American Geophysical Union. Stogryn, A. and Desargant, G.J. (1985) The dielectric properties of brine in sea ice at microwave frequencies, IEEE Transactions on Antennas and Propagation, AP-33, pp. 523–532. Taskjelle, T. et al. (2016) Spectral albedo and transmittance of thin young Arctic sea ice, Journal of Geophysical Research: Oceans, 121, pp. 540–553. Available from: https://doi.org/ 10.1002/2015JC011254. Thackeray, C.W. and Hall, A. (2019) An emergent constraint of future Arctic sea ice albedo feedback, Nature Climate Change, 9, pp. 972–978. Tiuri, M. et al. (1984) The complex dielectric constant of snow at microwave frequencies, IEEE Journal of Oceanic Engineering, 9, pp. 377–382. Tonboe, R., Andersen, A. and Toudal, L. (2003) Anomalous winter sea ice backscatter and brightness temperature, Danish Meteorological Institute, Scientific Report 03-13, Copenhagen, ISBN-10: 8-774-78489-7. Tschudi, M.A., Maslanik, J.A. and Perovich, D.K. (2008) Derivation ofmelt pondcoverageonArcticsea ice using MODIS observations, Remote Sensing of Environment, 112, pp. 2605–2614. Tucker III, W.B. et al. (1999) Physical characteristics of summer sea ice across the Arctic Ocean, Journal of Geophysical Research: Oceans, 104, pp. 1489–1504. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1981) Microwave remote sensing: Active and passive’, Vol. I, Microwave remote sensing fundamentals and radiometry, Norwood, MA: Artech House Inc., p. 225. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1982) Microwave remote sensing: Active and passive, Vol. II, Radar remote sensing and surface scattering and emission theory, Norwood, MA: Artech House Inc., p. 847.

Ulaby, F.T., Moore, R.K. and Fung, A.K. (1986) Microwave remote sensing: Active and passive, Vol. III, From theory to applications, Norwood, MA: Artech House Inc., p. 2058. Vachon, P.W. and Wolfe, J. (2011) C-band cross polarization wind speed retrieval, IEEE Geoscience and Remote Sensing Letters, 8(3), pp. 456–459. Vant, M.R., Ramseier, R.O. and Makros, V. (1978) The complex dielectric constant of sea ice at frequencies in the range 0.1–40 GHz, Journal of Applied Physics, 49(3), pp. 1264–1280. Vant, M.R. et al. (1974) Dielectric properties of fresh and sea ice at 10 and 35 GHz, Journal of Applied Physics, 45, (Chapter 2), pp. 4712–4717. Voronovich, A.G. and Zavorotny, V.U. (2011) Depolarization of microwave backscattering from a rough sea surface: Modeling with small-slope approximation, Proceedings of International Geoscience and Remote Sensing Symposium (IGARSS), pp. 2033–2036. Wang, L. et al. (2011) Integrated pan-Arctic melt onset detection from satellite active and passive microwave measurements, 2000–2009, Journal of Geophysical Research, 116 (D22103). Warren, S.G. (2019) Optical properties of ice and snow, Philosophical Transactions of the Royal Society, 277, 20180161. Warren, S.G., Roesler, C.S. and Brandt, R.E. (1997) Solar radiation processes in the East Antarctic sea ice zone, Antarctic Journal of the United States, 32(5), pp. 185–187. Wensnahan, M., Maykut, G.A. and Grenfell, T.C. (1993) Passive microwave remote sensing of thin sea ice using principle component analysis, Journal of Geophysical Research, 98, pp. 12,453–12,468. Wigneron, P. et al. (2006) Surface emission, In: Mätzler et al., eds. Thermal microwave radiation: applications for remote sensing, IET Electromagnetic Waves Series, Vol. 52, London, UK: The Institute of Engineering and Technology. Willmes, S., Nicolais, M. and Haas, C. (2013) The microwave emissivity variability of snow covered first-year sea ice from late winter to early summer: A model study, The Cryosphere, 7, pp. 5711–5734. Wiscombe, W.J. and Warren, S.G. (1980) A model for spectral albedo of snow, I: Pure snow, Journal of the Atmospheric Sciences, 37(12), pp. 2712–2733. Yang, H. and Weng, F. (2011) Error sources in remote sensing of microwave land surface emissivity, IEEE Transactions on Geoscience and Remote Sensing, 49(9), pp. 3437–3442. Zakhvatkina, N. et al. (2017) Operational algorithm for ice–water classification on dual-polarized RADARSAT-2 images, The Cryosphere, 11, pp. 33–46. Zhang, B. and Perrie, W. (2012) Cross-polarized synthetic aperture radar: A new potential technique for hurricanes, Bulletin of the American Meteorological Society, 93, pp. 531–541. Zwally, H.J. et al. (1983) Antarctic sea ice, 1973–1976: Satellite passive-microwave observations, NASA SP-459, Washington, D.C.: National Aeronautics and Space Administration, 206 pp.

10 Retrieval of Sea Ice Surface Information

10.1

10.2

Mechanically Generated Surface Deformation..................... 412 10.1.1 Rafted Ice ................................................................ 412 10.1.2 Ridged, Rubble, and Brash Ice................................ 413 10.1.3 Kinematic Processes: Convergence, Divergence, Shear, and Vorticity ............................. 417 10.1.4 Cracks and Leads .................................................... 421 Thermally Induced Surface Features .................................... 428 10.2.1 Surface Melt ............................................................ 428 10.2.1.1 Optical Observations ............................... 428

Sea ice surface features can be mechanically generated, thermally induced, or meteorologically driven. Mechanically generated surface deformation occurs when ice floes collide or ice sheets converge or diverge. Under compression, deformation is manifested in the form of rafted ice, pressure ridges, rubble fields, or brash ice. In this case, the surface is referred to as being deformed as opposed to leveled surface. On the other hand, under tension (divergence), cracks and leads are formed. Thermally induced surface features are manifested in the forms of surface melt, melt ponds, frost flowers, re-frozen leads in winter and rotten ice in spring and summer. Rotten ice is sea ice in an advanced state of disintegration, which usually takes a honey-combed shape. Meteorologically driven features are demonstrated in the form of snow cover or any other precipitation at the surface. Polynyas may be included under this category because they are driven by wind or upwelling warm ocean current (section 2.9.1). Remote sensing data can be used to identify most of these features. This is the subject of this chapter. Use of remote sensing for retrieval of sea ice geophysical parameters (concentration, thickness, temperature, etc.) is addressed in the next chapter. While identification of some surface features in SAR image can be an involved

10.2.1.2 Passive Microwave Observations ............. 432 10.2.1.3 Active Microwave Observations .............. 434 10.2.1.4 Airborne Photography............................. 437 10.2.2 Frost Flowers .......................................................... 438 10.3

10.4

Meteorologically Driven Surface Features............................ 442 10.3.1 Polynya Identification and Properties...................... 442 10.3.2 Snow Depth ............................................................. 444 References............................................................................. 448

process, simple threshold methods can do the job. For example, Dierking [2013] used airborne SAR data to set thresholds that separate deformed from leveled ice. In principle, it is feasible to use SAR for discriminating deformed from leveled ice because the former has remarkably higher backscatter. However, identification of other deformation aspects requires more work. For example, Bouillon and Rampal [2015] developed a method to produce sea ice deformation data set based on SAR-derived sea ice motion. This chapter covers methods of identification and characterization of a few surface features using optical, thermal, and radar remote sensing techniques. The presentation is grouped into three parts. The mechanically generated features include rafting, ridging, kinematic features, cracks, and leads. The thermally induced features include surface melt and frost flowers. Polynya and snow depth are included in the third part of meteorologically driven surface features. It should be noted that other studies use different categories of ice surface features. For example, Wright and Polashenski [2018] developed a method to classify surface from using optical imagery into three main categories: bare and snow-covered ice, melt ponds and submerged ice, and open water (OW). The authors present an open-source code for this classification.

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 411

412

SEA ICE

10.1. MECHANICALLY GENERATED SURFACE DEFORMATION This section starts by addressing rafting and ridging with brief information on brash and rubble ice. All these forms of surface deformation play an important role in determining the aerodynamic coupling between the ice and the atmosphere as well as the hydrodynamic coupling between the ice and the underlying ocean water. Ridges and rubble ice are also important features from climatic point of view because they are associated with large thickness that affects the heat release from the ocean and influence the total ice volume balance in the Arctic Basin. The section then proceeds with presentations on kinematic deformation processes (Convergence, divergence, shear and vorticity of the ice cover) and concludes with brief information on cracks and leads. 10.1.1. Rafted Ice If two thin ice sheets/floes (< 30 cm thick) collide, rafting is likely to occur (Figure 2.36). Rafted ice is visually identified in nature and can be visually identified in SAR images as well. In fact, this is the key feature used by ice analysts in ice monitoring centers and operators on ships to identify gray and gray-white ice types (section 2.8). In addition to its apparent shape in SAR images that mimics the shape in nature, rafting can be associated with very high backscatter in SAR image. The high backscatter may not be limited to the raised (and rough) border of the rafted pieces in the image but may extend across the entire area of the rafted ice piece. This point is presented here with detailed description of the radar scattering mechanism that triggers the extensive

high backscatter. It is worth noting that SAR is the only remote sensing tool that can identify rafting though it can only be identified through visual image analysis. No research has been pursued to develop a quantitative method for rafting identification in SAR images perhaps because it does not have impact on marine activities or climate studies. Physically speaking, the only significant aspect of rafting is the thickening of the ice cover. It is known that rafted ice can be identified by the raised edge of the top piece, which is visible in nature and SAR images. Yet, rafted ice can have remarkably high backscatter (around –4.0 dB) in SAR images as shown in Figure 10.1. The image was acquired by RADARSAT2 FP fine-beam mode, and the bright rafted ice is identified as such in the image analysis by CIS operators. A first attempt to explain such high backscatter is presented in Shokr et al. [2022] based on the power from each one of the three scattering mechanisms: SB, DB, and MB (see section 7.6.1.3). These are denoted as ps, pd, and pv [see equation (7.95) and the following text]. The high total backscatter (SPAN) from rafted ice may include high component of the DB scattering mechanism (pd). This is shown in the right panel of Figure 10.1, where pd is visibly high (–17 dB to –20 dB). Therefore, rafted ice can be identified using the high total backscatter power in conjunction with its high pd component (the latter is unique to the rafted ice). A physical explanation of the high SPAN and pd from rafted ice is presented in the following. Yet, it should be noted that analysts of SAR images usually check the wind history to confirm the rafting process. The sketch in Figure 10.2 shows two ice pieces, one overrides the other. This configuration is used to explain the high SPAN with its high components of ps and pd. The layer between the two pieces is most crucial for

Wi nd

Grey ice Th

in

Gr

iffi

th

Isl

ice

Cornwallis Island

an

d

SPAN (total power)

Double-bounce power

Figure 10.1 Total backscatter power (SPAN) image of an area in Lancaster Sound, Canadian Arctic acquired from RADARSAT-2 Quad-Pol on 14 November 2017 (left panel), showing a large area of rafted ice marked by the dotted ellipses. The DB power image (pd) is presented in the right panel and shows high pd from one rafted piece (courtesy of the CIS).

Surface scattering (SB)

D S c oub at l e te -b r i n ou g nc (D e B)

al

In

ci

de

nt

si

gn

tte r ca

ca

Ba

ck s Ba

ck s

tte r

RETRIEVAL OF SEA ICE SURFACE INFORMATION 413

2 1

sea water

Figure 10.2 Radar scattering mechanism from rafted ice. The top piece (2) has lost its dendritic configuration at its bottom surface when it slides over the bottom piece and the interface between the two pieces is filled with saline water or ice. Two scattering mechanisms are shown; SB scattering off both top and bottom surfaces of the rafted piece (2), and DB off the dihedral shape of the raised edge of piece (2).

explanation of radar scattering mechanisms in this case. Originally, the top piece (#2 in the figure) has dendritic surface at the bottom, but the dendrites are removed when this piece slides over piece #1. This leaves the layer between the two pieces filled with sea water or highly saline ice upon freezing. In either case, it constitutes a strong dielectric mismatch. Bailey, Feltham, Sammonds [2010] presented a one-dimensional thermal consolidation model for rafted ice. They proposed that it is composed of two or three layers separated by thin layers of ocean water, which freeze within 15 hours. Based on this configuration, the SB scattering mechanism component becomes active not only at the surface of the top piece but also at the highly reflective layer of the dielectric mismatch at the interface between the two ice pieces. Moreover, a DB component may be triggered by the raised edges of the rafted pieces as shown in Figure 10.2. Both components add to the high backscatter shown in the left panel in Figure 10.1. More data are needed to confirm the high backscatter with its high-DB scattering component, hence the validity of the above explanation. In-situ observations or measurements coincident with SAR acquisition in FP mode would be sufficient to achieve this purpose. 10.1.2. Ridged, Rubble, and Brash Ice Ridged, rubble, and brash ice result when compression stress raptures the ice. When two ice floes collide, a relatively mild impact may limit the damage to the edges, causing formation of raised edges by a few tens of centimeters (Figure 2.40). Heavy impact of thick ice floes (>30 cm thick) causes formation of ridges (Figure 2.42). If the impact of the crushing extends deep into the ice

floe/pack, then a rubble ice field is formed (Figures 2.47 and 2.48). These appear as extensive distribution of highly distorted (crushed) ice blocks with limited vertical accumulation. Rubble fields do not have identifiable structure as ridges and their keels are not deep. Brash ice is defined as accumulation of fragments of ice pieces. It is commonly found in thin ice between ice floes when compressed. Ridges are accumulations of ice blocks of tens of centimeters or above one meter thick (section 2.6.2). They exist in nature as either pressure or shear ridge. A pressure ridge may form when two ice sheets are pushed against each other (Figure 2.39 and 2.42). This type of ridge may exist anywhere within a dynamic pack ice, especially in marginal ice zones (MIZ). Shear ridges, on the other hand, may form when the two ice sheets move parallel to each other but with differential speed or when one sheet moves contiguous to stationary (fast) ice. This is commonly found at the offshore edge of landfast ice. Pressure ridges are usually thicker than shear ridges, but both are possible navigational hazard. Unlike rubble fields, ridges increase the mechanical load on the ship’s hull. Recall that the keel of a ridge is about four times larger than its sail (Figure 2.44). Therefore, when a ship encounters a ridge of typical 2 m sail height, it has to break through a keel of 8 m. SAR is the most suitable tool for detecting ridges, rubble, and brash ice. For one thing, SAR is sensitive to surface roughness of rubble, brash, and deformed ice as those features generate high backscatter yet with different dominant scattering mechanisms. Moreover, the topographic structure of the ridges may trigger particularly high power from the MB and DB mechanisms (section 7.6.1.3). The total backscatter (SPAN) from all those ice surface features is high, and this is the first parameter that should be checked to visually identify them in SAR images, followed by the power from each mechanism to pin down the right type. The shape is another parameter to be checked. Ridges usually appear as bright linear features in the image, while rubble and brash ice are also bright features but extend over area, not lines. An example of brash ice in SAR image is shown in Figure 10.3. This is an image of highly mobile ice in a section of the Beaufort Sea. Brash ice appears as bright patches of random shapes entrapped between dark (smooth) ice floes. This is usually the case as the water between floes tends to freeze, but the collision of the floes breaks the new ice into small fragments, which become mixed with water. While backscatter from brash ice is high, it is not a unique identifiable feature of this ice surface. As far as the authors of this book know, there is no SAR parameter that can uniquely identify brash ice. Contextual information in the image must be used. Brash ice usually appears between floes as small bright patches with

414

SEA ICE

Figure 10.3 RADARSAT-1 image of an area in the Beaufort Sea north of Alaska showing smooth first-year ice (FYI) floes (dark areas) with surrounding rubble and brash ice (bright areas). Brash ice is found between ice floes (courtesy of the CIS).

no identifiable geometry. It is caused by crushing thin ice between floes while in motion when crushed sea waves. The gray objects in Figure 10.3 are probably rough ice with rubble surface. Other features that are worth commenting on are the bright linear boundaries around the floes, which represent raised edges. Note also the bright linear features within the dark ice floes. These are edges of smaller floes, which amalgamate to compose the larger conglomerated floe. Ridges can be identified in SAR images based on their high backscatter pixels, but this should be formed in linear or curvilinear shape. Ridge identification can be achieved in SAR images using one of the following three approaches: visual image interpretation, application of an edge detection technique, or mapping a measure of convergence of ice sheets in sequential images. In the last approach, pairs of SAR images acquired within a short period (up to 3 days) should be used. Details about using the three approaches are given below. A difficulty may arise in quantifying the area of the ridge if its width is smaller than the pixel spacing in the SAR image (typically 100 m in the resampled ScanSAR images used in ice operational sea ice monitoring). If only part of the pixel is covered with a ridge, the entire pixel will appear as a ridge pixel. Therefore, areal information of the ridge cannot be accurately extracted using SAR images unless a fineresolution mode (5–25 m) is used. However, the image swath is always narrow in these modes. Visual identification of ridges in SAR images is more successful if SAR parameters (frequency, polarization, and incidence angle) and its viewing geometry are suitable

to enhance the visibility of the ridge. As for the viewing geometry (look angle), the ridge becomes more visible if its orientation is parallel (or nearly parallel) to satellite track. If the orientation of the ridge changes along its length, or it is surrounded by rubble ice it becomes difficult to detect. Effects of the radar frequency and incidence angle on edge detection are discussed below. As a rule of thumb, larger incidence angles (i.e., shallow viewing angles) are preferred for ridge detection. At these angles, the backscatter from the surface roughness is reduced and this accentuates the ridge because of its higher backscatter from the SB and possibly an additional component of power from the DB scattering. Melling [1998] concluded that the ERS-1 C-band operating at a small (steep) incidence angle (around 25 ) is only of marginal use for ridge detection, whereas an airborne X-band SAR with a larger incidence angle (between 40 and 70 ) enables unambiguous discrimination of ridges. As for radar frequency, the longer wavelength L-band (23 cm) delineates ridges better than the C-band (5.4 cm). Dierking and Busche (2006) compared coincident images from the L-band JERS-1 and the C-band ERS-1, and concluded that although the C-band is regarded as a reasonable choice for all-season ice monitoring capabilities, the L-band is superior for the specific task of mapping the surface deformation features. The comparison is presented in Figure 10.4. In comparing the two images it should be noted that C-band VV polarization of ERS-1 acquired images at relatively small incidence angle (23 ) compared to the L-band JRES-1 with its HH polarization and relatively large incidence angle (32 –38 ). Figure 10.4 demonstrates the better capability of the C-band (5.4 cm wavelength) in delineating details in zone (A), which is covered by NI types. However, the advantage of the L-band (22 cm wavelength) is apparent in delineating the brash ice (zone C) and the shear ridge (D). The better contrast between the brash ice and the surrounding FYI floes in the L-band image is because the surface of the latter becomes smooth compared to the long wavelength of the L-band. Hence, the FYI floes appear dark. Multi-year ice (MYI) floes appear darker in the L-band (zone B). This is because the wavelength of the L-band is much larger than the few millimeters scale of the air bubbles in MYI, hence volume scattering is reduced. Arkett et al. [2008] compared the L-band ALOSPALSAR images of different sea ice scenes in the Arctic and the east coast of Canada against near coincident RADARSAT-1 (C-band) ScanSAR Wide images. Figure 10.5 shows the comparison of a scene from Baffin Bay, acquired on 13 January 2008, by the two sensors with 5 hours difference between the satellite acquisitions. Both images were from ScanSAR mode, but the viewing angle of each segment in the image is unknown. The RADARSAT-1 image analysis from CIS identifies the ice field in this area as a mixture of thick FYI and

RETRIEVAL OF SEA ICE SURFACE INFORMATION 415

B

Shear line

B

C

C D

D

A

A

C-band ERS-1

L-band JERS-1

Figure 10.4 Comparison between C-band ERS-1 (VV) and L-band JERS-1 (HH) of a sea ice scene near the east coast of Greenland acquired on 7 January 1994. The width of the images is about 50 km and the pixel spacing 100 m. A, B, C, and D refer to new ice, an MYI floe, brash ice, and a shear zone, respectively. Note the visibility of the shear ridge in the L-band image. ERS-1 image © ESA, JERS-1 image © JAXA [Dierking and Busche, 2006 / with permission from IEEE].

Radarsat-1

ALOS-PALSAR

MYI

FYI

Figure 10.5 Sea ice scene from Baffin Bay, Canada, acquired on 13 January 2008, with the C-band HH RADARSAT-1 and the L-Band HH PALSAR, showing good contrast between FYI and MYI in the C-band and the excellent delineation of surface features in the L-band [Arkett et al., 2008 / from IEEE].

MYI. Once again, the ridging information and surface deformation in the L-band data are readily visible compared to the C-band image. The bright linear features in the MYI area, distinctively visible in the L-band image, are hummocks (former ridges). Incidentally, those

features are not visible at all in X-band SAR or the coarse-resolution Ku-scatterometer. Figure 10.5 also confirms the utilization of the L-band accentuating surface features of FYI as appears in the lower left area on the image.

416

SEA ICE

To continue with the visual capability to identify ridges in SAR image, it is worth introducing the potential use of the power from the DB scattering mechanism (pd). As shown in Table 9.8, pd is highest from ridges (data in this table are based on the cited study only). Figure 10.6 shows a segment of RADARSAT-2 Quad-pol image over Lancaster Sound, Canadian Arctic, with resolution around 10 m. The ridge is much more visible in the pd image with values around -19 dB, which contrasts well with the much smaller values from the surrounding ice. All ice types have lower pd except (perhaps) some spots of heavily deformed ice around the ridge. The pd image can be used for visual or quantitative identification of ridges using a simple threshold. It is more reliable than using the backscatter image because pd masks all ice around the ridge. Quantitative edge detection methods delineate ridge pixels based on the calculated spatial derivatives of the gray tone in the image. Potential edge pixels are identified based on the maximum of the first derivative and the zerocrossing of the second derivative. However, it should be noted that those pixels may not uniquely identify ridges. If they are spread over an area, they probably indicate rubble or brash ice. That is why the shape of the potential edge pixels must be employed as a complementary criterion. With these two criteria, the edge detection task usually encompasses four components: (1) a gradient operator to calculate the derivative at each pixel using an appropriate window size, (2) a threshold on the gradient to identify potential edge pixels, (3) a spatial filter to eliminate noisy (isolated) pixels, and (4) a method to connect the potential edge pixels after excluding the noise. Only pixels that form into a linear geometrical order are considered to compose the ridge while other pixels are just spurious noise. The last component can be

SPAN image

achieved using a Fast Fourier Transform (FFT) technique accompanied with a high-pass filter. Since only relative difference between pixel values is required, calibrated backscatter is not needed. Ridge identification in remote sensing images is mainly an image processing endeavor. It belongs to the computer science field more than sea ice physics field. One of the first studies that addressed this subject was conducted by Vesecky, Smoth, Samadani [1990] using the L-band SAR onboard Seasat satellite (see section 8.1). It uses what has become a traditional approach of image processing to identify and characterize ridges in radar imagery. It combines a threshold to identify bright pixels in the image and a line detection operator using a set of masks that are convolved with a 5 × 5 pixel window. This produces linear segments that can potentially form a ridge. The rest of the work involves examination of the connectivity of those segments to decide if they structurally make a ridge. As for ridges located within an ice floe, Zhou and Li [2000] combined the backscatter threshold with a supervised classification to identify ridges and brash ice in SAR images. As these two objects have equally high backscatter, they can be distinguished using an object delineation method to demarcate the ice floe boundaries. The high backscattering pixels located within an ice floe can then be uniquely identified as ridge pixels, while those located at the edges or between floes can be labeled as brash ice. Ridges can also be identified in aerial photography. Miao et al. [2016] used shadows on photographs to detect ridges in high-resolution aerial photos of MIZ near Barrow, Alaska. The authors used object-based classification scheme in a batch processing method to retrieve ridge attributes. Machine learning has become a popular tool in digital image recognition [Emmert-Streib et al., 2020].

DB (pd) image

Figure 10.6 A segment of RADARSAT-2 Quad-pol image over Resolute Passage acquired on 20 November 2017, showing ridge in the middle of the image. The SPAN image (left) is showing the ridge against the background ice, but the DB scattering power image (right) masks the background ice and keeps only the ridge.

RETRIEVAL OF SEA ICE SURFACE INFORMATION 417

It builds models based on training samples and makes decisions or predictions without a need to develop conventional algorithms to identify the objects or the desired classes in the imagery data. Identification of sea ice ridges using ASAR data along with other Earth observation sensors (Cryosat-2, COSMOS-Skymed, Landsat, and QuickBird) is presented in a report published by the Centre for Cold Ocean Resources Engineering, C-CORE [2012]. The report aimed at analyzing ridge-like features and correlated them to in-situ data in order to develop a methodology to routinely analyze high-resolution images for the extraction of extreme ridge feature. This work has been motivated by a growing interest in exploiting the oil and gas resources in the Arctic. Over 25% of the world’s petroleum reserves are believed to be in the Arctic region and other ice-rich marine environments. C-Core is a Canadian multidisciplinary R&D corporation established to address challenges faced in oil and gas development offshore the east coast of Canada and the Arctic. In the third approach, ridges can be identified based on maps of sea ice deformation. This is part of products from the RADARSAT Geophysical Processing System (RGPS), which is briefly described in the next section. The maps are gridded in polar stereographic projection and show spatial distribution of quantified ice deformation in the western Arctic section. Ridges are part of what is known as “linear kinematic feature” (LKF), which vary in length from a few kilometers to hundreds of kilometers. Linow and Dierking [2017] used the ice deformation map product to delineate ridges following an object-based detection approach.

10.1.3. Kinematic Processes: Convergence, Divergence, Shear, and Vorticity Forms of ice surface deformation are triggered by kinematic surface processes, depicting convergence, divergence, shear, and vorticity. All these processes are instigated by ice motion, which can be estimated from pairs of SAR images separated by 1–3 days. A pioneering work to identify the aforementioned kinematic surface processes has been implemented in the RGPS of Alaska Satellite Facility (ASF) [Kwok et al., 1998] and further developed in Kwok (2003). The products are offered within the NASA’s project Making Earth System Data Records for Use in Research Environments (MEaSUREs). They encompass a wide range of geophysical parameters, including cryospheric parameters. Sea ice parameters include motion maps sampled in Eulerian mode and surface deformation parameters, based on the kinematic features, sampled in Lagrangian mode. The data have contributed significantly to the development of new approaches for modeling the

mechanical behavior of sea ice and description of the seasonal and regional variability of sea ice deformation. To identify the evolving deformation of the sea ice surface, a Lagrangian representation is generated from sequential SAR images. In this representation, a uniform grid is established and the displacement of grid points between successive images are calculated [Lindsay and Stern, 2003, Herman and Glowacki, 2012]. Next, a deformation field at a spatial scale less than the grid spacing can be obtained. This approach has been implemented in MEaSUREs using sequential RADARSAT-1 (1995–2008) and Envisat ASAR (2008–2012). Formulation to calculate the deformation parameters is shown below. The system tracks 10 km × 10 km cells typically at 3-day intervals during an entire ice growth season. An example of deformation maps extracted from an animation provided by R. Kwok of JPL is shown in Figure 10.7. It shows the tracking of the grid points between sequential pairs of RADARSAT-1 images in a section of the Beaufort Sea. The dates and locations are not shown since the figure aims at showing the concept of distorted cells from one date to the next. Another example showing maps of the three kinematic parameters of surface processes (divergence/convergence, vorticity, and shear), along with the ice motion vector, is presented in Figure 10.8. These parameters are calculated from the velocity at the corner points of each cell using equations (10.1) to (10.3). The kinematic parameters are observed in areas away from the fast ice. Sea ice deformation is usually localized in high strain zones that may extend hundreds of kilometers, especially in the western section of the Arctic. These are also known as failure zones. Since ice is heterogeneous at the grid scale, such failure zones may intersect and propagate across the region, depending upon the boundary conditions and the applied wind stress. Hutchings, Heil, Hilber [2005] modeled these failure zones with a viscous-plastic sea ice model, using an isotropy rheology. The rheological model describes plastic failure of the sea ice. This work demonstrates the benefit of combining deformationdriven parameters from remote sensing with models that describe mechanical behavior of sea ice. Nevertheless, remote-sensing-based kinematics deformation parameters have not been used to their full potential to support studies in the field of sea ice mechanics. As the vertices of each cell move within a time step, the three kinematic properties: namely divergence/convergence rate (Ed), the shear rate (Es), and the vorticity (vrt) occurring within each cell can be calculated to characterize the response of the ice cover to stresses induced by wind and ocean currents. The strain rate vectors are shown in Figure 10.9. Following Rothrock [1986], the three parameters are defined as follows: Ed =

∂u ∂v + ∂x ∂y

(10.1)

418

SEA ICE

1

2

3

4

5

6

7

8

9

Figure 10.7 Lagrangian representation of ice deformation in a section of the Beaufort Sea. Successive frames (1–9) are 3-days apart, selected from an animation provided by Ron Kwok of JPL. The initial grid box is 50 km wide (images courtesy of Ron Kwok of JPL, RADARSAT images © MDA Ltd, Canada).

Es =

vrt =

∂u ∂v − ∂x ∂y

∂u ∂v + ∂y ∂x

2

+

∂u ∂v + ∂y ∂x

E t = E 2d + E 2s

(10.2) 2

12

(10.4)

and the azimuth of the strain rate is defined as:

12

(10.3)

where, u and v denote the velocity components along the x and y axes of the grid points, respectively. A positive value of Ed indicates a divergent motion (possibly generates openings in the ice cover), and a negative value indicates a convergent motion (possibly generates pressure ridges or rafting). In the differential operators, u and v can be regarded as the displacements assuming a unit interval between the times of acquisition of the image pairs, therefore ∂u and ∂v are calculated by comparisons of the displacements at the two ends of the x and y sides, respectively. The total deformation (strain) rate Et is defined as:

θ = tan− 1

Es Et

(10.5)

Values of θ = 0, π/2, and π correspond to pure divergence, pure shear, and pure convergence, respectively [Feltham, 2008]. The strain rate components are computed from approximations of the line integral around the boundary of each cell [Lindsay and Stern, 2003]: ∂u 1 = u dy ∂x A

1 n ui + 1 + ui yi + 1 − yi 2A i = 1

(10.6)

RETRIEVAL OF SEA ICE SURFACE INFORMATION 419 SIBERIA

SIBERIA

DIVERGENCE 0.04

ALASKA

SHEAR 0.08

ALASKA

0.00

0.04

–0.04

0.00

07035 – 07041 Feb 4 – Feb 10

07035 – 07041 Feb 4 – Feb 10

SIBERIA 101 VORTICITY 0.06

ALASKA

motion 1012

1020

1028

0.00

–0.06 07035 – 07041 Feb 4 – Feb 10

36 10

10 km d–1

Figure 10.8 Maps of 6-day average divergence/convergence, shear, vorticity, and ice motion annotated with sealevel pressure contours (interval 4 hPa). These are products from the RGPS. Data are computed at grid cell of 10 km and averaged over the period 4–10 February 2007. The solid circle in the motion map marks landfast ice area and the asterisk points to windy area where ice motion coincides with sea-level pressure contours [Kwok, 2010 / Reproduced from Cambridge University Press].

where, n and A are the number of vertices of the cell and its area, respectively. A similar equation can be constructed ∂u for . The area is computed as: ∂x A=

1 n xi yi + 1 − yi xi + 1 2i=1

(10.7)

Areal changes of the cells can be used as an indicator of the growth or transformation of young ice. The example shown in Figure 10.9 was produced in Yu et al. [2009]

from calculations of total strain rate using equation (10.4). The motion field was generated from two consecutive SAR images acquired on 11 and 14 December 2001, over the MIZ of the Bering Sea. The averaged azimuth θ over the scene shown in the figure is 79.2 , which is close to the value 90 for the case of pure shear. The mean shear Es [equation (10.2)] was found to increase by 13.2% per day, which was significantly higher than values in neighboring areas. This is explained by the presence of extensive landfast ice in this region and the fact that a shear zone separates landfast from the pack ice. The map in

420

SEA ICE 500

Y (100 m)

400

300

200

100 –1 15%day 15% day–1

100

200

300

400

500

600

700

X (100 m)

Figure 10.9 Deformation grid of sea ice cover on 11 and 14 December 2001, obtained from a pair of SAR images acquired on these 2 days. Total strain-rate vector (indicated by arrows), are overlain on the SAR image of 14 December. A scale of the deformation rate is shown at the lower left corner [Yu et al., 2009 / with permission from Taylor and Francis].

Figure 10.9 demonstrates the potential use of theoretical and in-situ observations of ice kinematics to generate maps mechanical properties of sea ice. Lindsay and Stern [2003] show an illustrative figure of the ice deformation produced from the RGPS, covering the Beaufort Sea and the area north of it in the Arctic Ocean. It is presented in Figure 10.10, which is a map of total deformation rate calculated from equation (10.4) for all cells of the available images from RADARSAT-1 scenes acquired during the last week of May 1998. The 10% of the cells with the largest total deformation rate are shown in black. The concentration of high deformation indicates quasi-steady LKFs. These are pixels with higher deformation rates compared to the surrounding environment. They may be interpreted as ridges if the width and length satisfy the shape criteria. The large area with no data in the Beaufort Sea is caused by advection of the ice cells that covered the sea at the time of initialization of the motion field in January 1998. Linow and Dierking [2017] identify and characterize LKFs in the polar regions. This was performed using an image analysis approach. This allowed development of a variety of new metrics for ice deformation as well as the temporal stability of individual features. Data are vectorized to obtain results on an object-based level, then a semantic post-processing is applied to determine the angle of junctions and between crossing structures.

North Pole

Beaufort Sea Alaska

Figure 10.10 Map of total deformation rates at the end of May 1998 produced by the RGPS. Calculations were performed using a series of RADARSAT-1 images initiated in January 1998. The 10% of the cells with the largest deformation rates are shown in black. Note the quasi-linear forms of large deformation with their high concentration at the right and bottom parts of the image [Lindsay and Stern, 2003, Figure 4 / with permission from the American Meteorological Society].

RETRIEVAL OF SEA ICE SURFACE INFORMATION 421

In another publication, Hutter, Zampieri, Losch [2019] introduced two algorithms to detect LKFs in sea ice deformation data and establish a data set for the entire observing period of the RADARSAT-1 in ASF (1995–2008). The algorithms are available as open-source code and applicable to any gridded sea ice drift and deformation data. The multiple small segments of KLFs reconnect into individual features based on their relative distance and orientation. The above discussions focus on retrieving surface deformations from SAR. However, optical remote sensing data have also been used, though not very often, to identify ridges and determine their statistics. Johannessen et al. [2007] demonstrate limited applications of these data and hint to a plan of using SAR and optical images in operational ice service to monitor ice ridges in the Northern Sea Route. 10.1.4. Cracks and Leads Another ice surface feature of interest to marine navigators and climate modelers is fractures in ice cover. This is manifested in the form of cracks and leads, which result from divergence of the ice cover under tensile failure of the ice in response to the stresses induced by the differential motions of the overlaying air and underlying water. A crack is any fracture in an ice cover which is typically less than 1 m wide (Figure 2.48). Its width varies from tens of meters to a few kilometers. Cracks can be detected in remote sensing data using optical sensors of sub-meter resolution. These include commercial satellites such as Ikonos, QuickBird, Worldview, and GeoEye. However, images from these sensors are expensive ($10–$18 /km2), hence not commonly used in scientific research. A more commonly used source of fine-resolution data is aerial photography, which is even more expensive but offers data where and when required. Detection of cracks is not practically important except for the fact that it can serve as an indicator of a possible lead formation (much like detection of cracks on ice sheets that may eventually cause calving of icebergs). An unfamiliar impact of cracks on sea ice is discussed in Moore et al. [2014]. The authors discovered that vigorous mixing in the air above cracks in sea ice in the Arctic pumps atmospheric mercury down to the exposed water surface. This causes toxic pollutants entering marine food system, which may affect the health of the fish and ultimately humans. Leads are openings navigable by surface vessel [MANICE, 2005]. Their identification and spatial statistics in polar regions are important for a few reasons. Leads are major sources of heat and moisture fluxes to the atmosphere. Since they have much less insulating capacity and a warmer surface temperature compared to the surrounding ice, they trigger ocean–atmosphere

heat exchange with rates of two orders of magnitude greater than the surrounding ice cover [Maykut, 1978]. Identification of lead and their distributions within the regional sea ice cover is therefore critical for weather and climate modeling [Worby and Allison, 1991]. Leads are also navigable marine routes, whether open or covered with thin ice. Moreover, they are important for wildlife. Seals, whales, polar bears, Antarctic penguins, and other animals rely on leads for access to oxygen and food. The Arctic and remote sensing communities have recently developed more interest in lead identification as their frequency of occurrence has increased as a result of the ongoing trend of ice thinning and mobility. Moreover, leads have been more useful for accurately estimating the sea ice freeboard needed to retrieve ice thickness from airborne and space-borne altimeters (section 11.4.3). Leads are formed from cracks or fissures as they open up progressively under synoptic-scale weather and oceanic forces. The opening takes place over a few days. Obviously, cracks and the subsequent possible formation of leads depend on the mechanical properties of the pack ice. They are more likely to occur when ice is relatively thin (up to the stage of thin and medium FYI as defined in Table 2.4). In the Central Arctic, leads cover 1–2% of the surface area in winter and a somewhat higher percentage during the summer. The impact of leads on the ice cover in the polar regions is different in winter than in summer. In winter, the OW in leads begins to freeze almost immediately when exposed to the cold atmosphere and becomes nearly completely frozen within a day or a few days. Accordingly, the surface of leads normally becomes covered with thin ice types (nilas or gray ice). In the summer, the lower albedo of the water surface in leads causes more absorption of solar energy than the surrounding ice. This accelerates the ice melting and widens the lead. Leads often branch or intersect, creating a complex network at the surface of the ice cover. Open leads release heat and moisture to the atmosphere, while refrozen leads continue to release heat. Leads form under geostrophic wind forcing. On a basin-wide scale in the Arctic, this occurs in bands of approximately uniform width, running parallel to the direction of the major stress in the ice sheet [Melior, 1986]. Physically speaking, leads are triggered by highpressure systems that produced a relatively warm atmospheric temperatures and a series of high southwesterly winds that drove an atmospheric/oceanic circulation known as the Beaufort Sea Gyre (BSG) [McLaren, Serreze, Barry, 1987]. They are more observed in the western region of the Beaufort Sea in winter. In the Antarctic region, katabatic wind continues to remove sea ice formed in coastal areas, resulting in opening leads and polynyas. Katabatic winds, also called fall winds, are caused by gravity pulling high density air from high elevation areas on top

422

SEA ICE

N

Opennings in ice

Banks Island

of mountains down slope to lower density air (katabatic is Greek word meaning–going down). Leads can be detected using medium-resolution satellite sensors (e.g., 250–1000 m resolution MODIS) or fine-resolution SAR. An example of lead distribution under the action of the BSG is shown in the series of the AVHRR image animation obtained in February 2013. Three frames from the daily animation, separated by 1-week interval, are shown in Figure 10.11. Major leads are geometrically shaped in linear or curvilinear forms (they are usually correlated with geostrophic wind as illustrated later). The scenario

of lead formation and propagation proceeds as follows. A high-pressure system was established over the western area during February–March 2013, causing relatively warm atmospheric temperature and a series of high southwesterly winds that drove the BSG. Cracks started to appear in late January and spread east toward Banks Island. The series of images shown in the figure demonstrates the effect of the frequent storms on exacerbating the initial fracturing (top image). Ice fractures advanced progressively eastward. Within a week, more leads were formed (apparently with re-frozen surface) and had extended about 500 km eastward. By the end of February (the bottom image), they advanced another 500 km, and the BSG detached the fast ice along the west coast of Banks Island. This opened a coastal polynya, which is not uncommon feature in this region. Width of leads varies over a wide range from a few meters to hundreds of meters. To allow their detection using remote sensing data, the lead’s width should be at least double the spatial resolution or the footprint of the sensor. Their length extends from a few hundred meters to a few tens of kilometers in length. Yet, some synoptic-scale leads may extend for hundreds of kilometers. Kwok et al. [2009] presented a cumulative distribution of lead width using data from the Arctic Ice Mapping (AIM) moorings, and an average of five submarine cruises during the Scientific Ice Expeditions (SCICEX) in 1993, 1996, 1997, 1998, and 1999 (Figure 10.12). A considerable

Consolidated ice sheet 1.0

Re-frozen leads

Fresh leads

Cumulative fraction

0.8

0.6

Avg of SCICEX cruises AIM:S1:9/03 to 8/04 AIM:S2:9/04 to 9/05

0.4

0.2 ICESat footprint

Figure 10.11 AVHRR images acquired in February 2013, with 1 week separating each image, showing leads formed in the Beaufort Sea. Note the propagation of leads eastward (from top to bottom images). The clockwise motion of the BSG is indicated in the top frame. Original open leads that appear dark in the top images became re-frozen with brighter signatures in the bottom image. Note the fresh coastal lead (coastal polynya) near Banks Island in the bottom image (frames from animation available from NASA’s Earth Observatory, annotated by the authors).

0.0 0

50

100

150

200

Lead width

Figure 10.12 Cumulative distribution of the width of leads in the Arctic from AIM and SCICEX programs (see text). Data were acquired during summer and winter. The arrow points to the 70 m footprint width of ICESat-1 [Kwok et al., 2009, Figure 2b / with permission from AGU].

RETRIEVAL OF SEA ICE SURFACE INFORMATION 423

fraction of the leads is smaller than 100 m in width and therefore cannot be detected by medium-resolution sensor (a few hundred meters). The widths of less than 20% of the Arctic leads are comparable to the 70 m nominal footprint of the laser altimeter GLAS on ICESat (see Table 8.6), which was used during its lifetime (2003–2010) for lead detection in order to infer the freeboard of the ice cover. Later, sea ice leads were used to produce regular products of ice freeboard from ATLAS laser altimeter onboard ICESat-2. Data are available through NSIDC website. In general, leads can be detected in VIS, TIR, or SAR images. In all cases they appear in the images as long quasi-linear or jagged narrow features. In optical imagery, leads and the surrounding ice areas appear with the same contrast as they do in nature; i.e., leads are darker due to their lower albedo. In TIR imagery, the contrast is maintained because of the significantly higher surface temperature of the lead compared to the much colder ice in polar winter (recall that the emissivity of both surface covers in the TIR region is almost equal as explained in section 7.2.3). While VIS channels can only be used during daylight polar season, they are preferred over the TIR during summer when the temperature difference between the ice and water surface is reduced. Wind does not generate as much wave action in leads as it does in OW, consequently thin ice grows in leads with a smooth surface under relatively steady conditions. Hence, in SAR imagery, leads appear darker than the surrounding ice even if the lead is covered with thin ice. Figure 10.13 shows a RADARSAT-2 SAR image overlaid on MODIS image of a sector in the Beaufort Sea with numerous leads

within the ice cover. More leads with more details are apparent in the fine-resolution SAR. More importantly, while all leads in the MODIS image appear with the same gray tone, they appear in different tones (backscatter) in SAR image. In fact, some leads demonstrate dark and bright tone with strong contrast. In SAR images, the dark and bright tones indicate water and re-frozen ice, respectively. The combination of the two tones in the same lead reflects the different stages of freezing within the lead. The brighter the tone the thicker the ice (Figure 9.27 shows the backscatter increase during the early growth phase of sea ice). Optical data do not show the stage of new ice growth in the lead. Earlier studies of lead detection (1980s) relied on visual interpretation of satellite images in the VIS region (0.4–1.1 μm) and TIR region (8–13 μm) to detect largescale leads in the Arctic. Barry et al. [1989] used VIS and TIR channels of the Operational Linescan System (OLS) onboard a few DMSP satellites (section 8.3) to manually detect leads at 0.6 km resolution in the Beaufort Sea in May 1983, based on the contrast between their signature against the surrounding ice. The authors also used geostrophic winds provided by the Arctic Ocean Buoy Program, referenced at a 1 × 5 grid to explore their correlation with large-scale lead geometry. Leads were found to be arranged roughly parallel to the geostrophic wind direction as shown in Figure 10.14 for leads that are greater than 300 m in width. However, the wind may not be able to open up leads in winter if ice becomes fully compacted. Another study that used VIS and TIR imagery to manually map the distribution of leads in the western Arctic is presented in Miles and Barry [1998]. A five-year lead climatology for Arctic ice is presented by utilizing Landsat multispectral scanner (MSS) images (80 m resolution) with the OLS images (600 m resolution). To quantitatively analyze the large-scale relationships between the wind field and the lead orientations, Barry et al. [1989] used the following relation:

rB = N

−1

cos δi i=1

Figure 10.13 A background of MODIS image of a section in the Beaufort Sea, acquired on 25 April 2016, with RADARSAT-2 image overlaid. Details of the leads are shown better in SAR image with the composite coverage of open and refrozen surfaces (image courtesy of Meng Qu of Wuhan University, China).

2

N

2

N

sin δi

+

1 2

(10.8)

i=1

where, δ is the difference between the lead orientation ϕ and the geostrophic wind direction θ, δ = (ϕ − θ); N is the number of examined grid cells. A large rB indicates that ϕ is correlated with θ and vice versa. Therefore, rB can be used as an indicator of the variability of lead patterns which can be attributed to atmospheric forcing. A summary of the methods for lead detection is presented in the following. One of the early quantitative approaches to identify leads in remote sensing images of ice cover involved examination of the probability

424

SEA ICE 90°W

90°E

1020 102 4

80° N

102 6

70

102

0

°N

10 16

102

4

H

Alaska

1 May 1985

Figure 10.14 Pattern of mean sea-level pressure (mbars) in the Beaufort Sea on 1 May 1985 (thick curves), which served as an indicator of geostrophic wind, and the lead distribution on the following day obtained from using the VIS and TIR channels of the OLS sensor onboard DMSP satellites (thin curved lines). Note the tendency of lead orientation to be arranged roughly parallel to the geostrophic wind direction, which is approximately parallel to the isobars [adapted from Barry et al., 1989].

distribution of albedo or brightness temperature from the entire image. Since each parameter exhibits a substantial difference between ice and OW, the distribution should be bimodal with separate modes representing leads (whether open or re-frozen) and the surrounding ice. In this case a simple threshold can be set to separate the two entities. However, this idealization may not be work if thick ice and water coexist in the lead, which is possible particularly in wide leads. Ice within a lead may be manifested in the form of re-frozen ice or floating thick ice pieces broken from the edge of the lead (see Figure 2.50). This situation may be revealed in SAR images but not in the case of a relatively coarse resolution sensor (e.g., AVHRR, MODIS, VIIRS, or PM sensors). A simple threshold approach will not be suitable in the latter case. Given the possibility of heterogeneous lead pixels in AVHRR, Lindsay and Rothrock [1995] used the threshold approach to detect subpixel-sized leads in a series of AVHRR imagery acquired throughout 1989 in the Arctic using observed thermal brightness temperature in winter and albedo in summer. For each cell of 200 km2 they calculated the surface temperature Tsfc using a split window equation, e.g., equation (7.55). In a heterogeneous pixel, a potential for OW based on brightness temperature, denoted δBT, is calculated based on a ratio of surface temperature using typical values from thick ice, Tice and OW, Tow: δBT

T sfc − T ice = , T ow − T ice

for T sfc > T ice

(10.9)

and δBT = 0,

for T sfc ≤ T ice

(10.10)

Similarly, another potential for OW, based on albedo, denoted δA is given by: δA =

αice − αsfc αice − αow

(10.11)

where, αsfc is the estimated surface albedo from AVHRR, αice and αow are the typical albedo of ice and OW, respectively. Lindsay and Rothrock [1995] used δA for summer months June, July, and August. When a threshold of δBT or δA is chosen, binary images of lead-like structures can be generated. Statistics of leads (e.g., length, width, orientation, and spatial frequency) can then be determined, but they will be sensitive to the selected threshold. Based on δBT or δA equal 0.1, the authors found that the average lead width in the Central Arctic reaches a minimum of 2.3 km in January and February and a maximum of 6 km in September. They also confirmed a previous finding by Wadhams [1988] that the distribution of lead width follows a power law (the power law distribution exhibits a negative exponential shape when plotted on a linear axis system and a linear shape when plotted on log-log axes) N w = aw − b

(10.12)

where, N(w) is the number of leads of width w per kilometer of track, and a and b are coefficients that can be determined from empirical data (the best fit of frequency of

RETRIEVAL OF SEA ICE SURFACE INFORMATION 425

occurrence of lead width from a set of lead width measurements). Wadhams [1988] used submarine sonar data to calculate a power law exponent. He found b = 2, in Fram Strait and 2.29 in Davis Strait. Using a threshold on satellite radiometric measurements to identify leads, produces acceptable results only when the leads are open (contains water only). Recent studies of leads in the Arctic have used more effective approaches. Of particular interest are methods adjusted to detect narrow leads and examine the ice/water contents within the lead. To account for the heterogeneity of the surface composition in the lead, Willmes and Heinemann [2015] introduced the concept of the pixel anomaly to separate the lead pixels from the surrounding thick ice. This concept became popular and was used later in a few studies of lead identification. The concept entails application of a median filter with appropriate window size at each pixel of the original image. The image can be of brightness temperature or physical temperature derived from TIR satellite data. The median value is assigned to each pixel. In the next step, the deviation of the pixel value in the median image from the values of the surrounding pixels in the original image is calculated. Those values are used to construct a new image, called “anomaly image.” The underlying theme is that thick ice outside the lead has more-or-less same brightness or physical temperature (since it is thick and stable during the winter season). Therefore, its anomaly values should be low. On the other hand, the brightness or physical temperature within the lead must be higher and more variable due to the heterogeneous surface conditions; i.e., coexistence of water and ice at

different growth stages. This means the anomaly values within the lead should be at the higher side using an appropriate threshold. A threshold on the anomaly can be used to discriminate the lead from the surrounding area. This approach was used in Willmes and Heinemann [2016] to map pan-Arctic lead distribution at 1 kilometer spatial resolution using anomalies of ice surface temperature (IST) product from MODIS. Results from 2003 to 2016 show no clear interannual trend in lead numbers or spatial distributions. Reiser, Willmes, Heinemann [2020] updated the algorithm of Willmes and Heinemann [2016] to map leads in the Arctic and Antarctic regions. The authors also did not find interannual trend in lead number or distribution and attributed this to the conservative cloud mask in the MODIS/MYD29 IST product and the disregard of rapid ice motion in the retrieval of daily leads. Qu et al. [2021] used the theme of the original algorithm by Willmes and Heinemann [2016] with a few modifications to map the leads in the Beaufort Sea in the month of April from 2001 to 2020. Surface temperature was calculated from MODIS TIR data instead of using the IST product. Anomalies in detected lead distributions were related to surface pressure and sea ice dynamic. However, similar to the two aforementioned studies, Qu et al. [2021] did not find identifiable trend of lead area during the two studied decades. Figure 10.15 shows an annual record of lead area during the study period. The lack of trend is obvious. Nevertheless, there is still a compelling hypothesis in favor of an increasing trend of the number and area of the leads in the Beaufort Sea, since ice is thinning and its mobility is enhancing.

April mean

Lead fraction

0.20

0.15

0.10

0.05

0.00 2001

2005

2010

2015

2020

Year

Figure 10.15 Interannual variability of lead area in the Beaufort Sea in April 2001–2020. Daily lead area fractions are illustrated using boxplots (courtesy of Xi Zhao, Sun Yat-sen University, China).

426

SEA ICE

Research on lead detection should continue using fineresolution SAR data from constellations such as the system on RCM, which warrant daily acquisition of data over the Arctic Basin. Contents of the lead were studied in Röhrs and Kaleschke [2012]. The authors developed a lead detection algorithm, where OW and thin ice are treated as a single entity and thick ice is another entity. This could be achieved using passive microwave (PM) data from 89.0 GHz and 18.7 GHz from AMSR-E. The ratio of the emissivity between these two frequencies is >1.0 for water or thin ice (new ice, nilas, and pancake). This allows discrimination of the two entities in the lead. Since retrieval of emissivity requires using radiative transfer modeling, the authors avoided this by showing that emissivity ratio is the same as the ratio of the brightness temperature between 89.0 GHz and 18.7 GHz. The theme is sound, but separation of thin ice from water in the lead is missing in this method. Furthermore, the resolution of PM imagery is too coarse to detect narrow leads prevalent in the Arctic. In the work of Röhrs and Kaleschke [2012] only leads wider than 3 km were detected. This is about 50% of the leads that are visible in MODIS images. High-resolution data needed to identify narrow leads (a few tens of meters width) can be obtained from airborne sensors. Tschudi, Curry, Maslanik [2002] used data from an airborne video camera and a down-looking PM radiometer, both mounted on the underside of the NSF/ NCAR C-130 aircraft. Data were obtained from the vicinity of the Surface Heat Budget of the Arctic Ocean (SHEBA) site during April–July 1998. The authors used an automated methodology to determine lead fraction, width, and orientation. They found that leads were mostly covered with thin ice. The width varied between 25 m and nearly half a kilometer, but leads narrower than 100 m accounted for 75% of the width distribution. Another interesting finding was that the lead fraction (percentage of lead area per unit area of ice cover) decreases exponentially with increasing lead width. The study confirmed, once again, that lead orientation was consistent with prevailing wind patterns and ocean circulation. Detailed survey of leads using airborne cameras is the best tool for geometric characterization of leads across the vast pack ice in the Arctic, but it is not a practical tool to map leads across a large region. Onana et al. [2013] developed an algorithm capable of detecting narrow leads and identifying their ice type composition. The authors used high-resolution airborne visible imagery acquired by the Digital Mapping System (DMS) with a spatial resolution between 0.1 and 2.7 m, depending on the aircraft altitude. The data were collected during NASA’s Operation Ice Bridge mission (OIB). The algorithm detects a wide variety of leads ranging from narrow meter-wide to wide leads of

hundreds of meters. The unique feature of this algorithm is its ability to account for the variation of optical signatures within the leads. Although the overall signature of a lead can be of low albedo (typical of the OW signature), it may have a dynamic pixel intensity range that varies across individual scenes. The range accounts for the different phases of re-freezing within the lead as well as the frost flowers that may cover its thin ice surface temporarily (section 10.2.2). The algorithm starts with a transformation, called lead vicinity transformation (LVT), to determine if leads exist in the image: LVT I x, y

= atan I x, y

0,

π 2

(10.13)

where, I(x, y) is the pixel intensity at position (x, y). The transformation acts to smooth I over ice floes and creates an intensity transition within the lead. This step is followed by another affine time-frequency transformation that generates localization around low intensity and low frequency lead pixels. It is called the minimal signal transform (MST). An example of the results from the two transformations is presented in Figure 10.16. The original image from the DMS airborne system is shown along with the results from applying the LVT followed by the MST. The original image shows a variety of signatures within the lead and the surrounding ice. The result from the application of the LVT transformation only (not shown in the figure) reveals less variability of the lead signature and therefore allows easier detection. The application of the LVT followed by the MST highlights the lead area with more uniform and homogeneous signatures and better contrast with the surrounding ice. An appropriate threshold can then be used to discriminate between the lead and its surrounding ice cover. The ice type classification within the lead is also shown Figure 10.16. The classification is based on a set of thresholds that was determined from the probability distribution of the fusion of minimal signal. Ice usually exists at the boundaries of the lead, while the core is usually covered with ice. SAR data were used to map leads at fine resolution (a few tens or hundreds of meters), though the swath is narrow. Murashkin et al. [2018] developed an algorithm to detect lead using dual-polarization Sentinel-1 SAR images. Murashkin and Spreen [2019] used a series of Sentinel-1A/B images to generate lead distribution in the Arctic with a temporal resolution of 1–3 days. In this work, the authors used the backscatter and texture features in the image to separate OW and thin ice combined from the rest of the ice types. Therefore, the area between ice floes is also classified as leads. The number of leads is therefore overestimated. An application of a shape criterion is necessary to recover the lead pixels and exclude other water areas.

RETRIEVAL OF SEA ICE SURFACE INFORMATION 427 (a)

(b)

200

0.3

180 0.25

160 140

0.2

120 100

0.15

80 0.1

60 40 20

0.05 1 cm ~70 m

0

(c)

(d)

250 0.9 0.8

200

0.7 0.6

150

0.5 0.4

100

0.3 0.2

50

0.1 0

0

Figure 10.16 (a) Original DMS airborne image of ice sheet traversed by a lead and (b) the same image after applying the LVT followed by the MST. It is much easier to detect the lead in (b) by applying a threshold. (c) another lead surrounded by sea ice and (d) the resulting classification of the image in (c) after applying the MST where the dark, medium, and light gray shades represent gray ice, thin ice, and OW, respectively. The arrows point to OW [adapted from Onana et al., 2013].

Currently, there are two datasets of lead distributions in the Arctic generated by the Integrated Climate Data Center at the University of Hamburg. The first is derived from AMSR-E and the second from CryoSat-2. The AMSR-E dataset was generated using a brightness temperature ratio between microwave frequencies to calculate thin ice concentration or lead area fraction. This was available during the lifetime of AMSR-E (2002–2011), at a resolution of 6.25 km. The CryoSat-2 dataset was derived from the Synthetic Aperture Interferometric Radar Altimeter (SIRAL), averaged monthly in a 100 km grid. More

comprehensive pan-Arctic datasets are yet to be generated at finer resolutions. As Arctic sea ice continues to decline and become thinner under the recent global warming, more leads tend to appear, particularly in the western section of the Arctic Ocean. The interest in studying the frequency, geometric characteristics, and contents of the lead has grown. A few scientific questions about modern-day leads in the Arctic are still under investigation [Serreze and Barry, 2011]. Has re-frozen or detached ice filled more area of the lead today compared to a few decades ago? Do leads trigger Arctic

428

SEA ICE

amplification? Is the trend of lead formation affected by variations of Arctic sea ice composition and seasonal cycle?

10.2. THERMALLY INDUCED SURFACE FEATURES 10.2.1. Surface Melt Ice surface melt is a significant climatic parameter. Its mapping at a synoptic scale across the Arctic is of interest to climatologists to assess the impact of the current climate change on the Arctic climate system. Surface melt also impacts operational marine activities since it is an indicator of ice decay. Mapping this parameter at regional scales is needed to support marine navigation because ice decay is associated with the reduction of ice loads on the ships. Since early 2000s, the Canadian Ice Service has incorporated surface melt as an additional entry in its ice mapping products [MANICE, 2005]. Surface melt affects interpretation of remote sensing data. It modulates the reflection and emission of radiation in all bands of optical and TIR spectra. Additionally, it modulates the microwave signature of the ice surface from passive and active sensors. Wet surface increases emission from all ice types but decreases the backscatter from MYI surface (Figure 9.18) as it masks the dominant volume scattering from this ice type. This effect usually leads to underestimation of ice concentration. Therefore, knowledge of surface melt fraction is required to improve the accuracy of ice concentration in the summer [Tschudi, Curry, Maslanik, 2001]. The main drive of ice surface melt is the advection of warm and humid air over the ice cover. This is associated with an increase in downwelling longwave radiation, a process that initializes melt onset [Mortin et al., 2016]. Early melt onset allows absorption of more solar radiation, hence triggers more melting through a positive feedback loop. Snow on sea ice is more susceptible to melting due to its higher melting point. Ice surface melt progresses through three phases: premelting (also called melt onset), advanced melting (also called ponding phase), and finally the ice decay phase. Definitions of these phases follow. The remote sensing community is particularly interested in the first phase because it leaves significant impact on the observations. The operational ice community is interested in the ponding and decay phases as marine navigation becomes more feasible during that time. Climatologists are interested in the three phases. Surface melt should monitored the throughout the three phases. Parameterizing the first two phases is relatively easy but difficult for the third phase of decay. The following discussions focus on

estimation and mapping of the melt onset and ponding phases. Only the Arctic region is addressed. In mapping these phases at a regional or synoptic scale using remote sensing data, a compromise between the swath, spatial, and temporal resolution of the sensor becomes a central issue. Coarse resolution sensors seem to be a perfect choice to satisfy the daily Arctic-wide mapping yet at the expense of the spatial accuracy. This is the case when using PM or scatterometer data (a few kilometers of spatial resolution at best). On the other hand, SAR offers the best resolution at a few tens or hundreds of meters, but the gaps between the satellite overpasses can be wide and the temporal resolution is also poor. Nevertheless, data from SAR constellation systems such as the European Sentinel-1A/B or the Canadian RCM provide promising solutions to overcome the poor temporal resolution. The pre-melting phase is characterized by a remarkable increase of wetness or degree of waterlogging in the snow cover. In the case of a snow-free ice surface, quasi-liquid films are formed on the surface. In the advanced melting phase, sporadic melt ponds appear at the surface (Figure 2.54). Their distribution is linked to the spatial variation of the snow depth and the uneven distribution of the surface salinity, which makes the absorption of the incoming solar radiation spatially different. Figure 10.17 shows snow-covered ice surface during the pre-melting phase and the same surface with melt ponds that appeared 6 weeks later. The photos were obtained using web cameras deployed to monitor air, ice, and ocean conditions within the USA–Japan joint project entitled “North Pole Environmental Observatory” (NPEO). During the ponding phase, the sea ice surface is transformed into a heterogeneous mixture of light blue-colored melt ponds and white interstitial snow/ice areas. In the last stage of ice melt (ice decay), the ice surface develops into what is known as “rotten ice.” This is the phase of ice disintegration where the surface becomes honey-combed with melt ponds covering more than 50% of the surface area as shown in Figure 10.18 [MANICE, 2005]. This is the least understood stage from remote sensing perspective, partly due to the lack of dedicated field observations but mostly because of the heterogeneous footprint that contains pond water and ice. In the following discussions, the onset of snow melt and, to a less extent, the ponding phases are addressed using optical, PM, active microwave, and airborne data. However, it should be noted that melt pond detection has been addressed using ICESat-2 [Tilling et al., 2020] but not introduced here. 10.2.1.1. Optical Observations Broadband albedo from melting ice surface and melting snow in the Arctic is presented in Table 9.1, and spectral

RETRIEVAL OF SEA ICE SURFACE INFORMATION 429

Figure 10.17 Two photographs of snow-covered sea ice in the Arctic (same site) during onset of melt period (left) and advanced melt (right) (from NOAA Arctic website).

Figure 10.18 Rotten FYI during the advanced decay phase, showing extensive melt ponds and interstitial snow/ice areas [MANICE, 2005 / from Canadian Ice Service].

albedo of ice types with different snow conditions are shown in Figure 9.3. Albedo from snow and ice during melting phases obtained from a combination of AVHRR visible channels and in-situ observations is presented in De Abreu, Arkett, Ramsay [2001]. The data were obtained during a field campaign within the CollaborativeInterdisciplinary Cryospheric Experiment (C-ICE’00), conducted over a 9-week period from 21 May to 25 July 2000, in the central Canadian Arctic Archipelago. Figure 10.19 demonstrates the evolution of the average top-of-atmosphere (TOA) reflectance calculated over a 3 × 3 pixel window. The effectiveness of using this measurement to identify the three phases of surface melting can be confirmed from the figure. During the pre-melting phase the reflectance from the ice surface decreases with respect to the measurements from the winter conditions. De Abreu, Arkett, Ramsay [2001] reported that the TOA reflectance from channel 2 decreased from 0.74 to 0.64 during the period 17 April to 14 June. This is due to the growth of snow grain and the presence of snow liquid water content in the upper layer of the snow [Warren, 1982]. In-situ observations indicated that melt ponds first appeared after 19 June. The reflectance continued to decrease during the melt pond formation period, showing a drop from 0.67 to 0.50 (channel 1) and 0.47 to 0.29 (channel 2) between 19 June and 25 June. The pond fraction increased from 0% to 45% during this period. Surface drainage was first observed on 24 June and resulted in a decrease in melt pond fraction. As the drainage progresses, the enlarged brine

SEA ICE

pockets interconnect to form vertical drainage pathways to the underlying water. Moreover, the ice becomes littered with cracks and seal holes. All these processes accelerate surface melt water to drain vertically through the ice, leaving much less surface area covered by water. De Abreu, Arkett, Ramsay [2001] also reported that the newly uncovered remnant ice surface is often dry, crumbly, and dull white in color. That is the reason for the increasing reflectance during the second week of July as shown in Figure 10.19. It should be mentioned that surface melt can also be identified by the increased difference between reflectance from AVHRR channels 1 and 2 when compared to the winter dry surface reflectance (almost equal from AVHRR 1 and 2, but not shown in the figure). Polashenski, Perovich, Courville [2012] conducted a series of observations on melting FYI and landfast ice near Barrow, Alaska in 2008, 2009, and 2010 to explore the seasonal evolution of melt pond coverage and its expected effect on decreasing the surface albedo. The albedo was measured along transect lines every 2.5 m using an albedometer that integrates albedo over wavelengths between 300 and 3000 nm. Spatially averaged albedo values calculated from the measurements are plotted in Figure 10.20. The snow-covered ice surface shows stable albedo around 0.75 in winter. A few days before the onset of ponding, albedo starts to drop sharply. A few days later it reaches a minimum around 0.25 when the pond coverage reaches maximum. Albedo starts to increase when meltwater begins to percolate through

connective porosity in the ice, exposing a nearly dry ice surface. Greater variability of albedo is shown during this period. In 2 weeks, the albedo stabilizes around values between 0.5 and 0.6. Figure 10.20 also shows the correlation between the percentage of the pond coverage and the measured albedo. It is obvious that the pond coverage is the primary driver of albedo decrease. Daily melt pond cover over sea ice in the Beaufort/ Chukchi Sea region was detected using measurements of surface reflectance from MODIS through the summer of 2004 [Tschudi, Maslanik, Perovich, 2008]. Results show a rapid increase of ponding areas from 10% to 40% of the total ice surface during the first 20 days of melt, followed by fluctuations through summer. Toward the end of the summer, a gradual decrease occurs, with the ponding area reaching 10%, in late August. Since MODIS observations are usually obtained from heterogeneous footprints, the author used the well-known linear decomposition equation to decompose the observation into components from ice and melt ponds. The daily fraction of melt pond coverage can then be obtained by solving a set of such equations: Rk =

i

ar , k i ik

10

90

AVHRR TOA reflectance (%)

0

Winter ice –10

70 60 Drained ice 50

Snow melt

40 AVHRR 1 30 AVHRR 2 20 10 0 4/15

Resolute air temp.

4/20

6–8

(10.14)

where, Rk is the observed surface reflectance from MODIS channel k, rik represents the typical spectral reflectance from the surface i by channel k, and ai is the fractional coverage of each surface. Four surface

100

80

ai = 1

Air temperature (°C)

430

Rotten ice Ponded ice

6–13 6–18 6–23 6–28

7–3

7–8

7–13 7–18 7–23 7–28

Date

Figure 10.19 Variation of TOA reflectance by AVHRR over fast FYI during phases of ice melt as determined by concurrent surface observations during the C-ICE’2000 field experiment in the Central Arctic. Air temperature was measured at Environment Canada’s weather station in Resolute Bay [De Abreu, Arkett, Ramsay 2001].

RETRIEVAL OF SEA ICE SURFACE INFORMATION 431 (a)

(b) 0.8

0.9 2008

0.8 0.7

2009

2010

0.6

2010

0.5 Albedo

Albedo

0.6 0.5 0.4

0.4 0.3

0.3

0.2

0.2

0.1

0.1 0.0 –10

2008

0.7

2009

0.0 –5

0

5

10

15

20

0

25

10

20

30

50

40

60

70

80

Pond coverage (%)

Days from onset of ponding

Figure 10.20 (a) Evolution of albedo from surface measurements around and after onset of melting, and (b) albedo versus percentage of melt pond coverage. The solid line represents the linear regression of the data. Data obtained from three years of observations of Arctic sea ice near Barrow, Alaska [Polashenski, Perovich, Courville, 2012, Figure 21 / with permission from AGU].

types were selected: melt ponds, OW, snow-covered ice, and white ice. The values of rik were obtained from in-situ measurements of reflectance in June 2004, near Barrow, Alaska. Three MODIS bands were used: band 1 (620–670 nm), band 2 (841–876 nm), and band 3 (459–479 nm). Equation (10.14) is solved for the

–160°

–140°

fractional coverage of each surface. An example of the daily melt pond fraction is shown in Figure 10.21. The pond fraction is higher at southern latitudes near the coast where the pack ice is separated from the landfast ice. It is lower inside the pack ice and within the archipelago.

0.0

70° Alaska

Pond Fraction

0.6

Clouds

80°

Figure 10.21 Reflectance from Channel 1 of MODIS (left) and the derived pond fraction using equation (10.14) (right). Data are for 13 June 2004, projected in 500 m EASE-grid format [Tschudi, Maslanik, Perovich, 2008, Figure 3 / with permission from Elsevier].

432

SEA ICE

In general, melt ponds appear on sea ice surface in the Arctic during the daylight season. Therefore, optical sensors can be used for their detection under cloud-free sky. Rösel [2013] presents a summary of using optical sensors to detect melt pond and presents an Arctic-wide, multiannual melt pond data for the years 2000–2011. 10.2.1.2. Passive Microwave Observations A few approaches exist to determine the onset of melt and re-freeze of Arctic sea ice from satellite PM data. They are based on examination of sequential daily parameters, derived from the observed brightness temperature, which are indicators of the melt or freezing onset. Thresholds are usually specified based on physical reasoning to delineate each event (onset of melting and re-freezing). PM data have been used to establish a long record of sea ice melt onset and freezing onset (1979–present) with fine temporal resolution (1 day) but coarse spatial resolution (12–25 km). Kunzi, Pail, Rott [1982] found that the presence of liquid water in the snow decreases the difference between brightness temperatures (Tb, 18h − Tb, 37h). This parameter is called horizontal range (HR). The physical explanation of using it entails the following. For dry snow, the smaller wavelength of the 37 GHz experiences more scattering from the overlaying snow, and therefore produces less brightness temperature [Forster et al., 2001]. As the ice surface or snow becomes wet, the emission at the 37 GHz becomes stronger than 18 GHz [Ulaby, Moore, Fung, 1982]. Hence, HR should be negative. Anderson [1997] developed an algorithm to detect the onset of snow or ice surface melt by monitoring the temporal record from HR using SMMR and SSM/I channels. The difference changes from positive in the case of dry snow in winter to zero or negative at the onset of snowmelt. Instead of using a single threshold on HR, Drobot and Anderson [2001] improved the algorithm of Anderson [1997] by using another scheme of monitoring the difference between the brightness temperature from the 18 GHz SMMR (or 19 GHz SSM/I) and the 37 GHz. The method, called advanced horizontal range algorithm (AHRA), proceeds as follows. HR is denoted ΔTb and calculated daily for any given gridded pixel: ΔT b = T b,18h − T b,37h

(10.15)

If it is greater than 4 K at a given point, then a dry snow winter condition can be assumed. If it is less than –10 K, then liquid water is assumed to be present in the snowpack and the algorithm classifies the day as a possible snowmelt onset date. If the difference is between 4 K and –10 K, the algorithm determines if snowmelt onset has occurred based on a 20-day time series of the difference. In this case the maximum and the minimum differences are

determined within the 10 days prior to the potential melt onset date, and the difference between these two extremes is calculated. The same process is performed for the 9 days after the potential melt date. The former number is subtracted from the latter number, and if the difference (Diff) is greater than 7.5 K (the threshold is based on empirical data), the algorithm assigns melt onset to that particular day. The process can be formulated in the following equation: Diff = max 2ΔT b − min 2ΔT b − max 1ΔT b − min 1ΔT b

(10.16)

where, the prefixes max2 and min2 are the maximum and minimum of ΔTb over the subsequent 9 days, and min1 and min1 are the corresponding extremes over the preceding 9 days. Results from this method compared favorably with other methods developed earlier to retrieve snow melt onset from passive and active microwave data. Drobot and Anderson [2001] show a map of the 20-year snowmelt onset date (1979–1998) in the Arctic (Figure 10.22). Snowmelt begins at the ice edge between Julian day 110 and 130 and progresses northward. Prior to this period snowmelt would appear along coastlines of many seas and bays. The total melt onset area nearly doubles from Julian day 150 through Julian day 170. The latest snow melt onset dates occur in the Lincoln Sea north of Greenland, where MYI is concentrated. The advantage of using PM is the availability of the daily coverage of the entire Arctic Basin, which allows extensive mapping of the onset of melt day. Belchansky, Douglas, Platonov [2004] used a modified version of the AHRA to estimate the regional and interannual variability of melt season in the Arctic from 1979 to 2001. They incorporated surface air temperature, obtained from the International Arctic Buoy Program/ Polar Exchange to circumvent two factors: the anomalous estimates arising from false Tb, and the relative insensitivity of the AHRA to the choice of the thresholds. They presented maps of the average sea ice melt onset, melt duration, and onset of freezing data for the entire Arctic Basin, averaged over two periods: 1979–1988 during a low index Arctic oscillation (AO), and 1989–2001 during a high index AO. The authors found a correlation between the ice melt duration and the seasonal strength of the AO. As expected, melt onset occurs early (first weeks in May) at the far east of the Arctic region (the Kara and Barents Seas) and far west (the Chukchi and Beaufort Seas). The duration of the surface in these two areas continues for 140–180 days, exceeding its duration in the Central Arctic (between 30 and 60 days) and the peripherals of the Central Arctic (between 60 and 140 days). Later, Bliss and Anderson [2018] used basically the same method of Drobot and Anderson [2001] with addition of

RETRIEVAL OF SEA ICE SURFACE INFORMATION 433

melt onset date 90

130

170

Figure 10.22 Mean snowmelt onset averaged from PM data from 1979 through 1998, using the algorithm presented in Drobot and Anderson, 2001. The melt onset is shown in terms of the day of the year (day 190 is July 9) according to the attached gray tone bar [Drobot and Anderson, 2001, Figure 5 / with permission from AGU].

surface air temperature and sea ice concentration data to estimate the onset of ice melt in the Arctic, since 1979. The method estimates the uncertainty of the product and identifies a few major interannual changes of the date of melt onset at key regions. An increasing trend toward early melt onset as large as −9.45 days/decade is found in the east Siberian Sea and −5.59 days/decade for the entire Arctic. Results also indicate that of Arctic-wide melt onset date using just surface air temperature of −1 C, threshold occurs ~11 days later than the estimates using PM data. Despite this bias, correlations between results from the two studies are generally good (≥0.6) for most of the Arctic region. Markus, Stroeve, Miller [2009] used a different approach where three indicators of surface melt onset, all derived from PM observations, are combined. All indicators are sensitive to different features of surface melt. The algorithm that employs this approach is called the PM Algorithm (PMA). A key theme is that emissivity increases sharply when snow becomes wet as moisture absorbs more solar energy, hence emits more microwave energy [Ulaby, Moore, Fung, 1986] (see section 7.9.3.2). The three indicators employed in the PMA are expressed by the following equations: ΔT b,37V = T b,37V i+ 1 − T b,37V i

(10.17)

where, Tb, 37V is the brightness temperature from the 37 GHz channel with vertical polarization and ΔTb,37V

is the absolute difference between observations from day i and the following day i + 1. The premise behind using this equation is that Tb,37V shows a significant increase in temporal variability (up to 30 K) when melt begins. The second indicator is: ΔGRice = GRice i − GRice i+ 1

(10.18)

where, GRice is the spectral gradient ratio between radiation from 37 and 19 GHz channels in vertical polarization, determined using equation (9.2). GRice takes negative values for snow-covered ice in winter (around −0.06 and −0.01 for MYI and FYI, respectively) and zero or positive values during the melt-freeze transition periods [Markus, Stroeve, Miller, 2009]. In order to account for the effect of ice concentration, the brightness temperature from ice in a heterogeneous footprint Tb,ice should be determined from the following equation, which can then be used to determine GRice: T b,ice = T b,obs − 1− C T b,ow C

(10.19)

where Tb,obs is the observed brightness temperature, C is ice concentration, and Tb,ow is the typical brightness temperature from OW. The third indicator is: P = T b,19V + 0 8 T b,37V

(10.20)

For dry snow on MYI, P has values less than 460 K. This value increases at the onset of melt [Smith, 1998].

