Sea Ice in the Arctic: Past, Present and Future [1st ed. 2020] 978-3-030-21300-8, 978-3-030-21301-5

Table of contents :
Front Matter ....Pages i-x
Introduction (Ola M. Johannessen)....Pages 1-8
Sea Ice in the Arctic Paleoenvironments (Leonid P. Bobylev, Martin W. Miles)....Pages 9-56
Marginal Ice Zone and Ice-Air-Ocean Interactions (Ola M. Johannessen, Stein Sandven, Richard Davy, Einar O. Olason)....Pages 57-91
Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries (Elena V. Shalina, Ola M. Johannessen, Stein Sandven)....Pages 93-166
Arctic Sea Ice Thickness and Volume Transformation (Elena V. Shalina, Kirill Khvorostovsky, Stein Sandven)....Pages 167-246
SAR Sea Ice Type Classification and Drift Retrieval in the Arctic (Natalia Y. Zakhvatkina, Denis Demchev, Stein Sandven, Vladimir A. Volkov, Alexander S. Komarov)....Pages 247-299
Sea Ice Drift in the Arctic (Vladimir A. Volkov, Alexandra Mushta, Denis Demchev)....Pages 301-313
Sea Ice Modelling (Matti Leppäranta, Valentin P. Meleshko, Petteri Uotila, Tatiana Pavlova)....Pages 315-387
Operational Forecasting of Sea Ice in the Arctic Using TOPAZ System (Laurent Bertino, Jiping Xie)....Pages 389-397
Current and Projected Sea Ice in the Arctic in the Twenty-First Century (Valentin P. Meleshko, Tatiana Pavlova, Leonid P. Bobylev, Pavel Golubkin)....Pages 399-463
Climate Change Impact on the Arctic Economy (Lasse H. Pettersson, Anton G. Kjelaas, Dmitry V. Kovalevsky, Klaus Hasselmann)....Pages 465-506
Annex: SAR Sea Ice Interpretation Guide (Ola M. Johannessen)....Pages 507-573
Back Matter ....Pages 575-575

Citation preview

Springer Polar Sciences

Ola M. Johannessen Leonid P. Bobylev Elena V. Shalina Stein Sandven Editors

Sea Ice in the Arctic Past, Present and Future

Springer Polar Sciences Series editor James Ford, Priestley International Centre for Climate, University of Leeds, Leeds, West Yorkshire, UK

Springer Polar Sciences Springer Polar Sciences is an interdisciplinary book series that is dedicated to research in the Arctic, sub-Arctic regions, and the Antarctic. In recent years, the polar regions have received increased scientific and public interest. Both the Arctic and Antarctic have been recognized as key regions in the regulation of the global climate, and polar ecosystems have been identified to be particularly susceptible to the ongoing environmental changes. Consequently, the international efforts in polar research have been enhanced considerably, and a wealth of new findings is being produced at a growing rate by the international community of polar researchers. Springer Polar Sciences aims to present a broad platform that will include stateof-the-art research, bringing together both science and humanities to facilitate an exchange of knowledge between the various polar science communities. The Series offers an outlet to publish contributions, monographs, edited works, conference proceedings, etc. Topics and perspectives will be broad and will include, but not be limited to: climate change impacts, environmental change, polar ecology, governance, health, economics, indigenous populations, tourism and resource extraction activities. Books published in the series will appeal to scientists, students, polar researchers and policy makers.

More information about this series at http://www.springer.com/series/15180

Ola M. Johannessen • Leonid P. Bobylev Elena V. Shalina • Stein Sandven Editors

Sea Ice in the Arctic Past, Present and Future

Editors Ola M. Johannessen Nansen Scientific Society Bergen, Norway Elena V. Shalina Nansen International Environmental and Remote Sensing Centre (NIERSC) and Saint Petersburg State University Saint Petersburg, Russia

Leonid P. Bobylev Nansen International Environmental and Remote Sensing Centre (NIERSC) Saint Petersburg, Russia Stein Sandven Nansen Environmental and Remote Sensing Center Bergen, Norway University Centre in Svalbard Longyearbyen, Svalbard, Norway

ISSN 2510-0475 ISSN 2510-0483 (electronic) Springer Polar Sciences ISBN 978-3-030-21300-8 ISBN 978-3-030-21301-5 (eBook) https://doi.org/10.1007/978-3-030-21301-5 © Springer Nature Switzerland AG 2020 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG. The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Foreword

Sea ice is recognised to have a significant influence on global climate. It has an impact on the heat exchange between the ocean and atmosphere, a critical influence on the high-latitude atmosphere and freshwater balance, and an essential role in regulating the global thermohaline circulation. “What happens in the Arctic doesn’t stay in the Arctic” has become an important catch phrase. The loss of sea ice is just one effect of the increasingly warmer Arctic, which is warming at a rate two to three times faster than the rest of the planet on average. The change in temperatures at the poles can have a major effect on all parts of the globe because of the interconnected nature of Earth’s climate system. Against this backdrop, the book Sea Ice in the Arctic: Past, Present and Future, edited by Ola M. Johannessen and colleagues, provides essential information and assesses the current situation of Arctic sea ice. Sketching out the long-term perspective, the book treats a wide range of relevant topics. For example, it discusses the reconstruction of paleo-records (spanning a reference frame for the exceptional nature of today’s changes), highlights more recent observations and methods (including measurements from space), and touches upon future trends like the use of artificial intelligence in sea ice classification and corollary effects like the impact of climate change on the Arctic economy. The European Space Agency (ESA) has been addressing Arctic sea ice through various projects like the Climate Change Initiative (CCI), which features sea ice as an Essential Climate Variable (ECV), and through sea ice monitoring data sets from CryoSat, SMOS, Sentinel-1, and Sentinel-3. They are fundamental to data assimilation into coupled models and for delivering operational services through the Copernicus Marine Environment Monitoring Service (CMEMS). The ESA remains committed to continue its endeavours in this regard, as the need for sustained sea ice and ocean data sets in the Arctic is recognised by Copernicus users. Three relevant Copernicus Sentinel high-priority candidate missions are currently being studied: an imaging passive microwave mission (for ice concentration), a polar ice and snow topography mission (for sea ice thickness), and an L-band synthetic aperture radar (SAR) mission as a complement to Sentinel-1. v

vi

Foreword

ESA’s activities in the domain of Arctic sea ice are just one element of a widescale effort to monitor and protect our planet Earth for future generations – a global endeavour by nature – requiring a holistic societal approach. This book is an important element of this endeavour, raising awareness, informing public debate, and sketching a way forward to tackle the strategic topic of Arctic sea ice. Director of Earth Observation Programmes European Space Agency Frascati, Italy

Josef Aschbacher

Contents

1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Ola M. Johannessen

1

2

Sea Ice in the Arctic Paleoenvironments . . . . . . . . . . . . . . . . . . . . . Leonid P. Bobylev and Martin W. Miles

9

3

Marginal Ice Zone and Ice-Air-Ocean Interactions . . . . . . . . . . . . . Ola M. Johannessen, Stein Sandven, Richard Davy, and Einar O. Olason

57

4

Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Elena V. Shalina, Ola M. Johannessen, and Stein Sandven

93

5

Arctic Sea Ice Thickness and Volume Transformation . . . . . . . . . . 167 Elena V. Shalina, Kirill Khvorostovsky, and Stein Sandven

6

SAR Sea Ice Type Classification and Drift Retrieval in the Arctic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247 Natalia Y. Zakhvatkina, Denis Demchev, Stein Sandven, Vladimir A. Volkov, and Alexander S. Komarov

7

Sea Ice Drift in the Arctic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301 Vladimir A. Volkov, Alexandra Mushta, and Denis Demchev

8

Sea Ice Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315 Matti Leppäranta, Valentin P. Meleshko, Petteri Uotila, and Tatiana Pavlova

9

Operational Forecasting of Sea Ice in the Arctic Using TOPAZ System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389 Laurent Bertino and Jiping Xie

vii

viii

Contents

10

Current and Projected Sea Ice in the Arctic in the Twenty-First Century . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399 Valentin P. Meleshko, Tatiana Pavlova, Leonid P. Bobylev, and Pavel Golubkin

11

Climate Change Impact on the Arctic Economy . . . . . . . . . . . . . . . 465 Lasse H. Pettersson, Anton G. Kjelaas, Dmitry V. Kovalevsky, and Klaus Hasselmann

12

Annex: SAR Sea Ice Interpretation Guide . . . . . . . . . . . . . . . . . . . . 507 Ola M. Johannessen

Afterword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575 Ola M. Johannessen

Contributors

Laurent Bertino Nansen Environmental and Remote Sensing Center, Bergen, Norway Leonid P. Bobylev Nansen International Environmental and Remote Sensing Centre (NIERSC), Saint Petersburg, Russia Richard Davy Nansen Environmental and Remote Sensing Center, Bergen, Norway Denis Demchev Arctic and Antarctic Research Institute, Saint Petersburg, Russia Nansen International Environmental and Remote Sensing Centre, Saint Petersburg, Russia Pavel Golubkin Nansen International Environmental and Remote Sensing Centre, Saint Petersburg, Russia Klaus Hasselmann Max Planck Institute for Meteorology, Hamburg, Germany Ola M. Johannessen Nansen Scientific Society (NSS), Bergen, Norway Kirill Khvorostovsky Russian State Hydrometeorological University, Saint Petersburg, Russia Anton G. Kjelaas Norwegian Scientific Academy for Polar Research, Nesoddtangen, Norway Alexander S. Komarov Environment and Climate Change Canada, Ottawa, ON, Canada Dmitry V. Kovalevsky Climate Service Center Germany (GERICS), HelmholtzZentrum, Geesthacht, Germany Matti Leppäranta Institute for Atmospheric and Earth System Research (INAR), University of Helsinki, Helsinki, Finland

ix

x

Contributors

Valentin P. Meleshko Voeikov Main Geophysical Observatory, Saint Petersburg, Russia Martin W. Miles NORCE Norwegian Research Centre, Bjerknes Centre for Climate Research, Bergen, Norway Institute of Arctic and Alpine Research University of Colorado, Boulder, CO, USA Alexandra Mushta Nansen International Environmental and Remote Sensing Centre, Saint Petersburg, Russia Einar O. Olason Nansen Environmental and Remote Sensing Center, Bergen, Norway Tatiana Pavlova Voeikov Main Geophysical Observatory, St. Petersburg, Russia Lasse H. Pettersson Nansen Environmental and Remote Sensing Center, Bergen, Norway Stein Sandven Nansen Environmental and Remote Sensing Center (NERSC), Bergen, Norway University Centre in Svalbard, Longyearbyen, Svalbard, Norway Elena V. Shalina Nansen International Environmental and Remote Sensing Centre (NIERSC) and Saint Petersburg State University, Saint Petersburg, Russia Petteri Uotila Institute for Atmospheric and Earth System Research (INAR), University of Helsinki, Helsinki, Finland Vladimir A. Volkov Nansen International Environmental and Remote Sensing Centre, Saint Petersburg, Russia Jiping Xie Nansen Environmental and Remote Sensing Center, Bergen, Norway Natalia Y. Zakhvatkina Arctic and Antarctic Research Institute, Nansen International Environmental and Remote Sensing Centre, Saint Petersburg, Russia

Chapter 1

Introduction Ola M. Johannessen

Fridtjof Nansen, the Norwegian oceanographer, was the first scientist to explore and study the interior part of the Arctic Ocean with the polar ship Fram’s drift across the Arctic Ocean during the period 1893–1896. Nansen carried out new and fundamental studies of the Arctic Ocean with the discovery of the Transpolar Drift Stream, which was hypothesised to exist by the Norwegian meteorologist Henrik Mohn. Furthermore, Nansen discovered that the Arctic Ocean was deep, more than 4000 m, and that the ice drifted to the right of the wind direction because of the Coriolis force, which later became the Ekman Spiral, a phenomena fundamental to wind-driven circulation in coastal and world oceans. Nansen also observed the oceanographical structure of the Arctic Ocean including the dead water and internal waves. Since Nansen’s Fram drift, the Arctic sea ice has decreased dramatically, particularly in recent decades and in particular during summertime. From the EuroGOOS ice information system, coordinated by the Nansen Center in Bergen, arctic-roos.org, one can follow the daily development of the ice extent and area since the end of 1978 when satellite microwave data became available. For example on 5 September 1979, the ice area was 6.4 million km2, while in 2012, the record minimum of 3.3 million km2 was reached on 12 September, see arctic-roos.org. In general, the negative trends in both summer and winter ice extent are caused primarily by the increasing emission of greenhouse gasses, mainly CO2. The yearly minimum in September has decreased by 10.5% per decade during the period 1979– 2018, or approximately 40% since 1979, while the yearly maximum in March has decreased by 2.6% per decade or 10% since 1979, and the annual mean by 4.4% per decade or 17% since 1979. See Chap. 4 for extensive analyses of the changes in seaice area and extent for the twentieth and twenty-first centuries, and Chap. 5 for the decreases in ice thickness and volume.

O. M. Johannessen (*) Nansen Scientific Society, Bergen, Norway e-mail: ola.johannessen@nansenscientificsociety.no © Springer Nature Switzerland AG 2020 O. M. Johannessen et al. (eds.), Sea Ice in the Arctic, Springer Polar Sciences, https://doi.org/10.1007/978-3-030-21301-5_1

1

2

O. M. Johannessen

Coupled climate models all project that the summer ice will melt during this century under the impact of an increasing atmospheric CO2 concentration in global warming scenarios, but the spread between these models is significant – see Chap. 10 for extensive analyses of coupled climate models for the Arctic and projection for the twenty-first century. However, natural modes of climate variability such as the Atlantic Multidecadal Oscillation (AMO) with a period of 60–80 years, the Pacific Multidecadal Oscillation (PDO) with a more irregular period of 10–15 years, and the yearly winter North Atlantic Oscillation (NAO) are modulating the anthropogenic warming and the variability of the sea-ice patterns. The book also deals with the past variability of the sea ice in Chap. 2, the processes in the marginal ice zone and in the boundary layer above and below the ice and their interaction in Chap. 3, the application of Synthetic Aperture Radar (SAR) for ice-type classification in Chap. 6, the ice drift in the Arctic Ocean and surrounding seas in Chap. 7, sea-ice modelling and forecasting in Chaps. 8 and 9, and the climate impact on economy and society – Chap. 11. The dramatic decreases in the ice extent, ice area, ice thickness and volume in the Arctic Ocean and surrounding seas motivated us to write this book, which was funded primarily by the European Space Agency, Earth Observation Programmes with Ola M. Johannessen as the project leader. Additional funding was granted by the Nansen Scientific Society, Bergen, Norway to the Nansen Center in St. Petersburg, Russia for completing the book and a travel grant for authors to participate in book meetings at the Nansen Center in St. Petersburg, from the Research Council of Norway Project, “Arctic cooperation between Norway, Russia, India, China and US in satellite Earth Observation and Education (ARCONOR)” coordinated by the Nansen Center in Bergen. The book would not be completed without the in kind contributions from all the authors. Richard Davy and Martin Miles also helped in the editing of several Chapters of the book. We also acknoweldge the contribution from the late Dr. Vitaly Yu. Alexandrov from the Nansen Centre in St. Petersburg to Chaps. 2, 4 and 5. The book consists of 12 chapters including a Preface, Afterword and an Annex (Chap. 12) dealing with a SAR Ice Interpretation Guide, summarized below including a map of the Arctic Ocean and the surrounding seas (Fig. 1.1).

1.1

Chapter 2: Sea Ice in the Arctic Paleoenvironments

The chapter begins with an overview of the geological time scales of Arctic paleoenvironments and the diverse types of proxies for reconstructing sea ice in the past based on paleoenvironmental records, primarily marine sediment cores retrieved from the shelves and the deep sea. There is then a summary of knowledge on Arctic sea ice in the distant geologic past, i.e., pre-Quaternary as well as the Pleistocene epoch of the Quaternary including its glacial and interglacial stages and the more recent geologic past, namely the Holocene epoch since the end of the most recent ice age around 12,000 years ago. Here the focus is on the early- to

1 Introduction

3

Fig. 1.1 Map of the Arctic Ocean. (Source: https://www.mapsland.com/world/arctic-region/largescale-political-map-of-arctic-region-with-relief-2009. Copyright: Mapsland)

mid-Holocene – when Arctic temperatures were as warm or warmer than they are at present-day – and the subsequent neoglacial cooling. The final part of the chapter presents in more detail Arctic sea-ice variability during the past millennium, which is of interest with respect to major climate transitions, and a time period for which there are relatively extensive, high temporal resolution data records from paleoenvironmental and historical archives, especially from the Atlantic Arctic. While natural variability on multidecadal and century time scales is large, the exceptional nature of the twentieth and twenty-first centuries is evident in the long-term paleo perspective.

1.2

Chapter 3: Marginal Ice Zone and Ice-Air-Ocean Interactions

The marginal ice zone is the region where the ice cover meets the open ocean. In this region, important mesoscale processes take place such as: (1) the generation of ice-ocean eddies causing the melting of the ice edge, (2) ice-edge wind-driven

4

O. M. Johannessen

upwelling important for marine productivity, (3) internal wave generation causing mixing of the water masses and (4) surface wave penetration into the ice cover, breaking it up and causing lateral ablation during summertime. All of these processes are reviewed including the future of the marginal ice zone, which may not exist during summertime because of melting of the ice cover in the twenty-first century forced by the increasing atmospheric concentrations of CO2. This chapter also presents a review of the key processes which govern the thermal and dynamical interaction between the sea ice, ocean, and atmosphere in the Arctic. It begins with a summary of the importance of sea ice in shaping the surface coupling in the Arctic. The review is then divided into the perspective of those coupled interactions which predominately affect the sea ice and atmosphere, and those which primarily affect the ocean and ice. It is reviewed what is known about these individual processes from observations and models, how important they are in shaping the surface conditions in the Arctic, and to what degree the key features of each process are understood. The chapter concludes with a summary of how surface coupling is expected to change in the future as the Arctic transitions towards a more seasonal ice cover.

1.3

Chapter 4: Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

This chapter describes datasets of the spatial and temporal distribution and variability of the sea ice in the Arctic. The earliest observations are a part of the Arctic Climate System Study (ACSYS) Historical Ice Chart Archive that contains maps showing ice edge position from ships, newspapers, and even letters and diaries that go back to 1553. The chapter gives an overview of the most important sources of ice information including sea-ice archives with historical data, different organizations’ portals that provide ice charts and satellite sea ice databases. Since 1979, satellites have provided a consistent continuous record of sea ice cover. Special attention is paid to sea-ice data sets for climate monitoring and model validation, which have the status of climate data records (CDR). The unprecedented Arctic sea ice decline observed in recent decades is described and analysed and compared with historical records.

1.4

Chapter 5: Arctic Sea Ice Thickness and Volume Transformation

This chapter deals with ice thickness, which is one of the most important variables of sea ice. It describes techniques for the measurement of ice-thickness, including drilling, the use of upward looking sonars, electromagnetic induction systems, airand space-borne lasers, radar altimetry, and the use of infrared and passive

1 Introduction

5

microwave satellite data. It describes data sets collected in different periods of time that can be used for the studies of sea-ice thickness change and for validation of new methods of sea-ice thickness retrieval including model simulations. Arctic-wide observations of sea-ice thickness are essential for estimating sea ice volume, the changes of which impact on regional heat and freshwater budgets, on patterns of atmospheric circulation in the Arctic and in general the climate. The chapter reviews present knowledge of sea-ice thickness and volume change including uncertainties. The negative long-term trend of sea-ice volume since 1979 has been simulated by the PIOMAS model that showed a 70% volume decrease in the September minima from 1979 to the present and a 22% decrease in the March maxima during the same period.

1.5

Chapter 6: SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

The chapter describes radar images for ice cover monitoring including the technique of calibration of Envisat, Radarsat-2 and Sentinel-1 SAR images. It introduces the technique of angular correction of SAR images, acquired in a wide swath mode from various satellites, which makes it possible to better separate different types of sea ice over the range of the entire SAR images. A description of the noise reduction for SAR images in HV-polarization is also presented. The chapter presents the main approaches to the automated classification of sea ice images. The ranges of the normalized radar cross section (NRCS) for some ice types are overlapped, which motivated the use of additional information of the texture in SAR images as input for the automatic classification algorithm. The application of neural networks (NN) and support vector machines (SVM) methods for sea ice classification of SAR images are discussed including a brief description of the methods. An algorithm for the automated identification of sea ice from SAR images using a model of backscatter NN and SVM is developed and validated.

1.6

Chapter 7: Sea Ice Drift in the Arctic

The chapter presents analysis of sea-ice drift in the Arctic Ocean. The first section describes the drivers determining the spatial distribution of sea ice in the Arctic Ocean such as wind regime, currents, water exchange, and freshwater runoff. The second section presents two main observational datasets for sea-ice drift study and analysis – the IFREMER and Pathfinder datasets. Since these datasets include a large amount of ice-drift vector information, it requires a special approach to operate with this information, so-called vectorial-algebraic analysis. Vectorial-algebraic analysis allows the initial information to be significantly compressed. For the sea ice drift speed vector this approach aggregates effects of variability of both its module and

6

O. M. Johannessen

direction. The fourth section describes the variability of large-scale sea ice circulation in the Arctic Ocean based on the multi-year satellite observational data and its connection with the synoptic atmospheric processes. Years with a well-developed anticyclonic gyre in the Canadian basin and the Transpolar current in the Eurasian part are analysed along with the identification of the synoptic atmospheric processes driving them.

1.7

Chapter 8: Sea Ice Modelling

The first and second sections introduce the theory of sea-ice geophysics necessary to understand mesoscale and large-scale models. Sea ice is described across the range of spatial scales. Then the fundamental laws of thermodynamics and dynamics are given. They can be treated separately, since in a large ice floe, thermodynamics is a one-dimensional problem while dynamics when integrated across sea-ice thickness becomes a horizontal problem. They are linked together by the ice conservation law. Forcing of sea-ice dynamics and thermodynamics is via atmospheric and oceanic boundary layers, where parameterizations of the exchange of momentum, heat, moisture, and salinity are of key importance. Sea-ice properties relevant to climate are primarily extent, concentration and thickness. Mesoscale and large-scale sea ice models treat sea ice as a continuum by thickness distribution. These models consist of a momentum equation, conservation laws of heat, salinity, and ice rheology. The main models used for climate investigations are CICE and LIM, which are reviewed including their parametrizations. Aspects of data assimilation are discussed where modelling is closely linked to sea ice remote sensing data. Then the performance of sea ice simulations by CMIP3 and CMIP5 models are discussed. However, CMIP5 inter-model spread in sea ice simulations does not appear to have been appreciably reduced as compared to that in CMIP3 models. Spatial distribution of annual mean ice thickness also varies considerably among the models. The sea-ice assimilation system PIOMAS is briefly described which provides a way of estimating sea-ice thickness and volume.

1.8

Chapter 9: Operational Forecasting of Sea Ice in the Arctic Using TOPAZ System

This chapter describes the Arctic Marine Forecasting Center (ARC MFC) which provides 10-day forecasts of ocean currents, sea ice, marine biogeochemistry and waves on a daily basis, and a 25-year reanalysis of the Arctic Ocean updated every year. The ARC MFC is powered by the TOPAZ configuration of the HYCOM model, coupled to the sea-ice model CICE, the ecosystem model ECOSMO, and assimilates the following data with the Ensemble Kalman Filter: along-track sea

1 Introduction

7

level anomalies, sea-surface temperatures, sea-ice concentrations, sea-ice drift, sea-ice thickness and in-situ temperature and salinity profiles. The ARC MFC constitutes the Arctic component of the European Copernicus Marine Environment Monitoring Services, including recent developments of wave–ice interactions and assimilation of sea-ice thickness data.

1.9

Chapter 10: Current and Projected Sea Ice in the Arctic in the Twenty-First Century

Coupled global models are the major objective tools available to provide future climate projections based on physical laws that govern ocean–atmosphere circulation and sea ice. All models forced with increasing greenhouse-gas (GHG) concentrations simulate persistent reductions of Arctic sea ice during the late twentieth century and in the twenty-first century. But the rate of projected sea ice loss is subject to considerable uncertainty resulting from different assumptions on future external forcing, internally generated variability and inter-model differences in parameterization of oceanic and atmospheric physics. According to sea-ice projections with the RCP4.5 emission scenario, 32% of the 56 CMIP5 ensemble members reach nearly ice-free conditions during September by the end of this century, while some show a nearly ice-free state as early as 2020. Recent studies have indicated that sea-ice seasonal variability in such regions as the Barents-Kara Seas and the East Siberian-Chukchi Seas appears to be major drivers of extremely cold weather over Eurasia and North America. For instance, the Barents-Kara Seas are an active area for rapid Arctic sea-ice decrease that favours intensive air-sea interaction and significant warming of the troposphere. However, causal relationships between sea-ice decrease and the atmospheric circulation anomalies that results in increased frequency of cold extremes are not fully understood. Finally, some studies suggested that the rapid decline of summer Arctic sea ice in recent decades may be partially due to an enhanced Atlantic heat transport into the Arctic. On decadal to multi-decadal time scales, internal variability could also temporarily counter changes in anthropogenic radiative forcing and cause a hiatus in the decline of Arctic sea ice.

1.10

Chapter 11: Climate Change Impact on the Arctic Economy

The chapter starts with a brief assessment of the current scientific and operational use of satellite Earth observations (EO) for monitoring of the Arctic Ocean and the surrounding seas, including the various data and applications discussed in this book.

8

O. M. Johannessen

Their relevance to international programs contributes to demonstrate the societal benefits and economic value of the use of EO monitoring in the Arctic. Direct impact of changes in the Arctic on the economy is discussed through an assessment of the potential use and exploitation of its natural resources. We have selected and limited our discussions to the current status and future of the oil-and-gas offshore industry, marine transportation and – to some extent potential fisheries – in the Arctic seas. The economic impact of these three areas of activities is also discussed from an economic modelling perspective. At the end of the chapter, the climate-related risks for the development of the economy in the Arctic are briefly discussed.

1.11

Annex (Chap. 12): SAR Sea Ice Interpretation Guide

This interpretation guide contains a selection of SAR images illustrating various sea-ice conditions and phenomena demonstrating their characteristic SAR signatures covering the European sector of the Arctic from the east coast of Greenland to the western Laptev Sea in Russia. The images were acquired at various times of the year during different weather conditions, in order to illustrate both short-term and longterm variability in SAR ice characteristics. The images include examples of the most important types of ice, such as multiyear ice, first-year ice, and various stages of young and new ice. Processes in the ice-edge region as well as the ice-edge response to wind variations are illustrated. The interpretation guide also includes examples of phenomena observed near the coast, such as landfast ice, shear zones and polynyas and processes taking place in the interior of the pack ice fields, such as leads, ice motion and break-up of floes. The purpose of this interpretation guide is to promote the use of SAR imagery in operational monitoring and forecasting including ship routing. Furthermore it is to encourage scientists and students to use SAR in their ice studies.

1.12

Afterword

The Afterword reflects on the past and present climate and briefly points out what needs to be done for improving the future projections of the sea ice and Arctic climate by coupled climate models.

Chapter 2

Sea Ice in the Arctic Paleoenvironments Leonid P. Bobylev and Martin W. Miles

In this chapter we consider the long-term natural variability and changes in Arctic sea ice, based on evidence from reconstructions using paleoenvironmental proxy records and historical records. Sea-ice reconstructions provide a long-term paleo perspective for understanding the changes in sea ice observed in the most recent decades. Here we provide an overview of past variability and changes in Arctic sea ice spanning a range of climate regimes, including previous periods when the climate was as warm or wamer than present day – possible analogs for the future of the Arctic sea-ice cover. Section 2.1 provides an overview of the geological time scales of Arctic paleoenvironments. Section 2.2 summarizes the diverse types of proxies and methods for reconstructing past sea-ice conditions, based on paleoenvironmental records, primarily marine sediment cores retrieved from the shelves and the deep sea. Section 2.3 summarizes knowledge on Arctic sea ice in the distant geologic past, namely the pre-Quaternary and the Pleistocene epoch of the Quaternary including its glacial and interglacial stages. Section 2.4 presents Arctic sea-ice variability in the more recent geologic past, namely the Holocene epoch since the end of the most recent ice age around 12,000 years ago. Here the focus is on the early- to mid-Holocene – when Arctic temperatures were as warm or warmer than at present-day – and the subsequent Neoglacial cooling. Section 2.5 presents in more detail Arctic sea ice variability during the past millennium of the Holocene, which is of interest with respect to major climate transitions, and a period for which there are relatively extensive, high temporal resolution data records from palaoenvironmental L. P. Bobylev (*) Nansen International Environmental and Remote Sensing Centre (NIERSC), Saint Petersburg, Russia e-mail: [email protected] M. W. Miles NORCE Norwegian Research Centre, Bjerknes Centre for Climate Research, Bergen, Norway Institute of Arctic and Alpine Research University of Colorado, Boulder, CO, USA e-mail: [email protected] © Springer Nature Switzerland AG 2020 O. M. Johannessen et al. (eds.), Sea Ice in the Arctic, Springer Polar Sciences, https://doi.org/10.1007/978-3-030-21301-5_2

9

10

L. P. Bobylev and M. W. Miles

and historical archives. Section 2.6 describes the paleo and historical evidence for sea-ice variability on multidecadal time scales. The material focused on the past millennium and most recent centuries is more comprehensive and up-to-date than found in the standard history of Arctic sea ice by Polyak et al. (2010). Section 2.7 presents a concise summary statement.