434

SEA ICE

For FYI, P is greater than 440 K for dry snow and drops below that value at the onset of melt. The PMA method uses the three parameters presented in equations (10.17), (10.18), and (10.20) after normalizing each value by the ranges representing the melt condition. For example, the range for ΔGRice when melt is potentially present is between 0.005 and 0.015. The normalized value of ΔGRice between these two limits is called a melt signature weight, denoted by WΔGRice. Similar signature weights are used for the other two parameters; WΔTb, 37V and WP. The sum of the three weights is calculated for each day: W = W ΔT b,37V + W ΔGRice + WP

(10.21)

The day with the greatest sum is considered to be the melt onset day. Markus, Stroeve, Miller [2009] divided the Arctic Basin into 10 regions and presented data on onset of melt and effective melt for each region. This is a reasonable approach because different regions are subjected to different climatic conditions; hence respond differently to the recent climate change. The NSIDC produces yearly maps of snow melt onset dates over Arctic sea ice from 1979 through 2017. Data were derived from Tb measurements obtained from SMMR, SSM/I, and SSMIS brightness temperature measurements. Statistics for each grid cell are available. This includes mean melt onset date, latest (i.e., maximum), and earliest (i.e., minimum) melt onset date, the difference between maximum and minimum onset dates, and the standard deviation of melt onset dates. One browse image is provided for each statistical field. Data are available through the link http://doi.org/10.5067/ A9YK15H5EBHK.

10.2.1.3. Active Microwave Observations Similar to the PM algorithms, melt onset from radar data is based on the same concept of searching for a change in a temporal daily signature record that identifies the melt onset. Daily coverage of the polar region from scatterometer data is available but at a coarse resolution of a few kilometers at best (section 8.5). On the other hand, SAR has the appropriate fine resolution of ten or hundred meters yet with limited swath, meaning less temporal resolution. However, SAR temporal resolution can be improved by using data from multi-systems or from constellation of same system (e.g., Sentinel-1 from the Canadian RCM). Physically speaking, the onset of the snow melt increases the dielectric constant of the overlaid snow. This results in increasing the backscatter from the FYI surface. On the other hand, since wet snow absorbs most of the incident radar signal, it masks the scattering from the sub-surface bubbly layer of MYI; hence the backscatter from MYI is expected to decrease in this case (Figure 9.18). In estimating onset of melt using SAR, only data from FYI are used. Mahmud et al. [2020] presented a graph of daily evolution of σ ohh from FYI during the freezing season of 2009, where the days of melt onset, pond onset, and break-up (decay) of existing sea ice as well as freeze-up onset in the next season are marked. The thresholds that define those events are shown for the L-, C-, and Ku-bands. The following discussions start with the use of scatterometer data, followed by SAR data. An example of SAR images that show the increase of σ ohh as the onset of melting of FYI progresses is shown in Figure 10.23. The two images of a section in the

Figure 10.23 RADARSAT-2 SAR image of a scene in the Beaufort Sea acquired on 22 April 2016 (left) and 28 April 2016 (right). The scene is covered by FYI with many openings; dark indicates water and white indicates young ice. The average backscatter in the 28 April scene is higher than the 22 April scene by 2.5 dB. The onset of melt fell between these two dates (courtesy of the CIS, RADARSAT2 is © of MDA Ltd., Vancouver, Canada).

RETRIEVAL OF SEA ICE SURFACE INFORMATION 435

Beaufort Sea, acquired on 22 and 28 April 2016, show an increase of 2.5 dB on average in the image of 28 April. This is attributed to snow wetness as the air temperature increased from -9.5 C to nearly -4 C between the two dates [M. Shokr, unpublished data]. The melt onset fell between these two dates. Ice dynamics opened a fresh lead covered with water, which appears in the dark tone in the image of 28 April. A few studies used the high temporal resolution (1–2 days) scatterometer observations to generate maps of melt onset over the Arctic, since the year 2000 [e.g., Howell et al., 2008, Drinkwater and Liu, 2000, Wang et al., 2011, Mortin et al., 2012, Mortin et al., 2014]. Howell et al. [2008] used scatterometer data from QuikSCAT to estimate sea ice onset of melt, freeze and duration of the melt within the Canadian Arctic Archipelago for the period 2000–2007. Their approach entails setting a threshold on the backscatter, but they treated snow on FYI and MYI differently. After confirming the increase of backscatter from FYI and its decrease from MYI as the melt onset approaches, Howell et al. [2008] estimated the date of melt onset for these two types using a threshold of absolute change in σo of greater than 2 dB from the stable winter value for each type. The threshold was estimated based on temporal evolution of σo from numerous sites within the Canadian Arctic Archipelago. The authors reported that changing the threshold by ±1 dB results in a change in the date of the snowmelt onset by only 1 or 2 days, which is considered a small error. Day-of-year 150 (end of May) was found to be the mean date of melt onset for both ice types. Integrated pan-Arctic melt onset from satellite scatterometer and radiometer is presented in Derksen et al. [2008]. Interannual variability of the melt onset in the Arctic Basin between 2000 and 2009 is demonstrated. Attempts to use C-band SAR data to identify the melt onset in the Arctic are presented in Kwok, Cunningham, Nghiem [2003], Yackel et al. [2007], and Mahmud et al. [2016]. Those attempts produced approximate melt onset data within a limited region because of the narrow swath of SAR and its coarse temporal revisits. Even when Mahmud et al. [2016] used 4457 images from RADARSAT-1 and RADARSAT-2 to estimate the melt onset within the northern section of the Canadian Arctic Archipelago, the authors noted that the temporal resolution of the imagery was still a limitation. The mean annual average melt onset date was day 164 (mid-June) with a standard deviation of 4 days and no trend was identified within the 18-year study period (1997–2014). Besides the insufficient temporal resolution, other restrictions hampered the ability of SAR in retrieving information about ice surface melt in general. These include the inconsistent viewing geometries of the sensor and the limited image availability across the Arctic domain. The latter restriction has been alleviated by using the frequent coverage of

Sentinel-1A/B images (almost daily) over the eastern section of the Arctic, which covers the waterways between Greenland and Ellesmere Island and the Lincoln Sea north of Ellesmere Island. Howell et al. [2019] used time series of Sentinel-1A/B and RADARSAT-2 to estimate melt onset over the eastern section of the Canadian Archipelago. The study used the same concept of examining the daily temporal evolution of SAR signature from the Extra Wide and Interferometric Wide mode of Sentinel-1A/B (total of 3109 images), in addition to the less frequent coverage of RADARSAT-2 ScanSAR wide mode (321 images) from March to August in the years 2016 and 2017. Figure 10.24 shows the spatial distribution of the sea ice melt onset in the Arctic Basin section around the Ellesmere Island. Melt onset data in this figure were generated using three sets of input: the SAR combination mentioned above, the ASCAT scatterometer, and the AMSR-2 PM observations. Compared to the use from SAR and scatterometer, PM data have limited use in narrow channels. On the other hand, SAR data provide most details. Note the earlier melt onset (orange color distribution) from SAR data in 2016 and 2017. In 2016 the polynya started at the northern section of Smith Sound, and in 2017 it started at the northern section of Robeson Channel (for the locations of the polynyas see Figure 2.64). Since the IST in the polynya is always higher than that in the stationary ice upstream of the ice arch, the melt onset date is expected to be earlier in the polynya area as confirmed in the map. Daily coverage of SAR can lead to estimation of melt onset day in narrow passages. SAR data in Figure 10.24 are the only data that provide melt onset information in Robeson Channel, which is only 30 km in width. These results highlight the promising use of SAR constellation data, to obtain pan-Arctic maps of ice melt phases at fine spatial resolutions. After the onset of melt, the ponding phase of sea ice surface starts. This phase has been studied and characterized in the Arctic and reported in several publications. Physically speaking, melt pond on FYI is different than that on MYI. Greater variability of pond area is found on FYI, typically between 0.1 in early summer and 0.8 in late summer [Fetterer and Untersteiner, 1998, Yackel, Barber, Hanesiak, 2000]. Less ponding fraction is associated with MYI, typically between 0.2–0.3 [Tschudi, Curry, Maslanik, 2001]. This is mainly due to the thicker snow cover on MYI. However, in a study area within the Resolute Passage in the Canadian Archipelago, Yackel and Barber [2000] found that the melt pond on MYI was 60 cm thick, which was deeper than FYI by about 20 cm. Data were obtained during an aerial survey between Julian days 181 and 184, in 1997. Melt ponds can be identified in SAR data based on their low backscatter if the pond surface is smooth. If roughened by the wind, it would appear as bright as the surrounding

436

SEA ICE

2016 80°0'0''W

60°0'0''W

100°0'0''W

80°0'0''W

60°0'0''W

100°0'0''W

80°0'0''W

60°0'0''W

80°0'0''N

80°0'0''N

100°0'0''W

(c) PMW Melt Onset

(b) ASCAT Melt Onset

(a) R2+S1A Melt Onset

80°0'0''W

80°0'0''W

80°0'0''W

2017

(d) R2+S1AB Melt Onset 100°0'0''W

80°0'0''W

(f) PMW Melt Onset

(e) ASCAT Melt Onset 60°0'0''W

100°0'0''W

80°0'0''W

60°0'0''W

100°0'0''W

80°0'0''W

DOY

60°0'0''W

90–100 100–110

80°0'0''N

80°0'0''N

110–120 120–130 130–140 140–150 150–160 160–170 170–180 180–190 190–200 200–210 80°0'0''W

80°0'0''W

80°0'0''W

Figure 10.24 Spatial distribution of melt onset in the eastern section of the Arctic, defined by the shown latitude and longitude, for the freezing seasons of 2016 and 2017. (a and d) data from multi-sensor SAR, (b and e) data from ASCAT, and (c and f ) data from PM sensors. Note the detailed definition of melt onset from SAR data in the Nares Strait (the long linear passage at the right side of the image) [Howell et al., 2019 / from Elsevier].

ice, which makes its identification difficult. What complicates the identification even further is the difference in backscatter of the pond between the near and far ranges of SAR image (ponds appear darker in the far range). Identification of melt pond in SAR images is strictly an image processing task, where a threshold on backscatter is used to separate the pond from the background ice. Using RADARSAT-1 data from a standard beam (25 m resolution) to study melt pond in the Arctic, Yackel and Barber [2000] have identified pond fraction between 13% and 34% in the Canadian Archipelago area. The authors also confirmed that incidence angle and surface wind speed explained at least 90% of the variability in σ o. Since melt pond area varies between a few square meters to a few hundreds of square meters, airborne cameras (see next section), in addition to commercial highresolution satellite optical sensors [see summary of those sensors in Yang, 2018], can be the best option to identify

small ponds. A few SAR systems can also achieve 1 m resolution with minimum noise such as TerraSAR-X, COSMO-SkyMed, and 1.5–3 m from RADARSAT-2 wide ultra-fine mode. However, those systems have not been used for melt pond detection but a sub-meter helicopter-borne SAR system was used by Kim et al. [2013] to survey melt ponds in the northern Chukchi Sea during summer of 2011. The system mapped the ice surface at 0.3 m resolution and compared the results against coincident information obtained from TerraSAR-X (6m resolution StripMap dual-polarization mode). This system was used successfully to derive fraction, size, and shape data of melt ponds as small as 10 m2 area. The study in Kim et al. [2013] has demonstrated the potential application of future ultra-fine SAR systems in mapping the ice surface melt phases. Such systems can hopefully be developed with wider swath (a few tens of kilometers).

RETRIEVAL OF SEA ICE SURFACE INFORMATION 437

Not much work has been pursued to study melt onset of sea ice in the Antarctic, partly because of the less sensitivity of this ice to climate change but also since there is no impact on operational activities. Drinkwater and Liu [2000] mapped the spatial pattern of melt onset of Antarctic sea ice between 1992 and 1997 using scatterometer data from the C-band (VV) active microwave instrument onboard ERS-1 and ERS-2, and the Ku-band (VV and HH) NSCAT. Antarctic sea ice typically retains snow cover year-round. A comprehensive review of snow on Antarctic sea ice is presented in Massom et al. [2001]. The melt onset in Drinkwater and Liu [2000] was calculated using the difference in backscatter between consecutive images (a few-days step). For each pixel, the melt onset date was determined when the backscatter value decreased from one image to the next by an amount > 0.5 dB, and the total decrease within the seven consecutive images window was >3 dB. The authors did not provide sensitivity of the melt onset day to the selection of the threshold, but they concluded that atmospherically driven surface melt in the Antarctic is far less extensive in time and space than in the Arctic. They also stated that “atmospherically induced surface melting and resulting albedo feedback is not the dominant sea ice removal mechanism. Instead, a combination of ocean heat flux and summer shortwave radiation absorbed by the ocean surface (in small lead fractions) rapidly disposes most of the sea ice cover (p. 1841).” This is an important difference of ice melting process between the Arctic and Antarctic regions. Melt ponds are less frequent in the Antarctic. 10.2.1.4. Airborne Photography Though expensive and of limited coverage, aerial photography is the most effective tool to detect and quantify the ponding phase of sea ice surface because ponds are visible in the photos. Moreover, the very fine resolution of the camera allows capturing a wide range of pond size on a spatial scale from a few meters to hundreds of meters. Digital and video cameras mounted on aircrafts with a motor drive to adjust the viewing angle have been used to survey the field [Rothrock and Thorndike, 1984, Tschudi, Curry, Maslanik, 2001] or helicopters [Perovich, Tucker, III, Ligett, 2002, Peng et al. 2010]. Depending on the camera’s lenses, a photograph that covers roughly 1 km2 can be acquired from an altitude of 2000 m. Photographs can be digitally processed to partition the scene into solid ice surface and pond surface (or leads). This also facilitates the tracking of pond evolution and the determination of their areal fraction and density. To enable quantification of the evolution of melt ponds, a program of aerial photography was carried out at the main site of SHEBA (section 6.6) in the Beaufort Sea between mid-May through early October 1998 [Perovich, Tucker, III, Ligett, 2002]. Helicopter-based digital and video cameras were used to acquire images of the ice

surface. The images were processed to partition the contents into four surface categories: snow cover and bare ice, ponded ice, newly formed young ice, and OW. Partitioning was performed based on selection of color thresholds derived from the color-distribution histograms of the images. Melt pond has typically a bluish appearance, which makes it distinct from the other three categories. The thresholds for the four categories were subjectively selected, but a quality check of the results was performed to ensure that the sum of the area percentages for each feature was nearly within 2% of the ideal 100%. Based on the image processing, Perovich, Tucker, III, Ligett [2002] concluded that melt ponds started in early June and grew rapidly in size to cover 5% to 20% of the imaged ice surface by mid-June. After that, the growth continued at a slower but steady rate until it reached its peak in early August. The two forces that determined the fractional area of the ponds were the rate of surface melt and the drainage of the water to the underlying ocean through cracks and holes. From a series of images obtained from a downward looking camera mounted on the underside of a research aircraft, flown over the SHEBA site, Tschudi, Curry, Maslanik [2001] found that the melt pond areal fraction reached a maximum of 34% on July 26. Obviously, the areal fraction from an airborne camera depends on the viewing geometry of the camera and for that reason the images should be normalized. In another study, Peng et al. [2010] analyzed aerial photographs of Arctic summer sea ice obtained during the Third Chinese National Arctic Research Expedition in 2008 (CHINARE2008). Over 9000 photographs were acquired by a helicopter-mounted Canon G9 camera. Sea ice was undergoing a transition period from the late melting phase to the early freeze-up phase. Using the photographs, the authors could identify three components of the surface: snow-covered sea ice, melt ponds, and open leads. Figure 10.25 shows two photographs acquired on 17 August and 20 August 2008, from an altitude of 113 m and 94 m, respectively. The three surface categories can visually be identified. Melt ponds have distinct shapes and spread across the ice surface forming a large complex network. They appear in nature with a blue turquoise color because of the presence of sea ice under the pond water. The snow-covered ice surface appears bright as expected. Leads appear much darker because of their low surface reflectance (no underlying ice). These shades were used to identify each surface in the color images using simple digital image processing techniques. As expected, the leads in the image of 20 August in Figure 10.25 are larger and more mature. A fraction of each component was estimated based on the differences of the brightness and colors. Aerial photography can provide physical details about ice surface melt, which can be used to enhance estimation of this feature at regional scale from satellite data.

438

SEA ICE

17 Aug. 2008, alt.133 m

20 Aug. 2008, alt. 94 m

Figure 10.25 Aerial photographs showing Arctic sea ice surface (light blue), melt ponds (bluish gray), and fully melted ice, i.e. seawater surface (dark). The photographs were obtained during CHINARE2008 field campaign (Peng et al., 2010 / Reproduced with permission from ELSEVIER).

10.2.2. Frost Flowers Frost flowers are fragile and highly salty ice crystals, which grow to a height of 10–30 mm on the surface of freshly formed sea ice and occasionally freshwater ice [Martin, Ducker, Fort, 1995]. Their origin is water vapor released from the ice surface, the sub-surface up through a porous ice layer, or sublimated directly from the surface. As the warmer wet air near the surface meets the overlying cold air, it becomes supersaturated and thus water may condense immediately under very cold surface air temperatures. This gives rise to frost crystal formation at the surface. The nucleus of the crystal could be any protrusion at the surface. The fragile crystals usually take on dendritic shapes, although they may also grow in rod-like morphologies; hence the name flower is given. Frost flowers typically form if the air temperature is colder than IST by at least 15 C, though this difference depends on the humidity of the air and can be reduced under a humid boundary layer. Frost flowers formation needs also windless conditions. If wind blows at the ice surface, the supersaturated layer is scrubbed from the surface, and the blowing snow also obscures the surface. Flowers are most commonly observed on thin ice types (particularly nilas) because their relatively warm surface allows for the required large temperature gradient within the air layer adjacent to the ice surface. When the ice grows too thick, less heat from the underlying warm water reaches the surface, hence the difference between the surface and air temperatures diminishes. Frost flowers no longer form under this condition. The substantially cold air temperature in the polar

regions enhances the chance of the formation of frost flowers. It is not uncommon to find frost flowers covering the surface of re-frozen leads in the Arctic region. On the other hand, frost flowers are hardly seen on thin ice surface in temperate regions (areas located between 40 –60 to the north-south of the Equator). The mechanism of frost flower formation is illustrated graphically in Figure 10.26. Profiles of atmospheric temperature (T) and vapor density (ρ) above the ice surface are shown, along with the saturation vapor density profile

ρ

ρ

T

sat

Frost flowers

Young sea ice

Figure 10.26 Schematic diagram of the temperature T and humidity (vapor density) ρ in the air above ice surface. The dashed curve indicates the saturation vapor density ρsat(T). A region of supersaturation can develop above the ice when ρ > ρsat. This is where frost flowers can grow.

RETRIEVAL OF SEA ICE SURFACE INFORMATION 439

(ρsat), which is a non-linear function of temperature as given by the following equation [Style and Worster, 2009]: ρsat T = ρsat T ∞

T ML 1 1 exp − − T∞ R T T∞ (10.22)

where, M is the molar mass of water, L is the latent heat of vaporization per unit mass, R is the gas constant, and T∞ is the temperature of the far-field atmosphere. When water vapor from ice reaches the relatively cold air above the ice, a region of supersaturation can develop above the surface, if the vapor density exceeds the saturation density (Figure 10.26). This is where frost flowers can grow around appropriate nuclei. Once formed, they usually wick up brine from the surface and therefore can have an extremely high salinity that may reach 100‰ or more [Martin, Ducker, Fort, 1995]. Frost flower crystals are similar to the hoar ice crystals at the snow base [Style and Worster, 2009] in the way of formation and salinity acquisition from the underlying ice. Ehlert [2012] presents a credible study on growth and melt of frost flowers and their interaction with the underlying sea ice. It included detailed measurements of the morphology, temperature and salinity evolution and their impact on the sea ice during both field and laboratory experiments. The study was conducted in a small Greenlandic Bay (72.79 N, 56.06 W) in March 2010. A photograph of almost one-day old flower is shown in Figure 10.27. Once formed, the flower grows by accumulation of stellar-shaped crystals. The base of the “old” crystals appears darker (hence thicker) than the “young” crystals at the top. It is interesting to note the sea ice surface below the flower, which features milky appearance and much less porosity than the surrounding surface. Ehlert [2012] recorded and presented a complete life cycle of a frost flower. It started to grow on a protrusion at the young ice surface. Its height increased from 0.5 cm to 4.8 cm within 3 hours. The melting process was induced after 16 hours, and the flower collapsed after approximately 22 hours. After 42 hours the flowers had melted completely, leaving a brine pool at the ice surface. According to the study, the source of nucleation of a frost flower is not snow or diamond-dust crystals but only ice platelets protruding at the surface. The author suggests that the physical processes driving the formation of frost flowers are still not fully understood. A photograph of frost flowers on nilas ice surface in the Barents Sea is shown in Figure 10.28. From a geophysical point of view, the extremely high salinity of frost flowers modifies the thermodynamic properties of the underlying ice surface greatly. They act as an insulator layer, making the ice surface in the presence of frost flowers warmer by 1–2 C than the bare

1 cm

Figure 10.27 Photograph of a frost flower that grew on natural sea ice in a bay in the Greenland Sea, on the night of 16 March 2010. Note the different gray shade between the base and the top of the flower, representing thicker and thinner ice crystals, respectively. Note also the small single crystals circled in the lower left corner from which other flowers can grow. The sea ice below frost flowers appeared milky and was almost free of air bubbles [courtesy of I. Ehlert, Max Planck Institute of Meteorology, Hamburg, Germany, from Ehlert, 2012, Figure 2.14].

surface [Martin, Ducker, Fort, 1995]. The size of the flowers and their concentration determine the average temperature of the underlying ice surface, hence the ice temperature under individual frost flower becomes different. Ehlert [2012] studied the morphology of sea ice under frost flowers by carefully sawing out the piece of sea ice on which the flower was grown. Optical and infrared pictures of a slab (2.5 cm width and about 4 cm depth) were taken to examine the structure and temperature fields as shown in Figure 10.29. The sea ice area immediately under a frost flower is delineated by the blue line in the optical photograph. The morphology of the ice in this area looks different than the surrounding ice. While the latter shows signs of associated columnar crystallographic structure (roughly speaking), the structure below the flower shows a random configuration with bright areas that probably indicate the presence of liquid. The size of the funnel in the upper part coincides with the former size of the flower on top of the ice, while the narrow channel to the bottom of the ice marks a possible presence of a brine channel. The infrared photograph reveals that the temperature within the funnel was lower

SEA ICE

Figure 10.28 Frost flowers that bloom on sea ice nilas surface in the Barents Sea. The bluish background is partly due to image processing as well as the reflection of the sky on the ice surface. The inset is a magnification of part of the area, yet at a different location (image courtesy of S. Kern, University of Hamburg).

0

4.0

–0.1 3.5 –0.2 3.0 2.5

–0.4

2.0

–0.5

1.5

–0.6

Temperature (°C)

–0.3

Height (cm)

440

–0.7

1.0

–0.8 0.5 –0.9 0.0 –1 0.0

0.5

1.0

1.5

2.0

2.5

0.5

1.0

1.5

2.0

2.5

Figure 10.29 Optical photograph (left) and infrared photograph (right) of a vertical section of laboratory-grown sea ice under a frost flower (not shown). Colder temperatures in the infrared photograph are correlated with the area under the frost flower, which is delineated by the blue lines in the optical photograph [courtesy of I. Ehlert, Max Planck Institute of Meteorology, Hamburg, Germany, from Ehlert, 2012, Figure 2.16].

RETRIEVAL OF SEA ICE SURFACE INFORMATION 441

by 0.6 C than the surrounding temperature. This suggests that the area under the frost flower was filled with salty liquid that remained at lower temperature because it had a lower freezing point than the less-saltier ice surrounding the flower. The impact of frost flowers on ice thermodynamics is justifiably neglected in weather and climate models. This is because frost flowers may persist for only a few hours or few days on thin ice surface before any wind or possible rise in air temperature wipes their fragile structure. Impacts of frost flowers on atmospheric chemistry are more important. Frost flowers have been recognized as the dominant source of sea salt aerosol in the Antarctic, containing bromine compounds such as bromine monoxide. The ozone-depletion is correlated with high concentration of these compounds. It has been proposed that frost flowers are part of the cause of tropospheric ozone-depletion events during the polar sunrise [Rankin, Wolff, Martin, 2002, Kaleschke et al., 2004]. With regard to microwave remote sensing, the interest in frost flowers follows from their modulation of the measured emission and radar backscatter. In PM imagery the highly saline frost flowers cause the brightness temperature to decrease, since the saline flowers are lossy medium to the emitted radiation from the underlying ice. However, the polarization ratio of the measurements is maintained within the same range as of flower-free surface [Grenfell and Perovich, 1994]. This means that the structure of the flower crystals does not affect the polarization of the emitted radiation. Therefore, frost flowers do not have an adverse effect on the estimation of ice concentration or any other parameter that can be estimated based on the polarization difference in PM data. The situation is different with the radar observations. Sea ice areas covered with frost flowers produce significantly higher backscatter than surrounding areas. This is because of the higher salinity and surface roughness in this case. An attempt to model the microwave scattering from frost flowers is presented in Pimsamarn [1997]. The remarkably high backscattering from frost flowers in SAR images and the fine resolution of those images make SAR the most suitable sensor for identification of frost flowers. A first attempt to characterize the radar backscattering from frost flowers was presented in Nghiem et al. [1997]. The authors measured the C-band backscatter from frost flowers growing on a surface of new ice in a freezing pool using a polarimetric SAR system. The frost flowers grew to 10–15 mm in clusters on top of a slush layer. The study suggested that an increase of approximately 5 dB can be used as an index to mark the areal coverage of frost flowers on new ice. A few polarimetric parameters were examined and Nghiem et al.

[1997] concluded that the flower-covered ice surface had a higher co-polarized backscatter ratio compared to the bare ice surface, but no systematic trend was observed in the cross-polarized backscatter ratio. This is understood since cross-polarization radar scattering is a manifestation of MB mechanism (section 7.6.1.3), of which rough frost flowers has no contribution. Isleifson et al. [2010] measured backscatter of newly formed sea ice during the Canadian Arctic Shelf Exchange Study (CASES) project in the Cape Bathurst polynya in the south eastern limit of the Beaufort Sea in the Canadian Arctic. They used a C-band polarimetric scatterometer system mounted on the port side of the Canadian Coast Guard ice breaker Amundsen at a height of about 8 m from the sea surface. Time series measurements were obtained from one particular station beginning at a time when snow was covering frost flowers and running through the time after the snow was removed and the frost flowers were exposed. Results of co- and cross-polarization backscattering measurements are shown in Figure 10.30. In absence of frost flowers, there is large separation between the two co-polarization signatures at all incidence angles with σ 0vv always a few decibels higher. Without frost flowers more rapid decrease of backscatter from all channels with incidence angle is observed. In the presence of the flowers, the backscatter from both co-polarization channels is effectively the same while it is several decibels lower from the crosspolarization channel as expected. It can be concluded that frost flowers increase the backscatter by 4.5, 8.3, and 7.0 dB from σ 0hh, σ 0vv, and σ 0hv, respectively. In the absence of frost flowers, there is evidently more rapid decrease of backscatter with incidence angle. Isleifson et al. [2010] also observed that when the ice was covered with frost flowers, ice thickness increased by only 1 cm during a 12-hour period. When the frost flowers were removed, it increased by 4 cm during the same period. This is attributed to the insulating effect of frost flowers caused by their high salinity and void contents. In a later study, Isleifson et al. [2014]) measured backscattering from laboratory-grown frost flowers in an outdoor pool using a surface-based polarimetric C-band scatterometer. The study presents a detailed description of the growth stages of frost flowers and a synopsis of the polarimetric signature at each stage. The stages are defined as initial formation, surface brine expulsion, frost flowers growth, and decimation. The initial formation stage lasted between 6 and10 hours. By the end of that period the surface was completely covered with frost flowers of an average height of 5 mm and the salinity of the ice surface was 10‰. Backscattering from all polarimetric

442

SEA ICE

channels was low at the beginning of this stage but increased sharply when frost flowers covered most of the area. For example, σ 0hh and σ 0vv increased from about −25 to −10 dB, while σ 0hv increased from −37 to −29 dB, all measured at an incidence angle of 25 . During the second stage that lasted between 10 and 15 hours, dark patches were observed on a few spots, and they grew to cover nearly the entire ice surface. The authors interpreted this as the area where brine was expelled upward to penetrate the frost flowers. The surface salinity was doubled after about 15 hours, and polarimetric backscattering measurements underwent a rapid decrease for all incidence angles until a local minimum occurred shortly after most of the frost flowers surface was covered with brine. This is perhaps the same stage at which Nghiem et al. [1997] observed a slush layer between the frost flowers as mentioned above. In the third stage of frost flower growth (hour 15–38) the flowers developed to an average height of 15 mm with a dendritic appearance. Their salinity reached a peak of 52‰ at hour 22. Polarimetric backscattering measurements showed stable values that varied with the incidence angle. For 25 incidence angle, σ 0hh and σ 0vv were almost equal (-16 dB and -19 dB, respectively) while σ 0hv was around -20 dB. The fourth stage of the decimation is characterized by a reduction of the frost flower height from 15 to 5 mm, occurring after hour 38. Warm air temperature and increasing solar radiation were responsible for the eventual elimination of the flower crystals.

(a)

10.3.1. Polynya Identification and Properties In section 2.9.1, polynya is considered an ice regime in the sense that it holds certain features that uniquely identify it. This includes limited ice types (only thin ice types exist), high ice reproduction, high ice mobility, and considerable sea water areal percentage. Basically, polynya is a regime of open drift ice with concentration that does not exceed 0.7. The two mechanisms of polynya formation are also presented in section 2.9.1. In that section an important note is made about the unfeasibility of polynya formation even in the presence of one or both of these two qualifying mechanisms. The polynya formation is not feasible if a source of ice flux continues to feed into the polynya area. For a polynya to form in this case, a natural obstruction must exist to block the ice flux. This may be considered as a sufficient condition for the polynya formation, in addition to the one or both of the aforementioned necessary conditions (strong wind and/or warm upwelling ocean water). In this section, retrieval of polynya extent, heat flux (HF) to the atmosphere and ice production (IP) are summarized. Nevertheless, polynya pixels must be identified first. Arch information is shown to be an important element to improve the polynya information retrieval. There are two criteria to identify polynya pixels in remote sensing data, both involve setting a threshold.

(b) 0

0

σ W σ VH σ

–5 –10

–10

HH

–15 –20 –25 –30 –35

–20 –25 –30 –35 –40

–45

–45 20

30

40

50

Incidence angle (deg.)

60

70

HH

–15

–40

–50 10

σ W σVH σ

–5

NRCS [dBm2/m2]

NRCS [dBm2/m2]

10.3. METEOROLOGICALLY DRIVEN SURFACE FEATURES

–50

10

20

30

40

50

60

Incidence angle (deg.)

Figure 10.30 Angular variation of backscatter coefficient measured using a ship-based C-band scatterometer. Measurements were from Arctic thin ice with (a) frost flowers at the surface and (b) after removal of frost flower. Third-order polynomials are fitted to the data points. The dashed line is the noise floor of the system at −40 dB [Isleifson et al., 2010, Figure 6 / with permission from IEEE).

70

RETRIEVAL OF SEA ICE SURFACE INFORMATION 443

The first sets a threshold on sea ice concentrations [Cavalieri and Martin, 1994, Barber et al., 2010, Cheng et al., 2019, Preußer et al., 2015] and the second on thin ice thickness (TIT) [Yu and Rothrock, 1996, Adams et al., 2013]. The premise of using ice concentration is justified since polynya should be associated with low concentration (typically < 0.7). The use of TIT is justified because polynya is mostly covered by thin ice (typically 5 cm). Otherwise, results may not be accurate because the observed emitted radiation becomes influenced by snow properties other than depth (section 7.9.3.2). For example, the model in equation (10.29) does not account for wet snow conditions which make the emissivity from both 19 and 37 GHz equal (approaching 1.0). In this case the gradient ratio approaches zero regardless of the snow depth. The increase of correlation length of grain size (an indicator of grain size) causes GR37V19V to become more negative [Powell et al., 2006], while densification of a snowpack causes GR37V19V to become less negative. Moreover, results from the model are not expected to be accurate in the case of snow-free or a very thin snow layer on the ice surface because the gradient ratio is almost zero from both channels. Retrieval of snow depth on MYI is difficult because of the ambiguity between the emitted signal from the ice surface and the snow cover. Markus et al. [2006] flagged MYI from the calculation of snow depth in the Arctic using AMSR-E data. Similarly, the operational snow-depth product from MODIS is generated only for the seasonal sea ice in the Arctic. In general, snow-depth algorithms using PM observations can be improved if the above-mentioned snow properties/processes are accounted for, namely density, grain size, hoar layer structure, wetness, presence of slush, and ice layering. A theoretical microwave modeling approach seems to be suitable to parameterize these effects, but a comprehensive study is yet to be accomplished. Even with the availability of a comprehensive model, its application will require input of parameters from the observed snow such as wetness and grain size. Coincident data of these parameters may not be available. In an attempt to study the uncertainties of radiometerbased snow-depth algorithms, Rostosky et al. [2020] explored the influence of geophysical parameters from ice, snow, and atmospheric properties on the depth retrievals using a Monte Carlo uncertainty estimation. MEMLS, SNOWPACK, and the Passive and Active Microwave TRAnsfer models were used. Simulations were based on in-situ observations obtained during the Norwegian Young Ice Experiment (N-ICE2015). The average uncertainty in potential snow-depth retrievals was found to be between 11% and 19% and increases with increasing snow-depth. For retrievals using lowerfrequency PM data (including 6.9 GHz), the strongest source of uncertainty were the unknown snow properties (because of the large penetration depth), while for

higher-frequency retrievals (including 36.5 GHz), the contribution of ice, snow properties, and clouds were equally strong sources of uncertainty. Li, Chen, Guan [2021] used MEMLS model to explore the sensitivity of Tb from the Chinese FengYun-3B/ MicroWave Radiometer Imager (FY3B/MWRI) to snow parameters and presented graphs of GR10.7V36.5V versus snow depth for different snow conditions. A graph is shown Figure 10.33 for data from FYI and MYI, with a variety of snow conditions identified by the codes at the side bar. Snow conditions (density, particle size, and correlation length) in relation to the color code shown in the figure are presented in Table 10.1 for FYI and MYI. Li, Chen, Guan [2021] considered two models of snow layers. In the first, a single layer is used and in the second the snow was layered every 5 cm. In the former case, the density and diameter of particles varied, while in the latter the same values as specified in Tonboe, Andersen, Pedersen [2006] were used. A few observations can be drawn from Figure 10.33. For FYI, the simulated GR10.7V36.5V is not sensitive to snow depth when the particle size is minimum (0.5 mm) (D1 set code). This means that snow depth cannot be estimated using this gradient ratio when the snow is fresh. For example, a 20 cm of fresh snow cannot be estimated shortly after its fall on the surface. Once on the surface, snow crystals metamorphose into different shapes influenced by wind, freeze-thaw, and sublimation. The figure shows the sensitivity of GR10.7V36.5V to thickness when snow grain size is 1.0 mm or higher. In this case, the sensitivity of GR10.7V36.5V to snow depth is effective up to a depth of approximately 40 cm. The situation is marginally different for snow on MYI. Variation in snow density does not affect snow-depth retrieval within the range 0–40 cm. Beyond this range, an increase in snow density causes increase of GR10.7V36.5V , but the values become stable, hence snow depth cannot be inferred. Figure 10.33 also shows the non-linear effects of the combination of the three parameters: particle size, snow density, and correlation length on the snow-depth estimates (see the codes in Table 10.1 in connection to Figure 10.33). The effect of snow wetness is not included in this dataset. Nevertheless, for the results to be useful for climate modeling, the wetness must be added as an input to the simulation. Snow depth could also be related to ice surface roughness. Sturm et al. [2006] utilized this relation to generate approximate maps of snow depth by identifying three ice types in aerial photographs in the Alaskan Arctic: smooth, rough, and moderately deformed ice. The snow depth was then inferred from roughness statistics related to these three types. Combined data from radar and laser altimeter system have been used to estimate snow depth. Radar altimeter

RETRIEVAL OF SEA ICE SURFACE INFORMATION 447 (a)

Simuated (GR10.7V36.5V)

0.05

D1–1 D1–2 D1–3 D1–4 D2–1 D2–2 D2–3 D2–4 D3–1 D3–2 D3–3 D3–4 D4–1 D4–2 D4–3 D4–4 5CM

0.00 –0.05 –0.10 –0.15 –0.20 –0.25 –0.30 –0.35 0

10

20

(b)

30 40 50 Snow depth (cm) -FYI

60

70

Simuated (GR10.7V36.5V)

0.05

D1–1 D1–2 D1–3 D1–4 D2–1 D2–2 D2–3 D2–4 D3–1 D3–2 D3–3 D3–4 D4–1 D4–2 D4–3 D4–4 5CM

0.00 –0.05 –0.10 –0.15 –0.20 –0.25 –0.30 –0.35

0

10

20

30 40 50 Snow depth (cm) -MYI

60

70

Figure 10.33 Variation of simulated GR10.7V36.5V versus snow depth on FYI (top) and MYI (bottom) for different snow conditions identified by the colors and the codes shown at the side. Explanation of the codes is presented in Li, Chen, Guan [2021], from which this figure is adapted.

Table 10.1 Snow properties associated with the color codes that appear in Figure 10.33. These are the settings for the MEMLS model used in Li et al. [2021]. First-year ice

Multi-year ice

Code

Particle size (mm)

Snow density (kg/m3)

Correlation length (mm)

Particle size (mm)

Snow density (kg/m3)

Correlation length (mm)

D1-1 D1-2 D1-3 D1-4 D2-1 D2-2 D2-3 D2-4 D3-1 D3-2 D3-3 D3-4 D4-1 D4-2 D4-3 D4-4

0.5 0.5 0.5 0.5 1.0 1.0 1.0 1.0 1.5 1.5 1.5 1.5 2.0 2.0 2.0 2.0

200 250 300 350 200 250 300 350 200 250 300 350 200 250 300 350

0.078 0.073 0.067 0.062 0.156 0.145 0.135 0.124 0.235 0.218 0.202 0.185 0.313 0.291 0.269 0.247

0.5 0.5 0.5 0.5 1.2 1.2 1.2 1.2 1.8 1.8 1.8 1.8 2.5 2.5 2.5 2.5

200 260 330 400 200 260 330 400 200 260 330 400 200 260 330 400

0.078 0.072 0.064 0.056 0.188 0.172 0.154 0.135 0.281 0.258 0.230 0.203 0.390 0.358 0.320 0.282

448

SEA ICE

measures the heights from the satellite orbit to the bottom of the snow cover, while laser altimeter measures heights to the top of the snow cover. Differencing the two heights provides the snow depth. The advantage of this method is the retrieval independence of the snow properties because the received signal is a function of the distance from the sensor to the reflective surface only. However, there is weak dependence of the position of the radar reflective layer within the snow, on snow properties. This new approach is presented in Kwok et al. [2020] with results of snow depth covering the period between 14 October 2018 and the end of April 2019. Snow depth ranges from 9 cm to 30 cm, with depth on MYI were found to be 2-3 times the depth on FYI. Compared to the reconstruction of the snow depth from ERA5, snow depth is systematically underestimated from the altimeter systems. However, spatial pattern and gradient compares well between the two methods. Sensitivity to space–time sampling was examined because the freeboard from the two altimeter satellites was not measured simultaneously. Impacts of ice deformation and biases of radar altimeter measurements due to scattering within the snow are also discussed in Kwok et al. [2020]. Only monthly averaged snow depth can be produced using combination of altimeters approach, which is useful as a climatic, but not as operational, dataset. 10.4. REFERENCES Adams, S.et al. (2013) Improvement and sensitivity analysis of thermal thin-ice thickness retrievals, IEEE Transactions on Geoscience and Remote Sensing, 51(6), pp. 3306–3318. Anderson, M.R. (1997) Determination of a melt onset date for Arctic sea ice region using passive microwave data, Annals of Glaciology, 25, pp. 382–387. Arkett, M. et al. (2008) Evaluating ALOS-PALSAR for ice monitoring: What can L-band do for the North American ice services?, Proceedings of International Geoscience and Remote Sensing Symposium (IGRASS), 5, pp. 188–191. Bailey, E., Feltham, D.L. and Sammonds, P.R. (2010) A model for the consolidation of rafted sea ice, Journal of Geophysical Research, 115(C004015). Available from: doi:10.1029/ 2008JC005103. Barber, D.G. et al. (2001) Sea-ice and meteorological conditions in Northern Baffin Bay and the North Water Polynya between 1979 and 1996, Atmosphere–Ocean, 39, pp. 343–359. Barber, D.G. et al. (2010) Physical processes within the North Water (NOW) polynya, Atmosphere–Ocean, 39(3), pp. 163–166. Barry, R.G. et al. (1989) Characteristics of Arctic sea ice from remote sensing data and their relationships to atmospheric processes, Annals of Glaciology, 12, pp. 9–15. Belchansky, G.I., Douglas, D.C. and Platonov, N.G. (2004) Duration of the Arctic sea ice melt season: Regional and interannual variability, 1979–2001, Journal of Climate, 17, pp. 67–80.

Bliss, A.C. and Anderson, M.R. (2018) Arctic sea ice melt onset timing from passive microwave-based and surface air temperature-based methods, Journal of Geophysical Research: Atmosphere, 123, pp. 9063–9080. Bouillon, S. and Rampal, P. (2015) On producing sea ice deformation data sets from SAR-derived sea ice motion, The Cryosphere, 9, pp. 663–673. C-Core (2012) Delineation of extreme ridges in high resolution satellite-based radar imagery, C-Core Report No. R-12-028941, Revision 1.0. Cavalieri, D.J. and Martin, S. (1994) The contribution of Alaskan, Siberian, and Canadian coastal polynyas to the cold halocline layer of the Arctic Ocean, Journal of Geophysical Research, 99(C9), pp. 18,343–18,362. Cheng, Z. et al. (2019) Heat flux sources analysis to the Ross Ice Shelf polynya ice production time series and the impact of wind forcing, Remote Sensing, 11, p. 188. Comiso, J.C., Cavalieri, D.J. and Markus, T. (2003) Sea ice concentration, ice temperature, and snow depth using AMSR-E data, IEEE Transactions on Geoscience and Remote Sensing, 41(2), pp. 243–252. De Abreu, R., Arkett, M. and Ramsay, B. (2001) Monitoring the seasonal decay of first year sea ice at the Canadian Ice Service with NOAA AVHRR, Proceedings of 11th Conference on Satellite Meteorology and Oceanography, Madison, Michigan, October 15–18. Derksen, C. et al. (2008) Integrated pan-Arctic melt onset detection from satellite scatterometer and radiometer measurements, Journal of Geophysical Research, 116(D22103). Dierking, W. (2013) Sea ice monitoring by synthetic aperture radar, Oceanography, 26(2), pp.100–111. Available from: http://dx.doi.org/10.5670/oceanog.2013.33. Dierking, W. and Busche, T. (2006) Sea ice monitoring by L-band SAR: An assessment based on literature and comparisons of JERS-1 and ERS-1 imagery, IEEE Transactions on Geoscience and Remote Sensing, 44(2), pp. 957–970. Drinkwater, M.R. and Liu, X. (2000) Seasonal to interannual variability in Antarctic sea-ice surface melt, IEEE Transactions on Geoscience and Remote Sensing, 38(4), pp. 1827–1842. Drobot, S.D. and Anderson, M.R. (2001) An improved method for determining snowmelt onset dates over Arctic sea ice using scanning multichannel microwave radiometer and Special Sensor Microwave/Imager data, Journal of Geophysical Research, 106, pp. 24,033–24,049. Ehlert, I. (2012) On the evolution of first-year sea ice, Reports on Earth System Science 122-2012, Max Planck Institute for Meteorology, Hamburg, Germany, ISSN 1614–1199. Emmert-Streib F. et al. (2020) An introductory review of deep learning for prediction models with big data, Frontier of Artificial Intelligence, 3(4). Available from: doi:10.3389/ frai.2020.00004. Feltham, D.L. (2008) Sea ice rheology, Annual Review of Fluid Mechanics, 40, pp. 91–112. Fetterer, F. and Untersteiner, N. (1998) Observations of melt ponds on Arctic sea ice, Journal of Geophysical Research, 103(C11), pp. 24,821–24,835. Forster, R.R. et al. (2001) The onset of Arctic sea-ice snowmelt as detected with passive and active microwave remote sensing, Annals of Glaciology, 33, pp. 85–93.

RETRIEVAL OF SEA ICE SURFACE INFORMATION 449 Grenfell, T.C. and Perovich, D.K. (1994) Analysis of surface based passive microwave observations during LEADEX 1992, Proceedings of Geoscience and Remote Sensing Symposium (IGARSS’94), 2, pp. 1005–1007. Haid, V. and Timmermann, R. (2013) Simulated heat flux and sea ice production at coastal polynyas in the southwestern Weddell Sea, Journal of Geophysical Research: Oceans, 118 (C5), pp. 2640–2652. Herman, A. and Glowacki, O. (2012) Variability of ice deformation rates in the Arctic and their relationship with basin-scale wind forcing, The Cryosphere, 6, pp. 1553–1559. Howell, S.A. et al. (2008) Changing sea ice melt parameters in the Canadian Arctic Archipelago: Implications for the future presence of multiyear ice, Journal of Geophysical Research, 113(C09030). Available from: doi:10.1029/2008JC004730. Howell, S.E.L. et al. (2019) Estimating melt onset over Arctic sea ice from time series multi-sensor Sentinel-1 and RADARSAT-2 backscatter, Remote Sensing of Environment, 229, pp. 48–59. Hutchings, J.H., Heil, P. and Hilber, W.D. (2005) Modeling linear kinematic features in sea ice, Monthly Weather Review, 132(12), pp. 3481–3497. Hutter, N., Zampieri, L., Losch, M. (2019) Leads and ridges in Arctic sea ice from RGPS data and a new tracking algorithm, The Cryosphere, 13, pp. 627–645. Available from: https://doi. org/10.5194/tc-13-627-2019. Isleifson, D. et al. (2010) C-band polarimetric backscattering signatures of newly formed sea ice during fall freeze-up, IEEE Transactions on Geoscience and Remote Sensing, 48(8), pp. 3256–3267. Isleifson, D. et al. (2014) A study on the C-band polarimetric scattering and physical characteristics of frost flowers on experimental sea ice, IEEE Transactions on Geoscience and Remote Sensing, 52(3), pp. 1787–1798. Johannessen, O. et al. (2007) Remote sensing of sea ice in the Northern Sea route: Studies and applications, Springer-Praxis Books in Geophysical Sciences, ISBN-10: 3-540-24448-4. Kaleschke, L. et al. (2004) Frost flowers on sea ice as a source of sea salt and their influence on tropospheric halogen chemistry, Geophysical Research Letters, 31(L16114). Available from: doi:10.1029/2004GL020655. Kim, D. et al. (2013) Melt pond mapping with high-resolution SAR: The first view, Proceedings of IEEE, 101(3), pp. 748–758. König-Langlo, G. and Augstein, E. (1994) Parameterization of the downward long-wave radiation at the Earth’s surface in polar regions, Meteorologische zeitschrift, N.F.3, Jg. 1994, H. 6, 3, pp. 343–347. Kunzi, K.F., Pail, S. and Rott, H. (1982) Snow-cover parameters derived from Nimbus-7 scanning multichannel microwave radiometer (SMMR) data, IEEE Transactions on Geoscience and Remote Sensing, GE-20, pp. 57–66. Kwok, R. (2003) RGPS Arctic Ocean sea ice deformation from SAR ice motion: Linear kinematic features Winter 1996–1997, Winter 1997–1998, Summer 1998, Polar Remote Sensing Group, Jet Propulsion Laboratory Report, JPL D-21524, 80 p. Kwok, R. (2010) Satellite remote sensing of sea-ice thickness and kinematics: A review, Journal of Glaciology, 56(200), pp. 1129–1140.

Kwok, R., Cunningham, G.F. and Nghiem, S.V. (2003) A study of the onset of melt over the Arctic Ocean in RADARSAT synthetic aperture radar data, Journal of Geophysical Research, 108. Kwok, R. et al. (1998) Sea ice motion from satellite passive microwave imagery assessed with ERS SAR and buoy motions, Journal of Geophysical Research, 103(C4), pp. 8191–8214. Kwok, R. et al. (2009) Thinning and volume loss of the Arctic Ocean sea ice cover: 2003–2008, Journal of Geophysical Research, 114(C07005). Available from: doi:10.1029/ 2009JC005312. Kwok, R. et al. (2020) Arctic snow depth and sea ice thickness from ICESat-2 and CryoSat-2 freeboards: A first examination, Journal of Geophysical Research: Oceans, 125(3), e2019JC016008. Lewis, L. et al. (1996) Springtime sensible heat, nutrients and phytoplankton in the Northwater Polynya, Canadian Arctic, Continental Shelf Research, 16(14), pp. 1775–1792. Li, L., Chen, H. and Guan, L. (2021) Retrieval of snow depth on arctic sea ice from the FY3B/MWRI, Remote Sensing, 2021, 13(8), p. 1457. Lindsay, R.W. and Rothrock, D.A. (1995) Arctic sea ice leads from advanced very high resolution radiometer images, Journal of Geophysical Research, 100, pp. 4533–4544. Lindsay, R.W. and Stern, H.L. (2003) The RADARSAT Geophysical Processor System: Quality of sea ice trajectory and deformation, Atmospheric Oceanic Technology, 20, pp. 1333–1347. Linow, S. and Dierking, W. (2017) Object-based detection of linear kinematics features in sea ice, Remote Sensing, 9(5), p. 493. Available from: https://doi.org/10.3390/rs9050493. Mahmud, M.S. et al. (2016) Detection of melt onset over the northern Canadian Arctic Archipelago sea ice from RADARSAT, 1997–2014, Remote Sensing Environment, 178(59–69), 56, pp. 6686–6698. Mahmud, M.S. et al. (2020) Seasonal evolution of L-band SAR backscatter over landfast Arctic sea ice, Remote Sensing of Environment, 251, p. 11,2049. MANICE (2005) Manual of standard procedures for observing and reporting ice conditions, Canadian Ice Service– Environment Canada, ISBN-10: 0-660-62858-9, Catalogue No. EN56-175/2005. Markus, T. and Cavalieri, D.J. (1998) Snow depth distribution over sea ice in the Southern Ocean from satellite passive microwave data, In: Jeffries, M.O., ed. Antarctic sea ice: physical processes, interactions, and variability, Antarctic Research Series, Vol. 74, American Geophysical Union, pp. 19–39. Markus, T., Stroeve, J.C. and Miller, J. (2009) Recent changes in Arctic sea ice melt onset, freeze-up, and melt season length, Journal of Geophysical Research, 114(C12024). Available from: doi:10.1029/2009JC005436. Markus, T. et al. (2006) Microwave signature of snow on sea ice: observations, IEEE Transactions on Geoscience and Remote Sensing, 44(11), pp. 3081–3090. Martin, S., Ducker, R. and Fort, M. (1995) A laboratory study of frost flower growth on the surface of young sea ice, Journal of Geophysical Research, 100, pp. 7027–7036.

450

SEA ICE

Massom, R.A. et al. (2001) Snow on Antarctic sea ice, Reviews of Geophysics, 39(3), pp. 413–445. Maykut, G.A. (1978) Energy exchange over young sea ice in the Central Arctic, Journal of Geophysical Research, 83(C7), pp. 3646–3658. McLaren, A.S., Serreze, M.C. and Barry, R.G. (1987) Seasonal variation of sea ice motion in the Canada basin and their implications, Geophysical Research Letters, 14, pp. 1123–1126. Melior, M. (1986) Mechanical behavior of sea ice, In: Untersteiner, N., ed. The geophysics of sea ice, NATO ASI Series, Series B, Physics, New York: Plenum Press, 146, pp. 165–281. Melling, H. (1998) Detection of features in first-year pack ice by synthetic aperture radar (SAR), International Journal of Remote Sensing, 19, pp. 1223–1249. Miao, X. et al. (2016) Object-based Arctic sea ice ridge detection from high spatial resolution imagery, IEEE Transactions on Geoscience and Remote Sensing, 13(6), pp. 787–791. Miles, M.W. and Barry, R. (1998) A 5-year satellite climatology of winter sea ice leads in the western Arctic, Journal of Geophysical Research: Oceans, 103(C10), pp. 21,723–21,734. Moore, C.W. et al. (2014) Convective forcing of mercury and ozone in the Arctic boundary layer induced by leads in sea ice’, Nature, 12924. Available from: doi:10.1038/ nature12924. Mortin, J. et al. (2012) Mapping of seasonal freeze-thaw transitions across the pan-Arctic land and sea ice domains with satellite radar, Journal of Geophysical Research, 117(C8), C08004. Mortin, J. et al. (2014) Extending the QuikSCAT record of seasonal melt–freeze transitions over Arctic sea ice using ASCAT, Remote Sensing of Environment, 141, pp. 214–230. Mortin, J. et al. (2016) Melt onset over Arctic sea ice controlled by atmospheric moisture transport, Geophysical Research Letters, 43, pp. 6636–6642. Murashkin, D. and Spreen, G. (2019) Sea ice leads detected from Sentinel-1 SAR images, Proceedings of IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 174–177. Murashkin, D. et al. (2018) Method for detection of leads from Sentinel-1 SAR images, Annals of Glaciology, 59, pp. 124– 136. Available from: https://doi.org/10.1017/aog.2018.6. Nihashi, S., Ohshima, K.I. and Tamura, T. (2017) Sea-ice production in Antarctic Coastal Polynyas estimated from AMSR2 data and its validation using AMSR-E and SSM/ I-SSMIS data, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 10(9), pp. 3912–3922. Nghiem, S.V. et al. (1997) A laboratory study of the effect of frost flowers on C band radar backscatter from sea ice, Journal of Geophysical Research, 102(C2), pp. 3357–3370. Ohshima, K.I., Nihashi, S. and Iwamoto, K. (2016) Global view of sea-ice production in polynyas and its linkage to dense/bottom water formation, Geoscience Letters, 3(1), p. 13. Ohshima, K.I. et al. (2020) Estimation of sea ice production in the Bering Sea from AMSR-E and AMSR2 data, with special emphasis on the Anadyr Polynya, Journal of Geophysical Research: Oceans, 125(7), e2019JC016023. Available from: https://doi.org/10.1029/2019JC016023

Onana, V. et al. (2013) A sea-ice lead detection algorithm for use with high-resolution airborne visible imagery, IEEE Transactions on Geoscience and Remote Sensing, 51(1), pp. 38–56. Parkinson, C.L. and Washington, W.M. (1979) A large-scale numerical model of sea ice, Journal of Geophysical Research: Oceans, 84(C1), pp. 311–337. Peng, L. et al. (2010) Sea ice surface features in Arctic summer 2008: Aerial observations, Remote Sensing of Environment, 114(4), pp. 693–699. Perovich, D.K., Tucker, III, W.B. and Ligett, K.A. (2002) Aerial observations of the evolution of ice surface conditions during summer, Journal of Geophysical Research, 107(C10), pp. 24.1–24.14. Pimsamarn, K. (1997) Electromagnetic scattering model for saline ice covered with frost flowers, ME thesis, Department of Electrical Engineering and Computer Science, Massachusetts Institute of Technology. Polashenski, C., Perovich, D., Courville, Z. (2012) The mechanisms of sea ice melt pond formation and evolution, Journal of Geophysical Research, 117(C01001). Available from: 10.1029/ 2011JC007231. Powell, D.C. et al. (2006) Microwave signature of snow on sea ice: Modeling, IEEE Transactions on Geoscience and Remote Sensing, 44(11), pp. 3091–3102. Preußer, A. et al. (2015) Multi-decadal variability of polynya characteristics and ice production in the North Water Polynya by means of passive microwave and thermal infrared satellite imagery. Remote Sensing, 7(12), pp. 15,844–15,867. Available from: https://doi.org/10.3390/rs71215807. Qu, M. et al. (2021) Spring leads in the Beaufort Sea and its interannual trend using Terra/ MODIS thermal imagery, Vol. 256, p. 11,2342. Rankin, A.M., Wolff, E.W. and Martin, S. (2002) Frost flowers: Implications for tropospheric chemistry and ice core interpretation, Journal of Geophysical Research, 107(D23), AAC 4–1–15. Reiser, F., Willmes, S. and Heinemann, G. (2020) A new algorithm for daily sea ice lead identification in the Arctic and Antarctic winter from thermal-infrared satellite imagery, Remote Sensing, 12(12), 1957. Available from: doi:10.3390/ rs12121957. Röhrs, J. and Kaleschke, L. (2012) An algorithm to detect sea ice leads using AMSR-E passive microwave imagery, Cryosphere, 6, pp. 343–352. Rösel, A. (2013) Detection of melt ponds on Arctic sea ice with optical satellite data, New York, Dordrecht, London: Springer Heidelberg. Rostosky, P. et al. (2018) Snow depth retrieval on Arctic sea ice from passive microwave radiometers—Improvements and extension to multiyear ice using lower frequency, Journal of Geophysical Research: Oceans, 123, pp. 7120–7138. Available from: doi.org/10.1029/2018JC014028. Rostosky, P. et al. (2020) Modeling of the microwave emission of snow on Arctic sea ice for estimating the uncertainty of satellite retrievals, Journal of Geophysical Research: Oceans, 125, 3, e2019jc015465. Rothrock, D.A. (1986) Ice thickness distribution measurements and theory, In: Untersteiner, N., ed. The Geophysics of Sea

RETRIEVAL OF SEA ICE SURFACE INFORMATION 451 Ice, NATO ASI Series, Vol. 146, New York: Plenum Press, pp. 551–575. Rothrock, D.A. and Thorndike, A.S. (1984) Measuring the sea ice floe size distribution, Journal of Geophysical Research, 89, pp. 6477–6486. Serreze, M.C. and Barry, R.G. (2011) Processes and impacts of Arctic amplification: A research synthesis, Global Planet Change, 77, pp. 85–96. Shokr, M. and Kaleschke, L. (2012) Impact of surface conditions on thin sea ice concentration estimate from passive microwave observations, Remote Sensing of Environment, 121, pp. 36–50. Shokr, M. et al. (2022) Observations from C-band SAR fullypolarimetric parameters of mobile sea ice based on radar scattering mechanisms to support operational sea ice monitoring, Canadian Journal of Remote Sensing, 48(2), pp. 197–213. Smith, D.M. (1998) Observation of perennial Arctic sea ice melt and freeze-up using passive microwave data, Journal of Geophysical Research, 103, pp. 27,753–27,769. Steele, M. and Flato, G.M. (2000) Sea ice growth, melt and modeling: A survey, In: Lewis, E.L. et al., eds. The fresh budget of the Arctic Ocean, NATO Science Series, Vol. 70, Dordrecht: Springer, pp. 549–587. Sturm, M. et al. (2006) Snow depth and ice thickness measurements from the Beaufort and Chukchi Seas collected during the AMSR-Ice03 campaign, IEEE Transactions on Geoscience and Remote Sensing, 44(11), pp. 3009–3020. Style, R.W. and Worster, M.G. (2009) Frost flower formation on sea ice and lake ice, Geophysical Research Letters, 36 (L11501). Available from: doi:10.1029/2009GL037304. Tilling, R. et al. (2020) Detection of melt ponds on Arctic summer sea ice from ICESat-2, Geophysical Research Letters, 47 (23), e2020GL090644. Tonboe, R., Andersen, S. and Pedersen, L.T. (2006) Simulation of the Ku-band radar altimeter sea ice effective scattering surface, IEEE Geoscience and Remote Sensing Letters, 3, pp. 237–240. Tschudi, M.A., Curry, J.A. and Maslanik, J.A. (2001) Airborne observations of summertime surface features and their effect on surface albedo during Fire/SHEBA, Journal of Geophysical Research, 106, pp. 15,335–15,344. Tschudi, M.A., Curry, J.A. and Maslanik, J.A. (2002) Characterization of spring time leads in the Beaufort/Chukchi Seas from airborne and satellite observations during FIRE/ SHEBA, Journal of Geophysical Research, 107(C10), p. 8934. Available from: doi:10.1029/2000JC000541. Tschudi, M.A., Maslanik, J.A. and Perovich, D.K. (2008) Derivation of melt pond coverage on Arctic sea ice using MODIS observations, Remote Sensing of Environment, 112, pp. 2605–2614. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1982) Microwave remote sensing: Active and passive, Vol. II: Radar remote sensing and surface scattering and emission theory, Norwood, MA: Artech House Inc., p. 847. Ulaby, F.T., Moore, R.K. and Fung, A.K. (1986) Microwave remote sensing: Active and passive, Vol. III: From theory to applications, Norwood, MA: Artech House Inc., p. 2058.

Vesecky, J.F., Smoth, M.P. and Samadani, R. (1990) Extraction of lead and ridges characteristics from SAR images of sea ice, IEEE Transactions on Geoscience and Remote Sensing, 28, pp. 740–744. Wadhams, P. (1988) The underside of Arctic sea ice imaged by side-scan sonar, Nature, 333, pp.161–164. Wang, L. et al. (2011) Integrated ocean-Arctic melt onset detection from satellite active and passive microwave measurements, 2000–2009, Journal of Geophysical Research, 116 (D22103). Warren, S.G. (1982) Optical properties of snow, Reviews of Geophysics and Space Physics, 20(1), pp. 67–89. Wiesmann, A. and Mätzler, C. (1999) Microwave emission model of layered snowpacks, Remote Sensing of Environment, 70(3), pp. 307–316. Willmes, S. and Heinemann, G. (2015) Pan-Arctic lead detection from MODIS thermal infrared imagery, Annals of Glaciology, 56(69), pp. 29–37. Willmes, S. and Heinemann, G. (2016) Sea ice wintertime lead frequencies and regional characteristics in the Arctic, Remote Sensing, 8(4). Available from: http://doi.org/10.3390/ rs8010004. Worby, A.P. and Allison, I. (1991) Ocean–atmosphere energy exchange over thin, variable concentration Antarctic pack ice, Annals of Glaciology, 15, pp. 184–190. Wright, N.C. and Polashenski, C.M. (2018) Open-source algorithm for detecting sea ice surface features in high-resolution optical imagery, The Cryosphere, 12, pp. 1307–1329. Yackel, J.J. and Barber, D.G. (2000) Melt ponds on sea ice in the Canadian Arctic Archipelago: Part 2. On the use of Radarsat2 Synthetic aperture radar for geophysical inversion, Journal of Geophysical Research, 105(C9), pp. 22,061–22,070. Yackel, J., Barber, D. and Hanesiak, J. (2000) Melt ponds on sea ice in the Canadian Archipelago 1. Variability in morphological and radiative properties, Journal Geophysical Research, 105(9), pp. 22,049–22,060. Yackel, J.J. et al. (2007) First-year sea ice spring melt transitions in the Canadian Arctic Archipelago from time-series synthetic aperture radar data, 1992–2002, Hydrological Processes, 21 (2), pp. 253–265. Yang, C. (2018) High resolution satellite imaging sensor for precision agriculture, Frontiers of Agriculture Science and Engineering, 5(4), pp. 393–405. Yao, T. and Tang, L. (2003) The formation and maintenance of the North Water Polynya, Atmosphere–Ocean, 41(3), pp.187–201. Yu, Y. and Rothrock, D.A. (1996) Thin ice thickness from satellite thermal imagery, Journal of Geophysical Research, 101 (C11), pp. 25,753–25,766. Yu, J. et al. (2009) Analysis of sea ice motion and deformation in the marginal ice zone of the Bering Sea using SAR data, International Journal of Remote Sensing, 30(14), pp. 3603–3611. Zhou, X. and Li, S. (2000) Concentration determination of sea ice pressure ridge in the Ross Sea based on radar backscatter and object delineation methods, Proceedings of International Symposium of Geoscience and Remote Sensing (IGARSS), Honolulu, pp. 1323–1325.

11 Retrieval of Sea Ice Geophysical Parameters

11.1

11.2

Sea Ice Type Classification ................................................... 454 11.1.1 Ice Classification from Optical and TIR Systems .... 456 11.1.2 Ice Classification from Passive Microwave Data..... 457 11.1.3 Ice Classification from SAR .................................... 458 11.1.3.1 Ice Classification from Single-Channel SAR......................................................... 460 11.1.3.2 Ice Classification from Dual-Channel SAR......................................................... 461 11.1.3.3 Ice Classification from Polarimetric SAR Data......................................................... 467 Sea Ice Concentration........................................................... 471 11.2.1 Ice Concentration from Optical and TIR Images....................................................... 472 11.2.2 Ice Concentration from Coarse-Resolution Microwave Observations ............................................................ 473 11.2.2.1 NASA Team (NT) Algorithm ................. 475 11.2.2.2 The Enhanced NASA Team (NT2) Algorithm ................................................ 476 11.2.2.3 The ASI Algorithm.................................. 478 11.2.2.4 ECICE Algorithm ................................... 479 11.2.2.5 Intercomparison of PM Algorithms......... 486 11.2.2.6 Sources of Error and Sensitivity of Ice Concentration Algorithms ....................... 490

Sea ice parameters needed from remote sensing data have different priorities between two users’ communities: the operational sea ice monitoring community and the climate research community. Priorities for the operational ice monitoring include ice types (based on age or equivalently stage of development), ice concentration, motion, ice mechanical properties, ice edge location, onset of ice decay, and ice topography. For climate studies the priorities include ice thickness, concentration, extent, surface temperature, albedo, emissivity, snow properties, ice kinematics (including motion), and onset of freezing and melting dates. Obviously, there is an overlapping set. Selected ice parameters are covered in this chapter, namely sea ice types (age-based), ice concentration,

11.2.2.7

Assessment of Ice Concentration Results Against Ice Charts ................................... 493

11.2.3 Ice Concentration from Fine-Resolution SAR ........ 496 11.3 11.4

Sea Ice Extent and Area ....................................................... 498 Sea Ice Thickness (SIT) ........................................................ 501 11.4.1 SIT from TIR Observations..................................... 503 11.4.2 SIT from PM Observations ..................................... 506 11.4.3 SIT from Altimeter Observations ............................ 510 11.4.4 SIT from SAR Observations.................................... 514

11.5

Ice Surface Temperature (IST).............................................. 517 11.5.1 IST from TIR Observations..................................... 517 11.5.2 IST from PM Observations ..................................... 520

11.6

Sea Ice Age ........................................................................... 522

11.7

Sea Ice Motion and Kinematics............................................ 524 11.7.1 Methods of Ice Motion Tracking ............................ 526 11.7.1.1 Motion Tracking Using Image Features.. 526 11.7.1.2 Motion Tracking Using Individual Sea Ice Floes .................................................. 528 11.7.2 Operational Ice Motion Products ............................ 532

11.8

References............................................................................. 533

extent, thickness, surface temperature, age, and motion. Methods of retrieval of these paramaters are addressed in some details. Retrieval of sea ice parameters from remote sensing data should take into consideration three issues. The first, and by far the most crucial, is the limited penetration of the incident or emitted radiation into the observed surface. Optical and IR signals penetrate the surface down to a few millimeters while microwave signal can penetrate deeper as shown in the data presented in section 9.5. The limited penetration means that the observed signal does not carry information about the bulk properties of the ice cover. This affects retrieval of information involving thickness such as thickness-based ice types. It also limits

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 453

454

SEA ICE

the definition of ice temperature retrieval to skin temperature or radiating layer temperature. In this case, subsurface features of the ice can be used as proxy indicators of bulk ice type/property. This is demonstrated, for example, in identifying multi-year ice (MYI) in radar data because it is characterized by bubbly subsurface layer, which engender distinguishably high backscatter. Another example is using the surface temperature to discriminate between seasonal and perennial ice is presented (section 11.5). The second issue is the influences of atmospheric constituents (e.g., gases, aerosols and clouds) on the observed radiation/scattering. This issue has been recognized since the inception of using satellite optical (and later microwave) remote sensing and several methods have been developed to account for such influences as outlined briefly in section 7.9.1. The third issue is the influence of the snow, especially when metamorphosed and masked the underlying ice. In this chapter, mathematical schemes of retrieval of key ice parameter from different categories of remote sensing data are presented. More details are included regarding the methods of ice concentration and thickness retrievals. When retrieval of same parameter is available from different categories of remote sensors, methods are grouped based on each category; namely optical, thermal infrared (TIR), passive microwave (PM) and radar. Potential and limitations of each sensor or retrieval method are discussed with comparison between methods when available. In addition, sample results from the methods are presented. Since this chapter is mainly about parameter retrieval methods, it is worth noting that well-established sea ice parameters have been included in daily or weekly products from a few agencies. The Ocean and Sea Ice Satellite Application Facility (OSI-SAF), operated jointly by the Norwegian and Danish Meteorological Institutes, produces daily maps of sea ice concentration, ice types (only FYI and MYI), and global sea ice edge delineation for the Arctic and Antarctic regions. The maps are produced at 10 km resolution in polar stereographic projection. Coarse-resolution maps of sea ice drift (on 62.5 km polar stereographic grid) are also produced every other day. OSI-SAF also offers ice motion at same coarse-resolution and medium-resolution (20 km) grid. Most data are available since 1979, when routine monitoring by satellites started. The US National Snow and Ice Data Center (NSIDC) produces a wide variety of sea ice parameters including daily maps of ice concentration from PM data, ice extent for selected areas in the polar regions from visible (VIS) and infrared observations, ice edge boundary in the northern hemisphere, and melt onset of the Arctic ice from a combination of observations. NSIDC also provides graphs showing trends and anomalies of monthly mean ice concentration and extent beginning in 1979

for both polar regions. Weekly ice age product is also available in an EASE-grid projection with 12.5 km resolution. NASA offers daily datasets of polar sea ice extent and surface temperature at 1 km resolution from MODIS satellite [Hall, Riggs, Salomonson, 2006]. The RADARSAT Geophysical Processor System (RGPS) at the Alaska Satellite Facility (ASF) produces maps of ice motion and deformation regularly at 3- and 6-day intervals using successive observations from the fine-resolution SAR data. Operational sea ice monitoring services, [e.g., the Canadian Ice Service (CIS)] generate different products of the aforementioned parameters based on visual analysis of satellite imagery data (especially SAR) and several sets of meteorological, oceanic and climatic data. Sea ice products such as concentration and age using different commonly used retrieval methods are also made publicly available through ftp sites from University of Bremen (UB) and Alfred Wegener Institute (AWI).

11.1. SEA ICE TYPE CLASSIFICATION From the viewpoint of climate studies, it is sufficient to use the three major ice types: young ice (YI), first-year ice (FYI), and multi-year ice (MYI) in a classification scheme. However, subcategories of YI may be important because heat flux from the underlying warm ocean in winter is sensitive to the ice thickness within this category. On the other hand, thicker ice types are more important from the viewpoint of operational ice monitoring. Therefore, subcategories of FYI such as thin, medium, and thick FYI are required because their different thickness ranges impact mechanical ice loading on marine structures differently. Deformed versus leveled ice is also important for the same reason. As explained in section 2.8, the most popular ice classification criterion is based on ice age (or equivalently ice thickness). This criterion is established by the World Meteorological Organization (WMO), and the scheme is known as “stages of sea ice development”. However, since remote sensing observations are engendered by electromagnetic (EM) wave interaction with a small thickness fraction of the top surface, remote sensing observations do not hold information about ice age or its bulk thickness. This information can be inferred only if the ice type leaves a distinct physical signature within the subsurface (penetration layer) that engenders the observation. Examples include the high salinity of newly formed ice, the smooth surface of nilas, the rafting of YI types, frost flowers on new and gray ice (GI) surface, the surface roughness of deformed ice, the flooded surface of degraded ice and the high porosity of MYI hummock ice. If the observations are less sensitive to such physical signatures, as pronounced in the cases of reflectivity from optical systems or

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 455

brightness temperature from TIR systems, then the use of the observations may be limited to discrimination between sea ice and water (with no information about ice types). In general, the sensitivity of SAR to most of the above-mentioned features is better than optical and TIR, and that is why this sensor has become the prime source of ice type classification for both operational and research purposes. However, to achieve the full potential of SAR for sea ice classification is to seek classes that fit the information implied in the backscatter from SAR. This implies pursuing SAR-driven ice classes rather than forcing SAR backscatter (or scattering mechanisms) to inform about an age-based ice classification, which SAR does not actually “see.” This point has been addressed in section 7.6.3.4 and illustrated again in this chapter (section 11.1.3.3) in the context of discriminating between hummocks and melt ponds of MYI. Aside from the new proposed criterion of SAR-based ice classification, this section pursues the traditional age-based ice classification using optical, TIR and microwave imagery data. Before embarking on addressing ice type classification using different remote sensing data, it is appropriate to demonstrate (through an example) the difference between using SAR and optical images in classifying certain ice types. The example in Figure 11.1 shows an image of

sea ice acquired by AVHRR channel 1 (0.58–0.68 μm) over the North Water polynya (NOW), north of Baffin Bay on 20 April 1998. The image is compared against the same scene acquired by RADARSAT-1 after resampling to nearly the same resolution as of AVHRR. The time difference between the two acquisitions was 35 minutes. The area south of the ice arc in the RADARSAT-1 image appears to have an elongated band of relatively bright signature with different shades. This is likely to be a mix of thin ice at different growth stages and open water (OW). The same area in the AVHRR image has uniform dark signature (low albedo). If one relies on the the AVHRR image it would be difficult to conclude the presence of YI in this area. The core area in the middle of the image is the North Water polynya. This appears with uniform low albedo in AVHRR image. More information is readily available in SAR image. Another difference is apparent at the bottom of the image (labeled as FYI). The brighter tone in AVHRR image (higher albedo) is probably caused by snow cover or clouds. SAR is not sensitive to either one of them, so it offers more details of the sea ice composition in this area. To conclude this part, it is worth noting that operational ice monitoring services still use visual analysis of remote sensing imagery, particularly SAR, in near-realtime to map ice types along with other ice information.

Clo

uds

nla Gr ee

nla

YI

Gr ee

&O W

Ell Isl esem an d ere

nd

nd

Ice arc

FY ice

AVHRR

Radarsat

Figure 11.1 Images of sea ice in the North Water polynya, north of Baffin Bay, Canadian Arctic. AVHRR channel 1 image (left) and RADARSAT-1 image (right) were acquired almost coincidentally on 20 April 1998. The contrast between FYI on one hand and OW and YI on the other hand is apparent in the AVHRR image. The red dots correspond to the same geographic point in both images (RADARSAT-1 image © CSA).

SEA ICE

11.1.1. Ice Classification from Optical and TIR Systems The two parameters that can be used for ice classification from optical and TIR sensors are albedo and surface temperature, respectively. Only ice types that are sensitive to either one or both parameters can be identified in the imagery data. Table 9.2 shows three categories of surface cover that can be identified by their fairly distinguished albedo; OW/nilas/grease-ice, gray/gray-white ice (GWI), and FYI (thin and thick). However, using albedo for ice classification is hampered by the presence of snow cover, which masks the albedo from the ice surface. On the other hand, using TIR can reveal the ice surface temperature (IST), which can be used to discriminate between three ice types: YI, FYI, and MYI. In general, for same air temperature, surface temperatures of these three ice types are different because of their different thickness ranges, which determine the amount of heat released from the warm ocean to the atmosphere. YI surface is the warmest and MYI surface is the coldest. Subcategories within any of the above three categories do not render unique albedo or surface temperatures that can assist in their classification from optical or TIR data, respectively. For example, none of these—thin, medium, or thick FYI—can be resolved from optical or TIR data. Similarly, leveled versus deformed ice surface cannot be distinguished. Ice classes available from optical and TIR sensors (once again, in the absence of snow cover) fit better the requirements of climate-related applications. Additionally, optical and TIR data can be used to discriminate between ice and OW. Yu and Rothrock [1996] found that the freezing temperature of Arctic sea water varied between 271.36 K and 271.45 K. Therefore, a threshold of 271.4 K is usually used to separate water from sea ice in TIR data. An early study to classify ice types using a combination of VIS and TIR data available from AVHHR was presented in Massom and Comiso [1994]. They used the NIR channel 2 (0.912 μm) and the TIR channel 4

(10.8 μm) to discriminate between four surfaces that existed in the Bering and Greenland seas in April and May 1988: OW, new ice (NI), YI (including light nilas), and FYI (intermingled with some MYI advected from the Central Arctic). The frequency distribution of the data from channels 2 and 4 for the four categories are shown in Figure 11.2. This cannot be considered a universal data set but it serves to point out the potential of using albedo and TIR in ice type classification. The albedo values of OW and NI are not distinguishable in the channel 2 data but their difference in physical temperatures is revealed in the measured radiance from the TIR channel 4 data (recall the equal and narrow range of emissivity from these surfaces in the TIR region as presented in section 7.2.3). The four surface types in Figure 11.2 are associated with distinguishable modes in the distribution of the TIR data and therefore can be identified using selected thresholds. This method, however, proved to be less reliable because of the contributions of clouds and atmospheric constituents to the observed signal. Methods to account for night time cloud affects are presented in Spangenberg et al. [2002] and Salomon, Tetzlaff, Karlsson [2007]. Distribution of water and ice types within a heterogeneous footprint is possible in optical or TIR data

6000 AVHRR channel 2 OW + New ice Frequency

Ice type maps are available in the form of daily ice chart products from a few operational national centers in Canada, USA, Norway, and Finland. As far as authors of this book know, no SAR operational analysis has used computer-assisted method to identify ice types in any remote sensing. No method has proven to be robust enough to satisfy the operational requirements. Visual image interpretation ensures the accuracy and robustness. Coarse Arctic-wide ice type mapping (in terms of spatial resolution and ice type categories) is generated daily or every second day using data from radiometers and scatterometer systems. These are available from NSIDC, OSI-SAF and University of Bremen among other centers. More information on the OSI-SAF product is shown in section 11.1.2.