2.1

Arctic Paleoenvironments

Arctic sea ice is a keystone indicator of ongoing anthropogenic global warming, which is amplified in the Arctic due to feedback mechanisms. Modelling studies predict further dramatic decline of the Arctic sea ice cover, and even the complete disappearance of summer sea ice in this century (ACIA 2004; Johannessen et al. 2004; Johannessen 2008; IPCC 2013). The observed recent changes in sea ice in the Arctic are caused by two major factors: (1) response of sea ice to anthropogenic increase in greenhouse gases (GHGs) (Johannessen 2008; IPCC 2013); and (2) natural variability of sea ice related to oceanic and atmospheric circulation patterns, i.e., internal climate-system variability (Bengtsson et al. 2004; Johannessen et al. 2004; Serreze et al. 2007; Stroeve et al. 2007; Miles et al. 2014; Ding et al. 2017). Natural variability on seasonal to multidecadal time scales plays a significant role in the recent changes. Whatever the anthropogenic climatic effects may be in the future, they will be superimposed on a climate system that responds to natural external and internal forcing factors (Bradley and Jones 1995). Thus, a major challenge consists of assessing of the degree to which sea ice is related to GHG forcing versus natural variability.The natural variability of the sea-ice distribution in the Arctic has been thoroughly investigated for the most of the twentieth century based on satellite and aerial reconnaissance data (e.g., Johannessen et al. 1995; Mahoney et al. 2008; Parkinson and Cavalieri 2008) and sea ice reconstructions (e.g., Bobylev et al. 2003; Johannessen et al. 2004; Zakharov 2004). However, these data cover only the latest period of the Arctic sea ice history. Over geological time periods, the Earth’s natural climate system has been driven by various external factors. Over millions of years, tectonic processes have affected the location and topography of the continents through plate tectonics and ocean spreading. Orbital changes – the so-called Milankovitch cycles – occurring over tens to hundreds of thousands of years and altering the amount of received solar radiation, have driven climate responses including the growth and decay of ice sheets at high latitudes. Changes in solar emissivity, occurred over decades, centuries, and millennia, have also affected the Earth’s temperature regime. Due to the influence of these external forces, global climate, Arctic climate and sea ice have experienced variability over time scales ranging from decades to millions of years (McBean et al. 2004). Different spans of time on the geological time scale are usually delimited by changes in the composition of strata, which correspond to them, indicating major geological or paleontological events, such as mass extinctions. The largest defined

2 Sea Ice in the Arctic Paleoenvironments

11

Geological Time Scale

50

Neoproterozoic

100

Proterozoic

1000

3000

Precambrian

10

0 0.0117

Miocene

20

Epochs Holocene

Pleistocene 2.58

Oligocene

30

40

50

250

Eocene

Permian Carboniferous

Neoarchean

3500

Pliocene

Triassic

300

Archean

Numericalage (Ma)

2500

Quaternary

Jurassic

200

Paleoproterozoic

2000

150

Cretaceous

350

Mesoarchean

400

Paleoarchean

450

Paleozoic

1500

Mesoproterozoic

Paleogene

Epochs

0 Quaternary

Neogene

Paleozoic

Periods Neogene

Paleogene

0

Cenozoic

Phanerozoic 500

Eras Cenozoic Mesozoic

Mesozoic

Eons Present 0

Pennsylva -nian

60

Paleocene

Mississippian

Devonian Silurian Ordovician

Eoarchean 500

4500

Hadean

4000

Cambrian

Fig. 2.1 Geological time scale showing numerical age of eons, eras, periods and epochs in million years before present (Mya – Mega-annum) (Adapted from the International Chronostratigraphic Chart, v. 2018/08, http://www.stratigraphy.org/index.php/ics-chart-timescale)

unit of time is the supereon, composed of eons. Eons are divided into eras, which are in turn divided into periods, epochs and ages (Fig. 2.1). The time scales of the eras, periods and epochs most relevant for the history of Arctic sea ice are specified in Table 2.1. The Arctic Ocean was formed in the Mesozoic Era, which is comprised of the Triassic, Jurassic and Cretaceous periods. The Mesozoic Era witnessed gradual rifting of the supercontinent Pangaea into separate landmasses that would eventually move into their current positions. Sea level began to rise during the Jurassic and the formation of new crust beneath the surface displaced sea level by 200 m higher than today. The sole major Mesozoic orogeny, which occurred in what is now the Arctic, was related to the opening of the Arctic Ocean. The size and position of the Arctic Ocean have changed as crustal plates have adjusted. By the Cretaceous, the present position of the Arctic Ocean was approached, but the ocean was only half as large as it is at present. Details of the pre-Cretaceous Arctic Ocean are unknown (Clark 1982). The climate of the Mesozoic was varied, alternating between warming and cooling periods, but overall, the Earth was hotter than it is today. From 120 to 90 million years ago (mya) during the Cretaceous period, the Arctic was significantly

12

L. P. Bobylev and M. W. Miles

Table 2.1 Time scales of the eras, periods and epochs referred to in this chapter. Numerical ages are from the International Chronostratigraphic Chart, v. 2018/08, http://www.stratigraphy.org/ index.php/ics-chart-timescale Era Cenezoic

Mesozoic

Numerical age (Mya) 66.0–present

252.2–66.0

Period Quaternary

Numerical age (Mya) 2.58–present

Neogene

23.03–2.58

Paleogene

66.0–23.03

Cretaceous Jurassic Triassic

145.0–66.0 201.3–145.0 252.2–201.3

Epoch Holocene Pleistocene Pliocene Miocene Oligocene Eocene Paleocene

Numerical age (Mya) 0.0117–present 2.58–0.0117 5.333–2.58 23.03–5.333 33.9–23.03 56.0–33.9 66.0–56.0

warmer than at present, and Arctic geography, atmospheric composition, ocean currents, and other factors were quite different from those at present (McBean et al. 2004). The climate of the Cretaceous remains widely disputed. There is a hypothesis that higher levels of CO2 in the atmosphere significantly decreased the world temperature gradient from north to south, and temperatures were about the same across the planet. Average temperatures were also higher than today by about 10
C. However, some data do not support this hypothesis, and it is also possible that temperature fluctuations have been sufficient for the presence of polar ice caps and glaciers. The Cenozoic Era is the current and most recent of the three Phanerozoic geological eras, following the Mesozoic Era. The Cenozoic is the era when the continents moved into their current positions. The global climate during the Paleogene, which comprised the first part of the Cenozoic, departed from the hot and humid conditions of the late Mesozoic Era and began a drying and cooling trend. However, the end of the Paleocene was marked by a significant global warming event, the Paleocene–Eocene Thermal Maximum (PETM) of 55.8 mya (Fig. 2.2). The climate became warm and humid worldwide with subtropical vegetation growing in Greenland and Patagonia, and warm seas circulated throughout the world, including the poles. Supercontinent Laurasia had not yet separated into three continents – Europe and Greenland were connected, North America and Asia were still intermittently joined by a land bridge, but Greenland and North America were beginning to separate. The modern basin of the deep Arctic Ocean began to develop in the Cenozoic Era, and there is the distinct difference between the Arctic Ocean of the Early Cenozoic and that of the Late Cenozoic. Climate evolution during the Eocene began with warming after the end of the PETM to a maximum during the Eocene Optimum at around 49 mya. During this period, little to no ice was present on Earth with an insignificant difference in

2 Sea Ice in the Arctic Paleoenvironments

13

Fig. 2.2 Generalized climate change during the Cenozoic Era of the last 65 million years, with emphasis on Polar Regions (Source: Wikepedia Commons)

temperature from the equator to the poles. However, a period of long-term cooling began since 49 mya, and ice began to reappear at the poles to the Eocene-Oligocene transition at 34 mya, when the Antarctic ice sheet began to expand rapidly. The large-scale development and build-up of ice sheets in the Arctic probably commenced during the period from 38 to 1.6 mya with ice accumulation facilitated initially by plate tectonics (McBean et al. 2004). The Oligocene is the third and final epoch of the Paleogene Period. During the Oligocene after the tectonic creation of Drake Passage, when South America fully detached from Antarctica, the climate cooled significantly due to the advent of the Antarctic Circumpolar Current, which brought cool deep Antarctic water to the surface. The formation of permanent Antarctic ice sheets during the early Oligocene and possible glacial activity in the Arctic may have influenced oceanic cooling. The oceans continued to cool as the Oligocene progressed. The Oligocene saw the beginnings of modern ocean circulation, with tectonic shifts causing the opening and closing of ocean gateways. The Neogene is the second period in the Cenozoic Era, following the Paleogene Period. The global climate became seasonal and continued its overall drying and cooling trend. The ice caps on both poles began to grow and thicken, and by the end of the period the first of a series of glaciations of the Ice Ages began. The Arctic region cooled in the Miocene due to the strengthening of the Gulf Stream current, when South America became attached to North America. Oceans cooled partly due to the formation of the Antarctic Circumpolar Current, and the ice cap in the southern hemisphere started to grow to its present form about 15 mya. The tectonically controlled widening of the Fram Strait in the Miocene (~17.5 mya) allowed commencing a two-way surface exchange between the Arctic Ocean and Norwegian and Greenland seas. The Fram Strait had attained sufficient width to support opposing

14

L. P. Bobylev and M. W. Miles

exit and entry surface currents at 14 mya (Jakobsson et al. 2007). In about 5.2–6.5 mya, global lowering of sea level caused maximal isolation of the Arctic Ocean from the World Ocean. This shallowing of the Arctic Ocean could have led only to greater warmth and salinity of its waters and the sea was ice-free (Danilov 1989). Continents continued to drift in the Pliocene, moving to positions only 70 km from their present locations. With the formation of the Isthmus of Panama, warm equatorial ocean currents were cut off and an Atlantic cooling cycle began, with cold Arctic and Antarctic waters dropping temperatures in the now-isolated Atlantic Ocean. In Early Pliocene, about 3.5–5 mya sea level rose and a major transgression took place throughout the Arctic Basin (Danilov 1989). During the Pliocene epoch climate became cooler and drier, and seasonal, similar to modern climate. The global average temperature in the mid-Pliocene (3.3–3 mya) was 2–3
C higher than today, global sea level 25 m higher. Marine deposits postdate the opening of the Bering Strait at slightly before 3 mya (Darby 2008). The World Ocean records indicate that ice began to accumulate on the continents of the Northern Hemisphere in Late Pliocene (about 2–3 mya) with the opening of the Bering Strait. According to Jansen et al. (2000), glaciers have existed in the areas surrounding the Nordic Seas since at least the late Miocene. Progressive intensification of glaciation seems to have begun from 3.5 to 3 mya, when the Greenland Ice Sheet expanded to northern Greenland (McBean et al. 2004). Jansen et al. (2000) document the stepwise inception of largescale glacial cycles in the Northern Hemisphere. The first step marks an expansion of the Greenland ice sheet at 3.3 mya, and the second at 2.74 mya, when main ice sheets in the Northern Hemisphere started to produce icebergs. The main establishment of Northern Hemisphere glaciation apparently occurred at 2.7 mya, when major peaks in ice-rafted debris (IRD) appear from the Northern Hemisphere ice sheets. The Quaternary Period, which is the most recent of the three periods of the Cenozoic Era, is characterized by overall low temperatures and especially large swings in climate regime related to changes in insolation modulated by Earth’s orbital parameters (Polyak et al. 2010). The Quaternary starts at the onset of the Northern Hemisphere periodic glaciations approximately 2.6 mya, when continental glaciers expanded from the Polar Region to 40
N latitude. The global climate system during the last 1.6 mya has been characterized by periodic climatic variations and has switched between interglacial and glacial stages, with further subdivision into stadials (shorter cold periods) and interstadials (shorter mild episodes) (Fig. 2.3). Climatic conditions during some interglacial periods were generally similar to those of today (McBean et al. 2004). Vast Northern Hemisphere ice sheets melted away during deglaciations, sea level rose about 120 m, atmospheric CO2 increased by about 100 parts per million by volume (ppmv), and interglacial climate emerged across the planet. The Late Quaternary glacial cycles varied between 80,000 and 120,000 years in length, with an average recurrence interval of about 100,000 years (Denton et al. 2010). The records within the last 1.2 mya document a shift toward glaciations of longer duration as well as an evolution toward more pronounced interglacial periods (Jansen et al. 2000).

2 Sea Ice in the Arctic Paleoenvironments

15

Fig. 2.3 Ice core data for temperature (red) and atmospheric CO2 (blue) related to glacial– interglacial cycles. Light red shading indicates warm periods (interglacials). Numbers refer to marine isotope stages (MIS) (Modified from EPICA Dome C ice core program, data from NOAA www.climate.gov)

The dramatic changes in sediment composition, biogenic preservation, and IRD abundance in the ACEX record and adjacent cores during the last 190,000 years are linked to advances and retreats of the Barents–Kara ice sheet (Polyak et al. 2010). The average air temperature in the warm half-year at 65
N during periods of maximum glaciation in about 187, 116, 72, and 22,000 years ago (kya) was lower than contemporary temperatures by 6.4, 6.5, 5.9, and 5.2
С, respectively (Barkov and Petrov 1994). The last interglacial – also called the Eemian – extended from the end of the penultimate glaciation about 130 kya until about 107 kya when the last glacial period began (McBean et al. 2004). According to most proxy data, the last interglacial was slightly warmer everywhere than at present (IPCC 2013). After the cooling event at about 107 kya, climatic conditions often changed suddenly, followed by several thousand years of relatively stable climate or a temporary reversal to warmth, but overall, there was a decline in global temperatures. Large ice sheets began to develop on all the continents surrounding the Arctic Ocean, and the global ice extent was at its greatest during the Last Glacial Maximum from ~24 to 21 kya. The glacio-eustatic decrease in sea level during this period amounted to about 120 m (McBean et al. 2004). The Pleistocene ice sheet covered North America from the Atlantic to the Pacific with a thickness of up to 2 miles (Fletcher 1968). The Eurasian ice sheet at its maximum extent covered Scandinavia

16

L. P. Bobylev and M. W. Miles

Fig. 2.4 Extent of the Last Glacial Maximum in Eurasia, ca. 19 kya. SBKIS Svalbard–Barents– Kara Seas Ice Sheet, SIS Scandinavian Ice Sheet, BIIS British–Irish Ice Sheet (Source: Hughes et al. (2016), updated from Mangerud et al. (2004))

and the Barents Shelf with its eastern margin located west of the Taymyr Peninsula (Fig. 2.4). The period since the Last Glacial Maximum (20 kya) is characterized by a warming trend with three major stages of deglaciation, interrupted by the cold periods of the Dryas stadials. Deglaciation stage, occurred between 18 and 20 kya, was followed by the Oldest Dryas, dated between approximately 18 and 15 kya. The second stage of deglaciation, occurred 12 to 15–16 kya, was followed by the Older Dryas during the period of ~11.7–12 kya. The third stage of deglaciation at about

2 Sea Ice in the Arctic Paleoenvironments

17

Fig. 2.5 Temperature variations during the Holocene based on various paleoenvironmental reconstructions. The anomalies are plotted with respect to the mid twentieth century average temperature (Source: Wikipedia Commons)

12 kya was suddenly interrupted by another and very short-lived cold event known as the Younger Dryas, which occurred between approximately 12.8 and 11.5 kya and led to a brief advance of the ice sheets. The Holocene is the present interglacial period, which has persisted for about 12 kya. The Arctic summer air temperatures during the warmest part of the period were as much as 2–3
C above present for much of the region, which was well above the interglacial average temperature for the rest of Earth (Fig. 2.5). Multiple proxy records indicating that early Holocene temperatures were higher than today and that the Arctic contained less ice, are consistent with a high intensity of orbitallycontrolled spring and summer insolation that peaked about 11 kya and gradually decreased thereafter. The warming phase after the end of the Younger Dryas was very abrupt and central Greenland temperatures increased by 7
C or more in a few decades (McBean et al. 2004). Arctic summer temperatures were warm enough to melt all glaciers below 5 km elevation, except the Greenland Ice Sheet, which was reduced moderately. The last major ice sheet disappeared from Scandinavia about 8000–7000 BC, while in North America the ice retreated completely at an even later date (Frolov et al. 2009). The continued Holocene climate warming, which culminated in the “Holocene Climatic Optimum” (HCO) of 5–9 kya, was characterized by a significant increase in mean air temperature, which was generally 2–3
C higher in summer compared to present conditions. However, this warming and its characterization in terms of the “Holocene Thermal Maximum” (HTM) (maximum temperatures within the HCO) has been shown by Kaufmann et al. (2004) to be time-transgressive (i.e., geographically asynchronous) in the Arctic. Therefore, the response of the sea ice to warming should be expected to reflect this regional variability – the same as for the subsequent

18

L. P. Bobylev and M. W. Miles

Neoglacial cooling that began roughly 4–5 kya, caused by decreasing orbital-driven summer insolation. During the last 2000 years, several climate periods can be identified in the Northern Hemisphere. Generally warm and dry period in Europe, known as the “Roman Warm Period” (RWP), lasted from ~200 BC to AD 400, was followed by cooler and wetter conditions during the “Dark Ages Cold Period” (DACP), ~400 to 600. Climatic improvement following the Dark Ages continued with another exceptionally warm period from ~600 to 760, which corresponded to the onset of the “Medieval Warm Period” (MWP) also known as the “Medieval Climate Anomaly” (MCA). The warming in Iceland preceded the rest of Western Europe, where the maximum warmth began ~ AD 800–850 (Patterson et al. 2010). The MCA spanned roughly 1000–1400 years (Henshaw 2003), and the average air temperature in this epoch was approximately 1.5
C higher than the subsequent cold period and slightly above late twentieth century temperatures (Frolov et al. 2009). The eighth to fourteenth centuries were characterized by favourable navigation conditions in the North Atlantic and Viking colonization of the Greenland coast and part of North America (Borisenkov 1995). The MCA was succeeded by the so-called “Little Ice Age” (LIA), which began around 1400 or some decades earlier (e.g., Moffa-Sánchez et al. 2014). The LIA lasted until about 1850 or some decades later, again depending on the region (Gribbin and Lamb 1978; IPCC 2013). During the late LIA in the 1880s and early 1900s, the coasts of Greenland and Iceland were frequently bound by ice, and the surface water temperature in the North Atlantic was 2–3
C lower than in the period of warming in the Arctic in the 1920s–1930s. Considering state of sea ice near the coast of Iceland, which is one of the most important indicators of climate change, the culmination of the LIA occurred in the period 1780–1820. The twentieth century is warmer than any other time in the last 500 years, but this warming did not start suddenly. A slight warming trend is observed from the end of eighteenth century, and significant changes in atmospheric circulation and climate took place during the 1820s (Zakharov 2004). Sea ice during these periods in the past millennium is discussed in Sects. 2.5 and 2.6.

2.2 2.2.1

Proxies for Reconstructing Sea Ice in the Arctic Marine Records

The sparseness of instrumental climate records prior to the twentieth century means that estimates of climate variability in the Arctic during past centuries must rely upon indirect “proxy” paleoclimate indicators, which may provide evidence for prior large-scale climatic changes (McBean et al. 2004). Direct indication of past sea-ice conditions is found in marine sediments, which may include particles entrained and dispersed by sea ice (Nørgaard-Pedersen 2009). Most of the ocean floor is covered with sediment of varying thickness, which, according their origin, can be classified into terrigenous, biogenous, and hydrogenous sediments. Terrigenous sediments

2 Sea Ice in the Arctic Paleoenvironments

19

originate on land and consist primarily of mineral grains that were eroded from continental shelf and continental rocks and transported to the ocean. The terrigenous sediment is the thickest on continental shelves. Biogenous sediments consist of shells and skeletons of marine animals and algae. Hydrogenous sediments consist of minerals crystallized directly from ocean water through various chemical reactions and make up only a small portion of the ocean’s sediments. A thorough understanding of sea-ice history depends on refining sea-ice proxies in sediment taken from strategically selected sites in the Arctic Ocean and along its continental margins (Polyak et al. 2010). The sediments of the Arctic Ocean floor record the natural conditions of the physical environment, climate, and ecosystems on time scales determined by the ability to sample them through coring and at resolutions determined by the rates of deposition. Sediment cores representing the long-term history of sea ice embracing millions of years can be found in the deep, central part of the Arctic Ocean, where the seafloor was not eroded during periods of lower sea-level and the passage of large ice sheets. Rates of sediment deposition in this area are on the order of centimeters or millimeters per thousand years, and sedimentary records may capture variations on timescales larger than millennial. Cores from the Arctic continental margins usually cover a relatively short time interval since the Last Glacial Maximum (LGM), i.e. the period between 26.5 and 19–20 kya; however, they provide high-resolution records that capture events on century or even decadal time scales (Polyak et al. 2010). Aside from sea-ice biomarker IP25 described below, the most direct proxies for the presence of ice are derived from sediment that melts out or drops from drifting ice. The presence of ice-rafted debris (IRD), defined as the coarse sediment fraction of >63 μm, in seafloor sediment cores indicate that icebergs, sea ice, or both have occurred at that location during a known time interval. Sea ice and glacial ice transport sand-sized sediment and their size for sea ice is usually restricted to about 200–250 mm (Reimnitz et al. 1993). Existence of only seasonal ice can be assumed if all grains require less than a year to reach the area under study. A perennial ice cover is required when significant numbers of ice-rafted grains come from sources of more than 1 year of typical ice drift time (Darby 2008). Sedimentation in the Arctic is predominantly iceberg-derived during glacial periods due to low sea levels and huge ice-sheet fronts surrounding the Arctic Ocean. Contribution of icebergs to sediment transport and deposition in the Arctic is limited during interglacials, whereas the role of sea ice is greatly enhanced due to existence of broad and shallow continental shelves, the major sites of sea ice formation (Polyak et al. 2010). The Laptev Sea is an important sea-ice factory as well as the dominant source of sediments for the areas, located within the main track of the Transpolar Drift (Eicken et al. 1997). Skeletal remains of microscopic organisms in bottom sediment such as foraminifers, diatoms, and dynocysts may indicate the sea ice conditions. Diatoms are single-celled phytoplankton, which are dependent on light, and therefore live in the uppermost part of the oceans. The diatoms are an excellent tool for paleoclimatic reconstructions in high latitude oceans, because of their high productivity and diversity (Patterson et al. 2010). Some diatoms live in or on sea ice. Certain diatom

20

L. P. Bobylev and M. W. Miles

species live only along the margins of sea ice and could therefore be useful in determining past sea-ice coverage (Nørgaard-Pedersen 2009). Remnants of diatoms and dinocysts have been used to infer the former presence of sea ice and even to estimate the length of the ice-cover season (de Vernal et al. 2013). Organisms that require open water can be used to identify intervals of diminished ice. Oxygen isotopes fractionate during sea-ice formation and melting, and δ18O variations in fossil calcite may record the intensity of these processes. The composition of specific organic biomarkers can also be used to characterize the environment in which it formed. A specific biomarker IP25 (ice proxy with 25 carbon atoms), is associated with diatoms living in sea ice, and reflects the occurrence of ice in spring (Polyak et al. 2010). This biomarker is stable for at least 10,000 years in marine sediments and has been used to reconstruct sea-ice cover at a number of sites in the Arctic (Nørgaard-Pedersen 2009). A strong correlation between documented sea ice occurrence and the IP25 proxy was demonstrated by Massé et al. (2008) for a site on the north Iceland shelf.

2.2.2

Terrestrial Records

In many places along the Arctic and subarctic coasts, evidence of the sea ice extent in the past is recorded in landforms such as coastal plains, marine terraces, and beaches. These coastal deposits represent a limited time span and geographic distribution but provide valuable information that can be compared with seafloor sediment records. The spacious coastal exposures enable abundant recovery of larger fossils such as plant remains, driftwood, whalebone, and relatively large molluscs (Polyak et al. 2010). Sea-ice conditions, especially the presence or absence of fast ice, may be inferred from the distribution of driftwood logs, found in raised beaches along the coasts of Canadian Arctic, Greenland, Svalbard, and Iceland. Most of the larch logs found come from the northern Siberia, whereas most of the spruce comes from the Mackenzie and Yukon Rivers (Polyak et al. 2010). Since the voyage takes several years and can only be accomplished if the wood is incorporated in sea ice, it is an indicator of multiyear ice. The changing proportions of larch and spruce indicate changes in the strength of the Transpolar Drift and the Beaufort Gyre. The larch from central Siberia is transported to Greenland by the Transpolar Drift, with a travelling time of 2–5 years. The driftwood from North America, predominantly spruce, has to go into the Beaufort Gyre and make a detour around Siberia before it can join the Transpolar Drift, increasing the traveling time to 6–7 years or more (Funder et al. 2011). Previous ice conditions can also be reconstructed from bowhead whale remains because seasonal migrations of the whale are determined by the oscillations of the sea ice (Polyak et al. 2010). Plant remains and pollen provide a necessary link to information about the past climate throughout Arctic and subarctic regions. The location of the northern tree line, which is presently controlled by the July 7
C mean isotherm, is a critical paleobotanic indicator for understanding Arctic paleoclimate including sea-ice

2 Sea Ice in the Arctic Paleoenvironments

21

conditions. Tree-ring and lake sediments from circum-Arctic sites reflect coastal climate conditions, which may also be linked to sea ice cover (Kinnard et al. 2011). Ice cores from glaciers and ice sheets preserve signals of atmospheric temperature, moisture source and marine aerosol loadings, which may be linked to sea ice conditions (Kinnard et al. 2011). Studies of the ice cores of Greenland showed significant increase of chemical elements concentration in the ice during cold periods, which is caused by increase of aerosol content in the atmosphere. Further, the isotopic composition data contained in ice cores on Arctic islands have been used to statistically calibrate sea-ice extent against sea-ice observational data, yielding an 800-year record for the western Nordic Seas (Macias-Fauria et al. 2009) and a 700-year record for the Barents–Kara Seas (Zhang et al. 2018). Historical records, while not technically paleoenvironmental records, are valuable for reconstructing past sea-ice variability. Historical records describe meteorological and climatic features such as sea-ice conditions with high temporal resolution, typically annual or even subannual. In Europe and northern North Atlantic regions, such records can extend several centuries. The Iceland inhabitants documented sea ice observations almost from the time of their settlement there in the tenth century. The southern Greenland became a Norse-populated region after discovery of Greenland by Erik the Red in 982 and until the 1400s (Thostrup and Rasmussen 2009). Explorers, whalers, sealers and fishermen operated in the Atlantic approach to the Arctic Ocean for more than 400 years, and ice information is available for most years since 1730 (Vinje 1999). Historical observations from other regions of the Arctic (e.g. Bering Sea, Alaska region) are generally more fragmented and do not extend back more a century.

2.3 2.3.1

Arctic Sea Ice in the Geologic Past Pre-quaternary Sea Ice

Knowledge on sea ice in the Arctic Ocean in the distant geologic past (i.e., Pre-Quaternary) has until recently been highly speculative, and based on physical reasoning rather than paleoceanographic data, apart from a few marginal sites. The temperate regime in high northern latitudes during the early Cenozoic has supported a hypothesis of an open Arctic Ocean. The open-water Arctic was one of the factors contributing to worldwide climate uniformity. The appearance of sea ice in the Arctic is connected with formation of halocline, which acts as a screen on the way of warm flows from ocean depth to the surface, determines existence of ice cover in the Arctic, its geographical distribution and stability (Zakharov 1995). The significantly reduced vertical heat flux through the halocline, varying from 4 to 15 kilojoule cm2 year1 in the central part of the Arctic Ocean, could not compensate heat outflux from open water to atmosphere. Prevalence of precipitation and fresh water inflow from the mainland over evaporation from ocean surface determines formation

22

L. P. Bobylev and M. W. Miles

of the halocline. Further evolution of the sea ice cover is connected with a development of self-oscillations in the system ocean-atmosphere-glaciers. Cooling at about 34 mya culminated in an abrupt decrease in temperature and atmospheric CO2, and sustained climate conditions lingered through the Neogene at about 23–16 mya (Polyak et al. 2010). This cooling may led to the formation of Arctic sea ice, but period of its appearance was not known (Barry 1989). Based on the study of the available oldest age sediment of middle Pliocene, Clark (1971) developed the theory that the Arctic Ocean has been frozen since at least the middle Pliocene and that the most significant change in the Arctic ice has been in its thickness. Thicker ice restricted photosynthesis and foraminifera productivity, whereas times of thinner ice with conditions like those of the present are represented by abundance peaks of foraminifera and concentrations of glacially rafted material. Clark (1982) surmised that the Arctic Ocean became ice covered in the middle Cenozoic and together with the Antarctic ice sheet provided significant areas of net heat loss to propel the atmospheric circulation. The probable Middle Cenozoic development of an ice cover, accompanied by Antarctic ice development and a shift of the Gulf Stream to its present position, led to the development of modern climates. Some other authors believed that sea ice in the Arctic appeared later. According to Herman et al. (1989), ice rafting in the Arctic Ocean started at least as early as 4.5–5 mya, and the development of perennial ice cover, expressed by the earliest records of sediments available, was dated about 0.9 mya. Herman believes that from about 5 to 3.5–3 mya the Arctic Ocean was cold, but ice-free, with strong vertical mixing. Approximately 2.5 mya, the waters are thought to have become vertically stratified, possibly with seasonal ice cover (Barry 1989). Shelf ice and icebergs also appear to have been present after 4 mya. According to Zakharov (1995) sea ice in the Arctic appeared approximately 0.7 mya with a formation of halocline. The Cenozoic paleoceanography of the central Arctic Ocean was essentially unknown prior to drilling of Arctic Coring Expedition (ACEX) because of the extremely limited availability of pre-Pleistocene sediments (Backman et al. 2008), and therefore knowledge of the long-term history of the perennial ice was limited (Darby 2008). The ACEX deep-sea drilling borehole on the Lomonosov ridge revealed that about 50 mya the Arctic Ocean was considerably warmer than it is today, with summer temperatures estimated as high as 24
C (Moran et al. 2006). The discovery of sand and coarser detritus in the ACEX core back to the mid-Eocene (45 mya) suggests some ice rafting back to this time (Moran et al. 2006; Darby 2008). An abundance of sea-ice-dependent fossil diatoms Synedropsis spp., concurrent with the first occurrence of sand-sized debris at ~47 mya and a doubling of its flux at ~46 mya, indicate the possible onset of seasonal drifting sea ice and perhaps some glaciers in the Arctic during the cooling that followed the Eocene thermal optimum (Moran et al. 2006; Stickley et al. 2009). Sea ice apparently became a feature of the Arctic by 47 mya, and consistently covered at least part of the Arctic Ocean for no less than the last 13–14 million years (Polyak et al. 2010). Based on analysis of a 2-million-year sea-ice record, Stickley et al. (2009) revealed the transition from a warm, ice-free environment to one dominated by winter sea ice. The onset of IRD in the ACEX cores at 47.5 mya marks the onset of

2 Sea Ice in the Arctic Paleoenvironments

23

shore-fast ice and/or frazil ice, forming in shallow areas of the marginal Arctic shelf, which entrained and transported terrigenous material to the central Arctic. The low abundance of IRD and the absence of sea-ice diatoms in this phase indicate the episodic sea-ice formation. The appearance of well-preserved, lamina-forming sea-ice diatoms and the concurrent disappearance of cosmopolitan diatoms at 47 mya mark the onset of seasonal sea-ice formation in the central Arctic. Sea ice was the dominant mode of ice transport before 46.15 mya, when 80–100% of the grains exhibit sea-ice textures. After 46.15 mya, peaks in iceberg-IRD broadly correspond to peaks in total IRD abundance, indicating the presence of small isolated glaciers in Greenland (Stickley et al. 2009). Krylov et al. (2008) found a consistent change in mineral assemblages of the ACEX record at about 13 mya, indicating shift in their source area from the Barents– Kara–western Laptev Sea region to the eastern Laptev–East Siberian Seas. Considering that sea ice is the main transporting media for sediments and a drift time from mineral source area in the eastern Laptev–East Siberian to the ACEX drill sites exceeds 1 year, even assuming a faster than probable drift, the authors suggested that the sea ice have survived a summer melting. This indicates that the development of a multiyear ice cover in the Arctic Ocean began at about 13 mya. During the rest of the Neogene climate was considerably warmer than in the Pleistocene. However, many data sources indicate presence of seasonal sea ice, but not the extensive perennial ice (Polyak et al. 2010). Warm periods during the mid-to-late Pliocene have been documented in the Arctic from northwest Alaska to northeastern Greenland (Polyak et al. 2010). The Mid-Pliocene Warm Period (MPWP) occurred approximately 3.3–3.0 mya, when atmospheric CO2 was between 300 and 450 ppm and warm conditions lasted long enough to approach an equilibrium state. Although paleo evidence for the MPWP is limited, reconstructions suggest generally reduced summer sea-ice cover in the Arctic Ocean during the MPWP (Dowsett et al. 2016) and biomarkers from the Iceland Plateau indicate seasonal sea-ice cover with occasional ice-free intervals (Fischer et al. 2018). During this warm interval, the East Greenland Current may have transported sea ice into the Iceland Sea and/or brought cooler and fresher waters that could be favourable for local formation of sea ice (Clotten et al. 2018). Evidence from Alaska indicates that the nearby part of the Arctic Ocean was not frozen at just before 3 mya (Darby 2008), and, according (Carter et al. 1986), the southern limit of seasonal sea ice in Alaskan waters was at least 1600 km farther north than at present. The sea ice in the Arctic Ocean was much more limited than today, and perennial ice was either severely restricted or absent.