4000 Light nilas + Young ice 2000

0

0.0

0.5

1.5 2.0 1.0 Radiance AVHRR channel 2

2.5

3.0

5000 AVHRR channel 4 Light nilas FY ice + young ice

4000 Frequency

456

3000

Open water

2000 New ice 1000 0 75

70

65

60

55

Radiance AVHRR channel 4

Figure 11.2 Histograms of AVHRR channel 2 (VIS) and channel 4 (TIR) radiances from a region in the Bering Sea during the spring of 1988, showing the separation of the four surfaces in the channel 4 data [Massom and Comiso, 1994, Figure 3, with permission from AGU].

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 457

because of the relatively large footprint, and this results in wrong classification. The idea of using multi-angular observation from visible bands to determine ice type was explored in Nolin, Fetterer, Scambos [2002]. The premise is that the scale of surface roughness determines the pattern of surface scattering. For example, ice surface is a forward scatterer at relatively fine roughness scales but becomes a backward scatterer when surface micro-topography dominates. The authors explored this theme using data from the multi-spectral Multi-angle Image SpectroRadiometer (MISR) onboard NASA’s Terra satellite. The spectral information is provided by four channels (three in the VIS and one in the NIR). The angular information is provided by nine cameras operating at different angles. One camera is pointed straight down (nadir direction), four pointed at different forward angles, and the other four at the same backward angles. This makes a total of 36 reflectance measurements. The study classified the TOA reflectance data obtained from the nine angular MISR channels using the standard unsupervised classification technique ISODATA [Tou and Gonzalez, 1974]. Results showed potential success in discriminating FYI from MYI as multi-angular signatures can be used in much the same way as multi-spectral signatures. The discrimination, however, was not as clear as it could have been from using SAR data. However, in summer, when surface flood renders SAR data useless for ice classification the multi-angular data from MISR provide key icemapping information. The 36 channels from a sensor such as MISR can be used better within a deep learning approach, which has received more attention in remote sensing sea ice classification recently. 11.1.2. Ice Classification from Passive Microwave Data PM data are not often used in sea ice type classification because of their coarse resolutions. Within a single footprint that measures a few kilometers or tens of kilometers, depending on the wavelength of the emitted radiation, the ice content is almost always heterogeneous. PM is more successful in estimating the partial concentration of major ice types, but not their spartial distribution, within the footprint. It is more successful in discriminating sea ice (regardless of the type) from open water. In general, PM data are capable of distinguishing between three surfaces: OW, FYI, and MYI based on the gradient ratio [equation (9.2)]. This parameter is positive for OW, negative for MYI, and takes on values close to zero for FYI (Figures 9.8 and 9.10). Figure 11.3 shows the distributions of Tb,37H , Tb,37V , GR37V19V , and GR22V19V from OW, FYI, and MYI. It can be seen that Tb,37V is not useful in separating OW from MYI and GR22V19V is not useful in separating FYI from MYI. The separation of the three

surfaces is possible when using Tb,37H or GR37V19V. Examination of distributions of the desired parameters for discrimination between a given set of ice types is important in order to decide on the suitable parameter(s) that can be used in the classification scheme. Another data set that shows the separability between FYI and MYI in Tb from AMSR-E and AMSR2 is presented in the scatter plots in Figure 11.4. This is a joint histogram of PM and scatterometer data. PM data are from AMSR-E and AMSR2 36.0 GHz channel with horizontal polarization. Scatterometer data are from the Ku-band QSCAT and C-band ASCAT (both from VV polarization). Note the tight clusters of MYI data compared to the wide clusters of FYI data, which extends across 20–40 counts of Tb. This wide range is probably caused by the sensitivity of Tb to variations of the snow cover over FYI. This sensitivity is not observed in the MYI data because the snow cover is not saline. Figure 11.4 highlights the advantage of combining the scatterometer with PM observations to distinguish between FYI and MYI. This combination is used in the Environment Canada’s Ice Concentration Extractor algorithm (ECICE) (section 11.2.2.4) and also presented in Zhang et al. [2019]. Information from the above two examples confirms that Tb,37H /or GR37V19V can be used to identify OW, FYI, and MYI surfaces (yet with no capability to identify other ice types). Thresholds can be used to achieve this purpose. If the observation is obtained from a heterogeneous footprint, the surface types and their partial concentration can be identified by decomposing the given observation into its components from each surface using the familiar linear decomposition equation, then the concentration of the required ice type can be calculated. For example, if the footprint is composed of ice and water, Tb can be composed into its components of Tb,ice and Tb,ow weighed by their concentrations Cice and Cwater, then the following equation can be used to calculate Cice, (noting that Cwater = 1 − Cice): C ice = T b− T b,ow

T b,ice − T b,ow

(11.1)

If two observations are available, e.g., Tb,37H and GR37V19V , then two ice types (e.g., FYI and MYI) can be resolved by solving a set of linear equations, where each equation decomposes one observation to its components from the different surface types. The concentration of OW (Cow) can be obtained by adding a third equation, namely an equality constraint, which dictates that the summation of the concentrations from the three surfaces must be 100%. This mathematical scheme produces concentrations of each surface type within the footprint but does not assign a certain single type to the footprint. This is more accurate than assigning one ice type to each pixel based on using a set of thresholds.

458

SEA ICE 0.040 0.035

OW MYI FYI

0.07

0.030

0.06

0.025

0.05

Probability

Probability

0.08

OW MYI FYI

0.020 0.015

0.04 0.03

0.010

0.02

0.005

0.01

0.000

0.00 100

150

200

250

100

150

Tb,37H

0.03

0.02

0.06

0.04

0.01

0.02

0.00

0.00

–0.10

–0.05

0.00

0.05

0.10

GR37V19V

250

OW MYI FYI

0.08

Probability

Probability

0.04

OW MYI FYI

200 Tb,37V

–0.04 –0.02

0.00

0.02

0.04

GR227V19V

Figure 11.3 Probability distribution of brightness temperature and gradient ratio of AMSR-E data from samples over OW, FYI, and MYI in the Arctic. Vertical dashed lines are tie points selected to represent the given parameter in most ice type algorithms. Best radiometric parameters for ice type classification are those with separate distributions from different ice types [adapted from Ye et al., 2019, Figure 3].

Hughes [2009] combined the seven channels of SSM/I to explore their potential use in an unsupervised classification scheme designed for operational use in the Norwegian Ice Service. The author incorporated data in a seven-dimensional parameter space of brightness temperatures from any PM system and examined their distribution. The K-means clustering technique [Hartigan, 1975] was used to partition the data into four clusters such that the within-cluster sum of squares is minimized. The technique starts by selecting initial centers of the clusters and assigning each observation to its closest cluster. After all points are assigned, cluster centers are updated and the process is repeated until there are no further tangible changes in the centers. Clusters are then labeled to ice types according to some ancillary information. Three ice types: YI, FYI, and MYI were used in addition to OW. Weekly maps were produced for three selected ice seasons.

Ice type classification (FYI versus MYI) plus OW in the Arctic is a product from OSI-SAF since 2005 using SSMIS PM and ASCAT scatterometer data [Aaboe et al., 2017]. The algorithm is based on a Bayesian approach that computes the probability of occurrence of the most likely ice type. Ice type maps from this product provide synoptic coverage useful for climate application. They are too coarse to be of practical value for operational marine users, yet they can be used as a baseline for detailed maps of local ice coverage. 11.1.3. Ice Classification from SAR Synthetic Aperture Radar (SAR) is the prime sensor for operational ice type classification, whether through visual image analysis or digital methods. In fact, the development of the early space-borne SAR systems, ERS, and RADARSAT, was mainly driven by requirements for

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 459 (a) –5

0.8 0.6

–15 0.4 –20

–25 140

0.2

160

180

(b)

200 Tb (K)

220

240

260

–5

1 0.8

–10 σo(dB)

0

0.6 –15 0.4 –20

–25 140

0.2

160

180

200 Tb (K)

220

240

260

Normalized Count

σo(dB)

–10

Normalized Count

1

0

Figure 11.4 Joint histograms of σ o and Tb across the Arctic sea ice cover from (a) combination of QuikSCAT/AMSRE and from day 351, 2004, and (b) ASCAT/AMSR-2 from day 351, 2014. The two clusters represent FYI (top left) and MYI (bottom right). Tb is from 36.0 GHz channel with horizontal polarization and σ o is from QSCAT Ku-band in VV polarization (courtesy of J. Zhang of SYSU, China).

operational sea ice monitoring, with ice classification set as a top priority. At the center of these requirements was the identification of navigationally hazardous ice types such as MYI and heavily ridged FYI. In the 1990s, funds were allocated to develop automated or semi-automated SAR-based systems of ice classification. The promise to move on from the visual analysis of airborne SAR images to an automated ice classification was heightened but unfortunately did not bear fruits later. Although most of the early attempts of ice classification in the 1990s were not successful, they identified a few issues that needed further development in terms of the sensor characteristics and analysis methodology. An example was the need to increase the dimensionality of SAR observations to resolve the issue of the backscatter overlapping from different ice types. This had led to the development of polarimetric SAR systems. However, it should be emphasized that so far no computer-assisted method of operational SAR data processing has been used. None of the many methods developed over the past four decades has achieved the

accuracy, robustness, and fast turnaround required by the operational environment. It is worth repeating that SAR data are more effective in identifying what may be called SAR-based, rather than age-based, ice types (section 7.6.3.4). For example, SAR can be used more successfully to identify leveled versus deformed ice, ridged versus rafted ice, frozen versus flooded ice, dry versus wet or metamorphosed snow cover, and frost-flower-covered ice surface. Nevertheless, efforts continue to use SAR in ice classification schemes following the traditional age-based WMO categories. This is accepted by marine operational community and, in some ways, by the climatology community. The material in this section covers retrieval of age-based ice types, yet a hint to SAR-based classification is introduced in section 11.1.3.3 in the context of discriminating hummock from melt pond ice in MYI. Pixel-based ice classification from the fine-resolution SAR can be used to examine the heterogeneous ice composition from a coarse-resolution data such as PM or scatterometer data [Shokr, Lambe, Agnew, 2008, Ye et al., 2016].

460

SEA ICE

A review of major approaches used in sea ice classification from SAR data are presented in Dierking [2013]. It addresses issues such as ice properties that influence the radar backscatter signal and the geophysical sea ice parameters that can be retrieved from SAR. Another review is presented in Zakhvatkina, Smirnov, Bychkova [2019]. It includes the main techniques used for ice charting in several national sea ice monitoring services with a critical assessment of advantages and disadvantages of the techniques. This leads to identifying technical abilities of SAR that improve the charting products and suggesting possible further development of SAR data to improve the products. It is worth noting that any automated classification method should account for the angular variability of sea ice types in the wide-scan mode of ScanSAR, which is commonly used in operational sea ice monitoring. This variability is ice-type dependent. Lohse, Doulgeris, Dierking [2020] introduced a supervised classification algorithm that incorporates the angular variation of backscatter across the SAR swath. Only ice types from winter data are presented because the inference of ice types in summer data is virtually unattainable as the flooded surface masks the underlying ice. The use of single- and multi-channel SAR is introduced first, followed by the methods that utilize the recently developed operational polarimetric SAR. The following applications target Arctic sea ice. Parallel work for Antarctic sea ice has not advanced. 11.1.3.1. Ice Classification from Single-Channel SAR First attempts to classify ice types in SAR images used a single-channel data and adopted common approaches of image processing techniques. These approaches are presented in several text books [e.g., Richards and Jia, 2005, Schowengerdt, 2006, Campbell and Wynne, 2011]. They encompass supervised and unsupervised classification techniques, data fusion methods to combine observations from different sensors, dimensionality reduction techniques (e.g., principal component analysis) as well as a few statistical tools. To overcome the difficulty of backscatter overlap from ice types, the classification benefited from using contextual information such as texture and shape of ice floes, or ancillary information such as ice climatology, wind, temperature fields, and recent history of the ice cover. In the early 1990s, the first common approach of SAR ice classification used a set of thresholds to separate ice types. The thresholds were established based on training data sets of backscattering from known ice types and employed later in a supervised classification algorithm. In many studies, this approach was implemented based on the gray tone in SAR images without much work to link it to ice properties that trigger the observed backscatter. This approach rendered useless for operational sea ice

monitoring environment because of lack of robustness (the validity was limited to the dataset from which the thresholds were selected). A well-known and widely used probabilistic approach of image classification is the Bayesian parameter estimation or maximum likelihood estimation (these are nearly identical methods but the approaches are slightly different). It has been used often in early ice classification studies and still used today to generate operational ice type maps (FYI, MYI, and OW) in the European OSI-SAF products. The approach is based on statistical properties of the radiometric measurements from each ice class. For a set of satellite observations, R, the conditional probability of occurrence of ice type (icej) (including sea water as a possible type) given an observation Ro is obtained from the equation: p icej Ro =

p icej p Ro icej N

(11.2)

p icek p Ro icek k=1

where, p(Ro| icej ) is the probability of occurrence of Ro given the ice type icej, p(icej ) is the a priori probability of icej, and N is the number of types. In the case where no prior knowledge or assumptions are available, p(icej ) can be set to the inverse of the number of ice types and OW; i.e., if the purpose is to assign the observation to one of three possible surface types, then all p(icej ) can be assumed one-third for each type. The above equation can be applied with any number of observations from multisensor sources. The left-hand side is determined and the pixel is assigned to the class icej that has the maximum conditional probability. This approach also failed to satisfy the robustness required by the operational environment because of the overlap between backscatter from different age-based ice types. In order to resolve this issue, a few studies augmented texture with SAR backscatter to increase the dimensionality of the data, hence facilitate the ice classification. One of the most popular texture quantification methods is called Gray Level Co-occurrence Matrix (GLCM), developed by Haralick, Shanmugan, Dinstein [1973]. Calculations start by quantizing the gray tone or the backscatter range in the image to 16, 32, or 64 discretized levels. It then proceeds with the calculations of cooccurrence matrix entries, where each entry represents the number of occurrences of two neighboring pixels (within the specified computational window) that have a specified difference of gray tone. Details of the method and the derived texture measures are included in Haralick, Shanmugan, Dinstein [1973]. Texture quantification is based on gray tone in SAR images, not the backscatter. Although a few studies claimed partial success of using texture measures in ice classification [Shokr, 1991,

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 461

Nystuen and Garcia, 1992, Smith, Barrett, Scott, 1995, Clausi and Yue, 2004], other studies found these measures not always suitable to uniquely identify most of the required ice types [Chen et al., 2020]. Texture has proven useful in discriminating MYI from leveled FYI or YI, but these two types can be easily discriminated from backscatter data. One issue related to the use of texture is the selection of the window size from which the texture is quantified. Texture is always presented in different scales even for the same surface. The window size in a texture quantification method should be selected to capture the desired texture scale, which may differ across the image and certainly between ice types. For example, deformed ice surface appears in SAR images with certain spatial distribution of backscatter (i.e., texture) that can be visually captured but only digitally captured if the window size is selected to be slightly larger than the average spatial distance between successive deformation pieces in the image. Too small or too large windows will miss the existing texture scale. The same argument applies to MYI with its spatial distributions of the hummock and melt pond surfaces, which generate different backscatter and therefore specific texture. If the window size is too small, the texture measure will not capture the apparent texture in the image. If it is too large the measure will rather reveal a smooth surface (i.e., overlook the natural texture). The use of adjusted dynamic window size seems appropriate to capture the dominant texture in different parts of the image. This idea has been addressed in Chen et al. [2020] within an ice classification scheme of MYI/FYI. Although the idea does not seem to be successful in this work, it is worth trying with other datasets. In general, although the use of texture adds an independent measure to augment the gray tone, it has proven to be of limited success in discriminating ice types that feature multiscale surface geometries and deformation. It can be successful if the ice type can visually be identified based on its unique texture such as pancake ice. Clausi and Deng [2003] combined tone and texture to achieve pixel-based segmentation of RADARSAT-1 images. This was to support of the operational analysis of SAR images at the CIS. They found that sequential image processing using the Markov Random Field (MRF) labeling model improved the image segmentation significantly. Combining tone and texture in an unsupervised “K-means” clustering algorithm improved ice–OW discrimination as well as FYI–MYI discrimination. An example of the results from their work is presented in Figure 11.5. The scene contains three ice types: GI, GWI, and MYI. The histogram of the image gray tone shows a unimodal distribution (Figure 11.5b), which makes it difficult to segment the image using a computer vision technique. As shown in Figure 11.5c, the gray tone

will not achieve proper image segmentation. Only two of the three ice classes are identified because MYI and GWI have similar gray tone values. The result from another segmentation approach using texture only (GLCM measures followed by K-means clustering) is shown in Figure 11.5d. In this case, GI and GWI are erroneously grouped together, and MYI is identified as a separate class. Figure 11.5e represents a segmentation based on a fused feature set (GLCM texture + tone) followed by K-means clustering. In this case all three ice classes are identified though not accurately. When the MRF label model is applied to the result in Figure 11.5e the best segmentation result emerges as presented in Figure 11.5f. The K-means method is used as an initial guess, and the MRF labeling algorithm is used for refinement purposes. Although the success of the MRF labeling algorithm is proved based on the shown image, it should be noted that the original SAR image (Figure 11.5a) shows MYI floes surrounded by fairly leveled ice. The recommended approach may not be suitable in the case of deformed or rough ice surfaces or other surface morphologies covered. It may not accurately classify surfaces covered with frost flowers or metamorphosed snow. One of a few sea ice classification algorithms that found a tentative way into applications in operational ice monitoring environment is presented in Clausi et al. [2012]. The algorithm, called MAp-Guided Ice Classification (MAGIC), is designed specifically to interpret SAR sea ice images using coincident ice charts provided by the CIS. Daily ice charts are manually created at the CIS based on visual analysis of SAR images and other ancillary data. A chart shows polygons, delineated by an ice analyst, where ice types, their partial concentrations, total concentration, and floe sizes located within the polygon are specified according to a given code [MANICE, 2005]. Each parameter is assumed to be uniformly distributed across the polygon. MAGIC uses this information as input, then generates a pixel-based ice map for each polygon, a product not feasibly produced manually. One of its primary features is a segmentation module, which segments each polygon into regions of uniform ice type then assigns a label for the region. The labeling technique uniquely models the spatial relationship of regions between the polygons in the form of a neighborhood system embedded in an MRF framework. MAGIC has been used at CIS to support the visual interpretation of RADARSAT images. 11.1.3.2. Ice Classification from Dual-Channel SAR Multi-channel SAR data offer better potential for ice classification. The data can be presented in the form of multi-frequency, multi-polarization, or fully polarimetric (FP) mode. A brief account of the first two approaches follows in this section, but only dual-channel SAR data

462

SEA ICE

(a)

(b)

(c)

(d)

(e)

(f)

Figure 11.5 (a) RADARSAT-1 SAR sea ice image of a northern area of Baffin Bay, acquired on 7 February 1998, (b) the backscatter histogram of the image in 8-bit gray tone scale, (c) segmentation based on gray tone only using gamma mixture model. (d) segmentation based on texture only using GLCM and K-means clustering, (e) segmentation based on fused texture and tone with K-means clustering, (f ) an improved segmentation with application of Gamma MRF model. Three ice types are marked in the segmented images by different gray shades: GI in black, GWI in gray, and MYI in white [Clausi and Deng, 2003 / Reproduced with permission from IEEE].

are presented here, and the FP (multi-channel) data uses are presented in the next section. So far, there has been no multi-frequency space-borne SAR system on the same platform. A few multi-frequency SAR studies used data from multi-platform SAR and scatterometer systems, but only when coincident data were available. One of the notable studies on using multi-frequency SAR for ice type classification from same platform has been presented in Brath, Kern, Stammer [2013]. The multi-frequency co-polarization data were acquired by a helicopter-borne scatterometer operated at four frequencies: S-, C-, X-, and Ku-band (see Table 7.2 for frequencies and wavelengths). Data were acquired during the German R/V Polarstern cruise ARKXXII/2 into the eastern Arctic Ocean in October 2007.

Classification was performed to identify four ice types: NI < 10 cm thick, GI between 10 and 15 cm thick, old ice (OI), and OW (no FYI was found during the expedition) using the supervised Wishart classifier [Lee et al., 1999]. The authors presented an extensive set of histograms of σ ohh and σ ovv for seven intervals of incidence angles for each ice class at each frequency. The unique contribution of the study was to determine the relative performance of each frequency and combination of frequencies for ice classification. Figure 11.6 shows the classification accuracy of each ice type and the overall classification accuracy from 15 single and combined frequency channels. These accuracies were calculated with respect to the accuracy from using the combination CXKu (a symbol for the combined C-, X-, and Ku-band),

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 463 50 N Gl 40

30

20

Difference [%]

Ol Ow all

10

SCXKu

CXKu

SXKu

SCKu

SCX

XKu

CKu

Cx

SKu

SX

SC

Ku

X

C

S

0

Figure 11.6 Relative classification accuracy of three ice types plus OW using Wishart classifier from multifrequency helicopter-borne scatterometer data (S-, C-, X-, and Ku-band). The combination CXKu is used as the reference from which relative classification accuracy (Difference %) is estimated [Brath, Kern, Stammer, 2013, Figure 14, with permission from IEEE].

which was assumed to be the “best accuracy” based on visual observations and video images of the scene. A few observations can be drawn from Figure 11.6. Using a single frequency, the classification accuracy improves as the frequency increases (except for the fact that Ku-band degrades the classification with respect to the X-band). The same conclusion applies in the case of the dual frequencies. It appears that the S-band deteriorates the classification accuracy perhaps because of its larger footprint (8.2 × 5.8 m2) compared to (3.7 × 2.6 m2), (2.0 × 1.4 m2), and (1.2 × 0.9 m2) from the C-, X-, and Ku-band, respectively. Classification of OI improves as more combinations of frequencies are used and when higher frequencies are included. The value of the results from this study has not yet been envisioned for a future space-borne SAR. Dual-polarization SAR data have been available from a few space-borne sensors such as Envisat, RADARSAT2, and ALOS PALSAR. Several investigations were conducted to explore the potential uses of this mode (also called polarization variety) from Envisat in ice classification. Geldsetzer and Yackle [2009] explored the potential of combining the co-polarization backscatter coefficients σ ohh and σ ovv to classify ice types and discriminate ice from OW. Twenty scenes of sea ice and OW in an area surrounding Cornwallis Island, Nunavut, Canada, obtained in April 2004, were used. Figure 11.7 shows two images of σ ohh and σ ovv in addition to their co-polarization ratio γ co = σ ohh/σ ovv. The backscatter of OW from σ ovv is considerably higher than σ ohh, but the reverse is true for the smooth

FYI. The co-polarization ratio from OW and thin ice surfaces is remarkably high. This is consistent with the polarization-dependent Fresnel reflection that predicts higher reflection in the horizontal polarization (section 7.2.2.1 and Figure 7.10). In the γ co image, the MYI floes cannot be identified and the smooth FYI has a relatively wide range of backscatter. The classification technique used in Geldsetzer and Yackle [2009] is rather simple but reveals information on the discriminating capability of the examined co-polarization channels. The threshold for separating ice of a given type (TH σo ) is derived from backscatter magnitude in dB. TH σo =

σ ol + sl + σ ou − su

2

(11.3)

where, σ ol , σ ou are the lower and upper limits of the mean backscatter coefficients, and sl and su are the standard deviations, respectively, for a given ice type. A similar threshold for separating ice types and OW, based on the co-polarization ratio γ co, is defined by an equation similar to equation (11.3), where subscripts u and l stand for upper and lower limits, respectively. TH γ = γ co − l + sl + γ co − u − su

2

(11.4)

Geldsetzer and Yackle [2009] found the classification accuracies of the MYI, rough FYI, smooth FYI, thin ice, and OW to be 99%, 32%, 89%, 67%, and 50%, respectively while the total classification accuracy was approximately 70%. Except for MYI and smooth FYI, the rest of the surface types could not be classified accurately. This

464

SEA ICE

(a)

(b)

σ0 vv

σ0hh (c)

(d) Griffith Island

OW Smooth FY ice MY

Thin ice

ice

Rough FY ice

γco

N

Figure 11.7 Envisat ASAR image of an area surrounding Cornwallis Island, Nunavut, Canada in 25 April 2004, from (a) σ ohh, (b) σohh, (c) γco. All images are enhanced using Gaussian stretch. The composite image in (d) is generated with RGB channels corresponding to σohh , σovv , and γco, respectively. Ice types marked in (d) were verified in the field [Geldsetzer and Yackel, 2009 / Reproduced from Taylor & Francis Group].

shows the limitation of both the dual co-polarization parameters and the simple thresholding technique. The availability of the Sentinel-1A/B SAR data free of charge through ESA’s Copernicus program has motivated many researchers to develop methods of ice type classification targeting the operational ice charting products. More than 100 Sentinel-1 SAR images are available from the Arctic Ocean per day. Recently, the deep learning (DL) approach utilizing dual-channel SAR of Sentinel-1 for ice classification has been commonly used. The most common DL architecture used in remote sensing data analysis is the conventional neural network (CNN). An exhaustive review of DL for SAR imagery is given in Zhu et al. [2020].

An algorithm of sea ice classification that uses CNN with input from Sentinel-1 SAR images is presented in Boulze, Korosov, Brajard [2020]. The neural network is trained on reference ice charts with an existing machine learning based on texture features and random forest classifier. The training used two datasets for retrieval of four classes: OW, YI, FYI, and MYI. The authors claim that the method outperforms existing random forest product for each type. The code is publicly available through https://github.com/nansencenter/ s1_icetype_cnn. An example of the classification results is presented in Figure 11.8, which shows the original SAR images in the two polarizations, σ ohh and σ ohv . Also shown are the product of the manual classification, which

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 465 σOhh

σO hv

OW New ice Young ice First-year ice Old ice 1.0 0.9 0.8 0.7 0.6 0.5

Operational ice chart

CNN ice chart

CNN class probability

Figure 11.8 Sentinel-1 σ ohh and σ ohv images acquired on 6 March 2018, at 07:43:19 GTM (top images), the corresponding NIC ice chart (bottom left), the neural network classification (bottom middle), and the probability of success of classification of each type (bottom right) [Boutze et al., 2020 / Reproduced from MDPI / CC BY 4.0].

appears in the ice chart from the US National Ice Center (NIC), and the classification using the CNN method. The σ ohh does not provide much information to discriminate between ice types. On the other hand, σ ohv provides more information, particularly the bright signature at the left side of the image, which indicates OI. The CNN classification is similar to the NIC visual classification because the training data come from the NIC charts. However, the CNN classification provides more details such as the numerous small OI floes within the FYI cover. Automated ice classification techniques provide details within the homogeneous ice type areas, which are usually identified in ice charts. Figure 11.8 also includes a map of uncertainty of the classification. This is an important piece of information that should accompany maps of any ice parameter retrieved from remote sensing observations. Another algorithm that uses a machine learning model for sea ice classification is presented in Park et al. [2020]. The proposed scheme aims at producing operational icemapping products through two processing phases: training and implementation. In the training phase, the weekly ice charts are reprojected into the SAR image geometry. In the implementation phase, the classifier is applied to the texture features from Sentinel-1 images. The application demonstrated a capability of identifying three surface types: OW, mixed FYI, and OI. Uchaev et al. [2021] proposed a technique for multifractal classification of sea ice types from SAR images. Results were compared against weekly ice charts from NIC. The first version of the

technique was implemented in the Sentinel Application Platform (SNAP), an open-source architecture developed by ESA to process SAR images. The next three steps are implemented in MATLAB application http://github.com/ UchaevC/GMAT. Operational ice charts provide gross information about ice types and their partial concentrations within subjectively identified polygons of a few kilometers or tens of kilometers size. While this is acceptable by the marine operation community, details about the number, locations and size of hazardous ice floes within the polygon may become critically needed. Moreover, operational ice charts usually provide conservative information, i.e., tending to overstress thicker and older ice types such as MYI [Shokr and Markus, 2006]. To identify individual MYI floes in SAR images, Chen et al. [2020] developed an automated algorithm using Sentinel-1 data acquired during the winter of 2017 in the Canadian Archipelago area. Images are classified as being composed of MYI and a combination of FYI, OW, and other ice types in one type called OW/OIT (other ice types). Backscatter alone or combined with texture cannot be used to identify MYI when surrounded by rough ice cover, which is usually the case in the western Arctic region. The method in Chen et al. [2020] identifies individual MYI floes using an extended-maximum operator, morphological image processing, and a few geometrical features (including the rounded contour of the MYI floe). Their classification was performed using texture employed within a neural

466

SEA ICE

network model. The interesting aspect of this work is the identification of MYI floes based on the above-mentioned criteria. This is a useful product for marine operators since identification of individual hazardous ice floes would complement the readily available information about the concentration of hazardous ice in a big area as provided in operational ice charts. Figure 11.9 compares the classification from the method in Chen et al. [2020] against four images: the original image of σ ohh from Sentinel-1A, the FYI/MYI classification from CIS’s ice charts, the MYI/FYI classification from an algorithm presented in Zhang et al. [2019], another classification/concentration of MYI from ECICE algorithm (see section 11.2.2.4). Pixel spacing from Sentinel-1 is 40 m, which is the same as in the output from Chen’s algorithm, grid resolution from CIS chart is 5 km and spatial resolutions from Zhang’s and ECICE algorithms is 4.45 km. The results from Chen’s method nearly mimic the distribution of individual MYI floes in the Sentinel-1 image (the bright spots). Maps from CIS and Zhang’s ice type methods are similar and mark a wide area of dominant MYI but without delineation of individual floes. Same is true, yet with less area of MYI in the data from ECICE algorithm. The advantage of using the fine-resolution SAR to identify individual floes is apparent compared to the other coarse-resolution products. Identification of MYI floe is also useful in assessing the ice type product from coarse-resolution data. In summer, backscatter cannot be used as a criterion to classify sea ice into its typical major classes of FYI and MYI because the wet snow or surface flood masks the physical features that can otherwise be used to discriminate between the types. Therefore, ice classification should be based on a more suitable criterion. Zhang et al. [2021b],

Sentinel = 1/A σ ohh

CIS chart

use a mini sea ice residual convolutional network, called MSI-ResNET, with input from a single and combinations of σ ohh, σ ovv, and σ ohv from the C-band Sentinel-1, then from the Chinese Gaofen-3 (GF-3) satellite to evaluate their performances in discriminating between the three summer time ice types: floating ice (FI), brash ice (BI), and OW. FI includes FYI and MYI floes. This category was chosen because the backscatter from FYI and MYI are similar in the melt season. BI is found in gaps between ice floes in the summer. The combination of the three backscatter observations can only be used in the GF-3 data because this sensor is fully polarimetric. Sentinel-1 data offer only dual-polarization (σ ovv and σ ohv in this case). Comparison between using the dual-polarization data σ ohh and σ ovv from Sentinel-1 and GF3 in the MSI-ResNET classification scheme [Zhang et al., 2021b] is presented in Figure 11.10. The figure shows a subset of Sentinel-1A image acquired on 17 June 2017, which nearly coincides with the full scene of GF-3 with its 27 km swath. The scene is classified into the three categories mentioned above. Classification from the two SAR sets provides nearly the same results, yet with more details of brash/OW discrimination in the GF-3 results. More results presented in Zhang et al. [2021b] compare the classification using different sets of input: single σ ohh , single σ ovv , dual (σ ohh and σ ohv ), and the combination of (σ ohh , σ ohv , and σ ovv ). The dual polarization of σ ovv and σ ohv produced a fair improvement over the use of σ ovv , and insignificant improvement over the use of σ ohv polarization only. The input combination of the three polarizations from GF-3 produced an improvement in the overall accuracy and the accuracy of each surface type. However, the fading of sharp boundaries between brash and floe ice reduced the classification accuracy especially in areas of small ice floes.

Zhang’s algorithm

ECICE algorithm

Chen’s algorithm

1 2 4

3

5

0

dB

–40

0% 1%–20% 21%–40%

41%–60% 61%–80% 91%–100%

MYI

MYI concentration (%)

FYI

MYI

0

FYI

100

Figure 11.9 Sentinel-1 image of Viscount Melville Strait, Canadian Arctic, acquired on 23 March 2017 (left), followed by the weekly ice chart from the CIS for 20 March 2017, the MYI/FYI classification from Zhang et al. [2019], the ice type concentrations from ECICE algorithm and the ice classification from Chen et al. [2020] algorithm (courtesy of Shiyi Chen, SYSU, China).

OW/OIT

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 467 N

N

Sentinel-1A HH+HV

N

GF-3 HH+HV

36′

36′

36′

30′

30′

30′

80°N 24.00′

18′

18′

km

12′ 0 8°W

5

18′

km

12′ 10

30′

0 7°W

30′

6°W

30′

8°W

5

10 30′

0 7°W

30′

6°W

30′

FI BI OW

km

12′ 8°W

5

10 30′

7°W

30′

6°W

30′

Figure 11.10 Sentinel-1A Extra Wide (EW) image (left), and the sea ice classification results using Sentinel-1 (middle) and GF-3 (right). Classification is based on using dual-polarization data σ ohh , and σ ohv as input to the MSI-ResNET method. FI, BI, and OW stand for floe ice, brash ice, and open water, respectively [Zhang et al., 2021 / Reproduced from MDPI / CC BY 4.0].

11.1.3.3. Ice Classification from Polarimetric SAR Data The first experimental FP multi-frequency airborne SAR system (AIRSAR) was onboard NASA’s DC-8 aircraft in 1988. It operated at C-, L-, and P-band (wavelength 5.6, 23.5, and 68 cm, respectively). Drinkwater et al. [1991] compiled statistics of polarimetric parameters (no polarimetric decomposition parameters as defined in section 7.6.3.2 were available from airborne SAR sensor) from the three bands to characterize the broad categories of FYI and MYI along with the subcategories of thin- and thick-FYI in the Beaufort Sea and the marginal ice zone (MIZ) in the Bering Sea during March 1988. Although the study did not encompass any classification scheme, data from existing ice types led the authors to conclude that various combinations of wavelength and polarization offer better capability to distinguish between sea ice of different fabrics or morphologies. This theme was not seriously pursued after the AIRSAR was decommissioned in 2004. However, it was continued later when spaceborne FP SAR system became available. Among the three sets of parameters derived from polarimetric SAR observations (section 7.6.3.2), only two sets have been explored for their potential in sea ice classification: the polarimetric parameters and the polarimetric decomposition parameters. The first constitutes parameters derived directly from elements of the scattering matrix [equations (7.82–7.87)]. These are based on the intensity and phase measurements of the received signal. The second set constitutes parameters that represent the power from the three scattering mechanisms: SB, DB, and MB [equations (7.92–7.95)] (for the definitions of the mechanisms see Figure 7.23). Those parameters are derived from

decomposition of second-order matrixes, e.g., the coherency matrix given in equation (7.80), constructed from elements of the scattering matrix [equation (7.74)]. Findings from selected studies of sea ice classification using each set are presented in the following. However, it should be noted that polarimetric parameters from the FP SAR mode are available only from narrow swaths, e.g., up to 50 km wide from RADARSAT-2. This is too small to be useful for operational applications, which usually utilize images from a wide-swath mode of 300–500 kilometers. In an early study, Scheuchl et al. [2001] examined the capability of using the three parameters derived from Cloude-Pottier decomposition (entropy, anisotropy, and alpha-angle) for classification of FYI and MYI (see section 7.6.3.2). In another study, Wakabayashi et al. [2004] found the scattering anisotropy and beta-angle parameters from the same decomposition to be sensitive to ice type differences. More interestingly, the low depolarization characteristics of OW, hence its low entropy, was found to be useful in ice–water discrimination. This is confirmed in Wakabayashi and Sakai [2010], and later in Shokr et al. [2022]. Based on Wishart distribution, Dabboor and Shokr [2013] developed a new likelihood ratio, called the Bayesian Likelihood Ratio Test (BLRT), for the supervised classification of polarimetric data. Elements of coherence matrix were used to classify OW, and three FYI surfaces: smooth FYI (SFYI), rough FYI (RFYI), and deformed FYI (DFYI). The proposed BLRT is based on the asymptotic classification error probability derived by Chernoff [1952]. It is complemented with another new spatial criterion that incorporates spatial contextual information in order to produce more homogeneous classes. The authors

468

SEA ICE

applied the method to classify four ice types in the area of Franklin Bay, Canadian Arctic, using FP C-band RADARSAT-2 data. Results show that the combined use of the BLRT and the spatial criterion produces homogeneous classification results with better accuracy than using the Wishart classifier. Moen et al. [2015] investigated the use of the commonly known Pauli image, with RGB composed of scattering elements components |HH-VV|, 2|HV|, and |HH+VV|, to develop an automatic segmentation and classification of Arctic sea ice in the C-band FP SAR data. They examined the robustness of the results under slightly varying environmental conditions and different viewing geometries. The segmentation was performed using an unsupervised mixture-of-Gaussian segmentation algorithm utilizing six features extracted from the polarimetric data. After segmentation, the segments were contextually smoothened. Poor classification results were obtained when the incidence angle was very different between scenes. Account for incidence angle must be considered when wide-swath data from the CP (compact polarization) SAR mode are used to develop automated ice classification algorithm. Dabboor et al. [2017] evaluated the potential application of FP SAR from C-band RADARSAT-2 and L-band ALOS-2 imagery for monitoring and classifying sea ice during dry winter conditions. Twelve polarimetric parameters were derived and their capabilities of discriminating between FYI and OI were examined. Feature vectors

of effective C- and L-band polarimetric parameters are selected based on the criterion of minimum distance between the two ice clusters in the feature space. Kolmogorov-Smirnov distance is used for this purpose (a non-parametric separability criterion which measures the maximum absolute difference between two cumulative distribution functions). For the C-band, the feature space constitutes the SPAN, σ ohh , and σ ohv , and for the L-band it constitutes the entropy and the SPAN. The random forest classification algorithm was used. Results indicate that C-band SAR provides high classification accuracy (98.99%) of FYI and OI in comparison to the accuracy obtained from using the L-band SAR (82.17%). However, the L-band was found to classify only the MYI floes as OI, while merging both FYI and secondyear ice (SYI) into one separate class. This is probably because the air bubbles in the SYI have not matured in size, hence become transparent to the L-band wavelength. This is contrary to C-band, which classifies both MYI and SYI as OI. A potential for discriminating SYI from MYI by combining C- and L-band SAR is implied. Figure 11.11 shows the classification results from Dabboor et al. [2017] overlaid on the regional ice chart of 27 April 2015 over Hudson Bay, Canada, produced by the CIS. Note the wider coverage of the L-band ALOS2. The noticeable observation is the connection of OI floes in the RADARSAT-2 data and their separability in the ALOS-2. This occurs in an area defined as 100% OI

OI Section

OI section

FYI Section FYI Section

Legend

Legend

CIS lce chart

CIS lce chart

Thick-First-Year Ice

Thick-First-Year Ice

Old Ice

Old Ice

Polygen Boundary

Polygen Boundary

Classification Results

Classification Results

Ol

Ol

FYI

FYI

Figure 11.11 Classification results of FYI and OI overlaid on sea ice chart from the CIS for 29 April 2015, over Hudson Bay. Classification was based on SAR data from FP mode using, (a) RADARSAT-2 and (b) ALOS-2 SAR [Dabboor et al., 2017].

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 469

coverage in the ice chart. Figure 11.11 confirms the same remark made earlier that ice type classification using the fine-resolution SAR data reveals unprecedented details on individual or conglomerated ice floes that cannot otherwise be identified in the conservative ice charts from operational ice centers (in the sense of exaggerating the older and thicker ice). Also, it is obvious from the figure that the L-band data miss identification of a few (or many) OI floes compared to the C-band. This is because of the much longer wavelength of the L-band (around 23 cm) with respect to the characteristic dimensions of air bubbles, the main scattering elements in the OI, which measure a few millimeters or sub-millimeters (Tables 5.2 and 5.3). Ice-type classification has been traditionally performed using SAR backscatter, from single-channel, dual-channels, and continued when the backscatter from the three orthogonal channels with the phase information became available from FP SAR (this is grouped in the set of polarimetric parameters). In addition to the backscatter (intensity and phase), ice types also trigger different scattering mechanisms, which can be used to identify at least some types. The power from the three scattering mechanisms (SB, DB, and MB) can be calculated explicitly using Yamaguchi decomposition [equation (7.95)] or implicitly through the entropy, anisotropy, and alpha-angle [equations (7.92)–(7.94)]. So far, the majority of the studies that explore the potential use of the scattering mechanism to identify sea ice types have used the latter set. A sketch showing how NI can be identified based on its strong power from the MB mechanism is shown in Figure 7.26. Another example that links rafted ice to its strong SB scattering is shown in Figures 10.1 and 10.2, and a third example that shows how ridged ice triggers strong DB scattering is shown in Figure 10.6. Hence the use of the relative power of the scattering mechanisms should be recognized as a tool to reveal certain ice surface features. A few studies have been recently undertaken to explore the potential of using power from the three radar scattering mechanisms to discriminate between ice types. Hossain et al. [2014] examined the use of scattering power from the SB, DB, and MB mechanisms for discriminating three sub-types of FYI during late winter; smooth ice (SI), rough ice (RI), and deformed ice (DI), all in Hudson Bay, Canada. Data were obtained from RADARSAT-2 FP mode. Results show that surface scattering (SB mechanism) dominates from all three types (SI ≈ 77.3%, RI ≈ 66.0%, and DI ≈ 61.1%) at incidence angle 29 and decreases with increasing the incidence angle or surface roughness. Volume scattering is found to be the second dominant mechanism (SI ≈ 19.1%, RI ≈ 32.2%, and DI ≈ 37.4%) at 29 and increases with increasing incidence angle. DB scattering is triggered by ridges or deformed surface. If the relative power from the three scattering

mechanisms for each ice type is calculated, the clustering of the ice type data in the 3D feature space can be examined to determine their separability and potential discrimination. Johansson et al. [2018] suggested combining the scattering entropy and the co-polarization ratio to separate the newly formed sea ice from OW and thicker sea ice in winter and spring seasons using any radar frequency: X-, C-, or L-bands. The authors observed a high correlation between scattering entropy values calculated from Cand L-band data. They concluded that dual-polarization (HH and VV) from the X-band can be directly used to complement the FP C- and L-band scenes to discriminate between new and older sea ice. Singha et al. [2018] developed an automated sea ice classification scheme and applied it to near-coincident FP acquisitions from the ALOS-2, RADARSAT-2, and TerraSAR-X/TanDEMX satellites. The scheme consists of two steps, polarimetric feature extraction and ingestion of feature vectors into a trained neural network classifier to achieve a pixelwise supervised classification. Eigenvalue decomposition parameters derived from the coherency matrix were found to be of less relevance compared to parameters derived from decomposition of the covariance matrix. Among the most useful features for classification are matrix invariantbased features (geometric intensity, scattering diversity, and surface scattering fraction). Classification results show that 100% of the OW is separated from the surrounding sea ice, while separation between NI and older ice classes scores at least 96.9% accuracy. This has been consistent with results from later studies [e.g., Shokr et al., 2022]. The first attempt to explore the potential of using the relative power from radar scattering mechanism, derived directly from Yamaguchi decomposition of the covariance matrix (section 7.6.3.2), in ice type discrimination is presented in Shokr et al. [2022]. A few roles that determine the potential use of the scattering mechanism in identifying a few ice types and surface features are presented in section 7.6.3.3. An example of using the power from radar scattering mechanism to discriminate between the two surfaces of MYI: hummocks and melt pond is presented in the following. This is the first attempt to discriminate between the two surfaces and it can only be performed using the scattering mechanism power, not the total power (traditional backscatter intensity). It is well known that MYI returns high backscatter from both co- and cross-polarization SAR. Its contrast against the background ice is better in the crosspolarization images. The high backscatter is triggered by strong MB scattering, triggered by the bubbly sublayer of hummock ice. The pending question would then be about the melt pond ice. If it does not trigger high backscatter, which is likely, can it be distinguished from the

470

SEA ICE

hummock surface based on the scattering mechanisms? The answer is yes, hummock and pond ice trigger dominant MB and SB scattering mechanisms, respectively and this can be used to identify the two surfaces as shown in the following example. Figure 11.12 includes an image of the total power (SPAN) from an MYI floe (left top panel), taken from a scene of sea ice in the Resolute Passage, Canadian Arctic, acquired by RADARSAT-2 FP mode on 21 October 2017. The content of the red box in this panel is enlarged in the rest of the panels to reveal seven polarimetric decomposition parameters: SPAN, entropy, anisotropy, surface scattering (SB), volume scattering (MB), double-bounce scattering (DB) and the ratio of power from MB/SB (for definitions see section 7.6.2.3). A few observations can be drawn from the micro-look of the contents in the red box. In the middle of the SPAN image, there is a large area of low backscatter, which is interpreted as frozen melt pond surface, surrounded by high backscatter around its boundary, which is interpreted as hummock surface. This should be an acceptable interpretation based on the fact the subsurface layer of hummock and melt pond ice are bubbly and bubble-free, respectively (see Figure 2.15). In Figure 11.12 the entropy, the scattering power of SB, MB (Ps and Pv, respectively), as well as the ratio (Ps/Pv)

are high from the hummock compared to the melt pond ice. This is mainly because of the numerous scattering elements (bubbles) in hummock ice and their absence in melt pond ice. Therefore, those parameters can be used to discriminate between hummock and melt pond ice surfaces. In nature, the majority of the surface of an MYI floe is melt pond ice. The well-known high backscatter from MYI originates from hummock surface only. Another striking observation from Figure 11.12 is the apparent high DB from the perimeter around the hummock area (though its absolute value is still low, around -19 dB). This is conceivable as hummock and melt pond ice usually coexist next to each other, with the boundary possibly forming semi-dihedral configuration. In the near future, the CP SAR mode will be the prime data source for operational sea ice monitoring. So far, only tentative studies using simulated CP data (based on RADARSAT-2 FP mode data) have been presented to reveal the operational advantages of using data from this mode. A first trial to explore the best CP parameters for ice classification has been presented in Dabboor and Geldsetzer [2013]. The authors applied a maximum likelihood classification approach using different combinations of CP parameters and determined the best combination based on the classification error. A few CP

Figure 11.12 An MYI floe as appears in the SPAN image of Resolute Passage, (top left panel), acquired by RADARSAT-2 FP mode on 20 November 2017. The red box shows a frozen melt pond ice in dark tone surrounded by a hummock surface in bright tone. Images of different scattering parameters of the area marked by the red box are shown in succeeding panels. Note the contrast between melt pond (at the core) and surrounding hummock surfaces in each parameter.

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 471

parameters outperformed the linear polarization parameters. More studies on CP data are expected to emerge as the RADARSAT Constellation Mission (RCM) data have become available. The CP mode on RCM acquires data in ScanSAR configuration, with up to 500 km swath. This is useful for operational applications. It would be useful to close this section on the following note. Radar scattering mechanisms of snow-covered sea ice are triggered by several factors. They include the enlarged snow grains, the preferential orientation of snow grains relative to radar look direction, elevated brine volume in the ice, the scattering elements within the subsurface penetration depth, the dendritic interface at the bottom of NI, the scale of surface roughness, surface deformation, etc. A few of those factors were measured in field studies where a scatterometer system was deployed. Links between polarimetric parameters (that imply intensity and phase of the received backscatter) to snow and ice conditions could be established. However, if the purpose is to link the snow and ice properties to power from radar scattering mechanisms (not just intensity and phase of the signal), observations from space-borne polarimetric SAR (FP or CP modes) must be used in conjunction with coincident in-situ measurements of snow and ice. This has not been accomplished so far. Scattering mechanism power parameters are generated from second-order forms of scattering matrix (section 7.6.3.2) where elements are spatially averaged. It is this condition that renders ground-based scatterometer data not as useful as space-borne SAR for generating the scattering mechanism power data.

11.2. SEA ICE CONCENTRATION Sea ice concentration (SIC) is defined as the fraction of ice cover within a given area of the ocean. This can be a footprint of a satellite sensor or a delineated polygon in a satellite image. It is the most feasibly retrieved parameter from microwave remote sensing data. Several algorithms have been developed to estimate this parameter, some are limited to estimate the total ice concentration (regardless of ice types), while others add estimations of ice type concentration (called partial ice concentration). The importance of total ice and partial ice concentration to the marine operators cannot be overemphasized. Moreover, climatologists have become increasingly interested in SIC in order to improve climate models by assimilating this parameter in climate models [McLaren et al., 2006, Caya, Buehner, Carrieeres, 2010, Tietsche et al., 2013]. In winter, SIC determines the latent and sensible heat as well as the longwave radiative exchange between ocean and atmosphere. In summer, SIC affects the amount of absorbed solar radiation by the ocean in polar regions. SIC can be also used in determining a number of

climate-related parameters at the grid scale of climate models. This includes surface albedo, emissivity, heat and moisture fluxes between the warm ocean and the cold atmosphere. Ice volume can also be determined by combining SIC with ice thickness. Additionally, when SIC is combined with sea ice growth rate, salt migration to the underlying ocean water can be estimated. SIC is used to define ice edge [Comiso et al., 2001] and ice extent [Zwally et al., 2002]. Stroeve and Notz [2018] used this parameter to make projections of decadal trends of Arctic sea ice. The most commonly used sensor for estimating SIC has been the PM, but more use of scatterometer data has been demonstrated recently. Both sensors provide wide coverage that allows mosaicking the data for the entire polar region on a daily basis. Retrieval using PM data is based on the sharp contrast between emissivity of water and ice (Figure 9.31). This is not always achievable in scatterometer data because the backscatter from the wind-driven rough water surface overlaps with that from sea ice surface. The disadvantage of using both PM and scatterometer systems resides in their large footprints (a few kilometers or tens of kilometers). This leads to the generation of information at a coarse resolution not suitable for marine operation purpose. Even more seriously, the content of the footprint is almost always heterogeneous, composed of a few ice types plus OW. Decomposition of the single observation (e.g., brightness temperature or backscatter) from a footprint to its components from the different ice types and OW is necessary in this case. In a heterogeneous footprint that comprises ice and OW, the following linear model decomposes the single observation Robs into its components from two entities (ice and water in this case): Robs = Ri C i + Rw C w

(11.5)

where, C is the concentration, Ri and Rw are the typical values of radiometric or scattering observations (called tie points) from ice and OW, respectively. The unknowns in these equations are Ci and Cw, but they follow the equality condition Cw + Ci = 1. Similar equations can be constructed with more observations from the same footprint to estimate concentrations of more ice surfaces (ice types can be included in this case). In general, the number of observations must less than the number of ice types by one. This is because the equality constraint must be part of the linear system. In equation (11.5), the larger the separation between tie points from ice types, the more feasible and accurate the calculation of Ci would be. Equation (11.5) can be applied to any observation from optical, TIR or microwave sensor. Microwave data (from PM or scatterometer) are preferred because of their insensitivity to atmospheric constituents. Lower frequency PM channels provide higher contrast of Tb between OW (low Tb) and sea ice

472

SEA ICE

(high Tb) (Figure 9.31). The two deviations from this favorable contrast are found in the cases of very thin ice (as it represents the transition from OW to developed ice) and flooded ice surface during the spring and summer. In both cases neither PM nor scatterometer data can be useful tools for estimating SIC. The fine resolution of SAR warrants the homogeneity of the pixel contents. Therefore, the task of estimating SIC is reduced to assigning one ice type to each pixel. SIC across a given area can then be determined as the summation of pixels with same ice types, and then the summation of all types determines the total ice concentration. A challenge in using SAR for this purpose is the wide range of backscatter from the OW when the surface is wind-roughened. This causes backscatter from water to overlap with backscatter from most ice types (the unknown category in Figure 9.17 refers mostly to OW). SIC is produced regularly on daily or weekly basis at a few centers in the form of maps or charts. A partial list of those centers is included in Table 11.1 along with links to access the products. The SIC from those vendors is estimated either digitally using PM observations, or manually using visual analysis of satellite imagery, particularly SAR, by expert analysts. The rest of this section starts with a brief summary of SIC algorithms that use optical and TIR data. This is followed by detailed presentations of the formulation and processing of four SIC algorithms, three are based on PM data: NASA Team (NT), its enhanced version (NT2), ARTIST Sea Ice (ASI), and the fourth is based on combination of PM and scatterometer data; the ECICE. This latter algorithm has not been presented in detail in any open literature; hence its presentation includes more details here. Limitations of each algorithm are pointed out. The description of the four algorithms is followed by three subsections. The first is on intercomparison of PM-based algorithms. The second is on sources of

error and sensitivity of the algorithms. The third addresses the validation of SIC algorithms against ice charts. The section is concluded with methods of estimating SIC from SAR and scatterometer systems. Development of SIC algorithms will continue in the future to utilize the technological development of the space-borne sensors, which offer new observations to improve parameter retrieval. Small satellite platform and more SAR constellations carrying CP mode are interesting developments. Data from Chinese constellation of ocean dynamic environmental satellites have potential for CIS retrieval [Lin and Jia, 2022]. 11.2.1. Ice Concentration from Optical and TIR Images Visible (VIS) and thermal infrared (TIR) spectral data (processed separately or combined) are used to estimate SIC. Ice and OW can be distinguished based on their difference in albedo and surface temperature. The advantage of using these data is their relatively finer spatial resolution compared to the PM data. This allows estimation of SIC in narrow passages and archipelago regions. At the negative side, the unavailability of the VIS data during the dark polar season and the contamination of both VIS and TIR observations by clouds and other atmospheric constituents limit their use. SIC products from these sensors are generated on an opportunity basis at a few centers of sea ice monitoring. The OSI-SAF products from the Norwegian Meteorological Institute (DNMI) include SIC maps based on AVHRR data from three channels (2, 3, and 4) [Godøy and Eastwood, 2002]. It uses a Bayesian approach as described by equation (11.2). Following the application of a cloud mask, the algorithm uses two parameters from AVHRR: albedo from channel 2 (A2) and the difference of brightness temperatures between channels 3 and 4 (T3 − T4). Sea ice has higher A2 and lower T4 than OW.

Table 11.1 National centers and institutions that produce SIC maps routinely. The links to access the data are shown. Temporal frequency and geographic regions of map production are not included. Center/institution Canadian Ice Service (CIS)

Link https://www.canada.ca/en/environment-climate-change/services/iceforecasts-observations.html https://osi-saf.eumetsat.int/products/sea-ice-products

EUMETSAT’s Ocean ad Sea Ice Satellite Application Facility (OSI-SAF) NANSEN Environmental and Remote Sensing Center National Center for Atmospheric research (USA) (NCAR) National Snow and Ice Data Center (USA) (NSIDC)

http://www.arctic-roos.org/ https://climatedataguide.ucar.edu/climate-data http://nsidc.org/data/seaice/

Technical University of Denmark (PolarView) University of Bremen (Germany) US National Ice Center (NIC)

http://www.seaice.dk/ https://seaice.uni-bremen.de/sea-ice-concentration/amsre-amsr2/ https://usicecenter.gov/Products

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 473

When A2 is not available during the dark season, it is replaced by the difference of TIR observations from channels 4 and 5 (T4 − T5). For ice–OW discrimination using A2 and (T3 − T4), equation (11.2) can be re-written as: p ice A2 , T 3 − T 4

= p A2 ice p T 3 − T 4 ice p ice p A2 ice p T 3 − T 4 ice p ice + p A2 ow p T 3 − T 4 ow p ow

(11.6) The a priori probabilities of ice and water p(ice)and p(ow) in the above equation can be obtained using climatological ice records or can be set to 0.5 in the absence of this information. The conditional probabilities can be obtained from pixels of known surface area (training areas). Godøy and Eastwood [2002] found that those probabilities can be approximated by a Gamma distribution. This distribution has the advantages of simulating the Gaussian distribution around its peak while reproducing the skewness in the data. px =

x−γ

α−1

exp − x − γ β ; α > 0; β > 0; x > γ βα Γ α (11.7)

where, α, β, and γ are distribution parameters, and Γ is the gamma function. The parameters were determined for each month and each surface using training data. Killie et al. [2011] extended this approach by using four AVHRR spectral features for pixel classification: channel 1 bi-directional reflectance, the ratio of the bi-directional reflectance from channel 2 and channel 1, the same ratio between channels 3 and 1, and the temperature difference between the AVHRR channel 4 brightness temperature and a modeled surface temperature. The Bayesian equation [equation (11.6)] is extended to combine the four satellite spectral features. Once again, each pixel is assigned to the class with the highest a posteriori probability. The conditional probability density of each class can be approximated using an appropriate standard distribution (e.g., normal or gamma distribution functions). Meier [2005] estimated SIC from AVHRR VIS/IR channels and compared it against estimates from PM data using four algorithms: BT, Cal/Val, NT, and NT2 (for designation see the opening of section 11.2.2). In the absence of solar radiation during winter, ice and water can be discriminated using surface temperature from a TIR channel. A threshold of 271 K was selected to assign a pixel to ice when it is below this temperature. This method assumes homogeneous contents of the footprint, which is not always a valid assumption, especially at locations near and within the ice edge. Figure 11.13 presents a case study from Meier [2005] showing SIC estimates from AVHRR compared to estimates from the aforementioned four PM algorithms. In the original AVHRR channel 2 data (top left panel), the ice cover is broken with

considerable OW between large and small floes. The white contour delineates the region of clear sky in the AVHRR image, for which the concentration can be determined. The AVHRR concentration map shows less concentration along the right side of the contour, where water is clearly visible in the image. Other than that, the concentration is nearly 100% as can be visually identified in the original image. The NT algorithm gives the lowest SIC estimate in this summer scene, and Cal/Val algorithm gives the highest estimate. Results from Bootstrap and NT2 algorithms are similar. The advantage of the using finer resolution of AVHRR is apparent in the figure. Violation of heterogeneity assumption of the footprint, implied in the above-described threshold method, causes an overestimation of SIC. However, a previous study by Emery et al. [1991] concluded that this error is small compared to errors from using the SSM/I for SIC. During summer the surface temperature of ice and water become similar. Then the reflectance from a VIS channel can be used not only to discriminate between the two surfaces but also as an indicator of the proportion of water and ice in each pixel. This requires estimation of mean values (tie points) of the reflectance from each surface using samples of homogeneous ice and water pixels. The error in SIC using this approach data was determined to be between 5%–20% (winter and summer) [Comiso and Steffen, 2001], 6.8%–15.1% (summer), and 8.6%–26.8% (winter) [Emery et al., 1991]. 11.2.2. Ice Concentration from Coarse-Resolution Microwave Observations PM observations are most suitable for SIC estimates for two reasons. First, water has much lower emissivity than ice, and therefore becomes radiometrically cold in the microwave region even if it is physically warmer than ice. Second, microwave emission from water is more biased toward vertical polarization, making a large polarization difference between vertically and horizontally emitted radiation, while ice produces much smaller difference (Figure 9.8). The polarization difference is evident at all frequencies but is more pronounced in data from lower frequencies. It is a prime criterion used in most SIC retrieval algorithms. While PM sensors acquire data at coarse resolutions (a few kilometers or tens of kilometers), they can be resampled into finer resolution of 2–5 km over the polar regions where many overlapping swaths exist per day (around 14 daily overpasses). Such reconstructed fineresolution maps are available from the Microwave Earth Remote Sensing Laboratory of Brigham Young University through the ftp link ftp://ftp.scp.byu.edu/pub/data. The techniques for the data reconstruction from PM sensors and scatterometer systems are presented in Early and Long [2001] and Long and Daum [1998], respectively. Recently,

474

SEA ICE

AVHRR Channel 2

BT 72%

Cal/Val 81%

Novaya Zemlya

Franz Joseph Land

AV 79%

0

NT2 74%

10

20

30

40

NT 68%

50

60

70

80

90

100

Figure 11.13 AVHRR Channel 2 scene of the Barents Sea on 16 June 2001(top left), and SIC maps from different algorithms (see text for algorithm labels). Average concentration is shown from each algorithm. The clear-sky region in AVHRR image, for which SIC is calculated, is outlined by the white contour. Black corresponds to coast line pixels for which no concentration could be obtained from the PM data. [Meier, 2005 / with permission from IEEE].

Meier and Stewart [2020] presented a technique for enhancing the resolution of gridded brightness temperature from PM sensors and used it to improve the retrieval of SIC. High-frequency PM radiometers (e.g., 85 GHz) provide finer resolution, but the emitted radiation becomes affected by atmospheric influences as described in section 7.9.1.2. These influences should be accounted for before using the data in an SIC algorithm. SIC retrieval algorithms from PM data include the following (the list is arranged in chronological order): NORSEX [Svendsen et al., 1983], NT [Cavalieri, Gloersen, Campbell, 1984], the University of Massachusetts– Atmospheric Environment Service (UMass-AES) [Swift

et al., 1985], the Bootstrap algorithm (BS) [Comiso, 1986], the Svendsen’s near 90 GHz algorithm (N90GHz) [Svendsen, Mätzler, Grenfell, 1987], the calibration-validation (Cal/Val) [Ramseier, 1991], the Bristol algorithm (BRI) [Smith and Barrett, 1994], the Atmospheric Environment Service–York University (AESYork) [Rubinstein, Nazarenko, Tam, 1994], the NASA thin ice algorithm [Cavalieri, 1994], the Technical University of Denmark hybrid (TUD) [Pedersen, 1998], the Enhanced NASA Team (NT2) [Markus and Cavalieri, 2000], the Arctic Radiation and Turbulence Interaction Study (ARTIS) Sea Ice (ASI) [Kaleschke et al., 2001], SEALION [Kern, 2004], and the ECICE [Shokr, Lambe,

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 475

Agnew, 2008]. Recently, Dworak et al. [2021] blended ice concentration from PM (AMSR2) and VIS/IR satellite data (VIIRS) to take advantage of the all-sky capability of AMSR2 and the fine resolution of VIIRS, both are resampled to a 1 km EASE-grid map. The blended product outperformed the product from each sensor though it is much closer to the product from AMSR2. While different algorithms vary in details, they can be grouped into four categories according to the data processing method. (1) Solving a set of linear algebraic equations, similar to equation (11.5), where each equation decomposes a given radiometric observation into components pertaining to the OW and each specified ice type within a heterogeneous footprint, weighed by their partial concentration. An example is the NT algorithm. (2) Searching a database of simulated observations, generated using a large number of possible ice concentrations, to find the nearest point to the given set of observations. The simulated observations usually include brightness temperature or a few derived parameters. The simulation is conducted using a radiative transfer model with ice concentrations and atmospheric conditions vary by fixed intervals. An example is the Enhanced NASA Team (NT2) algorithm. (3) Solving a simplified version of a radiative transfer equation that accounts for atmospheric influences and incorporates different ice concentrations. This approach can be used if the observations are affected by the atmosphere, e.g., a high-frequency PM channel. Examples include N90GHZ and ASI algorithms. (4) Finding an optimal solution that minimizes a cost function, which represents the difference between an actual and a modeled observation. The cost function includes SIC. The example is the ECICE algorithm. Mathematical schemes of selected algorithms representing the above approaches are presented in the next subsection. NT and the NT2 algorithms represent the first and second approaches, respectively. ASI algorithm represents the third approach. The fourth approach is represented by ECICE algorithm. Outlines of other algorithms; namely NORSEX, BS, N-90, Cal/Val, BRI and TUD are presented in Andersen [2000]. Earlier summaries of NORSEX, NT, Mass-AES, BS and AES-York algorithms are presented in Steffen et al. [1992].

11.2.2.1. NASA Team (NT) Algorithm The original version of this algorithm was developed by Cavalieri, Gloersen, Campbell [1984] to retrieve total SIC in addition to partial concentrations of FYI and MYI. It was applied to SMMR, and later adopted to calculate the same parameters from SSM/I observations [Cavalieri et al., 1991]. Another version of the algorithm (NT2) was developed to calculate concentrations of thin ice

and FYI [Cavalieri, 1994]. The algorithm solves a set of linear equations that take the form: T b,obs = C ow T b,ow + C FY T b,FYI + C MY T b,MYI

(11.8)

where, Tb,obs is the observed brightness temperature, Tb,ow , Tb,FYI , and Tb,MYI are the typical Tb from OW, FYI, and MYI, respectively. Here Cow, CFY, and CMY are the unknown concentrations of the three surfaces. Hence, we need two equations similar to equation (11.8). The third equation stipulates that the summation of the fractional concentration from all surfaces ci should equal to 1: n

ci = 1

(11.9)

i=1

The NT algorithm uses the polarization ratio PR18 and the gradient ratio GR37V18V in the RHS of equation (11.8) to form two equations. The general expressions of these ratios are presented in equations (9.1) and (9.2) and repeated here using the actual radiometric parameters used in the NT algorithm. PR18 =

T b,18V − T b,18H T b,18V + T b,18H

GR37V18V =

T b,37V − T b,18V T b,37V + T b,18V

(11.10) (11.11)

The use of the polarization ratio is necessary because of its significantly higher value for water compared to ice. The justification for using the gradient ratio is that it varies markedly in magnitude and sign between the three surfaces of OW, FYI, and MYI. After adding equation (11.9), the three equations can be solved simultaneously to obtain Cow , CFY , and CMY. This simple method does not account for surface conditions that cause significant deviation of the emitted radiation from the tie points. For example, the emitted radiation from YI types may vary over a wide range and therefore deviate significantly from the specified tie point Tb,YI. This behavior was confirmed in previous laboratory and theoretical studies. Grenfell and Comiso (1986) show that significant changes in ice surface emissivity and polarization ratio occur as ice grows up to 5 cm thick. More on the erratic radiometric behavior of YI surface under variable meteorological conditions is shown in Figure (11.21). Deviation of an actual observation from the tie point of certain ice type may lead to erroneous results of SIC. To overcome this shortcoming, another version, called NT thin ice, was developed by Cavalieri [1994]. Nevertheless, the simple approach of solving algebraic equations as described above does not have a provision to avert unrealistic solutions of SIC above 100% or below 0%. In these cases, the output concentration is

476

SEA ICE

and as concentration decreases the two gradient ratios will have different values, hence the points start tom devoate from line AB. In order to resolve the ambiguity between points of true SIC and points of surface effects, ΔGR is used. Figure 11.14c a narrow cluster along line AB. Points along that line represent 100% SIC with different gradient 11.2.2.2. The Enhanced NASA Team (NT2) Algorithm values corresponding to different ice types. As SIC The NT2 algorithm is presented in Markus and Cavadecreases, the two gradient ratios will have slightly differlieri [2000]. The algorithm has the same functional form ent values and hence the points start to deviate slightly as the NT algorithm, but with a different approach based from line AB. When surface effects start to be proon a search in a parameter space for a solution that satisnounced, points deviate even further from line AB toward fies the closest distance with a given observation vector. increased GR85H19H. The cloud of points labeled “C” in The algorithm is designed to account for the “surface Figure 11.14b corresponds to the cluster of points labeled effects” that cause underestimation of SIC by NT. Those “C” in Figure 11.14a. It is obvious then that the gradient effects include snow glaze and ice layering within the snow difference represented by equation (11.14) takes higher cover, which have proven to be problematic in estimating values for ice type C. The algorithm uses a threshold on SIC [Comiso et al., 1997]. They are grouped in what is ΔGR to designate points of “surface effects” (i.e., ice type referred to as the C-type ice. Therefore, in addition to the C). Figure 11.14d shows the location of ice types C with total SIC, NT2 outputs the partial concentration of either respect to types A and B in the ΔGR, PRR (19) space. one of two ice types: thick ice (including FYI and MYI comSince 85 GHz data are sensitive to atmospheric condibined) and thin ice, or thick ice and C-type ice. The selection tions, the algorithm implements a correction for the of the output is performed within the algorithm based on a atmospheric influence. The method involves calculations certain criterion (i.e., it is not the user’s choice). of the expected (modeled) brightness temperature from all The algorithm uses PR19 and GR37V19V in addition to SSM/I spectral channels using a forward microwave radithree more ratios, PR85, GR85H19H, and GR85V19V. From ative transfer model developed by Kummerow [1993] for these five ratios, the algorithm uses three derived paradifferent surfaces and atmospheric conditions. Two pairs meters defined as follows: of surface types were used to output model results: FYI PRR 19 = − GR37V 19V sin ϕ19 + PR19 cos ϕ19 (11.12) and thin ice, and then FYI and C-type ice. Twelve atmospheric conditions were selected ranging from clear skies to PRR 85 = − GR37V 19V sin ϕ85 + PR85 cos ϕ85 (11.13) winter cumulus cloudy conditions. Each condition was characterized in terms of cloud type, cloud liquid water ΔGR = GR85H19H − GR85V 19V (11.14) contents, base height, and top height. These conditions where PRR(19) and PRR(85) are called rotated polariza- characterize different atmospheres in summer and winter tion ratios and ΔGR is called the gradient difference. seasons in the Arctic. Figures 11.14 b and 11.14d show Figure 11.14 shows scatter plots of parameters derived the model’s data (gray circles) in the ΔGR-PRR(19) and from SSM/I data over Weddell Sea in the Antarctic, ΔGR-PRR(85) parameter spaces. It can be seen that acquired on 15 September1992. In Figure 11.14a the gray model results span the width of the observed clusters. circles represent the tie points of OW as well as ice types A clear atmosphere has the lowest ΔGR for each surface. A and B, which correspond to FYI and MYI types, More importantly, variation in atmospheric conditions respectively. The line AB represents 100% ice concentra- result in a wider range of PRR(85) than PRR(19). The tion and the distance from the OW point to that line is a range of PRR(19) is very narrow for OW and almost zero measure of ice concentration. The algorithm uses the for the two ice types presented in Figures 11.14c. This rotated polarization ratio, PRR(19), which results from means that PRR(19) is a better parameter to resolve total rotating the axes in Figure 11.14a by an angle ϕ19, defined SIC, while PRR(85) is a better parameter to resolve atmosas the angle between the GR axis and the AB line. As pheric conditions. It can also be seen in Figure 11.14d that such, PRR(19) becomes independent of ice type. Label PRR(85) is also nearly independent of surface effects, C indicates pixels with significant surface effects such as resulting in an almost vertical cluster of points from all glaze and ice layering and it is shown in Figure 11.14b ice types, with or without surface effects. Nevertheless, separated from the cluster of A/B. Surface effects increase the C-type ice appears radiometrically distinct from other both PR19 and ΔGR (because they cause significant ice types on the ΔGR axis. This highlights another reason decrease in Tb, H). Figure 11.14c shows a narrow cluster to use PRR(85) in the calculation scheme. In summary, the rationale behind using the three specalomg line AB with similar values of the two gradient ratios. It is reported in Markus and Cavalieri (2000) that tral ratios in the NT2 algorithm is as follows. First, the points along that line represent 100% ice concentration rotated polarization ratio from the 19 GHz channel is truncated at 100% if it exceeds this value or at 0% if it becomes negative. Nonetheless, the NT algorithm is still being used to provide partial concentrations of FYI and MYI, which is needed at this time as the MYI in the Arctic is replenished by FYI.

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 477 (a)

(b) 0.20

0.10

g sin rea Inc ather we

OW 0.15

0.05

ΔGR

GR37v19v

OW A

0.00

0.10 C

–0.05

C

0.05

0.15

0.00 0.00

B –0.10 0.00

(c)

0.05

A/B 0.10

0.20

0.25

0.30

PR(19)

0.05

0.10

0.20

0.25

0.30

PRR(19)

(d)

0.2

0.15

0.20

Inc r wea easing the r 0.15

OW

0.1

ΔGR

GR85v19v

OW 0.10

0.0 A

0.05

Ice C

B –0.1 –0.10

A/B 0.00

0.10

0.20

0.30

0.40

GR85H19H

0.00 0.00

0.05

0.10

0.15

0.20

PRR(85)

Figure 11.14 Scatter plots of SSM/I observations over the Weddell Sea, acquired on 15 September 1992, presented in different parameter spaces [see equations (11.12) to (11.14)]. Gray circles are tie points of OW and ice types A, B, and C (see definitions in the text). Definition of angle ϕ is also shown in (a). The gray circles in (b) and (d) are modeled ratios for the three pure surface types (A/B, C, and OW) with different atmospheric conditions. Increasing weather means more clouds. Deviation of the data from the line in (c) is an indication of decreasing ice concentration [Markus and Cavalieri, 2000, Figure 1, with permission from IEEE].

used to facilitate discrimination between sea ice and water. Second, the gradient difference is used to resolve the ambiguity between pixels with low SIC and pixels with significant surface effects, where observed brightness temperature is mainly caused by metamorphosed snow rather than the ice surface. Third, the rotated polarization ratio from the 85 GHz channel is used to avoid ambiguity between changes in SIC and changes in atmospheric conditions (due to a higher sensitivity of the 85 GHz channel to atmospheric variability). The calculation scheme of the NT2 algorithm is presented in Figure 11.15. The search

for the closest modeled data point to the given observation is shown in the block at the bottom right of the diagram. As mentioned earlier, modeled data represent output from using different combinations of concentrations and atmospheric conditions in the forward microwave radiative transfer model. The calculation steps are summarized as follows. 1. The forward microwave radiative transfer model is used to generate a set of modeled SSM/I brightness temperature data for four “pure” surfaces: FYI/MYI, thin ice, C-type ice, and OW. For each surface, the 12 different

478

SEA ICE Modeled SSM/I TB for linear combinations of four surface types and 12 different atmospheric conditions

Calculate pol. and grad. ratios for all concentrations (0–100) and all twelve atmospheric conditions:

Observed SSM/I TB

Calculate set of ratios: Robs = {PRR(19), PRR(85), ΔGR}

Rmod = {PRR(19), PRR(85), ΔGR} Atmosphere index

Find weather-corrected ice concentration by minimizing δR = sqrt (Robs – Rmod)2

?

Ice concentration Cfy

Modeled value Rmod Ice concentration Cth

? Observed value Robs

Figure 11.15 Schematic diagram of illustrating the calculation procedures in the NT2 algorithm.

atmospheric conditions are used to generate a total of 48 sets of brightness temperatures. Each set has 7 spectral points corresponding to the SSM/I spectral channels. 2. Each set is considered to be a set of tie points and used in equation (11.8) to calculate the modeled radiometric parameters, then the modeled input parameters [equations (11.12) –(11.14)] for different combinations of concentrations of the three ice types: thick ice, thin ice, and OW; or thick ice, C-type ice, and OW. The three ice types are grouped in different combination of concentration ratios from 0% to 100% in a step of 10%. 3. The calculated parameters are then expressed in the vector form: Rmod = [PRR(19), PRR(85), ΔGR]mod 4. For an actual set of brightness temperature observations, the same three parameters PRR(19), PRR(85), and ΔGR are calculated and expressed in the vector form: Robs = [PRR(19), PRR(85), ΔGR]obs 5. A search is conducted to find the closest Rmod value to the given Robs. The criterion used is the minimum error between the observed and the modeled observations represented as: δR =

Robs − Rmod

2

(11.15)

The algorithm selects the route to generate thin ice or C-type ice (in addition to thick ice) based on a threshold value of ΔGR. If ΔGR > 0.05, modeled data are generated using emissivity of thick ice (basically FYI emissivity) and thin ice along with the atmospheric index. If ΔGR < 0.05, modeled data are generated using C-type ice to replace thin ice.

11.2.2.3. The ASI Algorithm The Arctic Radiation and Turbulence Interaction Study (ARTIST) program was conducted in March and April 1998, in the area around Svalbard to measure sea ice parameters using ground-based and airborne instruments. Based on the measurements, the ARTIST Sea Ice (ASI) algorithm was developed to estimate SIC using the high spatial resolution of the 85 GHz channels of the SSM/I [Kaleschke et al., 2001]. This algorithm is an enhanced version of the N90 algorithm developed by Svendsen, Mätzler, Grenfell [1987], which incorporates a simplified version of the radiative transfer equation to account for atmospheric influences on the observations. It is based on the familiar notion that the polarization difference P is a function of SIC: P = CPs,i + 1− C Ps,w ac

(11.16)

where, C is the ice concentration, Ps, i and Ps, w are the polarization difference for ice and water, respectively. The term between the square brackets in the RHS represents the surface polarization difference Ps (i.e., the composition of P from ice and water contents within the observation cell). The term ac represents the atmospheric influence and is a function of ice concentration. Svendsen, Mätzler, Grenfell [1987] developed the following equation to determine ac: ac = Ps e − τ 1 1e − τ − 0 11

(11.17)

where, τ is the atmospheric opacity. This equation is valid for a horizontally stratified atmosphere with an effective temperature replacing the vertical temperature profile. Denoting the atmospherically corrected polarization

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 479

difference from homogeneous OW and consolidated sea ice cover P0 and P1, respectively, their expressions are given by: P0 = a0 Ps,w

(11.18)

P1 = a1 Ps,i

(11.19)

where, a0 and a1 are terms to account for atmospheric influence over OW and ice surfaces, respectively. Taylor expansion of equation (11.16) around C = 0 and C = 1 produces the following equations if higher terms are neglected: P C = a0 C Ps,i − Ps,w + P0 for C

0

P C = a1 C− 1 Ps,i − Ps,w + P1 for C

(11.20) 1

(11.21)

Substitution of the atmospheric correction terms from equations (11.18) and (11.19) into equations (11.20) and (11.21) leads to: C= C=

P −1 P0

P + P1

Ps,w Ps,i − Ps,w

P −1 P1

for C

Ps,w Ps,i − Ps,w

0

for C

(11.22) 1

(11.23)

Under Arctic conditions the term [Ps, w/(Ps, i − Ps, w)] was determined by Svendsen, Mätzler, Grenfell [1987] to be approximately -1.14. In order to retrieve all values of SIC, an interpolation between the two solutions given in equations (11.22) and (11.23) should be conducted. Kaleschke et al. [2001] selected a third-order polynomial for the interpolation between 100% OW and 100% sea ice: C = d 0 + d 1 P3 + d 2 P3 + d 3 P3

(11.24)

Using equations (11.22) and (11.23) along with their derivatives, the coefficients in equation (11.24) can be obtained by solving the following linear equation system: P30 P31 3P30 3P31

P20 P21 2P20 2P21

P0 P1 P0 P1

1 1 0 0

d3 d2 d1 d0

=

0 1 − 1 14 − 1 14

(11.25)

With the tie points P0 and P1, the coefficients d0 to d3 can be determined from equation (11.25), and therefore SIC can be determined from equation (11.24). It is possible that the output concentration may exceed 100% or become less than 0%. In this case truncation of the solution at these limits is enforced. The correct selection of the tie points is important as they also incorporate the correction for atmospheric influences. 11.2.2.4. ECICE Algorithm Estimation of ice type concentration from remote sensing data is hampered by two factors: (1) the difficulty of

representing each ice type by a single tie point (especially YI), and (2) the possibility of obtaining ice concentration values outside the feasible domain between 0% and 100%. This possibility arises from solving a set of deterministic equations for each observation [similar to equation (11.8)]. ECICE algorithm was developed to avoid these two difficulties [Shokr, Lambe, Agnew, 2008]. It is presented with its mathematical details in the following. ECICE is a generic algorithm. It can be applied to determine the concentration of any number of surfaces in a heterogeneous footprint using a minimum number of observations equal to the number of surfaces minus one. The reason for the “minus one” is the presence of the extra equation of equality constraint which stipulates that the summation of the concentrations of all surface types is 100%. For example, if the algorithm is applied to estimate the concentrations of FYI, MYI, and OW in a heterogeneous cell, the minimum number of observations must be two. Extra observations are recommended because, after all, it is an optimization technique that searches for solution that minimizes a cost function (i.e., not a deterministic solution of a set of equations). Input observations may originate from different channels of the same sensor or a combination of sensors regardless of their categories, i.e., the algorithm can combine input from VIS, TIR, passive and active microwave sensors. Special modules are incorporated to address the peculiarities of certain sensors (e.g., atmospheric and cloud influences in the case of the VIS or TIR data or wind effect on backscatter from OW in the case of radar backscatter). It is worth noting one exception regarding the application of the algorithm. If the number of surfaces is two (e.g., sea ice and OW) and the number of observations is one (i.e., the minimum number required in this case), then there is no point in using ECICE with its optimization scheme because it will produce the same results as solving the algebraic system (equations (11.26) and (11.27) in the following) (for confirmation see section 4.2). Three new concepts are employed in ECICE. The first is using a large number of tie points (instead of a single tie point) to represent each ice type. These are called characteristic radiometric values (CRVs), and are generated from the probability distribution of each radiometric parameter for each surface type. The distribution of each parameter from the CRV set should mimic the original probability distribution. For example, if three surface types are specified and three observations are used, then a set of CRV will constitute values representing each observation from the three types. This is repeated for each observation. Sets of CRVs are generated from the probability distributions of each observation for each surface type (an example is provided later). The probability distributions are input to the algorithm and a large number of CRV sets (typically, 1000–5000 as specified by the user)

480

SEA ICE

are generated using a random number generator for each pixel. The use of this large number leads to the construction of an equally large number of solutions. ECICE does not seek a single concentration solution from a set of deterministic equations. Instead, it seeks an optimal solution of a limited number of unknowns (i.e., concentrations of the given surface types) using a set of equations where the number of equations is larger than the number of the unknowns, hence the need for optimal solution. Since thousands of CRV sets are used (in fact, each CRV can be viewed as a set of tie point), a very large number of set solution can be generated for each pixel. One solution is constructed as shown below. The second concept entails inclusion of a set of inequality constraints in the formulation to restrain the output concentration within the feasible domain of solution, namely between 0%–100%. This is in addition to the usual equality constraint, which entails that the summation of the partial concentration from each surface must equal 100%. This concept leads to seek an optimal solution of a set of linear equations in addition to inequality constraints. Standard solutions for such formulation that include inequality constraints are available in text books. One method is selected as described below. The third concept is the possibility of assigning a confidence level (a measure of uncertainty) for the final set of ice concentrations at each pixel (a set of solution is composed of total ice concentration and partial concentration of each specified ice type). This is made possible because the method produces as many solutions as the number of CRVs used in the constructed equations. With 1000 sets of CRVs, the algorithm produces 1000 sets of solution. While one average solution is selected (the final answer), the 1000 solutions are used to generate the confidence level as explained below. More details of the formulation and calculations follow. The expected value of a radiometric observation j, denoted Rej, is calculated from a set of linear equations similar to equation (11.8). The linear model can be written in the form: n

Re,j =

ci T i,j

(11.26)

i=1

where, ci is the concentration of the surface type i (i = 1, …n) and n is the number of surfaces including OW. Ti, j is the typical value of the observation j from surface i. The number of observations must be equal to or higher than the number of ice types (i.e., not including OW). In addition to these equations, the formulation includes an equality constraint; the summation of the partial concentrations must equal 1: i=n

ci = 1 i=1

(11.27)

The ice concentration vector

c

is the solution that

minimizes the following cost function. The function represents the summation of the error between the observed and the expected value of each parameter employed in the formulation at each pixel. The elements of

c

are

the partial concentrations of each surface: f c

2

m

n

j=1

i=1

=

− Ro,j

ci T i,j

Nj

2

i=n

ci − 1

+ i=1

(11.28) where, Ro, j is the radiometric observation j, m is the number of observations, Nj is a normalization factor which is introduced to ensure that the error terms in equation (11.28) have the same order of magnitude for each surface type. The factor is the average of the observation j from all surfaces; i.e., Ti,j for all i. Equation (11.27) appears in the cost function since its exclusion would cause the matrix of the second derivatives of the function to be ill-conditioned or singular, making the problem difficult to solve (this matrix is needed to find the minimum of the cost function). The above formulation is not, however, complete. In order to guarantee that the solution for each partial concentration will fall within the feasible domain between 0% and 100%, the formulation includes the following inequality constraint for each ci : 0 ≤ ci ≤ 100

(11.29)

With equation (11.29), the problem turns to be seeking optimal solution for a set of equations that includes inequality constraints. In unconstrained optimization, the necessary and sufficient conditions for a minimum are defined by setting the first derivative of the objective function to zero while the second derivative should be greater than zero. In the case of multivariable problems, the gradient vector is zero and the eigenvalues of the matrix of second derivatives must all be greater than zero. On the other hand, in a formulation with inequality constraints the notion of Lagrange function L is introduced [Fletcher, 1987]: n

L c,λ = f c −

λi Ai c − bi

(11.30)

i=1

Here, Ai c − bi

is an active constraint, re-written in

vector form, which originates from the inequality constraint i,λi is the corresponding Lagrange multiplier, and n is the total number of active constraints, which is determined during the optimization process. Minimizing the Lagrange function [equation (11.30)] is equivalent to minimizing the objective function [(equation (11.28)]

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 481

subject to both the equality and inequality constraints, provided that the correct set of active constraints is found. The concept of active constraint is introduced here. Equality constraints are always active because feasible solution points include those that fall on the boundary formed by the constraint. An inequality constraint, on the other hand, may or may not be active. It is said to be active at a feasible point if that point lies on the boundary of the feasible domain of the solution. This condition will arise if the solution that minimizes the cost function is found to be violating the constraint. In this case, the solution is forced to be relocated to the boundary of the constraint (as will be further explained later). When an inequality constraint becomes active it will take the form of an equality constraint because the feasible solution is now located on the boundary of the inequality constraint. For example, ci = 0 (i = 1, 2, …n). The coefficients for the active constraints in equation (11.30) are Ai = [1, 1, 1, 1] and bi = 1. If an inequality constraint of ci has become active (e.g., ci = 0), then Ai = 0 for all i except for the component k, while bi = 0 for all i. For example, if c2 becomes active (c2 = 0), then Ai = [0, 1, 0, 0] and bi = 0. Note that, not all inequality constraints will be active at the same time. ECICE uses Newton’s method to find the solution that minimizes the Lagrange function [equation (11.30)]. The solution is obtained by taking a second-order Taylor expansion of the function around an initial feasible point and set the first derivative to zero: ∇L c, λ = 0 where∇ =

∇c ∇λ

(11.31)

The parameter λ is the set of Lagrange multipliers that satisfies the solution. The result is the following system of linear equations: ∇2 f −A

− AT 0

Δc λnew

=

− ∇f 0

(11.32)

In this system, A is the matrix formed by taking each Ai as a row. Solving this system yields a step Δc toward the solution, and a new estimate of the Lagrange multipliers λnew. As mentioned previously, the matrix of second derivatives in this system cannot be ill-conditioned, otherwise this system becomes numerically difficult to solve. The first and second derivatives that appear in equation (11.32) are computed analytically from the objective function [equation (11.30)]. The first derivatives of the Lagrange function take the form: ∂L ∂f = − ∂ci ∂ci

n

λi Aij

for j = 1, …n

(11.33)

i=1

∂L = − Ai c − bi for i = 1, l ∂λi

(11.34)

where, n is the number of active constraints. Note that if all active constraints are satisfied, then equation (11.34) evaluates to zero. Moreover, the first derivative of the Lagrange function with respect to concentration is not exactly the same as the RHS of the linear system, due to the presence of the Lagrange multiplier terms in equation (11.33). The reason for this is that the multipliers have been absorbed into the λnew term in equation (11.32). In the original Taylor expansion that produces Newton’s method, λnew = λ + Δλ, where ∇λ is a step from the current Lagrange multiplier estimate λ. The ECICE algorithm constructs the linear system [equation (11.32)] by moving the summation ni = 1 λi Aij to the LHS and combining it with the Lagrange multiplier step Δλ. The second derivatives of the Lagrange function are obtained as follows: ∂2 f c ∂2 L = ; ∂ci ∂cj ∂ci ∂cj ∂2 L = − Aij ; ∂λi ∂cj ∂2 L = 0; ∂λi ∂λj

i, j = 1, …n i, j = 1, …n i, j = 1, …n

(11.35)

(11.36)

(11.37)

The search for the optimum solution that minimizes equation (11.28) is described in the following. More details with a numerical example of the procedures demonstrated with graphs are presented in a technical report published by the Science and Technology Branch of Environment Canada, authored by M. Shokr. It is available upon request from the author. The search starts with an initial solution that satisfies the equality constraint [equation (11.27)]. It then implements the first iteration by finding the solution that minimizes equation (11.30) with the equality constraint as the only active constraint (i.e., λ = 0). This is achieved by solving the linear system [equation (11.32)]. The solution is an increment that should be added to update the initial solution: k c k + 1 = c k + Δc k λ k + 1 = λnew

(11.38)

The solution is then examined to find if it violates any of the inequality constraints. If it does not, then that will be the required solution. In most cases, however, computing the new solution of equation (11.38) will result in a violation of one or more of the inactive inequality constraints [equation (11.29)], causing the solution to become infeasible. To prevent this, a line search technique is applied so that equation (11.38) becomes: c k + 1 = c k + α k Δc k

(11.39)

482

SEA ICE

where, α(k) is the step length and Δc(k) defines the search direction. The step length is calculated for each ice type (or equivalently each inequality constraint) “i” as follows: ci αi

k

=

− Δci

k

1 if 0 ≤ ci

k+1

k+1

n1. The power of the method resides in using a number of the observations larger than the specified number of surfaces. As mentioned before, ECICE uses a large number of CRVs (> 1000) to retrieve an equally large number of possible solutions of the concentration vector for each footprint observation (pixel). The CRVs are selected using the input probability distribution function (pdf ) of the observation parameters. A random number generator is used for this task. The process of obtaining the CRVs for a certain parameter using its pdf is illustrated in Figure 11.16. The figure shows a pdf of a given observation O at the top panel and its cumulative distribution function F(O) at the bottom panel. Each random number

0.6

Probability

0.5 0.4 0.3 0.2 0.1 0 80–90 90–100 100–110 110–120 120–130 130–140 140–150

Cumulative probability

Radiometric values 1 Random number generated

0.8 0.6

Selected CRV

0.4 0.2 0 80

90

100

110

120

130

140

150

Radiometric values

Figure 11.16 Graphical illustration of the process to generate a set of CRVs to mimic the probability distribution of a given observation for a given ice type. An example of a coarse probability distribution is shown at the top panel, and the corresponding cumulative probability is shown by the broken line at the bottom (sketch by M. Shokr).

r is used as an index to determine the corresponding CRV as CRV = F −1(r). The use of the cumulative probability in generating the random number generation allows the product CRVs to mimic the given pdf of the observation. This means that the CRVs are generated to maintain the relative weight of the observed signal under all possible meteorological and oceanic conditions that impact the surface. This argument particularly applies to the YI, which is most sensitive to such meteorological conditions as shown later. When two or more of the used observations are correlated (e.g., backscatter from HH and VV polarizations), the same random number is used to generate the CRVs in order to maintain correlation in the generated CRVs. An example of the probability distributions of four parameters, Tb,36h and PR36 from AMSR-E 36.5 GHz channel as well as the backscatter coefficients σ 0hh and σ 0vv from QuikSCAT Ku-band, for OW, YI, FYI, and MYI surfaces in the Arctic are shown in Figure 11.17. Examination of the distributions from the given surface types is necessary prior to the implementation of ECICE. The more overlap of the distribution, the more uncertainty the concentration output will be. Selection of input parameters should be based on visual examination of the probability distribution functions to ensure reasonable separability of the parameters between the given surfaces. Input parameters should not be correlated. Since the optimal solution is obtained from numerous sets of CRVs applied to the same pixel, a large number of possible solutions (equal to the number of the randomly selected CRVs) is produced per pixel. The statistics of those possible solutions (their mode and median) should be examined to generate the final answer as well as an uncertainty measure (a confidence level of the final answer). The median is selected as being the final answer unless the mode of the concentration of at least one surface is 100%. In this case, the median has to be incremented to bring its value close to 100% concentration for the surface that has its highest mode at 100% concentration. Note that the implementation of the inequality constraints usually results in distribution of possible solutions with two modes at 0% and 100%. The increment is calculated as a fraction of the distance on the concentration scale between the median and the 100% concentration. The incremental fraction is simply the difference between the percentage frequencies at the 100% concentration of each surface. The final concentration outputs of ice types and OW are normalized such that the total concentration is 100%. This is achieved by dividing each concentration by the summation of the estimated concentrations of all surfaces nj = 1 cj . As mentioned above, the availability of many possible solutions of concentration vector allows for the

SEA ICE

.25 .20 .15 .10 .05

–.08 –.06 –.04

90 98 106 114 122 130 138 146 154 162 170 178 186 194 202 210 218 226 234 242 250 258 266

.00

Polarization ratio PR38

.26 OW .24 YI .22 FYI .20 MYI .18 .16 .14 .12 .10 .08 .06 .04 .02 .00 –32 –30–28 –26 –24 –22–20 –18 –16 –14 –12 –10–08 –06 –04 –02 00

Backscatter coeff. σ0hh (dB)

Probability of occurence

Brightness temperature Tb38h (K)

Probability of occurence

OW YI FYI MYI

.26 .28 .30 .32

Probability of occurence

.30

.26 .24 .22 .20 .18 .16 .14 .12 .10 .08 .06 .04 .02 .00 .18 .20 .22 .24

OW YI FYI MYI

.35

Probability of occurence

.40

–.02 –.00 .02 .04 .06 .08 .10 .12 .14 .16

484

.26 OW .24 YI .22 .20 FYI .18 MYI .16 .14 .12 .10 .08 .06 .04 .02 .00 –32 –30–28 –26 –24 –22–20 –18 –16 –14 –12 –10–08 –06 –04 –02 00

Backscatter coeff. σ0w (dB)

Figure 11.17 Probability distributions of four radiometric parameters for three ice surfaces (YI, FYI, and MYI) and OW in the Arctic. Brightness temperatures were obtained from AMSR-E and backscatter from QuikSCAT. Noise floor of backscatter is -36 dB (generated by M. Shokr).

determination of a confidence level (CL) associated with the final concentration answer. The CL is a measure based on the mean absolute deviation (MAD). The latter is a fraction (between 0 and 1), defined as the average of the absolute values of the difference between the final solution concentration and the concentration from each possible solution. CL = 1 −

MAD MADmax

(11.44)

where, MADmax is the maximum value of the MAD from all the trials using different CRVs. In the case of the total concentration vector, the average of the Euclidean distance between the final solution and the concentration produced by each trial is used. MADmax is 100% for individual ice type concentrations and 100%∗√2 for the total concentration. Higher CL values mean less variability from the results using different CRVs (i.e., more robustness of the ice concentration from the trials). The reverse is true. In a perfect situation when each trial produces the same concentration (i.e., MAD = 0), the confidence level reaches a maximum value of 1.

Most of the SIC algorithms use a filter to exclude OW pixels. They are usually based on gradient ratios from microwave observations. ECICE uses a filter similar to the one used in NT2 but with different thresholds. The filter identifies the pixel as OW if the following two inequalities are satisfied: GR36V19V > 0.05 and GR23V19V ≥ 0.03 (these are the frequencies used in AMSR-E channels but similar frequencies from the sensors can be used). An example of the ECICE output of ice type concentrations is presented in Figure 11.18 from using a combination of input from AMSR-E radiometer and QuikSCAT scatterometer. The pixel spacing of the input data of 4.45 × 4.45 km2, available from the BYU site mentioned above. Total SIC of 100% can be seen extending from the Greenland side into the center of the Arctic Ocean, with vast regions of OW extending from the Beaufort Sea to the Laptev Sea. At this time of year (17 October 2007), the ice cover is expanding and high concentrations of YI can be seen at the leading edge of the ice cover facing the Eastern Siberian and Chukchi Seas. High concentrations of thicker FYI trail the YI inside the expanding ice pack. The MYI is confined mainly to the North American

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 485 2007–10–17

Ice Concentration Total

Ice Concentration YI

65N

Chakchi Sea

2007–10–17 65N

70N

70N

75N

75N

80N

80N

85N

85N

Novaya Zemlva

4% 85N

12%

85N

80N

20%

80N

75N

28%

75N

70N

36%

70N

65N

44%

65N

FYI

52%

2007–10–17

Ice Concentration

Ice Concentration

2007–10–17

65N

60%

70N

68%

70N

75N

76%

75N

80N

84%

80N

85N

92%

85N

MYI

65N

100%

85N

85N

80N

80N

75N

75N

70N

70N

65N

65N

Figure 11.18 Ice concentrations output from ECICE using AMSR-E 36.5 GHz channel data (Tb,36h and PR36) and QuikSCAT backscatter (σ 0hh and σ 0vv) acquired on 17 October 2007. The four panel display total ice concentration, and partial concentration of YI, FYI, and MYI as shown by the labels. The color bar denotes the concentration percentage and the gray color denotes OW.

side of the Arctic Ocean with a stream extending toward Severnaya Zemlya and the Laptev Sea. The figure shows how the results from ECICE could assist in determining the area occupied by each ice type, especially the uncommon product of YI, which has an increasing extent as a result of the ongoing global warming.

Since ECICE is a generic algorithm in the sense of accepting any set of remote sensing input data, it can be used to assess the capability of different input data sets in discriminating ice types. Figure 11.19 shows results of MYI concentration on 14 January 2008 using AMSR-E only (Tb,36h and PR36) and combined with

486

SEA ICE (a)

(b)

(e)

4% 12%

Ellesmere Island

20%

80° lat.

28% 36% Greenland

44%

JJ

Nares Strait

75° lat.

52%

CIS Ice Chart Jan. 14 2008

H KK

Baffin Bay

60%

E

G DD W

DD

68%

DD

MM DD

76%

70° lat.

FF

84% MM

92% 100%

FF

HH

(c)

(d)

0.85 0.86 0.87 0.88 0.89 0.91 0.92 0.93 0.95 0.96 0.98 0.99 1.00

Figure 11.19 ECICE estimates of (a) MYI concentration from using AMSR-E observations only and (b) using AMSR-E combined with QuikSCAT data, (c) and (d) are the confidence level of the two estimates as shown in (a) and (b), respectively. The corresponding weekly CIS ice chart is shown in (e).

QuikSCAT (σ 0hh and σ 0vv ). Only when AMSR-E and scatterometer observations are combined, the MYI stream appears as verified in the weekly ice chart of CIS. AMSR-E observations alone do not capture the MYI in the Baffin Bay but succeed to partially identify it in the Lincoln Sea. The confidence level at which MYI concentration [equation (11.44)] is produced is shown in Figures 11.19c and 11.19d. The MYI has been identified at CL between 85% and 90%. This is rather low but can be attributed to the heterogeneous composition of ice within the observation cell. The uncertainty of any parameter estimation from using remote sensing should be attached to the generated estimate. This issue has been raised lately within the sea ice scientific community, given

the different results by different ice parameter retrieval algorithms. 11.2.2.5. Intercomparison of PM Algorithms Different SIC algorithms produce similar results of total sea ice concentration under stable, fully consolidated (i.e., 100% ice concentration) winter sea ice cover in the Arctic (less data are available from the Antarctic, hence this statement is not confirmed there). They differ when the concentration is less than 90% (i.e., in open drift and close pack ice regimes). They also differ significantly in dynamic ice regimes such as the MIZ within the polar regions and the ice cover within the north temperate zone (between 23 N and 66.5 N). The latter includes in ice-rich

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 487

seas in the Gulf of St. Lawrence in Canada and the Baltic Sea in Europe. This has stimulated studies on intercomparison of results from different algorithms and identification of the best algorithm for use in a given ice regime. In the absence of in-situ validation data, intercomparison has become the only tool that provides information on the relative suitability and accuracy of the algorithms. Andersen [2000] compared results from eight SIC algorithms using results from a radiative transfer model. The study found that the NT algorithm is least sensitive to geophysical noise over consolidated ice whereas the frequency mode of the BS algorithm was the most stable over OW. Belchansky and Douglas [2002] compared total SIC from NT and BS algorithms, both using SSM/I input data, against results from an algorithm that utilizes a segmentation approach of the measured backscatter from RADARSAT-1. Data were obtained in the regions of Kara Sea, Laptev Sea, and Barents Sea during October 1995 through October 1999. This is a period of NI formation and highly dynamic ice cover. The authors found that the mean differences between the PM and RADARSAT1 in SIC were around 5.5%, and the differences between NT and BS concentrations were no more than 3.1%, with higher estimates from the BS algorithm. Andersen et al. [2007] compared SIC from seven algorithms: BRI, Cal/ Val, BS, N90GHZ, NT, NT2, and TUD. They used data from the Arctic perennial ice cover during the cold season from 31 October to 31 December, 2003. The validation data included ship observations, the RGPS output and classified SAR imagery statistics. The authors showed that SIC from the seven algorithms return near 100% concentration in an area delimited by daily ensemble mean concentration > 90%. To repeat, usually all algorithms succeed in reproducing the 100% SIC under cold polar winter conditions when the sea ice cover matures to its stable thickness, e.g., FYI thickness > 80 cm. Once again, this statement applies with more confidence to Arctic sea ice more than the highly dynamic Antarctic ice. The agreement between algorithms in Arctic winter period does not apply to areas of 100% ice concentration if the ice cover features mainly YI types with thickness 70%. Kern et al. [2019] compared total SIC from 10 algorithms in the Arctic and Antarctic against reference wintertime data of ~100% concentration and global yearround ship-based visual observations of the sea ice cover. The SICCI-25km, SICCI-50km, and OSI-450 concentration products nearly match the 100% reference data, yet with very low biases of −0.4% to −1.0% (Arctic) and −0.3% to −1.1% (Antarctic). Products of a group that includes data using climate data record (CDR) appear to be mostly biased high in the Arctic with a range between +1.0% and +3.5%, while their biases in the Antarctic range from −0.2% to +0.9%. The NT and ASI product biases are different for the Arctic, +0.9% (NT) and −3.7 % (ASI), but similar for the Antarctic, −5.4% and −5.6%, respectively. These are small biases, which confirm that most SIC algorithms perform well under stable cold winter conditions in the polar regions. Estimation of SIC is more challenging in dynamic ice regimes, namely open drift (4/10 to 6/10 concentration) and close pack drift (7/10 to 8/10 concentration). These are found at the margins of the polar ice and at southern latitude waters in the northern hemisphere. Kern et al. [2019] also show that truncation of concentration at 100% can result in misleading statistics and favor data sets that systematically overestimate the concentration. Kern et al. [2020] compared the performance of same algorithms used in Kern et al. [2019], but under Arctic summer conditions. Gridded melt ponds derived from MODIS data were excluded. SIC product from MODIS was also used in the comparison. Unlike the algorithms’ products during the stable winter ice coverage, the differences in the concentration magnitudes reached 20%–25% for SICCI-25km, SICCI-50km, and OSI-450, NT and ASI, and increased to 30%–35% for NT2 products. The

study identified the mismatch between the observed surface properties and those represented by ice tie points as the most likely reason for differences between PM and MODIS SIC products in areas of high melt pond fraction. It concluded that all the 10 algorithms products are highly inaccurate during summer melt. Ye et al. [2019] compared results of MYI concentration from six algorithms against ice age product from NSIDC. The methods include NT, BS, and four versions of ECICE, combining QSCAT and SSM/I with different input sets. The versions are denoted as ECICE-QSCAT1 (input σ ohh ,σ ovv , GR37V19V,PR19), ECICE-QSCAT-2 (input σ ohh ,σ ovv , Tb,37H, Tb,37V), ECICE-SSMI-1 (input Tb, 19H, Tb,19V, Tb,37V, GR37V19V), and ECICE-SSMI-2 (input Tb,19V, Tb,37H, Tb,37V, GR37V19V). Figure 11.21 shows a RADARSAT-1 scene in the Laptev Sea, acquired on 2 March 2002 along with MYI concentration maps from the six algorithms. The wind was blowing northward with speed around 15 m/s, and the air temperature was between -20 C and -25 C. The northeast section in RADARSAT1 image is mostly covered with MYI with its high backscatter and visible floe structure, while the southeast section is covered by what appears to be FYI with rough or DI in some parts. Information from the SAR image agrees well with the sea ice age map, which shows FYI for areas A and C, and SYI in area B. Among all the SIC methods, ECICE-QSCAT-2 agrees best with the visual analysis of SAR image. From this method, MYI concentration map shows values of 80%–90% in area B, and below 30% in areas A and C. The methods that use radiometer data only (ECICE-SSMI, NT, and BT) produce non-uniformly distributed MYI with concentrations between 50%and 100% in area B, below 30% in area A, and around 60% in area C. All methods except ECICEQSCAT-1 show some concentrations on MYI in the southwest area, which is supposed to be covered by FYI (as shown in the output from ECICE-QSCAT-2). This is an indication of misidentification of FYI as MYI under high wind conditions. It is well known that SIC from PM data are less accurate in summer because of the presence of melt pond on ice surface or even the change of wetness of the snow cover. Kern et al. [2016] investigated the sensitivity of eight SIC PM algorithms to melt pond using data from summer of 2009 in the Arctic. The melt pond fraction was obtained from MODIS reflectance data. The study found that SIC from BRI and BS algorithms are linearly related to the MODIS melt pond fraction after June, while the N90GHz produces weaker relation and becomes more apparent later in the summer. The presence of melt ponds causes underestimation of SIC by 14%–26% from the BS algorithm. This reduces to 0% for melt pond fraction under 20%. The study presents first results on evaluation of selected SIC PM algorithms under summer conditions

SEA ICE ECICE-QSCAT-1

83 °N A

B

2 km

ECICE-QSCAT-2 100

2 km

100

80 82°N

B A

80 B

60 40

C

81°N

40

C 20

20

0 80°N

0 ECICE-SSMI-2

ECICE-SSMI-1 100

B 78°N

100

80

79°N

A C

2 km

80 B

60 40

C

B A C

115 °E

125°E

40

C

130°E

20

0

NSIDC-SIA

15 m/s 10 5 0 –5 –10 –15 –20 –25 –30

120 °E

NT

0 BT

100 4 B

3

A C

2 1

Sea Ice Age [years]

ECMWF-ERA5

110 °E

Air temperature [°C]

105 °E

60

A

20 100 °E

60

A

MYI concentration [%]

490

100

80 B

80 B

60

60

A

A

40 C

C

40

20

20

0

0

Figure 11.21 RADARSAT-1 image of a section in the Beaufort Sea acquired on 2 March 2002 (top left), and MYI concentrations from six SIC methods displayed in the two rightmost columns (see text to explain the labels). The ECMWF ERA5 reanalysis of 2 m air temperature with overlaid 10 m wind vectors is shown at the bottom left panel, followed by sea ice age from NSIDC (courtesy of Yufang Ye of SYSU, China).

though explanation of the different results from the algorithms is not addressed conclusively because the snow conditions were not defined. A modeling approach to simulate brightness temperature under wet snow and flooded ice surface conditions would provide more conclusive results. It is more logical to study the sensitivity of brightness temperature to those factors than stepping up to the sensitivity of the algorithm products. Kern et al. [2016] presented a table of typical ranges of changes of Tb due to change of snow wetness compiled from previous studies and based on data from selected summer months (but the wetness values were not shown). 11.2.2.6. Sources of Error and Sensitivity of Ice Concentration Algorithms There are two major sources of error involved in any SIC retrieval algorithm using PM data: the possible anomalous values of a given input observation and the assumptions involved in the method of calculations. A few assumptions are addressed within the presentations of the four methods in section 11.2.2. They include the selection of the tie points, the truncation of the

solution if it falls outside the feasible range of the concentration (0% to 100%), and the assumptions involved in the mathematical scheme. As mentioned earlier, the approach may involve one of the three schemes: solving a set of linear equations, using an optimization method, or searching a domain of modeled data. Anomalies of PM emission may be caused by one or more of the following factors: (1) weather conditions that trigger significant deviation of emissivity/or brightness temperature from specified tie point of a given surface (this is particularly pronounced in the case of YI types), (2) atmospheric emission that adds to the emitted radiation from the surface, and (3) system noise. Effects of atmospheric influences are discussed in section 7.9.1.2. In general, atmospheric effects are more pronounced over regions of OW and open drift zones (ice concentration 60%). Retrieved concentrations are almost always less accurate in these regions and in MIZs. This can be adequately improved using radiative transfer models. System noise is not a major factor, which is usually identified and accounted for prior to the calculations. More elaboration on the weather effects follows.

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 491

Weather may affect the ice surface but more so affects the snow cover. The effect is much more pronounced during the transition seasons in the polar region or during the entire freezing season in southern latitudes. Emissivity of sea ice increases sharply at the onset of freezing (Figure 9.32), and this sensitivity adds to the uncertainty in estimating SIC when the ice cover is thin during the early growth phase. What complicates the SIC estimation in this case is the almost instantaneous response of the ice cover to changes in meteorological conditions. Such change may trigger other changes in the composition of the radiating layer. The salinity of the ice surface decreases when air temperature increases because of the acceleration of brine drainage. This increases the emissivity as the emitted radiation becomes exposed to less absorption. Changes in snow properties and the engendered microwave emission lead to more serious anomalies in the observed signature. Snow metamorphism resulting from thaw/refreezing cycle causes decrease of emissivity. Surface glaze and layering within the snow cover decrease the Tb,h at low-frequencies, and hence increase the polarization difference [Mätzler, 1985]. Subsequently, the ice concentration will be underestimated [Comiso et al., 1997]. Under steep vertical temperature increase within the snow pack in the polar regions in winter (>25 C per meter) the sublimated vapor from the snow base condensates within a short distance above the base, forming a hoar layer with relatively large grains of 1–10 mm in diameter. This process is enhanced in the presence of dense ice layers within the snow because the layers impede the upward flux of the vapor. Snow grains scatter the energy and therefore reduce the emissivity. The use of Tb in an SIC algorithm in this case will lead to underestimation of the concentration. Moreover, the presence of snow hoar layer at any depth, and particularly at the snow base, tends to affect the gradient ratio (used in the NT algorithm) and consequently SIC estimates based on this ratio. The snow accumulation on NI will cause brine to be wicked up or the ice surface to be depressed below the sea level. In both cases the salinity of the snow will increase and its emissivity decrease. Same is expected when snow becomes wet because it absorbs more solar radiation. If the combined effects of emissivity decrease and ice/snow temperature increase, is a decrease in Tb, the ice concentration will be underestimated. Most of the anomalies observed in the results from ice concentration retrieval algorithms are caused by processes in the snow. A summary of the response of snow on sea ice to weather events is presented in Sturm and Massom [2010]. It emphasizes that cycles of temperature rise near or above freezing point followed by a drop to deep freezing produce ice lenses and layering within the snow. Same is expected when rain-on-ice freezes. Ice

layering comprises only 3% of the snow cover on east Antarctic sea ice in winter [Sturm and Massom, 2010]. However, these layers delay percolation of rain water and snow melt from further into the ice cover, causing more ice thickening and spreading out laterally. Garrity [1992] studied the effect of these layers on microwave emission and the calculated ice concentration. The author concluded that they have significant depolarization effect, which leads to underestimation of SIC. In their classification of snow on the ground, Colbeck et al. [1990] identified saturated snow to have 15% of liquid water. Any level above that turns the snow into what is defined as slush. Ice concentration retrieval algorithms underestimate the SIC in this case. As soon as the snow starts to fall on the ice surface, Tb fluctuates and will only stabilize when snow settles. This results in parallel fluctuations of the estimated ice concentration (Figure 11.20). This factor has never been considered in SIC algorithms. Another meteorological effect on the now cover is caused by the wind and known as wind slab. Strong wind above 10 m/s produces highly compacted snow layer (slab), which has fine grains (0.1–0.5 mm). This hard snow (which can only be cut with a saw) is used by the Inuit to construct igloos. The effect of this snow on the microwave emitted radiation is not known. The above discussions signify the point that weather effects on the snow are a major source of error in estimating SIC from microwave remote sensing data. An example of the swift and abrupt change of the polarization ratios PR19 and PR85 [equation (9.1)] and gradient ratio GR37V19V [equation (9.2)] from thin ice due to weather change is shown in Figure 11.22. Data points were derived from measurements by a ground-based radiometer system, which sampled emitted microwave radiation from simulated sea ice grown in an outdoor pool located on the premises of the National Research Council of Canada in Ottawa. The figures show samples obtained every 30 minutes during the period 22–31 December 2001, when ice grew from 3 mm to 90 mm. Ideally, all data should form a narrow cluster whose center point would be considered as tie point for this thickness range. Nonetheless, the figure reveals a wide range of scattered points with no apparent temporal trend that can be associated with thickness growth. For one thing, this confirms the impracticality of using a single tie point to represent the ice during this unstable growth stage. More importantly, it shows the sensitivity of this snow-covered ice to weather conditions, which are not depicted in this figure but shown in Figure 11.20. PR19 is more sensitive than PR85 to weather conditions (as manifested in the wider range of data points). The period during which the data in Figure 11.22 were sampled features one snow fall event (4 cm accumulation), slush and frozen slush at the ice surface, and air temperature variation between −1.5 C and

492

SEA ICE 0.14

GR37V19V

0.12 0.10 0.08

Gradient ratio

0.06 0.04 0.02 0.00 –0.02 –0.04 –0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26 Polarization ratio: PR(19) 0.14 GR37V19V

0.12 0.10 0.08

Gradient ratio

0.06 0.04 0.02 0.00 –0.02 –0.04 –0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.26 Polarization ratio: PR(85)

Figure 11.22 Temporal distribution of the polarization and gradient ratios during early growth of laboratorysimulated sea ice in an outdoor pool. Data points were sampled every 30 minutes using a ground-based radiometer between 22 to 31 December, 2001. Sea ice was growing up to 90 mm thick during this period. Note the wide scattering of the data points, especially from the 19 GHz data (top panel) (M Shokr, unpublished data).

−11.5 C. It is understood that no SIC algorithm can produce reasonable results under this kind of ice type and weather conditions. Tonboe, Andersen, Pedersen [2006] studied the sensitivity of NT, BS, and Near 90 GHz SIC algorithms to parameters of snow and ice commonly observed in the Greenland Sea. The study used a sea ice version of the Microwave Emission Model of Layered Snowpacks (MEMLS) model [Tonboe, Pedersen, Haas, 2010] to simulate Tb from snow-covered FYI, then the data were used in the SIC retrieval algorithms and the sensitivity to the snow/ice parameters were explored. The study concluded that the BS algorithm is the least sensitive to warm air spills in the Arctic Ocean. NT algorithm is not sensitive

to the grain size of the snow as much it is sensitive to its density; the estimated SIC decreases as the snow density increases. The sensitivity of ice concentration to perturbations in emissivity or brightness temperature (i.e., the deviation from their typical values) is usually examined by incrementing the brightness temperature by a step equivalent to the upper limit of the sensor’s noise (usually 1 K) and calculating the SIC from a given algorithm. Markus and Cavalieri [2000] used this approach to determine the sensitivity of the NT2 algorithm to variations in brightness temperature from the 3 SSM/I channels used in the algorithm (around 19, 37, and 90 GHZ). They conducted the calculations using three different ice concentration

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 493 (a)

(b)

100

90 80 Total ice concentration (%)

80 60 40 20 250

70 60

Tb,85h = 180°K

50

Tb,85h = 182°K Tb,85h = 218°K

40

Tb,85 h = 220 °K

30 20 10

245

Tb

240 ,85v

235 230

180

190

200

210

220

T b,85h

0 225

230

235

240

245

250

255

Tb,85h

Figure 11.23 Sensitivity of the calculated total ice concentration from ECICE with simulated input observations of Tb,85h and Tb,85v, selected at fixed intervals. Sensitivity of the output concentration to variations in Tb: (a) a 3D perspective and (b) cross sections of 2D display (M. Shokr, unpublished data).

values corresponding to low, medium, and high concentration of 32%, 51%, and 98%, respectively. Except for the 19 GHz horizontal observations (to which the ice concentration is insensitive), the authors found that a change in brightness temperature from any other frequency or polarization channel led to a change in the total ice concentration output between 3% and 7% counts. This was true for the three ranges of ice concentration. The same methodology was used to evaluate the sensitivity of the calculated ice concentration from ECICE (section 11.2.2.4) to Tb. Results from using observations from Tb,85h and Tb,85v to determine total ice concentration are presented in Figure 11.23. Probability distributions of the input parameters for ice and water were input to the algorithm. Tb,85h was varied between 180 K and 220 K with a step of 2 K, while Tb,85v was varied between 230 K and 250K with a step of 1K. The figure shows that total ice concentration increases linearly with Tb,85h but non-linearly with Tb,85v. Figure 11.23a shows that for any given value of Tb,85v, the ice concentration increases by an average of 1.5% per one degree increase in Tb,85h. The non-linear variation of the calculated total concentration with Tb,85v is demonstrated further in Figure 11.23b. At low values of Tb,85h, typical of the radiometrically cold OW, ice concentration generally increases at a rate of 0.67% per one degree increase in Tb,85v. At the higher end of Tb,85h (220 K in the figure), which is a typical value from surface layering and glaze within the snow cover [Comiso et al., 1997], an increase in ice concentration with increasing Tb,85v is observed up to approximately 245 K, followed by a decrease of 5% concentration per one degree increase in Tb,85v.

11.2.2.7. Assessment of Ice Concentration Results Against Ice Charts In order to evaluate the ice concentration accuracy from any algorithm, a set of “validation” data must be used. The term validation is used here with reservation as explained shortly. The most common sources of such data include shipboard observations, field measurements, airborne camera survey, operational ice charts and, to a lesser extent, ice motion maps. Shipboard observations can be limited by the perimeter of validity of the information around the ship due to the grazing angle of the view. Field measurements are the closest dataset to the truth. Hence, they can indeed be considered validation data. Unfortunately, they are unavailable in the polar region during polar night period and they are limited in the spatial coverage compared to the large footprint from PM radiometers, from which results need to be assessed. Airborne observations cover wider areas, but they are expensive and the quantitative retrieval of the data is subject to sources of error due to the instability of the flight conditions. Operational ice charts from national ice monitoring centers (Table 11.1) are usually used as validation data though they cannot strictly be considered as truth data because the information is generated subjectively and they are made deliberately conservative, i.e., tend to exaggerate thicker and older ice percentage. Both temporal and spatial resolution of the ice charts are much coarser than the footprint of the radiometer or scatterometer system from which the SIC is generated. Ice motion algorithms that identify divergence and convergence of the ice cover have also been used tentatively as a source of validation, yet with some limitations [Kwok, 2002]. The conclusion is

494

SEA ICE

that except for an extensive field measurements or airborne visual observations, which are considered to be the closest to the truth data, no other dataset can be considered truth data per se, which are needed for a validation task. Given the inherent errors and limitations in any set of the above-mentioned data set (except for in-situ measurements), a best recommendation to assess the accuracy of the retrieval SIC algorithm would be to compare results to concentration data from as many sources as possible and also conduct intercomparison between results from different algorithms. One would argue that is not strictly a validation approach. This is correct in the absence of truth data, but it would facilitate the validation at least in a gross sense. For this reason, one should avoid using the term “validation” and use “accuracy assessment” instead. In the rest of this section, a method for assessment of PM-based SIC product is used. It involves comparison of calculated SIC against the similar information in ice charts. In fact, using ice charts for this assessment is common in research studies. The method entails overlaying contours of footprints from a coarse-resolution sensor, presented in colors that represent the calculated SIC, onto ice charts, also colored to represent the estimated ice types and their concentration. That way, the subjective estimates in the ice charts can also be verified using the calculated estimates. Further details follow. The basic analysis unit in any operational ice chart is called ice polygon. This is delineated subjectively by an

to scan

latitude

(b)

Δy

normal

(a)

ice analyst in the remote sensing image to be analyzed. A code for ice types and their partial concentrations (among other information) within the polygon is assigned. The spatial distributions of the assigned ice types and their concentrations are assumed to be uniform across the polygon. This assumption can be assessed by using the SIC from the smaller satellite footprints when overlaid on the polygon. Therefore, in addition to assessing the calculated SIC against the information in the polygon, the former can also be used to improve sub-polygon information in the ice charts. The calculated SIC can reveal, for example, the heterogeneity of information, assessed by ice analyst, within the polygon. This point is illustrated in the following presentation. Overlaying footprints of a coarse-resolution sensor (e.g., a PM sensor) on an image from a fine-resolution sensor (e.g., an SAR sensor) is presented in Shokr and Moucha [1998] and shown in Figure 11.24a. The data from the two sources must be nearly coincident. The coarseresolution footprint is approximated as a set of latitude–longitude coordinates of 12 vertices of a dodecagon shape (Figure 11.24b). The vertices are calculated using the latitude–longitude coordinates of the center of each footprint, along with the dimensions of its major and minor axes. These data are provided in the digital file of the coarse-resolution remote sensing data. A simple spherical earth surface model is used to determine the orientation of the footprint [Shokr, 2004]. The spatial accuracy of the footprint location depends on the geolocation

r

ne

scan li α

θ Δx

a

longitude arc (constant latitude)

b SSM/I grid points from the coarse-resolution data Polygon points approximation of the elliptic FOV

Figure 11.24 (a) Constructed elliptical footprints of SSM/I overlaid on a background of fine-resolution RADARSAT1 image. The subpixel contents of the SSM/I footprint are visible in the fine-resolution image. (b) An SSM/I footprint is approximated by a dodecagon whose vertices are defined by latitude–longitude coordinates (image composed by M. Shokr).

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 495

accuracy of pixels in the imagery dataset. The accuracy is typically 2–5 pixel spacing for a fine-resolution sensor such as RADARSAT, and one pixel from a coarseresolution sensor such as SSM/I. Figure 11.25 shows how the method can be used to compare total SIC from ECICE and NT2 algorithms

(using SSM/I data) against SIC estimated by ice analysts from CIS using the RADARSAT-1 image shown in the top left panel. The scene is of the southern part of the Gulf of St. Lawrence, Canada, acquired on 18 March 2003. The RADARSAT-1 image was nearly coincident with SSM/I observations from which SIC was calculated. Ice

(a)

(b) Gulf of St. Lawrence

d

lan

und wfo e N

Nova Scotia

Total ice concentration from ECICE-85

Atlantic Ocean

(c)

4.00 8.00 12.00 16.00 20.00 24.00 28.00 32.00 36.00 40.00 44.00 48.00 52.00 56.00 60.00 64.00 68.00 72.00 76.00 80.00 84.00 88.00 92.00 96.00 100.00

(d)

Total ice concentration from NT2

4.00 8.00 12.00 16.00 20.00 24.00 28.00 32.00 36.00 40.00 44.00 48.00 52.00 56.00 60.00 64.00 68.00 72.00 76.00 80.00 84.00 88.00 92.00 96.00 100.00

Total ice concentration from ECICE-37

Figure 11.25 RADARSAT-1 images acquired on 18 March 2003 with (a) delineated CIS polygon boundaries. Total ice concentration presented in the form of colored footprint are shown in (b) using ECICE with input from SSM/I 85 GHz channel, (c) using NT2 algorithm with input from SSM/I 37 GHz channel, (d) using ECICE with input from 37 GHz channel. The footprints are presented as ellipses with their contours coded in colors to represent calculated total ice concentration from the relevant algorithm. Colors of polygons represent total ice concentration from CIS (M. Shokr, unpublished data).

4.00 8.00 12.00 16.00 20.00 24.00 28.00 32.00 36.00 40.00 44.00 48.00 52.00 56.00 60.00 64.00 68.00 72.00 76.00 80.00 84.00 88.00 92.00 96.00 100.00

496

SEA ICE

analysts delineate polygons in the RADARSAT-1 image, called image analysis polygons (IAPs), and assign ice types and concentrations to each polygon. The polygons are shown in yellow contours in Figure 11.25a but without the codes for the ice types and concentrations. However, they are shown in color in the rest of the panels. The colors reflect the subjectively estimated total SIC by ice analysts. According to the shown color bar, most polygons show 100% concentration, but a few polygons at the bottom right corner of the scene had concentrations between 20% and 70%. The elliptical footprints of the SSM/I 85 GHz channel are displayed in panel (b) while the footprints of SSM/I 37 GHz channel are displayed in panels (c) and (d). In panel (c), the color of the footprints represents the calculated ice concentrations from NT2. In panels (b) and (d) the colors represent the calculated ice concentration from ECICE using observations from the 85 GHz and 37 GHz channels, respectively. The input observations to ECICE are the horizontal and vertical brightness temperature as well as the polarization ratio from the relevant channel (Tb,37h, Tb,37v, and PR37). If the estimated concentration by the ice analyst from CIS and the calculated concentration from the algorithm are identical, then the footprints will have the same color as the background polygon and therefore become invisible. Otherwise, the difference between the colors of the footprint and the background polygon reflects the discrepancy between the subjective estimate by the CIS analyst (or the ice chart if used to overlay the footprints). Figure 11.25 shows that NT2 underestimates the total ice concentration assigned by CIS, while ECICE correctly reproduces it across most of the scene (keep in mind the different input and calculation methods between the two algorithms). Discrepancies between the two algorithms are more visible near the ice edge (bottom right quarter). OW is well identified by both algorithms, but ECICE concentrations from using the 85 GHz observation provide better definition of the transition between OW and sea ice at the ice edge. This is the advantage of using the finer footprint of the 85 GHz observations. Note that the footprints that represent the ECICE output in Figure 11.25 are obtained from swath data (i.e., not gridded). This preserves the integrity of the observations. Gridded data are resampled and therefore offer an average observation from overlapping footprints at each grid point. More overlap is encountered at the far end of the swath as demonstrated in Figure 7.17. The above-described technique can be used to assess the SIC output from the algorithm against information in operational image analysis product or ice charts. However, the technique can also be used to adjust the subjectively estimated ice types and their partial concentration

in image analysis or ice charts to match SIC calculations from the algorithms. This is illustrated in Figure 11.26. A single IAP is shown in the top left panel, overlaid on a RADARSAT-1 image. The egg code that includes information on the ice composition in the polygon [MANICE, 2005] along with the acquisition times of RADARSAT-1 and SSM/I are shown in the top right panel. The rest of the panels show four ice concentrations output from ECICE: total ice, NI, YI (including gray and GWI), and FYI. In this example observations from the 85 GHz channels (Tb85h, Tb85v, and PR85) were input to ECICE. The entire polygon is assigned total ice concentration of 9+ tenth divided among the three ice types shown by the egg code: one-tenth of FYI thick (code 4.), four-tenths of FYI medium (code 1), and five-tenths of FYI thin (code 7). The calculated total ice concentration agrees with the CIS estimates. The algorithm outputs zero NI, which again agrees with the CIS chart. However, while CIS estimate calls for three types of FYI in the polygon with no YI, ECICE shows appreciable value of YI concentration. The remarkable observation in the figure is the sharp split between the calculated YI and FYI concentrations across the dark line in the top left panel. This suggests that the shown polygon could have been further divided into two parts separated by the dark line in order for each part to have a more-or-less uniform spatial distribution of ice types and concentrations. This is the kind of adjustment that this method offers to support the operational analysis of SAR imagery in addition to its use in assessing the SIC product from the retrieval algorithms. 11.2.3. Ice Concentration from Fine-Resolution SAR While estimate of SIC has been well established from PM data, a need for its retrieval from the fine-resolution SAR data has been identified and justified based on two requirements. The first is the increasing demand for data in MIZs with their highly variable ice type cover at small spatial scales. In general, impacts of climate change on Arctic marine environment are becoming more pronounced, and the interaction between sea ice and ocean dynamics has been taking place at finer spatial scales. The second reason is the emerging importance of ice classification and concentration in narrow water ways along the potential future ship routes in the Arctic (i.e., the Northwest and the Northeast passages). These passages are being monitored daily using the coarse-resolution PM and scatterometer data though they do not offer the required fine resolution to monitor narrow water passages. Unlike SIC from PM sensors where ice type(s) (or OW) concentration are assigned for the entire footprint of the sensor, in SAR data each pixel is actually a footprint, therefore one ice type or OW is assigned to the pixel.

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 497

March 18 2003 9+ 1 4 5 4. 1. 7 5 5 4

RADARSAT: 21:50 GMT SSM/I: 21:45 GMT

4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72 76 80 84 88 92 96 100

Total ice concentration

YI concentration

NI concentration

FYI concentration

Figure 11.26 RADARSAT-1 image with a single ice analysis polygon from CIS image analysis (top left). Results of total ice concentration and concentrations of three ice types from (NI, YI, and FYI) from ECICE using the input satellite observations combination mentioned in the text. Note the non-uniform distributions of YI and FYI concentrations within the polygon. The dark line in the top left panel marks a possible split of the shown ice polygon into two parts [M. Shokr, unpublished data].

Hence ice concentration from SAR is achieved by counting the ice or water pixels within a given area. SAR ice concentration approaches have been traditionally based on using thresholds that separate ice from water surfaces in a backscatter/or texture parameter space. One of the main difficulty with this approach is the overlap of backscatter signatures between ice and OW (section 9.3.1). Low backscatter from SI surface overlaps with the backscatter from calm OW and high backscatter from ice surface overlaps with wind-roughened ocean surface. These overlapping signatures are more likely to observe in the case of co-polarization than cross-polarization SAR from any operational frequency [Eriksson et al., 2010].

According to Geldsetzer and Yackle [2009] crosspolarization backscatter from OW is almost independent of the wind-generated water surface roughness. Yet, Voronovich and Zavorotny [2011] argued otherwise and suggested that the combination of co- and cross-polarized data could potentially lead to improved algorithms for the retrieval of ice concentration. Karvonen [2012] developed an algorithm that uses the autocorrelation as a statistical texture measure to avoid the dependence of OW backscatter signature on wind speed. The algorithm was applied to RADARSAT-2 σ ohh from the ScanSAR mode. Although it has not been widely applied or verified, it carries potential because

498

SEA ICE

the autocorrelation is also insensitive to the incidence angle (the RADARSAT ScanSAR mode incidence angle spans a wide range from 20 to 50 ). The closest SARbased ice concentration algorithm to operational application is presented in Karvonen [2017]. It was developed to estimate ice concentration in the Baltic Sea, which is monitored by the Finnish Meteorological Institute (FMI), with a potential of application over the Arctic. The method used the Sentinel-1 SAR dual-polarized (σ ohh and σ ohv ) observations in EW mode (400 km swath) separately and then jointly with the AMSR2 data for the estimation of SIC. The major sea ice areas of the Baltic Sea would be covered by Sentinel-1 once in 1–2 days. Results of SIC from the combination of SAR and AMSR2 are compared to results from using each sensor alone. Karvonen [2017] uses also SAR texture features to improve the SIC estimates. The method includes the following steps: (1) preprocessing for image calibration, noise floor equalization, georectification and correction for incidence angle variation, (2) segmentation using the iterated conditional modes algorithm applied to the principal component image of the two SAR channels, and (3) estimation of the SIC for each segment using multi-layer perceptron neural network. The digitized FMI SIC charts have been used as training and reference data for evaluation of the algorithm. Evaluation of the method has shown that estimation from SAR data alone produces higher errors than AMSR2 alone or AMSR2 combined with SAR. Much of the bias is manifested in overestimation of SIC in OW or low SIC areas. However, SARbased method provides more detailed information on SIC than the ice charts. It is possible to extend the application of this method to the Arctic-wide ice cover when more SAR data become available from new SAR constellations (e.g., the Canadian RCM), but training data based on SAR (not radiometer sensors) should be used. That is because the sensitivity of SAR and radiometer observations to the wind-roughened sea water surface is different. The DL approach has gained in popularity for SAR SIC as it has for ice classification. It is reopening the prospect of automatic sea ice mapping (classification and concentration). Several studies have explored this approach for SIC using training data from ice charts. However, none has made its way to operational environment so far. Wang et al. [2016] used the CNN approach to estimate ice concentration from dual-pol RADARSAT-2 SAR data during the melt season. The absolute mean errors of the generated ice concentration maps are less than 10% on average when compared with ice charts data. Nevertheless, the validation of the charts’ product during the melt season has not been fully established. More details are demonstrated in the CNN product than shown in the ice charts. The CNN approach requires adjusting a large number of hyperparameters (e.g., patch size,

number of layers, number and size of filters). The tuning of these parameters is essential to ensure best performance. So far, it is not known if there is generic tuning that fits the entire melt season. Wang, Scott, Clausi [2017] tested the same algorithm but for ice concentration estimates during freeze-up in the Gulf of St. Lawrence, east coast of Canada. This is the most significant operational ice region in Canada, which features highly dynamic sea ice with variety of YI and FYI. The authors compared ice concentration estimates from the CNN to those from another neural network, namely the multi-layer perceptron (MLP) that uses a single layer of hidden nodes. They found CNN to be less sensitive to subpixel details than the MLP and produced less noisy ice concentration in closer agreement with that from image analysis charts. The multi-layer structure of the CNN enables image features to be captured more accurately. More applications of this approach are needed to identify the optimum CNN parameters. Ice charts in confined operational sea water areas such as the Gulf of St. Lawrence and Baltic Sea are well established during the freezing season, which implies that using the data as input to a DL approach should lead to promising results. de Gélis, Colin, Longépé [2021] explored the potential of a fully convolutional network (FCN) for the automatic estimation of SIC using a long record of SAR from Sentinel-1. The study intended to mimic the work of ice analysts, which takes into consideration contextual information within the image and ancillary meteorological and climatological data. This was supported by input data down-sampled at 200 m and an FCN architecture duly parameterized. A comprehensive database was generated with 1320 dual-polarized Sentinel-1 scenes collocated with ice charts produced by the Norwegian Meteorological Institute. The FCN model was shown to be evenly robust to sea ice seasonal variability and incidence angle. The intention is to produce sea ice during the entire freezing or melting periods. Wang and Li [2021] introduced an approach for discriminating sea ice from OW using Sentinel-1 dual-polarization SAR in EW mode. It is based on a modified U-Net architecture, a DL network data, and applied to 28,000 images acquired on 2019. Comparison of results to ice concentration obtained using AMSR2 showed an absolute difference of 5.55%, yet with superior details revealed in the SAR product, particularly in MIZ. Data are available at: http://www.dx.doi.org/10.11922/ sciencedb.00273. 11.3. SEA ICE EXTENT AND AREA Sea ice extent is defined as the area that has an ice concentration exceeding a certain threshold. The threshold set by the NSIDC is 15%. Therefore, ice extent calculated from any ice concentration retrieval algorithm is simply

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 499

the summation of the area of pixels that have their ice concentration >15% (whether the pixels are connected or not). Here, each pixel with concentration >15% is assumed to be fully covered with ice. Therefore, ice extent implies larger ice area than the actual ice area. Alternatively, ice area is defined as the multiplication of ice concentration percentage at each pixel and the pixel area, summed over all pixels. Therefore, ice area is a more

accurate representation of the ice cover because it takes into account the actual concentration at each pixel. However, the use of ice extent is preferred in order to compensate for the underestimation of ice concentration in the summer using PM observations when the ice surface is melting. Figure 11.27 shows ice area and ice extent produced by different algorithms. Two sets are presented; one from using the original tie points provided with the

14

14

UMass-AES NORSEX 13

13

NORSEX

12

Bristol

11

TUD 10

Ice extent (×106 km2)

Ice area (×106 km2)

UMass-AES

NASA team

9

12 Bristol & TUD 11 NASA team 10

9

Using original tie points

Using original tie points

8

8 1995

2000

2005

2010

1995

14

14

13

13

2000

2005

2010

Ice extent (×106 km2)

Ice area (×106 km2)

UMass-AES & TUD

12 UMass-AES & NORSEX Bristol & TUD

11

10

9

NASA team

Bristol & NASA team

12

11 NORSEX 10

9 Using adapted tie points

Using adapted tie points 8

8 1995

2000

2005

2010

1995

2000

2005

2010

Figure 11.27 Annual variations of sea ice area and extent in the Arctic, estimated from five ice concentration algorithms: NORSEX, NT, UMass-AES, Bristol, and TUD. Two sets are presented: one from using the algorithms with the original tie points provided with each algorithm (top row) and the second with new common tie points obtained within the framework of ESA CCI project (bottom row). The plots are modified versions from an original set (courtesy of L. Pedersen and N. Ivanova).

500

SEA ICE

original version of each algorithm and the other from using a universal (more consistent) set obtained from the Round Robin Data Package (RRDP). This package became the backbone of the products of ESA Climate Change Initiative (CCI). Ice extent is used in monitoring the annual variation of the polar ice; an important indicator of global warming in the Arctic. Several centers produce daily maps of polar ice concentration and extent regularly. The graph in Figure 11.28 shows evolution of daily ice extent of Arctic sea ice, generated by NASA Arctic Earth Observatory. The Arctic ice reaches its maximum in March and its minimum in September. In 2021, the minimum ice extent of 4.72 × 106 km2 was observed on 16 September. It was higher than the lowest extent on record, which measured 3.41 × 106 km2 in September 2012. Sea ice extent in September 2020 and 2019 were the second and third lowest on record at 3.74 × 106 km2 and 4.14 × 106 km2, respectively. The decadal decrease of ice extent in winter is not as much as it is found in September when the most significant interannual changes are observed. The reduction in the Arctic ice extent allows for more seasonal ice cover. This has reversed the proportion of seasonal to perennial ice coverage in the Arctic, with seasonal sea ice now covering more than two-thirds (up from one-third four decades ago) of the Arctic Ocean in late winters (section 13.3.3).

An ice extent product from a given center is usually generated using a certain algorithm, with input from PM observations. How would different algorithms affect the results? This question was addressed in Ivanova et al. [2014] where the authors calculated daily ice extent and ice area in the Arctic during 2012, using 11 PM ice concentration algorithms. Results are shown in Figure 11.29. The common trend shown by all algorithms is ascertained regardless of the obvious variability. The difference in sea ice area varies between 0.1 and 1.4 million km2 with the maximum occurs in the end of June. On the other hand, the difference in ice extent varies between 0.1 and 0.7 million km2 with the maximum occurs in the end of May. Ice extent maps are standard products from MODIS based on VIS, NIR, and TIR data. The algorithm employs three criteria [Riggs, Hall, Ackerman, 1999]. The first is the reflectance in the VIS and NIR wavelength, where snow has a much higher reflectance than water. Snow-covered ice is identified if reflectance from band 2 ≥ 0.11 and in band 4 ≥ 0.10 (band 2 is NIR with wavelength 841–876 nm and band 4 is VIS with wavelength 545–565 nm). The second criterion is the normalized difference snow index (NDSI), defined as the difference between reflectance from the VIS channel 4 and the NIR band 6 (wavelength 1628–1652 nm) divided by their summation.

Arctic sea ice extent (millions of km2)

16

14

12

10

8

6 2021 2012 (record low) 1981–2010 average ± 2 standard deviations

4

2 Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Jan

Figure 11.28 Daily records of the Arctic ice extent. Average extent during the period 1981–2010 (dashed-line), year of minimum summer extend on record (orange line) and record of 2021 up to September (red line) are shown. The shaded blue areas represent the ±1 and ±2 standard deviation of the 1981–2010 average (courtesy of NASA Arctic Earth Observatory by Joshua Stevens using data from NSIDC).

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 501 Extent

Area 16

16

14

14

12

12

10

10

ASI

Million km2

Bootstrap Bristol CalVal NASA Teram NASA Teram2 8

8

Near90GHz

6

6

NORSEX

4

4

2 Jan Mar May Jul Sep Nov 2012

2 Jan Mar May Jul Sep Nov 2012

NORSEX85H TUD UMass-AES

Figure 11.29 Ice extent and ice area calculated from 11 different microwave ice concentration algorithms in the Arctic during 2012. Ice area is more sensitive to differences in concentration output from different algorithms. The minimum ice cover is always observed around mid-September (courtesy of NANSEN Environmental and Remote Sensing Center, Norway).

NDSI =

Band 4 − Band 6 Band 4 + Band 6

(11.45)

In winter, snow-covered ice is identified if NDSI ≥ 0.4. The NDSI is used to ensure the detection of the snowcovered ice because it has high values due to the high and low reflectance in the VIS and NIR; respectively. The third criterion is the IST calculated from MODIS TIR (bands 31 and 32 with wavelength around 11 and 12 μm, respectively) using a version of the split-window equation [e.g., equation (7.45)] with an appropriate value of surface emissivity). A pixel is identified as being covered with sea ice if the calculated surface temperature is ≤ 271.4 K. 11.4. SEA ICE THICKNESS (SIT) Estimation of SIT is important for both marine operations and climate studies, yet in different aspects. Marine operators are more concerned with thick ice, which constitutes navigational hazardous. Therefore, ice monitoring services usually make special effort to produce accurate information about thick seasonal ice and perennial ice. If in doubt, the analyst would assign thicker or older ice type in order to make the estimate conservative. Thin ice types (i.e., YI < 30 cm thick) do not represent obstacles to most ship classes; therefore, accurate estimate of thin ice thickness is not as crucial for this purpose as estimates for thick ice. On the other hand, climate and weather models require detailed information on thin ice because it has major impact on ocean–atmosphere interactions. The sensible heat flux in thin ice is one or two orders of magnitude larger than that in thick ice [Maykut, 1978] and is proportional to the ice thickness. This

highlights the importance of accurate estimates of thin ice thickness for the calculation of thermal balance between ice and atmosphere in winter and energy balance at the ice surface throughout the year. Unfortunately, thermal and radiative properties of thin ice change significantly in response to meteorological forcing. Due to lack of timely information about ice thickness distribution in the Arctic, seasonal climatic data are usually used as input in most climate models. The spatial scale at which ice thickness information is required varies according to the application. Monitoring ice thickness at a tactical scale (a few hundreds of meters or a few kilometers) is important for the safety of marine navigation, while monitoring ice thickness at a synoptic scale (tens of kilometers) would be sufficient to improve the accuracy of the global and regional weather and climate models. Ice thickness is also needed to compute the total ice volume and its mass balance in the polar regions as well as the mass flux of ice advected out of the Arctic Basin. Estimation of ice thickness from remote sensing observations is a challenging task because sensors can “sense” the emitted or scattered radiation only from the top few millimeters or centimeters of the ice cover, depending on the wavelength. Therefore, observations do not usually carry information about the entire ice thickness. Advances in remote sensing tools and analysis techniques have allowed better estimates of ice thickness indirectly using one of the following three approaches: (1) heat and energy balance equations, which are used with TIR observations; (2) empirical equations that relate ice thickness and radiometric observations, which are used with microwave observations; and (3) the buoyancy law that allows inference of ice thickness from its freeboard

502

SEA ICE

measurements. The latter can be achieved using altimeter data. The first two approaches are used to estimate thickness of thin ice, while the third is more useful in estimating thickness of thick ice. The three approaches are presented in the rest of this section, but it would be useful to begin with a summary of the advantages and limitations of each approach. 1. Heat and energy balance equations can be used to estimate ice thickness from TIR observations, yet with a few limitations. The method allows thickness retrieval up to 50 cm thick and only under clear-sky condition. Cloud filtering is required as a preprocessing step but it is difficult to apply during the dark polar season. Atmospheric correction must also be applied as a preprocessing step (see atmospheric windows in the TIR region in Figure 7.29). Retrieval of ice thickness can only be achieved when the ice surface and the overlaid snow are dry. A filter to exclude areas of thick ice (> 50 cm think) must also be incorporated in the algorithm. In general, ice thickness retrieval using TIR is a lengthy process and requires input of several meteorological as well as ice and snow physical parameters. Since this is a physicsbased approach it should produce more accurate results if the assumptions for its application are fulfilled and the input parameters are accurate. 2. The empirical approach implies developing regression equation(s) relating ice thickness to radiometric/scattering observations or derived parameters. It is based on the premise that ice thickness can be related to the microwave brightness temperature or radar backscatter during the early ice growth phase. This is driven by the intervening process of brine drainage (remarkable decrease of sea ice salinity) during this phase. That is why this approach can be used to estimate ice thickness up to 15–20 cm using observations from any frequency of operational microwave radiometers (18 to 89 GHz) (Figure 9.32) or up to 50–100 cm from using the L-band radiometer (1.4 GHz) onboard Soil Moisture and Ocean Salinity (SMOS) satellite. The above-mentioned premise may become invalid during ice decay as the dominant processes do not lend themselves to a systematic thicknessdependent relation. Such processes include surface flooding, snow wetness, and snow metamorphism. However, Iwamoto, Ohshima, Tamura [2014] developed an empirical relation that links thin ice thickness to polarization ratios from PM data and confirmed its applicability to most Arctic polynyas from September to May between 2002 and 2011. Recall that thin ice grows and decays continuously during the lifetime of the polynya. To develop the regression equation, ice surface must be snow-free. Atmospheric correction must be applied if observations are obtained from a high-frequency PM channel (e.g., 85 or 89 GHz). Although this approach is simple, it may not produce robust enough data because regression

equations depend on regional and seasonal factors as well as the sensor’s characteristics of the regressed observations. 3. The buoyancy law is used with altimeter data (radar or laser) to convert the measured distance from the satellite orbit to the surface into ice (or ice plus snow) thickness. Using buoyancy law avoids the limitation of the other sensors caused by the limited penetration depth of the signal. However, the application of the buoyancy law requires accurate measurement of ice freeboard. Inaccurate estimate of freeboard (including snow depth in the case of laser altimeter) could lead to error in the ice thickness estimates (section 7.8). Clouds and atmospheric constituents should also be accounted for in the case of laser altimeter. Ice thickness maps using altimeter data are generated at poor temporal and spatial resolutions (a few tens of kilometers). However, thickness maps covering the entire polar region can be produced from composite passes acquired over a few months. The maximum thickness that can be estimated using altimeter data is around 5 m, though data are unreliable for small thickness (say, < 30 cm). The limitations of the above three approaches suggest that no single approach can cover the entire ice thickness range. Requirements of the operational ice monitoring entail mapping ice >30 cm thick at a spatial scale of a few hundred of meters with daily temporal resolution in highly marine traffic areas. On the other hand, climate modeling and monitoring requires mapping ice thickness starting from a few centimeters thick with no upper limit, at a spatial resolution of a few kilometers or tens of kilometers and temporal resolution of a few days or one week. This product requires an integrated approach of remote sensing data with modeling meteorological data. The material in this section is strictly relevant to retrieval of ice thickness using space-borne observations. It is worth noting, however, that other remote sensing devices can be used to measure ice thickness. The spatial profiles of ice draft have been measured using submarine sonar systems (ice sail height is 10%–15% of ice draft). This approach was used in a few studies [Rothrock, Yu, Maykut, 1999, Wadhams and Davis, 2000, Winsor, 2001] to determine the interannual variability of Arctic ice thickness using data from submarine cruises conducted between 1991 and 1997. Helicopter-borne EM sensor has also been used to measure SIT (Hass et al. [2009]). Sea ice has a very low electrical conductivity (0–50 mS/m) compared to seawater (2400–2700 mS/m), which is a very good conductor. If a low-frequency EM generated by a transmitting coil strikes a sea ice surface, it should penetrate the ice volume almost unaffected. When it reaches the underside of the sea ice, it generates eddy currents in the seawater. These currents induce a secondary EM field, which returns through the sea ice cover and can

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 503

be measured by the receiving coil. The strength of the secondary EM field is related to the distance from the receiving coil to the ice–water interface (the conductive surface). If the height of the transmitting coil above the snow–ice surface is known, then the ice thickness (including the depth of possible snow cover) can be determined. That height can be obtained using a laser altimeter installed on the same helicopter platform. Modeling SIT at synoptic scale with assimilation of satellite ice thickness data has also been presented in a few studies. Zhang and Rothrock [2003] present the thickness and enthalpy distribution (TED) sea ice model that determines ice thickness and explicitly simulates sea ice ridging. This is a dynamic-thermodynamic model that employs a three-layer thermodynamics and a viscous–plastic rheology for ice dynamics. The model estimated FYI thickness in the Arctic to be 1–3 m thick and in the Southern Ocean to be 1–2 m thick. One of the interesting simulation results is that the thickest ridged ice is 20 m in the Arctic and 16 m in the Southern Ocean. Miller, Laxon, Feltham [2005] examined the sea ice model of Los Alamos to compare its output thickness/draft in the Arctic Ocean with submarine observations during cruises between 1987 and 1997. They found that the model overestimated the thickness in the Beaufort Sea and underestimated it near the North Pole. The authors pointed out the need for a better representation of sea ice rheology on the continuum scale.

11.4.1. SIT from TIR Observations In this approach, the term ice thickness refers to the thermodynamically grown ice as opposed to mechanically accumulated thickness (i.e., as a result of compression forces, which lead to DI). TIR observations were the first remote sensing data used to calculate ice thickness. Calculations are limited to ice of thickness < 50 cm. Within this range surface temperature can be related to ice thickness. This approach cannot be pursued under clouds or fog. The assumptions involved in this approach are: (1) the temperature profile in ice and snow is linear, (2) ice thickness is uniform across the sensor’s footprint, and (3) the observed ice is in a state of thermal equilibrium with a constant underlying sea water temperature. A one-dimensional surface energy balance model is used to derive the ice thickness distribution. Heat flux through the ice is balanced by the atmospheric heat and radiation fluxes. A simple linear approximation of this balance is described by the following equation [Yu and Rothrock, 1996] assuming a constant value of surface temperature (whether snow or ice); i.e. the surface is in a state of thermal equilibrium: dn F r − F up l − Fl + Fs + Fe + Fc = 0

(11.46)

where, Fr is the net solar radiation absorbed by the snow dn and ice media, F up l and F l are the upwelling and downwelling longwave radiation (the blackbody radiation of the ice surface and the atmosphere, respectively); Fs , Fe , and Fc are the sensible (turbulent), latent and conductive heat fluxes, respectively. In the night passes Fr should be neglected. It is also possible to assume Fs = Fe = 0 as a first approximation. This assumption is in line with the finding by Maykut [1978] that these two terms are much smaller than the incoming longwave radiation. The term Fr is the difference between the penetrated shortwave radiation minus the transmitted radiation flux I0 through the ice and snow: F r = 1− αs F sw − I 0

(11.47)

where, αs is the surface albedo and Fsw is the incoming shortwave radiation. The transmitted radiation through the ice volume is given by the equation: I 0 = i0 1− αs F sw

(11.48)

where, i0 is the transmittance (this is the term Tλ in equation (7.13)). Ice thickness is implied in the term Fc in equation (11.46). Hence, it can be estimated when all other terms are specified, namely after substituting each term with its appropriate parameterization. Details of the method are presented in Yu and Rothrock [1996] and used by several authors including Rudjord, Tier, Solberg [2011]. It is summarized in the following. To calculate Fr, three terms in equation (11.48) have to be determined: αs, i0, and Fsw. Data from any visible channel can be used to derive αs. The value of i0 can be determined as a function of ice and snow thickness [Grenfell, 1979]. However, it would be better to use constant (mean) values of αs and i0 within four intervals of ice thickness H, as given in Rudjord, Tier, Solberg [2011]. Table 11.2 includes these values. Fsw can be estimated using an expression presented in Bisht et al. [2005] based on a parameterization by Zillman [1972]. F sw = S 0 cos 2 θ d

(11.49)

Table 11.2 Values of ice surface albedo (αs) and transmittance (i0) for four ranges of ice thickness (H). H ≤ 5 cm 5 cm ≤ H < 20 cm 20 cm ≤ H < 40 cm H ≥ 40 cm

αs

i0

0.091571 0.663315 0.777930 0.799825

0.641808 0.604537 0.254103 0.094115

Source: Adapted from Rudjord et al. [2011].