2.3.2

Quaternary Sea Ice

The Quaternary Period (beginning 2.58 mya) was characterized by generally low temperatures, except during interglacial stages. According to Carter et al. (1986) climates colder than the present one dominated since 2.14–2.48 mya, and maintained

24

L. P. Bobylev and M. W. Miles

perennial ice and precluded ice-free periods, except perhaps for brief intervals. Barry (1989) states that in the period between 2 and 0.7 mya the Arctic Ocean was probably ice-covered with intervals of more open, thin ice or even ice-free conditions alternating with conditions similar to the present. According to Herman et al. (1989), the Arctic Ocean was episodically or seasonally ice covered between 2.5 and 0.73 mya, and generally perennially ice-covered but with possible ice-free episodes since 0.73 mya. Clark (1971) also reports either very thin or episodically absent sea ice cover between 2 and 0.7 mya. Some other sources report the severity of ice conditions during glacial stages, indicated by the rarity of biological remains and non-deposition intervals due to especially solid ice. The Arctic environments were likely characterized by much thicker and more solid sea ice cover and also by invasions of vast Pleistocene ice sheets into the central part of the Arctic Ocean. The succeeding disintegrations of ice sheet were sometimes abrupt with incursions of giant icebergs armadas to the center of the Arctic Ocean (Polyak and Jakobsson 2011). According to (Polyak et al. 2010) a reduction in the amount of IRD in the ACEX record between ca 1.5–2 mya, which followed by slightly varying IRD abundance during both glacial and interglacial intervals until the Late Pleistocene may be an indicator of a rather stable ice cover. Sea ice was probably less prevalent during Quaternary interglacials and major interstadials, and the Arctic Ocean may even have been seasonally ice-free during some of these times. Melles et al. (2012) suggest that interglacials during Marine Isotope Stages (MIS) 11c, 31, 49, 55, 77, 87, 91, and 93 represent “super interglacials” in the Arctic throughout the Quaternary. However, marine cores from the Arctic basin still lack the comparable resolution and length to test for perennial versus seasonal sea-ice conditions during interglacial conditions over the Quaternary. An excellent Quaternary example of past warm climates with essentially modern geography is the Last Interglacial (LIG) between about 130 and 116 kya – also known as the Eemian and corresponding to part of MIS5 (Fig. 2.3). The climate and sea ice in the LIG can also be compared to the warmer peak interglacial MIS 11.3 (~410–400 kya) as noted above. Qualitative reconstructions of sea ice extent and concentrations suggest generally reduced extent. However, even during the LIG – with strong summer insolation – sea ice existed in the central Arctic Ocean during summer, whereas sea ice was significantly reduced along the Barents Sea continental margin and potentially other shelf seas (Stein et al. 2017). In the late Quaternary, especially mild ice conditions are inferred for the periods started about 130, 75, and 10 kya, respectively (Polyak et al. 2010). A much-reduced summer sea-ice (compared to present conditions) was indicated in the area north of Alert during the last interglacial and a younger warm interstadial from the discovery of abundant numbers of subpolar foraminifers in two cores (Mikkelsen et al. 2006). Ice-core evidence based on halogen chemistry for the LIG has been interpreted as suggesting that multi-year sea ice near Greenland was reduced, while the winter sea ice cover was not greatly changed (Spolaor et al. 2016). Some coastal exposures of interglacial deposits also indicate warmer than present water temperatures and thus, reduced sea ice. According to Brigham-Grette and Hopkins (1995), during the Eemian, the winter sea-ice limit in Bering Strait was at least 800 km farther north than today

2 Sea Ice in the Arctic Paleoenvironments

25

and during some summers the Arctic Ocean may have been ice-free. Based on analysis of sediment core from Lomonosov ridge, Hanslik et al. (2010) suggest more favourable conditions for primary production and a reduced sea ice cover at about 60 kya compared to Holocene. One important issue is that paleoceanographic histories need to be developed in the eastern and western parts of the Arctic Ocean separately due to considerably different sedimentary patterns related to the transpolar drift stream and the Beaufort Gyre (Polyak and Jakobsson 2011).

2.4

Arctic Sea Ice in the Holocene

In the central Arctic Ocean, no comprehensive spatial reconstructions of Holocene ice conditions have been generated because of low sedimentation rates and stratigraphic uncertainties (Polyak et al. 2010). According to Cronin et al. (2010), a reduction in perennial ice in the central Arctic Ocean occurred between 16 kya and until the end of the Holocene Thermal Maximum at 5 kya, and after that, the perennial sea ice cover increased again. Jakobsson et al. (2010) believes that the seasonal Arctic sea-ice cover was strongly reduced during most of the early-to-mid Holocene and there appear even to have been periods of ice-free summers in large parts of the Arctic Ocean. On the continental shelves, Holocene sedimentary records are relatively accessible and useful for reconstructing paleoceanographic environments and ice drift patterns on millennial to decadal scales due to relatively high sedimentation rates. There are indications of significant climate change during the early, middle and late Holocene and the past millennium (e.g., MCA and LIA – see chapter 2.5), based on variations in foraminiferal assemblages, δ18O values of foraminifera, coccoliths, diatoms, IP25 and IRD. Even with an abundance of relatively high-resolution data records available for the Holocene, no definitive reconstructions of Arctic sea ice have been produced even for the marginal shelves, partly due to challenges and inconsistencies in sea-ice proxy interpretation. There are however recent (e.g., Polyak et al. 2010) and ongoing (e.g., Seidenkrantz et al. 2016) synthesis efforts that involve more comprehensive compilations and consistent interpretation of multi-proxy records. An example of a multi-site reconstruction and comparison from Polyak et al. (2010) is illustrated in Fig. 2.6. Seidenkrantz et al. (2016) compiled more than 100 sea-ice proxy reconstructions from the Arctic Ocean and subarctic marginal seas to evaluate the variability of sea-ice cover during the Holocene. The reconstructions are primarily based on published data combined with some unpublished records of biological (diatoms, dinocysts, foraminifera, ostracods), sedimentological (IRD), and biogeochemical (IP25, PIP25) sea-ice indicators. Their initial evaluation indicates that winter sea ice was present throughout the entire Holocene, but the summer sea ice was probably somewhat reduced in some areas during the Holocene Climate Optimum (10–6 kya), although with some differences regionally. In the Nordic Seas and North Atlantic,

26

L. P. Bobylev and M. W. Miles

Fig. 2.6 Comparison of proxy reconstructions of Holocene sea-ice conditions in the Canadian and North Atlantic Arctic. (upper) Quartz content indicating the presence of Arctic drift ice on the north Iceland shelf (Moros et al. 2006). (middle) Duration of ice cover (months per year) in northern Baffin Bay based on dinocyst assemblages (Levac et al. 2001). (lower) IP25 fluxes indicating spring sea ice occurrence in the central Canadian Arctic Archipelago straits (Vare et al. 2009) (Source: Polyak et al. (2010))

minimum sea-ice conditions are seen 10–6 kya, whereas in the eastern Labrador Sea minimum sea ice occurred later, between 6 and 4 kya. Since about 4 kya, the sea-ice cover has increased, especially in the most recent millennia, i.e., the Neoglacial cooling onward. The Pacific sector of the Arctic (Beaufort, Chukchi, Bering and Okhotsk seas) show less variability during the Holocene, though it is noted that these records have poorer chronological control and resolution than those from the Atlantic sector (e.g., Seidenkrantz et al. 2016). These efforts toward a comprehensive Holocene Arctic sea-ice synthesis remain a work-in-progress; however for some Arctic regions, there are well-established results from previous studies, as described below.

2.4.1

North Atlantic Arctic

The eastern Nordic Seas and the adjacent Barents Sea show a well-developed earlyHolocene warming, which has been slowed in the more western areas. According to

2 Sea Ice in the Arctic Paleoenvironments

27

Spielhagen et al. (2011), the Holocene was characterized by a thermal maximum of Atlantic water at 10–9 kya and a cooling thereafter. The data show a minimum in ice cover at the end of deglaciation (10–11 kya), after which the area of ice increased and then reached another minimum around 6 kya (Polyak et al. 2010). The presence of quartz throughout the Holocene demonstrates that on multicentury timescales, drift ice has rarely been absent from this region, and over the last 5 kya it has been pervasive. Andrews et al. (2009) report that the North Atlantic circulation mode has probably been more meridional than zonal, implying a dominance of negative NAO-like conditions. Variations in the volumes of IRD in the subarctic North Atlantic indicate several cooling and warming intervals during late Holocene similar to the MCA and LIA from the last millennium. A major change, occurred ~1 kya when iron-oxide grains from a variety of sources around the Arctic Ocean reached Iceland, may be associated with an intensification of a positive AO-like climate mode (Andrews et al. 2009). Several studies have reconstructed Holocene variations of sea ice through Fram Strait and advected southward along the eastern Greenland toward the North Atlantic, as summarized in Briner et al. (2016). For example, based on a marine sediment core from eastern Fram Strait, Werner et al. (2013) reconstructed cooling from 5.2 kya to the present associated with increased polar water flux and southward and eastward advancing sea-ice margin. Sea-ice advection to Fram Strait was apparently synchronous with full postglacial flooding of the Arctic Ocean shelves ~5 kya. Farther downstream in the Denmark Strait, Jennings et al. (2002) related changes in benthic foraminiferal assemblages, isotopes and increased IRD fluxes to a shift toward cooler conditions at ~4.7 kya, while dinocyst data from the same area provide some information about sea-surface conditions, which were marked by very low salinity peaks until ~4.5 ka as the result of meltwater discharge, and also by increased seasonal sea-ice cover after 3.5 kya. Elsewhere on the southeast Greenland shelf, sea surface temperatures were reconstructed for the last 8000 years based on planktic foraminiferal assemblages (Jennings et al. 2011). These data also show more sea ice and/or cold conditions after 3.5 kya. These and other records from the region indicate increased sea-ice cover and overall colder conditions during this Neoglacial cooling.

2.4.2

Northern Greenland

Holocene variations in multiyear and fast sea ice from northern Greenland were analyzed using the abundance and origin of driftwood as signals of multiyear sea ice and its traveling route, and the occurrence or absence of beach ridges along the coast as signals of seasonally open water. Funder et al. (2011) found that deglaciation of the coastal plain at ~10 kya was followed by marine transgression, and until ~8.5 kya there was only sporadic driftwood, and beach ridges occurred only at the southernmost locality. Multiyear sea ice reached a minimum between ~8.5 and 6 kya, when its limit was located ~1000 km to the north of its present position at the coast of

28

L. P. Bobylev and M. W. Miles

Greenland. An Arctic Ocean sea-ice cover in summer was less than half of the record low 2007 level at 8 kya. The general build-up of sea ice began from ~6 kya, and its Holocene maximum during the LIA attained its present extent at ~ 4 kya (Funder et al. 2011). The increase in multiyear sea ice, culminated during the past 2500 years, is linked to an increase in ice export from the western Arctic and higher variability of ice-drift routes. Highest numbers of driftwood from Greenland coasts north of 80
N in the period between ca 5.5 and 3 kya, probably marks an increase in multiyear ice, and a succeeding minimum between ca 3 and 1 kya may indicate a further barring of the coast by fast ice (Polyak et al. 2010). When the ice was at its minimum in northern Greenland, it greatly increased at Ellesmere Island to the west. The abundance of driftwood on the northeast coast of Greenland shows that the shoreline was ice free from about 6 to 4 kya (Dyke et al. 1997).

2.4.3

Canadian Arctic

Reduced sea ice during the early Holocene, an accelerating increase in ice occurrence between 6 and 3 ka, and high but variable occurrence since then was identified in an IP25 record from the central Canadian Arctic Archipelago (Polyak et al. 2010). However, samples from ice cores to the north of Alert, representing the Holocene, lack the subpolar foraminifer species, and thus indicate a consistent thick perennial sea ice in accordance with present-day conditions (Mikkelsen et al. 2006). Reconstructions of sea-ice conditions in the Canadian Arctic Archipelago have been derived also from the spatial and temporal distribution of marine mammal bones in raised marine deposits. The annual migration of the bowhead is controlled primarily by changing sea-ice cover, because of the bowhead’s strong preference for ice-edge habitat (Dyke and England 2003). Bowhead bones most commonly found in this region in early Holocene indicate at least periodically ice-free summers. Fewer bones indicate higher ice coverage of these waters in the period from ca 9 to 5–6 kya despite a relatively warm climate. Paleogeographic maps based on the distribution of bowhead whale bones show heavy sea ice in Jones Sound at 6 kya, followed by open water conditions at 5, 4, and 3 kya and a return to more sea ice during the past 2000 years (Dyke et al. 1996). During the most recent warming event, about ~2700 years ago, ice cover in the Jones Sound region persisted only 2 months per year. Analysis of past sea ice conditions in the Canadian Arctic is based substantially on driftwood occurrence. During postglacial time the different patterns of driftwood penetration to the Canadian Arctic Archipelago and the mix of species involved are due to changes in the trajectory of the transpolar drift, which carries North American wood along its Polar branch and Russian wood along its Siberian branch (Dyke et al. 1997). The delivery and deposition of driftwood on northern Ellesmere Island is affected by the influx of driftwood into the Arctic Ocean from northward flowing

2 Sea Ice in the Arctic Paleoenvironments

29

rivers, the availability of sea ice to raft wood across the Arctic Ocean during its voyage, the presence of favourable ocean currents, and the lack of ice shelves and multiyear fast sea ice along the northern Ellesmere Island coast permitting driftwood deposition (England et al. 2008). Using radiocarbon ages obtained on samples of Holocene driftwood transported across the Arctic Ocean and deposited on northern Ellesmere Island, England et al. (2008) clarified the age of the Markham, Ward Hunt and Petersen ice shelves, composed of multiyear fast ice. The records indicate intervals when Arctic Ocean surface currents and seasonally ice-free fiords allowed driftwood to enter. The absence of driftwood during some periods may be due to shifts in trajectories of drifting ice from this area and/or presence of multiyear fast ice along the coast (England et al. 2008). Samples, collected inland of the Ward Hunt Ice Shelf, are in age from 9.2 to 5.5 kya, after which the entry of driftwood terminated to present. This hiatus in driftwood deposition marks the inception of multiyear fast ice across northern Ellesmere Island. However, driftwood deposition continued until 3.5 kya in Clements Markham Inlet and to present along northeast Ellesmere Island. By 1960, the combined area of ice shelves and ice islands was approximately 4600 km2, roughly half the estimated area of the “Ellesmere Ice Shelf” at the end of the LIA (England et al. 2008). Along the margins of Baffin Bay only sparse wood arrived until 6.75 kya, when abundance increased and was sustained until 4 kya, with a distinct peak in the period from 5.25 to 5 kya. Little wood arrived during the last 4000 years, except for the renewed abundance of the last 250 years (Dyke et al. 1997). Most wood in the region of the Western Channels arrived from 9 to 4.75 kya and from 3.5 kya to present, whereas in the region of the Central Channels driftwood arrived sparsely between 8.6 and 6 kya, and then suddenly increased to maximum between 6 and 5.75 kya. The maximum influx of driftwood in the central Canadian Arctic Archipelago occurred in the second half of the Holocene indicating a favourable circulation and considerable mobility of sea ice in the channels and shorelines at least periodically free of fast ice. In total, the parts of the Canadian Arctic Archipelago not accessible to driftwood from Baffin Bay display the following broad features: (1) initial incursion of wood just after 9 kya and sparse wood incursion until 6 kya; (2) a sharp peak of wood incursion from 6 to 5.75 kya, succeeded by a minimum between 5.25 and 5 kya; (3) moderately abundant wood incursion for the rest of the middle Holocene; and (4) maximum postglacial incursion during the last 2.5–3 kya (Dyke et al. 1997). In the northeastern Canadian Arctic Archipelago, a major mid-Holocene warm interval from ~6.5 to 4.5 kya, and the periodic occurrence of warmer than present intervals from 4.85 until at least 2.5 kya was identified from marine cores. This period is characterized with large oscillations in summer SST from 2 to 4
C cooler than present, to 6
C warmer. These temperature changes correspond to annual variations in sea ice cover ranging from 2 months more of heavy (>50%) ice to a 4–month extension of open water conditions compared to now. The warming took ~50 to 100 years and lasted ~300 years before replacement by colder intervals lasting ~200 to 500 years (Mudie et al. 2005).

30

2.5

L. P. Bobylev and M. W. Miles

Arctic Sea Ice in the Past Millennium

The last millennium of the Holocene is of particular interest because it covers two major climate transitions – from the MCA to the LIA and the termination of the LIA – as well as twentieth century changes that include the Early Twentieth Century Warming (ETCW) and the anthropogenic warming era. Here we cover the natural variability and changes in sea ice that may arise from external and internal variability on suborbital (non-Milankovitch) time scales, e.g., solar fluctuations, volcanic eruptions, and atmosphere–ocean circulation variability and interactions. Variations in sea ice distribution in different Arctic regions during the last millennium have been identified with greater spatial and temporal detail than those from earlier in the Holocene and farther back in geological time, for three reasons. First, marine core data records for the past millennium or so are: (1) relatively abundant and well distributed, (2) of generally higher temporal resolution (century to even sub-decadal scale) and (3) of better chronological age control than longer records. Second, composite baseline reconstructions of regional sea-ice extent have been developed from spatial networks of paleoenvironmental data (e.g., MaciasFauria et al. 2009; Kinnard et al. 2011). Third, detailed evidence from historical observations of sea ice is available for many regions and locations during the past several centuries. Regionally, there have been historical (and paleoenvironmental) reconstructions in the Atlantic sector than elsewhere in the Arctic. Sections 2.5.1 – 2.5.6 describe in some detail reconstructed sea-ice variability during the past several centuries for different regions, starting with the Russian Arctic seas and extending across the Atlantic and into the Canadian Arctic.

2.5.1

Russian Arctic Seas

The local indigenous peoples of the north, the Norse and the Slavs likely used the Arctic waters for navigation in the southwestern Kara Sea in the first millennium A. D., and even earlier in the southeastern Barents Sea (Armstrong 1984). A team of the Novgorodian voevode Uleb made the first known sea voyage eastward of the White Sea in 1032 and obviously reached Kara Gate and Yugorsky Shar (Belov 1977). The Novgorodian merchants seem to have crossed the Yamal Peninsula and exited to the lower reaches of the Ob’ River in the twelfth century. Throughout the thirteenth to fifteenth centuries, the pomors sailed in the Kara Sea passing to it through the southern Novozemelsky Straits and Matochkin Shar Strait. However, information about sea-ice conditions in the Arctic from these voyages is practically absent. More intensive exploration of the Northern Sea Route began from the second half of the sixteenth century. The ships under the command of Hugh Willowby and Stefan Burrow reached the eastern Barents Sea, a Dutch merchant Oliver Brunell together with Russian seafarers made a sea voyage from the Pechora mouth to the lower Ob’,

2 Sea Ice in the Arctic Paleoenvironments

31

while the expedition commanded by Arthur Pet and Charles Jackman sailed across Yugorsky Shar Strait to the Kara Sea (Vize 1939). In the end of the sixteenth century, the expedition headed by Willem Barents was able to round the northern tip of Novaya Zemlya where their ship was beset in ice and damaged. The voyages of pomors and manufacturers were made to Novaya Zemlya and Grumant and also to the Ob’ and Taz Rivers to Mangazeya, where between 1610 and 1619 not less than 16–17 ships called annually. Using historical data, Nazarov (1947) reconstructed ice conditions in the Kara Sea starting from 1550, however with significant spatial and temporal gaps. He described the ice cover using the following conditional units: much decreased ice cover (2), decreased ice cover (1), expanded ice cover (+1) and very expanded ice cover (+2). According to Barnett (1991) sea ice conditions in the sixteenth century were quite severe, and navigators usually halted near the Kara Gate at the southwest entrance to the Kara Sea. However, availability of sea routes to the Kara Sea depends mostly on ice distribution, and sailors penetrated into free of ice parts of the Kara Sea, when only one of four possible routes (through Yugorskiy Strait, Kara Gate Strait, Matochkin Shar Strait, or around Cape Zhelaniya) was available (Itin 1933). Normally, the entire Kara Sea south of 75
N is ice-free, when the seasonal ice minimum is reached by mid-September, and in extremely mild summers, part of the sea may become ice free as far north as 80
to 82
N. In unusually cool summers southern current moves sea ice along the eastern coast of Novaya Zemlya to the Kara Gate, where it does not melt completely. In those summers it also may be present in the central part of the southwest Kara Sea. Sea ice conditions in the Kara Sea in the seventeenth century were assessed for 12 years; nine of them had decreased and very decreased ice cover (Nazarov 1947). Borisenkov (1995) believes that the unprecedented navigation in the Arctic seas during the first half of the seventeenth century was possible due to favourable climate, and from the mid-seventeenth century on, universal climate cooling set in for almost 200 years, until the middle of the nineteenth century, resulting in heavier ice cover in the Arctic seas. Lestgaft (1913) refers to availability of the Kara Sea for navigation during sixteenth and beginning of seventeenth centuries, when Russian merchants constantly sailed there. The Kara Sea was completely inappropriate for sailings to Siberia in 18th and first half of nineteenth centuries, since sea ice in the Barents and Kara Seas significantly expanded in 1820s and 1850s (Itin 1933). Zakharov (2004) states that in the eastern Barents Sea, typical August ice conditions in eighteenth and nineteenth centuries correspond approximately to contemporary sea ice conditions in late June. Information on sea-ice conditions is available for almost each year from the end of 1860s, when transport, fishing and expedition vessels regularly sailed in the southwestern part of the Kara Sea (Nazarov 1947). Sea-ice conditions in the Kara Sea for the last three decades of the nineteenth century were characterized by large interannual variability and maximum severity in the early 1880s, in agreement with other records from the Barents and Nordic Seas. The Barents and Kara Seas had relatively little ice cover in most years of the 1870s, except for 1872–1873. Several very severe years were observed in the first

32

L. P. Bobylev and M. W. Miles

half of 1880s, when sea ice persisted near the Murmansk coast and in the Pechora mouth during the whole summer (Sibirtsev and Itin 1936). Decreased and much decreased ice cover is reported in the Kara Sea during the 1890s. However, in the period 1869–1911, sailing to the Kara Sea through Novozemelskie Straits was complicated or impossible every four of 10 years due to ice occurrence (Lestgaft 1913). Therefore sailing in this time was more difficult than in the period of unfavourable sea ice conditions in 1956–1965, when in September sea ice was not observed in Yugorskiy Shar Strait, once was observed in the Kara Gate and once in Matochkin Shar Straits (Zakharov 2004). Sailing around Cape Zhelaniya was possible in six of 10 years in the period 1895–1910 (Lestgaft 1913), whereas at present the probability of ice occurrence here amounts only to 10%. Based on analysis of temporal changes in the Kara Sea ice cover, Nazarov (1947) revealed a 100-year cycle in occurrence of series of years with increased ice cover: 1620–1629, 1720–1729, 1820–1829, and 1920–1929. However, (Sibirtsev and Itin 1936) argue that existence of century-long variations in ice cover cannot be scientifically proved taking into consideration only historical but not geological time intervals. In the central part of the Northern Sea Route ice conditions were considerably more severe during the Great Northern Expedition in 1733–1743 than in the seventeenth century and at present. Several ships of this expedition were repeatedly beset in ice, sailed to the destination point for several years, and could not pass along the coast of the Taymyr Peninsula, in spite of enormous efforts (Johannessen et al. 2006). The conquering of the eastern part of the Northern Sea Route began in 1633, and the Cossacks discovered the mouths of all major East Siberian rivers within 15 years. This eastward advance of Russian seafarers along the Eurasian Arctic coast ended with the saga of Semen Dezhnev, who discovered the strait between Asia and America in 1648. Before and after that sail ships could not navigate there. However, this sailing could not unambiguously indicate easy ice conditions, because eastward navigation of sail ships mostly depends on prevailing winds. Narrow area along the Chukchi coast eastward of Kolyma is free of ice, when southeastern winds prevail, and covered with ice when winds are from north-west. Indigeneous people report that onshore motion of sea ice in the area of Cape Shelagskiy changes to a period of offshore drift during 4–5 years. The average period of this phenomenon is almost the same as in Iceland (Sibirtsev and Itin 1936). Ice conditions change differently in different parts of the Eurasian Arctic. The so-called phenomenon of ice opposition, when difficult ice conditions in the west correspond to favourable ice conditions in the east, and vice versa, was indicated from historical data. For several years in the 1820s, 1850s, and 1880s expanded ice cover in the western region corresponded to decreased ice cover in the east. Thus, during the periods 1881–1886 and 1912–1917, increased ice cover in the Barents and Kara Seas corresponded to its decreased in the eastern seas (Sibirtsev and Itin 1936). Favourable ice conditions prevail in the central part of the periods, when favourable ice conditions in the east and severe ice conditions in the Kara Sea were observed.

2 Sea Ice in the Arctic Paleoenvironments

33

Fig. 2.7 Reconstructed sea-ice extent (SIE) in the Barents–Kara (B-K) sector of the Arctic (Source: Zhang et al. (2018))

Zhang et al. (2018) recently used high-resolution ice core and tree ring proxies for sea ice extent to reconstruct a time series of autumn ice extent over the Barents-Kara (B-K) sector of the Arctic from 1289 to 1993 (Fig. 2.7). Their results show that from the late thirteenth century to the end of eighteenth century, the ice was generally extensive, with large variability and a slightly increasing trend, reflecting the LIA background. The B-K sea-ice extent (SIE) began to decrease at the end of the eighteenth century, and a negative trend became significant during the second half of the nineteenth century and lasting into the 1930s–1940s. The 1930s–1940s was a period with a relatively low SIE, then the SIE had a short period of expansion from the 1940s–1970s. However, the B-K SIE has continuously and significantly diminished since the 1970s, and in recent years may have become the lowest it has been in a millennial perspective.

2.5.2

Nordic Seas

The Nordic Seas (Greenland, Iceland, Norwegian and the Barents Seas) (Fig. 2.8) are of particular importance to global climate. Here warm, saline water from the Atlantic flows into the Arctic Basin, and the fresher Arctic Ocean water is exported to the Atlantic Ocean in the East Greenland Current (EGC). Changes in the ocean– ice–atmosphere interactions in this region are of critical importance to the ventilation of the deep oceans and to the global thermohaline circulation (Vinje 2001; Divine and Dick 2006). Variations in drift-ice trajectories, caused by the atmospheric forcing, are associated with the North Atlantic (NAO) or Arctic (AO) oscillations. During a positive (warm) AO phase, the atmospheric circulation is zonal and although sea ice is

34

L. P. Bobylev and M. W. Miles

Fig. 2.8 The Nordic Seas including the Barents Sea and adjacent regions

exported through Fram Strait it tends to be re-circulated into the Greenland Sea. In the negative (cold) phase the circulation is more meridional with a blocking ridge over Greenland and a trough boundary along East Greenland. This configuration is ideal for the export of sea ice as far south as Iceland and the Faeroe Islands. Periods of extensive drift ice around Iceland result from the export of sea ice from the Arctic Ocean and to a lesser degree from icebergs from Greenland Ice Sheet outlet glaciers. During the Great Salinity Anomaly (GSA), the AO and the NAO achieved record negative values and ice invaded Iceland waters (Andrews et al. 2009). The north Icelandic shelf, situated between opposing atmospheric/oceanic fronts, is particularly sensitive to changes in North Atlantic climate regimes (Patterson et al. 2010).

2.5.2.1

Iceland

The first historical record was compiled by Koch (1945), which gives the incidence of ice off the coast of Iceland and extends back to ~865. It revealed less severe conditions during 1640–1670, just prior to a peak of LIA severity in Europe, and severe ice conditions during the 1690s and 1740s (Barry 1989). Later, climate historian H. Lamb further developed Koch’s record and produced some additions

2 Sea Ice in the Arctic Paleoenvironments

35

to it for the period from 1600 to 1975. Changes of sea ice conditions off the coasts of Iceland were further studied by Ogilvie (1984, 1986, 1992, 1995, 2005), Ogilvie and Jonsdottir (2000), and Ogilvie and Jonsson (2001). The occurrences of mild and cold seasons and sea ice in Iceland during the period ca. 865–1598 have been studied by Ogilvie (1995) based on analysis of historical documents. It is often assumed that warm temperatures and a lack of sea ice represented conditions favourable to sea voyage in the periods of Icelandic and Greenland settlement, and the Icelandic sagas generally indicated a favourable climate in that time. In the period following the settlement of Iceland by ~960, maximum temperatures remained at ~7.5 to 9.5
C, but minimum temperatures were on average almost 6.0
C lower than during the initial wave of settlement. Maximum summer temperatures then also decreased and reached only 6.0
C and 5.0
C by ~990 and ~1080, respectively. The 18O values from molluscs on the north Iceland shelf show a warming trend after ~1120, and by ~1250 the summer temperature maximum reached 10.0
C – the values observed during the period of first written record of Iceland 300 years earlier (Patterson et al. 2010). According to historical descriptions severe seasons occurred from the early 1180s to the first decade of the thirteenth century, whereas between 1212 and 1232 the climate probably improved. Severe seasons occurred sporadically over the next few decades, and there can be little doubt that the climate of the 1280s and 1290s was severe (Ogilvie 1995) and sea ice was commonly present (Masse et al. 2008). The general hypothesis on the warm climate in this period is supported by several data sources. IP25 abundances for 800–1300 are lower than for the subsequent 700 years. Diatom-based sea-surface temperature reconstructions and Northern Hemisphere temperature profiles also show warmer temperatures during this time (Sicre et al. 2008; Masse et al. 2008). Borehole temperatures from central Greenland ice cores show a period warmer than or as warm as today between 900 and 1200 (Ogilvie and Palsson 2003). The foraminiferal evidence and the sediment record suggest that relatively mild Atlantic Intermediate Water was the predominant water mass in the period 730–1100, and possibly there was less Arctic ice. Ogilvie (1984) describes climate in the fourteenth century as quite variable. A mild period is suggested in the early 1300s, with only one reference to sea ice in 1306, and harsh sea ice and severe weather in the beginning of the 1320s, but only two winters are described as severe over the next two decades. The climate appears to have become colder from the 1350s and the years of severe weather in the 1360s and 1370s are comparable with the last two decades of the thirteenth century. According to Patterson et al. (2010) cooling trend continued until ~1380, when the lowest summer and winter temperatures since settlement were recorded by the mollusc time-series record. Ogilvie (1984) believes that the period from 1380 to 1430 was probably comparatively mild, whereas (Patterson et al. 2010) report records of severe weather and sea ice in the late 1300s and early 1400s. There is some evidence for a mild climate between 1430 and 1560, in spite of a paucity of historical climate records for much of this period, particularly for the period ~1500 to 1561 (Ogilvie 1984). Relatively mild temperatures in the period from about 1412 to 1470 are confirmed by low IP25 abundances between 1400 and

36

L. P. Bobylev and M. W. Miles

Fig. 2.9 Decadal winter/spring thermal index for Iceland. The index is a number of unusually mild seasons minus the number of unusually cold seasons, with each number weighted by 1–4 depending on how many regions reported mild or cold conditions (Source: Ogilvie (1981))

1460 (Ogilvie and Jónsson 2001; Masse et al. 2008). Evidence for substantial changes in climate during the 1430–1560 period, including a 40 to 50-year period of more extensive sea ice cover, is indicated by rapid changes in the abundance of IP25 during the mid-late fifteenth century and by substantial oscillations observed in the diatom-based temperature record (Sicre et al. 2008). A comparatively harsh climate is suggested for the latter part of the sixteenth century (Ogilvie 1984; Ogilvie and Jonsdottir 2000). The 1560s were mainly cold, and sea ice was described several times, whereas the 1570s were mild with sea ice observed only once. During the 1580s climate was relatively severe with much sea ice, but the ice was not observed every year (Ogilvie 1995). A variety of data sources including travel accounts, early newspapers, diaries, the later annals written by individuals, and the governmental reports in all of Iceland’s districts, available since 1600, allows producing decadal and annual indices (Ogilvie 1999). Decadal winter/spring thermal index for Iceland, presented in Fig. 2.9, shows the general harshness of the climate and its variability from one decade to another. The coldest decades were the 1630s, the 1690s, the 1740s and the 1750s. The yearto-year variability was also considerable, especially in the 1710s and the 1730s. Severe weather in Iceland does not necessarily indicate the presence of sea ice, however, the correlation coefficient between the decadal ice severity index and the decadal winter/spring thermal index is 0.57 (Ogilvie 1984). The ice index represents the number of seasons (winter, spring and summer) with sea ice present off the coast of Iceland. Sea ice rarely appears in autumn, and it was not included in the index. Each number is weighted by the number of regions (i.e., north, south, east, west), where ice was reported (Ogilvie 1984). The Iceland ice index provides useful information on the ice extent variations, but with some limitations. A zero index indicates that the ice edge may be located at any distance