504

SEA ICE

where, S0 is the solar constant at the TOA (=136 Wm−2), θ is the solar zenith angle, and d is a function of θ and the surface water vapor pressure ea: d = 1 085 cosθ + ea 2 7 + cosθ × 10 − 3 + 0 1 (11.50) Here ea = f esa, where, f is the relative humidity, which can be assumed to be 90% and esa is the saturation vapor pressure in hectopascals (hPa), which is estimated by Maykut [1982] as a fourth-order polynomial of the air temperature Ta: esa = 2 780 × 10 − 6 T 4a − 2 691 × 10 − 3 T 3a + 0 979 T 2a − 158 638 T a + 9653 192

(11.51)

The second to the sixth terms in the LHS of equation (11.46) can be determined as described below. The upward and downward longwave radiations are given by the blackbody radiation (section 7.2.3) from the surface and atmosphere, respectively: 4 F up l = εi σT s

(11.52)

4 F dn l = εa σT a

(11.53)

where, σ is the Stefan-Boltzmann constant (= 5.6704 × 10−8 W/m2.K4). According to Yu and Rothrock [1996], the ice emissivity εi can be taken as 0.97 and emissivity of the atmosphere εa = 0.7855(1 + 0.2232C2.75), where C is the cloud fraction. The IST Ts can be obtained from any TIR sensor such as MODIS. The air temperature Ta at 2 m level can be obtained through either one of the following two approaches. The first is an approximate estimate by adding 0.4 C to the estimated surface temperature from a TIR sensor. Lindsay and Rothrock [1995] adopted a similar approach where the bias was a monthly varying climatological air/surface temperature difference. The second entails using reanalysis of meteorological parameters from any weather model such as the Global Environmental Multiscale (GEM) model of the Canadian Meteorological. The sensible and latent heat fluxes, Fs and Fe, which can be neglected as a first approximation, are estimated from the equations: F s = ρa cp C s u T a− T s

(11.54)

F e = ρa LC e u f esa − es0

(11.55)

where, ρa is the air density, cp is the specific heat of air, L is the latent heat of vaporization, Cs and Ce are the bulk transfer coefficients for heat and evaporation. Yu and Rothrock [1996] chose Cs = Ce = 0.003 for very thin ice and = 0.00175 for thicker ice. The term es0 in equation (11.55) is the saturation vapor pressure at the surface. The surface wind speed u at 2 m level above the surface can be obtained from the output of any weather model.

Finally, the conductive heat flux is estimated assuming linear temperature gradients through the snow and ice cover: Fc =

ki ks T f − T s ki h + ks H

(11.56)

where, ki and ks are the thermal conductivity of sea ice and snow, respectively, h is the snow depth, H is the ice thickness, Tf is the freezing temperature of the seawater derived from a simplified relationship, Tf = − 0.055Sw, where Sw is the sea water salinity. The snow conductivity Ks can be set to 0.31, while the ice conductivity can be calculated from: ki = k0 + βS T s− 273 15

(11.57)

where, k0 is the conductivity of pure ice (= 2.034 W/m.K) and β = 0.13 W/m, S is sea ice salinity measured in parts per thousands, and Ts is, as usual, the surface temperature (more accurate results can be obtained if the bulk ice temperature is considered). The salinity is modeled by the empirical relationship developed by Cox and Weeks [1974]: S = 14 24 + 19 39H for H ≤ 40 cm or S = 7 88 + 1 59H for H > 40 cm

(11.58)

Snow depth is a crucial parameter especially over thin ice because it changes the thermal fluxes and surface temperature significantly. Unfortunately, it cannot be determined accurately from remote sensing (see section 11.8) and concurrently with the TIR sensor passes. Therefore, an approximate empirical relationship between snow depth h and ice thickness can be used [Yu and Rothrock, 1996]: h = 0 for H ≤ 5 cm h = 0 05h for 5cm ≤ H < 20 cm

(11.59)

h = 0.1h for H ≥ 20 cm Values from equations (11.47) –(11.59) can be substituted in equation (11.46), which can then be solved for the unknown ice thickness H (this parameter appears in equation (11.56). As mentioned before, the application of this approach requires a cloud masking module to flag out cloudy areas. Various algorithms have been developed to flag out clouds from calculations [e.g., Stowe, Davis, McClain, 1999] though Yu and Lindsay [2003] reported that no algorithm was developed to detect clouds during the dark seasons in the polar regions. The technique also requires application of a “thick ice filter” to flag out thick ice. Rudjord, Tier, Solberg [2011] used PM spectral profiles of OW and ice types to identify an appropriate parameter for such filter. They recommended

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 505

using the inequality (Tb,89V/Tb,19V > 1) to keep only pixels of thin ice. The above approach has been used in a few studies. Drucker, Martin, Moritz [2003] used it to estimate thin ice thickness in a coastal polynya adjacent to the St. Lawrence Island in the Bering Sea. They used gridded AVHRR data at 1-km resolution with collocated measured meteorological data (air temperature and wind speed). Figure 11.30 is a composite image of the polynya region constructed from a near-coincident RADARSAT1 image (the gray area) and AVHRR-derived thin ice thickness (colored areas). The outer boundary of the polynya can be identified in the ice thickness map. Most of the thickness values under 20 mm were located in an active frazil region within the polynya. A set of upwardlooking sonar sensors (ULS) were placed in the locations

marked by the circles in the figure to verify the ice thickness results. The thickness from these sensors was around 8 mm. The figure shows that ice thickness increases away from the coast (toward the pack ice). This contradicts the findings from the ULS measurements. Drucker, Martin, Moritz [2003] provide a plausible explanation in terms of Langmuir circulation which causes the distribution of frazil at depth near the coast to become strong while ocean wave damps the circulation away from the coast, causing less frazil distribution at depth. Therefore, the thickness in this case does not strictly result from thermodynamic growth but rather from frazil deposit at the ice bottom. More recent studies on thin ice thermal thickness estimation using the above-described TIR approach have been published recently. Zeng et al. [2016] estimated the thickness distribution in the Bohai Sea during winter of

175 km

0

0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14 Ice thickness (m)

Figure 11.30 AVHRR-derived thin ice thickness (color) overlaid on a RADARSAT-1 SAR image (B&W) of a coastal polynya in the Bering Sea west of Alaska. Images were acquired on 9 January 1999. Black areas within the ice thickness map are cloud masked. The circles and the squares show locations of ULS (see text) and salinity sensors, respectively [Canadian Space Agency].

506

SEA ICE

2009–2010. The retrieved thickness agreed with in-situ observations from two offshore oil platforms with overall bias 1.4 cm and root-mean-square error (RMSE) 3.9 cm. Adams et al. [2012] estimated thin ice thickness distribution within the Laptev Sea polynya for the two freezing seasons 2007/08 and 2008/09. The average uncertainty was ±4.7 cm for ice thicknesses below 0.2 m. Shi et al. [2016] estimated thickness of level ice in the Liaodong Bay (the northern section of the Bohai Sea) in winter 2012–2013 using MODIS TIR data and reanalysis weather data from ECMWF. The overall RMSE of the thin ice thickness was 5.6 cm. 11.4.2. SIT from PM Observations PM radiometry has been used to estimate thickness of thin sea ice, provided it is snow-free, despite the two limitations that arise from its coarse spatial resolution. First, thin ice occupies relatively small areas in the polar regions, mostly in polynyas, leads or at the margin of pack ice. Therefore, the application cannot be extensive. Second, it usually coexists with OW and other thicker ice types within the footprint of the PM sensor. This heterogeneity makes the thickness retrieval difficult or impossible. Nevertheless, PM data have been used to estimate ice thickness in polynya [Iwamoto, Ohshima, Tamura, 2014], MIZs, and newly formed fast ice [Tamura et al., 2007]. To reiterate, the premise behind using the PM observations to estimate thin ice thickness entails that ice thickness can be related to the microwave brightness temperature during the early ice growth phase. This is because of the rapid desalination of the subsurface layer during this phase (section 2.5.5). Hence, ice thickness up to 15–20 cm can be retrieved from PM data. Many studies have shown strong negative correlation between the polarization ratio and the thickness of thin ice in absence of snow cover [Martin et al., 2004, Tamura et al., 2007, Tamura and Ohshima, 2011, Singh et al., 2011, Iwamoto et al., 2013]. The polarization ratio (PRf) at frequency f is given by equation (9.1), but it can also take the following form, which is commonly used in ice thickness retrieval algorithms: PRf = Rf − 1

Rf + 1

(11.60)

where, Rf = T b,fv T b,fh

(11.61)

The retrieval algorithm uses a regression equation that relates thin ice thickness to PRf (some algorithms use Rf instead). The equation is generated from a scatter plot of PRf against estimated ice thickness, which is usually obtained from another satellite retrieval (e.g., TIR-based retrieval from AVHRR or MODIS satellites). An example of such a scatter plot presented in Tamura and

Ohshima [2011] and reproduced in Figure 11.31. The data were obtained from three different polynyas in the Arctic: the North Water polynya, the Chukchi Polynya, and the Laptev polynya. The ice thickness H (up to about 15 cm) is correlated to the polarization ratio. Beyond that limit the thickness becomes insensitive to the polarization ratio. Two least square curve fittings (linear and exponential) are given based on data from Figure 11.31: H = − 2 055 PR85 + 0 1765 for 0 081 ≥ PR85 ≥ 0 0494 (11.62) This estimates ice thickness between 0.01 m and 0.075 m. H = − 4 565 PR37 + 0 3492 for 0 06 ≥ PR37 ≥ 0 0436 (11.63) This estimates ice thickness between 0.075 m and 0.15 m Note the larger thickness range that can be estimated using PR37. Equation (11.62) is used first for the prescribed range of PR85. Values of PR85 ≥ 0.0494 correspond to an ice thickness less than 0.075 m. For the rest of the pixels, equation (11.63) is used, but if the calculated thickness is < 0.075 m it is set to be 0.075. This means that the algorithm has discontinuity at h = 0.075. The same equations, but with different coefficients suitable for thin ice in the Antarctic Ocean, were used in Tamura et al. [2007]. Alternatively, the ice thickness H (in meter) can be calculated using the equation for the exponential curve fitting of the data, which is given by Martin et al. [2004]: H = exp 1 αPR + β − γ

(11.64)

This equation is applied with two sets of coefficients: one for PR85 from SSM/I (or PR89 from AMSR-E) and the other for PR37 from SSM/I (or PR36 from AMSRE). The coefficients are shown in Table 11.3 using regression of data sets from the study of Tamura and Ohshima [2011] and another study by Iwamoto et al. [2013]. They are specific for each region and for each pair of TIR data from which thickness has been retrieved. A new set of coefficients must be established if equation (11.64) is to be used for a different region or different sensors. As mentioned earlier, observations from highfrequency channels (85 or 89 GHz) must be corrected to account for atmospheric conditions, namely water vapor column, fog, and cloud liquid water. A standard method has been used in previous studies based on the examination of scatter plots of PR36 and PR89 of pixels under clear- and cloudy-sky conditions. An example of such plot is presented in Figure 11.32, following Iwamoto et al. [2013]. The clear-sky data are shown in black and the cloudy-sky and water-vapor-rich atmospheric data are

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 507 0.30

AVHRR ice thickness (m)

PR85

PR37

0.25

0.25

0.20

0.20

0.15

0.15

0.10

0.10

0.05

0.05

0.00 0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.00 0.00

0.02

0.04

0.06

0.08

0.10

0.30

0.30

AVHRR ice thickness (m)

PR85

PR37

0.25

0.25

0.20

0.20

0.15

0.15

0.10

0.10

0.05

0.05

0.00 0.00

0.12

0.02

0.04

0.06

0.08

0.10

0.12

0.00 0.00

0.02

0.04

SSM/I PR

0.06

0.08

0.10

0.12

SSM/I PR

Figure 11.31 Scatter plots of AVHRR-derived ice thickness and SSM/I PR85 and PR37. The solid black line in the top panels denotes the linear regression and in the bottom panels the exponential regression. The vertical lines with crossbars show the standard deviation of the plots with respect to the regression curve. The blue, red, and green points in the top panels represent data from the NOW polynya, Chukchi Polynya, and Laptev polynya, respectively. Regression lines are shown in the same color [Tamura and Ohshima, 2011, Figure 3, with permission from AGU].

shown in red. Data from the two groups overlap but the second set has smaller PR89. This means that the ice thickness tends to be overestimated on days that have clouds or high-integrated water vapor contents. The boundary that separates the cloudy and cloud-free data points is delineated by the solid curve in Figure 11.32. The equation of this curve (developed for the Chukchi Sea polynya data using AMSR-E data) is:

PR89 = 3 2 PR36

2

− 0 077 PR36 + 0 0066

(11.65)

A similar equation is presented in Tamura et al. [2007] for data from the Antarctic Ocean (developed for SSM/I data): PR85 = 4 492 PR37

2

− 0 1062 PR37 + 0 01336 (11.66)

508

SEA ICE

Table 11.3 Coefficients of the empirical equation (11.64) that relates ice thickness to the polarization ratio from PM. Reference Tamura and Ohshima [2011] Iwamoto et al. [2013]

Sensor

Region

SSM/I SSM/I AMSR-E AMSR-E AMSR-E AMSR-E AMSR-E AMSR-E

Arctic polynya Arctic polynya Polynya & MIZ Polynya MIZ Polynya & MIZ Polynya MIZ

0.20

PR89

0.15

0.10

0.05

0.00 0.0

0.05

0.10

0.15

0.20

PR36

Figure 11.32 Scatter plot of AMSR-E polarization ratio PR36versus PR89 for clear sky (black dots) and sky with cloud or significant water vapor contents (red dots). The solid curve is the boundary represented by equation (11.65) that separates the two groups. Pixels below that boundary and with PR36 > 0.05 are excluded from the ice thickness calculation [Iwamoto et al., 2013, Figure 2, with permission from Taylor and Francis].

The pixels below the curve defined by equation (11.65) with PR36 > 0.05 are considered to be influenced by clouds or atmospheric water vapor and, therefore, excluded from the calculations of thin ice thickness. It is obvious that this method minimizes the error due to clouds but does not eliminate it. Alternatively, a semi-empirical radiative transfer model can be used to account for the atmospheric influences as presented in Gloersen and Cavalieri [1986] and Kern [2001]. Comparison between thin ice thickness retrieved from SSM/I using equations (11.62) and (11.63) and the corresponding estimates using the TIR technique is presented in Tamura and Ohshima [2011]. Results are shown in Figure 11.33 using observations acquired over the North

Frequency 85 37 36 36 36 89 89 89

GHz GHz GHz GHz GHz GHz GHz GHz

α

β

γ

215.15 88.40 206 40 300 218 103 298

0.508 1.023 –5.4 1.8 –11.1 –3.0 1.1 –6.7

1.0395 1.113 1.02 1.14 1.01 1.03 1.07 1.02

Water polynya on 30 March 1998. The AVHRR data are mapped onto the SSM/I grid in the figure. The surface temperature derived from AVHRR is included as a validation tool for the ice thickness results. The authors reported that the linear equations [(11.62) and (11.63)] and the exponential equation (11.64) produced very similar results. An OW filter was applied to reject pixels with ice concentrations < 30%. The thin ice thickness maps from AVHRR and SSM/I in the figure generally agree, yet with two exceptions. The first is the higher ice thickness estimate from SSM/I at the opening of Jones Sound into the Baffin Bay. The second is the larger area of ice thickness definition in the AVHRR data. This can be partly explained by the finer resolution of the AVHRR, but it can also be caused by the rejection of the low-ice concentration pixels from the calculations using SSM/I data. A similar observation is made in Iwamoto et al. [2013] who compared ice thickness retrieved from the TIR channels on MODIS against retrievals from AMSR-E data (using the 89 GHz channel) in the Chukchi Polynya. While IST delineates the northern boundary of the polynya (cold MYI and thick FYI coexist upstream of the ice arch while warmer thin ice types exist downstream of the arch), it does not delineate the southern boundary between the polynya and the OW (below latitude 77 N). On the other hand, ice thickness maps show clearly the entire boundary of the polynya. The SMOS satellite was launched in November 2009 and has been providing data since the summer of 2010 [Mecklenburg et al., 2012]. It carries a radiometer called Microwave Imaging Radiometer with Aperture Synthesis (MIRAS). It features a multi-angular, dual-polarization (H and V) channels operating in the L-band (1.4 GHz and 21 cm wavelength). The spatial resolution is around 50 km. It is more suitable for ice thickness retrieval because of the deeper penetration of the L-band signal [Kerr et al., 2010]. The coarse footprint also makes it suitable for synoptic scale observations. A method of ice thickness retrieval from this sensor has been presented in Kaleschke et al. [2012]. The approach is different from the above-described empirical approach using the traditional microwave channels. Here, ice thickness is

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 509 (b)

(c)

(a)

Ice arch

78°N

(K)

78°N

78°N

(m)

0.15 60°W

60°W

250

0.5

240 74°N

0.0

80°W

74°N

80°W

80°W

Jones Sound

0.1

74°N

60°W

260

Baffin Bay 230

Figure 11.33 Estimates of ice thickness using AVHRR and SSM/I in the North Water polynya (a) and (b), respectively. The corresponding IST derived from AVHRR data is shown in (c) [Tamura and Ohshima, 2011, Figure 2/ reproduced with permission from American Geophysical Union].

calculated inversely from a model of brightness temperature as a function of sea ice and water emissivity. The model starts with the familiar equation that decomposes an observed brightness temperature Tb from a heterogeneous footprint into its components from ice and water, weighted by the ice concentration C; T b = Cεi T i + 1− C εw T w

(11.67)

where, ε is the emissivity, T is the physical temperature, and the subscripts “i” and “w” stand for ice and water, respectively. The large contrast between the ice and water emissivity (0.8–0.9 and 0.3–0.5, respectively) justifies the use of this equation. Atmospheric and ionospheric effects are insignificant in the case of the L-band [Reul et al., 2008]. The model proceeds by introducing expressions for εi and εw, which are functions of the permittivity. The emissivity of the ice is also a function of its thickness H and brine volume. A semi-empirical approximation for the incoherent solution of equation (11.67) is given by Kaleschke et al. [2012]: T b = T b,m − T b,m − T b,w exp γH

(11.68)

where, Tb,w and Tb,m and are the brightness temperatures of sea water and the mixture of the heterogeneous footprint (ice and water), respectively. The latter is given by: T b,m = C T b,ice + 1− C T b w

(11.69)

where, Tb,ice is the brightness temperature of the infinitely thick sea ice. Both Tb,ice and Tb.w are functions of the surface temperature and salinity. These parameters can be

determined using satellite observations from ice-free and thick-ice-covered footprints. The coefficient γ in equation (11.68) can be determined from a least squares linear regression between brightness temperature and ice thickness. The relation between ice thickness and Tb formulated in equation (11.68) is shown in Figure 11.34. The data points are determined for different bulk sea ice temperatures with a constant low salinity of 0.65‰ (typical of the ice in the Baltic Sea) and surface roughness equivalent to one-tenth of the ice thickness. The sensitivity of Tb to ice thickness continues up to 40 cm at ice temperature −5 C, and around 150 cm when the temperature drops to −20 C. Recall that the polarization ratios from the 37 and 85 GHz channels are sensitive to ice thickness up to about 10–20 cm (Figure 11.31). This demonstrates the advantage of using the L-band data for mapping thicker ice although the output is produced at coarser resolution of 50 km. Kaleschke et al. [2010] stated that for typical polar sea ice with bulk ice temperature of –5 C and salinity 8‰, the resulting Tb from the L-band would resemble that of sea ice in the Baltic Sea at a temperature of −0.5 C. The ice thickness can be derived from equation (11.68): H= −

1 T b,m − T b ln γ T b,m − T b,w

(11.70)

This equation has been used by Kaleschke et al. [2012] to derive ice thickness up to 50 cm from the SMOS observations when surface temperature is below −10 C.

510

SEA ICE 260

Brightness temperature [K]

240 220 200 180 160 140

Tice = –20° Tice = –10° Tice = –5° Tice = –2° Hmax

120 100 80 0.0

0.2

0.4

0.6

0.8

1.0

Ice thickness [m]

Figure 11.34 Variation of brightness temperature from the SMOS L-band sensor as a function of ice thickness [following equation (11.68)] for different bulk ice temperatures and constant bulk salinity of 0.65‰ and surface roughness = 0.1 of ice thickness [Adapted from Kaleschke et al., 2010].

The authors define the maximum ice thickness Hmax that can be retrieved for a given observational error δ as: H max = −

1 δ ln γ Δ

(11.71)

where, Δ = Tb, m − Tb, w, and δ is the uncertainty of the tie point Tb,ice. The accuracy of this retrieval has been estimated by the same authors. They reported that an uncertainty in the radiometric accuracy of ±1 C leads to a thickness uncertainty of ± 2 cm for H = 40 cm, and ±5 cm for H = 70 cm. However, due to its coarser resolution small polynyas and refrozen ice in leads and other openings may not be detected by the SMOS MIRAS. Therefore, a combination of the L-band microwave with other traditional bands may satisfy operational ice monitoring and climatic application needs. Kaleschke et al. [2010] show maps of retrieved ice thickness and provide indications of the ice types (FYI or MYI). 11.4.3. SIT from Altimeter Observations Satellite altimeter systems (laser or radar) are used to estimatie ice thickness. Unlike the TIR and PM sensors, an altimeter can estimate ice thickness within its full range from a few centimeters to a few meters. However, data of thin ice (smaller than, say, 20 cm) may not be reliable because of their high sensitivity to errors and uncertainties in estimating the ice freeboard as explained later. That is why thickness estimates from altimeter systems should

complement estimates of thin ice thickness from PM data (up to 30 cm thick). So far there is no satellite platform that carries both sensors. Fundamentals of ice thickness retrieval from altimeter systems are briefly presented in section 7.8. The platforms and sensors are described in section 8.6. In this section the discussions and presented data focus on: (1) freeboard estimation, (2) samples of Arctic ice thickness, and (3) sources of error in estimating ice thickness from altimeters. Ice thickness from altimeters is calculated based on an estimation of the freeboard of the ice sheet, which is usually measured by the same sensor. Sea ice is assumed to float in a state of hydrostatic equilibrium. In addition, calculations use estimates of snow depth and density derived from climatology database. Different amounts of snow usually accumulate on FYI and MYI [Warren et al., 1999]. The temporal resolution of the derived thickness from altimeter systems is rather poor. That is because the satellite overpasses provide measurements along the ground track of the satellite. This means that filling gaps to produce regional thickness maps requires data from numerous passes over a long period. This is not operationally acceptable. Therefore, the real value of the altimeter data resides in providing climate-related information on ice thickness and volume that reveals trends of interannual variability at a synoptic scale. There are two key sources of uncertainty in estimating SIT from altimeter data. The first is error in estimating the snow on ice, and the second is uncertainly in estimating

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 511

ice/snow freeboard (height of the surface above the local sea level). The first source is well addressed in Mallett et al. [2021]. The second is discussed here and its link to sea surface height (SSH) is demonstrated. In the early years of developing sea ice applications from altimeter data, rough generic estimates were used for freeboard. For example, Kwok et al. [2007] determined the mean freeboard of FYI and MYI in the Arctic to be 14 cm and 35 cm in the fall, and 27 cm and 43 cm in the winter. The increase in winter is caused by the 4 months of ice growth and snow accumulation (mainly due to snow in the case of MYI). A better estimate was adopted by interpolating measurements at ice floes from a number of lead observations along the altimeter’s orbital track [Farrell et al., 2009, Onana et al., 2013]. Recently, a growing number of studies have been undertaken to improve the accuracy of freeboard and SSH estimation. Landy et al. [2021] used an objective mapping approach to estimate SSH from all proximal lead samples within 105 km around the satellite orbital track. By examining the covariance of CryoSat-2 lead observations, the SSH zonal and temporal decorrelation length scales were obtained, and then used to determine an optimal SSH and error estimate for each sea ice floe location. The authors claim that this approach has improved the freeboard estimation at orbital crossovers by up to 20%. Zhang et al. [2021a] estimated the Arctic sea ice freeboard and from Envisat satellite altimetry during the period from 2002 to 2012. Residuals in the static geoid were removed by using a moving average technique, then determining the local SSH and sea ice freeboard from the Envisat elevation profiles. The authors validated the estimates using

January 2019

freeboard products from three sources: ESA’s CCI, the Alfred Wegener Institute (AWI), and the Operation IceBridge (OIB). For freeboard > 0.3 m, the estimates correlate better with OIB freeboard than with CCI and AWI data. Sea ice freeboard in the polar oceans has been included as science products (Level 3A) from the ICESat-2 (ATL10) [Kwok et al., 2021]. The first step in the method of calculation is the identification of the surface returns that could be used to estimate the height of the local sea surface. In earlier releases of this product the estimated SSH was biased (toward higher values) due to the presence of clouds, resulting of lower estimate of freeboard. This has been corrected in the latest release (Release 003) by using the more reliable specular leads (excluding the dark leads). Monthly composite freeboard over the Arctic Basin is shown in Figure 11.35 for January, June, and October, 2019. The red area (large freeboard) north of the Canadian archipelago is covered with MYI and the blue area at the margin is thin ice and MIZ. The distribution of freeboard matches that of ice thickness, hence (to some extent) ice type. Note the apparent higher freeboard in January. It is worth noting that melt ponds prevent the application of radar altimetry over sea ice during summer melt. The first radar altimeters flew onboard ERS-1 (launched in 1991), ERS-2 (launched in 1995), and Envisat (launched in 2000). They were used to obtain snap shots of SIT in the Arctic. Laxon, Peacock, Smith [2003] used data from ERS-1 and ERS-2 to compile the first Arctic-wide ice thickness map in winter (October– March), averaged over an 8-year period from 1993 to

June 2019

0.0

0.3

October 2019

0.6

Total freeboard (m)

Figure 11.35 Monthly mean freeboard in the Arctic in 3 months representing winter, late spring, and fall seasons. The thick ice north of the Canadian Arctic Archipelago has larger freeboard. Data obtained from ICESat-2 specular leads [Kwok, et al., 2021 / Reproduced from Copernicus Publications / CC BY 4.0].

512

SEA ICE

2001. This has been used as a reference map to estimate the change in Arctic ice thickness from more recent instruments. Data were not available in the MIZ or above the latitudinal limit of 81.58 N. The authors stated that accuracy of the ice thickness estimates was within 11 cm. Achievements of ICESat (operated from January 2003 to February 2010) are presented in Kwok [2010]. SIT was a product calculated during a number of data acquisitions campaigns from this satellite. A total of 10 campaigns were achieved by the end of February 2007. Each campaign spanned approximately a 33-day sub-cycle of the 91-day repeat cycle of the satellite’s orbit. An early example of ice thickness profile along a 160 km ground track north of Ellesmere Island in the Arctic is presented in Kwok, Zwally, Yi [2004] and reproduced in Figure 11.36. The coincident RADARSAT-1 image is included to provide a closer inspection of the sea ice feature along the ICESat track. It is resampled into 150 m pixel spacing to match sample spacing of ICESat. The freeboard was estimated using a reference of water surface obtained from open leads or leads covered with very thin ice. The primary ice types that appear in the RADARSAT-1 image are thick FYI and MYI. The dark areas are either OW or refrozen smooth thin ice in leads. Most of the thickness data in Figure 11.36 exceed 2 m and reach 6 m at some points, which means that the ice cover is mainly composed of MYI. This highlights an interesting

issue regarding SAR image interpretation. According to the rules of SAR image interpretation, MYI should appear as single or agglomerated ice floes with round contours. The appearance of the MYI in this figure, with the confirmation from its ice large thickness, defies the application of this rule. In other words, it is possible that the background gray tone in this image can be interpreted as deformed FYI. Admittedly, the interpretation of the background as being MYI can be mistaken as deformed FYI and this is not unique to this image. The authors of this book noticed similar appearance of MYI quite often in areas just north of the Canadian Arctic Archipelago, the Beaufort Sea when ice undergoes strong gyre action, and the Fram Strait (check section 9.3.2 and comments on Figure 9.22). The common feature between mobile sea ice in those regions is the crushing of the ice under enormous compression stress. This leads to the disappearance of the typical backscatter and shape of MYI. However, the appearance of such ice form in SAR images in those regions is usually correctly classified as MYI in operational ice charts. This is verified from the ICESat data in Figure 11.36. The most recent and frequently used radar altimeter to monitoring FI thickness in the polar oceans is ESA’s CryoSat-2 (launched on 8 April 2010). The first map of SIT obtained from CryoSat-2 mission over the entire Arctic Basin was produced by the Centre for Polar Observation and Modelling at the University College in

RADARSAT image of sea

ICESat track

Ice thickness (m)

0 –2 –4 –6 –8

New lead with thin ice 0

20

40

*Reference thickness estimated using ice age from RGPS

60

80

100

120

Along-track distance (km)

Figure 11.36 A RADARSAT-1 image acquired on 20 March 2003, north of Ellesmere Island in the Arctic region (top) and a near-coincident ICESat-derived thickness estimate (bottom). The dark objects in the RADARSAT image are new leads. The dashed-line in the top panel is the ICESat track. The dotted line in the bottom panel is the zero reference line of seawater surface. The red bands connecting the top and bottom panels highlight areas of thin ice or OW in the RADARSAT image, mostly found in leads [Kwok et al., 2004 / with permission from American Geophysical Union].

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 513

London, UK. It shows the average ice thickness at each observation footprint during the months of January and February 2011. Thickness from those maps served as a reference set to explore the interannual variability of Arctic SIT in later years yet with advanced processing of CryoSat-2 data. Tilling, Ridout, Shepherd [2018] developed an end-toend algorithm of CryoSat-2 processing steps to estimate SIT and subsequent ice volume in the northern hemisphere with support from complementary observations. This has been a continuous endeavor undertaken by the Centre for Polar Observation and Modelling (CPOM), part of the Natural Environment Research Council (NSERC), since the early 1990s. The authors found no significant bias in thickness estimates when compared with independent measurements. Xiao et al. [2020] developed a model based on least squares adjustment to estimate Arctic SIT with CryoSat-2 measurements, including freeboard estimates. The study presents a set of monthly Arctic-wide maps of ice thickness and another set of uncertainty of the thickness estimate from October 2018 to April 2019. It compared the results from the new method against products from other sources including the AWI, the CPOM, NASA Goddard Space Flight Center (GSFC), and OIB. As expected, large differences (a few centimeters or tens of centimeters) are found in areas covered with thin ice due to the poor accuracy of the calculated freeboard there. Monthly mean SIT over the Arctic region has been generated regularly from CryoSat-2 data by ESA’s CPOM (www.cpom.ucl.ac.uk/ csopr/seaice.php). Average SIT for December 2020 was 1.28 meters. This was the fourth lowest over the CryoSat-2 record and below the 2010–2020 average of 1.35 meters. ICESat-2 mission, launched in September 2018 with its sole instrument ATLAS (section 8.6), has been used to generate Arctic-wide ice thickness. Petty et al. [2020] calculated ice thickness by combining the measured freeboard from the same satellite with a new model of snow on sea ice. This involves snow depth and density data from the NASA Eulerian Snow on Sea Ice Model. The coarseresolution snow data (~100 km) are redistributed onto the high-resolution freeboards from ATLAS (~30–100 m) using relationships between snow depth and freeboard data collected by NASA’s OIB mission. ATLAS L3A sea ice freeboard (called ATL10) data sets are available from NSIDC. Petty et al. [2020] generated a gridded monthly thickness product and compared it with thickness obtained from CryoSat-2 estimates. ICESat-2 ATLAS show consistently lower thicknesses. This point is further demonstrated later although more work is still needed to explain the difference. Comparison between SIT from CryoSat-2 and ICESat is presented in Kim et al. [2020]. Since there was no

overlap between the two sensors (ICESat ended on 14 August 2010, and CryoSat launched on 8 April 2010), the authors used ice thickness data derived from the difference between vertically polarized emissivity (10.65 and 18.7 GHz channels) onboard AMSR-E to assess the difference between the products from the two altimeters (AMSR-E overlapped with each altimeter, and ended in October 2011). In other words, AMSR-E was used as a bridge connecting the data sets from the two altimeters. Results showed that ice thickness from ICESat was systematically lower than that from CryoSat-2 by about 50 cm. Validation against ice thickness from AMSR-E and also thickness product from OIB measurements suggested that the 50 cm difference was more likely caused by an underestimation of SIT from ICESat rather than an overestimation by CryoSat-2. NSIDC confirms that SIT measurements from ICESat-2 are considerably thinner (up to a meter thinner in some places) than CryoSat-2 estimates ((https://nsidc.org/nsidc-highlights/ 2020/06/frozen-drift-mosaic-story-part-i). This appears to be largely caused by the different sensitivities of the two retrieval methods to errors in input snow depth. Recall that the laser sensors onboard ICESat platforms operate at 532 and 1064 nm wavelength. This ensures reflection of the incident laser beam off the snow surface. The wavelength of the Ku-band radar altimeters is much larger (22,370,000 nm) compared to the laser band. This means that the radar signal can penetrate through the snow cover all the way to ice–water interface but only if the ice is fresh and dry. Otherwise, the penetration could be shorter if the snow becomes compacted, soggy, wet, or metamorphosed. If the radar signal is reflected off any depth before snow–ice interface, then the ice thickness would be overestimated. This explanation is not supported by the comparison of SIT maps across the Arctic region, estimated from ICESat-2 and CryoSat-2, and processed at NSIDC as shown in Figure 11.37. The figure shows that ICESat-2 produces larger ice thickness and captures the thicker ice extending across the Beaufort Sea, which agree with sea ice age data. Currently, research studies continue to provide more reasonable explanation for the difference between ice thickness estimated from CryoSat-2 and ICESat-2 data. More data on the variability of Arctic SIT during the past few decades are presented in section 13.3.2. Assessment of uncertainty in ice thickness retrieval from altimeters has been addressed in several studies and still an active field of research. Regardless of the uncertainties of the freeboard measurements discussed above, conversion of freeboard to ice thickness requires knowledge of the snow depth as well as the densities of snow, sea ice, and sea water. Uncertainties in these parameters inevitably lead to uncertainties in SIT estimations. It is difficult to obtain those parameters concurrently with

514

SEA ICE CryoSat-2 SMOS - 15 April 2021

ICESat-2 30 - April 2021

0.0

1.0

2.0 3.0 4.0 Sea ice thickness (m)

5.0

0.0

1.0

2.0 3.0 4.0 Sea ice thickness (m)

5.0

Figure 11.37 SIT distribution across the Arctic region, estimated from ICESat-2 and CryoSat-2 for data acquired on 30 April 2021 (courtesy NSIDC, available through https://nsidc.org/data/IS2SITMOGR4, with modifications by the authors).

altimeter measurements. The relative contribution of the error in each one of the above parameters to the estimated ice thickness is still an active research topic. Tonboe, Pedersen, Haas [2010] addressed the sources of uncertainty and errors in estimating ice thickness from radar altimetry using a scattering simulation and field data approaches. Field data revealed that the spatial variability of snow depth could produce spatial and temporal uncertainty of about 0.3 m in the estimated ice thickness. Another 0.3 m uncertainty is caused by unknown variability of snow density. In addition to the well-established factors that affect the ice floe buoyancy (namely; ice density, snow density and snow depth), the simulations assert that the radar penetration variability and preferential sampling errors are no less factors affecting the freeboard measurements than those affecting the ice floe buoyancy. The sensitivity of the estimated ice thickness to errors in snow depth can be determined simply from the first-order differentiation of the second term in the RHS of equations (7.96) and (7.97) for laser and radar altimeters, respectively. Based on this approach, Kim et al. [2020] suggested that an error of +5 cm in snow depth can cause a bias of approximately −35 cm in ice thickness from ICESat and only +15 cm from CryoSat-2. That is how an uncertainty in snow depth of 5 cm can cause a 50 cm relative bias between estimates from the two systems. Tilling, Ridout, Shepherd [2018] suggested that the prime factors that cause uncertainties of SIT and volume estimates from altimeters are snow depth and density. Their contribution should be considered in conjunction with their monthly variation. The authors concluded that errors of snow depth and density are likely to cause

overestimation of ice volume by an average of 10% and 6% of ice each month. They pointed out the importance of improving estimates of snow loading on sea ice. Ricker [2015] reported from observational evidence that the increase of freeboard correlates with snow accumulation events over the Arctic MYI. Once again, this may lead to overestimated SIT. Price et al. [2013] addressed an issue pertaining to the sea ice in the Antarctic, namely the formation of a platelet ice layer under the ice–water interface (Figure 5.5). This leads to overestimation of the measured freeboard and therefore ice thickness. Clouds and atmospheric constituents are additional sources of error that should be corrected for. Mu et al. [2018] assimilated ice thickness products from CryoSat-2 and SMOS in the Arctic regional ice-ocean model to produce Arctic-wide maps of ice thickness. The approach combines the skill of satellite thickness obtained during freezing seasons with model skill in the melting season, when satellite data are unavailable. The method has been proven to be reliable in producing ice thickness data comparable to data from in-situ observations, mass balance buoys, and NASA’s OIB. However, as expected, it fails to produce thickness of heavily deformed ice. This issue will continue to defy the use of any remote sensing technology or model scheme that operates at a footprint much coarser than the typical deformation scale. 11.4.4. SIT from SAR Observations Since SAR is more sensitive to desalination of the sea ice subsurface, it can be used to retrieve thermodynamically

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 515 (b)

(c) 3.0

–10

2.5

–15 –20 –25 HH HV VV

–30 –35

150 Estimated ice thickness (cm)

–5

Polarization ratio RVV/hh

Backscattering coefficient (dB)

(a)

2.0 1.5 1.0 0.5 0.0

0

30

60 90 120 Ice thickness (cm)

150

120

90

60

30

0 0

30

60 90 120 Ice thickness (cm)

150

0

30 60 90 120 Observed ice thickness (cm)

150

Figure 11.38 (a) L-band backscattering coefficient versus ice thickness data from airborne SAR over the Sea of Okhotsk in 1997, (b) co-polarization ratio versus ice thickness, and (c) observed versus estimated thickness from equation (11.73) [Nakamura et al., 2005 / IEEE].

grown undeformed thin sea ice (to avoid interference with surface roughness, which also affects backscatter). This is the favorable condition for land-fast ice. SIT retrieval from SAR is restricted to an upper limit of 15–20 cm for the same reason mentioned for the PM data, namely the rapid desalination of the ice during the early growth phase (see section 2.5.5.1). Desalination leads to sharp increase of the backscatter. Another major restriction on using SAR for SIT retrieval is the limited spatial and temporal coverage of the sensor. This is not acceptable to fulfill requirements of operational sea ice monitoring. SIT retrieval from SAR is mostly based on empirical equations. Correlation between ice thickness and backscatter coefficient (at different polarizations) or copolarization ratio Rhh/vv (σ 0vv σ 0hh ) has been recorded and used to estimate SIT. Nakamura et al. [2005] pursued this approach using data from an airborne X-band and L-band SAR system, called Pi-SAR. The spatial resolution was 1.5 m for the X-band and 3.0 m for the L-band. Data were obtained over the Sea of Okhotsk and results are shown in Figure 11.38 for L-band data (data from the X-band show very similar trends). While the backscatter is not sensitive to ice thickness beyond 30 cm (as expected), Rhh/vv continues its sensitivity up to 120 cm thick ice. Nakamura et al. [2005] justified this sensitivity as follows. The contribution of the surface roughness is assumed to be the same for both co-polarization returns. Consequently, any change in Rvv/hh should be attributed to changes in the dielectric constant or equivalently the salinity of the penetrating layer within the ice sheet. As salinity decreases with the sea ice growth, σ 0vv decreases more than σ 0hh, hence the decrease of σ 0vv σ 0hh. This explanation, however, is questionable because neither the X-band nor the L-band can penetrate through saline FYI any deeper than, say a few tens of centimeters. Moreover, the bulk salinity of ice stabilizes after 40 cm thickness

(Figure 3.6). Nevertheless, regression of the data in Figure 11.38 leads to the following equations, which were used to estimate the ice thickness H (in cm). They were validated against ground observations as shown in Figure 11.38c. H = 202 0 − 90 91Rvv

hh

for the X band

(11.72)

H = 177 9 − 76 92Rvv

hh

for the L band

(11.73)

where, Rvv/hh is the same as σ 0vv σ 0hh. The validation of the L-band equation shown in Figure 11.38c shows successful ice thickness retrieval up to ice thickness of 1 m. The L-band radar signature is often less affected by surface roughness and penetrates the ice cover more than C-band. Therefore, its use may offer an advantage. Moving from airborne to space-borne SAR, Nakamura et al., [2005] showed the same correlation using SAR backscatter from Envisat and ice thickness measurements along the track of the Japanese icebreaker RV Shirase in the Lützow-Holm Bay East Antarctic during midDecember, 2004. Ice thickness and snow depth data were obtained using a downward-looking camera. Digital images were analyzed to quantify these two parameters by measuring the cross section of broken ice pieces. Data are presented in Figure 11.39. Once again, Rvv/hh decreases as the ice thickness increases. The relation appears to be valid up to 2 m ice thickness. Note the limited thickness range of fast ice compared to the pack ice. Nakamura [2009] developed a linear and logarithmic regression equations relating ice thickness H and Rvv/hh: H = a1− Rvv

hh

a0

H = exp b1− Rvv

hh

(11.74) b0

(11.75)

Sets of coefficients a0, a1, b0, and b1 for pack ice and fast ice are listed in Table 11.4.

Polarization ratio Rvv/hh (dB)

516

SEA ICE 3.0

3.0

2.5

2.5

2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5 Pack-ice

0.0 0.0

0.5

1.0

1.5

2.0

Fast-ice

3.0

2.5

0.0 0.0

0.5

1.0

1.5

2.0

2.5

3.0

Ice thickness (m)

Ice thickness (m)

Figure 11.39 Correlation between the co-polarization ratio of backscattering from Envisat C-band SAR and ice thickness measured along a track of an icebreaker in the East Antarctic during mid-December, 2004. (a) Data from pack ice and (b) data from fast ice. The linear regression is the thick line and the logarithmic regression is the thin line [Nakamura, 2009 Figure 4, with permission from IEEE].

Table 11.4 Coefficients used in equations (11.74) and (11.75) [Nakamura et al., 2009 / IEEE]. Linear fit

Pack ice Fast ice

Logarithmic fit

a0

a1

b0

b1

0.600 0.472

2.442 2.229

0.648 0.602

1.794 1.765

Using PALSAR L-band data, Toyota et al. [2011] found correlation between backscatter measurements σ ohh (in dB) and in-situ measurements of ice thickness and surface roughness. The experiment was conducted in the southern area of the Sea of Okhotsk in February 2008. The surface roughness was measured using a helicopter-borne laser profiler. The following linear equation can be used to estimate the ice thickness (in m): H = 0 047 σ ohh + 1 012

(11.76)

The equation can be applied to estimate ice thickness in the range from 0.2 to 0.6 m, excluding DI. Toyota et al. [2011] found that the RMSE between measured and estimated ice thickness using the above equation is 0.04 m and the correlation coefficient between them is 0.86 significant at 99% level. The same study presents a linear regression equation between σ ohh and surface roughness. The correlation coefficient is 0.70 significant at 99% level in

this case. Equations (11.72) –(11.76) seem to offer a simple and promising method for SAR-driven ice thickness retrieval. However, no widespread applications have been reported. Robustness, uncertainty, and extensive validation of the results should be established. The presence of metamorphosed snow or other ice surface conditions that influence SAR observations, e.g., frost flowers or surface flooding may annul the application of the aforementioned equations. Moving from dual-polarization to FP SAR data, Shokr and Dabboor [2020] used RADARSAT-2 FP mode to estimate thickness of thin fast ice. The authors examined the correlation of the following parameters with thin ice thickness: (1) the traditional SAR backscatter σ ohh , the total backscatter (SPAN), the polarization ratio σ 0vv σ 0hh , the entropy (H), and alpha-angle (α) (section 7.6.2.3). The study found that all these parameters are correlated to ice thickness up to 30 cm thick. Empirical non-linear regressions were constructed to relate the ice thickness to each parameter. Once again, more data are needed to confirm the robustness of these relations. Figure 11.40 shows the temporal evolution of SIT in a fjord in the area of Resolute Passage, Canadian Arctic from 20 September to 10 October. The thickness at each pixel is calculated as the average of thickness estimated from using σ ohh , σ 0vv σ 0hh , H, and α. The figure demonstrates the gradual thickening of the fast ice in the fjord. More studies using

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 517 20 September 2017

27 September 2017

23 September 2017

30cm–35cm 20cm–30cm 20cm–25cm 15cm–20cm 10cm–15cm 5cm–10cm 0–5cm

30cm–35cm 20cm–30cm 20cm–25cm 15cm–20cm 10cm–15cm 5cm–10cm 0–5cm

10 October 2017

3 October 2017

30 September 2017

30cm–35cm 20 cm–30cm 20 cm–25cm 15cm–20cm 10cm–15 cm 5cm–10cm 0–5cm

30 cm–35 cm 20 cm–30 cm 20 cm–25 cm 15 cm–20cm 10 cm–15 cm 5cm–10 cm 0 –5 cm

30cm–35cm 20cm–30cm 20cm–25cm 15cm–20cm 10cm–15cm 5cm–10cm 0–5cm

30 cm–35 cm 20 cm–30 cm 20 cm–25 cm 15cm–20cm 10 cm–15 cm 5 cm–10 cm 0 –5 cm

Figure 11.40 Development of thickness of thin ice in a fjord in the Resolute Passage area in the Canadian Arctic, calculated from using RADARSAT-2 polarimetric parameters. Note the thickness increase with time until it stabilizes. The background is the image of σ ovv [Shokr and Dabboor, 2020 / with Elsevier].

SAR polarimetric decomposition parameters are expected to be undertaken in the future to explore the potential of ice thickness retrieval. This is particularly promising from using the CP SAR data as they become more readily available to users. CP data offer all of the above-mentioned parameters except for the entropy. However, a similar parameter can be derived from the data (Dabboor, M. Personal Communication). 11.5. ICE SURFACE TEMPERATURE (IST) IST is required for analysis of energy exchange between the ocean and atmosphere. It is also an approximate indicator of ice thickness/type since the ice surface becomes colder as ice thickens or ages. IST is part of MODIS data products because it can be estimated using TIR data. IST can also be estimated from PM observations. These two approaches are presented separately in the following to demonstrate their potentials and the range of applicability. 11.5.1. IST from TIR Observations TIR remote sensing is a practical tool for retrieving land or IST (mostly under clear sky) by using various data sets and methods. The theoretical background for IST retrieval from TIR observations is based on the fact that the emitted energy in this spectral region is proportional to the fourth power of the physical temperature of the radiating layer [equation (7.20)]. Therefore, a small

variation in the physical temperature of this layer would trigger large change in the emitted radiation (i.e., Tb). Algorithms that use TIR data to retrieve surface temperature (including ice surface) pertain to three categories: (1) single-channel algorithms (SC) [Jiménez-Muñoz and Sobrino, 2003], (2) mono-window algorithms (MW) [Qin, Karnieli, Berliner, 2001], and split-window (SW) algorithms [Key, 1997]. While most algorithms use radiative transfer equations (RTE) in one way or another to estimate parameters used in retrieving the surface temperature, the RTE can be used directly but it needs real-time atmospheric profile data. This method can produce most accurate results but atmospheric profile data are not easy to obtain, especially for polar regions. The SC- and MW-based algorithms need only specific atmospheric physical and radiative parameters, but they differ in the number of input parameters and the methods of atmospheric correction. A few versions of each category have been applied to different data from different TIR sensors because different sensors have different spectral response functions. The SC algorithm requires two input parameters, the surface emissivity and atmospheric water vapor contents. Because of fewer requirements of input data, it has become more popular in applications using Landsat, MODIS, ASTER, and AATSR. A few versions are available and researchers keep improving them [e.g., Jiménez-Muñoz et al., 2009]. The MW algorithm requires input of three parameters: mean atmospheric temperature, surface emissivity, and atmospheric transmittance. Two different forms for use with Landsat

518

SEA ICE

TM and Landsat-8 are presented in Wang, Lu, Yao [2019]. Li et al. [2019] presented an improved version of SC (called ISC) to retrieve surface temperature of glaciers using Landsat-8 band 10 data. The main steps include: (1) simulation of atmospheric radiative parameters by regression against the atmospheric water vapor content and the effective mean atmospheric temperature; (2) calculation of IST using Planck’s equation, instead of using Taylor’s approximation; and (3) implementation of an iterative scheme for IST calculation. The SC and MW methods have recently gained momentum mainly because Landsat-8 has only one functional TIR channel (11 μm). They are also useful if one of the thermal windows of other TIR satellite sensors fails (e.g., on VIIRS). So far, the only application of the SC technique to estimate IST is presented in Liu, Dworak, Key [2018]. The authors used the single-band (11 μm) onboard VIIRS to retrieve the IST. They used the single-channel IST equation developed in Yu, Rothrock, Lindsay [1995]: T s = a + b T 11

(11.77)

where, Ts is the IST in K, T11 is the observation from the 11 μm channel, and a and b are regression constants. Liu, Dworak, Key [2018] validated the above equation using IST measurements from an airborne infrared radiation pyrometer during the NASA’s IceBridge campaign in the Arctic. Results show that IST from using the single TIR band from the VIIRS has comparable performance to IST with the VIIRS dual-band (SW) method, with a bias of 0.22 K and RMSE of 1.03 K. Neither ISC method developed by Li et al. [2019] nor MW method has been applied to sea ice. Fan et al. [2020] assessed the accuracy of IST derived from two SC and three SW methods based on Landsat-8 TIR imagery at 100 m resolution over Arctic sea ice regions. Practically speaking, IST estimation using the 11 μm channel only is close enough to the truth [Key et al., 1997]. However, the SW technique is commonly used to map this parameter in routine operational products from a few satellite sources such as AVHRR, MODIS, VIIRS the Suomi National Polar-orbiting Partnership (S-NPP) satellite, and the Joint Polar Satellite System (JPSS) (section 8.2). The SW method uses two TIR bands (11 and 12 μm) because the 12 μm channel is more sensitive to water vapor than the 11 μm channel (section 7.4). Therefore, the difference between the two channels is a function of the absorptive response of the water vapor content in the atmosphere. The more water vapor contents, the higher the difference (or the ratio) between two observations. Usually, a regression equation between (Tb,11 − Tb,12) and IST can be developed as shown in Equation (7.55). Maslanik et al. [2001] used the SW technique [equation (7.55)] with observations from AVHRR TIR channels acquired over the western Arctic in April–July, 1998,

during the SHEBA project (see section 6.6) to produce a suite of parameters that included: all-sky surface temperature, broadband albedo, upwelling and downwelling shortwave radiation in addition to a few cloud products. A key observation from the study was the above-freezing IST obtained within the ice pack. Upon stratifying the AVHRR-derived IST as a function of SSM/I-derived ice concentration, the study concluded that heating of OW within the pack ice contributed to ISTs that could be several degrees above the freezing point even when a substantial amount of ice is present. In other words, the temperature of an area that contains ice and OW increases non-linearly with the OW fraction. A relevant conclusion from the study of Maslanik et al. [2001, p.15, 243] states that this increase “contrasts with the assumption typically used in climate models, where OW is prescribed to remain at the freezing point until all ice is melted within a grid cell.” Lin et al. [2015] retrieved IST in the Antarctic using a modified version of the SW technique. The authors introduced a polynomial fitting for atmospheric transmittance simulation. Results were validated against MODIS IST product (MOD29) and automatic weather station (AWS) dataset from the Zhongshan Station in the Antarctic. The bias was −0.61 K and the RMSE was 1.32 K for the Zhongshan Station dataset. The IST is a standard product from MODIS at 1 km resolution using the same SW equation, but the algorithm incorporates a capability for cloud filtering. Description of the algorithm and examples of the product are given in Hall et al. [2004]. An estimated IST selected from this study is shown in Figure 11.41. The scene shows ice cover in the northern Greenland Sea. Recent refrozen leads can be identified by their relatively high IST. In the absence of validation data, it can be assumed that the area of relatively high temperature (red color) at the bottom right corner of the figure is thin ice. The surface of thin ice is relatively warm because it is more readily affected by the heat conduction from the warmer seawater temperature at the underside of the ice sheet. In fact, this concept was used to identify thin ice types from the IST measurements [Stone and Key, 1993]. The white color area at the left side in the figure is thicker ice (probably MYI). The accuracy of the MODIS-derived IST was estimated by Hall et al. [2004] to be within 1.2–1.3 K, but Key et al. [1997] reported a wider range of 0.3–2.1 K from using the same SW technique. Although this may be considered as an acceptable accuracy for most applications, the limitation on using the TIR data remains to be the difficulty of identifying a cloud-free sky using an appropriate cloud mask algorithm. For example, it is not possible to estimate IST accurately in the presence of even thin clouds or fog. A possible alternative is to estimate IST from a near-coincident value under clear sky and apply it to

RETRIEVAL OF SEA ICE GEOPHYSICAL PARAMETERS 519

N

Refrozen water between ice floes

Sea ice surface temperature (K) 233 235

240

245

250

255

260

265

270 271.5

Figure 11.41 Terra MODIS product of IST of a scene in the northern Greenland Sea acquired on 12 March 2003. The center of the image is at 81.7 N and 1.0 E. Warmer temperature (bottom right and left) indicates refrozen leads or thin ice sheet [Hall et al., 2004 / with permission from IEEE].

the area under cloudy sky. Another possible alternative is by using PM data, separately or in conjunction with the TIR data, as described later. IST can be used as an approximate indicator of ice types in an area of heterogeneous ice cover. This is demonstrated in Shokr [2001] to identify ice types that appear in RADARSAT-1 image (Figure 11.42) in the North Water polynya in the Arctic. IST data in the figure were calculated from equation (7.55) using observations from the AVHRR TIR channels. Data in the figure are presented in the form of elliptic footprints of SSM/I 85 GHz overlaid on the background RADARSAT-1 image. The color of the footprint represents the IST. RADARSAT-1 image was acquired on 24 March 1998, 2 weeks after the formation of the ice bridge (arch) that blocked the ice influx from the Arctic Ocean. The time difference between the RADARSAT-1 and SSM/I was slightly over 1 hour. The IST can be used to verify the ice types in the RADARSAT-1 (100 m resolution) image. Cold thick MYI and FYI existed upstream of the arch

220 222 224 226 228 230 232 234 236 238 240 242 244 246 248 250 252 254 256 258 260 262 264 266 268 270

Figure 11.42 A RADARSAT-1 image of the NOW polynya acquired on 24 March 1998 at 100 m spatial resolution with superimposed ellipses that represent footprints of SSM/I 85 GHz channel. The footprints are color-coded to represent the IST calculated from the TIR channels of a near-coincident overpass of AVHRR. Cold temperatures correspond to thick ice or fast ice while warm temperatures correspond to thin ice in the active polynyas [Shokr, 2001 / from IEEE].

while thin ice (perhaps mixed with OW) covers a large area downstream from it. The high-surface temperature area in the middle of the polynya is covered with thin ice and OW. The two cold ice temperatures at the two sides of the polynya are covered with thick fast ice. As mentioned before, the SW equation [equation (7.55)] can be applied only to cloud-free TIR observations. However, Key and Wong [1999] presented an empirical equation to estimate the IST under cloudy sky Tcld if the temperature under clear sky Tclr is given along with three other parameters: the cloud optical depth τ, wind speed u, and total snow and ice thickness Hs. The Tclr can be obtained from either nearby pixel or past record. The equation is given here, but it should be noted that it has not been widely applied. T cld = b0 + b1 T clr + b2 μ + b3 μ2 + b4 lnτ + b5 ln τ

2

+ b6 u + b7 H s

(11.78)

where, μ is the cosine of the solar zenith angle. Cloud optical depth can be obtained from satellite data while the wind speed and snow and ice thickness can only be obtained from ancillary data. The coefficients bi are obtained by regressing ice temperature under cloud and under clear-sky conditions. Key and Wong [1999]

520

SEA ICE

determined the uncertainty of the results from equation (11.77) to be about 2 K. They also presented data that showed a linear increase of the IST with the longwave downwelling flux during the winter in the Arctic. In general, they found IST to be highest for low wind speeds as heat was not removed from the surface as readily through turbulent transfer.

11.5.2. IST from PM Observations PM data have been used to estimate the temperature of the radiating layer of the ice. If snow-covered, this layer may integrate snow as well. In this case, the term IST can be used loosely because the radiating layer may involve mostly or solely snow. This is different from the skin temperature estimated using TIR data, which is much closer to the actual surface temperature. The penetration depth (radiating layer) of the TIR and microwave radiation in sea ice measures in sub-millimeters and millimeters, respectively. IST from PM data is commonly assumed to be equivalent to the temperature of the snow–ice interface for FYI or a weighted temperature of the freeboard portion of MYI [Steffen et al., 1992]. With this in mind, the IST derived from PM observations is not directly comparable with that derived from the TIR observations. Maslanik and Key [1993] developed a technique that combines AVHRR and SSM/I data to derive more accurate estimates of IST by incorporating the estimated ice concentration from PM data into the estimated temperature from a heterogeneous footprint. It also allows estimating IST under cloudy sky. There are two main differences between the TIR and microwave observations that impact the way the surface temperature is retrieved from the observations. The first is that the radiances in the TIR spectral region are more affected by the atmosphere. That is the reason for using the SW technique to account for atmospheric contribution to the emitted radiation. The second, is the large difference in emissivity between OW and sea ice types in the microwave bands compared to the insignificant difference in the TIR band. For that reason, the concept of the “composite” emissivity to account for the heterogeneity of emissivity within the footprint from microwave sensor has been developed (the footprint of PM sensors typically measures a few kilometers or tens of kilometers compared to the few hundred meters from TIR). All algorithms that estimate IST from PM data were developed a few decades ago. The first algorithm was applied to SMMR data [Gloersen et al., 1992]. A simple relationship between temperature of the radiating layer TRL and the brightness temperature at 6.6 GHz vertical polarization is used:

T RL =

Tb 6 6V εice 6 6V

(11.79)

Ice emissivity εice= 0.96 is selected. The 6.6 GHz channel is chosen because of the deep penetration of the signal. This ensures that the derived temperature can represent the snow–ice interface temperature or the top layer of the ice surface. However, the very coarse footprint of this channel (150 km) means that its composition is almost always heterogeneous. The calculated temperature is sensitive to the heterogeneity and is more reliable when ice concentration exceeds 80%. For ice concentration 0.25 and L > λ. These models can be combined to simulate the sea ice surface scattering in the full range of the roughness spectrum or to simulate large-scale undulations of hummocks with superimposed small-scale roughness [Nghiem et al., 1995a]. The integration of these different parts of the roughness spectrum has also been achieved in one model; the IEM [Fung, 1994, Fung and Chen, 2010] and later improvements of that [Fung and Chen, 2004; Fung and Chen, 2010]. The improved IEM agrees with the small perturbation model for low frequency and smooth surfaces and it agrees with the Kirchhoff model for high frequencies and high roughness surfaces [Fung and Chen, 2004].

12.4. EXAMPLE OF THE IMPLEMENTATION OF AN ALTIMETER MODEL TO STUDY THE IMPACT OF SALINE SNOW ON THE BACKSCATTER Simple first-order radar scattering models can answer some interesting questions which can help improve sea ice thickness mapping applications using radar altimeters. In the following a radar altimeter scattering model is applied to FYI covered by saline snow. Sea ice thickness is derived from radar altimeter estimates of sea ice freeboard under the assumption of the

547

ice floe being in hydrostatic equilibrium and that the snow–ice interface is the primary/dominating scattering horizon. The term “scattering horizon” is used because in radar altimetry there are corrections so that the track-point can be projected into the snow/ice profile and aligned with the snow–ice interface. Even though the snow–ice interface dominates the track-point, it is still affected by the air–snow interface scattering. Both assumptions have been debated using both forward modeling and experimental evidence [Rothrock, 1986, Tonboe et al., 2021]. Even though radar scattering from sea ice is still an open research topic, forward modeling can help in understanding what to look for in data, planning efficient campaigns, and making sure that the most important physical parameters are measured in the field. Here we use a simple first-order backscatter model as an example to simulate the backscatter from the snow surface and the snow–ice interface as a function of snow temperature for saline FYI and saline snow. The input dataset to the model is a simplification of data collected during the MOSAiC campaign (section 6.11) as presented in Table 12.1. For example, the snow and ice layers are isothermal in this simulation and as the snow temperature Ts increases, the ice temperature Ti increases at a slightly slower rate (using equation 3.4): Ti =

k i H s T w − k s hi T s k i hs + ks hi

(12.5)

where, Tw is the water temperature (271.35 K), ki is the thermal conductivity of ice (2.1 W/mK), ks is the thermal conductivity of snow (0.3 W/mK), hs is the snow depth (0.03 m) and hi is the ice thickness (0.83 m). The flat-patch area of both the snow surface and the snow–ice interface are the same at 0.5% throughout the simulation, the snow density is 320 kg/m3, and the ice density is 930 kg/m3. In reality, the snow and sea ice density changes as a function of temperature, but here it is kept constant. The surface and interface backscatter, σ 0, is a function of the flat-patch area, F, the Fresnel reflection coefficient at nadir, R(0), the height of the satellite H (730 km), the

Table 12.1 Input to the radar backscatter model.

1: Snow layer 2: Ice layer

Temperature, T [K]

Flat-patch area, F [m2/m2]

Density [kg/m3]

Layer thickness, d [m]

Scatterer correlation length (exp.), pec [mm]

Salinity [ppt]

240.0 250.0 260.0 246.3 254.3 262.3

0.005

320.0

0.03

0.15

7.8

0.005

930.0

0.83

0.15

9.0

Permittivity at 240 K, 250 K, 260 K 1.47 + 1.50 + 1.54 + 3.30 + 3.39 + 3.58 +

0.01j 0.02j 0.02j 0.08j 0.09j 0.12j

548

SEA ICE

speed of light, u (3e8 m/s), and τ is the pulse length (3.125 ns): H uτ

(12.6)

We tested a geometric optics model with an exponential surface correlation function and a surface roughness slope of 0.02 m, which is within the valid range of this model at Ku-band (13.6 GHz), and it yielded nearly the same backscatter as equation 12.6 with a flat-patch area of 0.5%. The permittivity of the saline snow is simulated in Mätzler [1998] as a cloud of spherical liquid brine droplets in a background of non-saline snow. The non-saline snow is simulated as a cloud of snow-flakes in a background of air. The ice permittivity is simulated with the same model [Mätzler, 1998] and here, the background is pure ice and the inclusions are spherical brine pockets. The procedure is the same as in Barber et al. [1998]. To compute the total backscatter as well as the backscatter from the snow surface and the ice surface, we use the first-order radiative transfer model described in Tonboe et al. [2006]. The total backscatter is: n

σ 0i

=

σ surf i

+

backs T 2i σ vol i

1

2 i = 0 Li − 1

T 2i − 1

(12.7)

where, L is the loss function of scattering and absorption within the snow layer and T the interface transmission backs coefficient (L0 = T0 = 1) for the first layer. σ vol is i the volume backscatter coefficient which is small compared to the surface scattering. σ surf is either the scattering i from snow surface or the snow–ice interface. Figure 12.2 shows the total backscatter, the backscatter from snow surface, and the backscatter from the snow–ice interface as a function of snow and ice temperature, and using Table 12.1 as input to the model. The precipitation of sodium chloride in the brine pockets and droplets at a temperature of −22.9 C (250.25 K) is clearly affecting the brine volume and the dielectric loss. The snow brine volume is higher than 14% in the shaded blue region for temperatures above 270 K. In this case, the assumption of uniformly distributed cloud of brine droplets in a background of fresh snow, which is used in the snow permittivity model, may not be valid with such high level of brine content. The brine will start to drain toward the bottom of the snow-pack. As temperature approaches the melting point, brine forms in channels rather than pockets. Hence, the pocket assumption becomes less valid. Snow brine volume of 14%, marks approximately the transition between the snow pendular, where the model is valid, and the funicular regime, where the model is not valid [Denoth, 1980]. The simulated backscatter from the snow surface and the ice surface are equal in magnitude at a temperature of about 265 K and the snow–ice interface is dominating

Total backscatter Snow surface snow-ice interface

27.5 Backscatter [dB]

2

σ 0 = 0 9F R 0

30.0

25.0 22.5 20.0 17.5 15.0 12.5 10.0 240

245

260 250 255 Snow temperature [K]

265

270

Figure 12.2 Total backscatter and backscatter from snow–air interface and snow–ice interface versus snow temperature. The vertical yellow line is at the precipitation temperature (−22.9 C) of sodium chloride. The blue shaded region is where the snow brine volume is higher than 14%, which is approximately the transition from the pendular to the funicular regime in snow.

the total backscatter for colder temperatures, while the snow surface is dominating at warmer temperatures. For FYI with a saline snow cover and cold temperatures below 265 K, the simulations show that the backscatter from snow–ice interface dominates the total backscatter measured with the altimeter. Conversely, for warm temperatures above 265 K, the snow surface backscatter dominates the total backscatter. The transition between these two ends is a function of snow depth, surface roughness, and snow salinity. When scattering from the snow–ice interface dominates, then the projected track-point is close to the snow–ice interface, and when scattering from the snow surface dominates, the projected track-point is closer to the snow surface. This has some important consequences for the way the sea ice freeboard is estimated and interpreted [Tonboe et al., 2021]. In radar altimetry it is normally assumed that the dominating scattering surface coincides with the snow–ice interface when the radar wave penetration time in the snow has been corrected for the lower propagation speed [Laxon, Peacock, Smidth, 2003]. The location of the scattering surface is very important to know when the radar freeboard is converted to sea ice thickness. 12.5. EXAMPLE OF COMBINING ATMOSPHERIC, OCEAN, AND SEA ICE EMISSION MODELS TO SIMULATE THE NOISE IN SEA ICE CONCENTRATION ESTIMATES Sea ice concentrations (SIC) derived from satellite brightness temperatures (Tb) are sensitive to the geophysical noise. These are generated, for example, from

MODELING MICROWAVE EMISSION AND SCATTERING FROM SNOW-COVERED SEA ICE

atmospheric water vapor, wind speed over the ocean, and the temperature of emitting layer of ice [Tonboe et al., 2022]. Some processing facilities (e.g., EUMETSAT OSISAF) use numerical weather prediction data and forward models to reduce the sensitivity of the Tb to atmospheric noise before computing the SIC [Andersen et al., 2006]. Depending on the combination of channels used as input, different SIC algorithms have different sensitivities to these geophysical noise sources. When using these estimated SIC data for sea ice climate data records it is important to: (1) find SIC algorithms with low sensitivity to noise that we cannot correct for, and (2) correct for noise to avoid introducing artificial trends from the noise sources. 12.5.1. Snow in the Emission Models Snow grain size is a parameter used in thermodynamic models to compute the shortwave extinction in the snow and it is further related to the microwave scattering magnitude. According to the temperature gradient metamorphism model for dry snow [Marbouty (1980)], the mean grain size diameter (Dm) growth is a function of temperature gradient and density. Baunach et al. [2001] and Jordan, Andreas, Makshtas [1999] indicate that the growth rate is also a function of the initial grain size. However, as a first approximation, Marbouty’s model gives a good fit to observations. For conversion to other structural parameters, it is an advantage that growth rate is independent of initial size (see below). The shortwave extinction in the snow is a function of the snow density and the optical grain diameter (D0) [Brun et al., 1992]. Since the growth rate is independent of initial size, Dm can be set equal to the diameter of equally sized spheres (D). D is related to D0 by the following expression [Mätzler, 2002]: D = D0 1− ν

(12.8)

where, ν=

ρsnow ρice

(12.9)

The emission model uses the exponential correlation length (pec), a structural parameter of scatterer size and distribution [Mätzler, 1998]. Mätzler [2002] analyzes different relationships between the snow structure, grain size, and the correlation length. D0 is related to pec for rounded grains, i.e., pec = F D0

(12.10)

where, F is 0.16 for snow grain sizes modeled with snow thermodynamic and mass model SNTHERM (described in Jordan [1991]), 0.3–0.4 for Crocus grain sizes and types, and 0.16 for fine-grained snow and 0.25 for medium-

549

grained snow using a one-dimensional scattering model (Crocus is described in Brun et al. [1989, 1992]). It is further suggested to include the snow density, i.e., pec = 0 5D0 1− ν

(12.11)

For realistic snow densities between 150–400 kg/m3, the factor 0.5D0(1 − ν) is 0.28–0.41, i.e., the factor F in the equation 12.10. These relationships are not strictly valid for new snow and depth hoar where the snow grains are not spherical [Mätzler, 2002]. There are no snow metamorphosis relationships in terms of pec, so the conversion between terms is needed thus assuming that the grain size and pec evolve similarly during metamorphosis. If the grain size growth model is not dependent on initial size, then the conversion factor F is not critical. 12.5.2. The Combined Sea Ice Thermodynamic, Atmospheric, Ocean, and Sea Ice Emission Models Here, an example of implementing a sea ice emission model for studying sea ice concentration cice algorithms and sensitivities is introduced. Sea ice thermodynamic model is combined with an atmospheric, ocean, and sea ice emission models to simulate top-of-the-atmosphere (TOA) Tb. The model combination and the input to cice algorithms are described in Tonboe et al. [2022]. Impacts of each geophysical noise source on the sea ice concentration estimate are determined. Here, we follow the procedure in Tonboe et al. [2022]: 1. The multi-layer snow and sea thermodynamic model are described in Tonboe [2010] and Tonboe, Dybkjær, G., Høyer [2011]. It takes meteorological parameters as input to compute detailed one-dimensional snow and ice profiles. Input includes: 2 m air temperature [K], relative humidity [%], 10 m wind speed [m/s], downwelling shortwave and longwave radiation [W/m2], and precipitation [kg/m2]. Output for each layer includes: temperature [K], density [kg/m3], thickness [m], snow grain size or correlation length [mm], snow and ice type [new snow, recrystallized snow, FYI, MYI], ice salinity [‰], and snow liquid water content [m3/m3]. 2. The sea ice emission model is the sea ice version of MEMLS, which is described in Tonboe et al. [2006]. The input to the model is the same as the output from the multi-layer snow and sea ice thermodynamic model (example profile is shown in Table 12.2). The output is: Tb, emissivity, emitting layer temperature. 3. The atmospheric and ocean emission models are modified versions of the Wentz and Meissner [2000] ocean emission model, which can also be used as an atmospheric model over sea ice with emissivity and emitting layer temperatures from the sea ice emission model as input. The input to the model is 2 m air temperature [K], 10 m wind

550

SEA ICE

Table 12.2 The initial MYI profile input to the emission model. Layer 4 is subdivided into 49 layers, each 0.05 m thick, to resolve the temperature profile.

Layer

Temperature [K]

Density [kg/m3]

Layer thickness [m]

Correlation length (exponential) [mm]

270 270 270 270

300 900 900 910

0.05 0.05 0.15 2.25

0.35 1.25 0.85 – 0.35 0.25

1 2 3 4

speed [m/s], atmospheric water vapor [kg/m2], atmospheric cloud liquid water [kg/m3]. Over sea ice the surface emissivity, e, and the emitting layer temperature, Teff, which are input to the model are computed by the sea ice emission model (see above). The output is Tb at the TOA. 4. The sea ice concentration algorithms use the simulated Tb as input to compute the sea ice concentration. Figure 12.3 shows the snow depth and ice thickness of the MYI profile as a function of the thermodynamical model timesteps of 6 hours. The profiles in the blue zones at the beginning and in the end of the simulation are excluded from further analysis. The first 120 simulations are excluded because the model needs to spin-up before realistic simulations for MYI can be reached. The blue zone at the end of the simulation is excluded because the temperature is reaching the melting point and snow and ice melting processes are not well represented in the thermodynamic model. Snow precipitation events occur mostly during the beginning (Fall) and end (Spring) of the simulation. However, there are also two larger mid-winter precipitation events. The ice grows by about

1

Air Snow

–1

250

Ice

–2 Water

–3 –4 0

200

400 600 800 Timestep [6h]

1000

1200

Figure 12.3 The simulated snow depth and MYI thickness as a function of model timestep of 6 hours using the thermodynamic model. The shaded zones at the beginning and the end of the simulation are not included in the analysis.

Type

0.00 0.50 1.00 2.50

old snow sea ice sea ice sea ice

0.66 m from 2.45 m at the beginning of the simulation to 3.11 m at the end of the simulation. Snow depth increases from 0.05 m to 0.37 m. The detailed snow and ice profiles from the thermodynamic model are input to the sea ice emission model for computing the sea ice self-emission (emissivity and emitting layer temperature). These are input to the atmospheric model together with water vapor and cloud liquid water and air temperature to compute the TOA Tb for sea ice. Over open water, the ocean and atmospheric models are used directly. The simulated TOA, Tb are compared with tie-points for open water, FYI, and MYI in Figure 12.4. The simulated brightness temperatures are within ±1 standard deviation of the tie-points, except for Tb from the MYI at 89 GHz where the simulated Tb are higher than the tie-points. This indicates that the simulated correlation length in the top of the snow cover is smaller than average for MYI. The tie-points are typical signatures of ice and open water [Ivanova et al., 2015]. Figure 12.5 shows the simulated brightness temperatures at 6.9, 18.7, and 36.5 GHz from MYI self-emission

Brightness temperature [K]

Depth [m]

0

Salinity [ppt]

FY

225 MY 200 TMY v

175 150

OW

TMY h TFY v

125

TFY h

100

TvOW

75

ThOW 20

40 60 Frequency [GHz]

80

Figure 12.4 The mean simulated Tb as a function of electromagnetic frequency are shown together with tie-points for FYI, MYI, and open water. The error bars are ±1 standard deviation of the simulated variability.

MODELING MICROWAVE EMISSION AND SCATTERING FROM SNOW-COVERED SEA ICE

Brightness temperature [K]

300

551

T TOA 7v

275

Surf T 7v

250

TOA T 19v

225

Surf T 19v TOA T 37v

200

Surf T 37v

175 150 125 100 0

200

400

600 800 Timestep [6h]

1000

1200

Figure 12.5 Simulated MYI Tb at vertical polarization as a function of model timestep of 6 hours for both surface TOA Surf

emission (Surf ) and when including the atmosphere (TOA) (T 7 19 37GHz ).

and at TOA. At 6.9 GHz there is almost no difference between the TOA and the surface self-emission. This difference is increasing to an average of 3 K at 18.7 GHz, where the TOA Tb is higher than the surface self-emission. The peak in Tb near timestep 500 is an air temperature warming event. Figure 12.6 shows the estimated cice using the simulated Tb shown in Figure 12.5. Ideally, these cice should be at SIC = 1 and the tie-points can be adjusted to achieve a mean SIC of 1. However, it is clear that the simulated SIC varies throughout the simulation as a function of the physical variables which are input to the emission models.

The simulated cice from the one channel 7 GHz algorithm as a function of snow surface and snow–ice interface temperatures are shown in Figure 12.7. The cice estimate is indeed sensitive to both snow surface and snow–ice interface physical temperature and lines have been fitted to the two clusters. The slope of the red line is 0.0036/K (snow surface temperature vs cice) and the slope of the blue line is 0.005/K (snow–ice interface temperature vs cice). The one channel algorithms and the spectral gradient algorithms are sensitive to snow and ice temperature, but this is not the case for algorithms using the polarization difference or polarization ratio [Tonboe et al., 2022]. Instead, the polarization algorithms are sensitive to snow

Ice concentration [1/100]

1.3 OneCh-7 (Surf) OneCh-7 (TOA) Bootsrap-F (Surf) Bootsrap-F (TOA) Bootsrap-P (Surf) Bootsrap-P (TOA)

1.2

1.1

1.0

0.9 0

200

400

600 800 Timestep [6h]

1000

1200

Figure 12.6 Simulated sea ice concentration for MYI as a function of a model timestep of 6 hours. Each of the three algorithms, one channel 7 GHz, Bootstrap-F, Bootstrap-P are shown for surface emission (Surf ) and with the atmosphere included (TOA).

Sea ice concentration (OneCh.7)[1/100]

552

SEA ICE 1.3

ice salinity, the simulated emission can be a function of ice thickness only, and it is then possible to invert the model to be able to compute the thin ice thickness in the Arctic as a function of L-band radiometer brightness temperatures [Kaleschke et al., 2012]:

Snow surf. temperature Snow ice interface temperature

1.2 1.1

T obs = T I − T I − T O exp − γd 1.0 0.9 0.8 200

220

260

240

280

300

Temperature [K]

Figure 12.7 The simulated cice for MYI as a function of snow surface (red +) and snow–ice interface (blue +) temperature for the one channel 7 GHz algorithm. The red and blue solid lines are the best linear fit to the clusters of points cice onechannel7GHz = 0 0036T s + 0 20 red and

Sea ice concentration (Bootstrap-P)[1/100]

cice onechannel7GHz = 0 0050T si − 0 20 blue

1.20

Snow surf. density

(12.12)

This type of inversion was implemented using SMOS data by Kaleschke et al. [2010, 2012] and later further developed by Tian-Kunze et al. [2014], and now these data are produced on a regular basis and combined with the sea ice thickness derived independently from the radar altimeter on CryoSat-2 [Ricker et al., 2017]. The inversion of sea ice emission models using statistical retrieval theory is an active research topic. So far, the sea ice forward models have been empirical [Pedersen, 1994, Kilic et al., 2019, 2020, Scarlat et al., 2020]. These empirical models relate the sea ice concentration, sea ice type, snow surface temperature, and ice thickness to satellite microwave radiometer Tb. For example, the empirical model relating sea ice concentration and first-year (FY) and multi-year (MY) ice types to the surface Tb is presented in Scarlat et al. [2020]: T B,surf = C OW ϵOW T OW + C FY ϵFY T FY + C MY ϵMY T MY (12.13)

1.18

1.16

1.14

1.12 50

100

150

200

250

300

350

Snow surf. density [kg/m3]

Figure 12.8 The sea ice concentration using the Bootstrap-P algorithm as a function of simulated snow surface density. The solid blue line is the best linear fit to the cluster of simulated snow surface densities [cice(Bootstrap − P) = 0.00024 SSD + 1.2].

surface density as shown in Figure 12.8. As the snow density increases, the sea ice concentration computed with the Bootstrap-P algorithm decreases by −0.0002/(kg/m3). 12.6. INVERSE MODELING When volume scattering in the sea ice can be ignored, which is the case in the microwave L-band at 1.4 GHz, then emission models can be simplified significantly. In the extreme case, and under assumptions about the sea

where, ϵFY and ϵMY are static emissivities for each polarization and frequency. The emissivities do not have to be static and can be included in the list of parameters to be retrieved if there are enough independent satellite observations (channels) [Pedersen, 1994]. There should be fewer physical parameters (m) to estimate, for example, temperature, snow depth, and ice thickness than the number of satellite channels (n), for example, Tb at 6 GHz, 10 GHz, etc. and different polarizations. Optimal estimation minimizes a cost function by iterating over a forward model to simulate the satellite observations (e.g., Tb at different frequencies and polarizations) and the linear tangent model (M) to compute the sensitivity of Tb to the physical parameters. M is an n x n matrix of partial derivatives derived from the forward model. In principle, it could be any of those models described in this chapter. Covariance matrices for the measured Tb (n × n measurement error covariance matrix Se) and the physical variables (m × m covariance matrix of physical parameters Sp) are used for constraining the next iteration and for quantifying the covariance of the different retrieved parameters. After convergence and the criteria for the cost function have been satisfied, the physical state vector (p) is considered to represent the set of physical parameters that results in sufficiently close results of the forward model to the satellite observations (Tb ).

MODELING MICROWAVE EMISSION AND SCATTERING FROM SNOW-COVERED SEA ICE First guess, p0

Measured TB

553

Sp and Se

Enter loop No, run another loop

Simulate TB: TB(sim)i = F(pi)

Linear tangent model, Mi

Compute S: Si = (Sp–1 +Mti Se–1Mi)–1

Compute next p: pi +1 = pi + Si (Mti Se–1(TB – TB(sim)i) + Sp–1(p0 – pi))

Convergence

Print p and the S diagonal

Yes, exit loop

Figure 12.9 Flow chart showing the optimal estimation process schematically.

The estimation of the physical state vector, pi +1, using the state vector from the former iteration, pi, is: pi + 1 = pi + S M ti S e− 1 T b− T b sim i + S p− 1 p0− pi (12.14) If the model is non-linear, then M must be evaluated at every iteration. Mt is the transposed M and S p− 1 and S e− 1 are the inverse covariance matrices. S is the retrieval uncertainty of S i = Tb(sim)s are then:

S p− 1

+

−1 M ti S e− 1 M i

T b sim i = F pi

The simulated

(12.15)

Optimal estimation is shown schematically in the flow chart in Figure 12.9. It was initially developed for retrieval of atmospheric temperature and composition from satellite Tb [Rogers, 1976] and has been applied for retrieving cice and associated error source parameters (water vapor in the atmosphere, sea-surface temperature, etc.) as well [Pedersen, 1994]. None of the major numerical weather prediction centers assimilate AMSU-A channels 1–5 (23.8, 31.4, 50.3, 52.8, 53.6 GHz) over land or sea ice. This leaves big holes in the data assimilated at high latitudes.

Even for channels 6–8 (54.4, 54.9, 55.5 GHz) the sea ice emissivity and the emitting layer temperature must be specified as 0.02 and 4 K, respectively, which is challenging and the development of new forward models for sea ice is therefore urgent. Optimal estimation can be used as a testbed for evaluating forward models before implementation in data assimilation. 12.7. REFERENCES Andersen, S. et al. (2006) Improved retrieval of sea ice total concentration from spaceborne passive microwave observations using numerical weather prediction model fields: An intercomparison of nine algorithms, Remote Sensing of Environment, 104(4), pp. 374–392. Barber, D.G. et al. (1998) The role of snow on microwave emission and scattering over first-year sea ice, IEEE Transactions on Geoscience and Remote Sensing, 36(5), pp. 1750–1763. Baunach, T. et al. (2001) A model for kinetic grain growth, Annals of Glaciology, 32, pp. 1–6. Bitz, C.M. and Lipscomb, W.H. (1999) An energy-conserving thermodynamic model of sea ice, Journal of Geophysical Research, 104(C7), pp. 15,669–15,677. Brown, G.S. (1982) A theory for near normal incidence microwave scattering from first-year sea ice, Radio Science, 17(1), pp. 233–243.

554

SEA ICE

Brun E. et al. (1989) An energy and mass model of snow cover suitable for operational avalanche forecasting, Journal of Glaciology, 35(121), pp. 333–342. Brun, E. et al. (1992) A numerical model to simulate snow-cover stratigraphy for operational avalanche forecasting, Journal of Glaciology, 38(128), pp. 13–22. Carlström, A. (1997) A microwave backscattering model for deformed first-year ice and comparisons with SAR data, IEEE Transactions on Geoscience and Remote Sensing, 35 (2), pp. 378–391. Carlström, A. and Ulander, L.M.H. (1995) Validation of backscatter models for level and deformed sea-ice in ERS-1 SAR images, International Journal of Remote Sensing, 16(17), pp. 3245–3266. Denoth, A. (1980) The pendular–funicular liquid transition in snow, Journal of Glaciology, 25(91), pp. 93–97. Dierking, W. (2000) RMS slope of exponentially correlated surface roughness for radar applications, 38(3), pp. 1451–1454. Available from: doi: 10.1109/36.843040. Dierking, W., Petterson, M.I. and Askne, J. (1999) Multifrequency scatterometer measurements of Baltic sea ice during EMAC-95, International Journal of Remote Sensing, 20(2), pp. 349–372. Drinkwater, M.R. (1989) LIMEX ‘87 Ice surface characteristics: Implications for C-band SAR backscatter signatures, IEEE Transactions on Geoscience and Remote Sensing, 27(5), pp. 501–513. ECMWF2018. Available from: https://www.ecmwf.int/sites/ default/files/elibrary/2018/18157-observations-and-analysissea-surface-temperature-and-sea-ice-nwp-and-climate-applications-ws.pdf (accessed July 2022). Elachi, C. and van Zyl, J. (2006) Introduction to the physics and techniques of remote sensing, John Wiley and Sons., Inc. English, S., Lean, P. and Geer, A. (2020) How radiative transfer models can support the future needs of Earth-system forecasting and reanalysis, Journal of Quantitative Spectroscopy and Radiative Transfer, 251, p. 10,7044. Available from: https:// doi.org/10.1016/j.jqsrt.2020.107044. Fetterer, F.M. et al. (1992) Sea ice altimetry, In: Carsey, F.D., ed. Microwave remote sensing of sea ice, Geophysical Monograph, Vol. 68, Washington D.C.: American Geophysical Union, pp. 111–135. Fuhrhop, R. et al. (1997) Study of passive remote sensing of the atmosphere and surface ice, Berichte aus dem Institut für Meereskunde an der Christian-Albrechts Universität, Kiel 297, pp. 241. Fung, A.K. (1994) Microwave scattering and emission models and their applications, Norwood, MA: Artec House, 592 p. Fung, A.K. and Chen, K.S. (2004) An update on the IEM surface backscattering model, IEEE Geoscience and Remote Sensing Letters, 1(2), pp. 75–77. Fung, A.K. and Chen, K.S. (2010) Microwave scattering and emission models for users, Boston, London: Artec House, ISBN-13: 978-1-60807-037-4. Fung, A.K. and Eom, H.J. (1982) Application of a combined rough surface and volume scattering theory to sea ice and snow backscatter, IEEE Transactions on Geoscience and Remote Sensing, 20(4), pp. 528–536.

Fung, A.K. and Eom, H.J. (1985) A study of backscattering and emission from closely packed inhomogeneous media, IEEE Transactions on Geoscience and Remote Sensing, GE-23(5), pp. 761–767. Ivanova, N. et al. (2015) Inter-comparison and evaluation of sea ice algorithms: Towards further identification of challenges and optimal approach using passive microwave observations, The Cryosphere, 9, pp. 1797–1817. Available from: https://doi. org/10.5194/tc-9-1797-2015. Johnsen, K.P. and Heygster, G. (2000) Interference effects in freshwater and sea ice In: Mätzler, C., ed., COST Action 712, Radiative transfer models for microwave radiometry, Final Report. European Commission EUR 19543 EN, pp. 149–162. Jordan, R. (1991) A one-dimensional temperature model for a snow cover, Technical documentation for SNTHERM.89, SP 91-16, CRREL. Available from: (https://erdc-library. erdc.dren.mil/jspui/bitstream/11681/11677/1/SR-91-16.pdf). Jordan, R.E., Andreas, E.L. and Makshtas, A.P. (1999) Heat budget of snow-covered sea ice at North Pole 4, Journal of Geophysical Research, 104(C4), pp. 7785–7806. Kaleschke, L. et al. (2010) A sea-ice thickness retrieval model for 1.4 GHz radiometry and application to airborne measurements over low salinity sea-ice, The Cryosphere, 4, pp. 583– 592. Available from: https://doi.org/10.5194/tc-4-583-2010. Kaleschke, L. et al. (2012) Sea ice thickness retrieval from SMOS brightness temperatures during the Arctic freeze-up period, Geophysical Research Letters, 39(5). Available from: doi: 10.1029/2012GL050916. Kang, E.J. et al. (2021) Implementation of a 1-D thermodynamic model for simulating the winter-time evolvement of physical properties of snow and ice over the Arctic Ocean, Journal of Advances in Modeling Earth Systems, 13, e2020MS002448. Kilic, L. et al. (2019) Estimating the snow depth, the snow–ice interface temperature, and the effective temperature of Arctic sea ice using Advanced Microwave Scanning Radiometer 2 and ice mass balance buoy data, The Cryosphere, 13, pp. 1283–1296. Available from: https://doi.org/10.5194/tc-131283-2019. Kilic, L. et al. (2020) Ice concentration retrieval from the analysis of microwaves: A new methodology designed for the Copernicus imaging microwave radiometer, Remote Sensing 12, 1060. Available from: https://doi.org/10.3390/ rs12071060. Landy, J.C., Tsamados, M. and Scharien, R.K. (2018) A facetbased numerical model for simulating SAR altimeter echoes from heterogeneous sea ice surfaces’, IEEE Transactions on Geoscience and Remote Sensing, 57(7), pp. 4164–4180. Landy, J.C. et al. (2020) Sea ice roughness as a key source of uncertainty in CrySat-2 ice freeboard retrievals, Journal of Geophysical Research: Oceans, 125, e2019jc015820. Lavergne, T. et al. (2019) Version 2 of the EUMETSAT OSI SAF and ESA CCI sea-ice concentration climate data records, The Cryosphere, 13, pp. 49–78. Available from: https://doi. org/10.5194/tc-13-49-2019. Laxon, S., Peacock, H. and Smidth, D. (2003) High interannual variability of sea ice thickness in the Arctic region, Nature, 425 (6961), pp. 947–950.

MODELING MICROWAVE EMISSION AND SCATTERING FROM SNOW-COVERED SEA ICE Laxon S.W. et al. (2013) CryoSat-2 estimates of Arctic sea ice thickness and volume, Geophysical Research Letters, 40, pp. 732–737. Available from: doi:10.1002/grl.50193. Löwe, H. and Picard, G. (2015) Microwave scattering coefficient of snow in MEMLS and DMRT-ML revisited: The relevance of sticky hard spheres and tomography-based estimates of stickiness, The Cryosphere, 9, pp. 2101–2117. Available from: https://doi.org/10.5194/tc-9-2101-2015. Manninen, A.T. (1992) Effects of ice ridge properties on calculated surface backscattering in BEPERS-88, International Journal of Remote Sensing, 13(13), pp. 2469–2487. Available from: doi: 10.1080/0143116920890482. Marbouty, D. (1980) An experimental study of temperature gradient metamorphism, Journal of Glaciology, 26(94), pp. 303–312. Miernecki, M. et al. (2020) Effects of decimetre-scale surface roughness on L-band brightness temperature of sea ice, The Cryosphere, 14, pp. 461–476. Available from: https://doi. org/10.5194/tc-14-461-2020. Mätzler, C. (1987) Applications of the interaction of microwaves with the natural snow cover, Remote Sensing Reviews, 2(2), pp. 259–391. Mätzler, C. (1998) Improved Born approximation for scattering of radiation in a granular medium, Journal of Applied Physics, 83(11), pp. 6111–6117. Mätzler, C. (2002) Relation between grain-size and correlation length of snow, Journal of Glaciology, 48(162), pp. 461–466. Mätzler C. and Wiesmann, A. (2014) Microwave emission model of layered snowpacks, Documentation for MEMLS, Version 3, Institute of Applied Physics–University of Bern Sidlerstrasse, 5, 3012 Bern, Switzerland, 26 pp. Nghiem, S.V. et al. (1995a) Polarimetric signatures of sea ice 1. Theoretical model, Journal of Geophysical Research, 100(C7), pp. 13,665–13,679. Nghiem, S.V. et al. (1995b) Polarimetric signatures of sea ice, 2, Experimental observations, Journal of Geophysical Research, 100(C7), pp. 13,681–13,698. Paul, S. et al. (2018) Empirical parametrization of Envisat freeboard retrieval of Arctic and Antarctic sea ice based on CryoSat-2: progress in the ESA climate change initiative, The Cryosphere, 12, pp. 2437–2460. Available from: https://doi. org/10.5194/tc-12-2437-2018. Pedersen, L.T. (1994) Merging microwave radiometer data and meteorological data for improved sea ice concentrations, EARSeL Advances in Remote Sensing, 3(2-XII), pp. 81–89. Picard, G., Sandells, M. and Löwe, H. (2018) SMRT: An active– passive microwave radiative transfer model for snow with multiple microstructure and scattering formulations, (v1.0), Geoscience Model Development, 11, pp. 2763–2788. Available from: https://doi.org/10.5194/gmd-11-2763-2018. Proksch, M., Löwe, H. and Schneebeli, M. (2015) Density, specific surface area, and correlation length of snow measured by high-resolution penetrometry, Journal of Geophysical Research: Earth Surface, 120, pp. 346–362. Available from: doi:10.1002/2014JF003266. Proksch, M. et al. (2015) MEMLS3&a: Microwave emission model of layered snowpacks adapted to include backscattering, Geoscience Model Development, 8, pp. 2611–2626. Available from: https://doi.org/10.5194/gmd-8-2611-2015.

555

Ricker, R. et al. (2017) A weekly Arctic sea-ice thickness data record from merged CryoSat-2 and SMOS satellite data, The Cryosphere, 11, pp. 1607–1623. Available from: https:// doi.org/10.5194/tc-11-1607-2017. Rogers, C.D. (1976) Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation, Reviews of Geophysics and Space Physics, 14(4), pp. 609–624. Rostosky, P. et al. (2020) Modeling the microwave emission of snow on Arctic sea ice for estimating the uncertainty of satellite retrievals, Journal of Geophysical Research: Oceans, 125, e2019JC015465. Available from: https://doi.org/10.1029/ 2019JC015465. Rothrock, D.A. (1986). The thickness distribution–Measurement and theory (p. 551–575), In: Untersteiner, N. ed. The geophysics of sea ice, NATO ASI Series, Series B: Physics, Vol. 146. Scarlat, R.C. et al. (2020) Sea ice and atmospheric parameter retrieval from satellite microwave radiometers: Synergy of AMSR2 and SMOS compared with the CIMR candidate mission, Journal of Geophysical Research: Oceans, 125, e2019JC015749. Stogryn, A. (1987) An analysis of the tensor dielectric constant of sea ice at microwave frequencies, IEEE Transactions on Geoscience and Remote Sensing, GE-25(2), pp. 147–158. Stroeve, J.C. et al. (2006) Impact of surface roughness on AMSR-E sea ice products, IEEE Transactions on Geoscience and Remote Sensing, 44, pp. 3103–3116. Tian-Kunze, X. et al. (2014) SMOS-derived thin sea ice thickness: Algorithm baseline, product specifications and initial verification, The Cryosphere, 8, pp. 997–1018. Available from: https://doi.org/10.5194/tc-8-997-2014. Tonboe. R.T. (2010) The simulated sea ice thermal microwave emission at window and sounding frequencies, Tellus, Series A: Dynamic Meteorology and Oceanography, 62(3), pp. 333–344. Tonboe, R.T., Dybkjær, G. and Høyer, J.L. (2011) Simulations of the snow covered sea ice surface temperature and microwave effective temperature, Tellus A: Dynamic Meteorology and Oceanography, 63(5), pp. 1028–1037. Available from: doi:10.1111/j.1600-0870.2011.00530.x. Tonboe, R.T. et al. (2006) Sea ice emission modelling, In: Mätzler, C. et al., eds. Thermal microwave radiation–Applications for remote sensing, IET Electromagnetic Waves Series 52, London, UK. Tonboe, R.T. et al. (2016) The EUMETSAT sea ice concentration climate data record, The Cryosphere, 10, pp. 2275–2290. Available from: https://doi.org/10.5194/tc-10-2275-2016. Tonboe, R.T. et al. (2021) Simulated Ka- and Ku-band radar altimeter height and freeboard estimation on snow-covered Arctic sea ice, The Cryosphere, 15, pp. 1811–1822. Available from: https://doi.org/10.5194/tc-15-1811-2021. Tonboe, R.T. et al. (2022) Simulated geophysical noise in sea ice concentration estimates of open water and snow-covered sea ice, IEEE Journal of Selected Topics in Applied Earth Observation and Remote Sensing, 15, pp. 1309–1326. Available from: doi: 10.1109/JSTARS.2021.3134021. Tsang, L. and Ishimaru, A. (1987) Radiative wave equations for vector electromagnetic propagation in dense non-tenuous media, Journal of Electromagnetic Waves and Applications, 1(1), pp. 59–72.

556

SEA ICE

Ulaby, F.T., Moore, R.K. and Fung, A.K. (1981) Microwave remote sensing—Active and passive, Microwave Remote Sensing Fundamentals and Radiometry, Vol. I, Reading, Massachusetts: Addison-Wesley. Ulaby, F.T. et al. (2014) Microwave radar and radiometric remote sensing, Ann Arbor: University of Michigan Press. Ulander, L.M.H. and Carlström, A. (1991) Radar backscatter signatures of Baltic sea ice, Proceedings of the IGARSS’91, Vol. 3, Helsinki, Finland, pp. 1215–1218. Vant, M.R. (1976) A combined empirical and theoretical study of the dielectric properties of sea ice over the frequency range 100 MHz to 40 GHz, Technical Report,

Carleton University and Environment Canada, 1976, p. 476. Wentz, F.J. and Meissner, T. (2000) AMSR Ocean Algorithm, Version 2, Report number 121599A-1, Remote Sensing Systems, Santa Rosa, CA, 66 pp. Wiesmann, A. and Mätzler, C. (1999) Microwave emission model of layered snow packs, Remote Sensing of Environment, 70, pp. 307–316. Winebrenner, D.P. et al. (1992). Microwave sea ice signature modelling, In: Carsey, F.D., ed. Microwave remote sensing of sea ice, Geophysical Monograph, Vol. 68, Washington D.C.: American Geophysical Union, pp. 137–175.

13 Impacts of Climate Change on Polar Ice

13.1

The Inconvenient Truth of Global Warming: How is it Manifested in The Polar Region? ......................................... 560

13.4.1.1

13.2

Sea Ice Regimes in the Two Polar Regions .......................... 562 13.2.1 Geographic Differences Between the Two Polar Regions and Their Impacts on Sea Ice... 562 13.2.2 Differences in Sea Ice Characteristics Between the Two Polar Regions.............................. 564 Changes of Polar Sea Ice in Response to Global Warming.. 565 13.3.1 The Arctic and Antarctic Ice Extent........................ 565 13.3.2 The Arctic and Antarctic Ice Thickness and Volume 569 13.3.3 The Arctic Sea Ice Age ............................................ 572 13.3.4 The Arctic Sea Ice Dynamics .................................. 575 13.3.5 The Antarctic Icebergs............................................. 576

13.4.1.2

13.3

13.4

Coupling Between Polar Sea Ice and Environmental Factors.................................................................................. 577 13.4.1 Interaction of the Arctic Sea Ice with the Environment .............................................. 578

By definition, climate change refers to the long-term variation of climate and weather patterns. It may happen due to natural causes (e.g., changes in the sun’s activities, variations in the earth’s orbits, emissions from volcanoes), or anthropogenic causes (e.g., burning fossil fuels, destroying forests, emission of greenhouse gases). The recent episode of global warming is attributed to the bloom of the industrial revolution in the nineteenth century. Specifically speaking, it has been caused by the increasing emission of carbon dioxide and other “greenhouse” gases (e.g., methane, nitrous oxide, and hydrofluorocarbons). These gases are especially effective at absorbing the outgoing longwave radiation emitted by the earth. Therefore, an increase of their concentration in the atmosphere causes warming of the planet’s surface while cooling the upper atmosphere. From 1990 to 2019, the total warming effect from greenhouse gases emission increased by 45%, of which the warming associated with carbon dioxide alone increased by 36% (United States Environmental

13.4.1.3 13.4.1.4 13.4.1.5

13.5

Atmospheric Factors that Contribute to Changes in the Arctic Sea Ice .................. 578 Enhanced Arctic Warming due to Changes of Sea Ice Cover ........................ 578 Arctic Warming due to Sea Ice Advection Out of the Arctic Basin .......... 580 Interaction of the Arctic Sea Ice with Wind... 582 Mutual Interactions Between the Arctic Sea Ice Cover and Oceanic Forcing......... 584

13.4.2 Interaction of the Antarctic Sea Ice with the Environment .............................................. 585 13.4.2.1 Interaction of the Antarctic Sea ice with Atmospheric Factors ............................... 585 13.4.2.2 Interaction of the Antarctic Sea Ice with Oceanic Forcing....................................... 586 13.4.2.3 Interaction Between the Antarctic Sea Ice, Ice Shelves, and Icebergs................... 587 References............................................................................. 590

Protection Agency, Climate Indicators web site). This has enhanced the climate forcing, a term that refers to a change in the earth’s energy balance, leading to either warming or cooling. Although the term climate change has been publicly circulated and become a political issue in the past few decades, the phenomenon is not strictly connected to our age. Roughly 18,000 years ago, during a peak period of the last ice age, much of the earth’s land was covered with ice before a natural change of climate melted the ice and exposed the landscape we see today. During that time there were no apparent seas or rivers as the entire landscape was covered with ice. Most of Canada was covered by ice caps, which were essentially joined together forming a huge ice sheet called Laurentide (Figure 13.1). Valleys in the western mountains in the present-day province of British Columbia and the Yukon Territories that include the Canadian Rocky Mountain (part of the Rockies in the USA) were full of ice forming the Cordilleran ice sheet. The maximum ice thickness of the Laurentide ice

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 557

558

SEA ICE (a)

(b) ARCTIC

GREENLAND

GR

EE

NL

AN

T

EE SH

A LER DIL COR

E IC

D

IES

HEE

R OC K

ES

N IC

LAURENTIDE ICE SHEET

T

18,000 YEARS AGO

PRESENT DAY

Figure 13.1 Schematic depiction of upper half of North America (a) 18,000 years ago; and (b) today (sketch by Nirmal K. Sinha).

sheet was probably around 4 km; that of the Cordilleran ice sheet may have been close to 2 km. Greenland was covered with another ice sheet that carries its name, and remains covered till today. Figure 13.1 shows the map of Canada and the northern part of what is known today as the United States 18,000 years ago, compared with the same map in recent time after the ice retreated. Most of the ice from the landscape of the North America, including Canada of today, was gone by 5,000 years ago, except for a few ice caps. The Arctic Ocean, seas, lakes, and rivers became exposed, and sea ice and fresh water ice became known to the few inhabitants. Greenland, however, remained to be under ice. This situation remained effectively unchanged until the bloom of the industrial age as pointed out earlier. Global warming started to make impacts on many aspects of life on the earth, particularly the polar regions. The remaining ice in those regions continues to melt, raising sea level and developing the river system in North America. Together, the Antarctic and Greenland ice sheets contain more than 99% of the freshwater ice on Earth. If the Greenland ice sheet were to melt or move into the ocean, global sea level would rise by 6.5 meters. If same happens to the west Antarctic ice sheet it would rise by approximately 8 meters. The Antarctic glaciers have lost large percentages of their areas and volumes since the end of the Little Ice Age (1300–1860). If all the ice covering Antarctica, Greenland, and mountain glaciers around the world were to melt, sea level would rise by about 70 meters. Then, low-elevation land areas such as the state of West Bengal of India and Bangladesh, and most coastal cities and island countries of the world will be under water, as depicted in Figure 13.2. Climate change is a term coined with the current trend of global warming. Scientifically speaking, climate change is a more comprehensive term, which incorporates several physical, biophysical, social, and political impacts of

Figure 13.2 Conceptualized image about sea level if it rises as a result of melting of all glaciers and ice caps in the world (by Nirmal K. Sinha).

global warming. These impacts include boosting water vapor in the atmosphere (which lead to more frequent heavy rain and hurricanes), increasing chances of hot days (which lead to more frequent droughts), rising sea water level (which leads to destructive flooding), reducing the ice cover in the polar regions, among several other aspects [Letcher, 2021]. In the following discussions, the terms climate change and global warming are used interchangeably to address their impact on polar ice. Some politicians prefer the use of the term climate change or “climate consequences” over the term global warming in order to avert discussing the troubling issue of anthropogenic causes of global warming, hence ease

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

the impact on the public sphere. Some express skepticism over global warming, claiming that its mitigation would hurt the economy. Climate change sounds more controllable and less emotional. Thus, the term has been widely used in the media, political circles, scientific communities, and even incorporated in the name of the international authority on the subject, namely the Intergovernmental Panel on Climate Change (IPCC). This organization was established in 1989, under the auspices of the United Nations to provide a scientific view of climate change and its political and economic impacts. While climate change is also incorporated in the title of this chapter, global warming is used more in the following text because it fits better the root cause of the presented phenomenon, namely changes of sea ice cover in the polar regions. While addressing the impacts of global warming on the two polar regions, it is important to keep in mind that the Arctic is an extensive territory of sea ice. On the other hand, the Antarctic is a massive ice cap that terminates with ice shelves, while the sea ice, mixed with floating icebergs, is located around the peripheral of the continent. With this theme in mind, it is conceivable to envision the impact of climate change on the Arctic sea ice to be manifested in shrinking and thinning of the sea ice cover and replacing older (perennial) ice with seasonal ice. On the other hand, in the Antarctic region, climate change impacts first and foremost the ice sheet and ice shelves from which icebergs calve off their edges. This is manifested in the recent increase of calved icebergs, leaving their noticeable impacts on sea ice. The above themes are central to the presentation in this chapter. They can be rephrased in the following statements. The study of icebergs in the Antarctic region is as relevant to climate change as the study of the sea ice in the Arctic region. As the reduction of sea ice in the Arctic (extent and thickness) has become a research priority recently, the priority for the Antarctic should include the impacts of the increasing number and dynamics of icebergs on sea ice. An obvious impact is the release of freshwater as a result of basal and side melting of icebergs. This affects sea ice formation and its subsequent properties. Each polar region has its own peculiarities (geographyand climate-wise). Therefore, researchers should not try to “force” conclusions about impacts of climate change on the Arctic sea ice onto the Antarctic sea ice. Unlike the former, the latter has not thinned or shrunk simply because it is exposed to very different geographic, climatic, and environmental conditions, which offset the action of global warming. The material in this chapter revolves around this theme. Unfortunately, data about the Antarctic sea ice are limited compared to data about the Arctic sea ice (section 13.3). This has made knowledge about the Antarctic sea ice incomplete and inconclusive so far. It is appropriate to emphasize here that the sea ice in the Arctic region is far more sensitive to climate change

559

than the ice in the Antarctic. Its response is also more important partly because most of the world’s population inhabit the northern hemisphere. Guarino et al. [2020] found that the Arctic temperature during the last interglacial period (130,000–116,000 years ago) was 4 C – 5 C higher than the pre-industrial era. The authors showed the sensitivity of the Arctic sea ice to climate change using the UK Hadley Centre climate model (HadGEM3) with incorporation of a comprehensive sea ice melt pond scheme. They predicted that the Arctic will be ice-free by 2035. This is indeed an alarming news, though it should be checked as time progresses. Johannessen et al. [2020] presented a review on the Arctic sea ice: past, present, and future. NOAA publishes online series of “Arctic Report Card” [NOAA, 2021]. Massom and Stammerjohn [2010] presented a summary of the Antarctic physical and ecological implications of the recent changes of sea ice. With regard to sea ice in interactive global climate models, the formidable problem is the extreme sensitivity of sea ice to thermodynamic forcing by the atmosphere and the ocean. To overcome this problem, climate models usually tune a few parameters (e.g., sea ice albedo), to achieve reasonable results of ice thickness and extent. Obviously, this is not a reasonable approach. As this subject is outside the scope of this book, the reader may refer to publications of climate models that include sea ice components [Langehaug et al., 2013, Stroeve et al., 2014, Hunke et al., 2015, Mackie et al., 2020, Blockley et al., 2020]. The chapter starts by addressing key differences between sea ice in the two polar regions in terms of types, duration, advection, and physical properties (section 13.2). These differences regulate the differences in response to climate change. Recent decadal changes in sea ice (in terms of ice extent, thickness, volume, age, dynamics, and presence of icebergs) are addressed in section 13.3, with more focus on the Arctic region where more significant changes have been observed. Reasons for the observed change are then presented in section 13.4 for the Arctic and Antarctic separately, with emphasis on the peculiarities of each region. The coupling between the polar sea ice and the environment is summarized in order to identify the role of each environmental forcing on the observed changes. The coupling between the Arctic ice cover and the atmosphere is much stronger because the ice is effectively entrapped within the enclosed Arctic Basin under much colder winter temperatures. Conversely, the Antarctic sea ice is not exposed to such cold temperatures because it is formed in a temperate region, far from the polar core. As the ice becomes free to drift away into the warmer region in the Southern Ocean, it nearly completely melts in summer. Only limited areas of perennial ice remain in the Weddell Sea and Ross Sea as explained in detail later.

560

SEA ICE

13.1. THE INCONVENIENT TRUTH OF GLOBAL WARMING: HOW IS IT MANIFESTED IN THE POLAR REGION?

Diff. from avrg. temperature (°C)

In 2006, 5 years after the IPCC issued its third report on climate change confirming the presence of global warming and its highly damaging future impacts, the documentary film An Inconvenient Truth was released. It featured the former vice president of the United States, Al Gore, presenting a series of slides showing evidence that the increase in global temperature over the last 100 years was caused by the increasing emission of greenhouse gases. The documentary ended with Gore calling for taking appropriate actions to cut down on the release of greenhouse gases (particularly CO2). This could be achieved by planting more vegetation, minimizing consumption of fossil fuels, and switching to renewable energy sources. In 2007, Al Gore and IPCC were awarded the Nobel Peace Prize for their efforts to raise the public awareness about the anthropogenic causes of the current global warming. The first notable minimum record of the Arctic sea ice extent that caught the public’s attention (and to some degree, scientists’ attention) was set in September 2007 (4.16 × 106 km2). The record of polar sea ice extent has been established, since 1979, when satellite data became available. The message in Gore’s documentary has been reasserted through numerous research articles and subsequent IPPC reports. It is clear and simple: recent global warming is a fact and it is a human-induced phenomenon. With this in mind, the near-surface air temperature (nSAT) records over the polar regions (i.e., temperature measured at a height of 1.5 to 2 meters above land, ocean, or ice cover) should be examined first. Then, the following two questions must be addressed. Do the records reveal a trend of temperature increase? And if so, how is it compared to the global temperature increase? Obviously, both polar regions are colder than the rest of the globe because they receive much less of sunlight. Historical records of nSAT across the entire Arctic and Antarctic regions, provide answers to the above questions.

The record of the Arctic data (Figure 13.3) is more relevant to the change of sea ice because the temperature is spatially averaged over the Arctic Ocean, where the ice cover extends. Figure 13.4 shows the annual mean temperature anomalies over the two polar regions as well as the entire globe. For the Arctic, no warming trend is abserved before 1980 and a strong trend (about 0.8 C/decade) is observed after 2000. On the other hand, the record of the Antarctic data (Figure 13.4) follows nearly the same trend of the global temperature change. However, it represents temperature integrated over the extensive ice cap rather than the peripheral of the continent where sea ice is found. This means that the record may not be of direct use to explain the scenario of sea ice change because the temperature at the marginal areas of the Antarctic is remarkably warmer than the average across the continent. To put it differently, comparing the Arctic to the global nSAT would reveal information on the impact of global warming on the Arctic, but this is not likely true for the Antarctic because the sea ice is not precisely located in the polar area. Figure 13.3, adapted from Ballinger et al. [2021], shows integrated nSAT across the Arctic region (north of 60 N) obtained from many weather stations located on the Arctic lands. Global temperature is also shown. The data were compiled by the Met Office and the University of East Angelia, United Kingdom. Prior to 1980, the record does not show an identifiable trend; where the Arctic temperature can be warmer or colder than the rest of the world. However, more fluctuations of the Arctic temperature are observed compared to global temperature. Since 1980, the global and Arctic temperature records exhibit an increasing trend at nearly the same rate until 2000. Starting 2000, the rate of warming in the Arctic has been ~ 0.75 C/decade, nearly 2.5–4.0 times steeper than the global rate. This is known as the Arctic amplification (AA) phenomenon, which points to positive feedback within the ice/ocean/atmosphere system. Serreze and Barry [2011] present a suite of intertwined causes for the AA. It includes changes in sea ice extent, concentration, thickness, mobility, which affect the atmospheric oceanic heat transfer. The strong warming is also

2

Arctic Global

1 0

1981–2010 average

–1 –2 1900

1920

1940

1960 Year

1980

2000

2020

Figure 13.3 Anomaly of the Arctic and global nSAT relative to 1981–2010 average, covering 22 years (1900–2021) [adapted from Ballinger et al., 2021].

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

Annual surface temperature anomaly (°C)

4 3

561

Arctic Antarctic Global

2 1 0 –1 –2 –3 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 Year

Figure 13.4 Anomalies of near-surface air temperature of the Arctic, Antarctic regions and the entire globe between 1880 and 2018 relative to the average temperature during the period 1951–1980 (credit NASA/GISS based on Global Historical Climatology Network V3 data).

attributable to the cloud cover and atmospheric water vapor, both alter the longwave radiation. Less ice concentration means more openings in the ice cover, allowing more absorption of solar energy by the exposed sea water, hence raising the local air temperature. More on the role of the AA on sea ice retreat and thinning is presented in section 13.4.1. It is interesting to note the start of the sharp trend of warmer Arctic air temperature in the early 2000s (Figure 13.3), which coincide with the first notable lowest record of summer ice extent observed in 2007 (section 1.4.2). It has since become clear that this temperature increase was not an anomaly. But the compelling question becomes how has it impacted the ice extent and thickness in a way that triggers its amplification? This question is addressed in the sections 13.3 and 13.4. By the end of the twenty-first century, a continuous global warming of 2 C above pre-industrial levels would cause the Arctic Ocean to be ice-free in the summer, once in every 3 years [IPCC, 2019]. This will have significant biophysical, economic, and climatic impacts on the northern hemisphere. Parallel prediction for the Antarctic and Southern Hemisphere are yet to be reported. The Antarctic may not respond to the global warming in the same way as the Arctic region because it is covered by a huge ice sheet based on land. Therefore, there is no source of heat release to the atmosphere similar to the situation in the Arctic where ocean under the ice releases heat that amplifies the warming trend. Records of nSAT in the Antarctic are sparse, brief, and have been controversial. They are obtained from some 30 permanent weather stations (only two are in the interior of the continent), a number between 100 and 200 automated weather stations (AWS), and more recently satellite observations. Most of the atmospheric parameters about Antarctica

and surrounding sea ice region are derived from global atmospheric reanalysis sources, which start in 1979 when comprehensive satellite observations began. The controversy revolves around the question of whether the Antarctic region is warming (as a few climate models predict) or cooling. Figure 13.4 is a historical record comparing the nSAT across the Arctic and Antarctic. Data are generated by NASA’s Goddard Institute for Space Studies (GISS) based on updated Global Historical Climatology Network V4. Unlike the Arctic, the Antarctic has largely no clear trend of warming. This observation was used in the past to refute the argument of climate change but not pursued anymore. Figure 13.4 is useful in highlighting the difference in temperature trends between the two polar regions in general but fails short of providing the needed details about the nSAT records at specific locations within the massive area of the Antarctic (14.2 × 106 km2), and more importantly over sea ice. In other words, it does not provide information on the interannual or decadal changes of the temperature at local regional scales. Other data sets provide more information on the Antarctic temperature record. The Australian Arctic Program estimated the mean annual temperature of the interior of Antarctica to be around −57 C, but the average temperature in some coastal areas can reach −10 C in the warmest parts (https://www.antarctica.gov.au). Certain peripheral areas have demonstrated warming trends, at least for certain periods. For example, the mean nSAT of the Antarctic Peninsula (AP) has increased by nearly 3 C in the past 50 years [Turner et al., 2005]. This is comparable to the warming scenario of the Arctic. It is interesting to note the geographic similarity of the AP to the Arctic in that both feature enclosed sea ice regime. Carrasco, Bozkurt, Cordero [2021] analyzed nSAT data

562

SEA ICE

from eight Antarctic stations in the AP and showed that warming trends were identified, though they ended in some locations. Other coastal areas of the Antarctic have shown indications of rising nSAT, though at a milder rate. In a study by Comiso [2000], the author examined records of nSAT from 21 weather stations, and surface temperature from satellite infrared sensors during the period 1979–1998. Warming trends were found at most of the coastal locations in the west Antarctic, while cooling trends were found mostly in the east Antarctic sector and the locations of the two weather stations in the interior of the continent. The rate or warming or cooling ranges between 0.01 C and 0.07 C per year, with the AP warming at a rate of 0.05 C. Steig et al. [2009] presents an interesting map of the spatial variation of the decadal rate of change of the Antarctic air surface temperature from 1957 through 2006. It was generated using data obtained from weather stations and satellite observations. The map (Figure 13.5) shows temperature increase of about 0.20–0.25 C/decade in the West Antarctica in sharp contrast to the much less warming trend in nearly all parts of the East Antarctica (0.05–0.10 C/decade). It is interesting to note that the higher rate of warming in the West Antarctica occurred at southern latitudes nearer to the pole. The map shows virtually no temperature variation over the sea ice that surrounds the continent. More icebergs are calved from ice shelves in the West Antarctica. An update of the map in Figure 13.5 is recommended. It is not appropriate to draw conclusions about temperatures in the Antarctic region based on local measurements obtained at specific times. For example, on 18 March 2022 (beginning of the austral fall), a few weather stations in Antarctica registered a record high

temperature. Concordia station, an Italian-German station located at (75.1S,123.4E), registered −12.2 C, which is about 21 C warmer than the average. Vostok station, a Russian station located at (78.5S,106.9E), recorded −17.7 C, which is above average by 15 C. The temperature from the coastal Terra Nova Base, an Italian station located at (74.8S,164,5E), was far above freezing (7 C). It is possible that these extreme warm temperatures are just random variability with no connection to global warming [W. Meier, personal communication]. More spatiotemporal data are needed to ascertain any possible trend. Air temperature data across the Antarctic and the surrounding ice should be updated regularly using multiple sources and with interpolation. For the purpose of studying sea ice in the Antarctic it would be appropriate to examine a record of surface air temperature reanalysis over the entire ice cover. Bozkurt et al. [2020] studied recent trends of nSAT in key regions of the Antarctic sea ice using observations and reanalysis data. For the question posed in the title of this section: How is global warming manifested in the temperature of polar regions?, the direct answer is that it is strongly felt in the Arctic region, triggering what is known as the AA scenario. This phenomenon is driven by interactions between the ice cover and key environmental factors (section 13.4.1). On the other hand, global warming is not manifested over the Antarctic sea ice, which is mainly located at the border between the polar and temperate regions (except for the Weddell and Ross seas). The rest of the chapter elaborates on this answer. 13.2. SEA ICE REGIMES IN THE TWO POLAR REGIONS Differences in the geography of the two polar regions and their sea ice characteristics are presented. They provide clues to understand the different impacts of global warming on the sea ice in both regions. 13.2.1. Geographic Differences Between the Two Polar Regions and Their Impacts on Sea Ice

0.0

0.05

0.10

0.15

0.20

0.25

Temperature change per decade (°C)

Figure 13.5 Map of air surface temperature trend across Antarctica from data obtained during 1957–2006 from weather stations and AVHRR data (image is in public domain; courtesy of Trent Schindler, NASA Goddard Space Flight Center Scientific Visualization Studio).

Generally speaking, the Arctic is a geographic region surrounding the North Pole. Yet, it has a few definitions depending on the particular aspect of interest, e.g., land cover, ice cover, treeline, biomass, temperature, permafrost, and homelands of northern indigenous people. If defined in terms of solar radiation, Arctic is the region that has at least one day without daylight in the winter, and at least one day without night in the summer. According to this criterion, the Arctic region encompasses all lands, seas, and ice caps north of latitude 66.5 N, which is called the Arctic Circle (Figure 13.6). However, if defined according to the temperature, frost line, or sea ice cover, the border

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

563

Figure 13.6 Map of the Arctic region bounded by the Arctic Circle. Four major gates of sea ice advection are indicated by arrows. Three gates allow ice outflow, and one gate at Bering Sea allows ice and water inflow to the Arctic.

Geographically speaking, most of the Arctic region is covered by water, of which the Arctic Ocean has the biggest basin. This ocean is the smallest of the world’s five oceans with area nearly 14 × 106 km2. It is also the shallowest with an average depth of 1.1 km and deepest part of about 5.5 km at the floor of Fram Strait. The Arctic Ocean and the surrounding seas, including Barents, Kara, Laptev, East Siberia, Chukchi, Beaufort, and Greenland, are nearly landlocked by parts of Siberia, Alaska, Canadian High Arctic islands and Greenland. The Arctic Ocean and most of those seas are covered with ice during winter. In most areas of the Arctic region the ice is removed through melt during summer. However, since the summer temperature at the core of the Arctic Ocean around the North Pole (north of Greenland and the Canadian archipelago) exceeds the ice melting point by only a few degrees, the ice can only partially melt. The only other means of ice removal from this landlocked ocean is by migration to southern areas through specific gates. Figures 13.6 show four gates at Fram Strait, Robeson Channels, the Canadian Arctic Archipelago (CAA), and Bering Strait. Three more gates are identified in Figure 13.23 (marked 2, 5, and 6). Ice advection through these gates continues all year-round, yet a considerable portion of the ice remains entrapped within the Arctic Basin. This is the most consequential effect of the landlocked configuration of the Arctic Basin. Fram Strait is the route of the highest ice flux. Wei, Zhang, Wang [2019] estimated its average ice volume

during 1979–2012 to be around 3216 km3/y, which represent 12.7% of the 34-year average volume through all gates. Ice flux through Bering Strait, the only gateway to the Pacific Ocean, has not been estimated regularly but the volume flux was estimated in Travers [2012] to be around 190 km3/y. The authors used data from an array of subsurface moored Acoustic Doppler Current Profilers (ADCPs) and other instruments deployed in the Bering Strait from 2007–2008. Robeson Channel is a narrow gateway (30 km wide) between Greenland and Ellesmere Island, which allows ice flux from the Lincoln Sea to the Baffin Bay. However, the ice advection is sometimes blocked in the winter by the formation of a natural arch-shaped ice structure (section 2.9.1). The arch usually collapses in late spring or early summer, allowing for the resumption of the ice flux. Kwok et al. [2010] estimated the annual ice flux through Robeson Channel when it is not blocked in the winter to be about 10% of the average annual amount of ice discharge through the much wider Fram Strait (400 km wide). The gateways through CAA allow limited ice flux during summer and usually iceblocked in the winter. Description of those gateways is presented in Howell et al. [2013]. The interest in those gateways is not so much in facilitating ice migration from the Arctic Basin but in blocking the Northwest Passage for polar marine traffic. Hu and Myers [2014] used a coupled ocean and sea ice model forced by the IPCC climate scenario to study the evolution of ice within the CAA. The simulation

564

SEA ICE

thinning of sea ice (65% and 75%, respectively) during the period 2020–2060. If this continues to happen, the Arctic will be mostly covered by seasonal ice, similar to the Antarctic sea ice cover. The entrapped ice may circulate within the Arctic Basin for several years, driven by the Beaufort Sea Gyre (BSG), or for just a few years driven by the action of the Transpolar Drift Stream (TDS) (patterns of these systems are shown in Figure 11.45). This perennial ice lasts typically for 3–7 years and usually grows to thickness between 2–4 m. However, more lifespan and thicker ice are also possible. Perennial ice used to cover two-thirds of the Arctic region up to 1980s. It has been diminishing gradually under the current global warming trend. Today, it occupies nearly one-third of the Arctic area. The decrease of perennial ice triggers a positive feedback cycle. Less perennial ice contributes to higher local atmospheric temperature and this leads to more presence of seasonal ice. The annual cycle of freeze and melt of sea ice in the Arctic varies with location. The onset of freezing of the ocean surface can occur as early as mid-August in the High Arctic. Then, the ice cover starts expanding in September and continues throughout the harsh winter when the temperature at the core area of the Arctic hovers around −35 C. The decline of ice cover starts generally in April, but May is a reasonable time for melt onset across the Central Arctic. Ice melting continues partially through summer even when average temperature only rises to 5 C. Because of the cold winters, sea ice will keep reforming every year and cover the entire Arctic region even as global warming continues. The Arctic temperature will always be below (sometimes deep below) the freezing temperature in winter, no matter what happens to the warming of the global temperature. Hypothetically speaking, if the Arctic temperature rises above the freezing temperature in winter (a scenario associated with ice-free Arctic), the average global temperature will have exceeded 50 C, and this is when most irreversible damage to cells starts. This means many aspects of life on Earth will be annihilated before the Arctic will be ice-free in winters. The argument of a possible ice-free Arctic Ocean all yearround is neither foreseeable nor realistic. Global warming is affecting the characteristics of the ice cover but not threatening its presence in winter. The Arctic is expected to be partially ice-free during summer only. Unlike the Arctic sea ice, which extends from the North Pole all the way south to the border of the Arctic Circle at 66.5 N latitude, the Antarctic sea ice fringes the Antarctic continent. In the east sector, it starts at the Antarctic Circle latitude of 66.5 S (Figure 13.7). This means that ice freezes and melts at higher temperatures (i.e., lower latitude south) than the Arctic ice. Nevertheless, the Antarctic sea ice also exists within the Weddell, Ross, Bellingshausen, and Amundsen Seas, all located within the Antarctic Circle. The sea ice regime in the Antarctic is divided into five sectors (Figure 13.7), representing

different geographic and climatic conditions, which leave impacts on the sea ice characteristics [Zwally et al., 2002]. Around Antarctica, the ocean current and atmospheric circulation causes ice to drift away from the coastline into warmer water in the summer (no entrapment by geographic features). This ensures nearly complete ice melting during summer, making most of the ice cover in the Antarctic seasonal. This does not warrant extensive presence of perennial ice, which is found mainly in two areas of the Antarctic region: Weddle Sea and Ross Sea. It is interesting to note the geographic similarity between two ice-covered areas in the Arctic and Antarctic regions. The Kara/Barents Seas in the eastern Arctic are similar to the entire zone of the east Antarctic in that both are open to the wide ocean. Sea ice in these areas features seasonal types and possesses the characteristics of the marginal ice zone (MIZ). On the other hand, the semi-enclosed areas of the Weddell and Ross seas in the Antarctic resemble the enclosed area of the Arctic Ocean. Perennial and thick ice are common features in these areas. The perennial ice in the Weddell Sea is also maintained by the circulating pattern of the Weddell Sea Gyre, which hinders the ice advection to the South Atlantic Ocean. 13.2.2. Differences in Sea Ice Characteristics Between the Two Polar Regions The different responses of sea ice in the two polar regions to the recent climate change reside (at least partly) in the different ice characteristics between the two regions. The uniquely different species that live in each polar region, namely the polar bear in the Arctic and the penguins in the Antarctic tell something about the sea ice there. Polar bears live in the Arctic because they need stable platforms of sea ice to travel on and hunt their food. Their main diet is seals and they catch them when they come out to the surface through breathing holes. Polar bears keep switching between sea ice and islands, with not much need for sea water. The thick ice in the Arctic provides the right platform for polar bears to pursue their life and eating habits. Penguins, on the other hand, spend most of their time on land, but it has to be near coastal waters because they catch fish from the ocean. This is readily available in the Antarctic because the wind pushes sea ice offshore and opens numerous coastal polynyas. Some species spend long months at sea but none live in or need sea ice. The frequent coastal polynyas in the Antarctic provide the right habitat for the penguins. Differences between ice characteristics in the Arctic and Antarctic are mainly driven by their reversed geographic configurations as outlined above and also by atmospheric, oceanic, and glaciological factors as explained later. In the meantime, it is appropriate to outline those differences and their natural driving forces. While comprehensive information about the Arctic sea ice characteristics has been compiled from numerous

60

Queen M

°W

an ce O r an to di ec In s

a Se ell r d ed to W sec

aud Lan

d

Enderby Land

Weddell Sea Ronne Ice shelf

B & A Seas sector

East Antarctica

South pole

90°E

Ross Sea

W

0°

Ro s se s S ct ea or

Victoria Land

°W

160

Antarctic circle 66.5°S. latitude

W e O ste ce rn an P se ac ct ific or

Ross ice shelf

75°S. latitude

13

nd

Mary Byrd Land

Wilkes La

West Antarctica Bellingshausen and Amundsen Sea

565

20° E

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

Figure 13.7 Map of Antarctica showing the main seas of land and ocean sectors.

navigation and scientific expeditions since the Cold War era, many aspects of the Antarctic sea ice (physical and otherwise) are still poorly understood. Limited sources of information about the Antarctic sea ice include the book by Jefferies [1998] and the review paper about east Antarctic ice by Meiners et al. [2016]. Table 13.1 summarizes the main characteristics of sea ice in the two regions. The sources of the information include a few websites about polar sea ice (e.g., NSIDC, NOAA, NASA, and ECMWF).

13.3. CHANGES OF POLAR SEA ICE IN RESPONSE TO GLOBAL WARMING The ongoing global warming is expected to impact polar sea ice. Some impacts are measurable such as ice extent, thickness, age, while others can be qualitatively observed/described such as impacts on marine operations, development of cracks on ice shelves and calving of icebergs. All impacts can be predicted using modeling

approaches. Concurrently, global warming impacts atmospheric and oceanic processes in the polar regions, which subsequently affect the sea ice cover. Therefore, sea ice is directly impacted by global warming and indirectly impacted by the influence of global warming on the atmosphere and ocean. Combining these two aspects, the net impact on sea ice can be amplified or neutralized. This point is used to explain the contrasting response of sea ice in the two polar regions to the global warming; namely the remarkable thinning and shrinking of the Arctic ice versus the apparently stable sea ice in the Antarctic. A review of changes in the Arctic region is presented in Meier et al. [2014]. Parallel reviews of the Antarctic sea ice are presented in Turner et al. [2009a], and Pope et al. [2017]. 13.3.1. The Arctic and Antarctic Ice Extent Sea ice extent (SIE) in the two polar regions is a geography-related issue. In the Arctic, SIE is limited by the surrounding landmasses, while in the Antarctic it

566

SEA ICE

Table 13.1 Main characteristics of sea ice in the two polar regions. Arctic sea ice

Antarctic sea ice

Ice is formed in and around the North Pole in the coldest regional area. As a result, the Arctic ice receives less solar energy, mainly because of the longer dark season; hence it is more likely to grow thicker. Crystallographic structure features more columnar grains as ice grows in less turbulent conditions. Much of sea ice is entrapped within the Arctic Basin with only small part is drifted outside to southern seas through specific gates. As a result, not all ice melts in a typical summer, hence multi-year ice (MYI) exists. Ice floes are more prone to converge and collide, hence more ridges and thicker ice is expected. Ice thickness in winter is typically 2–4 m. Maximum ice volume is about 0.05 × 106 km3. Presence of ridges and thicker ice facilitates longer survival of summer ice. Leads are formed intensively in areas of active ice motion such as Beaufort Sea. These are offshore areas. Ice distribution is geographically asymmetric, with much more ice in certain longitudes than others. Since Arctic Ocean is surrounded by land, less precipitation occurs, hence snow on sea ice is relatively thin except near ice edge. Numerous rivers in Canada and Russia feed the Arctic Ocean with freshwater, hence this cold water allows more sea ice growth. Relatively limited number of ice shelves exists, with virtually no effect on sea ice growth. Average minimum and maximum ice extent are 6.5 × 106 km2 and 15.5 × 106 km2, respectively. Mechanism of formation of platelet ice is arguable. It is more common to use the term frazil ice carried by convection to stick at the bottom of the ice. Since 1979, observed change in sea ice is large, and the length of the melt season has increased by 38 days.

can expand easily into the temperate region of the surrounding Southern Ocean. This makes the average SIE in the Antarctic approximately 20% greater than that in the Arctic. However, most of the Antarctic sea ice does not survive the summer melt season. Maximum and minimum extents of sea ice in the Arctic occur in February/ March and September, respectively; while in the Antarctic they occur in August/September and February, respectively. Overall, the Arctic SIE has declined by 12.8% per decade. This rate is likely to be unprecedented at least in the past 1,000 years [IPCC, 2019]. No parallel decline has been observed in the Antarctic as explained in more details in the following. Figure 13.8 is a record of minimum and maximum extents of the Arctic and Antarctic sea ice cover spanning the years 1979–2021. It is based on data provided in spreadsheets from the Arctic Sea Ice News tools of NSIDC: https://nsidc.org/arcticseaicenews/sea-ice-tools/. The data

Ice is formed at latitudes further north from the South Pole, in relatively warmer area. As a result, sea ice receives more solar energy, hence ice growth is relatively slower and ice cover is generally thinner. Crystallographic structure features more frazil and snow ice grains as ice grows in more turbulent conditions. Sea ice expands freely into the open South Ocean, resulting in higher drift speeds. As a result, nearly all sea ice melts in a typical summer, hence less MYI exists, and only in specific areas, namely the Weddell Sea, Ross Sea, and a few coastal areas as landfast ice. Although ice drift is faster, ridge formation is less, often because ice is mostly thin. Ice thickness in winter is typically 1–2 m. Maximum ice volume is nearly half the maximum volume in the Arctic. When the relatively thin ice float northward into warmer water in the summer, it melts fast. Leads usually exist in areas of thin ice and more notably in coastal rather than offshore areas. Roughly symmetric ice distribution around Antarctica. The Antarctic is more open to the wide ocean; hence moisture is more rapidly available and snow on sea ice is thicker and may become flooded with ocean water. Ice shelves break often, resulting in huge icebergs floating and affecting the dynamics of sea ice. Numerous ice shelves exist and the freshwater resulting from basal melt affects sea ice growth. Average minimum and maximum ice extent are 3.1 × 106 km2 and 18.8 × 106 km2, respectively. Melting of the bottom of ice shelves produces supercooled fresh water, which spills out and freezes under coastal sea ice, forming platelet ice crystals (Figure 5.5). Since 1979, observed change in sea ice is small with no change in the duration of the melt season.

are produced by the NSIDC Sea Ice Index (https://nsidc. org/data/G02135/versions/3). A few observations from the graphs in Figure 13.8 are discussed in the following. The Arctic region shows different behaviors between the records of minimum SIE (in September) and the maximum SIE (in February/March). The average of the minimum during the period 2000–2021 is 36% less than what it was during 1980–2000. It decreased at a rate of nearly 95,000 km2/y, which is triple the rate between 1980–2021. This is remarkably higher than the rate of decrease of the maximum extent (in February/March), calculated as 36,000 km2/y. This means that atmospheric warming impacts the ice extent in the summer more than it does in winter. Stroeve and Notz [2018] studied seasonal and interannual changes of the Arctic sea ice in detail starting from 1979. The study found that decadal sea ice loss during winter months in the Arctic has increased from −2.4%/decade (from 1979 to 1999) to −3.4%/decade

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

567

20 18

Area (km2) × 106

16 14 12

maximum ice extent (February/March)

10

minimum ice extent (September)

8 6 4 2 0 1975

1980

1985

1990

1995

2000

2005

2010

2015

2020

2025

2010

2015

2020

2025

20 18

Area (km2) × 106

16 14 12

maximum ice extent (August/September)

10

minimum ice extent (February)

8 6 4 2 0 1975

1980

1985

1990

1995

2000 Year

2005

Figure 13.8 Interannual variability of maximum and minimum SIE (1975–2021) across the Arctic (top) region and Antarctic (bottom) regions. The trend lines are shown for segments from 2000–2021 in the Arctic data (figure prepared by M. Shokr from data made available by NSIDC, website http://nsidc.org/data/seaice.index/).

(from 2000 onward). Close examination of minimum ice extent of the Arctic ice in Figure 13.8 suggests that the Arctic minimum SIE may have recovered slightly lately (2015–2021) from its decline trend. Yet, this has to be confirmed by adding data from more years in the future. In fact, the SIE in February 2022 (not shown in the figure) was 14.61 × 106 km2, ranking fourteenth lowest in the satellite record. It was 400,000 km2 above the record low set in 2018, but still 690,000 km2 below the 1981 to 2010 average. Figure 13.8 shows also that the difference between maximum and minimum ice extent in the Arctic increases at a rate of 10,000 km2 per decade. The summer ice loss increases at a higher rate than the winter ice loss. Warmer temperature in summer has a direct impact on ice melting, while warmer temperature in winter only delays the ice growth, stimulates more openings in the ice cover, and shrinks the SIE at the margins of the ice cover. This limits the loss of SIE in winter. The statistically significant downward trends of the maximum and minimum Arctic SIE suggest a correlation between the two. However, observations suggested that they are not correlated. For example, after the lowest ice extent in September 2012 (3.39 × 106 km2), the maximum extent in the winter of

2013 was not among the ten lowest maximum extents. Similarly, the winter of 2017 has the lowest maximum in the satellite record, but the summer minimum ranked at the tenth lowest maximum extent. W. Meier plotted detrended data of the maximum versus minimum extent in the past 40 years (https://nsidc.org/arcticseaicenews/ 2020/03/no-record-breaker-maximum/). The plot shows random distribution with no correlation whatsoever, namely the minimum extent is determined by the summer weather conditions, which are independent of winter conditions. The Antarctic circumpolar SIE (Figure 13.8) has shown a slight increasing trend during the satellite period (1979– recent), yet with exceptions observed in recent years as presented later. This trend has been widely reported and conformed to reprocessed data that addressed inconsistency issues in the time series [Comiso et al., 2017]. However, Eisenman, Meier, Norris [2014] pointed out that the ice extent has been overestimated because of the way the satellite data were processed using the Bootstrap algorithm. The study suggested that the reported increasing trend was statistically insignificant. According to NSIDC data in Figure 13.8, the ice extent varies slightly around 18.75 × 106 km2 in winter and 2.82 × 106 km2 in

568

SEA ICE

summer. This means that nearly 85% of the Antarctic sea ice melts in summer. The remaining 15% forms perennial ice, which usually concentrate in the Weddell and Ross seas, as mentioned above. However, data in this figure do not reveal the situation within each one of the five sectors shown in Figure 13.7. Sea ice cover around Antarctica varies by region and season due to different dynamic growth, climate conditions, and retreat process, especially in the Ross, Bellingshausen-Amundsen, and Weddell seas [Parkinson and Cavalieri, 2012]. An overall positive trend in SIE was estimated to be 17,100 ±2300 km2/y, but much of the increase occurred in the region of Ross Sea. SIE is indeed associated with regional variability, where some regions show increase and others are show decrease [Meier and Markus, 2015]. In The National Academies of Sciences, Engineering, and Medicine [2017] dataset obtained for the period 1979–2015, the Antarctic-wide SIE data are decomposed by the five sectors shown in Figure 13.7. All sectors show a positive trend in SIE except the Bellingshausen and Amundsen sector, which shows a strong negative trend, since 2007. In contrast, a strong positive trend is shown in the Ross Sea sector. Such local variation in trends can readily occur from natural variability of the atmosphere, ocean, and sea ice system. Comiso et al. [2017] explored the variability of SIE in the Antarctic against the surface ice temperature for 34-year period starting in 1981. The study showed a strong inverse correlation of −0.94 during the freezing season and lesser correlation of −0.86 during the melt

season. With one-month lag in surface temperature (i.e., the ice extent becomes the stimulus and the surface temperature is the response), the results showed stronger correlation of −0.96 and −0.98 during the freezing and melt seasons, respectively. The inverse correlation is expected since ice cover delays the heat release from the warm ocean to the colder atmosphere. However, it would be more interesting to generate a correlation function between nSAT and SIE, contingent on successively increased time lag (e.g., 2–5 days time interval). This will be an effective tool to study the coupling between SIE and nSAT. Parkinson [2019] examined a 40-year record (1979–2018) of satellite observations to quantify changes in the Antarctic SIE. The study confirmed the gradual increase of the SIE by a negligible percentage between 1979 and 2014, yet with substantial spatial and interannual variation. It also revealed an exceptional decrease during the period 2015–2017 at a rate far exceeding the widely publicized ice decay in the Arctic. This comparison is shown in the graph of the Arctic and Antarctic SIE anomalies during the period 1979–2020 available from NSIDC (Figure 13.9). While the decreasing trend of the Arctic SIE has been sustained since 1979, the variation of the Antarctic SIE reveals a different story. A mild trend of increase, yet with large variability is observed between 1979 and 2012, followed by a sharp increase that peaked in 2014. This is followed by a sharp decrease that sets a low record in 2017. After 2017 there have been rebounds but nowhere near the peak of 2014. Spatial distribution of interannual variation of SIE around the Antarctic has not been well established, but

Standardized anomaly (# st.dev. rom the average of 1981–2010)

5 4 3

Arctic Antarctic

2 1 0 –1 –2 –3 –4 –5 1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008 2011 2014 2017 2020

Figure 13.9 Anomalies of the Arctic and Antarctic SIE expressed in terms of number of standard deviation from the average of 1981–2010. Thin lines indicate monthly anomalies and thick lines indicate 12-month running means. Note the steep rate of the Arctic ice decline and the very mild rate of increase in the Antarctic ice extent. A sharp rise followed by sharp decline are observed in the Antarctic data during the period 2012–2017 (the graph is provided by NSIDC, https://nsidc.org/cryosphere/sotc/sea_ice.html).

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

there are indications that it has increased in the Ross Sea and decreased in the Weddell Sea. It is too early to relate the behavior of the Antarctic-wide SIE to the current global warming. More data are needed to monitor the changes which link them to environmental causes. A key theme to take from the above discussions is the importance of examining the SIE data following a geography- and climate-based sectorization of the polar regions. The region-wide trend of ice extent (or thickness) may not reveal the different local trends (sometime reverse trends) within certain sectors. For example, ice trend in the Kara/Barents Seas in the Arctic does not represent the Arctic-wide trend. Similarly, most of the increase in the Antarctic sea ice so far, has occurred in the Weddell and western Ross seas only [The National Academies of Sciences, Engineering, and Medicine, 2017]. In general, the Antarctic sea ice is less understood compared to the Arctic sea ice. Its behavior in response to the global warming is far less understood. Stroeve and Notz [2018] explored the relationship between pan-Arctic SIE and the global anthropogenic CO2 emission. The study found that the annual loss of sea ice per ton of CO2 emission ranges from 1 m2 in winter to more than 3 m2 throughout summer. This is an interesting estimation because it links the source of the current global warming (anthropogenic greenhouse gas emission) directly to the SIE, bypassing all intermediate impacts of the emission on the environment. Based on a linear interpolation of this trend, the study predicts that the Arctic Ocean will become ice-free through August and September for an additional 800 ± 300 GT of CO2 emissions, and through July to October for an additional 1400 ± 300 GT of CO2 emissions. It should be noted that the global emission of CO2 remains at around 38 GT level annually though it decreased to 31.4 GT in 2020 during the industrial lockdown due to the outbreak of COVID-19. A study on the impact of the decreased CO2 emission on the Arctic sea ice concentration during the same period is presented in Chen et al. [2021]. 13.3.2. The Arctic and Antarctic Ice Thickness and Volume Quantification of sea ice thickness (SIT) and its interannual trend is more difficult than SIE. This is due to the limited spatial coverage of satellite data from which thickness can be determined [Lindsay and Schweiger, 2015] and the fact that SIT should be estimated at a relatively fine resolution in order to warrant a reasonable level of ice cover homogeneity within the resolution cell. At present, the prime tool for estimating large-scale coverage of SIT in the polar regions is the space-borne altimeters (radar and laser). Characteristics of these sensors are presented in section 8.6 and the principles of SIT retrieval are

569

presented in section 11.4.3. It is worth repeating that the unknown snow cover or inaccurate estimates of ice freeboard are the main difficulties that hamper accurate retrieval of SIT. Also, estimation of thin ice thickness from altimeter data is less accurate than thick ice because of the uncertainty of freeboard measurement in this case. SIT information about the Arctic ice in winter is more crucial than SIE or SIC. This is because SIE continues to cover nearly the entire Arctic Ocean, unimpacted by climate change, and SIC remains close to 100% in the core area of the Arctic. At the same time, SIT (as well as ice volume), being much impacted by climate variability, has featured significant spatial variation but with an identifiable declining trend. The reduction of SIT and the increase of openings in the sea ice cover in the Arctic during winter are aspects of the impacts of global warming and they both contribute to changes in local atmospheric parameters and weather phenomena. Both aspects trigger positive feedback loops. Recently, interannual variations of SIT and SIE in the Arctic have received more attention than the Antarctic because of Arctic environment has greater climatic and economic impacts on the global climate. In a limited way, the Antarctic sea ice is important for the formation of bottom water and global ocean circulation. Sporadic data from selected locations in the Arctic have confirmed the progressive decline of SIT, yet with significant spatial and interannual variability. Using combination of altimeter, scatterometer, and submarine data, Kwok [2018] reported that between the submarine period (1958–1976) and CryoSat-2 (2011–2018), the average SIT near the end of the melt season decreased by 2.0 m (i.e., 66% over six decades) in six regions. The decline of SIT in the MIZs of the Arctic is sharper than the decline in the core area [Mallett et al., 2021]. Mesoscale data of SIT at fine resolutions suitable for marine operations are not readily available. This will continue to be an issue, especially if near-real-time data are needed by users. What is readily available at this time is regional scale SIT maps retrieved from altimeter data, though at coarse spatial and temporal resolutions. Though the record is rather short, data are still useful for climate studies as they offer clues on the interannual variability and the spatial distribution of the Arctic SIT. Research on improving SIT retrieval from altimeter data has been (and will continue to be) active along with validation of the products. Validation has been conducted using SIT from upward-looking sonar or airborne systems (e.g., NASA’s Operation IceBridge program). Maps that show the difference between the pan-Arctic SIT obtained in January 2015 and January 2022 from CryoSat-2 data are presented in Figure 13.10. The maps are produced daily on a 25 km grid as 30-day averages for the months between September and May. The sharp

570

SEA ICE January 2015

0.00

0.50

January 2022

1.00

1.50 2.00 2.50 Ice thickness (m)

3.00

3.50

Figure 13.10 Change of the Arctic-wide SIT calculated from CryoSat-2 data between January 2015 and January 2022, at 25 km resolution. The significant reduction of thickness north of the Canadian archipelago is noticeable (adapted from maps provided by the ESA’s Centre for Polar Observation and Modelling Data Portal, University College of London, www.cpom.ucl.ac.uk/csopr/seaice.php).

drop of thickness within the 7 years in Figure 13.10 is particularly noticeable north of the Canadian archipelago and east of Greenland. This is a clear indication of the first-year ice (FYI) replacing MYI as the latter is being lost at a higher rate recently (see Figure 13.13) due to melting and/or possibly higher advection out of the Arctic Basin (the latter point is investigated in section 13.4.1.3). When Mallett et al. [2021] used snow data with realistic variability instead of conventional climatology data to retrieve ice thickness from CryoSat-2 data, they found that mean ice thickness declines between 60%–100% faster across the Arctic. Milder reduction is observed in the Eurasian zones. Progress in data processing will improve the results and longer altimeter data products will assure further understanding of the climate/ice interaction system. Regular products of SIT in the Arctic are generated by NSIDC using CryoSat-2 SIRAL and PM SSMIS. Maps are available from September 2010 to present, and cover the region north of 55 N, at 25 km × 25 km spatial resolution and 30-day temporal resolution (https://nsidc.org/ data/RDEFT4/versions/1). NASA produced same maps from ICESat-2 at same spatial and temporal resolutions from November 2018 to April 2021, covering area between 60 N and 88 N (https://nsidc.org/data/IS2SITMOGR4/versions/2. So far, most of the studies that use these records focus on data processing, product validation, and sources of error [e.g., Tilling, Ridout, Shepherd,

2018] (see also section 11.4.3 on the uncertainty of thickness measurements). More studies are needed to present the results in a context of the Arctic sea ice/ climate interaction. It is worth repeating that accurate snow loading must be used to produce reliable SIT maps, hence correctly linking them to the Arctic warming. A multisensory platform with sensors combining altimeter and PM for measuring snow depth would be useful. In order circumvent the short duration of the altimeter records and the uncertainty of SIT retrieval, an alternative approach to estimate SIT is by using ocean/ice/atmosphere models, with assimilation of satellite observations. Labe, Magnusdottir, Stern [2018] used the Pan-Arctic Ice Ocean Modeling and Assimilation System (PIOMAS) [Zhang and Rothrock, 2003] and the Community Earth System Model Large Ensemble Project (LENS) (https://registry.opendata.aws/ncar-cesm-lens) to estimate the spatial and temporal variability of the Arctic SIT and volume. The PIOMAS model was developed at the Applied Physics Laboratory, University of Washington. Useful historical records of the Arctic SIT and its anomalies, obtained from the model and satellite observations, are constructed by Z. Labe of Colorado State University and made available as open source through the website https://sites.uci.edu/zlabe/arctic-seaice-volumethickness/. Figure 13.11 shows monthly variability of ice thickness from the 1990s to February of 2022. Regardless of the validation issue, data are showing

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

571

Ice thickness (m)

2.5

2.0

1980s 1990s 2000s

1.5

2010s 1.0

2022

0.5 Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

Figure 13.11 Arctic-wide SIT calculated using PIOMAS model during the 3 decades of 1990s, 2000s, and 2010s. More data are available through the website https://sites.uci.edu/zlabe/arctic-sea-ice-volumethickness/ (courtesy of Zachary Labe, Department of Atmospheric Sciences, Colorado State University) .

gradual decrease of SIT through the three decades. Maximum thickness occurs in the beginning of the melt season (June). It dropped from approximately 2.3 m in the 1990s to approximately 1.6 m in 2010s. The minimum thickness occurs at the beginning of the freezing season (October/ November). It dropped from approximately 2 m to 1.0 m. Visual analysis of Figure 13.11 shows also larger decadal rate of SIT reduction during summer (August–October) than winter (January–April). Figure 13.11 shows also larger decadal rate of SIT reduction during summer (August–October) than winter (January–April). As regional information is more important than the Arcticwide sea ice information, it would be useful to break down the SIT using the same modeling approach into four zones: west Arctic, central Arctic, east Arctic (including the MIZ), and the Eurasian. Stroeve and Notz [2018] presented a figure showing a trend of SIT decline across the Arctic Ocean from using the PIOMAS model (started in 1980) with addition of CryoSat-2 data (starting from 2010) from selected areas. It appears that the average SIT decreased from about 3.0 m in the mid-1980s to about 2.2 m in 2017. The actual rate of decline started in the mid-1990s. The thickness reduction reflects the long-term shift from the thick MYI to the thinner seasonal ice. Sea ice volume combines SIE and SIT. Therefore, it is more directly tied to climate forcing than the extent or thickness alone. However, the Arctic sea ice volume cannot be observed regularly because satellite observations and field measurements are limited in space and time (recall that pan-Arctic coverage from altimeter systems occurs every 30 days or so). Once again, the modeling approach with assimilation of satellite observations can

be used to generate records/maps of sea ice volume regularly. The PIOMAS model was used in the Polar Science Center to achieve this purpose. Daily sea ice volume anomalies from 1 January 1979 to 1 January 2022 are computed relative to the average over 1979 to 2020, in order to remove the annual cycle. Figure 13.12 is a record of sea ice volume anomalies, which shows a strong and systematic trend of decline. The Arctic sea ice volume is decreasing at a rate of 2,500 km3/decade (~25% per decade). According to the same data set, the mean ice volume varies between a minimum of 11,500 km3 in September to a maximum of 28,000 km3 in April. The minimum ice volume on record was 2,400 km3, set in the fall of 2017. In January 2022, the average volume (17,000 km3) was the 9th lowest on record. For 2022, the mean monthly ice volume was 39% below the maximum in 1979, and 20% below the mean value for 1979–2021. So far, SIT estimates of the Antarctic sea ice are not generated regularly, hence it is not known whether or not there is long-term trend of either thickness or volume. A need for additional information on the Antarctic SIT has been identified [IPCC, 2013]. However, a few studies have shown that changes in the Antarctic SIT are insignificant. This is demonstrated in Kurtz and Markus [2012] using ICESat observations during 2003–2008 and Worby et al. [2008] using ship-based observations during 1981–2005 (note the relatively old dates of this information). Xei et al. [2013] used the ICESat altimeter to estimate the SIT in the Bellingshausen/Amundsen seas during the period of 2003–2009 and reported an overall increasing trend of 0.03 m/y but with no statistical significance (p = 0.11). Distribution of interannual SIT shows wide

SEA ICE

5

5

0

0

–5

–5

–10

–10

2022

2020

2018

2016

2014

2012

2010

2008

2006

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

Anomaly Trend: –3.0±1.0 [1000 km3/Decade] Last Day: 2022–01–31 Version: IC–SST 2.1

1980

Ice volume anomaly relative to 1979–2021 (×103 km3)

572

Figure 13.12 The Arctic sea ice volume anomaly from PIOMAS model updated once a month. The trend for the period 1979–present is shown and the shaded areas show one and two standard deviations from the trend. Error bars indicate the uncertainty of the monthly anomaly (data are in public domain, courtesy of Polar Science Center, University of Washington, psc.apl.uw.edu/research/projects/arctic-sea-ice-volume-anomaly/).

range of mean thickness and standard deviation. This might be associated with ice deformation rather than impact of climate change. Spatial maps of mean effective SIT in the Antarctic in the four annual seasons, generated for the period 1992–2010, are presented in Holland et al. [2014]. Data were obtained using the c62r general circulation model of the Massachusetts Institute of Technology (MITgcm). Effective ice thickness is defined as the product of ice thickness and the ice area concentration (i.e., ice volume per unit area of ocean). The data show nominal thickness of approximately 1 m all around Antarctica, with maximum extent in July–September and minimum in January–March time windows. Thick ice (close to 2 m) is found attached to the coast in the eastern section of the AP, the coast of the Bellingshausen/Amundsen seas area and the southern part of the Weddell Sea. The limited area of thick ice and the fact that it is mostly fastened to land or ice shelves explains the very mild rate of the Antarctic sea ice volume increase, 28.7 km3/y (0.4% per year). Kacimi and Kwok [2020] presented data on the Antarctic sea ice freeboard and thickness from combined observations of ICESat-2 and CryoSat-2. The data spanned 8 months of austral winter between 1 April and 18 November, 2019. While the data do not provide information on decadal changes of SIT (as intended for the

presentation in this section), they are useful in showing the spatial distribution of SIT. Thick ice (4–5 m) is found in coastal areas of two sectors only: the Bellingshausen/ Amundsen seas and the Weddell Sea (Figure 13.7). It is possible that at least part of this ice is landfast. To conclude, since less is known about the Antarctic sea ice compared to Arctic sea ice, the next two sections will focus on changes of ice age and dynamics in the Arctic region only. 13.3.3. The Arctic Sea Ice Age The prime impact of global warming on the Arctic ice cover is not the decline of its extent but the reduction of its thickness and the loss of perennial ice. Ice will continue to cover the entire Arctic Ocean in winter, but the replenishment of the perennial ice by seasonal ice is the new climate change reality (another new reality is the appearance of MIZs in areas that previously featured consolidated and close compact ice). For those who have passion for the Arctic environment, thick, and MYI are indicators of “healthy” Arctic sea-ice cover. Unfortunately, this aspect is being deteriorating. Sea ice age is an important indicator of changes in dynamic and thermodynamic ice regimes in the Arctic. Younger ice makes the entire ice cover more vulnerable to break up, deformation and drift, which enhances ice

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

573

MYI extent (×106 km2)

4.5 R-squared: 0.8543 RMSE: 0.3172

4.0 3.5 3.0 2.5 2.0 1.5 2003

2005

2007

2009

2011

2013

2015

2017

Figure 13.13 Interannual variation of mean extent of MYI in the Arctic, averaged in February from 2002 to 2017 (courtsey: Zhilun Zhang, Sun Yat-Sen University, China).

dynamics, melting and migration outside the Arctic region. Gradual disappearance of aged sea ice has stimulated studies on sea ice age and promoted new products. Ice age is assigned on a yearly basis. FYI is identified as ice of less than one year while MYI is binned 2–10 years, but usually the ice older than 5 years is binned into the “5+” age category. Tschudi, Stroeve, Stewart [2016] used ice motion data to estimate the ice age and ICESat to retrieve ice thickness (more on methods to estimate ice age is presented in section 11.6). The authors then related ice age to ice thickness as shown in the Table 13.2. Data are averaged between February/March 2004–2008. Note the lack of monotonic rate of thickness increase with age, perhaps, because of the limited data set used or the sequential processes of bottom ice melt followed by ice bottom accretion as described in section 6.2.4. Kwok [2018] used data from scatterometer systems and found that the Arctic has lost more than 2 × 106 km2 of MYI (regardless of its age) between 1999 and 2017, that is more than 50% of the MYI that existed in the 1990s. The study suggested that MYI variability explains 85% of the variance in the anomalies in the Arctic sea ice volume. Trends of decline of Arctic MYI are strong indication of the sensitivity of the ice cover to climate forcing. Similar results are confirmed in another study [Zhang et al., 2019] using scatterometer and radiometer data during the period 2003–2017 with results averaged over the month of February as shown in Figure 13.13. Nearly 2 × 106 km2 of MYI extent was lost during the 15-year study period. This represents 4.2% loss per decade relative to the mean value of the MYI extent during this period. Maslanik et al. [2007] pointed out, for the first time, the increasing loss of the older (thicker) MYI in the Arctic

compared to younger (thinner) MYI. The study projected an increasing loss of total sea ice as a result of more loss of older MYI (pointing to a positive feedback cycle). In the mid-1980s, the percentage of the young MYI (i.e., secondand third-year) was 35%. This ratio has increased to 58% by mid-2000s. The authors suggested that the decline of thicker (older) MYI, combined with the enhanced mobility and the more openings in the ice cover, should boost the overall ice loss through migration to southern latitudes. This theme is pursued in later studies as explained in section 13.4.1.3. Weekly maps of spatial distribution of ice age in the Arctic are available through NSIDC link https://nsidc. org/data/nsidc-0611. Data are offered at spatial resolution 12.5 km × 12.5 km with temporal coverage from January 1984 to 31 December 2021 (at the time of writing this chapter). Ice age is estimated based on ice motion, which is derived from a suite of PM and TIR sensors in addition to drifting buoys [Tschudi, Meier, Stewart, 2020]. Example of quick-look version of these maps for the period 12–18 March in 1985 and 2021, are shown in Figure 13.14. Note that the entire Arctic region is icecovered in March. In 1980s, seasonal ice used to cover a small area around the extensive MYI cover. Today, it covers the entire area of the eastern Arctic while thick MYI is confined to a small area, north of the CAA. Most of the MYI cover is second year as shown in the map. Note also that the maps do not resolve the small scale MYI floes in the Baffin Bay, hence showing the area with exclusive FYI cover. Data are also missing within the narrow channels of the CAA, an area which is the most important for determining the future of the Northwest marine passage route. In spite of missing such fine-scale

Table 13.2 Mean ice thickness of different ice age categories in the Arctic [data from Tschudi, Stroeve, Stewart, 2016]. Ice age Thickness (m)

1st year

2nd year

3rd year

4th year

5th year

1.56

2.20

2.51

2.64

2.99

574

SEA ICE

EASE-Grid Sea Ice Age, v4.1

EASE-Grid Sea Ice Age (QL)

Mar 12 – 8, 1985

Mar 12 – 18, 2021

Age (years) 0–1 1–2 2–3 3–4

Age (years) 4+

0–1 1–2

NASA NSIDC DAAC University of Colorado Tschudi, Meler, Stewart

2–3 3–4

4+

NASA NSIDC DAAC University of Colorado Tschudi, Meler, Stewart

Figure 13.14 Spatial distribution of four age-based categories of the Arctic sea ice, calculated between 12–18 March for the years 1985 and 2021. Note the remarkable reduction in the perennial ice (with permission from W. Meier, NSIDC).

details, the maps still hold valuable synoptic scale climatic information. Interannual evolution of MYI area of each age category is shown in Figure 13.15. The rate of ice loss is virtually the same for each category older than 1 year, reflecting the fact that ice advection out of the Arctic Ocean does not favor certain ice age category. This makes the 4+ year category nearly diminishing at present. Interestingly enough, this

category comprised nearly 30% of the total ice cover in 1980s, but decreased to nearly 3% in 2021. Longer record of these data in the future will be one of the best tools to reveal the impact of climate change on the Arctic ice. It should be noted that a new ice age algorithm is presented in Korosov et al. [2018], which uses daily sea ice drift and concentration from OSI-SAF of EUMETSAT to generate individual ice age fractions in each pixel. This method has

100% 90%

Percentage ice coverage

80% 70% 60% 50% 40% 30% 20% 10% 0% 1985

1990 Ice age

1995 0–1 y

2000 1–2 y

2005 2–3 y

2010 3–4 y

2015

2020

4+ y

Figure 13.15 Time series of percentage of age-based sea ice categories in the Arctic from 1985–2021. The inset shows the data domain in purple shading (with permission from W. Meier, NSIDC).