2 Sea Ice in the Arctic Paleoenvironments

37

Fig. 2.10 A sea-ice index for Iceland showing variations in the incidence of sea ice off the coasts during the period 1600–1850. The data used are from a variety of historical sources. After ca. 1700, the main data source is sheriffs letters (Source: Ogilvie and Jonsdottir (2000))

to the northwest from it, and the index maxima do not indicate how far beyond Iceland the ice edge has advanced (Lassen and Thejll 2005). Figure 2.10 shows the variations of sea-ice index for Iceland from 1600 to 1850, reconstructed from a variety of reliable historical sources (Ogilvie and Jonsdottir 2000). The Icelandic sources suggest a cooling trend around the end of the sixteenth century and the beginning of the seventeenth century. The first decade of the seventeenth century does not appear to have been extremely cold, but it experienced more sea ice than any other in the period 1601 to 1690, and sea ice is recorded in the north in 1602, 1604, 1605, 1608, 1610. The 1610s appear as mild but with three severe sea-ice years. From 1618 to 1624 mild winters occurred one after the other. The 1620s were rather cold, and six severe sea-ice winters were recorded in the north (Ogilvie 1995). The 1630s appear to have been especially cold, and particularly severe years all over Iceland were 1633 and 1639. Sea ice was occurred in 1633, 1636, 1638, 1639, and in winter 1638–1639 it reached the eastern and southern coasts. The mild period from ~1641 to 1670 coincided with the period of least ice in the seventeenth century (from ~1631 to ~1681). During the 1640s sea ice was recorded only twice, during the 1650s was not reported at all, and little ice was observed in the 1660s and 1670s (Ogilvie 1995; Ogilvie and Jonsson 2001). Although Iceland’s shores remained relatively ice-free, the 1670s and 1680s saw a return to a colder regime. The sources recount four occurrences of sea ice during the

38

L. P. Bobylev and M. W. Miles

1680s. The 1690s are considered to be the coldest decade for the seventeenth century in the northern hemisphere (Ogilvie and Jónsson 2001), and also a decade with the most sea ice in the seventeenth century (Ogilvie 1995). Lamb (1979) used reports of sea ice and fisheries records to conclude that sea ice and Arctic waters nearly reached the Faroe Islands in the last two decades of the seventeenth century. The period from 1701 to 1730 was the mildest of the century with particularly mild decade 1701 to 1710, but sea ice occurred in 1701, 1703, 1706, and 1708. The climate in 1710s was colder than in the previous decade, but sea ice showed a marked decrease. Four severe-ice years were recorded in the 1720s. The 1730s were cold, and winter of 1737 was the worst since the winter of 1633. The 1740s and 1750s were very cold, similar to the 1630s and 1690s. Sea ice was recorded for 7 years in the 1740s and for 5 years in the 1750s (Ogilvie 1995). The period of very cold climate, 1731 to 1760, was followed by the milder 1760s, but sea ice distribution was similar to the 1750s, and ice was reported in several years. The 1770s were colder than the previous decade, and incidence of sea ice, reported in every year except of 1779 and 1780, was as great as in the 1690s and 1740s. The 1780s were the coldest decade in the entire series from 1601 to 1800, and possibly even from 1501, with the harsh conditions compounded by volcanic activity. This decade is also characterized with the greatest extent and duration of sea ice, which was present every year. In the years 1782–1785 the ice was extensive off the northern, northwestern, and eastern coasts, and present from the winter through to summer (Ogilvie 1995). The 1790s appear to have been as cold as the 1780s, but sea-ice incidence over this decade was quite extensive. The years 1780–1800 were the iciest period around Iceland possibly even from 1500. The climate of the nineteenth century was highly variable with a succession of cold and mild decades. The 1810s, 1830s, and 1880s were comparatively cold, whereas there was no ice off the Icelandic coasts from 1840 to 1855 (Ogilvie and Jonsdottir 2000). After that and to the end of the century there was frequent ice again, although its incidence does not seem to have been as heavy as in the earlier part of the century. Further clusters of sea-ice years occurred again from ca. 1864 to 1872. Several very severe sea-ice years were observed in the 1880s, and some sea-ice years occurred in the 1890s. In agreement with the historical data, describing the latter part of the nineteenth century as the coldest, IP25 abundances are similarly variable during this period with the highest abundances observed in the more recent sediments (Masse et al. 2008). In general, the greater part of the evidence supports the notion that the climate of the north Atlantic region from ca. 1250 to ca. 1900 was at least slightly colder than the twentieth century overall (Ogilvie 1995; Ogilvie and Jonsdottir 2000). The coldest centuries of the last millennium in the northern hemisphere were the nineteenth and, particularly, seventeenth centuries. The seventeenth century was on the average only 0.5–0.8
C below that for the 1961–1990. During the early and late decades of the seventeenth century, much sea ice was present, but from ca. 1640 to ca. 1680, there appears to have been little sea ice off Iceland (Ogilvie and Jonsson 2001). The sea ice visited the coasts of Iceland more frequently in the eighteenth than in the seventeenth century. During the period 1600–1850, the decades with most ice

2 Sea Ice in the Arctic Paleoenvironments

39

present were the 1780s, the 1810s, and the 1830s. The 1840s saw a return to a regime with very little ice. From about 1850 to around 1900, the incidence of sea ice off Iceland was apparently relatively heavy. In the twentieth century sea ice off Iceland has occurred less frequently as compared with the entire period 1600 to 1900 (Ogilvie 1999; Ogilvie and Jonsson 2001).

2.5.2.2

Greenland and Barents Seas

Several studies deal with changes of sea ice conditions in the whole area of the Nordic Seas, which in addition to Iceland includes the Greenland, Norwegian and Barents Seas. Macias-Fauria et al. (2009) reconstructed an eight-century-long history of sea-ice conditions in this region based on the combined tree-ring and ice-core data from northern Scandinavia and Svalbard, calibrated against historical observations from Vinje (2001). The Macias-Fauria et al. (2009) reconstruction exhibits large variability on decadal to multidecadal time scales, as well as revealing the unprecedented low sea-ice extent in the twentieth century. Analysis of climatic changes in ice edge position in the Nordic Seas was made based on observations from April through August, spanning the period 1750–2002 (Vinje 1999, 2001; Vinje and Goosse 2003; Lassen and Thejll 2005; Divine and Dick 2006). Vinje (2001) found a highly variable, but substantial reduction of ~33% in the April ice extent in the Nordic Seas over the period 1864–1998 with the average period of variations of 12–14 years. Nearly half of this reduction took place before 1900. The April ice extent reduction over the period 1864–1998 was been greater by 46% in the area to the west of 10
E as compared with 24% in the eastern area. This reduction is ascribed to a dominant oceanic effect, corresponding in net rise in the upper layers of the Atlantic water of about 0.5
C. A strong correlation was observed between the NAO-scaled winter circulation and the subsequent April ice extent: negative in the Nordic Seas, and positive for Newfoundland–Labrador Sea. The maximum correlation of 0.62 during cooling periods illustrates the correspondence between increased ice extents in the Iceland Sea (Ogilvie 1992) and in the Barents Sea (Vinje 1999) during the extreme expansion around 1800 (Vinje 2001). Divine and Dick (2006) found oscillations in ice cover with periods of about 60–80 years and 20–30 years, superimposed on a continuous negative trend. The lower frequency oscillations are more prominent in the Greenland Sea, while higher frequency oscillations are dominant in the Barents Sea. Barents Sea ice-edge position in August in the second half of the eighteenth century exhibits values close to the recent ones. The August ice edge in the sector between 20
and 45
E, covering the western Barents Sea between Svalbard and Franz Josef Land and the adjacent part of the Arctic Ocean was analyzed by Vinje (1999), who found large variations in the ice edge location from year to year, as well as on centennial and decadal scales (Fig. 2.11). The extreme northern ice edge positions in this sector coincide with increased meridional circulation and an increased influx of water from the Norwegian Sea entering the Arctic Ocean north of Svalbard. After a marked change in the

40

L. P. Bobylev and M. W. Miles

Fig. 2.11 Time series of the August ice edge in the sector between 20
and 45
E, covering the western Barents Sea between Svalbard and Franz Josef Land and the adjacent part of the Arctic Ocean. The gray bars indicate open water. Extensions of white area down to 74
N indicate lack of observations. The upper line connects the 10-year mean ice-edge positions, except for the period 1680–1740 where (due to sparse data coverage) a 60-year mean was used. The lower line shows the deviation (right scale) from the Northern Hemisphere mean temperature 1961–1990 (Source: Vinje (1999))

annual melt-back around 1920, the long-term average annual melt-back rate tripled as compared with the period 1800–1920. This marked change coincides with increased winter temperatures at Svalbard during the Early Twentieth Century Warming (ETCW) in the Arctic, which was strong enough to impact the hemispheric temperature. A very high negative correlation (0.87) was found between the 10-year mean Barents Sea August ice extent and the Northern Hemisphere mean temperature (NHMT) back to ~1860 (Vinje 1999). Vinje also studied temporal variations of the NHMT using the time series of the 10-year mean ice edge location as a proxy and by making a backward extrapolation. He found considerable changes in the NHMT in the past centuries; a fall of the order of 0.6
C from about 1580 to 1650, a rise of the order of 0.6
C from about 1650 to 1760, and a marked fall of a similar order over three decades, from about 1760 to 1790. After a net rise of 0.2
C over the period 1790–1910, temperature increases of a similar rapidity as referred to above occurred from about 1910 to 1940 and again from about 1970 to 1990. In general, the August ice edge series indicates that the NHMT has experienced an increase of about 0.7
C from the minima around 1650 or 1790 until today (Vinje 1999). Changes of April sea ice extent for the period 1850–2001 in the Barents Sea and the Arctic Ocean region, extending from Svalbard to Novaya Zemlya and southward to the Kola Peninsula and covering the main pathway of the warmer water flow from the Norwegian Sea, were studied in (Shapiro et al. 2003). They found that the mean ice edge position retreated north-eastward over the 152-year period, with the greater retreat seen in the changes from the 1850–1899 sub-period to the 1900–1949 sub-period, and then from the 1900–1949 sub-period compared to

2 Sea Ice in the Arctic Paleoenvironments

41

Fig. 2.12 Ice extent observations (103 km2) in the Nordic Seas regionally, mostly based on historical observations in ship logbooks and Icelandic coastal observations

the 1950–2001 sub-period. Vinje and Goosse (2003) found, that the ice extent variations in the Iceland, Greenland and Barents Seas are quite similar. A minimum extent was around 1750 with slightly more ice than during around 2000, and a maximum in between, over the period 1800 to 1850 (Fig. 2.11). Variations of ice extent with a secondary minimum around 1720–1750, a maximum between 1800 and 1850, and the subsequent reduction correspond negatively to variations in the solar irradiance, with the proxy temperature series and the solar cycle length closely related to radiation. Lassen and Thejll (2005) suggested that a large part of the temperature and sea ice variability over the recent 150 years may be caused by variations in solar radiation. The ice extent reduction in the Nordic Seas since 1800 amounts to 50%, with a maximum (60%) and a minimum (23%) reduction in the Barents Sea and the Iceland Sea, respectively. Time series of the April – August ice extent in the Greenland Sea, published by Lassen and Thejll (2005) and in the Greenland and in the Barents Seas by Vinje (April), show a decrease of more than 30% since the last decades of the nineteenth century (Fig. 2.12).

42

L. P. Bobylev and M. W. Miles

Fig. 2.13 The “Storis” summer extent index for each year during the period 1820–2001 (Source: Schmith and Hansen (2003))

2.5.3

Greenland

Multi-year ice exported through Fram Strait is carried southward by the East Greenland Current (EGC) along the entire length of the east Greenland shelf and around Cape Farewell and further along the southern part of west Greenland, where it is termed “Storis” (Buch 2000). Historical records of the “Storis” along the west coast of Greenland exist from the 1820s onward as monthly descriptive summaries (Speerschneider 1931). Figure 2.13 shows “Storis” summer extent index during the period 1820–2001, calculated by Schmith and Hansen (2003). They standardized the time series of “Storis” extent for each of the months May, June, and July by subtracting the long-term average and dividing by the long-term standard deviation and the average of these three standardized time series was taken. The “Storis” summer extent index defined in this way has long-term average and standard deviation equal to 0 and 1 respectively. Hill (2008) compared ice indexes for Iceland, Newfoundland as well as the “Storis” ice index for the period from 1810. Qualitative analysis shows that in all three regions ice cover in the twentieth century reduced, as compared with the nineteenth century. Significant year-to-year variations can be identified in all three records, which are different from one region to another. More recent compilations and syntheses of these and other sea-ice records from the region have shown common characteristics of decadal to multidecadal variability – see Sect. 2.6. The variability of sea ice in the EGC during the past millennium and farther back recently has been reconstructed from high-resolution (decadal) sediment cores from the northeast and southeast Greenland shelf (Miettinen et al. 2015; Perner et al. 2015; Kolling et al. 2017). Substantial fluctuations in the intensity of the EGC and sea ice are evident, including several large positive anomalies that arose abruptly. The most prominent of these was a near-century anomaly beginning in the late 1200s around the onset of the LIA.

2 Sea Ice in the Arctic Paleoenvironments

2.5.4

43

Newfoundland and the Gulf of St. Lawrence

A historical record of the sea ice incidence on the Scotia Shelf and the Gulf of St. Lawrence of Atlantic Canada has been compiled on an annual basis from the early 1800s (Hill 1999; Hill et al. 2002). Sea ice in this area typically forms in the western and northern coastal zones of the Gulf of St. Lawrence during December and by the end of January may be exported out of it under the influence of wind and ocean currents. The ice usually reaches its greatest areal extent during March. About 25,000 ice records, spanning the years 1769–1962, were used in the plotting of monthly ice charts for the region. Ice records included ice patrol and shipping reports, local newspapers, lighthouse records and others. Information for early years is sporadic but is continuous from 1817. The ice extent value for the year is an average of the months for which there are data, so it may represent just 1 month if that is all that is available, or an average of up to 4 or 5 months. A historical record of sea ice distribution for the east coast of Newfoundland from 1810 to present was compiled by Hill (1999) for the winter months, January to April. The earliest ice observations were made in 1527, but annual records are available only from 1810. Starting from 1923 ice data were obtained from the International Ice Patrol annual bulletins, and prior to this year come from shipping journals, log books, diaries, as well as from gazettes. Most of those years prior to 1859 are limited to two, or even one, map, and starting from 1859, separate charts for each of the winter months, February, March and April were composed (Hill 1999). Ice extent east of Cabot Strait is characterized by a high degree of interannual variability with changes from near minimum to near maximum values in just 2 or 3 years. The years 1922 and 1923 appear to have been the most severe for the entire record while the approximate 30-year period around the 1860s also appears to have been severe but not quite so extensive. Moving averages of ice extent east of Cabot Strait and east of Newfoundland are shown in Fig. 2.14. Comparison of the two data

Fig. 2.14 Time series of ice extent east of Cabot Strait (left axis) and east of Newfoundland (right axis) showing a 7-year moving average

44

L. P. Bobylev and M. W. Miles

sets show significant differences in the higher frequency ranges due to regional climatic differences, but there is quite a degree of similarity in the longer-term trends. Both plots show an increase in ice severity in the mid- to late nineteenth century with ameliorating conditions at the turn of the century. Both then show a return to more severe conditions in the 1910s and 1920s and since then there has been an unsteady but parallel return to lighter conditions.

2.5.5

Canadian Arctic

The last millennium in the Canadian Arctic presents a generally colder interval in which there appears to be no major, sustained reductions in sea ice cover but some slightly warmer and colder intervals that roughly correspond to the MCA and LIA, respectively (Mudie et al. 2005). Expansion of the Thule culture into the High Arctic coincided with warmer climatic conditions from 1100 to 800 years ago, followed by a retreat during the onset of more severe cold and sea-ice conditions. The North Water Polynya record show a relatively warm summer interval (+1
C increase) and reduced sea ice cover from ~1100 to 1400 and it shows the 1600–1845 heavy sea-ice cover event, with a corresponding summer temperature decrease of 0–5
C. Decadalscale quantitative records of changes in sea surface temperature in summer and winter, and in sea ice cover (as months with >50% sea ice) have been obtained using dinocyst assemblages as proxies. According to Grumet et al. (2001), the Baffin Island region experienced a relatively cold period beginning in the 1600s, and this cooling is supported by other ice core data from the eastern Canadian Arctic, climate records derived from laminated sediments, and the former extent of LIA glaciers. Cold climatic conditions were recorded during the 1800s. The authors mentioned a significant difference with the Icelandic sea-ice record, which shows a dramatic increase in sea-ice beginning in the late eighteenth century, approximately 40 years prior to sea-ice expansion documented in the Baffin record. The Icelandic record also shows a rapid decrease in sea-ice in the middle of the nineteenth century during a period of sea ice expansion in the Baffin Bay region. This inverse relationship may be explained by different response of these regions to shifts in the NAO. Icelandic sea ice varies when the NAO pattern is more zonal, whereas the meridional component influences sea-ice conditions in the North Atlantic. During positive NAO modes, the jet stream is positioned more poleward, and the Icelandic Low is deeper, such that sea ice in the Greenland/Barents Sea region is reduced and sea-ice extent increases in the Davis/ Labrador Sea region. The Baffin record shows a decrease in sea-ice coverage during the warmer period of the eighteenth century relative to the cooler seventeenth and early nineteenth centuries. Historical sea-ice records also come from reports of British whaling ships crushed by sea ice in Davis Strait and northern Baffin Bay, but these events cannot be used quantitatively. The milder years occurred from 1853 to 1881, when the ships of Kane, Hayes, Nares and Young returned safely via Nares Strait and Lieutenant

2 Sea Ice in the Arctic Paleoenvironments

45

Greely sailed into western Hall Basin. In the following 2 years, however, heavy ice prevented transit of supply ships and Greely lost 19 of his 26 men from starvation by the spring of 1884.

2.5.6

Hudson Bay

The Hudson’s Bay Company regularly dispatched a flotilla of ships from England to the Hudson Bay starting from 1668 to 1669. The ships’ logbooks dating back to 1751 are preserved in the Company’s archives (Catchpole and Faurer 1985). Catchpole and Faurer (1983) analyzed the descriptive information in the logbooks to yield annual indices of summer sea-ice severity in Hudson Strait for the period 1751–1870. Catchpole (1995) produced indices of summer ice severity in Hudson Bay, Hudson Strait and the western margin of the Labrador Sea. To calculate daily and annual ice indices, numerical values of ice severity were assigned to following ice categories: zero for “Ice absent”, one for open ice, two for “Closed Ice 1” and three for “Closed ice 2”. Annual indices of summer sea-ice severity in Hudson strait for the period 1751–1870 are shown in Fig. 2.15. According to Catchpole (1995), there was no year in which records for Hudson Strait, Hudson Bay west, and Hudson Bay east yielded indices ranking among the highest ten values, and severe ice indices in both two records were derived only in 1816 and 1836. The highest index of summer sea ice severity in Hudson Strait occurred in 1816, probably due to dust veil emitted into the stratosphere by the eruption of Mount Tambora in 1815 (Catchpole and Faurer 1985). In 1816 open ice lingered more than 1 month later than last-ice dates for the period 1964–1979 in

Fig. 2.15 Annual indices of summer sea ice severity in Hudson strait, 1751–1870. These indices are based on (a) the annual duration (days) of the westward passages of the Hudson’s Bay Company ships through the strait, and (b) the annual frequencies of occurrence (days) of closed-ice conditions during these westward passages. These data are corrected for the effects of seasonal changes in ice severity arising from variations in the dates of commencement of the westward passages (Catchpole and Faurer 1983). The interruption in 1839–1841 is due to missing logbooks (Source: Catchpole and Faurer (1985))

46

L. P. Bobylev and M. W. Miles

Hudson Bay. Evidence from Western Europe and eastern North America demonstrates that exceptionally cool weather was widespread in the summer of 1816. These results are consistent with findings obtained in a study of summer sea-ice conditions in the Labrador Sea during the nineteenth century (Newell 1990). Historical evidence from the Labrador coast north of 5500 N shows that the most severe summer ice conditions in this region throughout the nineteenth century occurred in 1816. An historical study of sea ice in Davis Strait in the nineteenth century made no reference to conditions in 1816 but did note the prevalence of comparatively severe ice conditions in 1817 (Speerschneider 1931). During several periods, as in 1813–1817, 1835–1836, 1843–1845, 1854–1855, 1859, and 1865–1866, severe ice indices occurred consecutively. Catchpole and Hanuta (1989) have identified a tendency for severe summer sea ice in Hudson Strait and Hudson Bay to follow major volcanic eruptions in the period 1751 to 1889. This condition occurred after the eruptions of Lakagigar in 1783, Tambora in 1815, Coseguina in 1835, Sheveluch in 1854, Askja in 1875, and Krakatau in 1883. The virtual lack of coincidence between the severe ice indices in the three records may cast doubt on the validity of these data, however it is consistent with the nature of summer ice dispersal processes in the bay and the strait.

2.6

Multidecadal Variability in Sea Ice

The existence of low frequency oscillation in sea ice and arctic climate was first suggested by Polyakov et al. (2003), based on twentieth century observations from the Russian Arctic seas. The multidecadal signal appeared to be strongest in the Kara Sea, with a weaker signal propagating eastward to seas in the Siberian Arctic. Frankcombe et al. (2010) used these data to demonstrate that the Kara Sea ice is significantly correlated with the Atlantic Multidecadal Oscillation (AMO) – a widely used index of Atlantic Multidecadal Variability (AMV). Their analysis revealed that a high AMO index precedes a minimum in sea-ice extent and the lag is consistent with the propagation of warm Atlantic water into the region. Multidecadal fluctuations are also apparent in century-scale records from other regions, including a composite index of four historical sea-ice records for the northern North Atlantic (Wood et al. 2010). The twentieth century observations show sea-ice fluctuations that are in qualitative agreement with the expected sign of the change, i.e. reductions during warm (AMO positive phase) periods such as the ETCW (Drinkwater et al. 2013 and references therein) and increases during cold (AMO negative phase) periods (Wood et al. 2010). A time series (1899–1971) of the duration of the presence of “Storis” on the shelf off southwestern Greenland also shows an AMO-like signal. Analyses of oceanographic indices and sea ice over approximately the last century and a half (Schmith and Hansen 2003), as well as high-resolution paleo proxy data of sea ice from shelves and fjords in southeastern Greenland (Andresen et al. 2011) and west Greenland (Lloyd et al. 2011), indicate that local

2 Sea Ice in the Arctic Paleoenvironments

47

Fig. 2.16 Linking multidecadal fluctuations in sea ice to North Atlantic sea-surface temperatures (SSTs). Original non-smoothed time series (gray) and multidecadal 50 to 120-year component (blue) reconstructed from wavelet decomposition: (top) Atlantic Multidecadal Oscillation (AMO) index. (upper middle) Fram Strait ice export reconstructed from historical arctic ice extent along SW Greenland (Schmith and Hansen 2003). (lower middle) Icelandic sea-ice severity index (1600–1870) (Oglivie 2005) and sea-ice incidence index (1880–2000). (bottom) Western Nordic Seas sea-ice extent proxy reconstruction (Macias-Fauria et al. 2009) (Modified from Miles et al. (2014))

conditions display the AMO warm and cold phases, including apparent effects on the calving of the major marine glaciers that terminate in the fjords. However, strong observational evidence of a robust sea ice–AMO connection has been elusive until recently, largely because century-scale datasets for sea ice and sea-surface temperature (SST) are not commensurate to tracking more than one complete cycle of multidecadal variability. More robust evidence for a multidecadal signal in Arctic sea ice thus must be derived from longer multi-century historical reconstructions, as well as from high-resolution paleo-proxy reconstructions. Recent advances in developing longer, multi-century sea-ice time series from longer historical (e.g., Divine and Dick 2006) and high-resolution paleoenvironmental proxies (e.g., Macias-Fauria et al. 2009) have made it possible to more robustly identify and track multidecadal sea-ice variability and establish linkages to the AMO. Possible connections between reconstructed sea-ice variability and the AMO have been suggested (Divine and Dick 2006; Macias-Fauria et al. 2009); however, these authors only pointed out the similar time scales. Multicentury sea-ice time series from the Greenland Sea and Iceland have since been compiled and explicitly compared with the AMO (Miles et al. 2014). Comparisons of sea-ice variability and the AMO are shown in Fig. 2.16, focusing on three

48

L. P. Bobylev and M. W. Miles

Fig. 2.17 Reconstructed Western Nordic Sea winter sea-ice extent, 1200–2000 (aqua, right y-axis, data from Macias-Fauria et al. 2009) and paleo reconstruction of an Atlantic Multidecadal Oscillation (AMO) temperature index, 600–2000 (blue, left y-axis, data from Mann et al. 2009)

independent sea-ice records from the Greenland Sea region, together with two AMO indices. The sea-ice records are the Fram Strait ice export reconstruction (Schmith and Hansen 2003), a revised historical sea-ice series from Iceland (Oglivie 2005), and a paleo-proxy reconstruction for the western Nordic Seas (Macias-Fauria et al. 2009). The AMO records used for comparison were an instrumental-based series and the proxy reconstruction based on tree rings (Gray et al. 2004). A longer AMO reconstruction (Mann et al. 2009) based on a proxy network also compares well with the western Nordic Seas reconstruction (Macias-Fauria et al. 2009) (Fig. 2.17). Lastly, an independent sea-ice proxy reconstruction (Halfar et al. 2013) from the Labrador Sea spanning over 600 years also exhibits multidecadal variability, with amplified magnitude since the nineteenth century, according to Moore et al. (2017). An important question is whether the AMO-related variability evident in the multi-century sea-ice series from the Greenland Sea and Iceland is evident in even longer paleo proxy sea-ice records. Millennial and multi-millennial proxy reconstructions are based predominantly on material from marine sediments, which means that there are relatively few records with temporal resolution is high enough to identify multidecadal variability. There are however some pertinent high-resolution records from the north Iceland shelf (Moros et al. 2006; Massé et al. 2008) and the northeast Greenland shelf (Perner et al. 2015; Kolling et al. 2017). The presence of multidecadal fluctuations in sea-ice cover has been identified using spectral analysis of the northeast Greenland shelf IP25 record (Kolling et al. 2017) and the north Iceland shelf quartz data record (Moros et al. 2006; Miles et al. 2014), the latter exhibiting a multidecadal signal through several millennia, albeit weaker in some intervals. This quasi-persistence is similar to the modulation of AMO-related signals found by Knudsen et al. (2011) in other high-resolution paleo records from the North Atlantic region (Knudsen et al. 2011), notably δ18O from an ice core from Greenland and reconstructed SSTs from north of Iceland. In summary, these studies show that multidecadal variability of sea ice is a common feature, at least in the Atlantic sector. This suggests that the sea ice decline in the recent decades is in part due to natural variability on the multidecadal time scale.

2 Sea Ice in the Arctic Paleoenvironments

2.7

49

Summary

This chapter has presented an overview of sea ice in the Arctic paleoenvironments, which demonstrates the large degree of natural sea-ice variability across a range of temporal and spatial scales. While not presenting an exhaustive or completely comprehensive collection of data records, we have illustrated many different ways that sea-ice reconstructions have been made from diverse paleo and historical data records and used to study past variability and changes – and interactive linkages within the climate system. Although natural variability on millennial, centennial, multidecadal and decadal time scales appears to be a robust feature in many records, especially in the North Atlantic Arctic, the exceptional state of the sea-ice cover of the twentieth and twenty-first centuries is clearly evident in the long-term paleo perspective.

References ACIA. (2004). Arctic climate impact assesssment: summary assessment report. Cambridge: Cambridge University Press. Online athttp://www.amap.no/acia Andresen, C. S., Straneo, F., Ribergaard, M. H., Bjørk, A. A., Andersen, T. J., Kuijpers, A., Nørgaard-Pedersen, N., Kjær, K. H., Schjøth, F., Weckström, K., & Ahlstrøm, A. P. (2011). Rapid response of Helheim Glacier in Greenland to climate variability over the past century. Naure Geoscience, 5, 37–41. https://doi.org/10.1038/NGEO1349. Andrews, J. T., Darby, D., Eberle, D., Jennings, A. E., Moros, M., & Ogilvie, A. (2009). A robust, multisite Holocene history of drift ice off northern Iceland: Implications for North Atlantic climate. The Holocene, 19, 71–77. https://doi.org/10.1177/0959683608098953. Armstrong, T. (1984). In search of a sea route to Siberia, 1553–1619. Arctic, 37(4), 429–440. Backman, J., Jakobsson, M., Frank, M., Sangiorgi, F., Brinkhuis, H., Stickley, C., O’Regan, M., Løvlie, R., Pälike, H., Spofforth, D., Gattacecca, J., Moran, K., King, J., & Heil, C. (2008). Age model and core-seismic integration for the Cenozoic Arctic. Coring expedition sediments from the Lomonosov Ridge. Paleoceanography, 23(1), PA1S03. https://doi.org/10.1029/ 2007PA001476. Barkov, N. I., & Petrov, V. N. (1994). Study of the polar regions paleoclimate from glacier deep cores. Arctic and Antarctic Research, 67–68, 10–24 (in Russian). Barnett, D. (1991). Sea ice distribution in the Soviet Arctic. In L. W. Brigham (Ed.), The Soviet Maritime Arctic (pp. 47–62). London: Belhaven Press. Barry, R. G. (1989). The present climate of the Arctic Ocean and possible past and future states. In Y. Herman (Ed.), The Arctic seas climatology, oceanography, geology, and biology (pp. 1–48). New York: Van Nostrand Reinhold Company. Belov, M. I. (1977). Following in the tracks of polar expeditions. Gidrometeoizdat, Leningrad (in Russian), 142 p Bengtsson, L., Semenov, V. A., & Johannessen, O. M. (2004). The early 20th century warming in the Arctic – a possible mechanism. Journal of Climate, 17, 4045–4057. Bobylev, L. P., Kondratyev, K. Y., & Johannessen, O. M. (2003). Arctic environment variability in the context of global change. Chichester: Springer-Praxis, 471 p. Borisenkov, Y. P. (1995). Documentary evidence from the U.S.S.R. In R. S. Bradley & P. D. Jones (Eds.), Climate since A.D. 1500 (pp. 171–183). London: Routledge.