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

not yet been incorporated in an operational product in OSI-SAF, but a limited comparison with the NSIDC product has revealed a few improvements. 13.3.4. The Arctic Sea Ice Dynamics As mentioned in the previous section, the remarkable depletion of thick and old sea ice in the Arctic, gives rise to expansion of thinner (seasonal) ice during winter. Subsequently, the mobility of the ice cover is enhanced [Spreen, Kwok, Menemenlis, 2011]. This possibly boosts mass ice outflow and increases formation of leads, ridges, and other forms of ice deformation. Kashiwase et al. [2017] analyzed satellite data (1979–2014) along with a simplified ice/ocean coupled model and found that as thin ice extent increased interannually, divergent ice motion enhanced and this triggered large-scale feedback which amplified summer sea ice anomalies (and possibly winter time anomalies). The magnitude of divergence of the Arctic ice cover has doubled since 2000. Zhang et al. [2021] presented data of the spatiotemporal variability of ice motion in the Arctic region from 1979 to 2019, based on the sea ice motion vector product from NSIDC polar pathfinder 25 km EASE grid (Figure 13.16). A trend of increasing ice drift speed is detected, especially in the past 2 decades. Statistically significant positive trend (p < 0.001) is observed in all seasons, with highest rate in autumn (0.13 km/day/year), followed by the spring, then winter, and the lowest in summer (0.04 km/day/year). These numbers are close to results from

previous studies. Using the International Arctic Buoy Program (IABP) data (also started in 1979), Olason and Notz [2014] estimated the rate of increase of ice motion in autumn to be 0.11 km/day/year. Rampal, Weiss, Marsan [2009] estimated the rate in summer to be 0.05 km/day/year. The differences may be caused by the different temporal records used in each study. Nevertheless, all data confirm the interannual increasing trend of ice motion in the Arctic, regardless of the season, and this can be attributed to the recent trend of ice thinning. Once again, statistics of ice motion data over the entire Arctic may not reveal details that characterize some individual regions. Zhang et al. [2021] examined the ice motion trends in eight subregions and found considerable spatial heterogeneity. Positive trends with values higher than the average over the entire Arctic are found in the Beaufort, Chukchi, and East Siberia Seas. No significant trend was identified in summer ice drift in the Kara and Laptev Seas (obviously, this is an open drift ice regime). By examining the Arctic ice motion maps during October–April for 36 winters between 1979 and 2015, Kaur, Ehn, Barber [2018], identified the boundaries of the circulation regimes of Beaufort Gyre (BG), transpolar drift (TPD) (see their patterns in Figure 11.45), and ice motion systems in the Kara Sea (KS). The study also presents regression analysis of the ice drift speed anomalies, which shows statistically significant positive trends in BG, TPD, and KS and statistically insignificant negative trends associated with weak speeds north of the CAA, Chukchi Sea and the Siberian Shelf.

10 Spring: Y = 0.07685X-148.6 (R2 = 0.7389, P < 0.001)

Seasonal sea ice speed (km/day)

9

Summer: Y = 0.04297X-81.25 (R2 = 0.6067, P < 0.001) Autumn: Y = 0.1265X-245.9 (R2 = 0.8706, P < 0.001) Winter: Y = 0.1075X-208.8 (R2 = 0.8745, P < 0.001)

8

7

6

5

4

3 1979

1984

1989

1994

575

1999 Year

2004

2009

2014

2019

Figure 13.16 The Arctic sea ice drift speed between 1979 and 2019, decomposed by seasons. The linear trend equations are shown by the solid lines (courtesy of Fanyi Zhang, Wuhan University, China).

576

SEA ICE

Global warming minimizes the likelihood of ice arch formation at key ice outflow gates in the Arctic [Howell et al., 2013] (see more information about ice arches in section 10.3.1). Ice arch brings ice motion upstream the arch to stagnation. Therefore, with less likelihood of arch formation and enhanced Arctic ice dynamics, the sea ice composition in southern latitudes may include more pieces of the Arctic ice. This possibility was realized as reported in Barber et al. [2018]. The study reported in-situ observations of MYI and thick FYI floes floating, for the first time, in the Canadian waters northeast the coast of Newfoundland, about 3000 km south of the Lincoln Sea and the CAA (i.e., from where they originate). The floes were observed from the bridge of the Canadian ice breaker Amundsen in the spring of 2017. Barber et al. [2018] examined the pathway of MYI floes from the Lincoln Sea all the way to the coastal communities in Newfoundland. The ice takes 21 weeks to complete this path. The study also discussed how the increasing mobility of the Arctic ice may produce ice hazards in southern seas. 13.3.5. The Antarctic Icebergs The inconclusive warming trend of air temperature over Antarctica (Figure 13.4) is associated with no identifiable trend observed over the sea ice (Figure 13.5). This, along with the absence of a conclusive trend of interannual variability of SIE (Figure 13.8), suggests that other factors might have influenced sea ice and offset the direct influence of global warming. This includes atmospheric and oceanic forcing as well as drifting icebergs (section 13.4.2). This section presents data on the increasing number of icebergs observed during the recent episode of global warming. While data on sea ice type and thickness distributions in the Antarctic are sporadic or non-existent in some parts, iceberg data are readily available. Icebergs calve from ice shelves. A brief definition of ice shelves and the calving process are introduced first. Glacier and ice streams rest on land but they continuously flow onto coastal areas and eventually take the form of huge ice mass, known as ice shelves. They float in the ocean, yet remain attached to their original ice mass. Ice shelves surround 75% of the Antarctic coastline, covering an area approximately 14 million square kilometers. This is close to the 17.1 million square kilometers area of Russia, the largest-size country of the world. This landscape contrasts the Arctic region where ice shelves exist in a few identifiable areas, mainly along the coasts of Greenland, the northern section of Ellesmere Island, and parts of Russian Arctic. Ice shelves act as buttresses to support glaciers. Without them, the backpressure is removed and glaciers calving into the ocean accelerates. This would potentially increase sea level rise. Similarly, sea ice acts as buttresses to support ice shelves and delays their possible collapse. Nevertheless, an increasing number of collapses of ice shelves,

mainly located in the AP but also in the east Antarctica, has been recorded in the last 30 years. Collapse of ice shelves generates icebergs of different sizes from small to giant as well as numerous pieces of chunky ice, all float mostly within the sea ice cover. It is impossible to estimate their number at any time. News about calving of huge icebergs in the Southern Ocean usually circulates in scientific reports and news media. Such icebergs have areal extent of thousands of square kilometers. For example, the biggest iceberg (A76), broke off on 26 May 2021 from the western side of Ronne Ice Shelf (see the geography in Figure 13.7), was 4,320 km2 in size. Only a small number of massive icebergs (10–15) currently floating around Antarctica compared to the numerous small ones. The average rate of iceberg mass calving was estimated by Qi et al. [2021] as 955.4.3±51.4 Gt/y and the spatial distribution of the calving rate around Antarctica is shown in Figure 8 in the same paper. The underlying cause of the increasing number of ice shelf collapse is commonly linked to the ongoing global warming. Both warm air above the shelf and warm water under the shelf contribute to shelf melting. Warm air melts the surface, forming ponds of meltwater where water trickles down through small cracks. As it deepens, cracks erode, expand and eventually lead to iceberg calving. The underlying warmer ocean water, which are linked to warming atmospheric circulation patterns, melts the ice shelf from below. This makes it more vulnerable to cracking and, therefore, furnishes additional mechanism of ice shelf collapse and iceberg calving. Enhanced iceberg calving has been attributed to varying atmospheric and oceanic conditions [Liu et al., 2015, Massom et al., 2018]. A database of the Antarctic iceberg tracking paths is available through the website of the Brigham Young University (BYU) Microwave Earth Remote Sensing (MERS) (http://www.scp.byu.edu/data/iceberg/database1.html). It uses several scatterometer observations: ESA’s ERS-1/2 and ASCAT, ISRO’s OSCAT, NASA’s NSCAT, and SeaWinds on QSCAT (see section 8.5 for details on the sensors). The dataset is supplemented with weekly iceberg tracks from the USA National Ice Center (NIC). Tracks of all icebergs with area >5 km2 during 1992–2019 are shown in Figure 13.17. Note their intensive presence in the Weddell Sea and its northern area. More than 90% of all icebergs, regardless of their origin, pass through the Weddell Sea [Stuart and Long, 2011]. Moreover, the meridional currents close to Antarctica usually direct the icebergs poleward, which explains the concentration of the paths around the coastline. Also shown in Figure 13.17 is the anomaly of sea ice concentration in September 2019 relative to years 1981–2010. The point that should be drawn from this figure is the high population of icebergs within the sea ice regime around Antarctica. This is not the case in the Arctic region where icebergs are mainly calved from the east and west coasts

IMPACTS OF CLIMATE CHANGE ON POLAR ICE (a)

(b)

0° 30° W

0° 30° W

30° E

60° W

60° E

30° E

60° W

60° E

90° E 90° W

90° W

120° E

120° W

90° E

120° E

120°W

150° E

150° W

577

150° E

150° W 180°

180° –100

–75

–50

–25

0

25

50

75

100

Sea ice concentration (%)

Figure 13.17 Tracks of tabular icebergs larger than 5 km2 around Antarctica during the period 1992–2019 (left) (credit BYU data archive) and anomaly of maximum sea ice concentration in September 2019 relative to 1981–2010 (right) (credit Copernicus data archive).

of Greenland and take routes mostly in open seas, away from the sea ice cover within the Arctic Basin. Tournadre et al. [2016] examined the iceberg size distribution spanning the period 1992–2014 using altimeter data. The study combined small (< 8 km2) and large (>100 km2) areal size icebergs and found the size distribution to follow a power law with slope −1.52 ± 0.32. This means that the majority of icebergs are of small size. They are mainly fragments of larger icebergs and can either be trapped in sea ice and later released losing their connection to the parent iceberg. In fact, the vast area of the ocean around Antarctica is populated by this kind of small icebergs. England, Wagner, Eisenman [2020] found that small icebergs account more for Southern Ocean freshwater than large icebergs. 13.4. COUPLING BETWEEN POLAR SEA ICE AND ENVIRONMENTAL FACTORS A few forms of coupling between sea ice cover and environmental phenomena are addressed in this section. As mentioned earlier, their combination has enhanced the warming of the Arctic region while neutralizing it over the Antarctic region. Recall that the anomalies of average air temperature over the Antarctic are significantly small compared to the Arctic region (Figure 13.4). The couplings between sea ice and environment parameters should be carefully considered in relation to following question. Why the Antarctic sea ice does not

respond to the global warming the same way the Arctic sea ice does (namely by shrinking and thinning)? Attempts to answer this question have stimulated many research studies using observations and regional climate modeling. Some results are highlighted in this section. It is fitting to mention here that Serreze and Meier [2019] introduced an interesting relevant theme. The seasonal cycle of the SIE in the Arctic has become increasingly pronounced recently, and it starts to look more like the observed cycle in the Antarctic. This cycle features ice growth during the freezing season and almost complete disappearance throughout summer, leaving small amount of perennial ice cover. The Arctic ice deviates from this pattern because of two reasons. The first is the landlocked geographic configuration, which traps the ice within the Arctic Basin for several years with limited routes of “escape.” The second is the much colder temperature across the Arctic Ocean, which delays or ceases the complete ice melting. If time comes when all MYI and most of the thick ice are drifted out of the Arctic region as a result of the current enhanced ice mobility, then the sea ice will become mostly seasonal during the freezing season. This scenario appears to be ongoing as thick FYI and MYI are increasingly exported out of the region and replenished by thinner ice. If this scenario continues, the similarity between annual sea ice cycle in the Arctic and Antarctic will be approached (mostly seasonal ice in winter and almost complete melt in summer). This will have significant impact on the Arctic and the northern hemisphere regions

SEA ICE

as the air temperature will rise, causing major weakening of the Arctic amplification atmospheric pattern. With this in mind, the impending question should not be about why the Antarctic sea ice has not responded to global warming in the same way the Arctic sea ice has (by shrinking and thinning). It should rather be about how (and why) the warming trend has rendered the behavior of the Arctic ice to approach that of the Antarctic ice, namely bolstering its seasonality and abandoning its perennial feature. Briefly stated, global warming is relaxing the entrapment of the old ice within the Arctic Basin. This is a key process that feeds into the AA phenomenon and accelerates changes of the Arctic sea ice cover in a way that approaches the annual cyclic pattern of the Antarctic sea ice (i.e., nearly complete melting by end of the summer before refreezing the following year). 13.4.1. Interaction of the Arctic Sea Ice with the Environment The commonly circulated knowledge that links the Arctic sea ice loss to increasing emission of CO2 (an anthropogenic cause) has been confirmed in several publications [e.g., IPCC, 2013, Notz and Stroeve, 2016]. In a modeling study, Stroeve and Notz [2018] developed a linear relationship between the Arctic sea ice loss and the cumulative anthropogenic CO2 emission (see last paragraph in section 13.3.1). However, it should be noted that regardless of the origin of global warming, the recent loss of the Arctic sea ice triggers the AA phenomenon, which intensifies the Arctic warming. The AA responsible for the warming of the Arctic region at 2–3 times faster than the rest of the globe. The parallel coupling between the ice cover and atmospheric/oceanic environment in the Antarctic has triggered negative feedback that has neutralized the impact of anthropogenic-drived global warming on the sea ice in that region (see section 13.4.2). Undoubtedly, the anthropogenic emission affects the sea ice cover directly through warming but also indirectly through modulation of atmospheric and oceanic systems. The following discussions highlight the complexity of the interactions of many factors that contribute to the steep decline of the Arctic sea ice. Clues to identify the constructive and destructive interactions that lead to amplification or stabilization of the ice extent and thickness have been offered through many studies. However, the picture is still incomprehensible; lacking more subtle connections between elements of this complicated system. 13.4.1.1. Atmospheric Factors that Contribute to Changes in the Arctic Sea Ice Interannual trends of sea ice areal decline and surface air temperature rise in the Arctic are already demonstrated by combining the information from Figures 13.4 and 13.8. But is the atmospheric temperature the only factor that

25 Contribution of each mechanism (%)

578

All mechanisms

20 15 Surf. albedo Water vapor

10 5

Clouds

Oceanic heat transfer

0 –5

Surface wind

Figure 13.18 Contribution of each one of the shown five factors (in %) to the variability of sea ice area in the Arctic. The summation is 25%. The rest of the contribution (75%) is attributed to atmospheric temperature fluctuations [adapted from Olonscheck, Mauritsen, Notz, 2019].

alters the sea ice cover? Olonscheck, Mauritsen, Notz [2019] addressed this question by using the CMIPS-phase 5 model to estimate the relative weight of a few factors (each factor is decoupled from other factors) that contribute to the interannual variability of the sea ice cover in the Arctic. Those are albedo, cloud cover, atmospheric water vapor, surface wind, and poleward oceanic heat transfer. The percentage contribution of each factor and the summation of their contributions are presented in Figure 13.18. All factors cause decrease of the ice extent except the surface wind. Among the five factors, surface albedo has the largest contribution (11.5%) to the sea ice variability, while the oceanic heat transfer has a negligible contribution of less than 2%. The total contribution of the five factors is 25%. The rest (75%) is the contribution of atmospheric temperature fluctuations. This is the dominant factor that impacts the sea ice variability. Olonscheck, Mauritsen, Notz [2019] confirmed that this is consistent across observations from satellite data and simulations from global climate models. As atmospheric temperature is established as being the key driver for sea ice variability, it is important to explore the reverse effect, namely what sea ice variability does to the fluctuations of atmospheric temperature at local scales. This is addressed in the following section. 13.4.1.2. Enhanced Arctic Warming due to Changes of Sea Ice Cover The Arctic-wide surface air temperature is controlled by large-scale phenomena including (but not limited to)

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

enhanced heat transfer by warm Atlantic water, downward longwave radiative flux due to increasing cloudiness, development of sea ice, and formation of highand low-pressure systems. The latter is a result of the varying pattern of solar radiation, which is the driving force of atmospheric circulation. Over the mid-to-high latitudes of the Northern Hemisphere, the atmospheric circulation is referred to as Arctic Oscillation (AO). While all those factors control the large-scale patterns of atmospheric parameters (including air temperature), modulation of those parameters at local scales in the Arctic are triggered by development and changes of sea ice, e.g., new ice formation, ice thickening, snow accumulation, openings in the ice sheet, and redistribution of ice types, particularly thick and old ice in response to wind forcing. The following discussions highlight the effects of a few sea ice surface features on the local air temperature. Sea ice melting caused by atmospheric warming exposes more sea water area. This triggers lower surface albedo and more heat release to the atmosphere. Consequently, it sets off a positive feedback cycle that contributes to the AA phenomenon, especially in the colder Arctic areas, e.g., the core area compared to the marginal zones. Ice albedo feedback through opening in the ice cover has been quantified in Kashiwase et al. [2017] using satellite data (1979–2014) along with a simplified ice– ocean coupled model. The study concludes that divergent ice motion has doubled, since the year 2000 and that has triggered large-scale feedback that amplifies summer sea ice anomalies. The study presents quantitative evidence that heat input to the atmosphere through fractures in the sea ice cover is the primary driver of seasonal and interannual variations in the Arctic sea ice retreat. It is well known that leads within sea ice cover, whether open of refrozen, increases the local air temperature

Freeze Trend

579

significantly. Lüpkes et al. [2008] studied the effect of open leads on local air temperature in the Arctic under clear sky using 1D atmospheric model coupled with snow/sea ice model. The study showed that for ice concentration >90%, even a small change by 1% would cause an increase of near-surface temperature by 3.5 C. Ye et al. [2022] showed an interesting reverse phenomenon. Higher concentration of MYI inversely correlated with local air temperature. This is justified based on the significant difference between thermal conductivity of MYI (~1.80 W/m.K) compared to FYI (~2.2 W/m. K). Additionally, MYI accumulates thicker snow than FYI, with its thermal conductivity an order of magnitude less than sea ice (Table 3.1). While the study Ye et al. [2022] has empirically proven the concept of MYI influence on local air temperature using an empirical approach, the modulation of Arctic-wide temperature as a result of complete removal of MYI remains to be determined using a modeling approach. Arctic warming causes delay in both sea ice formation in autumn and ice melting in the spring. The delay contributes to the atmospheric temperature through another positive feedback cycle. Using satellite observations, Stroeve and Notz [2018] found that the accelerated loss of sea ice during the freezing season is driven by lengthening of the melt season. The melt onset has occurred 3 days earlier per decade on average during the period 1979–2017. Moreover, the delay in the onset of freezing is happening 7 days later per decade during the same period. The study also presented spatial maps of trends in freeze-up melt onset across the Arctic (Figure 13.19). The most susceptible areas are those at the margins of the Arctic region. An additional minor factor that contributes to the AA phenomenon is the melting of the permafrost that

Melt Trend

10 0

Days/decade

20

–20 –10

Figure 13.19 Trends in the Arctic sea ice freeze-up and melt onset in days per decade, estimated from microwave satellite observations during 1979–2017 period. Negative and positive scales mean earlier and later days relative to average during 1981–2010 period, respectively [Stroeve and Notz, 2018 / IOP Publishing / CC BY 3.0].

580

SEA ICE

surrounds the sea ice domain. Since higher air temperatures accelerate the degradation of frozen ground on permafrost lands adjacent to the sea ice, more of carbon gases that are trapped in the permafrost could be released. This adds to the CFC gases over the Arctic, which can be another factor accelerating the AA phenomenon.

13.4.1.3. Arctic Warming due to Sea Ice Advection Out of the Arctic Basin A possible secondary cause for the recent decline of the Arctic sea ice would be the acceleration of ice outflow from the Central Arctic to surrounding open seas through specific gateways, particularly the Fram Strait (400 km wide). To examine this hypothesis, a few studies have been conducted to explore the variability of the sea ice export through Fram Strait. The purpose is to identify a possible trend of an increasing ice flux as climate change progresses and, if so, whether it is linked to the recent decreasing trend of Arctic-wide SIE (Figure 13.8). Wei, Zhang, Wang [2019] used the high-resolution global ocean–sea ice model MITgcm-ECCO2 to explore the variability of total sea ice flux through Fram Strait (without identification of ice type and thickness) during 1979–2012. The study found that 12.7% of the Arctic ice is advected away from the Arctic Basin through this strait. It suggested also that ice flux decreased starting mid-1990s because of a shift in atmospheric circulation from conventional Arctic Oscillation to Arctic Rapid Change Pattern. This changed the main source of ice outflow. More importantly, the study concluded that the recent loss of the Arctic sea ice is not triggered by sea

ice outflow through Fram Strait as much as it is triggered by ice melting in the Arctic Ocean. A similar conclusion was drawn from a study on sea ice volume export though Fram Strait between 2003 and 2008 [Spreen et al., 2009]. Ice thickness data were obtained from ICESat laser altimeter, and ice area and drift data were obtained from AMSR-E 89 GHz observations. Results showed slight difference of monthly ice volume flux between the 2 years, where the flux in 2008 was 33 km3/month less. This is a small difference compared to sea ice volume export estimates obtained during the 1990s and it is well within the natural variability. This means no significant change in total sea ice volume flux through Fram Strait has occurred in response to the recent Arctic warming (once again, note the old relatively date of the study). Smedsrud et al. [2017] analyzed a longer record of sea ice export through Fram Strait for the period 1935–2014 using satellite data and surface pressure from nearby stations. The data record showed that the annual mean ice areal export was about 880,000 km2, yet with large interannual variability. This amounts to nearly 10% of the total ice export through all gates in the Arctic Basin. Results confirm the finding from other studies that ice drift speed is maximum in winter (averaged 11–19 km/day) and minimum in summer (averaged 2.5–6.0 km/day). But what about the trend of areal ice flux? Is there an increasing trend caused by sea ice thinning and enhanced mobility as hypothesized? Figure 13.20 (adapted from the same study) provides information to address this question. Data are normalized by the mean values from 1979–2014.

Anomalies of normalized values of ice area

3.0 2.0 1.0 0.0 –1.0 –2.0 –3.0 Areal export through Fram Strait (March-August) –4.0 –5.0 1935

Arctic-wide SIE in mid-September 1945

1955

1965

1975

1985

1995

2005

2105

Figure 13.20 The SIE in mid-September (green) and the areal export of sea ice through Fram Strait in the following spring from March through August (red). Data are normalized by subtracting mean value and divide by standard deviation. Smoothed curve produced by filtering with a 20-year-cutoff 8th order Butterworth filter. The 1979–2014 trends of both data sets are shown as solid straight lines [Smedsrud et al., 2017].

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

Data of sea ice area across the Arctic in mid-September reveals a pronounced decline trend. A parallel trend of sea ice flux through Fram Strait in the following spring is not apparent because the variability is quite high. The shown regression line of the areal ice flux through Fram Strait in the figure (red line) is associated with high variability. The slope of this line is less than the slope of the trend line of SIE. Therefore, even if the data of areal flux through Fram Strait is considered to form a trend, the trend will not match the rate of decrease of SIE (note that ice advection occurs through other gates although Fram Strait is the major one). Smedsrud et al. [2017] suggested that the positive anomalies over the last 36 years are caused by stronger geostrophic winds, estimated to increase by 6% per decade, which was triggered by an increase in surface pressure on Greenland. In brief, the hypothesis that ice flux out of the Arctic Basin would increase as a result of ice thinning and increasing mobility under the Arctic warming is not supported by the shown data. Hence, it is possible that only thick ice export is accelerating, which enhances the presence of thin ice (and mobility) in the Arctic. Yet, it is not possible to determine the flux of a thick ice category, but it is possible to determine the flux of MYI as it can be identified by a few algorithms such as ECICE (section 11.2.2.4). Is the MYI flux out of the Arctic Basin showing an increasing trend during the recent period of global warming? This question is addressed in the next paragraph. Kuang et al. [2022] estimated the Arctic MYI and total ice flux for the winters of 2002/2003 through 2020/2021 through the six gates marked in Figure 13.21; (1) Bering –180°E

–135°E

135°E

1

–90°E

2 3

90°E

6 4

–45°E

5

45°E

0°

Figure 13.21 Map of the Arctic with outflow gates of sea ice marked.

581

Sea, (2) Lancaster Sound, (3) North Water, (4) Fram Strait, (5) Svalbard/Franz Josef Land, and (6) Franz Josef/Severnaya Zemelya. The study used the ECICE algorithm (section 11.2.2.4) to estimate the MYI and total ice flux. Results of interannual variability of MYI and total ice flux area, both from Fram Strait then from all gates, are shown in Figure 13.22. Different combinations of active and passive microwave observations in different periods are used in ECICE to generate the data. The combinations are labeled by the satellite names. There is no apparent trend in the data except for the slight decreasing trend of total ice flux from all gates, starting in the freezing season 2007/08, immediately after the first minimum summer ice extent record in 2007. Data exhibit high variability with a large drop in the 2017/18 season. The total ice flux from all gates is 200.66 ×103 km2 and from Fram Strait is 117.389 ×103 km2, i.e., 58.5% of the total flux is advected through Fram Strait. The remarkable observation is the higher ratio of MYI flux from Fram Strait compared to the total ice flux (0.71). When all gates are included, the ratio becomes half as much, namely 0.36. This means that most of the MYI is advected through Fram Strait. From the data in Figure 13.22, the answer to the question posed at the end of the previous paragraph is that no interannual increasing trend of MYI has been established through the Fram Strait gate or all gates combined during the study period. Figure 13.23 shows the monthly average of total ice and MYI area flux through all six gates and MYI area flux through Fram Strait only. Two observations are worth noting. The first is that almost half of MYI advected out of the Arctic Basin occurs through Fram Strait. The second is that the advection of seasonal ice increases sharply during October–December. This is the period when the ice cover across the Arctic is loose, with concentration below 70% (in general). During this period, much of the significant coverage of open drifted young ice can be easily advected outside the Arctic Basin. After this period, the ice extent thickens and stabilizes and so does the advection. Using moored sonar data, Hansen et al. [2013] presents a dataset of SIT evolution in Fram Strait. It shows contrast between the thickness during the 5 years after the first minimum ice extent on record (2007) against the previous 1990s data. The authors highlighted three remarkable changes in ice thickness after 2007: (1) the thickness of MYI was reduced by 32%, (2) the MYI modal peak width was reduced by 25%, and (3) the fraction of ridged ice with thickness >5 m was reduced by 50%. Given the general thinning of the Arctic sea ice, the reduction of MYI thickness in Fram Strait is expected as shown in Hansen et al. [2013] but the results do not prove it as a response to increasing flux of thick ice through Fram Strait. The conclusion from the above discussions is that the Arctic sea ice outflow through the available gates does

582

SEA ICE

700

ASCATAMSRE

QSCAT-AMSRE

Areal ice flux ×103 (km2)

ASCAT-SSMIS

800

600

ASCAT-AMSR2

500 400 300 200 Total sea ice flux through Fram Strait

100

18/19

19/20

20/21

18/19

19/20

20/21

17/18

16/17

15/16

14/15

13/14

12/13

11/12

10/11

09/10

08/09

07/08

06/07

05/06

04/05

03/04

02/03

0

MYI sea ice flux through Fram Strait

Freezing season (year)

Areal ice flux ×103 (km2)

1200

ASCAT-SSMIS

1400 ASCATAMSRE

QSCAT-AMSRE

ASCAT-AMSR2

1000

800

600

400

Total sea ice flux of all gates

200

17/18

16/17

15/16

14/15

13/14

12/13

11/12

10/11

09/10

08/09

07/08

06/07

05/06

04/05

03/04

02/03

MYI sea ice flux of all gates

0

Freezing season (year)

Figure 13.22 Interannual variability of MYI and total ice flux through Fram Strait (top) and same from the six gates shown in Figure 13.22 (bottom). The labels indicate the satellite input observations used to generate the data from the ECICE algorithm (data courtesy of Yufang Ye and Huiyan Kuang of SYSU).

not appear to impact the retreat of Arctic-wide ice cover in terms of thickness or extent. The reduction of MYI in the Arctic does not seem to be a consequence of increasing ice migration out of the Arctic Basin. It is more likely that the decline of MYI in the Arctic is caused by less ice surviving the summer melt. More studies are needed to confirm this point and to address the issue of MYI budget in the Arctic over the past 3–4 decades. 13.4.1.4. Interaction of the Arctic Sea Ice with Wind The interaction of wind with the sea ice cover is manifested in two aspects. The first, which is more sensible, is the effect on ice motion and deformation. This has been

realized in early studied. For example, Thorndike and Colony [1982] found that most of the changes in ice motion in the Arctic could be explained by the geostrophic wind. The second aspect is the wind effect on the near-surface temperature. Wind blows over areas of different temperatures over the ice cover. Warmer air exists over leads and colder air exists over thick ice. The wind contributes to the restoration of a temperature balance across the region. There is a rule of thumb that links wind and sea ice motion through what is known as the wind factor. It comprises a constant ratio of ice motion/wind speed as well as a turning angle that determines the motion direction relative to the wind direction. These two factors have been updated

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

583

160

120 100 80 60

April

Mar.

Feb.

Jan.

0

Total ice-all gates MYI ice-all gates MYI-Fram Strait Dec.

20

Nov.

40

Oct.

Ice area flux ×103 (km2)

140

Figure 13.23 Monthly average of total ice and MYI area flux from the six gates shown in Figure 13.21, and MYI flux through Fram Strait gate only. Data were obtained during 2003–2021 (courtesy of Yufang Ye and Huiyan Kuang of SYSU).

recently in Maeda, Kimura, Yamaguchi [2020] while taking into consideration the impact of global warming on the Arctic ice motion. Nevertheless, ice motion still depends on other factors that may override the role of the wind (e.g., ocean current, ice thickness, and internal stresses within the pack when the ice concentration is high). Kwok, Spreen, Pang [2013] found only negligible trends in wind speed in the Central Arctic between 1982 and 2009 while ice motion was invariably active. The dynamics of the ice motion can be assessed at a variety of spatial and temporal scales, and they incorporate different relations to the wind [McNutt and Overland, 2003]. The scales cover individual-floe (3 C. Deser and Teng [2008] noted that trends of changes in AO trigger parallel changes of sea ice concentration at regional scale, notably over MIZs during winter and summer. However, the study did not confirm that the observed overall decline of the Arctic sea ice since 1979 could be attributed to changes in AO. Using a very long record (168 years) of sea ice area in six Arctic regions during 1850–2017, Cai et al. [2021] found that the key drivers to the accelerated decline of summer Arctic sea ice during the last several decades was the global warming combined with simultaneous local warming exerted by the AO, NAO, Atlantic Multidecadal Oscillation, and Pacific Decadal Oscillation. Lei et al. [2020] found that sea ice deformation exhibits high intermittence and suggested that deformation takes place mostly during extreme wind events. The authors used atmospheric reanalysis to explore relationship between changes in sea ice kinematics (with links to ice thickness and concentration) and the wind forcing within the AO. They found that ice deformation at large

584

SEA ICE

spatial scale is more susceptible to changes in the atmospheric circulation pattern, a conclusion that was already introduced in Kwok [2006]. Yamagami, Matsueda, Tanaka [2017] found that Arctic cyclones have become more frequent in recent years and predicted that such events would further increase the ice mobility and deformation. The BSG, a wind-driven ocean current located in the west Arctic, keeps the sea ice drifting for many years, hence the area features MYI and also thick FYI (mostly mechanically thick). The impact of sea ice loss on the intensity of the BSG has not been conclusive. Nevertheless, Armitage et al. [2020] presented observational estimates of the BSG and concluded that the recent loss of sea ice and accelerated ocean currents would lead to an increasingly energetic gyre with eddies playing a greater role in its stabilization. This means that western Arctic region will continue to host old and deformed ice although it has lost sizable amount of old ice. The fresh water supply from the river network surrounding the area will continue to increase. Inversely, the impact of Arctic ice cover on the AO, including surface air temperature, has been addressed in a few studies. Prinsenberg et al. [1997] reported local response of the circulation to changes of sea ice in subArctic seas, while Deser, Walsh, Timlin [2000] found same response to change of sea ice east of Greenland. Results from a modeling study by Fan, Huang, Zhou [2019] suggested more occurrence of negative AO in response to the largest sea ice loss in 2007. The role of the Arctic sea ice in modifying surface air temperature, precipitation patterns, and storm track behavior is summarized in Budikova [2009]. In all of the above-mentioned studies the Arctic-wide ice cover was considered as single entity with no distinction between ice types or surface features such as fractures, ridging, or polynyas. Ice with different thickness and surface topography should modulate local wind in different ways. The overall picture can be simplified, but the details are quite complicated.

13.4.1.5. Mutual Interactions Between the Arctic Sea Ice Cover and Oceanic Forcing While ice outflows from the Arctic Ocean are always heading to the Atlantic Ocean through Fram Strait and the waterway network within the CAA, there are three gateways that feature seawater inflow into the Arctic Ocean. The first is the Pacific-Arctic gateway of Bering Strait (through which 80 × 104 m3/s of water enters the Arctic) (Figure 13.6). It varies seasonally in volume, salinity, and temperature [Woodgate, Aagaard, Weingartner, 2005]. The second is the Atlantic gateway through Fram Strait (700 × 104 m3/s of water enters the Arctic) combined with the Barents Sea (100 × 104 m3/s of water enters the Arctic). This is usually warmer and saltier water

[Beszczynska-Möller et al., 2011]. Recirculation of the current in the Fram Strait usually returns nearly half of the water volume to the ocean. The third is the much smaller, Eurasian and Russian rivers, through which 10 × 104 m3/s of water enters the Arctic. Temperature and salinity of the seawater inflow affect the sea ice growth and extent. Growth is delayed by intrusion of warmer and saltier water. While the water volume provided by the Pacific inflow is relatively small compared to that of the Atlantic, the heat provided by the former is higher. Hence, the effect of the warm oceanic water on sea ice growth can be demonstrated better using data from the Bering Sea. Sea ice concentration anomaly was observed in the MIZ around the Pacific water in the winter of 2017. Wang, Liu, Zhang [2021] found that the area with anomalous sea ice reduction corresponded to the area where the warm Pacific Summer Water was observed. The temperature of the mixing layer (25–40 m under the surface) started to increase since late May 2016, and the convective mixing of the seawater brought heat to the upper layer. The authors suggested that the water pathways through the Bering Strait were acting as a conduit for heat entering into the Chukchi Sea. The temperature and heat content of the inflow were sufficiently high in September 2017, to reduce ice growth in the following winter by as much as 0.94 m. While changes in oceanic forcing have minor effect on sea ice growth, the impact of global warming on modulation of inflow of warm water into the Arctic Basin is yet to be quantified. Reciprocally, sea ice affects the ocean current in the North Atlantic, which is part of the global thermohaline circulation (THC) system (section 1.4.4). To reiterate the main feature of this system, the circulation starts at the eastern Arctic region (Greenland Sea) when sea ice forms and brine continues to drain heavily, causing the underline water to be saltier and denser. This water sinks and drives currents in the upper 100 m of the ocean. To close the THC, equatorial ocean areas (mainly the Atlantic and Indian oceans) feed warm water back to the Arctic and this affects the sea ice formation. So, what happens to the THC, and hence to the Arctic sea ice, as global warming continues? As seasonal sea ice melts, the upper layer of the ocean becomes saltier and this causes the ocean current to be faster and more turbulent. However, as an inevitable result of global warming, if sea ice ceases to form or the Greenland glaciers rapidly melt, the upper layer of the Atlantic Ocean will gradually be fresher (less salty), and this causes the part of the THC to slow or eventually stop altogether. This effect will be more pronounced in the Atlantic meridional overturning circulation (AMOC), which is part of the THC circulation. Rahmstorf et al. [2015] used a multi-proxy temperature reconstruction for AMOC and suggested that it has weakened after 1975. The study

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

predicts more weakening as further melting of Greenland glaciers continues in the coming decades. Obviously, the role of sea ice on the decline of the AMOC is minor compared to that of glacier melt though the relative weight has not been quantified in a study. A well-known projected consequence of the complete shutdown of the THC entails that the average temperature in north Europe will cool by 5–10 C. Sea ice/THC interaction constitutes negative feedback. As the THC weakens, air will be cooler and therefore sea water can freeze faster.

13.4.2. Interaction of the Antarctic Sea Ice with the Environment In spite of the expected consequences of global warming and the results from climate models that project a decreasing trend of sea ice in both polar regions, the Antarctic SIE has not shown a decline trend similar to the Arctic (section 13.3.1). Interactions between the Antarctic sea ice on one hand and meteorological, glaciological, and oceanic factors on the other hand should provide clues to explain the difference between the sea ice behaviors in the two polar regimes in response to global warming. Glaciological factors are manifested in aspects of interactions between sea ice, ice shelves, and icebergs. Apparently, those factors are yet to be presented better in the Antarctic sea ice coupling modeling. Possible physical explanations for the deviation of the Antarctic sea ice from the expected decreasing trend under global warming, compiled from several research studies, are presented in Parkinson et al. [2019]. They include links to the ozone hole, El Niño–Southern Oscillation (ENSO), the Interdecadal Pacific Oscillation, and basal fresh meltwater from the ice shelves. Some of those explanations were rejected by later studies. Parkinson et al. [2019] also indicated that the extremely rapid decline of the Antarctic ice extent between 2014–2017 (see Figure 13.9) deserved further studies into the atmospheric and oceanic forcing. Holland and Kwok [2012] suggested changes in atmospheric circulations, temperature, wind stress, ocean currents, and precipitation as being main driving forces for the Antarctic sea ice change. Opposing trends of different factors in different sectors determine the overall behavior of the Antarctic sea ice. The rest of this section addresses aspects of interactions of the Antarctic sea ice with atmospheric forcing, oceanic forcing, and ice shelves/icebergs. The latter constitutes a major factor that changes the sea water temperature and salinity as well as the local ocean currents. This factor is unique to the Antarctic because the number of ice shelves there far exceeds their number in the Arctic (section 13.3.5). Before proceeding, it is worth recalling that the Antarctic ice stretches into warmer latitudes and this leads to substantial melting in summer. This point

585

is important because, unlike the Arctic, even if the Antarctic ice melts completely in the summer, it will barely impact the regional climate. 13.4.2.1. Interaction of the Antarctic Sea ice with Atmospheric Factors The Antarctic sea ice has demonstrated a strong response to winds and atmospheric variability, including large-scale climate variability patterns such as ENSO and the Antarctic Oscillation, which is also called Southern Annular Mode (SAM) [Holland and Kwok, 2012, Raphael and Hobbs, 2014]. The unexpected behavior of Antarctic sea under current global warming has been attributed to change in atmospheric circulation patterns that induces atmospheric cooling [Serreze and Meier, 2019]. Those changes appear to soften the impact of global warming on the Antarctic sea ice. Moreover, large areas of lowpressure systems usually move across the sea ice cover all year-round. They are caused by the topography of Antarctica. Those systems leave their signature on the formation of sea ice, particularly at the ice edge, which features pancake ice. Using satellite-tracked sea ice motion between 1992 and 2010, Holland and Kwok [2012] found statistically strong correlation between ice drift and local winds in most sectors of the Antarctic. The study concluded that wind-driven ice advection is the dominant driver of ice concentration trends around the West Antarctica, whereas wind-driven thermodynamic processes dominate elsewhere. This conclusion raises the question of possible links between wind patterns in the Antarctic region and global anthropogenic forcing. This is still an open scientific question. Raphael and Hobbs [2014] argued that in order to understand the effect of atmospheric circulation pattern on sea ice in the Antarctic, regional SIE data should be considered rather than the extent over the entire Antarctic sea ice. Also, the effect should be examined with reference to seasons defined by ice (e.g., ice formation and retreat seasons), not by atmospheric seasonal cycle. The study of Raphael and Hobbs [2014] uses sea ice concentration data in specific regions and statistically link the ice retreat period with geopotential height data. This allows linking local atmospheric circulation pattern to the variability of sea ice. Circulation patterns of primary importance to sea ice variability have been identified. El Niño and La Niña are opposite weather patterns caused by fluctuations in ocean temperatures in the east-central equatorial Pacific Ocean. They both constitute the ENSO phenomenon. The two events occur every 2 to 7 years, on average, but not on a regular basis. Each event lasts for 9–12 months. El Niño leads to more ice formation in the eastern side of the AP (i.e., Weddell Sea) and less ice on the western side. La Niña leads to the opposite conditions. More on the effects of these two events on the Arctic sea ice is presented in Scott et al. [2019]. Yuan

586

SEA ICE

[2004] studied the effect of the Antarctic Dipole (AD) on the Antarctic sea ice. The AD is a predominant climate mode persisting for 3–4 seasons after being triggered by the ENSO forcing. Using a statistical model, the study reveals a link between the large interannual variability in the Antarctic sea ice concentration to such atmospheric forcing at a seasonal time scale. It demonstrates that sea ice anomalies are predictable with a simple statistical model. Similar to the AO, the Antarctic Oscillation (i.e., SAM) features a large-scale see-sawing atmospheric mass that oscillates between the pole and the mid-latitudes (NASA’s Earth Observatory site: https://earthobservatory.nasa.gov/features/SeaIce/page4.php). The oscillations influence wind, temperature, and storm tracks. Consequently, they influence SIE. During positive phases of the SAM, prevailing westerly winds encircle Antarctica. The strengthening of these winds isolates much of the continent and tends to have an overall cooling effect though it causes dramatic warming on the AP. Winds may drive the ice away from the coast in some areas and toward the coast in others. The ozone hole that develops over Antarctica intensifies a vortex of winds that circles the South Pole. This vortex isolates the Antarctic atmosphere from the rest of the planet. Turner et al. [2009b] suggested that the increase of the Antarctic SIE (largest in autumn) was primarily a result of stronger cyclonic atmospheric circulation, which was mainly a result of stratospheric ozone depletion. This suggestion, however, was rejected by Sigmond and Fyfe [2010]. 13.4.2.2. Interaction of the Antarctic Sea Ice with Oceanic Forcing Oceanic circulation provides a clue to explain differences in sea ice behavior between the two polar regions. In the Arctic, the north of the Atlantic Ocean is open to the warmer waters reaching from the south following the THC (Figure 1.8). This water hinders the sea ice formation in areas in the North Atlantic. In contrast, in the Antarctic there is ocean current, known as the Antarctic Circumpolar Current (ACC), which tends to flow around the continent clockwise in a west-to-east direction (Figure 13.24). This circumpolar pattern is facilitated by the lack of any landmass connecting with Antarctica. Wind also follows same pattern, hence both ACC and wind act as a barricade that blocks warmer air and water coming from the north. This helps maintaining the sea ice cover but more importantly the massive ice sheets. Part of the reason the Antarctic sea ice has remained nearly unchanged under the current global warming is because, since 1950, the surface of the Southern Ocean has warmed at a slower rate (0.02 C per decade) than the global ocean (0.08 C per decade) [The National Academies of Sciences, Engineering and Medicine, 2017]. At

0° 30°W

30°E

60°W

60°E

90°W

90°E

120°E

120°W

150°W

150°E 180°

Figure 13.24 Pathway of the ACC (rotating clockwise). Modern coastlines are overlaid on top of the gray continental background (courtesy of the Open University, UK).

the south side of the ACC shown in Figure 13.24, the SST has warmed only 25% as much as elsewhere around the global ocean. Armour et al. [2016] analyzed oceanographic observations and general circulation model simulations and emphasized the delayed warming south of the ACC and enhanced warming to the north (Figure 13.25). The striking result from this study is about the response of Southern Ocean to greenhouse gas forcing. The study suggested that the upwelled deep ocean waters are warmed at centennial or longer timescale. Mechanisms for this slow warming are proposed. One mechanism entails inefficient deep ocean heat uptake due to deep mixed layers and enhanced upper ocean stratification [IPCC, 2013]. A positive feedback cycle between sea ice cover and ocean heat flux is proposed in Goosse and Zunz [2014]. An increase in sea ice cover causes increase in freshwater input to the ocean during the spring melt. This, in turn, reduces heat flux from deep ocean to the surface, which results in cooler and fresher surface water. Accordingly, sea ice formation and extent are maintained in the following season. This scenario is based on the same principle of reduction/delay of upwelled warm deep water (described above). Ice shelf melting is a third input added to trigger this mechanism, as explained later. While most climate model simulations have predicted decline of sea ice in the Antarctic (similar to the decline of the Arctic ice), Rackow et al. [2022] presented a model that succeeded to simulate the stable Antarctic SIE in

IMPACTS OF CLIMATE CHANGE ON POLAR ICE

587

melt of ice shelves is 1325±235 Gt/y, which exceeds the calving flux of 1089±139 Gt/y. Each process is further explained in the following. The net freshwater produced from melting ice shelves in the Antarctic region is equivalent to about 75% of the annual mean precipitation in the shelf areas. A few models have been used to better understand the effect of ice shelf melting on the Antarctic sea ice formation. One of the early models was presented in Hellmer [2004]. It used two configurations with and without caverns beneath major ice shelves. Results show that meltwater flux due to deep basal melting contributes to sea ice increase in two ways: (1) cooling and freshening the surface waters, and (2) stabilizing the shelf water column in front and downstream the ice shelf. In the absence of freshwater from caverns, sea ice becomes thinner. Bintanja et al. [2013] simulated processes associated with ice shelf melt in a coupled ocean/ice model to determine how the freshwater melt impacted sea ice in the Antarctic region. The authors argued that the water column around the shelf is stabilized between warm deep water and cool and fresh surface water. Results have shown that –0.04 –0.02 0.0 0.02 0.04 this is the main driver for the observed increase of SIE, °C/decade more than any atmospheric factor. However, local changes in atmospheric patterns likely govern regional Figure 13.25 Annual mean sea surface temperature per decade. sea ice trends. Pauling et al. [2016] raised an important The black contour represents the maximum winter sea ice extent point: the rate of rain falling in the Southern Ocean is [Armour et al., 2016]. about 25,000 GT/y, which is nearly 500 times greater than the rate of freshwater melt from ice shelves that Bintanja September. The high-resolution model accounts for et al. [2013] found sufficient to cause sea ice to expand. Southern Ocean mesoscale eddies and projected that sea This casts doubt on results and identifies a continuing ice would not decline before the mid-twenty-first century. need to model the impact of freshwater melt on sea ice. The authors argued that representation of ocean eddies in In Pauling et al. [2017], the authors used CMIPS model the model is necessary to simulate the enhanced equator- with input of 45 GT/y of freshwater from ice shelves, ward heat transport, hence delaying sea ice decline. In i.e., same order of mass balance of the Antarctic ice sheets other words, this oceanic process has become more effi- and shelves, and found it was sufficient to offset the cient at moderating the anthropogenic warming around expected negative trend in sea ice area due to anthropoAntarctica. genic warming. The reciprocal effect of sea ice on glaciers and ice shelf is 13.4.2.3. Interaction Between the Antarctic Sea manifested in the following three aspects: (1) suppressing the ocean waves, which otherwise accelerate breakup Ice, Ice Shelves, and Icebergs As mentioned earlier, the Antarctic ice shelves fringe of the shelf, (2) providing cooler air over the ice shelves 75% of Antarctica’s coastline. They are fed by glacier when the air originating over the ocean passes over the ice through fast-flowing ice streams, snow fall, and, to ice, (3) acting as buttress of ice shelves. Given these much less extent, freezing of ice to their underside. Two impacts, the increasing number of ice shelf collapse processes that possibly delay the response of the Antarctic observed recently, especially in the AP, may be partly sea ice to global warming are: (1) the melting of ice shelves attributed to the decline of sea ice at coastal locations (basal and side melting), and (2) the floating and melting of Antarctica. This possibility warrants more research of icebergs, calved from ice shelves, within the sea ice studies, and it is easy to verify using a reasonable record cover. Both processes cause cooling of the sea water sur- of satellite remote sensing data. Interaction of icebergs with sea ice is manifested in two face and at depth. Cold water sinks and reinforces oceanic forcing. The result is manifested in increasing the sea ice processes. The first is thermodynamically driven, triggrowth and the formation of platelet sea ice structure gered by the continuous melting of the iceberg, and (see section 5.3.3.8). Rignot et al. [2013] found that basal the second is dynamically driven, triggered by the

588

SEA ICE

motion of the iceberg. The thermodynamic aspect is demonstrated in the continuous side and basal melting, which makes the surrounding sea water colder and fresher. This affects the formation, growth, extent, thickness, and the crystallographic structure of sea ice. The mechanical aspect is manifested in modification of the sea ice cover along the iceberg motion path. The two aspects are further discussed in the following. Impacts of iceberg melt on climate of the Southern Ocean and Antarctic sea ice are reviewed in Merino et al. [2016]. The authors modeled iceberg freshwater flux with improved representation of vertical profiles of ocean currents and climatological iceberg fresh water flux. They concluded that iceberg melt increases sea ice formation and volume, but about 10% annually yet with a few local deviations. Sea surface temperature also decreases as a result of increasing iceberg melting while drifting through sea ice. The reverse effect of sea ice cover on the dynamics and melting of icebergs is presented in Li et al. [2018]. The authors analyzed the evolution of two tabular icebergs using a continuous record of multisource radar remote sensing images (Envisat ASAR and Cryosat-2). They observed two patterns of iceberg area decrease while drifting. In austral summer, iceberg area shows a gradual decrease at a rate between 8.0 and 16.0 km2/month due to edge wasting, sidewall melting, and wave erosion. The freeboard also decreases at different rates at different locations (between −0.02 and −0.82 m/month). In general, the subject of iceberg interaction with the Antarctic sea ice is still open for more scientific research using longer remote sensing records and accurate in-situ measurements of sea ice surface temperature and salinity. Jongma et al. [2009] presented a global climate model coupled with a dynamic thermodynamic iceberg model and concluded that both fresh water and cooling flux increases the sea ice area by 12% and 6%, respectively. Rackow et al. [2017] complemented a global ocean seaice model (focused on the Southern Ocean) with an iceberg drift and decay model. The authors presented a comprehensive study of iceberg dynamics and melting under different sea ice, wind, and oceanic conditions and through different seasonal and interannual variability. The study highlighted information about the different dynamics and trajectories of small and giant icebergs and found that freshwater flux from melting icebergs partly compensated for the brine rejection from sea ice in the annual mean. Icebergs are tracked under the influence of winds, currents and sea surface tilt, and sea ice forcing. They cause the ice cover to deform or turn into open water. The ice at the windward side is crushed and thickens while open water is formed at the lee side. This modifies the ice thickness distribution. If an iceberg becomes grounded, a polynya or fast ice is formed depending on the dominant wind direction. Hunke and Ackley [2001] confirmed the

presence of polynyas downstream of large grounded icebergs in the Weddell Sea. Grounded icebergs appear to be responsible also for unusual large fast ice formation when sea ice becomes attached (fastened) to the side of the berg. Brunt, Sergienko, MacAyeal [2006] explored this intuitive speculation by tracking the effect of massive tabular icebergs that entered the southwest Ross Sea, in early 2001, and remained immobile for 4 years. Extensive fast ice was formed around the icebergs and along the Victoria Land coast (for geographic location, see Figure 13.7) with coverage of about five times larger than observed prior to the arrival of the icebergs. Brunt, Sergienko, MacAyeal [2006] concluded also that the interplay between iceberg location, movement, and local storms determines the fast ice variability. Hunke and Comeau [2011] incorporated an iceberg dynamic component in the Los Alamos sea ice model (CICE) to study anomalies of sea ice cover in the presence of icebergs. The study found that the net effect of icebergs on sea ice cover would be insignificant if considered within the context of the overall climatic influence on ice production. It may be important only for small scale processes (e.g., ridging, openings, and fast ice formation). Li et al. [2020] presented Sentinel-1 SAR images that showed fast ice with its offshore end attached to a grounded iceberg in front of Filchner ice shelf in the Antarctic region (the other end is attached to the land). When the iceberg started to move, the fast ice detached. This is shown in Figure 13.26 in which the offshore end of the fast ice is shown to extend and become attached to the A23A iceberg when it was grounded until 2006 (the cyan color contour). After that, the iceberg started drifting slowly northward and the area between it and the fast ice was filled with mobile ice. The fast ice remained bonded at its south edge to the ice shelf. When the iceberg drift accelerated in 2010, the fast ice bond failed under influence of other forces as shown by the green contour in the image. The gap was filled later and ice resumed its attachment to the ice shelf. The underlying theme is that icebergs can play a significant role in the development of landfast ice in the Antarctic. Aspects of sea ice influence on iceberg include speed, drag, or motion capturing under high sea ice concentration and strength. On the other hand, iceberg motion may cause sea ice deformation, polynya formation, or development of fast ice. So far, many observational and modeling studies have been dedicated to explore the impacts of sea ice on the iceberg dynamics. The opposite interaction (i.e., the influence of the iceberg on the characteristics of the surrounding sea ice cover) has received less attention. Given the view that the current increase of icebergs in the Antarctic region is a manifestation of global warming and that they apparently affect the sea

IMPACTS OF CLIMATE CHANGE ON POLAR ICE 40°W

589

35°W 2006

74°S

2007 2008 2009 2010 2011 2016

76°S

78°S

2017

55°W

50°W

45°W

Figure 13.26 Progression of the outer contour of landfast ice during 2006–2011 and 2016–2017, north of Filchner ice shelf in Antarctica overlaid on a Sentinel-1 image acquired on 23 November 2016. The ice shelf A23A is shown in the image and its motion track northward is displayed in the inset. The north edge of the landfast ice was attached to the iceberg until it moved north in 2010. After that it detached from the iceberg (image courtesy of Xinqing Li of Sun Yat-sen University, China).

Number of calving events

250

125

0 Year

20

20

05

20

06

/0

6

20

07

/0

7

8

20

20

08

/0

09

/0

9

20

10

/1

0

/1

20

11

1

20

12

/1

2

13

/1

3

20 20 20 20 20 20 14 15 16 17 18 19 / / / / / / 4 15 16 17 18 19 20

/1

Figure 13.27 Number of calved icebergs in the Antarctic region between 2005–2020. The bluish color marks a period of weak El Niño and the pinkish color marks a period of strong El Niño (figure courtesy of Yan Liu, Beijing Normal University).

ice cover, more modeling studies are needed to address this point. It would be useful to conclude this section with information that links the number of icebergs calving in the Antarctic region to the El Niño south oscillation phenomenon (section 13.4.2.1). Qi et al. [2021] published a database on number, distributions, and characteristics of circumpolar

Antarctic iceberg calving. Data were derived from satellite observations that measure every calving event larger than 1 km2 between August 2005 and August 2020 (this is a finer resolution than the database generated at BYU, from which a sample is displayed in Figure 13.17). The technique is described in the published study. A total of 1975 calving events were detected.

590

SEA ICE

Figure 13.27 shows the variation of the annual number of calving events during the study period. Interannual variability can be seen linked to the strength of the El Niño. When the El Niño was weak in January 2007, causing low sea surface temperature anomaly, the number of iceberg calving was low (the blue stripe in the figure). Conversely, when the El Niño was strong during the period 2014–2015, featuring high sea surface temperature anomaly in the equatorial Pacific Ocean, the number of calving peaked (the pink stripe in the figure). 13.5. REFERENCES Armitage, T.W.K. et al. (2020) Enhanced eddy activity in the Beaufort Gyre in response to sea ice loss, Nature Communications, 11(1), 761. Available from: doi: 10.1038/s41467-02014449-z. Armour, K.C. et al. (2016) Southern Ocean warming delayed by circumpolar upwelling and equatorward transport, Nature Geoscience, 9, pp. 549–554. Available from: doi:10.1038/ ngeo2731. Ballinger, T.J. et al. (2021) Surface air temperature, In: Moon, T.A., Druckenmiller, M.L. and Thoman, R.L., eds. Arctic Report Card 2021. Available from: https://repository.library. noaa.gov/view/noaa/34308. Barber et al. (2018) Increasing mobility of high arctic sea ice increases marine hazards off the East Coast of Newfoundland, Geophysical Research Letters, 45, pp. 2370–2379. Available from: http://doi.org/10.1002/2017GL076587. Beszczynska-Möller, A. et al. (2011) A synthesis of exchanges through the main oceanic gateways to the Arctic Ocean, Oceanography, 24, pp. 82–99. Available from: doi:10.5670/ oceanog.2011.59. Bintanja, R. et al. (2013) Important role for ocean warming and increased ice-shelf melt in Antarctic sea ice expansion, Nature Geoscience, 6(5), pp. 376–379. Available from: doi: 10.1038/ ngeo1767. Blockley, E. et al. (2020) The future of sea ice modeling: where do we go from here?, Bulletin of the American Meteorological Society, 101(8). Bozkurt, D. et al. (2020) Recent near-surface temperature trends in the Antarctic Peninsula from observed, reanalysis and regional climate model data, Advances in Atmospheric Sciences, 37, pp. 477–493. Brunt, K.M., Sergienko, O. and MacAyeal, D.R. (2006) Observations of unusual fast-ice conditions in the southwest Ross Sea, Antarctica: Preliminary analysis of iceberg and storminess effects, Annals of Glaciology, 44, pp.183–187. Budikova, D. (2009) Role of Arctic sea ice in global atmospheric circulation: A review, Global and Planetary Change, 68, pp. 149–163. Cai, Q. et al. (2021) Accelerated decline of Arctic sea ice during 1850–2017 and the amplified Arctic warming during the recent decades, Environment Research Letters, 16, 034015. Carrasco, J.F., Bozkurt, D. and Cordero, R.R. (2021) A review of the observed air temperature in the Antarctic Peninsula.

Did the warming trend come back after the early 21st hiatus?, Polar Science, 28, 100653. Chen, Z. et al. (2021) Fingerprint of COVID-19 in Arctic sea ice changes, Science Bulletin, 66(20). Available from: doi:10.1016/j.scib.2021.06.009. Comiso, J.C. (2000) Variability and trends in Antarctic surface temperatures from in situ and satellite infrared measurements, Journal of Climate, 13, pp. 1674–1696. Comiso, J.C. et al. (2017) Positive trend in the Antarctic sea ice cover and associated changes in surface temperature, Journal of Climate, 30(6), pp. 2251–2267. Deser, C. and Teng, H. (2008) Evolution of Arctic sea ice concentration trends and the role of atmospheric circulation forcing, 1979–2007. Journal of Geophysical Research, L02504. Available from: doi:10.1029/2007GL032023. Deser, C., Walsh, J.E. and Timlin, M. (2000) Arctic sea ice variability in the context of recent atmospheric circulation trends, Journal of Climate, 13, pp. 617–633. Ding, Q. et al. (2017) Influence of high-latitude atmospheric circulation changes on summertime Arctic sea ice, Nature Climate Change, 7, pp. 289–295. Available from: doi:10.1038/ NCLIMATE3241. Eisenman, I., Meier, W.N. and Norris, J.R. (2014) A spurious jump in the satellite record: Has Antarctic sea ice expansion been overestimated, The Cryosphere, 8, pp. 1289–1296. England, M.R., Wagner, T.J.W. and Eisenman, I. (2020) Modeling the breakup of tabular icebergs, Science Advances, 6, eabd1273. Fan, T., Huang, F. and Zhou, X. (2019) Response of the dominant modes of atmospheric circulation in the Northern Hemisphere to a projected Arctic sea ice loss in 2007, Journal of Ocean University of China, 18(3), pp. 589–595. Goosse, H. and Zunz, V. (2014) Decadal trends in the Antarctic sea ice extent ultimately controlled by ice-ocean feedback, The Cryosphere, 8(2), pp. 453–470. Available from: doi:10.5194/ tc-8-453-2014. Guarino, M.V., et al. (2020) Sea-ice free Arctic during the last interglacial supports fast future loss, Nature Climate Change, 10, pp. 928–932. Hansen, E. et al. (2013) Thinning of Arctic sea ice observed in Fram Strait: 1990–2011, Journal of Geophysical Research: Ocean, 118(10), pp. 5201–5221. Hellmer, H.H. (2004) Impact of Antarctic ice shelf basal melting on sea ice and deep ocean properties, Geophysical Research Letters, 31(10), L10307. Available from: doi:10.1029/ 2004GL019506. Holland, P.R. and Kwok, R. (2012) Wind-driven trends in Antarctic sea ice drift, Nature Geoscience, 5, pp. 872–875. Available from: doi:10.1038/ngeo1627. Holland, P.R. et al. (2014) Modeled trends in Antarctic sea ice, Journal of Climate, 27(10), pp. 3784–3801. Howell, S.E.L. et al. (2013) Recent changes in the exchange of sea ice between the Arctic Ocean and the Canadian Arctic Archipelago, Journal of Geophysical Research: Oceans, 118. Available from: doi:10.1002/jgrc.20265. Hu, X. and Myers, P.G. (2014) Changes to the Canadian Arctic Archipelago sea ice and freshwater fluxes in the twenty-first century under the Intergovernmental Panel on climate change

IMPACTS OF CLIMATE CHANGE ON POLAR ICE A1B climate scenario, Atmosphere–Ocean, 52(4), pp. 331–350. Hunke, E.C. and Ackley, S.F. (2001) A numerical investigation of the1997–98 Ronne Polynya, Journal of Geophysical Research: Oceans, 106, pp. 22,373–22,382. Hunke, E.C. and Comeau, D. (2011) Sea ice and iceberg dynamic interaction, Journal of Geophysical Research: Ocean, 116, C05008. Available from: doi:10.1029/2010JC006588. Hunke, E.C. et al. (2015) CICE: The Los Alamos sea ice model documentation and software user’s manual version 5.1, LA-CC06-012, Los Alamos, NM, Los Alamos National Laboratory. IPCC (2013) Climate change 2013: The physical science basis, In: Stocker, D. Qin et al., eds. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press. IPCC (2019) Climate change and land, Special Report on the ocean and cryosphere in changing climate, Cambridge, UK: Cambridge University Press. Available from: doi.org/ 10.1017/9781009157964. Jefferies, M.O., ed. (1998) Antarctic sea ice: Physical processes, interactions and variability, Antarctic Research Series, Vol. 74, Washington, D.C., USA: American Geophysical Union, 407 p. Johannessen, O.M. et al., eds. (2020) Sea ice in the Arctic: Past, present and future, Springer Polar Series, Springer International Publishing. ISBN-13: 978-3-030-21300-8, 575 p. Available from: doi: 10.1007/978-3-030-21301-5. Jongma, J.I. et al. (2009) The effect of dynamic-thermodynamic icebergs on the Southern Ocean climate in a three-dimensional model, Ocean Modelling, 26(1), pp. 104–113. Kacimi, S. and Kwok, R. (2020) The Antarctic sea ice cover from ICESat-2 and CryoSat-2: Freeboard, snow depth, and ice thickness, The Cryosphere, 14(12), pp. 4453–4474. Available from: https://doi.org/10.5194/tc-14-4453-2020. Kashiwase, H. et al. (2017) Evidence for ice-ocean albedo feedback in the Arctic Ocean shifting to a seasonal ice zone, Scientific Reports, 7, 8170. Available from: https://doi.org/ 10.1038/s41598-017-08467-z. Kaur, S., Ehn, J.K. and Barber, D.G. (2018) Pan-arctic winter drift speeds and changing patterns of sea ice motion: 1979–2015, Polar Records, 54(5-6), pp. 303–311. Korosov, A.A. et al. (2018) A new tracking algorithm for sea ice age distribution estimation, The Cryosphere, 12(6), pp. 2073– 2085. Available from: https://doi.org/10.5194/tc-12-2073-2018. Kuang, H. et al. (2022) Arctic multiyear ice areal flux and its connection with large-scale atmospheric circulations in the winters of 2002–2021, Remote Sensing, 14, p. 3742. Available from: https://doi.org/10.3390/rs14153742. Kurtz, N.T. and Markus, T. (2012) Satellite observations of Antarctic sea ice thickness and volume, Journal of Geophysical Research: Oceans, 117(C8). Available from: doi:10.1029/ 2012JC008141. Kwok, R. (2006) Contrasts in sea ice deformation and production in the Arctic seasonal and perennial ice zones, Journal of Geophysical Research, 111(C11), C11S22. Kwok, R. (2018) Arctic sea ice thickness, volume and multiyear ice coverage: Losses and coupled variability, Environmental Research Letters, 13, p. 105005.

591

Kwok, R., Spreen, G. and Pang, S. (2013) Arctic sea ice circulation and drift speed: Decadal trends and ocean currents, Journal of Geophysical Research, 118(C5), pp. 2408–2425. Kwok, R. et al. (2010) Large sea ice outflow into Nares Strait in 2007, Geophysical Research Letters, 37, L03502. Available from: https://doi.org/10.1029/2009GL041872. Labe, Z., Magnusdottir, G. and Stern, H. (2018) Variability of Arctic sea ice thickness using PIOMAS and CESM large ensemble, Journal of Climate, 31, pp. 3233–3247. Langehaug, H.R. et al. (2013) Arctic sea ice decline and ice export in the CMIP5 historical simulations, Ocean Modeling, 71, pp. 114–126. Lei, R. et al. (2020) Annual cycles of sea ice motion and deformation derived from buoy measurements in the Western Arctic Ocean over two ice seasons, Journal of Geophysical Research: Oceans, 12(6), e2019JC015310. Letcher, T.M., ed. (2021) The impacts of climate change: a comprehensive study of physical, biophysical, social and political issues, Elsevier Publisher, ISBN-10: 0-128-22373-1, 1338 p. Li, T. et al. (2018) Monitoring the tabular icebergs C28A and C28B calved from the Mertz Ice Tongue using radar remote sensing data, Remote Sensing of Environment, 2016, pp. 615–635. Li, X. et al. (2020) The spatio-temporal patterns of landfast ice in Antarctica during 2006–2011, and 2016–2017 using highresolution SAR imagery, Remote Sensing of Environment, 242, 111736. Lindsay, R.W. and Schweiger, A. (2015) Arctic sea ice thickness loss determined using subsurface, aircraft, and satellite observations, Cryosphere, 9, pp. 269–283. Available from: https:// doi.org/10.5194/tc-9-269-2015. Liu, Y. et al. (2015) Ocean-driven thinning enhances iceberg calving and retreat of Antarctic ice shelves, Proceedings of the National Academy of Sciences of the United States of America (PNAS), 112, pp. 440, 3263–3268. Available from: doi:10.1073/pnas.1415137112. Lüpkes C. et al. (2008) Influence of leads in the sea ice on the temperature of the atmosphere boundary layer during polar night, Geophysical Research Letters, 35, L03805. Available from: doi:10.1029/2007GL032461. Maeda, K., Kimura, N. and Yamaguchi, H. (2020) Temporal and spatial change in the relationship between sea ice motion and wind in the Arctic, Polar Research, 39, p. 3370. Mackie, S. et al. (2020) Sea ice formation in a coupled climate model including grease ice, Journal of Advances in Modeling Earth Systems, 12(8), e2020MS002103. Mallett, R.D.C. et al. (2021) Faster decline and higher variability in the sea ice thickness of the marginal Arctic seas when accounting for dynamic snow cover, The Cryosphere, 15, pp. 2429–2450. Maslanik, J.A. et al. (2007) A younger, thinner Arctic ice cover: Increased potential for rapid, extensive sea-ice loss, Geophysical Research Letters, 34, L24501. Available from: doi:10:1029/2007/GL032043. Massom, R.A. and Stammerjohn, S.E. (2010) Antarctic sea ice change and variability–Physical and ecological implications, Polar Science, 4(2), pp.149–158.