50

L. P. Bobylev and M. W. Miles

Bradley, R. S., & Jones, P. D. (1995). Climate since A.D. 1500: Introduction. In R. S. Bradley & P. D. Jones (Eds.), Climate since A.D. 1500 (pp. 1–16). London: Routledge. Brigham-Grette, J., & Hopkins, D. M. (1995). Emergent-marine record and paleoclimate of the last interglaciation along the northwest Alaskan coast. Quaternary Research, 43, 154–173. Briner, J. P., McKay, N. P., Axford, Y., Bennike, O., Bradley, R. S., de Vernal, A., Fisher, D., Francus, P., Frechette, B., Gajewski, K., Jennings, A., Kaufman, D. S., Miller, G., Rouston, C., & Wagner, B. (2016). Holocene climate change in Arctic Canada and Greenland. Quaternary Science Reviews, 147, 340–364. Buch, E. (2000). A monograph on physical oceanography of the Greenland waters (Scientific Report 00-12). Copenhagen: Danish Meteorological Institute, 245 p. Carter, L. D., Brigham-Grette, J., Marincovich, L., Jr., Pease, V. L., & Hillhouse, J. W. (1986). Late Cenozoic Arctic Ocean sea ice and terrestrial paleoclimate. Geology, 14, 675–678. Catchpole, A. J. W. (1995). Hudson’s Bay Company ships’ log-books as sources of sea ice data, 1751–1870. In R. S. Bradley & P. D. Jones (Eds.), Climate since A.D. 1500. London: Routledge. Catchpole, A. J. W., & Faurer, M. A. (1983). Summer sea ice severity in Hudson Strait, 1751–1870. Climatic Change, 5, 115–139. Catchpole, A. J. W., & Faurer, M. A. (1985). Ships’ log-books, sea ice and the cold summer of 1816 in Hudson Bay and its approaches. Arctic, 38(2), 121–128. Catchpole, A. J. W., & Hanuta, I. (1989). Severe summer ice in Hudson Strait and Hudson Bay following major volcanic eruptions, 1751 to 1889 A.D. Climatic Change, 14, 61–79. Clark, D. L. (1971). Arctic Ocean ice cover and its late Cenozoic history. Geological Society of America Bulletin, 82, 3313–3324. Clark, D. L. (1982). The Arctic Ocean and Post-Jurassic paleoclimatology. In Climate in Earth history: Studies in geophysics (pp. 133–138). Washington, DC: The National Academies Press. Clotten, C., Stein, R., Fahl, K., & De Schepper, S. (2018). Seasonal sea ice cover during the warm Pliocene: Evidence from the Iceland Sea (ODP Site 907). Earth and Planetary Science Letters, 481, 61–72. Cronin, T. M., Gemery, L., Briggs, W. M., Jr., Jakobsson, M., Polyak, L., & Brouwers, E. M. (2010). Quaternary sea ice history in the Arctic Ocean based on a new ostracode sea-ice proxy. Quaternary Science Reviews, 29(25–26), 3415–3429. https://doi.org/10.1016/j.quascirev.2010. 05.024. Danilov, I. D. (1989). Geological and paleoclimatic evolution of the Arctic during late Cenozoic time. In Y. Herman (Ed.), The Arctic seas climatology, oceanography, geology, and biology (pp. 759–760). New York: Van Nostrand Reinhold Company, 888 p. Darby, D. A. (2008). Arctic perennial ice cover over the last 14 million years. Paleoceanography, 23, PA1S07. https://doi.org/10.1029/2007PA001479. de Vernal, A., Gersonde, R., Goossec, H., Seidenkrantz, M. S., & Wolff, E. W. (2013). Sea ice in the paleoclimate system: The challenge of reconstructing sea ice from proxies – an introduction. Quaternary Science Reviews, 79, 1–8. Denton, G. H., Anderson, R. F., Toggweiler, J. R., Edwards, R. L., Schaefer, J. M., & Putnam, A. E. (2010). The last glacial termination. Science, 328, 1652–1656. https://doi.org/10.1126/science. 1184119. Ding, Q., Schweiger, A., L’Heureux, M., Battisti, D. S., Po-Chedley, S., Johnson, N. C., BlanchardWrigglesworth, E., Harnos, K., Zhang, Q., Eastman, R., & Steig, E. J. (2017). Influence of highlatitude atmospheric circulation changes on summertime Arctic sea ice. Nature Climate Change, 7, 289–295. Divine, D. V., & Dick, C. (2006). Historical variability of sea ice edge position in the Nordic Seas. Journal of Geophysical Research, 111, C01001. https://doi.org/10.1029/2004JC002851. Dowsett, H., Dolan, A., Rowley, D., Moucha, R., Forte, A. M., Mitrovica, J. X., Pound, M., Salzmann, U., Robinson, M., Chandler, M., Foley, K., & Haywood, A. (2016). The PRISM4 (mid-Piacenzian) paleoenvironmental reconstruction. Climate of the Past, 12, 1519–1538.

2 Sea Ice in the Arctic Paleoenvironments

51

Drinkwater, K., Miles, M. W., Medhaug, I., Otterå, O. H., Kristiansen, T., Sundby, S., & Gao, Y. (2013). Atlantic multidecadal oscillation: Its manifestations and impacts with emphasis on the region north of 60
N. Journal of Marine System, 128. https://doi.org/10.1016/j.jmarsys. 2013.11.001. Dyke, A. S., & England, J. (2003). Canada’s most northerly postglacial bowhead whales (Balaena mysticetus): Holocene sea-ice conditions and polynya development. Arctic, 56, 14–20. Dyke, A. S., Hooper, J., & Savelle, J. M. (1996). A history of sea ice in the Canadian Arctic archipelago based on the post glacial remains of the bowhead whale (Balaena mysticetus). Arctic, 49, 235–255. Dyke, A. S., England, J., Reimnitz, E., & Jette, H. (1997). Changes in driftwood delivery to the Canadian Arctic archipelago: The hypothesis of postglacial oscillations of the transpolar drift. Arctic, 50, 1–16. Eicken, H., Reimnitz, E., Alexandrov, V. Y., Martin, T., Kassens, H., & Viehoff, T. (1997). Sea-ice processes in the Laptev Sea and their importance for sediment export. Continental Shelf Research, 17(2), 205–233. England, J. H., Lakeman, T. R., Lemmen, D. S., Bednarski, J. M., Stewart, T. G., & Evans, D. J. A. (2008). A millennial-scale record of Arctic Ocean sea ice variability and the demise of the Ellesmere Island ice shelves. Geophysical Research Letters, 35, L19502. https://doi.org/10. 1029/2008GL034470. Fischer, H., et al. (2018). Palaeoclimate constraints on the impact of 2
C anthropogenic warming and beyond. Nature Climate Change, 11, 615. Fletcher, J. O. (1968). Changing climate. Santa Monica: Rand Corp, 27 p. Frankcombe, L. M., von der Heydt, A., & Dijkstra, H. A. (2010). North Atlantic multidecadal climate variability: An investigation of dominant time scales and processes. Journal of Climate, 23, 3626–3638. Frolov, I. E., Gudkovich, Z. M., Karklin, V. P., Kovalev, E. G., & Smolyanitsky, V. M. (2009). Climate change in Eurasian Arctic shelf seas. Centennial ice cover observations. Chichester: Praxis Publishing Ltd, 164 p. Funder, S., Goosse, H., Jepsen, H., Kaas, E., Kjær, K. H., Korsgaard, N. J., Larsen, N. K., Linderson, H., Lyså, A., Möller, P., Olsen, J., & Willerslev, E. (2011). A 10,000-year record of Arctic Ocean sea-ice variability—view from the beach. Science, 333, 747. https://doi.org/10. 1126/science.1202760. 747–750. Gray, S., Graumlich, L., Betancourt, J., & Pederson, G. (2004). A tree-ring based reconstruction of the Atlantic Multidecadal Oscillation since 1567 A.D. Geophysical Research Letters, 31, L12205. https://doi.org/10.1029/2004GL019932. Gribbin, J., & Lamb, H. H. (1978). Climatic change in historical times. In J. Gribbin (Ed.), Climatic change (pp. 68–82). London: Cambridge University Press. Grumet, N. S., Wake, C. P., Mayevski, P., Zielinski, G. A., Whitlow, S. I., Koerner, R., Fisher, D. A., & Woollett, J. M. (2001). Variability of sea-ice extent in Baffin Bay over the last millennium. Climatic Change, 49, 129–145. Halfar, J., Adey, W. H., Kronz, A., Hetzinger, S., Edinger, E., & Fitzhugh, W. W. (2013). Arctic sea-ice decline archived by multicentury annual-resolution record from crustose coralline algal prox. Proceedings of the National Academy of Science, 110, 19737–19741. https://doi.org/10. 1073/pnas.1313775110. Hanslik, D., Jakobsson, M., Backman, J., Bjorck, S., Sellen, E., O’Regan, M., Fornaciari, E., & Skog, G. (2010). Quaternary Arctic Ocean sea ice variations and radiocarbon reservoir age corrections. Quaternary Science Reviews, 29, 3430–3441. Henshaw, A. (2003). Climate and culture in the North: The interface of archaeology, paleoenvironmental science, and oral history. In S. Strauss & B. Orlove (Eds.), Weather, climate, culture (pp. 217–232). Oxford: Berg. Herman, Y., Osmond, J. K., & Somayajulu, B. L. K. (1989). Late Neogene Arctic paleooceanography: Micropaleontology, stable isotopes and chronology. In Y. Herman (Ed.),

52

L. P. Bobylev and M. W. Miles

The Arctic seas climatology, oceanography, geology, and biology (pp. 581–655). New York: Van Nostrand Reinhold Company, 888 p. Hill, B. T. (1999) Historical record of sea ice and iceberg distribution around Newfoundland and Labrador, 1810–1958. In Proceedings of the workshop of Arctic climate system study: Sea-ice charts of the Arctic, Seattle, WA, USA, 5–7 August 1998, (pp. 13–15). WMO/TD No. 949 IAPO Publication No. 3. Hill, B.T. (2008). Recent sea-ice records from Labrador and Newfoundland. In Proceedings of ICASS VI Conference, Nuuk, Greenland. Hill, B. T., Ruffman, A., Drinkwater, K. (2002) Historical record of the incidence of sea ice on the Scotian shelf and the Gulf of St. Lawrence. Ice in the Environment. In: Proceedings of the 16th IAHR International Symposium on Ice Dunedin, New Zealand, 2nd–6th December 2002 (pp. 16–24). New Zealand: International Association of Hydraulic Engineering and Research. Hughes, A. L. C., Gyllencreutz, R., Lohne, Ø. S., Mangerud, J., & Svendsen, J. I. (2016). The last Eurasian Ice Sheets – a chronological database and time-slice reconstruction, DATED-1. Boreas, 45(1), 1–45. https://doi.org/10.1111/bor.12142. Intergovernmental Panel on Climate Change (IPCC) Climate Change. (2013). Fifth assessment report. Cambridge, UK: Cambridge University Press. Itin, V. (1933). Sea ice routes in the Soviet Arctic. Moscow: Publishing House of Soviet Asia, 110 p (in Russian). Jakobsson, M., Backman, J., Rudels, B., Nycander, J., Frank, M., Mayer, L., Jokat, W., Sangiorgi, F., O'Regan, M., Brinkhuis, H., King, J., & Moran, K. (2007). The early Miocene onset of a ventilated circulation regime in the Arctic Ocean. Nature, 447, 986–990. https://doi.org/10. 1038/nature05924. Jakobsson, M., Long, A., Ingólfsson, Ó., Kjær, K. H., & Spielhagen, R. F. (2010). New insights on Arctic quaternary climate variability from palaeo-records and numerical modelling. Quaternary Science Reviews, 29, 3349–3358. Jansen, E., Fronval, T., Rack, F., & Channell, J. E. T. (2000). Pliocene-Pleistocene ice rafting history and cyclicity in the Nordic Seas during the last 3.5 Myr. Paleoceanography, 15(6), 709–721. Jennings, A. E., Knudsen, K. L., Hald, M., Hansen, C. V., & Andrews, J. T. (2002). A mid-Holocene shift in Arctic sea-ice variability on the East Greenland Shelf. Holocene, 12(1), 49–58. Jennings, A., Andrews, J., & Wilson, L. (2011). Holocene environmental evolution of the SE Greenland shelf north and south of the Denmark Strait: Irminger and east Greenland current interactions. Quaternary Science Reviews, 30(7–8), 980–998. Johannessen, O. M. (2008). Decreasing Arctic sea ice mirrors increasing CO2 on decadal time scale. Atmospheric and Oceanic Science Letters, 1, 51–56. Johannessen, O. M., Miles, M., & Bjørgo, E. (1995). The Arctic’s shrinking sea ice. Nature, 376, 126–127. Johannessen, O. M., Bengtsson, L., Miles, M. W., Kuzmina, S. I., Semenov, V. A., Alekseev, G. V., Nagyrnyi, A. P., Zakharov, V. F., Bobylev, L. P., Pettersson, L. H., Hasselmann, K., & Cattle, H. P. (2004). Arctic climate change: Observed and modeled temperature and sea ice variability. Tellus, 56A, 328–341. Johannessen, O. M., Alexandrov, V., Frolov, I. Y., Sandven, S., Bobylev, L. P., Pettersson, L. H., Kloster, K., Smirnov, V. G., Mironov, Y. U., & Babich, N. G. (2006). Remote sensing of sea ice in the Northern Sea route: Studies and applications. Chichester: Springer, 472 p. Kaufmann, D. S., 31 others. (2004). Holocene thermal maximum in the western Arctic (0–180
W). Quaternary Science Reviews, 23, 529–560. Kinnard, C., Zdanovicz, C. M., Fisher, D. A., Isaksson, E., de Vernal, A., & Thompson, L. G. (2011). Reconstructed changes in Arctic sea ice over the past 1,450 years. Nature, 479, 509–513.

2 Sea Ice in the Arctic Paleoenvironments

53

Knudsen, M. F., Seidenkrantz, M.-S., Jacobsen, B. H., & Kuijpers, A. (2011). Tracking the Atlantic Multidecadal Oscillation through the last 8000 years. Nature Communications, 2, 178. https:// doi.org/10.1038/ncomms1186. Koch, L. (1945). The East Greenland ice. Meddelelser om Grønland, 130(3), 1–374. Kolling, H. M., Stein, R., Fahl, K., Perner, K., & Moros, M. (2017). Short-term variability in late Holocene sea ice cover on the East Greenland Shelf and its driving mechanisms. Palaeogeography, Palaeoclimatology, Palaeoecology, 18, 336–350. https://doi.org/10.1016/j. palaeo.2017.06.024. Krylov, A. A., Andreeva, I. A., Vogt, C., Backman, J., Krupskaya, V. V., Grikurov, G. E., Moran, K., & Shoji, H. (2008). A shift in heavy and clay mineral provenance indicates a middle Miocene onset of a perennial sea ice cover in the Arctic Ocean. Paleoceanography, 23. https://doi.org/10.1029/2007PA001497. Lamb, H. H. (1979). Climatic variation and changes in the wind and ocean circulation: The Little Ice Age in the Northeast Atlantic. Quaternary Research, 11, 1–20. Lassen, K., & Thejll, P. (2005). Multi-decadal variation of the East Greenland Sea-Ice extent: AD 1500–2000 (Danish Meteorological Institute Scientific Report 05-02). Copenhagen: Danish Meteorological Institute, 13 p. Lestgaft, E. F. (1913). Ice of the Arctic Ocean and a sea route route from Europe to Siberia. St. Petersburg: Imperial Academy of Science (in Russian), 28 pp. Levac, E., de Vernal, A., & Blake, W. J. (2001). Sea-surface conditions in northernmost Baffin Bay during the Holocene – palynological evidence. Journal of Quaternary Science, 16, 353–363. Lloyd, J., Moros, M., Perner, K., Telford, R. J., Kuijpers, A., Jansen, E., & McCarthy, D. (2011). A 100 yr record of ocean temperature control on the stability of Jakobshavn Isbrae, West Greenland. Geology, 39, 867–870. Macias Fauria, M., Grinsted, A., Helama, S., Moore, J., Timonen, M., Martma, T., Isaksson, E., & Eronen, M. (2009). Unprecedented low twentieth century winter sea ice extent in the Western Nordic seas since a.D. 1200. Climate Dynamics, 34, 781–795. Mahoney, A. R., Barry, R. G., Smolyanitsky, V., & Fetterer, F. (2008). Observed sea ice extent in the Russian Arctic, 1933-2006. Journal of Geophysical Research, 113, C11005. https://doi.org/ 10.1029/2008JC004830. Mangerud, J., Jakobsson, M., Alexanderson, H., Astakhov, V., Clarke, G. K. C., Henriksen, M., Hjort, C., Krinner, G., Lunkka, J.-P., Möller, P., Murray, A., Nikolskaya, O., Saarnisto, M., & Svendsen, J. I. (2004). Ice-dammed lakes and rerouting of the drainage of northern Eurasia during the last glaciation. Quaternary Science Reviews, 23, 1313–1332. Mann, M. E., Zhang, Z., Rutherford, S., Bradley, R. S., Hughes, M. K., Shindell, D., Ammann, C., Faluvegi, G., & Ni, F. (2009). Global signatures and dynamical origins of the Little Ice Age and Medieval Climate Anomaly. Science, 326, 1256–1260. https://doi.org/10.1126/ science1177303. Massé, G., Rowland, S. J., Sicre, M.-A., Jacob, A. J., Jansen, E., & Belt, S. T. (2008). Abrupt climate changes for Iceland during the last millennium: Evidence from high resolution sea ice reconstructions. Earth and Planetary Science Letters, 269(3–4), 565–569. https://doi.org/10. 1016/j.epsl.2008.03.017. McBean, G., Alekseev, G., Chen, D., Førland, E., Fyfe, J., Groisman, P.Y., King, R., Melling, H., Vose, R., Whitfield, P.H. (2004). Arctic climate: Past and present. Chapter 2 of the ACIA Scientific report 2004, (pp. 21–60). Melles, M., Brigham-Grette, J., Minyuk, P. S., Nowaczyk, N. R., Wennrich, V., DeConto, R. M., Anderson, P. M., Andreev, A. A., Coletti, A., Cook, T. L., Haltia-Hovi, E., Kukkonen, M., Lozhkin, A. V., Rosen, P., Tarasov, P., Vogel, H., & Wagner, B. (2012). 2.8 million years of Arctic climate change from Lake El’gygytgyn, NE Russia. Science, 337, 315–320. Miettinen, A., Divine, D. V., Husum, K., Koç, N., & Jennings, A. (2015). Exceptional ocean surface conditions on the SE Greenland shelf during the Medieval Climate Anomaly. Paleoceanography, 30, 1–17.

54

L. P. Bobylev and M. W. Miles

Mikkelsen, N., Nørgaard-Pedersen, N., Kristoffersen, Y., Lassen, S. J., & Sheldon, E. (2006). Radical past climatic changes in the Arctic Ocean and a geophysical signature of the Lomonosov Ridge north of Greenland. Geological Survey of Denmark and Greenland Bulletin, 10, 61–64. Miles, M. W., Divine, D. V., Furevik, T., Jansen, E., Moros, M., & Ogilvie, A. E. J. (2014). A signal of persistent Atlantic multidecadal variability in Arctic sea ice. Geophysical Research Letters, 41, 463–469. Moffa-Sánchez, P., Hall, I. R., Barker, S., Thornalley, D. J. R., & Yashayaev, I. (2014). Surface changes in the eastern Labrador Sea around the onset of the Little Ice Age. Paleoceanography, 29, 160–175. Moore, G. W. K., Halfar, J., Majeed, H., Adey, W., & Kronz, A. (2017). Amplification of the Atlantic Multidecadal Oscillation associated with the onset of the industrial-era warming. Scientific Reports, 7, 40861. https://doi.org/10.1038/srep40861. Moran, K., Backman, J., Brinkhuis, H., Clemens, S. C., Cronin, T., Dickens, G. R., Eynaud, F., Gattacceca, J., Jakobsson, M., Jordan, R. W., Kaminski, M., King, J., Koç, N., Krylov, A., Martinez, N., Matthiessen, J., McInroy, D., Moore, T. C., Onodera, J., O’Regan, A. M., Palike, H., Rea, B., Rio, D., Sakamoto, T., Smith, D. C., Stein, R., St. John, K., Suto, I., Suzuki, N., Takahashi, K., Watanabe, M., Yamamoto, M., Farrell, J., Frank, M., Kubik, P., Jokat, W., & Kristoffersen, Y. (2006). The Cenozoic palaeoenvironment of the Arctic Ocean. Nature, 441, 601–605. Moros, M., Andrews, J. T., Eberl, D. D., & Jansen, E. (2006). Holocene history of drift ice in the northern North Atlantic: Evidence for different spatial and temporal modes. Paleoceanography, 21, PA2017. https://doi.org/10.1029/2005PA001214. Mudie, P. J., Rochon, A., & Levac, E. (2005). Decadal-scale sea ice changes in the Canadian Arctic and their impacts on humans during the past 4,000 years. Environmental Archaeology, 10, 113–126. Nazarov, V. S. (1947). Historical variation of ice conditions in the Kara sea. Izv Vses Geogr Obsch (Leningrad), 79(6), 653–655 (in Russian). Newell, J. P. (1990). Spring and summer sea ice and climate conditions in the Labrador Sea, 1800Present. PhD thesis. University of Colorado, 332 p. Nørgaard-Pedersen, N. (2009). Tracking ancient sea ice. Nature Geoscience, 2, 743–744. Ogilvie, A. E. J. (1981) Climate and society in Iceland from the Medieval Period to the late eighteenth century. Ph.D. Dissertation. Norwich: University of East Anglia. Ogilvie, A. E. J. (1984). The past climate and sea-ice record from Iceland, part 1: Data to A.D. 1780. Climatic Change, 6, 131–152. Ogilvie, A. E. J. (1986). The climate of Iceland, 1701-1784. Jökull, 36, 57–73. Ogilvie, A. E. J. (1992). Documentary evidence for changes in the climate of Iceland, A.D. 1500 to 1800. In R. S. Bradley & P. D. Jones (Eds.), Climate since A.D. 1500 (pp. 92–115). London: Routledge. Ogilvie, A. E. J. (1995). Documentary evidence for changes in the climate of Iceland, A.D. 1500 to 1800. In R. S. Bradley & P. D. Jones (Eds.), Climate since A.D. 1500 (pp. 171–183). London: Routledge. Ogilvie A. E. J. (1999). Historical sea-ice records from Iceland ca. AD 1145 to ca. 1850. In Proceedings of the Workshop on sea-ice charts of the Arctic, Seattle, WA, USA, 5–7 August 1998, WMO/TD No. 949 IAPO Publication No. 3. Ogilvie, A. E. J., & Jónsdóttir, I. (2000). Sea ice, climate, and Icelandic fisheries in the eighteenth and nineteenth centuries. Arctic, 53(4), 383–394. Ogilvie, A. E. J., & Jonsson, T. (2001). “Little Ice age” research: A perspective from Iceland. Climatic Change, 48, 9–52. Ogilvie, A. E. J., & Palsson, G. (2003). Mood, magic, and metaphor: Allusions to weather and climate in the sagas of Icelanders. In S. Strauss & B. Orlove (Eds.), Weather, climate, culture (pp. 217–232). Oxford: Berg.

2 Sea Ice in the Arctic Paleoenvironments

55

Oglivie, A. E. J. (2005). Local knowledge and travellers’ tales: A selection of climatic observations in Iceland. In C. Caseldine et al. (Eds.), Iceland—modern processes and past environments (pp. 257–287). Amsterdam: Elsevier. Parkinson, C. L., & Cavalieri, D. J. (2008). Arctic sea ice variability and trends, 1979–2006. Journal of Geophysical Research, 113, C07003. https://doi.org/10.1029/2007JC004558. Patterson, W. P., Dietrich, K. A., Holmden, C., & Andrews, J. T. (2010). Two millennia of North Atlantic seasonality and implications for Norse colonies. Proceedings of the National Academy of Science, 107, 5306–5310. https://doi.org/10.1073/pnas.0902522107. Perner, K., Moros, M., Lloyd, J. M., Jansen, E., & Stein, R. (2015). Mid to late Holocene strengthening of the East Greenland current linked to warm subsurface Atlantic water. Quaternary Science Reviews, 129, 296–307. Polyak, L., & Jakobsson, M. (2011). Quaternary sedimentation in the Arctic Ocean: Recent advances and further challenges. Oceanography, 24(3), 52–64. https://doi.org/10.5670/ oceanog.2011.55. Polyak, L., Alley, R. B., Andrews, J. T., Brigham-Grette, J., Cronin, T. M., Darby, D. A., Dyke, A. S., Fitzpatrick, J. J., Funder, S., Holland, M., Jennings, A. E., Miller, G. H., O'Regan, M., Savelle, J., Serreze, M., John, K. S., Whiter, J. W. C., & Wolff, E. (2010). History of sea ice in the Arctic. Quaternary Science Reviews, 29, 1757–1778. Polyakov, I., Alekseev, G. V., Bekryaev, R. V., Bhatt, U. S., Colony, R., Johnson, M. A., Karklin, V. P., Walsh, D., & Yulin, A. V. (2003). Long-term ice variability in arctic marginal seas. Journal of Climate, 16, 2078–2085. Reimnitz, E., McCormick, M., McDougall, K., & Brouwers, E. (1993). Sediment export by ice rafting from a coastal polynya, Arctic Alaska, USA. Arctic and Alpine Research, 25, 83–98. Schmith, T., & Hansen, C. (2003). Fram Strait ice export during the nineteenth and twentieth centuries reconstructed from a multiyear sea ice index from southwestern Greenland. Journal of Climate, 16, 2782–2791. Seidenkrantz, M.-S., de Vernal, A., Goosse, H., Klein, F., Solignac, S., van Nieuwenhove, N., Pearce, C., Caissie, B., Belt, S., Sha, L., Cronin, T. M., Stein, R., Macias-Fauria, M., & DeNinno, L. H. (2016). Holocene Northern Hemisphere sea-ice distribution – proxy data reconstruction and modelling. Geophysical Reseach Abstracts, 18, EGU2016–EG10461. Serreze, M. C., Holland, M. M., & Stroeve, J. C. (2007). Perspectives on the Arctic’s shrinking sea-ice cover. Science, 315, 1533–1536. Shapiro, I., Colony, R., & Vinje, T. (2003). April sea ice extent in the Barents Sea, 1850–2001. Polar Research, 22(1), 5–10. Sibirtsev, N., Itin, V. (1936) The Northern Sea route and the Kara expeditions. Novosibirsk: West Siberia Publishing house (in Russian), 231 p. Sicre, M.-A., Yiou, P., Eiríksson, J., Ezat, U., Guimbaut, E., Dahhaoui, I., Knudsen, K.-L., Jansen, E., & Turon, J.-L. (2008). A 4500-year reconstruction of sea surface temperature variability at decadal time-scales off North Iceland. Quaternary Science Reviews, 27, 2041–2047. Speerschneider, C. I. H. (1931). The state of the ice in Davis Strait, 1820–1930. Med. Danske Met. Inst., Nr. 8 (Copenhagen), 53 p. Spielhagen, R. F., Werner, K., Sorensen, S. A., Zamelczyk, K., Kandiano, E., Budeus, G., Husum, K., Marchitto, T. M., & Hald, M. (2011). Enhanced modern heat transfer to the Arctic by warm Atlantic water. Science, 331, 450–453. Spolaor, A., Vallelonga, P., Turetta, C., Maffezzoli, N., Cozzi, G., Gabrieli, J., Barbante, C., GotoAzuma, K., Saiz-Lopez, A., Cuevas, C. A., & Dahl-Jensen, D. (2016). Canadian Arctic sea ice reconstructed from bromine in the Greenland NEEM ice core. Scientific Reports, 6, 33925. Stein, R., Fahl, K., Gierz, P., Niessen, F., & Lohmann, G. (2017). Arctic Ocean sea ice cover during the penultimate glacial and the last interglacial. Nature Communications, 8, 373. Stickley, C. E., John, K. S., Koc, N., Jordan, R. W., Passchier, S., Pearce, R. B., & Kearns, L. E. (2009). Evidence for middle Eocene Arctic sea ice from diatoms and ice-rafted debris. Nature, 460, 376–379. https://doi.org/10.1038/nature08163.

56

L. P. Bobylev and M. W. Miles

Stroeve, J. C., Holland, M. M., Meier, W., Scambos, T., & Serreze, M. (2007). Arctic sea ice decline: Faster than forecast. Geophysical Research Letters, 34, L090501. https://doi.org/10. 1029/2007GL029703. Thostrup, L. & Rasmussen, R. O., (2009). Climate change and the North Atlantic. NORA, 128 p. Vare, L. L., Massé, G., Gregory, T. R., Smart, C. W., & Belt, S. T. (2009). Sea ice variations in the central Canadian Arctic Archipelago during the Holocene. Quaternary Science Reviews, 28, 1354–1366. Vinje, T. (1999). Barents Sea-ice edge variation over the past 400 years. In: Proceedings of the Workshop on sea-ice charts of the Arctic, Seattle, WA, USA, 5–7 August 1998, WMO/TD No. 949 IAPO Publication No. 3, pp. 4–6. Vinje, T. (2001). Anomalies and trends of sea-ice extent and atmospheric circulation in the Nordic Seas during the period 1864-1998. Journal of Climate, 14, 255–267. Vinje, T., & Goosse, H. (2003). Ice extent variations during last centuries (Final ACSYS Science Conference). St. Petersburg: Arctic and Antarctic Research Institute. Vize, V. Y. (1939) Seas of the Soviet Arctic. Moscow: Izd. Glavsevmorputi (in Russian), 567 pp. Werner, K., Spielhagen, R. F., Bauch, D., Hass, H. C., & Kandiano, E. (2013). Atlantic water advection versus sea-ice advances in the eastern Fram Strait during the last 9 kya: Multiproxy evidence for a two-phase Holocene. Paleoceanography, 28(2), 283–295. Wood, K. R., Overland, J. E., Jonsson, T., & Smoliak, B. V. (2010). Air temperature variations on the Atlantic–Arctic boundary since 1802. Geophysical Research Letters, 37, L17708. https:// doi.org/10.1029/2010GL044176. Zakharov, V. F. (1995). Sea ice in the climate system. Arctic and Antarctic Research, 69, 15–26 (in Russian). Zakharov, V. F. (2004). Chapter 4: Changes of the Arctic Sea ice state. In G. V. Alekseev (Ed.), Formation and dynamics of the contemporary Arctic climate (pp. 112–159). St. Petersburg: Arctic and Antarctic Research Institute. Zhang, Q., Xiao, C., Ding, M., & Dou, T. (2018). Reconstruction of autumn sea ice extent changes since AD1289 in the Barents-Kara Sea, Arctic. Science China Earth Sciences. https://doi.org/ 10.1007/s11430-017-9196-4.

Chapter 3

Marginal Ice Zone and Ice-Air-Ocean Interactions Ola M. Johannessen, Stein Sandven, Richard Davy, and Einar O. Olason

3.1

Marginal Ice Zone. A Brief Review

Ola M. Johannessen and Stein Sandven The section review major experiments carried out in the Marginal Zone (MIZ), (Sect. 3.1.2), and the mesoscale processes along the ice edge in the Fram Strait, Greenland and Barents Seas such as the Polar Ocean Fronts in the Greenland and Barents Seas (Sect. 3.1.3), ice-ocean eddies (Sect. 3.1.4), ice edge upwelling (Sect. 3.1.5), internal waves in the MIZ (Sect. 3.1.6), the future of MIZ (Sect. 3.1.7) before concluding with a Summary of the section (Sect. 3.1.8).

3.1.1

Introduction

The Marginal Ice Zone (MIZ) is the crucial region in which the polar air, ice and ocean interact. The exchanges which take place in the MIZ profoundly influence hemispheric climate system (Johannessen et al. 2004, 2016) and have also significant effects on fisheries (Drinkwater et al. 2014), shipping, petroleum exploration and production, tourism and naval operation (Johannessen et al. 1992, 2003, 2007). The O. M. Johannessen (*) Nansen Scientific Society, Bergen, Norway e-mail: ola.johannessen@nansenscientificsociety.no S. Sandven Nansen Environmental and Remote Sensing Center, Bergen, Norway University Centre in Svalbard, Longyearbyen, Svalbard, Norway e-mail: [email protected] R. Davy · E. O. Olason Nansen Environmental and Remote Sensing Center, Bergen, Norway e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2020 O. M. Johannessen et al. (eds.), Sea Ice in the Arctic, Springer Polar Sciences, https://doi.org/10.1007/978-3-030-21301-5_3

57

58

O. M. Johannessen et al.

Fig. 3.1 Arctic sea ice concentration from satellite microwave data (SSMIS), (a) 15 March 2017 and (b) 15 September 2017. (From http://arctic-roos.org)

MIZ is located where sea ice and open ocean meets. When the wind is off the ice cover, strips of ice bands are drifting away from the major ice cover causing the ice edge region to be very diffuse, stretching kilometres by kilometres off the main ice cover. However, when the wind is on the ice cover the ice edge is very sharp, also when the wind is parallel to the ice edge when the ice edge is located to the right of the wind direction caused by higher wind stress over the ice cover (Smith et al. 1970) resulting in Ekman transport to the right forming a sharp ice edge causing upwelling along the ice edge (Johannessen et al. 1983). Contrary, when the ice edge is to the left of the wind direction, ice bands are formed and transported off the ice edge due to Ekman transport to the right (Fenell and Johannessen 1989). In addition, ice edgeocean eddies which are abundant along the ice edge cause rapid melting of the ice edge due to the eddy transport of ice out in the warmer water (Johannessen et al. 1987a). The width of the MIZ ice cover also varies significantly with the on ice wind generated surface waves and ocean swells (Wadhams et al. 1986, 2018; Williams et al. 2017). The waves and swells can penetrate up to more than 100 km into the ice cover and breaking it up into floes. During summer time when the surface water between the broken-up ice floes due to wave penetration get warmer, side ablation causes additional melting, enhancing the melt of the ice cover. So far the important effects of waves, swells and ice edge-ocean eddies causing a decrease of the ice cover in the MIZ are not included in either regional or coupled climate models. The MIZ has also large seasonal and interannual variations. Figure 3.1a, b shows the winter and summer location of the ice cover and MIZ in the Arctic Ocean which in recent years has open up the ice free Northern Sea Route to more shipping traffic during the warm season (Johannessen et al. 2007). The air-sea-ice processes are reviewed in Sect. 3.2 in this Chapter.

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

3.1.2

59

Studies in MIZ

Up to the end of the 1970s only sporadic field investigations had occurred in the MIZ in Fram Strait, Greenland and Barents Seas. However, during 1979–1992 the following six major international experiments took place in these regions both during summer and winter time: NORSEX 79 (NORSEX Group 1983), MIZEX 1983–1984 (MIZEX Group 1986), MIZEX 87 (MIZEX Group 1989), SIZEX 89 (Johannessen et al. 1989) and SIZEX 92 (Johannessen et al. 1993; Sandven et al. 1999). These experiments were all interdisciplinary with the overall objectives to improve our understanding of the air-ice-ocean-acoustic-chemical and biological interactive processes in the MIZ in order that new knowledge of these processes could form a base for improved modelling both regional and to be parameterized in coupled climate models. Furthermore, an important objective was also to validate passive and active microwave sensors for MIZ such as the NORSEX algorithm (Svendsen et al. 1983) for ice edge detection and ice concentration from the Scanning Multichannel Microwave Radiometer (SMMR) of the NIMBUS 7 NASA satellite and the application of Synthetic Aperture Radar (SAR) for MIZ process studies (NORSEX Group 1983). SIZEX 89 was a prelaunch ESA ERS1 experiment (Johannessen et al. 1989), while winter SIZEX 92 was the first experiment for the validation of the SAR images in the MIZ from the ERS 1 satellite, which was launched in 1991, for ice edge detection, ice type discrimination, ice concentration and ice velocities (Johannessen et al. 1993; Sandven et al. 1999). The concept and strategy of all these large international experiments were to collect data from icebreakers and ice – strengthen vessels equipped with helicopters operating in the ice edge region and inside the ice pack, from oceanographical research vessels operating in the open water adjacent to the ice edge, from an array of drifting Argos buoys placed on ice floes equipped with oceanographic and meteorological sensors, bottom moored buoys, remote sensing aircraft and several satellite systems. For example, MIZEX 84 is still the largest experiment ever carried out in the MIZ, employing 7 research vessels, 8 aircraft, 4 helicopters, drifting and moored buoys, the use of two satellite systems including more than 200 scientists and technicians from 11 nations (Johannessen 1987). Some of these data were analysed in real time, in particular the remote sensing observations and the Argos drifting buoys, which enabled us to guide the different platforms to the active mesoscale process studies, for example the ice edge-ocean eddies which was one of the focus of the program. An extensive review of the NORSEX 79 – MIZEX 83, 84, 87 and SIZEX 1989 and references therein can be found in Johannessen et al. (1992) who also was the leader of all these experiments. This paper also includes validation and application of passive and active remote sensing methods used to study the ice edge processes, while SIZEX 92 in the Barents Sea (Johannessen et al. 1993) focused on validation of ERS1 SAR, also with Ola M. Johannessen as the leader. MIZ studies were also carried out in the Bering Strait and the Western part of the Arctic Ocean, but in the following it is selected to focus only on mesoscale processes in the MIZ in the Fram Strait, the Greenland and Barents Seas such as ocean fronts, ice edge-ocean eddies, ice edge jets and vortex pairs, ice edge

60

O. M. Johannessen et al.

upwelling and internal waves. A brief comment about the future of the MIZ under a global warming scenario will end this section.

3.1.3

Ocean Fronts

In the Barents and Greenland Seas ocean polar fronts are topographical controlled (Johannessen and Foster 1978; Johannessen 1986). In these ocean frontal regions Polar and Atlantic water masses converge causing strong horizontal temperature, salinity and density gradients including strong horizontal and vertical current shear. When the wind is blowing the ice across these sharp temperature ocean fronts into warmer water, the ice will melt and therefore these fronts are limiting boundaries for the ice edge. The Polar Ocean front in the MIZ in the Fram Strait and Greenland Sea is meandering strongly, Fig. 3.2, which is the region where eddies, jets and vortex pairs are formed where the cold Polar Water (blue colour) is converging with the warm Atlantic Water (red and yellow) coming from the south and recirculate southwards in the western part of the Fram Strait creating the Polar Ocean front in the Greenland Sea. The vertical section across the Polar Front in the Greenland Sea, Fig. 3.3, shows that the temperature change in the surface layer is in the order of 2 degree over a distance of few kilometres, Fig. 3.3a, including changes of 0.4 psu in salinity, Fig. 3.3b and 2.0 sigma-t in the density, Fig. 3.3c. The calculated geostrophic current, Fig. 3.3d, gave a south flowing current of 20 cm/s in the Polar Water and a north flowing weak current of 2 cm/s in the Atlantic Water, resulting in a significant horizontal and vertical shear causing baroclinic and barotropic instabilities generating eddies, jets and vortex pairs as seen in Fig. 3.2. However, in other parts of this frontal region the current and shear can be much stronger since the East Greenland Current can reach more than 50 cm/s (Johannessen et al. 1994). These changes across this ocean front also cause strong gradient in the speed of sound in the water masses which has a large effect on acoustic propagation, important for naval applications (Melberg et al. 1991). The Polar Ocean front in the Barents Sea surrounding the Bear Island stretching up to Svalbard and into the Barents Sea has even stronger temperature gradient of up to 3–5
C over 1–2 km caused by the inflow of the warm Atlantic Water into the Barents Sea, converging with the cold Polar Water. This Polar Ocean front is topographical controlled by the bathymetry and is highly correlated with the 100 m depth contour, both during winter and summer, in particular south of Bear Island (Johannessen and Foster 1978). During winter time when the ice cover often approaches this front, and if blown across this Polar front into the warm Atlantic water it will rapidly melt. Therefore this Polar front is the limiting boundary for the ice cover in the Barents Sea. During the warm season of the year the ice is melting

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

61

Fig. 3.2 NOAA AVHRR image from July I 1984 during the MIZEX 84 experiment. The image combined visual (channel 2) data over ice with IR data (channel 4) in open water. The dark blue indicates cold water along the ice edge, while light blue, red and yellow shows warmer water (> 2
C). The numbers indicate ice tongues similar to the SAR observations in MIZEX 87. (From Johannessen et al. 1994)

and moving northwards in the Barents Sea resulting in large seasonal variations of the MIZ, see Fig. 3.1. In this frontal region generation of ice edge jets, vortex pairs and eddies also take place along the ice edge (Johannessen et al. 1992) and when propagating out into the warm Atlantic Water during the winter time the ice will rapidly melt, again indicating the ice edge ocean eddies are important “melters” of the ice edge.

62

O. M. Johannessen et al.

(a)

(b)

125 250

0

Distance (km)

20 Distance (km)

40

500

40

375

20

500

0

Pressure (dbar)

375

40

500

500

20 Distance (km)

0

250

375

375 0

Pressure (dbar)

250

Pressure (dbar)

250

Pressure (dbar)

40

125

125

125

20 Distance (km)

0

0

0

0

(d)

(c)

Fig. 3.3 CTD vertical section down to 500 m across the Polar front at 78.75
N obtained on 31 March 1987 from “I/B Polarcircle,” showing (a) temperature in degree C, (b) salinity in psu, (c) density in sigma-t and (d) geostrophic speed on cm/s. (From Sandven et al. 1988)

3.1.4

Ice Edge-Ocean Eddies, Jets and Vortex Pairs

Ice edge-ocean eddies were investigated in the NORSEX 79 experiment (Johannessen et al. 1983) where for the first time two SAR images were obtained in the MIZ north of Svalbard by the JPL SAR flown by the NASA CV 990 on 19 September and 1 of October (NORSEX Group 1983). The SAR image from 1 of October, Fig. 3.4, during calm wind condition showed a meandering ice edge with a scale of 20–40 km where three cyclonic ice edge eddies, E1-E3, extended out from the ice edge with a scale of 5–15 km. The Eddy E4 is located off the ice edge in the open ocean and may have propagated out from the ice edge. Simultaneously, in situ temperature and salinity observations were obtained on 1 of October by the icebreaker “I/B Polarsyssel” confirming the oceanographic structure of these eddies both in the surface and subsurface layers. Geostrophic current calculation and drifting Argos buoys showed that the orbital velocities of theses eddies varied between 5 and 10 cm/s while they propagated westwards with the mean current of 10–20 cm/s which is about 10–20 km/day. Comparison of these eddy data set with instability theories indicated that these eddies were generated by barotropic instability mechanism caused by an ice-ocean jet along the ice edge creating strong horizontal shear in the current. During the pilot Summer MIZEX 83 one of the focuses was to study the eddy system over the Molloy Depth, a bathymetric depression approximately in the centre of the Fram Strait, reaching down to 5500 m with a horizontal scale of about 100 km. Previous investigations using a limited data set (Vinje 1977; Wadhams and Squire 1983) had indicated a 60 km eddy over this depression, which is located in the ice edge region. Our hypothesis based on preliminary modelling in 1982 before the

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

63

Fig. 3.4 Jet Propulsion Laboratory synthetic aperture radar (1.215 GHz) image of the marginal ice zone on October 1, 1979. (From NORSEX Group 1983)

MIZEX 83, later published by Smith et al. (1984), indicated that this eddy, named Molloy Depth Eddy, was trapped over this deep depression, which is shown in the combined visual and infrared satellite image, Fig. 3.5 (Johannessen et al. 1984). Here the surface temperature is superimposed on the bathymetry, the yellow is the Atlantic water of 5
C, while the blue and dark blue is the polar water of 0–1
C. The ice edge, green colour, is diffuse due to low wind speed. At the boundary of the 60 km cyclonic Molloy Depth Eddy smaller eddies with scales of 10-20 km are generated due to baroclinic instabilities. The image also shows jets and a vortex pairs,dark blue, extending out from the ice edge in the centre of the large Molloy Depth Eddy and also to the north of the Molloy Depth Eddy. Other images show that some of these smaller eddies propagated southwards along the ice edge where they interacted with the ice edge (Johannessen et al. 1987a). In situ deep oceanographic sections in a star pattern verified this cyclonic Molloy Depth Eddy to large depth (Johannessen et al. 1984) and thereby supporting our hypothesis that the Molloy Depth Eddy was topographical trapped by the deep bathymetric depression to more than 5500 m.. The ice-edge-ocean eddy study was expanded in Summer MIZEX 84 (Johannessen 1987) with an integrated investigation all along the ice edge from north of Svalbard southwards into the Greenland Sea. Real time analysis from several remote sensing aircraft and satellite systems were used to steer the different ships and icebreakers to the ice-edge ocean eddies, where the in situ observations

64

O. M. Johannessen et al.

Fig. 3.5 Combined IR/Visual TIROS-II image from the Fram Strait on 11 July 1983. Bathymetric contours for every 1000 m are given by white lines. (From Johannessen et al. 1984)

took place. The ice-edge-ocean eddies were abundant along the ice edge as shown in Fig. 3.6, which is a visual and an infrared satellite image (Johannessen et al. 1987a). There are several remarkable features to notice from Fig. 3.6, showing that five eddies were present with scales of 30–40 km, separated with about 50 km along the ice edge. The ice floes are converging in the centre of these eddies indicating that the eddies are not in geostrophic balance. Perhaps most important is that the warm Atlantic water in the core of this cyclonic eddies, which penetrate into the ice edge and sweeping the ice out from the ice edge out into the warm Atlantic water where it melts. Ablation measurement from the core of the ice floes in the centre of one of these eddies gave an underside melt rate of 30–40 cm/day while similar measurements inside the ice edge only gave values of 2–3 cm/day. With the eddy spacing of 50 km and assuming that half of the eddies were ice covered with ice thickness of initially 1.5 m, this caused the ice edge to melt with 1–2 km/day (Johannessen et al. 1987a), thus a very important process to melt the ice cover, not yet included in either regional models or parameterized in climate models. Our integrated data set show that the spin-up time of these ice ocean eddies were about 3 days, the orbital velocity was 30–40 cm/s, southwards advection of 10–15 km/day and life time of at least 20 days. In situ oceanographic measurements showed that these eddies extended to 800–1000 m depth. Comparing with theory shows that both barotropic and baroclinic were the instability mechanisms including the topographically trapped eddy over the Molly Depth. All together 14 ice-edge ocean eddies were studied in detailed during the MIZEX 83–84 in the region between 78
and 81
N, most of them with a cyclonic rotation (Johannessen et al. 1987b and references therein). During the Winter MIZEX 87 ice edge ocean eddies and ice tongues ending in vortex pairs with typical 20–40 km scale and life time of several days were studies in the East Greenland Sea between 76
and 79
north during the period March 28 to April 8. A unique set of airborne SAR images were observed each day with two

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

65

Fig. 3.6 NOAA-7 AVHRR images obtained on 4 July 1984. The left image is from the visible band, and the right image is simultaneously obtained from the infrared (IR) band. The resolution of both images is 1 km. Five eddies (numbered 1 through 5) are clearly observed in the IR image and three eddies can be distinguished in the visible band. In the IR image yellow is the warmest temperature (4
C), while red, light blue, and black represent decreasing temperatures. (From Johannessen et al. 1987a)

STAR aircrafts with the INTERA X-band SAR with 12 m resolution which was optimized for ice imaging and therefore not giving any information from the open ocean. The SAR images were downlinked to the icebreaker “I/B Polar Circle” in real time (Shuchman et al. 1988), allowing us to steer the icebreaker “I/B Polar Circle” and the open ocean Research Vessel “R/V Håkon Mosby” to eddies, ice jets and vortex pairs to collect oceanographic observations and also allowing us to drop drifting Argos buoys on the ice and in the open ocean in the area of study. The open ocean Argos buoys were equipped with sails below to give representative observations of the surface layer current. When the wind speed is moderate, the mesoscale movement of the ice edge is mirroring the ocean circulation, which is demonstrated in Fig. 3.7 (Sandven and Johannessen 1989), where the daily time series of SAR images show several major events. This integrated data set, including modelling were analyzed in a comprehensive paper by Johannessen et al. (1994 and references therein), which is briefly summarized below. The first event started on 28 of March, Fig. 3.7 when two ice edge jets propagated eastward from the major ice edge, a northerly cyclonic one which ended in a vortex pair and later developed into an eddy with scale of 20 km and a large anticyclonic eddy with a 40 km scale, Fig. 3.7 a and b. After these 2 days the wind increased to

66

O. M. Johannessen et al.

(a)

(b) 2W

0W

(c)

2W

(d)

6W

2E

(e) 2W

2W

(f)

6W

(g) 2W

(h) 6W

4W

80 N

79 N 79 N

78 N

78 N 77 N 28/29 March 2100-0130 UT

28 March 0230-0430 UT

(i) 80 N

(j) 6W

2W

31 March 0130-1700 UT

30 March 0830-1230 UT

(k) 10 W

6W

79 N

(l) 10 W

6W

6W

(m) 2W

03 April 1700-2200 UT

02 April 1030-1400 UT

01 April 1630-1900 UT

31 Mar/01 Apr 2200-0200

(n) 6W

6W

2W

79 N 78 N

78 N 77 N

77 N 76 N 04 April 1800-2200 UT

05 April 0900-1300 UT

06 April 1230-1530 UT

05 April 1800-2100 UT

07 April 1100-2030 UT

08 April 1400-1730 UT

Fig. 3.7 Time series of SAR images. Low backscatter from open water and new-frozen ice, high backscatter from ice. (From Sandven and Johannessen 1989)

(b)

(a) 5 m/s

10 km

Fig. 3.8 (a) A zoomed SAR image with resolution of 12 m obtained on March 31 at 78
000 N where a tongue of ice is evolving. The bold white arrow indicates drift of buoy 5068. (b) SAR image from the same area obtained 1 day later (April 1). The second image shows how the tongue is stretched and the development of a vortex-pair. (From Johannessen et al. 1994)

more than 10 m/s from northeast erasing the signatures of these eddies. However, several oceanographic sections show their existence for several days (Johannessen et al. 1994). In the second event a strong off ice jet started to grow south of the anticyclonic eddy on 31 of March, Fig. 3.7e, f, extending 40 km eastward out from the ice edge ending into a vortex pair, Fig. 3.8a, b, which were reproduced in a

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

67

modelling study (Johannessen et al. 1994). The third event was the development of a major anticyclonic eddy on 3 and 4 of April, Fig. 3.7h, which on 5 of April, Fig. 3.7j, grew into a vortex pair. In situ data from drifting Argos buoys and oceanographic observations showed that this anticyclonic eddy extended to depth of more than 1500 m. After 5 of April the ice edge region became very complex with several eddies, jets and vortex pairs interacting with each other. As seen from this unique time series, daily SAR images are needed in order to study the mesoscale variability of the ice edge region. The intense variability is confined to a 50–100 km wide belt along the ice edge, which is also located in the region of the Greenland Polar Ocean front. In a modelling study of the ice edge-ocean region it is shown that the jets, vortex pairs and the eddies observed can be explained by baroclinic and barotropic instabilities (Johannessen et al. 1994). In summary, the ice edge during the winter MIZEX 87 was located in approximately the same region as during the summer MIZEX 84 at least for this high latitude, probably controlled by the position of the Greenland Polar Ocean front between the south flowing Polar Water and the Modified Atlantic Water which is primarily topographically trapped by the shelf break in the Greenland Sea. Along this Polar Ocean front jets, eddies and vortex pairs are generated caused by barotropic and baroclinic instabilities (Fig. 3.3). These eddies have a life time from few days to 20–30 days and they are advected southwards with the mean current of 15–20 km per day. The orbital speed is varying from 10 to 40 cm/s. The ice edge jets and ice edge eddies are transporting the ice out into the warmer Modified Atlantic Water where it melts, and thereby causing loss of the ice from the main ice cover, however this loss is replaced with new ice advected from the Arctic Ocean by the East Greenland Current. A modelling study indicated that ice edge eddies along the 2000 km ice edge in the Greenland Sea could cause a loss of the ice cover of 300 km3 per year, which is approximately 12% of the annual transport of 2600 km3 per year past 79
north in the Fram Strait (Johannessen et al. 1994). Since there exist a strong temperature gradient in these eddies this will significantly also causing strong speed of sound gradient resulting in complex acoustic propagation (Melberg et al. 1987). Furthermore, the horizontal shear in these ice edge eddies is causing strong interaction between the ice floes generating a high level of ambient noise even during calm wind conditions (Johannessen et al. 2003). Both these two phenomena are important for naval applications.

3.1.5

Ice Edge Upwelling

Wind-driven coastal upwelling is a well-known phenomena but it was also hypothesed to take place along the long stretches of the ice edges in the Polar Oceans in a theoretical study by Gammelsrød et al. (1975). Based on their hypothesis two field experiments were carried out along the ice edge north of Svalbard in the Arctic Ocean, the first on December 1977 and the second one during the NORSEX

68

O. M. Johannessen et al.

Fig. 3.9 Density structure in sigma-t from CTD sections I and III, and wind speed during the first major easterly wind event, where CTD sections I and III are indicated. Stations indicated at the bottom of the sections and the distance in km on the top. (From Johannessen et al. 1983)

Experiment in 1979 (Buckley et al. 1979; Johannessen et al. 1983), both with Ola M. Johannessen as the leader. These two field experiments verified for the first time that wind-driven ice edge upwelling actually take place along the ice edge but also showed several differences from the simple analytical theory by Gammelsrød et al. (1975) where the ocean was unstratified with a non-moving ice cover, however improved in a more realistic model study including a moving ice cover by Røed and O’Brien (1983). Still later an improved comprehensive analytical study was carried out by Fennel and Johannessen (1998 and references therein), using the observational results from Johannessen et al. (1983) as input to the model, which included a moving ice edge and a stratified water column of Polar Water in the surface above the Atlantic Water. In the following we will briefly summarize these investigations. Buckley et al. (1979) based on oceanographical sections perpendicular to the ice edge established for the first time that wind-driven upwelling take place along the ice edge caused by wind blowing along the ice edge with the ice edge to the right of the wind direction. Since the drag coefficient is larger over the rough ice (Smith et al. 1970) compared to over water, the wind stress is larger over the ice cover which result in an Ekman transport to the right of the wind direction in the ice edge region causing divergence and upwelling along the ice edge. The study was expanded in the NORSEX 79 Experiment (Johannessen et al. 1983), where an integrated data set was observed including airborne SAR imaging, drifting Argos buoys on the ice with current meters suspended below including oceanographical and meteorological observations. An example of the wind-driven ice edge upwelling is shown in Fig. 3.9 under an easterly wind situation where the wind speed increased to more than 10 m/s resulting in an uplift of the isopycnals of 6–7 m including a northerly displacement of the ice edge of 15 km. This integrated data set qualitatively agreed well with the wind-driven theory by Røed and O’Brien (1983). A comprehensive analytical model (Fennel and Johannessen 1998) using different wind stresses over the ice cover with a moving ice edge and stratified ocean was also in general agreement with the observations by Johannessen et al. (1983). Wind-driven upwelling along the long stretches of ice edges in the Polar Oceans are very important for the upward fluxes of nutrients which in turn control the dynamics of the marine ecosystem particular during the long season of intense

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

69

light. Furthermore, during the windy winter time in Polar regions wind-driven ice edge upwelling is lifting the heavier water up to the surface exposing it to strong cooling resulting in deep water formation. Wind-driven ice edge upwelling is therefore important both during the summer and winter seasons.

3.1.6

Internal Waves in the MIZ

Internal waves are a common phenomena in coastal and ocean regions, and they are very important for mixing and turbulence, in particular when they break. In the Arctic Ocean they were first observed during Fridtjof Nansen’s Fram Drift during 1893–1896 (Nansen 1902), however, these observations were preceded by the first theory of internal waves which was published by Stokes in 1847, (Munk 1981). Since the Nansen observations several investigations have been carried out in the Arctic Ocean, see e.g. the review by Morison (1986). One major difference between the internal wave spectra for the ice-covered Arctic Ocean and the open ocean is that the energy level for the Arctic Ocean is much less when compared with the open ocean, which so far has not been explained. However, no internal wave experiment in the MIZ close to the edge was reported before the one carried out during MIZEX 83 (Sandven and Johannessen 1987), north of Svalbard in the MIZ of the Arctic Ocean. In this experiment three thermistor chains were deployed from drifting ice floes between 15 and 65 m, each 5 m with 2 min sampling in the thermocline region in a triangle with sides of 500–800 m during a 7 day period in June- July 1983. The thermistor chains were deployed by the “I/B Polarbjørn” which also determined the position of the array by acoustic sensors with an accuracy of a few meters. Oceanographic and meteorological observations including current measurements were also carried out from the “R/V Polarbjørn”, Fig. 3.10. The internal wave observations, Fig. 3.11, show a broad band of frequencies dominated by a wave train with frequencies of 2–3 cycles per hour (cph) with amplitudes of 5–6 m, starting at 11:00 on 3 July. The wave train arrived at different times at the different thermistor chains giving a group speed of 5 cm/s in a westsouthwest direction. Further analysis combined with internal wave theory gave the result that the phase velocity was 10–15 cm/s with wavelengths of 100–200 m. The estimated wavelengths and phase velocities agreed with the dispersion relation for the first-mode internal waves while the generation mechanism for these internal waves was uncertain but could be the interaction between the tidal flow and the shallower Yermark Plateau, Fig. 3.10. The energy spectra for the 7 day observational period was compared with the Garett-Munk spectrum (Munk 1981) and was lower by a factor of 2–3 compared with the open ocean, but higher than the energy level in the Arctic Ocean, thus the MIZ can be seen as a transition zone between the ice covered Arctic Ocean and the open ocean with reference to the energy level of internal waves. However, the reason for this is not clear and needs further study, both from a theoretical and an experimental point of view.

70

O. M. Johannessen et al. 81o40’N 900

81o30’N 500 1000

900 700

1100

81o20’N 900 900

81o10’N 800

900

700 900 1000

1100

81o00’N 5o E

6o E

7o E

8o E

9o E

Fig. 3.10 Chart of the area on the Yermak Plateau north of Svalbard where the internal wave experiment took place. The heavy line is the trajectory of the Polarbjørn with date marks in June and July. The arrows indicate the mean current at 40 m measured by an Aanderaa current meter. When the tidal components are removed the highest mean velocity (0.15 m/s) is observed at the slope of the seamount. The depth contours are in meters. (From Sandven and Johannessen 1987)

During the MIZEX 84 a Side Looking Airborne Radar image in the ice edge region indicated that internal waves with wavelengths of 1–2 km could be generated by an ice-ocean eddy, hypothesised by Johannessen et al. (1986). In order to test this hypothesis an internal wave experiment was carried out during the winter MIZEX 87 Experiment along the ice edge in the Greenland Sea, a region abundant with ice ocean eddies (Johannessen et al. 1987a). The experiment again included a comprehensive integrated observational scheme with an array of three thermistor chains deployed on ice floes with Argos positioning buoys in a triangle with few km scale, a string of current meters, daily airborne SAR images by two airplanes (see Fig. 3.7) and in situ oceanographical and meteorological observations from two ships, the “I/B Polarcircle” and “R/V Håkon Mosby” during 28 March to 8 April (Johannessen et al. 1994, 2018). The internal wave observations showed two major events detected by the drifting array, the first one occurring before midnight on 30 March with a dominant frequency of 3 cph and the second on 31 March in the morning with a dominant frequency of 1 cph, with amplitude 3–15 m, marked A and B in Fig. 3.12a. The ice drifting array passed a large anticyclonic eddy which was present to the east of the array, Fig. 3.12c, d, but due to strong north-easterly wind, the eddy signature was erased on the 31 March, but still present in the ocean since it was observed in an oceanographical section, Fig. 3.12e.

3 Marginal Ice Zone and Ice-Air-Ocean Interactions July 3

71 July 4

T 46

T 47

T 518

Fig. 3.11 Contour plot of temperature as a function of time and depth for the three thermistor chains. A period of 24 hours is selected (from July 3, 1100 UTC to July 4, 11 UTC) when most of the thermocline was displaced above 65 m. The isotherms are plotted with 0.10
C intervals. The top layer contains cold water of 1.7
C, and the temperature increases with depth. The arrows indicate the passage of the wave train. (From Sandven and Johannessen, 1987)

The signature of the internal waves was present in the SAR image on 30 of March, shown in detail in Fig. 3.13a, as slicks of ice band with wave length of 0.75–1 km. The intrinsic group velocity which is the same as the phase velocity for these nondispersive waves, the current velocity and the absolute group velocity relative to the bottom are shown in Fig. 3.13b, showing that the internal waves are propagating in a westward direction with a northerly component with approximately 20 cm/s away from the anticyclone eddy with approximately the same wavelength as seen in the SAR image, Fig. 3.13a. When the absolute group velocity vectors are extrapolated backwards they are hitting the anticyclonic eddy and therefore supporting our hypothesis that eddies can be the source of generating internal waves, for the first time shown here. Since eddies are a very common phenomena in all coastal and ocean areas our hypothesis should continue to be investigated both theoretically and experimentally in order to see how important the eddies are as sources of internal waves in the global ocean.

3.1.7

The Future of MIZ

The ice cover in the Arctic Ocean is decreasing both during winter and summer. The first spatial projection that the ice cover will melt during summer time under a

72

O. M. Johannessen et al.

Fig. 3.12 Observations of internal waves and ice-ocean eddies. (a) Isotherms between 1.8 and 1.8
C (shown with an interval of 0.1
C) in the surface layer of the cold Polar Water obtained each 2 min by the triangle (plotted above with 3-h interval) with thermistor chains. The arrival times for the two local wave crests detected during 30–31 March are marked as Case A and Case B. (b–e). Synthetic Aperture Radar (SAR) images from 28 to 31 March. Red dot marks the deployment of the array on 30 March, red arrow indicates the flow of the ice in the anticyclonic eddy; the red box shows the area where the internal waves are observed (see Fig. 3.13). Red dots in E show the array on 31 March at 00:00, 06:00, 12:00 and 15:00. Yellow lines are the East (E)—West (W) oceanographic section obtained on 31 March (Fig. 3.3) and the North (N)—South (S) section obtained on 2 April. Schematically, the E–W geostrophic profile is shown by a curved yellow line and the circular white arrow shows the anticyclonic eddy. (From Johannessen et al. 2018)

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

73

Fig. 3.13 Imagery of internal waves and group velocity calculations. (a) SAR image of internal wave train with the wavelengths of 750–1000 m propagating at the western boundary of the anticyclone eddy (shown for 30 March by the slicks of curving bands of ice floes); for the location !

see red box in Fig. 12D. (b) The triangle array T1-T2-T3 at 00:00 and 06:00 of 31 March, C is the !

g

intrinsic group velocity of the internal wave train, V is the average current velocity in the 0

!

thermocline region, C is the absolute group velocity relative to the ground. (c) SAR image of g,0

31 March. Yellow dots are the array positions at times 00 and 06 shown along with the absolute !

group velocity vectors, C . Inserted is the part of the SAR image from 30 March, which includes the g,0

box shown in Figs. 3.3a and 3.2a. Dotted yellow lines from the array positions are backward extrapolation directions of the absolute group velocities hitting the internal waves on 30 March. (From Johannessen et al. 2018)

doubling of CO2 was published by Johannessen et al. (2004) using two coupled climate models from the Max Planck Meteorological Institute in Hamburg and the Hadley Center in UK. As seen from Fig. 3.14, a dramatic decrease of the summer ice is projected under a doubling of CO2 (Fig. 3.14b, d) with less decrease during winter time (Fig. 3.14a, c), see Chap. 10 and references therein which gives the present condition of the ice projection in this century. Furthermore, based on annual linear correlation between the ice extent and the CO2 for the period 1961–2007 of R ¼ 0.95 and R2 ¼ 0.90, suggesting that 90% of the decreasing ice cover could be explained by the increasing CO2. The projection using the A and B scenarios of IPCC (2007) in the regression equation, indicated that the ice extent will be reduced much faster than the ensemble projection from the IPCC 2007 climate models, for the first time published by Johannessen (2008). However one should have in mind that these studies do not take natural variability into account such as e.g. the Atlantic Multidecadal Oscillation (AMO) with a period

74

O. M. Johannessen et al.

Mar 2001-2010

Sept 2001-2010

a

b

Mar 2081-2090

Sept 2081-2090

c

d

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Fig. 3.14 ECHAM4-modelled Northern Hemisphere sea-ice concentration in late winter (March) from (a) 2001–2010 and (c) and 2081–2090, and in late summer (September) from (b) 2001–2010 and (d) 2081–2090. The model has been run using the IPCC IS92 emission scenario comparable to IPCC SRES scenario B2. (From Johannessen et al. 2004)

of approximately 60-80 years, which strongly influences the Arctic surface air temperature, Johannessen et al. (2016). The AMO peaked around 1940 and again around 2015 before starting to decline, Frajka-Williams et al. (2017). This could mean that the AMO for the next 30–40 years could contribute to cooling of the Arctic with more ice, however in competition with the CO2 warming, difficult to project. However, after this 30–40 years period the AMO will probably increase again and together with the CO2 warming cause a dramatic increase in both Arctic temperature and an acceleration towards an ice free Arctic during the whole summer season, only time will show.