592

SEA ICE

Massom, R.A. et al. (2018) Antarctic ice shelf disintegration triggered by sea ice loss and ocean swell, Nature, 558, pp. 383–389. Available from: doi:10.1038/s41586-018-0212-1. McNutt, S.L. and Overland, J.E. (2003) Spatial hierarchy in Arctic sea ice dynamics, Tellus A, 55, pp. 181–191. Meier, W. and Markus, T. (2015) Remote sensing of sea ice, In: Tedesco, M., ed. Remote sensing of the cryosphere, New York, NY: Wiley Blackwell, pp. 248–272. Meier, W.N. et al. (2014) Arctic sea ice in transformation: A review of recent observed changes and impacts on biology and human activity, Reviews of Geophysics, 51, pp. 185–217. Available from: doi:10.1002/2013RG000431. Meiners, K.M. et al. (2016) SIPEX-2: A study of sea-ice physical, biochemical and ecosystem processes off East Antarctica during spring 2012, Deep-Sea Research Part II-Topical Studies in Oceanography, (131), pp. 1–6. Merino, N. et al. (2016) Antarctic icebergs melt over the Southern Ocean: Climatology and impact of sea ice, Ocean Modeling, 104, pp. 99–110. NOAA (2021) Arctic report card: Update for 2021, NOAA online publication. Available from: https://arctic.noaa.gov/ Portals/7/ArcticReportCard/Documents/ArcticReportCard_ full_report2021.pdf Notz, D. and Stroeve, J. (2016) Arctic sea ice loss directly follows cumulative anthropogenic CO2 emissions, Science, 354. Olason, E. and Notz, D. (2014) Drivers of variability in Arctic sea ice drift speed, Journal of Geophysical Research: Oceans, 119(9), pp. 5755–5775. Available from: https://doi.org/ 10.1002/2014JC009897. Olonscheck, D., Mauritsen, T. and Notz, D. (2019) Arctic sea ice variability is primarily driven by atmospheric temperature fluctuations, Nature Geoscience, 12, pp. 430–434. Overland, J.E. and Wang, M. (2010) Large-sale atmospheric circulation changes are associated with the recent loss of Arctic sea ice, Tellus A: Dynamic Meteorology and Oceanography, 62(1), pp. 1–9. Available from: doi:10.1111/j.16000870.2009.00421.x. Parkinson, C.L. (2019) A 40-y record reveals gradual Antarctic sea ice increases followed by decreases at rates far exceeding the rates seen in the Arctic, PNAS, Proceedings of the National Academy of Sciences of the USA, 116(29), pp. 14,414–14,423. Parkinson, C.L. and Cavalieri, D.J. (2012) Antarctic sea ice variability and trends, 1979–2010, The Cryosphere, 8, pp. 871– 889. Available from: doi:10.5194/tc-6-871-2012. Pauling, A.G. (2017) Time-dependent fresh water input from ice shelves: Impacts on Antarctic sea ice and the Southern Ocean in an Earth System Model, Geophysical Research Letters, 44, pp. 10,454–10,461. Available from: https://doi.org/10.1002/ 2017GL075017. Pauling, A.G. et al. (2016) The response of Southern Ocean and Antarctic sea ice to freshwater from ice shelves in an earth system model, Journal of Climate, 29(5), pp. 1655–1672. Pope, A. et al. (2017) Synthesis community review of Southern Ocean satellite data needs, Antarctic Science, 29(2), pp. 97–138. Prinsenberg, S.J. et al. (1997) Interaction between atmosphere, ice cover, and ocean off Labrador and Newfoundland from

1962–1992, Canadian Journal of Aquatic Sciences, 54, pp. 30–39. Qi, M. et al. (2021) A 15-year circum-Antarctic iceberg calving dataset derived from continuous satellite observations, Earth System Science Data, 13, pp. 4583–4601. Rackow, T. et al. (2017) A simulation of small to giant Antarctic iceberg evolution: Differential impact on climatology estimates, Journal of Geophysical Research: Ocean, 122(4), pp. 3170–3190. Rackow, T. et al. (2022) Delayed Antarctic sea-ice decline in high-resolution climate change simulation, Nature Communication, 13(637). Available from: https://doi.org/10.1038/ s41467-022-28259-y. Rahmstorf, S. et al. (2015) Exceptional twentieth-century slowdown in Atlantic Ocean overturning circulation, Nature Climate Change, 5(5), pp. 475–480. Rampal, P., Weiss, J. and Marsan, D. (2009) Positive trends in the mean speed and deformation rate of Arctic sea ice, 1979–2007, Journal of Geophysical Research: Oceans, 114, C05013. Available from: doi:10.1029/2008JC005066. Raphael, M.N. and Hobbs, W. (2014) The influence of the largescale atmospheric circulation on Antarctic sea ice during ice advance and retreat seasons, Geophysical Research Letters, 41. Available from: doi: 10.1002/2014GL060365 Rignot, E. et al. (2013) Ice-shelf melting around Antarctica, Science, 341(6143), pp. 266–270. Available from: https://doi.org/ 10.1126/science.1235798. Rigor, I.G., Wallace, J.M. and Colony, R.L. (2002) Response of sea ice to the Arctic oscillation, Journal of Climate, 15(18), pp. 2648–2663. Scott, R.C. et al. (2019) Meteorological drivers and large-scale climate forcing of West Antarctic surface melt, Journal of Climate, 32, pp. 665–684. Serreze, M.C. and Barry, R.G. (2011) Processes and impacts of Arctic amplification: A research synthesis, Global Planet Change, 77, pp.85–96. Serreze, M.C. and Meier, W. (2019) The Arctic’s sea ice cover: trends, variability, predictability, and comparisons to the Antarctic, Annals of New York Academy of Science, 1436(1), pp. 36–53. Available from: doi:10.1111/nyas.13856. Sigmond, M. and Fyfe, J.C. (2010) Has the ozone hole contributed to increased Antarctic sea ice extent?, Geophysical Research Letters, 37, L18502. Smedsrud, L.H. et al. (2017) Fram Strait sea ice export variability and September Arctic sea ice extent over the last 80 years, The Cryosphere, 11, pp. 65–79. Available from: doi:10.5194/ tc-11-65-2017. Spreen, G., Kwok, R. and Menemenlis, D. (2011) Trends in Arctic sea ice drift and role of wind forcing: 1992–2009, Geophysical Research Letters, 38, L19501. Available from: http://doi. org/10.1029/2011GL048970. Spreen, G. et al. (2009) Fram Strait ice volume export estimated between 2003 and 2008 from satellite data, Geophysical Research Letters, 36(19), L19502. Steig, E.J. et al. (2009) Warming of the Antarctic ice-sheet surface since the 1957 International Geophysical Year, Nature, 457(7228), pp. 459–462.

IMPACTS OF CLIMATE CHANGE ON POLAR ICE Stuart, K.M. and Long, D.G. (2011) Tracking large tabular icebergs using the SeaWinds Ku-band microwave scatterometer, Deep Sea Research, Part II, Topical Studies in Oceanography, 58(11), pp. 1285–1300. Available from: doi:10.1016/j. dsr2.2010.11.004 Stroeve, J. and Notz, D. (2018) Changing state of Arctic sea ice across all seasons, Environment Research Letters, 13(10), p. 103001. Stroeve, J. et al. (2014) Using records from submarine, aircraft and satellites to evaluate climate model simulations of Arctic sea ice thickness, The Cryosphere, 8(5), pp. 1839–1854. Available from: https://doi.org/10.5194/tc-8-1839-2014. The National Academies of Sciences, Engineering, and Medicine (2017) Antarctic sea ice variability in the Southern Ocean-climate system, Washington, D.C.: The National Academies Press. Available from: doi: 10.17226/24696. Thorndike, A.S. and Colony, R. (1982) Sea ice motion in response to geostrophic winds, Journal of Geophysical Research: Oceans, 87(C10), pp. 5845–5852. Tilling, R.L., Ridout, A. and Shepherd, A. (2018) Estimating Arctic sea ice thickness and volume using CryoSat-2 radar altimeter data, Advances in Space Research, 62(6), pp. 1293–1225. Tournadre, J. et al. (2016) Antarctic icebergs distributions 1992–2014, Journal of Geophysical Research: Oceans,121, pp. 327–349. Available from: doi:10.1002/2015JC011178. Travers, C.S. (2012) Quantifying sea-ice volume flux using Moored Instrumentation in the Bering Strait, M.Sc. thesis, College of the Environment: School of Oceanography, University of Washington, Washington, USA. Tschudi, M.A., Meier, W.N. and Stewart, J.S. (2020) An enhancement to sea ice motion and age products at the National Snow and Ice Data Center (NSIDC), The Cryosphere, 14(5), pp. 1519–1536. Tschudi, M.A., Stroeve, J.C. and Stewart, J.S. (2016) Relating the age of arctic sea ice tom its thickness, as measured during NASA’s ICESat and IceBridge campaigns, Remote Sensing, 8, p. 457. Available from: doi.10.3390/re8060457. Turner, J., et al. (2005) Antarctic climate change during the last 50 years, International Journal of Climatology, 25(3), pp. 279–294. Turner, J. et al. (2009a) Antarctic climate change and the environment, Scientific Community on the Antarctic Research, Scott Polar Research Institute, Cambridge, UK. Turner J. et al. (2009b) Non-annular atmospheric circulation change induced by stratospheric ozone depletion and its role in the recent increase of Antarctic sea ice extent, Geophysical Research Letters, 36(L08502). Wang, Y., Liu, N. and Zhang, Z. (2021) Sea ice reduction during winter of 2017 due to oceanic heat supplied by Pacific water in the Chukchi Sea, West Arctic Ocean, Frontier of Marine Science, 8, 650909. Available from: doi:10.3383/ fmars.2021.650909.

593

Wei, J., Zhang, X. and Wang, Z. (2019) Re-examination of Fram Strait sea ice export and its role in recently accelerated Arctic sea ice retreat, Climate Dynamics, 53, pp. 1823–1841. Wettstein, J.J. and Deser, C. (2014) Internal variability in projections of twenty-first century Arctic sea ice loss: Role of the large-scale atmospheric circulation, Journal of Climate 27, pp. 527–550. Woodgate, R.A., Aagaard, K. and Weingartner, T.J. (2005) A year in the physical oceanography of the Chukchi Sea: Moored measurements from autumn 1990–1991, Deep-Sea Research, Part II, 52(24-26), pp. 3116–3149. Worby, A.P. et al. (2008) Evaluation of AMSR-E snow depth product over East Antarctic sea ice using in situ measurements and aerial photography, Journal of Geophysical Research: Oceans, 113(C5). Available from: doi:10.1029/ 2007JC004181. Wu, B., Wang, J. and Walsh, J.E. (2006) Dipole anomaly in the winter Arctic atmosphere and its association with sea ice motion, Journal of Climate, 19, pp. 210–225. Xei, H. et al. (2013) Sea ice thickness estimation from ICESat altimetry over the Bellingshausen and Amundsen seas, 2003–2009, Journal of Geophysical Research, 118, pp. 1–16. Available from: doi:10.1002/jgrc.20179. Yamagami, A., Matsueda, M. and Tanaka, H.L. (2017) Extreme Arctic cyclone in August 2016, Atmospheric Science Letters, 18(7), pp. 307–314. Ye, Y. et al. (2022) Exploring the effect of Arctic perennial sea ice on modulation of local air temperature, Advanced in Climate Change Research, 13(4), pp. 473–488. Yuan, X. (2004) ENSO-related impacts on Antarctic sea ice: A synthesis of phenomenon and mechanisms, Antarctic Science, 16(4), pp. 415–425. Zhang, X. (2008) Recent radical shifts of atmospheric circulations and rapid changes in Arctic climate system, Geophysical Research Letters, 35(L22701). Available from: doi:10.1029.2008GL035607. Zhang, J. and Rothrock, D.A. (2003) Modeling global sea ice with a thickness and enthalpy distribution model in generalized curvilinear coordinates, Monthly Weather Review, 131, pp. 845–861. Zhang, Z. et al. (2019) Arctic sea ice classification using microwave scatterometer and radiometer data during 2002–2017, IEEE Transactions on Geoscience and Remote Sensing, 57 (8), pp. 5319–5328. Zhang, F. et al. (2021) Arctic sea ice motion change and response to atmospheric forcing between 1979 and 2019, International Journal of Climatology, 1–23. Available from: https://doi.org/ 10.1002/joc.7340. Zwally, H.J. et al. (2002) Variability of Antarctic sea ice 1979–1998, Journal of Geophysical Research, 107. Available from: doi:10.1029/2000JC000733.

INDEX Note: Page numbers in italics indicate figures, those in bold indicate tables. AA phenomenon, 561–562, 579–580 a axis, 194 Abel’s equation, 130–131 absorption of EM wave, 296 of microwave signal in snow, 337 of radiation by atmospheric gases, 325, 325–326 accumulated thawing degree-days (ATDDs), 78 active microwave sensors, 286–287, 287 surface melt observations, 434, 434–437, 437–438 advanced horizontal range algorithm (AHRA), 432 Advanced Microwave Scanning Radiometer (AMSR), 339, 340, 353, 354, 398, 427 ice surface temperature, 521, 521 sea ice concentration, 483–486, 484–486 sea ice extent (SIE), 580 sea ice thickness (SIT), 506–508, 508, 508, 509, 513 snow depth, 445, 445–446 Advanced SAR (ASAR), 314, 355, 384–385 identification of sea ice ridges using, 417 sea ice classification, 464 Advanced Scatterometer (ASCAT), 323, 323, 358, 358–359, 435, 436 Advanced Topographic Laser Altimeter System (ATLAS), 360, 423, 513 Advanced Very High- Resolution Radiometer (AVHRR), 269–271, 275, 279, 327 albedo observations, 365, 370 ice surface temperature, 518–521, 521 leads, 422, 422, 424 SAR image compared, 455, 455 sea ice concentration, 473, 474 sea ice thickness, 505, 505–508, 507 sensors, 352, 352 surface melt, 429–430, 430 advection, 62, 87–88, 90 aerial photography ridge identification by, 416 surface melt, 437, 438 age-based sea ice classification, 321–322 age of sea ice. see sea ice age (SIA) agglomerate ice (R type ice), 201, 203, 211–212, 212, 220 aging of sea ice, 29, 259–260 airborne SAR (AIRSAR), 314–315, 318, 467 air bubbles, 47–50, 48–49, 54, 56, 56, 58, 58, 63, 66, 81, 81–82, 194 geometric characteristics, 242–244, 243, 244 in hummock ice, 223–226, 227 in melt pond ice, 227–228 in MYI, 81–82 in seasonal ice, 220–221, 223 aircraft, landing on sea ice, 20–21, 20–21 air dissolved in seawater, 33 air temperature, near-surface (nSAT), 560–561, 560–562 air volume fraction, 124–125 albedo, 13, 23, 301–303, 364–370 atmospheric spherical, 302 black-sky, 301 broadband, 301–302, 364–367, 365, 366, 366–367, 370, 428 ice decay, 77

ice surface melt, 428–430, 431, 437 open leads, 76 scattering, 299 seawater, 302, 335 sediment-laden sea ice, 368, 369 snow, 192, 302, 335–336, 336, 444 spectral, 301, 364–368, 367–368, 370, 428–429 Surface Heat Budget of Arctic Ocean (SHEBA) field experiment, 271–272, 272 white-sky (diffuse), 301 algae, 17, 76, 87, 230–232, 231–232 ALOS-2 SAR, 468, 468–469 ALOS-PALSAR, 314–315, 318, 414, 464 altimeters, 24, 323–325, 324 laser, 323–325, 324 radar, 308, 323–324, 324 sea ice thickness (SIT) observations, 502, 510–514, 511–512, 514 sensors, 359–360, 360 amorphous, 194 anatomy, ice, 10 anchor ice, 200 anisotropic reflectance factors (ARFs), 302 anisotropy, 319, 319–320 annual ice. see seasonal ice Antarctic Circumpolar Current (ACC), 586, 586 Antarctic Oscillation, 585–586 Antarctic region climate change impact, 558–559 geography, 564, 565 growth rate of ice, 41 icebergs, 98, 99, 576–577, 577 ice of glacier origin, 95–96, 99 ice sheet, 558, 561 ice shelves, 96 latitude, 1 marginal ice zone, 94 platelet ice, 203 sea ice characteristics, 564–565, 566 sea ice extent, 565–569, 567–568 sea ice growth, 39 sea ice motion, 533, 533 sea ice salinity, 115, 115 sea ice thickness (SIT), 2, 571–572 snow cover, 45 superimposed ice, 40 surface air temperatures, 561, 561–562 Antarctic sea ice interaction with environment, 585–590 atmospheric factors, 585–586 oceanic forcing, 586–587 between sea ice, ice shelves, and icebergs, 589–590 Arctic amplification (AA), 561–562, 579–580 Arctic Ice Dynamic Joint Experiment (AIDJEX), 250, 250–252 Arctic Ice Mapping (AIM) program, 422, 422 Arctic Oscillation (AO), 579, 583–584, 586 Arctic Radiation and Turbulence Interaction Study (ARTIST), 478 Arctic region climate change impact, 558–559 geography, 562–564, 563

Sea Ice: Physics and Remote Sensing, Second Edition. Mohammed Shokr and Nirmal K. Sinha. © 2023 John Wiley & Sons, Inc. Published 2023 by John Wiley & Sons, Inc. 595

596

INDEX

Arctic region (cont’d) historical and community synopsis, 4–8 ice shelves, 96, 96 latitude, 1 marginal ice zone, 93–94 sea ice age, 572–575, 573, 573–574 sea ice characteristics, 564–565, 566 sea ice dynamics, 575 sea ice extent, 565–569, 567–568 sea ice growth, 39 sea ice motion, 532, 532–533 sea ice salinity, 115, 115 sea ice temperature profiles, 109 sea ice thickness, 2, 14, 569–571, 570 snow cover, 45 surface air temperatures, 13, 560–561, 560–561 wind patterns driving ocean circulation, 524 Arctic sea ice interaction with environment, 578, 578–585 atmospheric factors, 578 sea ice cover and oceanic forcing, 584–585 warming due to sea ice advection, 580–582, 580–583 warming due to sea ice cover changes, 578–580, 579 wind, 582–584 area of sea ice, 498–501, 499, 501 ASI algorithm, 478–479 Atlantic meridional overturning circulation (AMOC), 584–585 Atmosphere Environment Services (AES), 252–253, 253–254, 256, 260 atmospheric correction, 302, 325–327 commercial software packages for, 327 for passive microwave observations, 328–330 atmospheric factors changes in Antarctic sea ice, 585–86 changes in Arctic sea ice, 578 atmospheric oscillation (AO), 14 atmospheric transmission windows, 285, 286, 325, 325–326 backscatter, 376–395 co-polarization data, 378, 379, 384–387, 385, 386, 389, 391, 394 databases from single-channel SAR, 378–384 dual polarization data, 384–387 example of the implementation of an altimeter model to study the impact of saline snow on backscatter, 547, 547–548, 548 frost flowers, 441–442, 442 fully polarimeter data, 387–394, 393–394 models for level ice, 544–545 models for ridged ice, 545 snow influence, 334–341, 336, 338 backscatter coefficient, 310, 312, 316, 322, 324, 335, 336, 338, 338 basal plane, 194 Bayesian Likelihood Ratio Test (BLRT), 467 Beaufort Sea Gyre (BSG), 68, 80, 564, 575, 584 Beer-Lambert law, 299 Bernal-Flower rules, 198 bidirectional reflectance distribution function (BRDF), 300–302 biomass accumulation at ice bottom, 230–232, 231–232 birefringence, 153–155, 154, 164 blackbody radiation, 285, 298, 298, 543 blue ice, 365, 365 borehole indentation, 22, 79, 80 borehole strength, 82 Born approximation, 543, 546 Bottom Ice, 202 Bragg scattering, 317, 318, 321, 331–332, 332, 382, 387 brash ice, 413–414, 414–415, 416 breathing hole, seal, 16, 16, 77, 564 Brewster angle, 152, 152, 168, 295, 295 brightness, 156–157 brightness temperature, 297–299 dry snow depth and, 337 microwave data, 370–373, 370–376, 375–376 PM remote sensing, 305, 306, 308, 329–330 sea ice classification and, 458, 458 from snow, 337, 339, 341 snow grain size and, 339

snow wetness and, 339, 341 surface melt, 432–434 thermal infrared sensing, 304 brine dielectric constant of, 136 inclusions, geometric characteristics of, 232–234 mobility through subgrain boundaries, 66–67 permittivity, 142 thermal conductivity of, 126–128, 128 brine channel, 47, 48, 54, 56, 56–57, 60, 65–67, 67, 77 brine drainage, 9, 47, 48, 59, 65, 67, 67 brine expulsion, 62–65 brine flushing, 66 brine-layer spacing, 234, 236–239, 238–240 brine pockets, 31, 41, 47, 48, 50, 52, 54–58, 54–58, 61–64, 66, 79, 81–82, 194 EPS in, 232 formation, contents and distribution in sea ice, 54–58 in frazil ice, 218 geometric characteristics of brine pockets in first-year ice, 236–242, 238–242 migration, 61–62 in seasonal ice, 220, 223 brine (platelet) spacing, 194 BSG, 524 bubbles. see air bubbles bubbly ice, 171, 200, 223–224, 224 bulk salinity, 115, 115–116 buoyancy, 324, 501–502 calcite, 154, 154 calibration, of sensor data, 289 cameras, for photographing thin sections, 167–168 Canada historical and community synopsis, 4–8 map of the Canadian Arctic Archipelago, 5 Canadian Ice Service-Automated Sea Ice Tracking System (CIS-ASITS), 527–528, 529–530 capillary waves, 331 carbon dioxide, dissolved in seawater, 33 c axis, 152, 152, 155, 156, 194 CERTSAT, 380 characteristic radiometric values (CRVs), 479–480, 483, 483 chemical etching, 183–188, 185–187 circularly polarized light, 172–173 CIS Community Ice-Ocean Model (CIOM), 128 classification of ice, 84–85, 85–87, 200–216 freshwater ice, 200, 201, 202 R, 201, 203, 211–212, 212 S1, 201, 202, 205–207, 206–207 S2, 201, 202, 207–210, 208–210 S3, 201, 202, 211, 211–212 S4, 201, 202, 204–205, 205 S5, 201, 203, 204–205, 205–206 T1, 201, 203–204, 204 classification of sea ice, 454–471 age-based versus SAR-based and scattering-based, 321–322 from optical and TIR systems, 456, 456–457 from PM data, 457–458, 458–459 from SAR, 458–471, 462–468, 470 clear ice, 200 climate change, 1. see also global warming icebergs, 4, 11 impacts on polar ice, 557–590 terminology, 558 climate forcing, 557 climatology and sea ice, 13–14 Cloude-Pottier eigen decomposition, 318 cloud liquid water (CLW), 328–330 cloud masking, 327–328 cloud radiation, 271 coherency matrix, 316–320 Cold Regions Research and Engineering Laboratory (CRREL), 4, 275, 277–278, 377, 379, 401

INDEX 597 Collaborative-Interdisciplinary Cryospheric Experiments (C-ICE), 335, 429, 430 columnar crystal, 195 columnar-grained ice (S3 type) crystallographic classification, 201, 202, 211, 211–212 crystallographic structure, 218–219, 220–221 fabric diagram, 214–215, 215 columnar-grained with c axis horizontal and random (S2 ice) crystallographic classification, 207–210, 208–210 double-microtoming technique (DMT), 209, 209, 210 fabric diagram, 215, 215 columnar-grained with c axis vertical (S1 ice) crystallographic classification, 201, 202, 205–207, 206–207 fabric diagram, 214–215, 215 complex permittivity. see dielectric constant compositional supercooling, 50, 50–51 composition of sea ice, 9 compression, 68, 68–70, 73, 76 conductivity, 134–135 Confederation Bridge, Canada, 19, 19 congelation ice, 38–39 conical scanners, 288, 288–289 conservation of energy, 296 Constellation of Small Satellites for Mediterranean Basin Observation (COSMO-SkyMed), 357 convection current, 34, 34 conventional neural network (CNN) for sea ice classification, 464–465, 465 for sea ice concentration, 498 convergence, 414, 417–418, 419 Coordinated Eastern Arctic Experiment (CLEAREX), 275–276 Copernicus Imaging Microwave Radiometry (CIMR), 354 co-polarization, 462–464, 469, 497, 515 Cordilleran ice sheet, 558, 558 Coriolis force, 67, 99 corneal birefringence, 155 Couloumb’s law, 134–135 covariance matrix, 316, 318, 320 CP SAR (compact polarimetric SAR), 315–316, 316 CP (compact polarization) SAR, 468, 470–472, 517 cracks, 74–75, 74–76, 421–422, 430 cross-polarization, 469, 498 cross-polarization backscatter, 378, 384–388, 386 cross-polarized light, 168–172, 169 cross-polarized/scattered light viewing of thin sections, 169–172 cross-polarizers, 157, 164, 165, 173 cross-track scanners, 288, 288 CRREL Experiment (CRRELEX), 377, 379 CryoSat-2, 360, 427, 512–514, 514, 570–572 crystal definition, 194 optic axis of, 152 crystal defects definition, 194 dislocations, 194–195 point defects, 194 stacking faults, 195 crystalline texture, 195 crystallographic structure of perennial sea ice, 221–230 hummock ice, 223–226, 224–227 melt pond ice, 226–240, 228–230 crystallographic structure of seasonal sea ice, 217–221 agglomerate ice (R type), 220 air entrapment, 220–221, 223 columnar-grained ice (S3 type), 218–219, 220–221 frazil ice (S5 type), 217–218, 218–219 cubic ice (Ic), 198 damaged zones, 197 DB. see double-bounce (DB) scattering mechanism deep learning (DL) sea ice classification, 464 sea ice concentration, 498

deformation, 63 dislocations, 79 ice classification, 84–85 wind effect on, 583–584 dendritic ice-water interface, 51–52, 52 density ice, 10, 121–122, 121–123, 199–200 seawater, 32, 32–33 depolarization correlation coefficient, 318 depolarization of the radar signal, 290, 292, 312–314, 330, 338 desalinization during sea ice growth, 58–67 deuterium, 30, 153 diatoms, 17 dielectric constant, 134–136 of brine, 136 complex, 135 field measurements of, 142–145, 142–146 dielectric mixing models, 136 dielectric properties, 134–146 dielectric constant of brine, 136 dielectric mixing models, 136–142 field measurements of dielectric constant, 142–145, 142–146 overview, 134–136 Digital Mapping System (DMS), 426, 427 directional solidification (DS), 216 dislocation etch pit, 174–177, 175–176, 181, 181, 185 dislocation lines, 177 dislocations definition, 194–195 glide, 177, 185 pile up, 63–64 slipping basal, 64 divergence, 417–418, 419, 421 dog teams, 21, 22 Doppler shift, 311, 332 double-bounce (DB) scattering mechanism, 312–313, 313, 317–318, 320–321, 321, 332, 332 from rafted ice, 412–413, 413 sea ice classification, 469–470 double-microtoming technique (DMT), 233 overview, 158–159 precautions for thin sectioning by, 163 S2 ice, 209, 209, 210 section examples, 162–164 for thin sectioning of ice, 158 for thin sectioning of snow, 161–163 double refraction, 153–155 dual-channel SAR, ice classification from, 461–466 Earth Resources Technology Satellite (ERTS) programs, 349 ECICE algorithm, 466, 466, 479–486, 581 edge detection, 414, 416 Eigen decomposition, 319 eigenvector decomposition, 387 elasto-delayed-elastic-viscous (EDEV) model, 22, 63, 192 Electrically Scanning Microwave Radiometer (ESMR), 23, 350, 370, 370, 372 electromagnetic (EM) wave, 151, 152 absorption, 296 brightness temperature and, 297–299 circular, 290–291, 290–291 elliptical, 290–291, 291, 292 emission, 296–297, 297 linear, 290–291, 290–291 orientation, 291, 291, 292 penetration depth, 299–300 polarization, 290–292, 312 processes and properties, 289–300 propagation, 135 reflection, 293–295, 294–295 scattering, 296, 326, 326 transmission, 295 ellipticity of EM wave, 291, 291, 292 El Niño-Southern Oscillation (ENSO), 585, 589, 589–590

598

INDEX

emission models inversion modeling, 552–553 sea ice, 543–544, 548–552 emissivity, 293, 296–299, 297, 329–330, 333, 339 brightness temperature and, 297–299 data in the microwave bands, 395–403 PM sensing, 305, 307–308 thermal infrared sensing, 303–305 emitted radiation, 296–297, 306 EM spectral regions, 284–286, 285–286 energy balance ice thickness estimation and, 501–503 at the ice-water interface, 42 ENSO (El Niño-Southern Oscillation), 585, 589, 589–590 entropy, 319, 319 environmental interactions, 577–590 Antarctic sea ice, 585–590 Arctic sea ice, 578, 578–585 ENVISAT, 359, 515, 516 EO satellites geostationary, 287 polar orbiting, 286–287, 287 sun-synchronous, 287, 287 equiaxed crystal, 195 ERS. see European Remote Sensing (ERS) satellites etching chemical, 183–188, 185–187 process, 176–188 etch pit, 173–178, 175–176, 181, 181–187 EUMETSAT, 287 European Centre for Medium-Range Weather Forecasts (ECMWF), 542 European Remote Sensing (ERS) satellites, 351, 355, 358, 359 backscatter data, 378–380, 380–382 scatterometer, 323, 323 sea ice thickness, 511 surface feature images, 414, 415 extent of sea ice, 498–501, 499–501 extinction position, 195 extracellular polymeric substances (EPS), 232 extraordinary ray, 152 fabric diagram, 195, 214–216, 215 far infrared region (FIR), 285, 285 Fast Fourier Transform (FFT) shift theory, 527 fast ice, 84 field expeditions, 3–4, 249–280 Arctic Ice Dynamic Joint Experiment (AIDJEX), 250, 250–252 high arctic experience with ice of land origin, 262–266 Ice Exercise (ICEX), 277, 277–278 Labrador Ice Margin Experiment (LIMEX), 266–267, 266–268 marginal ice zone (MIZ) experiments, 274–276 Mould Bay experiments (1981-1984), 252–262 Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC), 278–280, 279 multi-year rubble field around Hobson’s Choice ice island, 265–266, 265–266 Norwegian Young Sea Ice Experiment (N-ICE), 272–274, 273 Sea Ice Monitoring and Modeling Site (SIMMS) program, 268–270 Surface Heat Budget of Arctic Ocean (SHEBA), 270, 270–272, 272 Ward Hunt Ice Shelf (WHIS), 262–264, 263–265 First Nations, 8 first-year ice (FYI), 11 aging to MYI, 259–260 air bubbles in, 58, 58, 242–243 albedo, 364–367, 368 brine drainage, 61 cracks in, 74, 74–75 density, 122, 122–123 dielectric properties, 138, 139–140, 140, 142–143, 143, 144 emissivity, 395, 397, 397–398, 398, 402 geometric characteristics of brine pockets in, 236–242, 238–242 ice aging, 80–83 ice decay, 76–79, 80 inclusions, 47, 48

microwave brightness temperature, 370, 370–373, 371, 373 microwave penetration depth, 403, 403–405, 404–405 Mould Bay field experiments, 254–262, 255, 261–262 permittivity, 139–140, 140, 143–144 physical and electrical properties, 109 radar backscatter, 378, 379, 380–381, 382, 385, 385, 385–389, 388–389, 390, 391, 391–392, 394 ridging, 72, 72, 72–73 salinity, 113, 115, 115–116 snow depth, 445–446, 447, 448 thermal conductivity of, 126–127, 129, 129 floating ice, 8–12, 84 Formvar, 174, 177–179, 182, 184 forward model inversion, 541, 552–553 forward models gross features of, 542 input to, 545–547 optima estimation and, 552–553, 553 overview, 541–542 primary input parameters, 545–546 secondary input parameters, 546 tertiary input parameters, 546–547 foxes, Arctic, 16 FP SAR (fully polarimetric SAR), 315–316, 316, 467–471 fractures in ice cover, 74–75, 74–76 frazil ice, 200 accumulation of, 36, 36 algal growth, 230 crystallographic classification, 201, 204–205, 205–206 crystallographic structure, 217–218, 218–219 initial formation, 36 lateral growth, 37 microwave penetration depth, 403, 403, 405, 405 platelet ice, 203, 203, 213–214, 214 randomly oriented (S4), 201, 204–205, 205 salinity, 115, 120, 128 vertically oriented (S5), 201, 204–205, 205–206 freeboard, 501–502, 510–514, 511, 520 Freeman-Durden decomposition, 320 freezer burn, 174 freezing degree days (FDD), 42–44, 43 freezing point, 32, 32–34 freezing process, 33–35 freshwater density, 32, 32 freezing processes, 33–34 salinities, 31 specific heat capacity, 131 thermal conductivity, 126 freshwater ice classification, 200, 201, 202 latent heat of fusion, 133–134 Fresnel equations, 293–294, 295, 297 Fresnel reflection model, 333 frost flowers, 438–440, 438–442, 442 fully convolutional network (FCN) model, 498 FYI. see first-year ice (FYI) Gaofen-3 (GF-3) satellite, 357, 466, 467 gases, dissolved in seawater, 33 gas inclusions, 47. see also air bubbles GEM weather model, 339, 340 general circulation model (GCM), 126, 270–271 geometric characteristics of air bubbles, 242–244, 243, 244 of brine pockets in first-year ice, 236–242, 238–242 of crystalline structure, 232–236, 234–237 geophysical parameters of sea ice, 453–533 classification of sea ice, 454–471 concentration, sea ice (SIC), 471–498 extent and area of sea ice, 498–501, 499–501 ice surface temperature (IST), 517–521, 519, 521 motion and kinematics, 524–533 retrieval of, 453–533 sea ice age (SIA), 522–524, 523 thickness of sea ice (SIT), 501–517

INDEX 599 Geoscience Laser Altimeter System (GLAS), 360 Geostationary Operational Environmental Satellite (GOES), 287 geostationary orbit, 287 geostrophic wind, 421–422, 422 glacier ice air bubbles in, 48 cracks, 48–49 microstructure, 212–213, 213, 262 glaciers effect of sea ice on, 587 ice of glacier origin, 95–99, 96 glaze, 333 Global Environmental Multi-scale (GEM), 328, 382, 504 Global Navigation Satellite System Reflectometer (GNSS-R), 280 global warming, 557–590 manifested in polar regions, 560–562 terminology, 558 global warming influence on polar sea ice, 565–577 Antarctic icebergs, 576–577, 577 Arctic sea ice age, 572–575, 573–574 Arctic sea ice dynamics, 575, 575–576 sea ice extent (SIE), 565–569, 567–568 sea ice thickness (SIT), 569–572, 570–571 sea ice volume, 571–572, 572 goniometer, 165, 166, 166 Gore, Al, 560 grain boundaries definition, 195 in frazil ice, 218, 219 geometric characteristics, 233, 239, 242 large-angle, 196 low-angle, 196 thermally etched, 176–182, 178, 180–182 thin section viewing issues, 163 grains geometric characteristics, 233–236, 235, 237 in sea ice, 53–54, 53–55, 193–194 grain size, in snow, 324, 334, 335–339, 338 granular ice, crystallographic classification of, 203–204, 204 gravity drainage of brine, 65 gravity waves, 331–332, 332, 331–332 gray ice, 220, 222 albedo, 366, 366–367 backscatter, 386 Gray Level Co-occurrence Matrix (GLCM), 460, 461, 462 grease ice, 36 albedo, 364–365, 366, 367 backscatter, 379 emissivity, 399, 399 salinity, 121 Great Lakes, ice coverage of, 2 great salinity anomaly, 16 greenhouse gases, 557, 560, 569, 583, 586 Greenland ice sheet, 558, 558 growth, of sea ice, 2, 37–47 lateral, 37–38 salinity loss, 58–67 simplified models of, 41–45 superimposed ice, 39–40 thermodynamic, 40–47 vertical, 38–39 growth direction of ice crystals, 199 Hadley Centre climate model (HadGEM3), 559 Hai Yang (HY) satellites, 358, 359–360 HDO, 30 heat budget, 271–272, 272 heat capacity, 131 heat exchange, 13, 126 heat flux, 13–14 in model of sea ice growth, 42, 42, 44 oceanic, 46 polynyas and, 442–443

heat of fusion, 13 heat transfer, 42–43 hexagonal crystals, 196, 196 hexagonal ice (Ih) density, 199 growth direction of crystals, 199 Miller indices for, 198, 198–199 High Arctic region, 4–7, 262–266 Himalayan cryosphere, 1, 2 Hobson’s Choice ice island, 253, 263–266, 263–266 hot-plate technique, 158, 159 hummock ice air bubbles, 243, 398 crystallographic structure, 223–226, 224–227 density of, 122, 122–123 emissivity, 396, 398, 400–401, 400–401 Hobson’s Choice ice island, 264, 264–265 microwave penetration depth, 403, 403 permittivity of, 145, 145–146 radar backscatter, 381, 387–388, 388 radar scattering mechanisms and, 321, 321 hummocks, 47, 48, 77, 81, 81–83, 85 hydrogen isotopes, 30 ice-albedo feedback loop, 364–365 ice arch, 88–90, 90–91 icebergs, 95–99 Antarctic, 4, 576–577, 577 bulbous bows, 12, 12 calving, 4, 11, 96, 99, 589, 589–590 charting of, 11–12 classification, 97 interaction with sea ice and ice shelves, 587–590, 589 response to global warming, 576–577 shapes, 97 size, 96–97 icebreakers, 3, 18, 19, 19–20, 84, 88, 98 ice charts, 456, 460–461, 464–466, 465–466, 468, 468–469, 486, 493–494, 496, 498, 512 ice classes, 84–85, 85–87 ice class ship, 84 Ice Cloud and Elevation Satellite (ICESat), 360, 512, 512, 514 sea ice extent (SIE), 580 sea ice thickness, 570–572 ice congelation, 29 ice crystals dendritic or stellar growth, 35, 35–36 forms of, 197–198 frazil ice, 36 growth direction, 199 initial formation of, 35–36 types, 196 ice decay, 76–79, 77–80 ice deformation, 67–76, 74–75, 74–76 compression and shear forces, 68, 68 leads, 75–76, 75–76 rafting, 68, 69, 69–70, 70, 72–73 ridging, 68, 70–73, 70–73, 72 rubble field, 68, 73–74, 73–74 spatial scales, 68 ice density, 10, 121–122, 121–123, 199–200 ice drift, 251, 259, 262–263, 263, 265, 268, 271, 273–274, 276–278 ice dynamics Arctic Ice Dynamic Joint Experiment (AIDJEX), 250, 250–252 Norwegian Young Sea Ice Experiment (N-ICE), 273 ice edge, 94–95, 95 Ice Exercise (ICEX), 277, 277–278 ice flux out of the Arctic Basin, 580–582, 581, 583 ice formation, initial, 33–36 ice hinges, 9 ice island, 11, 49, 49, 95, 98, 98–99 air bubbles, 49, 49 hazards, 262 Hobson’s Choice, 253, 263–266, 263–266 man-made, 262 ice layering, 337, 341 ice lenses, 333, 337–338, 341

600

INDEX

ice motion Arctic Ice Dynamic Joint Experiment (AIDJEX), 251 in Arctic region, 575, 575–576 wind effect on, 582 ice numeral, 86 ice of glacier origin, 95–99, 96 ice of land origin, field study experience with, 262–266 ice point, 196 ice polygon, 494–495, 494–496, 497, 530 ICESat. see Ice Cloud and Elevation Satellite (ICESat) ice shelf, 95–96, 96 interaction with sea ice and icebergs, 587–590 Ward Hunt Ice Shelf (WHIS), 262–264, 263–265 ice strength, 77, 79, 80 ice surface temperature (IST), 303–304, 517–521 from PM observations, 520–521, 521 from TIR observations, 517–520, 519 ice tongue, 95 ice-water interface compositional supercooling, 50, 50–51 dendritic, 51–52, 52 heat exchange, 126 salinity across, 120–121 salt rejection at, 59–61 temperature at, 110–111 imaging radar sensing, 308–322 multichannel SAR, 313–315 principles of imaging radar, 308–313 SAR polarimetry, 315–322 impurities, 9, 34 impurity entrapment, 13 inclusions in ice, 31, 41, 47–67, 48–49, 52, 54–58, 54–58, 66, 66, 79, 81, 81–82. see also air bubbles; brine pockets compositional (constitutional) supercooling at the ice-water interface, 50–51 dendritic ice-water interface, 51–52, 52 dielectric properties of ice, influence on, 136–140, 137, 139–140, 142, 145 grains and subgrains, 53–54, 53–55 in seasonal ice, 220–221, 223 incoming solar radiation (ISR), ice decay and, 77–78 An Inconvenient Truth (film), 560 indentation tests, 22, 264, 265–266, 266 Indian National Satellite System (INSAT), 287 inference of colors, 156 infrared observations, influences of atmosphere on, 325–328, 326 Integral Equation Model (IEM), 544, 547 integrated field of view (IFOV), 289 integrated water vapor (IWV), 328–330 interference colors, 157, 157–158, 164 Intergovernmental Panel on Climate Change (IPCC), 14, 559, 563 internal stress, ice motion and, 524–525 International Arctic Buoy Program (IABP), 271, 532–533, 575 Inuit, 4–8, 7 dog teams, 21, 22 knowledge about sea ice, 11 names for sea ice, 10–11 inverse modeling, 552–553 irradiance, 286, 300, 302, 303 isotopic composition of water, 30–31 ivuniit, 11 Japanese Earth Resources Satellite (JERS-1), 351, 414, 415 Joint Altimetry Satellite Oceanography Network (JASON) series, 359 Joint Polar Satellite System (JPSS), 353 katabatic wind, 421 Kerr electro-optic effect for water, 153 kinematics processes, 417–421, 418–420 sea ice motion and, 524–533 Kirchhoff’s law, 296, 297

laboratory techniques, 149–188 etching processes, 176–188 ice thin sectioning, 158–164 overview, 149–151 sublimation of ice, 173–176 viewing and photographing thin sections, 164–173 Labrador Ice Margin Experiment (LIMEX), 266–267, 266–268 Lagrange function, 481–482 Lagrangian particle, 260 Lagrangian representation, 417, 418 lake ice, air bubbles in, 48–49, 49 Lambertian surface, 300–301 land-based ice, microstructure of, 212–213, 213 landfast, 9 Landsat, 250–251, 269, 275, 423 La Niña, 585 laser altimeter, 323–325, 324 laser image detection and ranging (LIDAR), 323 laser ranging, 251 latent heat of sea ice, 133–134, 134 latent heat polynya, 87 lateral ice growth, 37–38 lattice, 196 lattice straining, 196 Laurentide ice sheet, 557, 558 leads, 75–76, 75–76, 421–428, 422–425, 427 effect on local air temperature, 579 width, 422, 422–423, 426 light circularly polarized, 172–173 cross-polarized, 168–172, 169 parallel-polarized, 168–169, 169 polarized, 151–153, 163 scattered, 169–172, 169–172 speed of, 151, 154 linear kinematic feature (LKF), 417, 420–421 liquid zone migration, 61 Little Ice Age, 558 loss factor, 135 Low Arctic region, 4–5 LOW-resolution atmospheric TRANsmission (LOWTRAN), 305 machine learning, 416–417, 464–465 Making Earth System Data Records for Use in Research Environments (MEaSUREs), 417 MAp-Guided Ice Classification (MAGIC), 461 marginal ice zone (MIZ), 39, 92–94 field expeditions, 274–276 radar backscatter data, 385 ridges, 413 marginal ice zone Department Research Initiative (MIZ-DRI), 276 marginal ice zone observations and processes experiment (MIZOPEX), 275–276 marine biology, sea ice in, 16–17 marine navigation, sea ice in, 17–19, 18 Markov Random Field (MRF) labeling model, 461, 462 maximum cross-correlation (MCC), 526, 526–527 MB. see multiple-bounce (MB) scattering mechanism Medium Resolution Imaging Spectrometer (MERIS), 370 melting ice aging, 81, 81–83 ice decay, 76–79, 77 melting point, 196 melt pond ice, 223 air bubbles, 243 crystallographic structure, 226–230, 228–230 emissivity, 396–397, 400–401 microwave penetration depth, 403, 403 radar scattering mechanism and, 321 melt ponds, aerial photography of, 437, 438 meteorologically driven surface features, 442–448 polynya, 442–444, 444 snow depth, 444–448, 445, 447 meteorology, sea ice in, 14–15

INDEX 601 microorganisms, in sea ice, 17 microscope, polarizing, 165–166 microstructure, 10, 11, 12, 193–194 microwave brightness temperature, 370–373, 370–376, 375–376 microwave data. see also active microwave; passive microwave (PM) observations seawater influences, 331–333 snow influences, 336–341 microwave emission, 285, 285, 286 emissivity in microwave bands, 395–403 modeling, 541–553 Microwave Emission Model for Layered Snowpacks (MEMLS), 274, 444, 446, 542–544 Microwave Imaging Radiometer with Aperture Synthesis (MIRAS), 508, 510 Microwave Model (MWMOD), 329–330, 397, 543 microwave penetration depth, 403–407 microwave sensors, 23, 285 active, 23 passive, 23 Mie scattering, 326, 326, 543 migration, of brine pocket, 61–62 Miller indices for hexagonal Ice, 198, 198–199 models/modeling altimeter model to study the impact of saline snow on backscatter, 547, 547–548, 548 forward models, 541–542, 545–547 inverse, 552–553 large-scale sea ice modeling, 542 microwave emission and scattering from snow-covered sea ice, 541–533 sea ice backscatter models for level ice, 544–545 sea ice backscatter models for ridged ice, 545 sea ice emission, 543–544, 548–552 of sea ice growth, 41–45 for sea ice thermal microwave emission, 543–545 to simulate the noise in sea ice concentration estimates, 548–552 Moderate Resolution Imaging Spectroradiometer (MODIS), 352–353 ice surface temperature, 517–518, 519 leads, 422–423, 423, 425–426 sea ice concentration, 489 sea ice thickness (SIT), 504, 506, 508 snow depth, 446 spectral albedo and, 367–368 surface melt, 430–431, 431 MODIS Cloud Mask, 328 molecular composition of water, 30 monocrystalline materials, 193 morphology of ice, 197–200 Mould Bay experiments (1981-1984), 252–262 aging of sea ice (from FYI to MYI), 259–260 sea ice conditions, 254–259, 255–259 second-year ice profile, interface between old and new ice in, 260–261, 260–262 site, resources, and logistics, 252–254, 253–254 Multi-angle Image SpectroRadiometer (MISR), 457 Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC), 278–280, 279 Multi-functional Transport Satellite (MTSAT), 287 multiple axis photometer (MAP), 173 multiple-bounce (MB) scattering mechanism, 312–313, 313, 317, 320–321, 321, 469–470 multi-year ice (MYI), 2 air bubbles in, 58, 242–244, 243, 244 albedo, 364–367, 368 Arctic ice flux, 581–582, 582–583 Arctic sea ice age, 572–575, 573–574 bubble formation in, 81–82 crystallographic structure, 221–230 density, 122, 122–123 dielectric properties, 138, 140, 141, 142, 145, 145–146 emissivity, 397, 397–398, 398, 400, 400 first-year ice (FYI) aging to MYI, 259–260 ice aging, 80–83 inclusions, 47, 48 microwave brightness temperature, 370, 370–373, 371, 373

microwave penetration depth, 403, 403–405, 404–405 Mould Bay field experiments, 254, 259–260 permittivity, 140, 141, 145, 145–146 physical and electrical properties, 109 radar backscatter, 378–381, 379, 382, 385, 385, 385–388, 388–389, 391, 391–392, 394, 395 radar scattering mechanisms and, 321, 321 ridging, 72, 72 rubble field around Hobson’s Choice ice island, 265–266, 265–266 salinity, 113, 115 snow depth, 445–446, 447, 448 surface during melt season, 81 SYI distinguished from, 83 thermal conductivity, 127–128 top and bottom surface profiles of MYI in the Nares Strait, 82, 83 mush ice, salinity of, 120–121 National Research Council (NRC) field experiments, 252–253, 256, 260, 264 National Snow and Ice Data Center (NSIDC), 276 ice type maps, 456 sea ice age, 489, 490, 522–523, 523, 573, 575 sea ice concentration, 454, 472 sea ice extent, 454, 498, 500, 566–568, 567–568 Sea Ice Index, 566 sea ice motion, 525, 528, 530, 532–533, 532–533 sea ice thickness, 513, 514, 570 near-infrared region (NIR), 285, 285 near-surface air temperature (nSAT), 560–561, 560–562, 568 new ice (NI) emissivity, 397, 397–398, 399 physical and electrical properties, 109 radar backscatter, 379, 381, 386 nilas, 37, 37–38, 220, 222 albedo, 364, 366, 366–367 backscatter, 378–379, 385, 385 emissivity, 397, 399, 399 microwave penetration depth, 404, 404 Nimbus program, 349–350 non-selective scattering, 326, 326 normalized difference snow index (NDSI), 500–501 Northeast Passage (NEP), 18, 18–19 North Water polynya (NOW), 87, 88, 90, 455, 507, 519 Northwest Passage (NWP), 2, 5, 10, 18, 18–19, 82 Norwegian Young Sea Ice Experiment (N-ICE), 272–274, 273, 446 NSCAT, 380 NT2 algorithm, 476–478, 478 NT algorithm, 475–476 nuclei, 35 ocean currents, 584, 586, 586 ice deformation and, 68 ice motion and, 524 ocean foam, 331–332 oceanic forcing Antarctic, 586–587 Arctic, 584–585 oceanography, sea ice in, 15, 15–16 Ocean Scatterometer (OSCAT), 323, 358, 358–359 ocean–sea ice–atmosphere (OSA) system, 269 Odden sea ice, 15, 15, 92 offshore structures, sea ice and, 19, 19–20 open water albedo, 366, 367, 369 backscatter data, 378–379, 381–382, 381–386, 382, 384, 385, 389, 390, 391, 393–394 brightness temperature, 370–373, 371–373, 373 emissivity, 396–399, 397, 398, 400 Operational Linescan System (OLS), 423, 424 optical data/observations ice classification from, 456, 456–457 influences of atmosphere on, 325–328, 326 SAR compared, 455, 455 sea ice concentration from, 472–473 seawater in, 330–331

602

INDEX

optical data/observations (cont’d) snow influence, 335–336 optical properties, 151–158 optical reflectance, 364–370 optical retardation, 155–157, 156–157, 163 optical sensing, 300–303, 303 optical sensors, 23 optic axis, 152, 152, 155, 196 optimal estimation, 552–553, 553 ordinary ice (Ih), 196. see also hexagonal ice (Ih) birefringence of, 153–155, 154 oxygen isotopes, 30–31 pack ice, 39, 84 ice edge, 92, 94–95, 95 leads in, 75 marginal ice zone (MIZ), 93 ridge zones, 71, 71 strength of, 68 palm-technique for thin sections, 158, 168 Pan-Arctic Ice Ocean Modeling and Assimilation System (PIOMAS), 570–571, 571–572 pancake ice, 37–38, 38, 90–92, 91–92 parallel-polarized light, 168–169, 169 passive microwave (PM) observations atmospheric correction for, 328–330 ice surface temperature (IST), 520–521, 521 ocean surface roughness, influence of, 332–333 remote sensing, 305, 305–308, 306–308 sea ice classification, 457–458, 458–459 sea ice concentration (SIC), 473–496 sea ice thickness (SIT), 506–510, 507–510 surface melt, 432–434, 433 passive microwave (PM) sensors, 287, 287 historical perspective, 350–351 modern, 353–354, 354 Pauli target vector, 316–317 penguins, 16, 564 perennial ice, 2, 84, 108, 221–230 permafrost, melting of, 579–580 permittivity, 134–135, 139–141, 139–145, 143–145 phase-correlation matching technique, 527–528 Phased Array type L-band Synthetic Aperture Radar (PALSAR), 355–356, 516 ALOS-PALSAR, 314–315, 318, 414, 464 phase diagram of ice, 33, 33, 197, 197–198 phase shift, 156, 291, 292 photographing ice thin sections, 164–173 Photosynthetically Active Radiation (PAR), 17, 270, 303 phthalate, 161–162 physical properties. see properties of sea ice physics, sea ice in, 12–13 Planck’s equation, 285, 298, 298 platelet ice formation, 203, 203 interface, 196 microstructure, 213–214, 214 structure, 196 plate tectonics, 21–22 PM. see passive microwave (PM) observations PM Algorithm (PMA), 433 PM SIC algorithms, 476–493 ASI algorithm, 478–479 assessment of ice concentration results against ice charts, 493–496 comparison of algorithms, 486–490 ECICE algorithm, 479–486 error sources, 490–492 NT2 algorithm, 476–478, 478 NT algorithm, 475–476 sensitivity of, 492–493, 493 point defects, 194 polar bears, 16, 564 Polar Continental Shelf Program (PCSP), 6, 250, 263–264, 264, 269, 269–270

polarimetric decomposition parameters, 387, 389, 390, 391, 391–394 Polarimetric Scanning Radiometer (PSR), 444 polariscope, 153, 153, 165–169, 166–167, 169, 196 polarization degree of, 291, 293, 312 of EM wave, 290, 290–292 of radar signals, 311–312 vertical and horizontal polarization signals, 312, 312 polarization signature, 317–318, 318 polarized light, 151–153, 163 polarizer, 150, 152–153, 153, 156–157, 162, 164–168, 169, 171–172, 172 polarizing filter, 152, 165–168, 197 polarizing microscope, 165–166 polar molecule, water as, 121, 121, 153 Polar-Orbiting Environmental Satellites (POES) series, 352, 352 polar orbiting satellites, 286–287, 287 polar regions. see also Antarctic region; Arctic region geographic differences, 562–564, 563, 565 manifestation of global warming in, 560–562 sea ice characteristics, differences in, 564–565, 566 sea ice regimes, 562–565 polycrystalline ice structure, 191–244 biomass accumulation at bottom of ice, 230–232, 231–232 classification of natural ice, 200–216 definition, 197 examples, 216–230 geometric characteristics of air bubbles, 242–244, 243, 244 geometric characteristics of brine pockets in FYI, 236–242, 238–242 geometric characteristics of crystalline structure, 232–236, 234–237 information contents in, 234–244 laboratory techniques for revealing, 149–188 morphology of ice, 197–200 perennial sea ice, 221–230 seasonal sea ice, 217–221 stereographical projection (fabric diagram), 214–216, 215 terms and definitions, 191–197 polycrystalline solids, 193 polygonization, 63, 64, 197 polynya, 16, 86–90, 88–91 coastal, 88, 89 identification and properties, 442–444, 444 latent heat, 87–88 open-ocean, 87 optical and SAR images compared, 455, 455 sensible heat, 87–88 polyvinyl alcohol (PVA), 165–166 polyvinyl formal (Formvar), 174, 177 Pond Inlet, 253 Practical Salinity Unit (PSU), 31 precipitation of salts, 31–32, 32, 57–58 pressure ridges, 68, 70–72, 71, 413, 418 pressure (compression) ridges, 70 primary ice, freshwater, 200, 201, 202 primary production, 17, 230, 232 properties of sea ice, 107–146 air volume fraction, 124 brine volume fraction, 123–124 bulk salinity, 115, 115–116 density, 121–122, 121–123 dielectric, 134–136, 137–145 equations to determine parameters, 109 pure ice volume fraction, 124 salinity, 113–121 salinity profile, 116–121, 117–120 solid salt volume fraction, 124 temperature profiles, 108–113, 110–114 thermal, 126–134, 128–134 typical values of key physical parameters, 107–108, 108–109 volume fraction of constituents, 123, 123–126 pure ice volume fraction, 124 PVD (Polder and Van Santen) model, 136–137, 142, 144–145 pycnocline zone, 34, 34 qavvaq, 11 qinu, 10–11

INDEX 603 QuikSCAT, 339, 340, 341, 358, 358–359 scatterometer, 323, 323 sea ice concentration, 484, 484–486, 486, 489 radar altimeter, 308, 323–324, 324 radar backscatter. see backscatter radar cross section (RCS), 293, 309–310 radar equations, 309, 309–311 Radar Imaging Satellite (RISAT), 357 radar pulse, 308–309 coherency, 311 illustration of, 309 polarization, 311–312 scattering mechanisms, 312–313, 313 RADARSAT, 252, 267, 269, 277, 308, 314–315, 318 ice classification, 461, 462 optical image compared, 455, 455 radar backscatter data, 377, 380, 382, 382–383, 384, 385–387, 389, 390, 391, 393–394, 394 rafted ice images, 412, 412 ridged, rubble, and brash ice, 414–416, 416 sea ice classification, 467–471, 468 sea ice concentration, 487, 489, 490, 495, 495–498, 497 sea ice motion tracking, 528, 529 sea ice thickness, 505, 505, 512, 516, 517, 519 sensors, 355–358, 356 single look concept, 311 surface deformation, 417, 418, 420, 420–421 surface melt, 434, 435–436 RADARSAT Constellation Mission (RCM), 355, 357–358, 471 RADARSAT Geophysical Processing System (RGPS), 417, 419– 420, 420 radar scattering coefficient, 293 radar scattering mechanisms. see scattering mechanisms radar sensors, 23 radar system, monostatic, 309 radiance, 286, 298, 298, 300, 302, 303 radiation power, 543 radiative processes in atmosphere, 325–330 in seawater, 330–333 in snow over ice, 333–341 radiative transfer, 543 radiative transfer equation (RTE), 307, 517 rafted ice, identification and characterization, 412–413, 412–413 rafting, 37, 68, 69–70, 69–70, 72–73 randomly oriented (S4) frazil ice crystallographic classification, 201, 204–205, 205 three-dimensional schematic diagram, 205 Rayleigh scattering, 155, 326, 326, 406, 543 real aperture radar (RAR), 287, 308–309 recrystallization, 197 reflectance, 293, 300–302, 303, 327, 364–370 leads, 437 surface melt, 429–431, 430–431 top-of-atmosphere (TOA), 429 reflection, 152, 152, 293–295, 294–295 scattered (diffuse), 293 spectral (mirror), 293 reflectivity, 293–294, 295, 296–297 refraction, 152 double, 153–155 refractive index, 152, 154, 293–294, 330, 364, 395 rejection of solutes/salts, 31, 36–37, 40, 47, 50, 54, 58–61 remote sensing altimeter systems, 323–325, 324 atmospheric influences on, 325–330 fundamentals, 283–341 general principles, 284–289 imaging radar sensing, 308–322, 309, 311–313, 316, 318–319, 321 optical sensing, 300–303, 303 overview, 22–24, 283–284 passive microwave (PM) sensing, 305, 305–308, 306–307 scatterometer systems, 322–323, 323 seawater influences on, 330–333

snow on ice, influences of, 333–341 thermal infrared sensing, 303–305, 304 replicating ice surfaces, 183–187 re-radiation, 296–297 ridges backscatter models/mechanisms for ridged ice, 321, 321, 545 identification and characterization, 413–418, 414, 416, 420–421 pressure, 413, 418 shear, 413–414, 415 ridge zones, 71 ridging, 68, 70–73, 70–73, 72 pressure ridges, 68, 70–72, 71 shear ridges, 68, 70 Rigsby stage, 165–166, 166, 214–215 Ross Ice Shelf, 4 rotten ice, 77, 411, 428, 429, 430 R type ice, 201, 203, 211–212, 212, 220 rubble ice, 68, 68, 73–74, 73–74 Hobson’s Choice ice island, 264–266, 264–266 identification and characterization, 413–414, 414, 416 S1 ice crystallographic classification, 201, 202, 205–207, 206–207 fabric diagram, 214–215, 215 S2 ice crystallographic classification, 207–210, 208–210 double-microtoming technique (DMT), 209, 209, 210 fabric diagram, 215, 215 S3 ice crystallographic classification, 201, 202, 211, 211–212 crystallographic structure, 218–219, 220–221 fabric diagram, 214–215, 215 S4 ice crystallographic classification, 201, 202, 204–205, 205 three-dimensional schematic diagram, 205 S5 ice crystallographic classification, 201, 203, 204–205, 205–206 crystallographic structure, 217–218, 218–219 three-dimensional schematic diagram, 205 salinity, 113–121 algal growth and, 230 brine volume fraction, 123–124 bulk, 115, 115–116 latent heat of fusion, influence on, 133–134, 134 loss during sea ice growth, 58–67 microwave penetration depth and, 403–405, 405 of near shore water, 202 seawater, 31–32, 32 specific heat, influence on, 131–133, 133 surface, 114 temperature dependence of brine salinity, 125, 125–126 temporal evolution of, 120, 120 salinity profiles, 116–121, 117–120 salt precipitation, 57–58 salt rejection, 15, 40, 47, 50, 54, 58–61 from bulk ice, 61–67 initial rapid at ice-water interface, 59–61 salts in sea ice, 31, 32 SAR sea ice classification, 458–471, 462–468, 470 sea ice concentration (SIC), 496–498, 497 sea ice thickness (SIT), 514–517, 515–517 SAR (synthetic aperture radar) backscatter databases from single-channel, 378–384 concept, 310–311, 311 detecting ridges, rubble, and brash ice, 413 dual polarization, 384–387 field experiments, 252, 256, 267–271, 273–276, 279 frost flowers, 441 full polarization, 387–394, 393–394 historical perspective, 351 lead detection, 422–424, 426 modern systems, 355, 355–358 multichannel, 313–315

604

INDEX

SAR (synthetic aperture radar) (cont’d) multi-frequency, 314 multi-polarization, 314 optical images compared, 455, 455 radar backscatter data, 376–394, 382, 392 rafted ice images, 412, 412–413 ridged, rubble, and brash ice, 413–417 sea ice classification, 321–322, 458–471 single channel, 313 SLC (single look complex), 315, 332 surface melt, 428, 434, 434–436, 436 SAR, ice classification from, 321–322, 458–471 dual-channel SAR, 461–466, 463–467 overview, 458–460 polarimetric SAR, 467–471, 468, 470 single-channel SAR, 460–461, 462 SAR interferometry, 311 SAR polarimetry, 315–322 formulation of polarimetric measurements, 316–317 linking radar scattering mechanisms to ice features, 320–321, 321 polarimetric parameters derived from the FP SAR data, 317–320 polarization modes, 316, 316–316 sastrugi, 85 satellite sensors, 349–360 altimeter sensors, 359–360, 360 historical perspective, 349–351 modern imaging radar, 355–358 optical, 352–353 scatterometer systems, 358, 358–359 thermal infrared (TIR), 352–353 Satellite with ARgos and ALtiKa (SARAL), 359 SB. see single-bounce (SB) scattering mechanism Scanning Multichannel Microwave Radiometer (SMMR), 308, 350–351, 354, 432, 434 surface melt, 432, 434 scanning radiometers conical, 288, 288–289 cross-track, 288, 288 ScanSAR, 311, 315, 378, 414, 435, 471, 497–498 Scat-SeaIce composite color scheme, 394, 394 scattering in the atmosphere, 326, 326 dense media volume scattering, 543 of EM wave, 296 Lambertian, 300 of microwave signal in snowpack, 336 Mie, 326, 326, 543 non-selective, 326, 326 Rayleigh, 155, 326, 326, 406, 543 scattering-based ice classification, 321–327 scattering matrix, 316–318 scattering mechanisms, 312–313, 313 double-bounce (DB), 312–313, 313, 317–318, 320–321, 321, 332, 332, 412–413, 413, 469–470 linking to ice features, 320–321, 321 multiple-bounce (MB), 312–313, 313, 317, 320–321, 321, 469–470 from ocean wave, 331–332, 332 from rafted ice, 412–413, 413 sea ice classification, 469–470 single-bounce (SB), 308, 312–313, 313, 318, 320, 321, 321, 469–470 scattering power, 313, 317, 320–321 scattering vector, 316 scatterometer systems, 322–323, 323, 358, 358–359 fixed fan-beam configuration, 322–323, 323 rotating fan-beam (pencil beam) configuration, 323, 323 Schmidt net, 214 Scientific Ice Expeditions (SCIEX), 422, 422 sea ice age (SIA), 522–524, 523 Arctic, 572–575, 573, 573–574 response to global warming, 572–575, 573–574 Sea Ice and Terrain Assessment (STAR), 312 sea ice characteristics, 8–12 sea ice concentration (SIC), 471–498 from coarse-resolution microwave observations, 473–496 from fine-resolution SAR, 496–498, 497

map vendors, 472 modeling, 548–552 from optical and TIR images, 472–472, 474 overview, 471–472 sea ice cover, enhanced Arctic warming due to changes of, 578–580, 579 sea ice extent (SIE), 498–501, 499–501 Antarctic, 585–586 Arctic, 580, 580–582 response to global warming, 565–569, 567–568, 580, 580–582 Sea Ice Monitoring and Modeling Site (SIMMS), 268–270, 327, 335, 338 sea ice motion, 524–533 Antarctic region, 533, 533 Arctic region, 532, 532–533 operational ice motion products, 532–533 overview, 524–525 tracking using image features, 526, 526–528, 529–530 tracking using individual sea ice floes, 528–531, 531 sea ice regimes, 85–99 ice edge, 94–95, 95 ice of glacier origin, 95–99, 96 marginal ice zone, 92–94 overview, 85–86 pancake, 90–92, 91–92 polynyas, 86–90, 88–91 sea ice thickness (SIT), 501–517 from altimeter observations, 502, 510–514, 511–512, 514 Antarctic, 2, 571–572 Arctic, 2, 14, 569–571, 570 drownings, 11 FYI, 82 growth (see growth, of sea ice) ice classification by, 84 ice decay, 76–79, 79 interannual evolution, 82 for landing aircraft, 20, 20–21 model of sea ice growth, 41–45, 43–44 MYI, 82–83, 83 overview, 501–503 from PM observations, 506–510, 507–510, 508 pressure ridges, 70–72 response to global warming, 569–572, 570–571 sampling, 3, 3 from SAR observations, 514–517, 515–517 from TIR observations, 503–506, 505 sea ice types based on ice concentration, 84–85, 86 based on ice form, 84, 86 based on surface feature, 85, 87 classification overview, 454–456 sea ice volume, 571–572, 572 seals, 16, 16–17, 564 Search and Rescue aircraft, 20, 21 sea salt aerosol, 441 seasonal ice, 2, 84 air entrapment in, 220–221, 223 crystallographic structure of, 217–221 properties of, 108 Seasonal Ice Zone Experiment (SIZEX), 378, 381 sea surface temperature (SST), 303–304 seawater albedo, 335 density, 32, 32–33 freezing processes, 33–35 in the microwave data, 331–333 in the optical and thermal infrared data, 330–331 radiative processes, 330–333 salinity, 31–32, 32 SeaWinds scatterometer, 323, 323 secondary cryospheric regions, 1 secondary ice, freshwater, 200, 201, 202 second-year ice (SYI), 2 air bubbles, 244, 244 ice aging, 80, 82–83 interface between old and new ice, 260–262, 260–262 Mould Bay field experiments, 256–262, 256–262 SYI-FYI interface, 260–262, 260–262

INDEX 605 second-year ice profile, interface between old and new ice in, 260–261, 260–262 sensible heat polynya, 87–88 sensors illumination source for, 287, 288 overview, 23 passive versus active, 287 scanning versus non-scanning, 287 Sentinel satellites, 354, 355, 356–357, 360, 426, 428, 434–435 icebergs, 588, 589 ice motion, 531 sea ice classification, 464–466, 465–467 sea ice concentration, 498 shear, 417–419, 419 shear ridges, 54, 63, 68, 68, 70, 74, 76, 413–414, 415 shift function, 192 shortwave IR (SWIR), 285, 285 side-looking airborne radar (SLAR), 253, 254, 267, 275 siku, 11 sikuaq, 11 sinaaq, 11 single-bounce (SB) scattering mechanism, 308, 312–313, 313, 318, 320, 321, 321, 469–470 single-channel SAR, ice classification from, 460–461 single look complex, 315, 332 Sinha’s rule, 110, 111 skeletal layer, 52, 52, 197 Snell’s law, 155 snow albedo, 335–336, 336, 365, 367, 367–368, 369 density, 334, 334, 337 DMT for thin sectioning, 161–163 in emission models, 549 grain size, 324, 334, 335–339, 338 hoar layer, 333–334, 335, 338 metamorphism, 300, 303, 305, 314, 322, 333–335, 491, 502 in microwave data, 336–341 microwave penetration depth, 403–404, 403–407, 404, 406 in optical and thermal infrared data, 335–336, 336 oxygen isotopic composition, 31 stratigraphy of snow on Arctic sea ice, 130 temperature profiles, 108–113, 111–114 thermal conductivity, 45, 129–131, 131–132 thick sections, 161 typical values of key physical properties, 107–108, 108 snow cover, 4 effect on sea ice, 45, 45–46 ice decay, 77, 79, 79 oxygen isotope content, 40 superimposed ice, 40 weather effect on, 491 wind slab, 491 snow-covered sea ice emissivity, 395–396, 399 microwave penetration depth, 403, 406 snow depth effect on microwave data, 337 mapping, 4–46 observations, 444–448, 445, 447 snowflakes, 35 snow ice crystallographic classification, 201, 203–204, 204 three-dimensional (3D) schematic diagram, 204 Snow Model Radiative Transfer (SMRT) project, 544 snow on sea ice microwave data, influence on, 336–341 physical and radiative processes, 333–341, 334, 336, 338 snow water equivalent (SWE), 269 snow wetness effect on microwave data, 339–341 emissivity and, 401–402 Soil Moisture and Ocean Salinity (SMOS), 349, 354, 354 sea ice thickness, 509–510, 510, 514 solid earth sciences, sea ice in relation to, 21–22 solid salt volume fraction, 124

Southern Annular Mode (SAM), 585–586 South Temperate Zone (STZ), 39–40 SPAN (total power), 387–388, 389, 390–391, 391–392, 394, 395–396, 412, 412–413, 416, 470, 470 Special Sensor Microwave Imager (SSM/I), 306, 306, 337, 350–351, 354, 354, 372–373, 372–373 backscatter data, 383, 384 ice surface temperature, 518, 520 sea ice concentration, 487, 489, 496 sea ice thickness (SIT), 504, 506–508, 508, 509 snow depth, 445 surface ice concentration, 473, 477, 477–478 Special Sensor Microwave Imager Sounder (SSMIS), 351, 353, 354, 570 specific heat of sea ice, 131–133, 133 split window technique, 304 stacking faults, 195 stamps, 8, 9 Stamukhi, 70 standard mean ocean water (SMOW), 31 stereographical projection, 214–216, 215 stoichiometric composition of water, 30 Stokes vector, 291–292, 312 stress around brine pockets, 62, 63 ice deformation, 67–76 structural aspects of ice, general terms for, 193–194 subgrain boundaries, 30, 53–56, 53–55, 58, 61, 63–64, 64, 66, 66–67, 79 algal growth along, 231–232, 232 definition, 195–196 geometric characteristics, 233–234, 236, 239, 242–243 thermally etched, 176, 178–184, 181–183 subgrains geometric characteristics, 233–236, 235, 237 in sea ice, 53–54, 53–55, 194 sublimation, 46 definition, 197 ice, 173–176 snow, 333–334, 335, 438 sublimation pits, 174–176, 175–176, 180 sun glint, 331, 331 sun-synchronous orbit, 287, 287 supercooling, 32, 34, 36, 50–51 superimposed ice, 39–40, 220, 262, 264, 274 freshwater, 200, 201, 202 surface ablation, 46–47 surface air temperature, global warming and, 560–561, 560–562 surface-based radiometers (SBR), emissivity estimates from, 396 surface deformation, 412–428 cracks and leads, 421–428, 422–425, 427 kinematic processes, 417–421, 418–420 rafted ice, 412–413, 412–413 ridged, rubble, and brash ice, 413–417, 414–416 surface energy, 177–178 surface features mechanically generated deformation, 412–428 meteorologically driven, 442–448 thermally induced, 428–442 Surface Heat Budget of Arctic Ocean (SHEBA), 270, 270–272, 272, 426, 518 surface melt, 46–47 active microwave data, 434, 434–437, 437–438 airborne photography, 437, 438 albedo, 428–430, 431, 437 brightness temperature, 432–434 optical observations, 428–432, 430–431 passive microwave data, 432–434, 433 phases, 428–429, 429 reflectance, 429–431, 430–431 surface roughness emissivity and, 395, 401, 402 ocean, 330–333 radar backscatter and, 378, 387, 391, 391 surface topography, AIDJEX and, 251 swell, 332, 332 SYI. see second-year ice (SYI) Synthetic Aperture Interferometric Radar Altimeter (SIRAL), 427

606

INDEX

T1 ice crystallographic classification, 201, 203–204, 204 fabric diagram, 215, 215 three-dimensional (3D) schematic diagram, 204 Television and Infrared Observations Satellite (TIROS), 349, 350 temperature profiles in ice and snow, 108–113 TerraSAR-X, 356–357 thaw hole, 77, 77 thermal conductivity, 4 in model of sea ice growth, 42–43 of sea ice, 13, 126–129, 128–129 of snow, 45, 129–131, 131–132 thermal diffusivity, 126 thermal etching, 54, 54, 63, 63–64, 161–162, 173, 176–183, 178, 180–183 thermal infrared (TIR) observations, 335–336 ice surface temperature (IST), 517–520, 519 lead detection, 423, 424 sea ice classification, 456, 456–457 sea ice concentration (SIC), 472–473, 474 sea ice thickness (SIT), 503–506, 505 seawater in, 330–331 surface temperature data, 303–305, 304 thermal infrared region (TIR), 285, 285 thermal infrared (TIR) sensors, 23, 286, 352–353 thermally induced surface features frost flowers, 438–442 surface melt, 428–437 thermal properties of sea ice, 126–134, 128–134 latent heat, 133–134, 134 specific heat, 131–133, 133 thermal conductivity, 126–129, 128–129 thermal state of ice, 9–10, 12, 192–193 thermodynamic ice growth, 40–47 growth rate, 41 models, 41–45 oceanic heat flux, effect of, 46 snow on ice, effect of, 45–46, 46 surface ablating, effect of, 46–47 thermohaline circulation (THC), 15, 15, 584–585 thick ice albedo, 366, 366 emissivity, 402 microwave brightness temperature, 375 thickness and enthalpy distribution (TED) sea ice model, 503 thickness of ice. see also sea ice thickness (SIT) albedo and, 503, 503 altimeter data, 323–325, 324 Arctic Ice Dynamic Joint Experiment (AIDJEX), 251 bulk salinity and, 115 emissivity and, 398–399, 399 microwave penetration depth, 405, 405 thick sections, 158, 158, 160, 161 thin ice albedo, 366, 367 backscatter, 372–374, 375–376, 377–379, 385, 385, 391–394 emissivity, 398–400 microwave brightness temperature, 372–374, 375–376 thin sectioning techniques, 158–164 double-microtoming technique (DMT), 158–163 hot-plate, 158, 159 optimum thickness for sections, 163–164 overview, 158–159 palm-technique, 158, 168 thin sections, viewing and photographing, 164–173 circularly polarized light, 172–173 cross-polarized light, 168–169, 169 overview, 164–165 parallel-polarized light, 168–169, 169 polariscope, 165–169, 166–167, 169 scattered light, 169–172, 169–172 tie points, 372–374, 373, 375 TIE region emission, 296–299 tilt boundaries, 63, 63–64, 197 Titanic, 97

total backscatter power. see SPAN (total power) transmission of EM wave, 295 transmittance, and ice thickness, 503, 503 transpolar drift (TPD), 575 Transpolar Drift Stream (TDS), 68, 524, 564 Transportable Microwave Active Spectrometer (TRAMAS), 378 transportation platform, sea ice as a, 20–21, 20–21 tritium, 30 tsunamis, 331–332 tuvaq, 11 unmanned airborne systems (UASs), 276 upward-looking sonar sensors (ULS), 505 Variation Arctic/Antarctic Sea Ice Algorithm 2 (VASIA2), 489 vertical ice growth, 38–39 vertically oriented (S5) frazil ice crystallographic classification, 201, 204–205, 205–206 crystallographic structure, 217–218, 218–219 three-dimensional schematic diagram, 205 Visible Infrared Imaging Radiometer Suite (VIIRS), 353 visible spectrum (VIS), 285 volume fraction of sea ice constituents, 123–126 air, 124–125 brine, 123–124 pure ice, 124 solid salt, 124 temperature dependence of, 125, 125–126 vorticity, 417, 419 walruses, 16 Ward Hunt Ice Shelf (WHIS), 49, 262–264, 263–265 water atomic and the molecular structure of, 121, 121, 153 molecular composition, 30 wave action marginal ice zone (MIZ), 93–94 pancake ice formation and, 92 wave equation, 135 wavelength, 151 wave propagation, 135 waves classes based on their wavelengths, 331–332 internal, 332 scattering from, 331–332, 332 weather forecasting, 6, 14–15 weather models, 14–15 weather stations, 6–7, 7 Weddell Gyre, 80 wind geostrophic, 421–422, 424 ice deformation and, 68 interaction of the Arctic sea ice with, 582–584 lead formation, 421–423, 424, 426 patterns driving Arctic circulation, 524 polynya, 87–88 sea ice motion, 524–525 wind scatterometer, 287 wind slab, 491 windspeed over open ocean, correction for, 328–330 Wulff net, 214 Yamaguchi decomposition, 316, 318, 320, 387, 469 young ice (YI) emissivity, 395, 399–400 microwave brightness temperature, 374 Mould Bay field experiments, 255, 255 Norwegian Young Sea Ice Experiment (N-ICE), 272–274, 273 physical and electrical properties, 109 salinity, 113 thermal conductivity, 127

WILEY END USER LICENSE AGREEMENT Go to www.wiley.com/go/eula to access Wiley’s ebook EULA.