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

75

In a study by Johannessen (2011) the ice extent in September was for the first time also correlated with the ln(CO2/CO2,0) concentration for the period 1901–2010 where CO2,0 is the reference level in 1901.The reason for correlating with ln(CO2/ CO2,0) is that this is how the reflected long wave radiation is affected by the CO2 in the atmosphere (Arrhenius 1896). The correlation was very high with R2 ¼ 0.8, which indicate that approximately 80% of the September ice decline in the last few decades could be explained by the increasing CO2. Using the regression equation between the ice cover and CO2, the CO2 concentration must reach a value of approximately 500 ppmv for an ice-free Arctic Ocean during September to occur. If the current rate of increase of CO2 concentration continue, this will occur approximately around 2050–2060, which is similar result as obtained by the ensemble projection of sea ice decline by IPCC (2013) using the RCP8.5 scenario. A similar calculation for the winter, indicated that the CO2 concentration must reach a level of several thousand ppmv, which fortunately is totally unrealistic. From these studies one can conclude that the length of the MIZ in the Arctic will dramatically decrease and probably be totally absent during summer time after the middle of this century, unless a drastic cut in the global CO2 emission will take place. The winter ice will not disappear, although the length of the MIZ will be less.

3.1.8

Summary

In the late 1970s and into the 1990s mega science integrated atmospheric-ice-ocean experiments were carried out in the MIZ in the Fram Strait, Greenland and Barents Seas with focus to improve the understanding of the mesoscale processes along the ice edge. We have shown that the Polar Ocean Front between the cold Polar Water and the warm Atlantic Water limits the extension of the ice edge, in particular during the winter time, that the ice ocean eddies are important for melting the ice edge, that ice edge upwelling is very important for biological productivity during the light season of the year and for precondition of deep water formation along the ice edge during the cold winter season. Furthermore we show for the first time that an anticyclonic eddy is the source for internal wave generations. Finally we comments on the future of the MIZ which probably will disappear during the summer time if the CO2 emission is not drastically reduced.

3.2

Interaction of Sea-Ice with the Ocean and Atmosphere

Richard Davy and Einar Olason In this section we review the myriad processes which govern the interaction between the atmosphere, ocean, and sea-ice in the Arctic. These coupling processes are important on spatial scales down to a few meters and on timescales of less than an

76

O. M. Johannessen et al.

hour, all the way up to affecting the pan-Arctic climate and climate sensitivity on the timescale of decades. We first review the coupling processes from the perspective of the atmosphere, then from the perspective of the ocean. In each of these sections we discuss both the limitations in our knowledge of the individual processes, and the subsequent effect on our ability to accurately model the Arctic climate. Finally, we discuss how the importance of these different processes may change in the future as we move towards more seasonal ice cover in the Arctic.

3.2.1

The Role of Sea Ice in Coupling the Ocean and Atmosphere

Sea-ice acts as an insulator in the exchange of heat, water, and particles between the ocean and atmosphere. As such it plays a crucial role in governing the thermal and dynamical properties of the Arctic atmosphere and ocean. Thick sea ice can thermally decouple the atmosphere from the ocean. This enables the atmosphere to cool radiatively to temperatures far below that of the ocean underneath. This large temperature difference can lead to very strong fluxes from the ocean when the cold air from over the ice passes over open water, which occurs at the marginal ice zone (MIZ) and over leads and polynyas in the Arctic interior. But even thin sea-ice acts as a good insulator, substantially reducing exchange between the ocean and atmosphere. We will review the numerous small-scale processes, as well as the large-scale dynamics, which act to couple the ocean, sea-ice, and atmosphere. Our understanding of surface coupling has proved a limitation in efforts to simulate the Arctic environment due to a combination of the fact that there are physical phenomena unique to this region which have remained relatively unstudied; and that there are limited in-situ observations. The Arctic is a challenging place to conduct field campaigns. Consequentially there is relatively little in-situ data for testing and constraining physical descriptions of surface coupling, compared to lower latitudes. This is a strong motivation behind recent and planned field campaigns like N-ICE and MOSAIC (https://www.mosaic-expedition.org/). One of the longest running in-situ observation campaigns from over-ice was the SHEBA campaign (Uttal et al. 2002). This campaign shaped a lot of contemporary understanding about surface coupling in the Arctic. In contrast to the limited observations over sea-ice there have been numerous observations in more accessible locations in the Arctic, such as in the surroundings of Svalbard, in Russia, Norway, and Alaska. These long-running observations have provided insight into important aspects of the atmospheric dynamics (Oltmans et al. 2012; Mazzola et al. 2016); surface fluxes (Ferrari et al. 2005; Westermann et al. 2009); and both atmospheric and oceanic processes inside fjords (Svendsen et al. 2002; Cottier et al. 2005). The Arctic remains a challenging region for models of all scales from weather prediction and seasonal forecasting to climate projection and prediction (IPCC 2007, 2013). There are a couple of reasons for this. Firstly, many of the important

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

77

processes in the Arctic system occur at scales far smaller than these models resolve. Therefore, these processes need to be well understood in terms of their physics, their interaction with other processes, and their effect on the resolved scales. And secondly, because our understanding of the physics of surface coupling in the Arctic is not as well-constrained through observation as it has been in other parts of the world. There are numerous small-scale features that are important in the ocean-iceatmosphere coupling: the thickness of the ice; the presence of melt-ponds; soot on the snow and ice; the thermal and dynamic roughness of the ice; the representation of turbulence under stable stratification; the self-organization of turbulence in the convection over leads and polynyas; cloud microphysics – especially in mixedphase clouds; interaction in clouds between the microphysics, radiation and turbulence; the dynamics of the sea-ice; and the effects of snow on the sea-ice (Fig. 3.15).

N

L

K J

A M

I F

C

H

E

O B

D

G P Q

R

Fig. 3.15 A schematic depiction of some of the most important small-scale features and processes which govern interactions between the ocean, sea ice, and atmosphere. (a) large-scale atmospheric advection of heat, momentum, moisture and particulates; (b) turbulence under stable stratification typical of cloud-free, over-ice conditions; (c) melt ponds; (d) snow on ice; (e) aerodynamic roughness of the ice; (f) deformation of ice; (g) formation of new sea-ice in leads and polynyas; (h) the formation of sea-smoke from the large moisture fluxes over leads; (i)– convective turbulence over leads and self-organization into turbulent structures; (j) the formation of an internal boundary layer due to advection; (k) microphysics and radiation absorption within mixed-phase clouds; (l) turbulence within clouds; (m) Precipitation over ice; (n) shortwave radiation and surface/cloud albedo; (o) longwave radiation emitted from the surface and partially absorbed and reemitted from clouds; (p) hydrodynamic roughness of ice; (q) turbulence in the ocean mixed layer; (r) the halocline, a layer of relatively fresh water

78

O. M. Johannessen et al.

3.2.2

Atmospheric Processes

3.2.2.1

Structure of the Atmospheric Boundary Layer

The structure of the atmospheric boundary layer in the Arctic both determines the atmospheric effects on the sea-ice and is strongly governed by the interaction with sea-ice and the ocean. For the 6 months of the year when there is little or no incoming solar radiation there is a net cooling of the atmosphere as the snow/ice surface radiatively cools to space. This surface cooling means the air above the ice becomes stably-stratified. These conditions can persist for long periods allowing the formation of strongly stable -stratified boundary layers. Because these conditions only exist in the high latitudes they are scarcely observed and as such remain poorly understood in terms of the physics, and consequentially present a challenge for models (Holtslag et al. 2013). Turbulence in stable stratification has been a persistent challenge in atmospheric science. Many of the assumptions used in turbulence parameterization, for example that turbulence is isotropic, are no longer valid in strongly-stable stratification because the buoyancy flux greatly suppresses vertical mixing. Classically it was also assumed that there existed a critical Richardson number – the ratio of buoyancy flux to wind shear – above which turbulence could no longer be maintained by shear stress. This cutting-off of turbulence led to a problem of runaway cooling in models as in the absence of turbulent mixing the surface became decoupled from the air and was free to radiatively cool without any thermal exchange with the atmosphere. However, numerous observational works have demonstrated that this does not occur in reality and so a variety of adjustments to turbulence parameterizations were introduced to resolve this issue. The problem with this was that these adjustments did not address the fundamental limitations of the underlying theory of turbulence. However, recently there has been some progress towards understanding and simulating the different regimes of turbulence found under strongly stable, and long-lived stable stratification (Zilitinkevich et al. 2012). Atmospheric models of all scales, from single-column models through weather forecast models to climate models have had persistent problems representing stablystratified boundary layers (Holtslag et al. 2013; Mahrt 2014). These problems have resulted in biases in mean properties and the sensitivity of the surface climate to changes in thermal forcing (Davy and Esau 2016; Pithan and Mauritsen 2014). There have been several model intercomparison programs to address the challenges of modelling the atmospheric boundary layer both specifically related to the Arctic and more generally to the stable boundary layer (Beare et al. 2006).

3.2.2.2

Clouds

Clouds over sea ice – created from fluxes through leads or from advection of warm and moist air – have a strong influence on the thermal structure of the atmosphere

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

79

and the melting of sea ice. The presence of clouds strongly alters the radiative balance throughout the atmospheric column. Clouds reflect incoming shortwave radiation thereby reducing surface heating. But they also increase the trapping of longwave radiation close to the surface which reduces the ability of the surface to radiatively cool. In the presence of clouds we typically have near-neutral stratification in the atmospheric boundary layer and it is relatively warm near the surface. Mixed-phase clouds – clouds containing both liquid and solid water – are commonly found over-ice in the Arctic, and the persistency of mixed-phase clouds is particular to the unique conditions of the Arctic (Morrison et al. 2012). This has contributed to the poor representation of these types of clouds within numerical models, which is a large source of uncertainty for both weather prediction and climate projection in the Arctic. The properties of the sea-ice strongly affect the presence and type of clouds in the Arctic. At the sea ice edge, the strong fluxes from the ocean stimulates the formation of cumulous clouds. These can be responsible for much of the precipitation over the sea-ice and nearby land. While clouds over the sea-ice in the central Arctic are more stratiform, whether they are formed from advected air-masses or from the fluxes of moisture through leads in the ice. The changes in frequency of occurrence of these different types of clouds is seen in regions where sea-ice has been disappearing, such as in the Barents-Kara sea region (Chernokulsky et al. 2017). These clouds can form from the convection over leads. The large temperature difference and ultra-low humidity in the atmosphere can trigger strong latent and sensible heat flux as the air passes from over-ice to over open water. The flux of heat and moisture over leads depends on the temperature difference between the atmosphere and ocean, but also strongly depends on the width of the leads (Esau 2007). Over narrow leads (width of a few hundred meters) the fluxes form a single convective column while over wider leads (more than a kilometre wide) the convection can self-organise into cells. This self-organization of turbulence into structures can greatly enhance surface fluxes compared to what would be expected for the given air-sea temperature differences.

3.2.2.3

Atmospheric Drag

The surface provides drag to the atmosphere and so momentum is transferred from the atmosphere to the sea ice and to the ocean. While the ocean can also contribute, the atmosphere is the principle driver of large-scale motion of sea ice. This has motivated decades of effort to obtain observational estimates of the atmospheric drag from in-situ and satellite measurements (Smith et al. 1970). How much momentum is transferred from the atmosphere depends upon the wind speed, the thermal stratification of the atmospheric boundary layer, and the aerodynamic roughness of the surface. The roughness is determined by the physical features of the surface: the presence of pressure ridges and floe edges in the ice; the heterogeneity in the composition of the ice; and the presence of melt ponds or other open water such as leads. The least drag is found over the smoothly homogenous ice of the Arctic interior, and higher drag over the MIZ and other regions where there has been

80

O. M. Johannessen et al.

significant deformation of the ice. However, surface roughness is difficult to characterise because there is a very high degree of heterogeneity in the Arctic related primarily to the spatial distribution of the deformation of ice. In the MIZ this is further complicated by the interaction with ocean waves as surface stress between the atmosphere and ocean drives the formation of ocean waves which further contribute to the deformation of the ice and increased surface roughness. During storms these ocean waves can be responsible for the break-up of ice hundreds of kilometres from the sea-ice edge (Kohout et al. 2014). While this non-linear relationship between wind speed and surface roughness is relatively well established for open water, the more complex interaction between wind speed, waves, and sea-ice is an area of ongoing research (Williams et al. 2013).

3.2.2.4

Melt Ponds

Melt ponds have become increasingly common in the Arctic summer and have a large impact on both the dynamic and thermodynamic surface roughness. But they also significantly reduce the surface albedo allowing more shortwave radiation to be absorbed, accelerating the melting of ice in the Spring and Summer. This forms a positive feedback as increased melting due to the presence of melt ponds drives the formation of more melt ponds. This is an important feedback on seasonal timescales; indeed, the number and size of melt ponds found in the springtime is a good predictor of the minimum sea-ice extent in September (Schröder et al. 2014). The changing properties of the sea ice, going from multi-year ice with a high surface roughness to younger, thinner ice with a smooth surface and lower roughness means that these two effects of melt-ponds are related (Landy et al. 2015).

3.2.2.5

Ice Rheology

It has been observed that the speed of sea-ice drift has increased in the last few decades (Rampal et al. 2009). Changes in sea-ice drift velocity are due to a combination of changes in the wind speed, ocean currents, and the ice rheology (Rampal et al. 2009; Kwok et al. 2013). As the ice thins it becomes mechanically weaker and more sensitive to atmospheric drag. However, younger ice tends to also be physically smoother and the reduced surface roughness means there is less momentum transfer from the atmosphere to the ice. However, there is a negative feedback on the atmosphere: as the drag imposed on the atmosphere decreases, the wind speed increases, which consequently increases the momentum transfer from the atmosphere to ocean. There have been several modelling studies which have sought to quantify the effect of these different mechanisms, but modelling studies are limited by the question of how to specify the roughness of the ice across different ice types and conditions, both with respect to the atmosphere and the ocean.

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

3.2.2.6

81

Snow Cover

The presence of snow on sea-ice can dramatically alter the surface energy budget. Snow increases the albedo of the surface, reducing the absorption of downwelling shortwave radiation. The rapid warming of the near-surface air in the Arctic in recent decades has increased the precipitation in the region, due in part to the increased availability of, and atmospheric holding-capacity for, local moisture sources (Kopec et al. 2016). When the snow on the sea ice melts it can form melt-ponds on the surface of the ice. This also has a very large impact on the sea ice (see above).

3.2.2.7

Impact on Large-Scale Flow

Changes in the properties of the sea ice – it’s extent, concentration, and thickness – can also affect the large-scale flow in the atmosphere, which in turn affects the sea-ice. Large-scale pressure anomalies in the atmosphere drive atmospheric circulation and can contribute to inter-annual variation and extremes in the sea-ice extent. Changes in atmospheric circulation during the summer have been shown to substantially contribute to the reduction in the September minimum in sea ice extent during the satellite observation era (Ding et al. 2017). Atmospheric circulation patterns, along with polar low activity, were also likely to have made significant contributions to the record minima in sea-ice extent in 2007 and 2012 (Wang et al. 2009; Parkinson and Comiso 2013; Zhang et al. 2013), although attribution of individual records remains uncertain. These persistent pressure anomalies can arise from the removal of sea ice from a given region. This can be a direct effect as the increased fluxes from the ocean reduces static stability in the atmosphere, altering the tendency towards given circulation patterns. Indeed, it has been shown that a reduction in sea ice extent can lead to a more frequent occurrence of a negative Arctic Oscillation pattern (Nakamura et al. 2016). The increased open water in the Arctic ocean also increases moisture availability and consequentially snow cover in the high latitudes of the continental interior of Asia. This further supports the negative phase of the Arctic oscillation. This pattern can enable an increase in the atmospheric transport of heat from lower latitudes that further enhances Arctic warming, creating a positive feedback on Arctic warming.

3.2.3

Ocean Processes

Both large-scale processes and local processes play an important role in governing the interaction between the ocean, sea ice, and atmosphere. These range from largescale ocean heat transport from the Atlantic to the formation and maintenance of the halocline. As the ocean has a heat capacity orders of magnitude larger than the sea

82

O. M. Johannessen et al.

ice, ocean influence on the ice is bound to be large. In particular the advection of warm waters from the south will determine the long term position of the ice edge, while the presence of the halocline insulates the ice from the vast storage of heat found in the Arctic Ocean interior caused by the Atlantic Water in the Arctic Ocean. The advection of oceanic heat to the Arctic and the maintenance of the halocline are thus probably the most important topics in terms of ice-ocean interaction in the Arctic.

3.2.3.1

Ocean Heat Transport

The main transport of oceanic heat into the Arctic comes from the North Atlantic, with one branch entering the Arctic via the Barents Sea and the other entering through the Fram Strait. The Fram Strait branch flows along the Spitsbergen shelf in the West-Spitsbergen current, bringing relatively warm Atlantic waters, along with Arctic intermediate and deep waters from the Nordic Seas to the ocean west and north of Svalbard. Once through the Fram Strait this western branch of Atlantic inflow subducts below the halocline, flowing along the continental shelf north of Spitsbergen and Franz-Josef’s Land, towards the Laptev Sea. The mean total inflow into the Arctic along the West Spitsbergen Current has been estimated to be between 7 and 12 Sv (Rudels et al. 2008; Schauer et al. 2008). The local oceanographic conditions make accurate assessments of the flow difficult (Rudels 2012). Unlike the other passageways mentioned below, the Fram Strait is also a gateway for large amounts of export from the Arctic ocean with estimates of 9 to 14 Sv mean total outflow through the strait (Rudels et al. 2008; Schauer et al. 2008). In addition to liquid water export there is an estimated export of about 2900 km3/year of sea ice through the Fram Strait (Vinje 2001). The other main pathway for export out of the Arctic is through the Canadian Arctic Archipelago. The estimated liquid water export through the Archipelago is 1.6 Sv and an estimated export of ice is 410 km3/year (Curry et al. 2014). The eastern branch of the Atlantic inflow into the Arctic Ocean flows northwards form the Atlantic, along the Norwegian coast into the Barents Sea. The amount of warm Atlantic water entering the Barents Sea in this way has been shown to strongly influence the ice extent in the Barents Sea, driving the “Atlantification” of the Arctic waters (Onarheim et al. 2014; Polyakov et al. 2017). The Atlantic inflow then exits the Barents Sea along the northern coast of Nova Zemlya where it flows below the surface halocline along the continental shelf break and joins the western branch of Atlantic inflow in flowing into the Laptev Sea. The flow of Atlantic water into the Barents Sea via the Barents Sea Opening has been measured since 1995 showing a mean inflow of 1.8 Sv without any significant trend. Interannual variations in the flow are significant, with the 12-month moving average values ranging from 0.8 to 2.9 Sv over the period of 1995–2006 (Skagseth et al. 2008). On the Pacific side of the Arctic, warm water enters through the Bering Strait flowing along the Alaskan coast. This flow brings relatively warm surface waters

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

83

towards Point Barrow, where most of it starts to flow below the halocline and along the shelf break into the Beaufort Sea. At annual mean flow of 0.8 Sv the Pacific water inflow is substantial, despite being much less than the Atlantic water inflow (Woodgate et al. 2005). On the whole, the most important influence ocean heat transport has on the ice is to influence the ice extent. This effect has probably been best quantified in the Barents Sea (Onarheim et al. 2014), while also being clearly visible east and north of Svalbard and in the Beaufort Sea. Once the warm Atlantic and Pacific waters subduct below the thermocline very little heat is transferred from them to the ice. As such, the Arctic halocline acts as an insulating layer between the warm Atlantic Layer and the ice.

3.2.3.2

Halocline

The Arctic halocline is a layer of fresh and cold surface waters in the Arctic which lies between the ice and the more saline Atlantic water masses below (Aagaard et al. 1981). The halocline is demarcated by constant low salinity (between about 32 and 34 psu) reaching depths of about 100 m. Below this there is a sharp transition towards the higher salinities of the Atlantic Layer (about 36 psu). The halocline has historically been divided into the upper and lower haloclines (e.g. Jones and Anderson 1986; Anderson et al. 2013), with the upper halocline being of a Pacific origin and mainly confined to the Canadian basin and the lower halocline of an Atlantic origin and more spread over the Arctic ocean. The lower halocline is primarily maintained by the substantial inflow of fresh water from a number of large rivers discharging into the Arctic. These are most notably the Lena, Ob, and Yenisei rivers in Russia. This inflow acts to replenish the fresh water of the halocline. Ice formation in large coastal and flaw polynyas, primarily in the Laptev Sea also contribute the maintenance of the halocline. Here offshore winds break open the ice at the coast or land-fast ice edge, forming a polynya. In the polynya, new ice forms while the brine enriched waters that remain sink towards the bottom of the ocean contributing to the intermediate and deep-water masses in the Arctic Ocean. The newly formed ice is exported into the central Arctic, where some of it will melt during the summer partially replenishing the fresh halocline layer. The upper halocline is maintained in a similar way as the lower halocline by freshwater inflow from the McKenzie river in Canada. In addition to riverine input, the upper halocline is also maintained by freshwater input from the Pacific Ocean via the Bering Strait inflow. These Pacific waters exhibit a strong seasonal cycle, with fresh and warm summer waters that spread near the surface of the western Arctic and winter waters with similar properties as the halocline waters of Atlantic origin. The Pacific Winter Waters influence the Arctic halocline by injection of cold, saline shelf waters and the cooling and freshening of the Atlantic waters (Aagaard et al. 1981; Aagaard and Carmack 1989).

84

3.2.3.3

O. M. Johannessen et al.

Leads

Leads are a ubiquitous feature of the Arctic ice cover. Their size ranges from a few meters in width to a few kilometres and they can extend for as long as several hundred kilometres. Sea ice is well known to be an excellent insulator, with Arctic sea ice reducing the potential flux of heat from ocean to atmosphere by two orders of magnitude in winter (e.g. Maykut 1986; Andreas et al. 1979). When a lead opens, especially in winter, a cold atmosphere is therefore exposed to the relatively warm ocean. This causes large amounts of heat loss from ocean to atmosphere and rapid ice growth in the lead accompanied with the expulsion of brine into the ocean below. Data from various field campaigns (e.g. Morison and McPhee 1998), as well as the results of numerical models (e.g. Smith et al. 2002) give a very consistent picture of brine release below leads. When the ice velocity is small or moderate salt plumes form below the lead and sink to the bottom of the mixed layer. The plumes cannot penetrate the halocline, but instead spread horizontally along the top of the halocline, reducing the depth of the mixed layer. When the ice velocity is sufficiently large turbulences start forming along the lead edge and distribute the rejected brine throughout the mixed layer. Lead formation therefore acts to maintain the halocline, in general, seeing as brine enriched water is deposited at the base of the halocline but freshwater, in the form of ice, remains at the surface.

3.2.3.4

High-Shear Events

It is well established that shear at the ocean surface drives upwelling through Ekman pumping. Shear is also the main mode of deformation of sea-ice and it has been postulated that shearing of the ice cover may drive upwelling in the ocean below. Observations of this phenomenon remain scarce (McPhee et al. 2005) and estimations of its effects on the ocean-ice system are consequently difficult. The current generation of ice models are also unable to reproduce the large amount of shear observed in the Arctic ice cover, although considerable advances have recently been made in this regard (Rampal et al. 2009).

3.2.4

The Future of Ocean-Ice-Atmosphere Interaction

As the climate warms the Arctic sea-ice extent becomes smaller, and the ice becomes thinner and more seasonal. This will mean more heat input from the ocean into the atmosphere due to reduced insolation, especially in autumn and early winter when the sea-ice extent is most reduced (Comiso et al. 2008; Serreze and Stroeve 2015). Thin first year ice is inherently more fragile than thick multi-year ice so the future ice cover can be expected to fracture more readily than the current, or the pre-industrial one. As a consequence, we can expect processes related to lead formation to be enhanced; resulting in larger ocean-atmosphere heat loss through leads along with

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

85

more ice formation and brine expulsion within the leads. While the latter can be expected to strengthen the halocline, we can expect the increased seasonality of the ice cover to act as a counterweight to this, as brine released from ice forming in the autumn may mix evenly enough into the waters below to erode the underlying halocline. In contrast, several of the processes which are thought to drive or enhance Arctic amplification – the more rapid warming in the Arctic compared to the hemisphericaverage – are expected to become less important as the sea ice is removed. The less frequent occurrence of shallow, stably-stratified boundary layers means the extra heat added at the surface will be spread deeper into the atmosphere, reducing the amount of surface warming per unit of added heat (Davy and Esau 2016). As the summer sea-ice extent becomes reduced and the frequency of melt ponds increases we can also expect the positive feedback associated with sea-ice and melt-pond albedo to be reduced as there is simply a lower sea-ice extent; although the thinner ice tends to favour the formation of melt ponds. And while there will be an increased moisture-bearing capacity of the warmer air, contemporary climate models show that more of this atmospheric water vapour will precipitate as rain due to the rapid warming in the atmospheric boundary layer (Bintanja and Andry 2017).

3.2.5

Summary

We have reviewed the most important processes which govern the interactions between the domains of the atmosphere, sea ice, and ocean in the Arctic. This gives us a clear picture of the challenges we face in trying to understand Arctic climate and climate change – there are a huge range of processes whose features can be determined at very small spatial and temporal scales from the microphysics, turbulence and radiation interactions in mixed phase clouds to the turbulent mixing in the long-lived stable boundary layers occurring during the Arctic winter. This presents a huge challenge for scientists to observe, understand, and model these processes; one which will likely take many more years of in-situ and remote observation, theoretical work, and modelling effort in order to be able to make accurate and robust predictions about the future state of the Arctic on timescales from days (for weather forecasting) to centuries (for climate projections).

References Aagaard, K., & Carmack, E. C. (1989). The role of sea ice and other fresh water in the Arctic circulation. Journal of Geophysical Research: Oceans, 94, 14485–14498. https://doi.org/10. 1029/JC094iC10p14485. Aagaard, K., Coachman, L. K., & Carmack, E. (1981). On the halocline of the Arctic Ocean. Deep Sea Research Part A. Oceanographic Research Papers, 28(6), 529–545.

86

O. M. Johannessen et al.

Anderson, L. G., Andersson, P. S., Björk, G., Peter Jones, E., Jutterström, S., & Wåhlström, I. (2013). Source and formation of the upper halocline of the Arctic Ocean. Journal of Geophysical Research: Oceans, 118, 410–421. https://doi.org/10.1029/2012JC00829. Andreas, E. L., Paulson, C. A., William, R. M., Lindsay, R. W., & Businger, J. A. (1979). The turbulent heat flux from arctic leads. Boundary-Layer Meteorology, 17, 57–91. https://doi.org/ 10.1007/BF00121937. Arrhenius, S. (1896). On the influence of carbonic acid in the air upon the temperature of the ground. Philosophical Magazine and Journal of Science, Series 5, 41(251), 237–276. Beare, R. J., Macvean, M. K., Holtslag, A. A., Cuxart, J., Esau, I., Golaz, J. C., Jimenez, M. A., Khairoutdinov, M., Kosovic, B., Lewellen, D., & Lund, T. S. (2006). An intercomparison of large-eddy simulations of the stable boundary layer. Boundary-Layer Meteorology, 118(2), 247–272. Bintanja, R., & Andry, O. (2017). Towards a rain-dominated Arctic. Nature Climate Change, 7, 263–267. Buckley, J. R., Gammelsrød, T., Johannessen, J. A., Johannessen, O. M., & Røed, L. P. (1979). Upwelling; oceanic structure at the edge of the arctic ice pack in winter. Science, 203(4376), 165–167. Chernokulsky, A. V., Esau, I., Bulygina, O. N., Davy, R., Mokhov, I. I., Outten, S., & Semenov, V. A. (2017). Climatology and interannual variability of cloudiness in the Atlantic Arctic from surface observations since the late nineteenth century. Journal of Climate, 30, 2103–2120. Comiso, J. C., Parkinson, C. L., Gersten, R., & Stock, L. (2008). Accelerated decline in the Arctic Sea ice cover. Geophysical Research Letters, 35(1), L01703. Cottier, F., Tverberg, V., Inall, M., Svendsen, H., Nilsen, F., & Griffiths, C. (2005). Water mass modification in an Arctic fjord through cross-shelf exchange: The seasonal hydrography of Kongsfjorden, Svalbard. Journal of Geophysical Research: Oceans, 110(C12), C12005. Curry, B., Lee, C. M., Petrie, B., Moritz, R. E., & Kwok, R. (2014). Multiyear volume, liquid freshwater, and sea ice transports through Davis Strait, 2004–10. Journal of Physical Oceanography, 44, 1244–1266. https://doi.org/10.1175/JPO-D-13-0177.1. Davy, R., & Esau, I. (2016). Differences in the efficacy of climate forcings explained by variations in atmospheric boundary layer depth. Nature Communications, 7, 11690. Ding, Q., Schweiger, A., L’Heureux, M., Battisti, D. S., Po-Chedley, S., Johnson, N. C., BlanchardWrigglesworth, E., Harnos, K., Zhang, Q., Eastman, R., & Steig, E. J. (2017). Influence of highlatitude atmospheric circulation changes on summertime Arctic sea ice. Nature Climate Change, 7(4), 289. Drinkwater, K. F., Miles, M., Medhaug, I., Otterå, O. D., Kristiansen, T., Sundby, S., & Gao, Y. (2014). The Atlantic multidecadal oscillation: Its manifestations and impacts with special emphasis onthe Atlantic region north of 60
N. Journal of Marine Systems, 133, 117–130. https://doi.org/10.1016/j.jmarsys.2013.11.001. Esau, I. N. (2007). Amplification of turbulent exchange over wide Arctic leads: Large-eddy simulation study. Journal of Geophysical Research: Atmospheres, 112(D8), D08109. Fennel, W., & Johannessen, O. M. (1998). Wind forced oceanic responses near ice edges revisited. Journal of Marine Systems, 14(1–2), 57–79. Ferrari, C. P., Gauchard, P. A., Aspmo, K., Dommergue, A., Magand, O., Bahlmann, E., Nagorski, S., Temme, C., Ebinghaus, R., Steffen, A., & Banic, C. (2005). Snow-to-air exchanges of mercury in an Arctic seasonal snow pack in Ny-Ålesund, Svalbard. Atmospheric Environment, 39(39), 7633–7645. Frajka-Williams, E., Beaulieu, C., & Duchez, A. (2017). Emerging negative Atlantic multidecadal oscillation index in spite of warm subtropics. Scientific Reports, 7(1), 11224. https://doi.org/10. 1038/s41598-017-11046-x. Gammelsrød, T., Mork, M., & Røed, L. P. (1975). Upwelling possibilities at the ice edge, homogeneous model. Marine Science Communications, 1, 115–145. Holtslag, A. A. M., Svensson, G., Baas, P., Basu, S., Beare, B., Beljaars, A. C. M., Bosveld, F. C., Cuxart, J., Lindvall, J., Steeneveld, G. J., & Tjernström, M. (2013). Stable atmospheric boundary layers and diurnal cycles: Challenges for weather and climate models. Bulletin of the American Meteorological Society, 94(11), 1691–1706.

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

87

IPCC: Climate Change 2007. (2007). In S. Solomon, D. Quin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, & H. L. Miller (Eds.), The physical science basis. Contribution of working group I to the fourth assessment report of the intergovernmental panel on climate change. Cambridge: Cambridge University Press. 996 p. IPCC: Climate Change 2013. (2013). In T. F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, & P. M. Midgley (Eds.), The physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. Cambridge., 1535 p: Cambridge University Press. https://doi.org/10. 1017/CBO9781107415324. Johannessen, O. M. (1986). Brief overview of the physical oceanography. In B. G. Hurdle (Ed.), The nordic seas (pp. 104–127). New York: Springer. Johannessen, O. M. (1987). Introduction: Summer marginal ice zone experiments during 1983 and 1984 in the Fram Strait and the Greenland Sea. Journal of Geophysical Research, 92(C7), 6716–6718. https://doi.org/10.1029/JC092iC07p06716. Johannessen, O. M. (2008). Decreasing arctic sea ice mirrors increasing CO2 on decadal time scale. Atmospheric and Oceanic Science Letter (AOSL), 1(1), 51–56. Johannessen O. M. (2011). Arctic climate – Present and future perspectives. Given at the symposium to honour the scientific works of Klaus Hasselmann on his 80 years birth day, 9–10 November 2011, Hamburg, Germany. Can be downloaded from Research Gate. Johannessen, O. M., & Foster, L. A. (1978). A note on the topographically controlled oceanic polar front in the Barents Sea. Journal of Geophysical Research, Ocean Front Issue, 83(C9), Ser. 30. https://doi.org/10.1029/JC083iC09p04567. Johannessen, O. M., Johannessen, J. A., Morison, J., Farrelly, B. A., & Svendsen, E. A. (1983). Oceanographic conditions in the marginal ice zone north of Svalbard in early fall 1979 with emphasis on mesoscale process. Journal of Geophysical Research, 88(C5), 2755–2770. https:// doi.org/10.1029/JC088iC05p02755. Johannessen, O. M., Johannessen, J. A., Farrelly, B. A., Kloster, K., & Shuchman, R. H (1984). Eddy studies during MIZEX 83 by ship and remote sensing observations. In Proceedings of the IGARSS 84 Symposium, ESA SP-215, Strasbourg, 27–30 August 1984. Johannessen, O. M., Johannessen, J. A., Sandven, S., & Davidson, K. L. (1986). Preliminary results of the Marginal Ice Zone Experiment (MIZEX) summer operations. In B. G. Hurdle (Ed.), The Nordic Seas (pp. 665–679). New York: Springer Verlag. Johannessen, O. M., Johannessen, J. A., Svendsen, E., Shuchmann, R. A., Campbell, W. J., & Josberger, E. (1987a). Ice-edge eddies in the Fram Strait marginal ice zone. Science, 236(4800), 427–429. https://doi.org/10.1126/science.236.4800.427. Johannessen, J. A., Johannessen, O. M., Svendsen, E., Shuchman, R., Manley, T., Campbell, W. J., Josberger, E. G., Sandven, S., Gascard, J. C., Olaussen, T., Davidson, K., & Van Leer, J. (1987b). Mesoscale eddies in the Fram Strait marginal ice zone during the 1983 and 1984 Marginal Ice Zone Experiment. Journal of Geophysical Research, 92(C7), 6754–6772. https:// doi.org/10.1029/JC092iC07p06754. Johannessen, O. M., Johannessen, J. A., Sandven, S., Olaussen, T., Geiger, K., Skagseth, O., Kovacs, Z., Karpuz, R., Haugan, P. M., Bittner, K., Samuel, P., Pettersson, L. H., Kloster, K., Johannessen, C. E., Sagen, H., Nytun, K., Svendsen, F., Shuchman, R., Onstott, R., Gaboury, S., Sutherland, L., Gloersen, P., Davidson, K., Helvey, R., Spiel, D., Blake, W., Woehler, K., Skupniewicz, C., Borrmann, S., Campbell, W.J., Josberger, E. G., Danes, F., Grenfell, T., Wensman, M., Horrigmoe, G., Jargensen, K. A., Kristiansen, E., Blankvov, A., Linder, B., O’Hara, S., Lamont, J. A., Bendiksen, B., Engelsen, I., Foche, K., Starke, K., Preller, R., Rottier, P., & Horn, D. (1989). SIZEX 89 Experiment report. NRSC Technical Report no. 23. Johannessen, O. M., Campbell, W. J., Shuchman, R. A., Sandven, S., Gloersen, P., Johannessen, J. A., Josberger, E. G., & Haugan, P. M. (1992). Microwave study programs of air-ice-ocean interactive processes in the seasonal ice zone of the Greenland and Barents seas. In F. Carsey (Ed.), Microwave remote sensing of sea ice (Geophysical Monograph 68) (pp. 261–289). Washington, DC: American Geophysical Union.

88

O. M. Johannessen et al.

Johannessen, O. M, Sandven, S., Campbell, W. J., & Shuchman, R. (1993). Ice studies in the Barents Sea by ERS-1 SAR during SIZEX 92. In Proceedings of the First ERS-1 symposium – space at the service of our environment, Cannes, France, 4–6 November 1992, ESA SP-359 (pp. 277–282). Johannessen, O. M., Sandven, S., Budgell, W. P., Johannessen, J. A., & Suchman, R. (1994). Observation and simulation of ice tongues and vortex-pairs in the marginal ice zone. In O. M. Johannessen, R. Muench, & J. E. Overland (Eds.), The polar oceans and their role in shaping the global environment (The Nansen Centennial Volume, Geophysical Monograph 85) (Vol. 1994, pp. 109–136). Washington, DC: AGU. Johannessen, O. M., Sagen, H., Sandven, S., & Stark, K. V. (2003). Hotspots in ambient noise caused by ice-edge eddies in the Greenland Sea and the Barents seas. IEEE Journal of Oceanic Engineering, 28(2), 212–228. Johannessen, O. M., Bengtsson, L., Miles, M. W., Kuzmina, S. I., Semenov, V. A., Alekseev, G. V., Nagurny, A. P., Zakharov, V. F., Bobylev, L. P., Pettersson, L. H., Hasselmann, K., & Cattle, H. P. (2004). Arctic climate change: Observed and modelled temperature and sea ice variability. Tellus, Dynamic Meteorology and Oceanography. Series A, 56A(4), 328–341. Johannessen, O. M., Alexandrov, V. Y., Frolov, I. Y., Bobylev, L. P., Sandven, S., Pettersson, L. H., Kloster, K., Smirnov, V. G., Mironov, Y. U., & Babich, N. G. (2007). Remote sensing of sea ice in the northern sea route: Studies and applications. Chichester: Springer–Praxis Publishing. 477 p. Johannessen, O. M., Kuzmina, S. I., Bobylev, L. P., & Miles, M. W. (2016). Surface air temperature variability and trends in the Arctic: New amplification assessment and regionalasation. Tellus A, 68, 28234. https://doi.org/10.3402/tellusa.v68.28234. Johannessen, O. M., Sandven, S., Chunchuzov, I. P., & Suchman, R. (2018). Observations of internal waves in the marginal ice zone generated by an anticyclonic eddy: A case study. Tellus A: Dynamic Meteorology & Oceanography, in review. Jones, E. P., & Anderson, L. G. (1986). On the origin of the chemical properties of the Arctic Ocean Halocline. Journal of Geophysical Research, 91, 10759–10767. Kohout, A. L., Williams, M. J. M., Dean, S. M., & Meylan, M. H. (2014). Storm-Induced Sea-ice breakup and the implications for ice extent. Nature, 509(7502), 604. Kopec, B. G., Feng, X., Michel, F. A., & Posmentier, E. S. (2016). Influence of sea ice on Arctic precipitation. PNAS, 113(1), 46–51. Kwok, R., Spreen, G., & Pang, S. (2013). Arctic sea ice circulation and drift speed: Decadal trends and ocean currents. Journal of Geophysical Research: Oceans, 118(5), 2408–2425. Landy, J. C., Ehn, J. K., & Barber, D. G. (2015). Albedo feedback enhanced by smoother Arctic sea ice. Geophysical Research Letters, 42(24), 10714–10720. https://doi.org/10.1002/ 2015GL066712. Mahrt, L. (2014). Stably stratified atmospheric boundary layers. Annual Review of Fluid Mechanics, 46, 23–45. Maykut, G. A. (1986). The surface heat and mass balance. In N. Untersteiner (Ed.), The geophysics of sea ice (NATO ASI Series) (pp. 395–463). New York: Springer. Mazzola, M., Tampieri, F., Viola, A. P., Lanconelli, C., & Choi, T. (2016). Stable boundary layer vertical scales in the Arctic: Observations and analyses at Ny-Ålesund, Svalbard. Quarterly Journal of the Royal Meteorological Society, 142(696), 1250–1258. https://doi.org/10.1002/qj. 2727. McPhee, M., Kwok, R., Robins, R., & Coon, M. (2005). Upwelling of Arctic pycnocline associated with shear motion of sea ice. Geophysical Research Letters, 32, L10616. https://doi.org/10. 1029/2004GL021819. Melberg, L. E., Johannessen, O. M., Connors, D. N., Botseas, G., & Browning, D. (1987). Modeled acoustic propagation through an ice edge Eddy in the East Greenland Sea Marginal Ice Zone. Journal of Geophysical Research, 92(C7), 6857–6868. https://doi.org/10.1029/ JC092iC07p06857. Melberg, L. E., Johannessen, O. M., Connors, D. N., Botseas, G., & Browning, D. (1991). Acoustic propagation in the western Greenland Sea frontal zone. Journal of Acoustical Society America, 89(5), 2144–2156.

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

89

MIZEX Group 86, Johannessen, O. M., Horn, D. A., Augstein, E., Baggeroer, A. B., Burns, B. A., Campbell, W. J., Davidson, K. L., Duckworth, G. L., Dyer, I., Farrelly, B. A., Grenfell, T., Heiberg, A., Hibler, W. D., III, & Johannessen, J. A. (1986). MIZEX East 83/84, the summer marginal ice zone program in the Fram Strait/Greenland Sea. EOS Transactions, American Geophysical Union, 67(23), 513–517. including cover. MIZEX Group 87, Johannessen, O. M., Shuchman, R. A., Johannessen, J. A., Sandven, S., Olaussen, T. I., Petterson, L. H., Horn, D. A., Bell, D. L., Grenfell, T. C., Brightman, R. I., Davidson, K. A., Guest, P. S., Gow, A. J., Meincke, J., Perovich, D. K., Quadfasel, D., Smith, S. L., Smith, W. O., Sutherland, L. L., & Tucker, W. B. (1989). MIZEX East 1987: Winter marginal ice zone program in the Fram Strait and Greenland Sea. EOS Transactions, American Geophysical Union, 70(17), 545–555. https://doi.org/10.1029/89EO00131. Morison, J. (1986). Internal waves in the Arctic Ocean: A review. In N. Untersteiner (Ed.), The geophysics of sea ice (NATO ASI Series (Series B: Physics)). Boston: Springer. https://doi.org/ 10.1007/978-1-4899-5352-0_20. Morison, J. H., & McPhee, M. G. (1998). Lead convection measured with an autonomous underwater vehicle. Journal of Geophysical Research: Oceans, 103, 3257–3281. https://doi. org/10.1029/97JC02264. Morrison, H., De Boer, G., Feingold, G., Harrington, J., Shupe, M. D., & Sulia, K. (2012). Resilience of persistent Arctic mixed-phase clouds. Nature Geoscience, 5(1), 11. Munk, W. (1981). Internal waves and small scale processes. In B. A. Warren & C. Wunsch (Eds.), Evolution of physical oceanography scientific surveys in honor of Henry Stommel (pp. 264–291). Cambridge, MA: The MIT Press. Nakamura, T., Yamazaki, K., Honda, M., Ukita, J., Jaiser, R., Handorf, D., & Dethloff, K. (2016). On the atmospheric response experiment to a Blue Arctic Ocean. Geophysical Research Letters, 43(19). Nansen, F. (1902). The Norwegian polar expedition 1893–1896. Scientific Results, 3, 346–351. NORSEX Group, Johannessen, O. M., Campnell, W. J., Farrelly, B. A., Johannessen, J. A., Johannessen, O. M., Svendsen, E., Kloster, K., Horjen, I., Mätzler, C., Campbell, W. J., Crawford, J., Harrington, R., Jones, L., Swift, C., Delnore, V. E., Cavalieri, D., Gloersen, P., Hsiao, S. V., Shemdin, O. H., Thomson, T. W., & Ramseier, R. O. (1983). Norwegian remote sensing experiment in a marginal ice zone. Science, 220(4599), 781. Oltmans, S. J., Johnson, B. J., & Harris, J. M. (2012). Springtime boundary layer ozone depletion at Barrow, Alaska: Meteorological influence, year-to-year variation, and long-term change. Journal of Geophysical Research: Atmospheres, 117(D1 4), D00R18. Onarheim, I. H., Smedsrud, L. H., Ingvaldsen, R. B., & Nilsen, F. (2014). Loss of sea ice during winter north of Svalbard. Tellus A: Dynamic Meteorology and Oceanography, 66(1), 23933. https://doi.org/10.3402/tellusa.v66.23933. Parkinson, C. L., & Comiso, J. C. (2013). On the 2012 record low Arctic sea ice cover: Combined impact of preconditioning and an August storm. Geophysical Research Letters, 40(7), 1356–1361. Pithan, F., & Mauritsen, T. (2014). Arctic amplification dominated by temperature feedbacks in contemporary climate models. Nature Geoscience, 7(3), 181. Polyakov, I. V., Pnyushkov, A. V., Alkire, M. B., Ashik, I. M., Baumann, T. M., Carmack, E. C., Goszczko, I., Guthrie, J., Ivanov, V. V., Kanzow, T., Krishfield, R., Kwok, R., Sundfjord, A., Morison, J., Rember, R., & Krishfield, R. (2017). Greater role for Atlantic inflows on sea-ice loss in the Eurasian Basin of the Arctic Ocean. Science, 356(6335), 285–291. Rampal, P., Weiss, J., & Marsan, D. (2009). Positive trend in the mean speed and deformation rate of Arctic Sea ice, 1979–2007. Journal of Geophysical Research: Oceans, 114(C5), C05013. Roed, L. P., & O’Brien, J. J. (1983). A coupled ice-ocean model of upwelling in the marginal ice zone. Journal of Geophysical Research, 88(C5), 2863–2872. Rudels, B. (2012). Arctic Ocean circulation and variability-advection and external forcing encounter constraints and local processes. Ocean Science. Rudels, B., Marnela, M., & Eriksson, P. (2008). Constraints on estimating mass, heat and freshwater transports in the Arctic Ocean: An exercise. In Arctic-subarctic ocean fluxes (pp. 315–341). Dordrecht: Springer.

90

O. M. Johannessen et al.

Sandven, S., & Johannessen, O. M. (1987). High-frequency internal wave observations in the marginal ice zone. Journal of Geophysical Research, 92, 6911–6920. Sandven, S., & Johannessen, O. M (1989). Ice edge motion during MIZEX 87. In Proceedings of IGARSS ‘89, Vancouver, Canada, July 10 – 14, 1989, Vol. 3, 1535-1538, IEEE Conference Publications 1989. Sandven, S., Kovacs, Z., Geiger, C., Olaussen, T. I., & Johannessen, O. M. (1988). MIZEX’87: CTD Data Report from Polar Circle. NRSC Technical Report no. 13. Sandven, S., Johannessen, O. M., Miles, M. W., Pettersson, L. H., & Kloster, K. (1999). Barents Sea seasonal ice zone features and processes from ERS 1 synthetic aperture radar: SIZEX 1992. Journal of Geophysical Research, 104(C7), 15843–15857. Schauer, U., Beszczynska-Möller, A., Walczowski, W., Fahrbach, E., Piechura, J., & Hansen, E. (2008). Variation of measured heat flow through the Fram Strait between 1997 and 2006. In Arctic-subarctic ocean fluxes (pp. 65–85). Dordrecht: Springer. Schröder, D., Feltham, D. L., Flocco, D., & Tsamados, M. (2014). September Arctic Sea-ice minimum predicted by spring melt-pond fraction. Nature Climate Change, 4(5), 353. Serreze, M. C., & Stroeve, J. (2015). Arctic sea ice trends, variability and implications for seasonal ice forecasting. Philosophical Transactions of the Royal Society A, 373(2045), 20140159. Shuchman, R. A., Sutherland, L. L., & Johannessen, O. M. (1988). Geophysical information on the winter marginal ice zone obtained from SAR. In Proceedings of IGARSS’88 symposium, international geoscience and remote sensing symposium (pp. 1111–1114). Edinburgh, 1988. Remote Sensing: Moving Toward the 21st Century 2. Skagseth, Ø., Furevik, T., Ingvaldsen, R., Loeng, H., Mork, K. A., Orvik, K. A., & Ozhigin, V. (2008). Volume and heat transports to the Arctic Ocean via the Norwegian and Barents seas. In Arctic–Subarctic Ocean Fluxes (pp. 45–64). Dordrecht: Springer. Smith, D. C., IV, Morison, J. H., Johannessen, J. A., & Untersteiner, N. (1984). Topographic generation of an eddy at the edge of the, East Greenland Current. Journal of Geophysical Research, 89(C5), 8205–8208. Smith, S. D., Banke, E. G., & Johannessen, O. M. (1970). Wind stress and turbulence over ice in the Gulf of St. Lawrence. Journal of Geophysical Research, 75(15), 2803–2812. https://doi.org/10. 1029/JC075i015p02803. Smith, D. C., Lavelle, J. W., & Fernando, H. J. S. (2002). Arctic Ocean mixed-layer eddy generation under leads in sea ice. Journal of Geophysical Research: Oceans, 107(C8), 3103, 17-1–17-17. https://doi.org/10.1029/2001JC000822. Svendsen, E., Kloster, K., Farrelly, B., Johannessen, O. M., Johannessen, J. A., Campbell, W. J., Gloersen, P., Cavalieri, D., & Mätzler, C. (1983). Norwegian remote sensing experiment: Evaluation of the Nimbus 7 scanning multichannel microwave radiometer for sea ice research. Journal of Geophysical Research, 88(C5), 2781–2791. https://doi.org/10.1029/ JC088iC05p02781. Svendsen, H., Beszczynska-Møller, A., Hagen, J. O., Lefauconnier, B., Tverberg, V., Gerland, S., & Azzolini, R. (2002). The physical environment of Kongsfjorden–Krossfjorden, an Arctic fjord system in Svalbard. Polar Research, 21(1), 133–166. Uttal, T., Curry, J. A., Mcphee, M. G., Perovich, D. K., Moritz, R. E., Maslanik, J. A., Guest, P. S., Stern, H. L., Moore, J. A., Turenne, R., & Heiberg, A. (2002). Surface heat budget of the Arctic Ocean. Bulletin of the American Meteorological Society, 83(2), 255–275. Vinje, T. E. (1977). Sea ice studies in the Spitzbergen-Greenland area, Landsat Rep. E77-10206, U.S. Dep. of Commer., Natl. Tech. Inf. Serv., Springfield, Va. Vinje, T. (2001). Fram Strait ice fluxes and atmospheric circulation: 1950–2000. Journal of Climate, 14, 3508–3517. Wadhams, P., & Squire, V. A. (1983). An ice-water vortex at the edge of the East Greenland Current. Journal of Geophysical Research, 88(C5), 2770–2781. Wadhams, P., Squire, V. A., Ewing, J. A., & Pascal, R. W. (1986). The effect of the marginal ice zone on the directional wave spectrum of the ocean. Journal of Physical Oceanography, 6(2), 358–376.

3 Marginal Ice Zone and Ice-Air-Ocean Interactions

91

Wadhams, P., Aulicino, G., Parmiggiani, F., Persson, O. G., & Holt, B. (2018). Pancake ice thickness mapping in the Beaufort Sea from wave dispersion observed in SAR imagery. Journal of Geophysical Research, 123(3), 2213. https://doi.org/10.1002/2017JC013003. Wang, J., Zhang, J., Watanabe, E., Ikeda, M., Mizobata, K., Walsh, J. E., Bai, X., & Wu, B. (2009). Is the dipole anomaly a major driver to record lows in Arctic summer sea ice extent? Geophysical Research Letters, 36(5), L05706. Westermann, S., Lüers, J., Langer, M., Piel, K., & Boike, J. (2009). The annual surface energy budget of a high-arctic permafrost site on Svalbard, Norway. The Cryosphere, 3(2), 245. Williams, T. D., Bennetts, L. G., Squire, V. A., Dumont, D., & Bertino, L. (2013). Wave–ice interactions in the marginal ice zone. Part 1: Theoretical foundations. Ocean Modelling, 71, 81–91. Williams, T., Rampal, P., & Bouillon, S. (2017). Wave-ice interactions in the neXtSIM sea-ice model. The Cryosphere, 11(5), 2117. https://doi.org/10.5194/tc-11-2117-2017. Woodgate, R. A., Aagaard, K., & Weingartner, T. J. (2005). Monthly temperature, salinity, and transport variability of the Bering Strait throughflow. Geophysical Research Letters, 32(4), L04601. https://doi.org/10.1029/2004GL021880. Zhang, J., Lindsay, R., Schweiger, A., & Steele, M. (2013). The impact of an intense summer cyclone on 2012 Arctic sea ice retreat. Geophysical Research Letters, 40(4), 720–726. Zilitinkevich, S. S., Elperin, T., Kleeorin, N., Rogachevskii, I., & Esau, I. (2012). A hierarchy of energy- and flux- budget (EFB) turbulence closure models for stably stratified geophysical flows. Boundary-Layer Meteorology, 146, 341–373.

Chapter 4

Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries Elena V. Shalina, Ola M. Johannessen, and Stein Sandven

Observations and Global Coupled Climate Model (GCM) projections suggest that the Arctic is the region where the warming is amplified and sea ice cover is decreasing (Johannessen et al. 2004, 2016; Johannessen 2008; Bekryaev et al. 2010; Kumar et al. 2010; Miller et al. 2010; Screen and Simmonds 2010; Serreze and Barry 2011; IPCC 2014; Onarheim et al. 2018). Arctic sea ice cover has declined over the past few decades with the most dramatic reduction of the ice extent occurring in summer (Comiso et al. 2008; Stroeve et al. 2012; Perovich et al. 2015; arctic-roos.org) including the decrease of sea ice thickness (Rothrock et al. 2008; Kwok and Rothrock 2009). Furthermore, there has been a decrease in the area covered by multiyear ice (Johannessen et al. 1999; Rigor and Wallace 2004; Maslanik et al. 2007; Nghiem et al. 2007; Tschudi et al. 2016a). In this chapter we will review available data sets describing the past and present temporal and spatial distribution of sea ice in the Arctic, available data sources, and recent findings from sea ice studies. The chapter is divided into two sections; the first describes different sources of information about sea ice and the second gives a brief overview of changes in sea ice extent that have been observed since 1900. The data sets include satellite-based sea

E. V. Shalina (*) Nansen International Environmental and Remote Sensing Centre (NIERSC) and Saint Petersburg State University, Saint Petersburg, Russia e-mail: [email protected] O. M. Johannessen Nansen Scientific Society, Bergen, Norway e-mail: ola.johannessen@nansenscientificsociety.no S. Sandven Nansen Environmental and Remote Sensing Center, Bergen, Norway University Centre in Svalbard, Longyearbyen, Svalbard, Norway e-mail: [email protected] © Springer Nature Switzerland AG 2020 O. M. Johannessen et al. (eds.), Sea Ice in the Arctic, Springer Polar Sciences, https://doi.org/10.1007/978-3-030-21301-5_4

93

94

E. V. Shalina et al.

Table 4.1 Data sets described in Chap. 4 Short description of the data set NSIDC and NOAA NCDC passive microwave satellite brightness temperature data NSIDC sea ice concentrations retrieved from passive microwave satellite data, (1) using NASA Team algorithm, (2) using Bootstrap algorithm, 25
25 km grid Arctic ROOS sea ice concentrations retrieved from passive microwave satellite data using NORSEX algorithm, 25
25 km grid EUMETSAT’s OSI SAF sea ice concentrations and sea ice types data set, 10
10 km grid University of Bremen AMSR-E and AMSR2 sea ice concentrations, 6.25
6.25 km grid Chinese FengYun-3-based sea ice concentrations, 12.5
12.5 km grid The multi-sensor and multi-services-based sea ice extent, 1 km and 4 km grid cell sizes The multi-sensor sea ice concentrations, 4
4 km, for model assimilation Synthetic Aperture Radar data sets Scatterometer data sets MODIS sea ice extent and sea ice temperature data set Arctic and Antarctic Research Institute sea ice charts Canadian Ice Service sea ice charts and climatologies National Ice Center Arctic sea ice charts and climatologies Norwegian Meteorological Institute sea ice charts Danish Meteorological Institute sea ice charts Ice chart WMO standards ACSYS historical ice chart archive, 1553–2002 Ice edge positions in the Nordic Seas in summer, 1750–2002 data set Arctic sea ice charts from Danish Meteorological Institute, 1893–1956 Zakharov sea ice data set, 1900–2000 Walsh and Chapman sea ice data sets Hadley Centre global sea ice and sea surface temperature data set NOAA/NSIDC Climate Data Record (CDR) of sea ice concentration EUMETSAT OSI SAF Climate Data Record of sea ice concentration ESA CCI Climate Data Records of sea ice concentration

Section 4.1.1.1 4.1.1.1 4.1.1.1 4.1.1.1 4.1.1.1 4.1.1.1 4.1.1.2 4.1.1.2 4.1.1.3 4.1.1.3 4.1.1.4 4.1.2.1 4.1.2.2 4.1.2.3 4.1.2.4 4.1.2.5 4.1.2.6 4.1.3.1 4.1.3.2 4.1.3.3 4.1.3.4 4.1.3.5 4.1.4.1 4.1.4.2 4.1.4.3 4.1.4.4

ice data sets covering the period from the beginning of the satellite era in the late 1970s to the present (Sect. 4.1.1), ice charts developed by different organizations (Sect. 4.1.2), historical data records that allow us to look further back into the past (Sect. 4.1.3) and climate data records (Sect. 4.1.4). The second Sect. 4.2 deals with variations of sea ice cover observed in the twentieth and twenty-first centuries. It briefly describes sea ice conditions in the beginning of the twentieth century (Sect. 4.2.1), then the event of the early warming and subsequent cooling in the Arctic (Sect. 4.2.2) and finally, the changes in sea ice cover observed over the period covered by satellite data, from 1979 to the present (Sect. 4.2.3). The sea ice data described in the chapter are listed in Table 4.1.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

4.1

95

Data Sets on Arctic Sea Ice

4.1.1

Satellite-Based Sea Ice Data Sets

4.1.1.1

Sea Ice from Satellite Passive Microwave Data

Present knowledge of the temporal and spatial distribution of Arctic sea ice and its changes in the last decades of the twentieth century and in the beginning of the twenty-first century is mostly based on satellite passive microwave data, (see e.g. the Eurogoos/Nansen Center Ice Information System “arctic-roos.org”). The first passive microwave remote sensing systems were launched on the Russian Cosmos 243 and Cosmos 384 satellites in 1968 and 1970, respectively. The Cosmos 243 was designed to measure the earth’s microwave emission and to develop techniques for the determination of the geophysical parameters of the atmosphere and the earth surface. The instrumentation consisted of four radiometric receivers with four wavelengths: 0.8, 1.35, 3.4 and 8.5 cm (Gorbunov and Kutuza 2018). Cosmos 384 continued the studies started by the Cosmos 243 microwave experiment. These missions demonstrated receiving information of geophysical parameters from space (Harvey and Zakutnaya 2011). An important step forward was the US panoramic scanning Electrically Scanning Microwave Radiometer (ESMR) on the Nimbus-5 satellite launched in 1972. The primary objectives of the ESMR instrument were to retrieve liquid water content of clouds from brightness temperatures measured over the oceans, sea ice and open water in the Polar Regions, as well as surface composition and soil moisture (Parkinson et al. 1987). After the ESMR period 1973–1976, a more advanced NASA satellite instrument, the Scanning Multichannel Microwave Radiometer (SMMR) was operated on Nimbus-7 for 9 years, 1978–1987 (Gloersen et al. 1984, 1992). A similar instrument, the Special Sensor Microwave Imager (SSM/I) onboard US DMSP satellites followed the SMMR and has provided continuous measurements for more than 25 years, establishing the first continuous time series for global sea ice research (Johannessen et al. 1995; Cavalieri et al. 1997), followed by many studies. Crucially, this dataset was used by Johannessen et al. (1999) to show that the ice cover was in transformation with less multiyear ice. The successor of SSM/I, the Special Sensor Microwave Imager/Sounder (SSMIS) is an enhanced system that combines and extends the imaging of the three previously separate DMSP microwave sensors: the SSM/T-1 temperature sounder, the SSMI/T-2 moisture sounder and the SSM/I. Passive microwave observations provide a long homogeneous series of data, however with coarse spatial resolution of 25 km and are therefore most suitable for large-scale global monitoring of sea ice and for validation of climate models. A new generation of microwave radiometers was the Advanced Microwave Scanning Radiometer – EOS (AMSR-E), developed jointly by NASA and NASDA, and launched on the EOS Aqua satellite in 2002. The main advantage of the AMSR-E in comparison with the SSM/I was its improved spatial resolution of

96

E. V. Shalina et al.

Table 4.2 Passive microwave measurements: satellites and sensors

Satellite Altitude, km Inclination angle, degrees Ascending equator crossing time (local time) Sensor Frequencies

Earth incidence angle, degrees Swath width, km

DMSP-F8/F11/-F13 856/853/850 98.8

DMSPF16/-F17/F18 848/848/850 98.8

SMMR 6.6, 10.7, 18, 21, 37 GHz

6:00 a.m./ 5:00 p.m./ 5:43 p.m. SSM/I 19.3, 22.3, 36.5, 85.5 GHz

9:32 p.m./ 5:36 p.m./ 8.00 p.m. SSMIS 19.3, 22.3, 36.5, 91.65 GHz

50.2

53.1

780

1400

Nimbus-7 955 99.1 12:00 p.m.

Aqua 705 98

GOM-W1 700 98

1:30 p.m.

1:30 p.m.

53.1

AMSR-E 6.9, 10.7, 18.7, 23.8, 36.5, 89.0 GHz 55

AMSR2 6.9, 7.3, 10.7, 18.7, 23.8, 36.5, 89.0 GHz 55

1700

1445

1445

12 km. Today the instrument AMSR2 (the successor of AMSR-E) provides measurements that enable continuation of sea ice studies with even higher resolution of 6.25 km. ESMR Sea Ice Data Set The Nimbus-5 ESMR sensor provided the earliest global all-weather and all-season satellite-based information of the Arctic sea ice. Data is available from 12 December 1972 to 31 December 1976. The ESMR was a crossscan instrument, with a resolution of approximately 30 km, and a frequency of 19.35 GHz. Some visible and infrared satellite data were available from the TIROS satellite system before the launch of ESMR but since the polar regions are either not illuminated by sunlight or covered by clouds for quite a long period in each year, these data did not make a consistent record, however they are useful support data for special studies because of higher spatial resolution. The data collected by the ESMR instrument demonstrated the efficiency of passive microwave satellite observations for sea ice monitoring. The sea ice concentration was retrieved from December 1972 to December 1976 and published as an atlas (Parkinson et al. 1987). Later this data set was included in the National Snow and Ice Data Center data archive (Parkinson et al. 2004, link https://nsidc.org/data/ NSIDC-0077). SSMR-SSM/I-SSMIS Sea Ice Data Sets Successors of the ESMR instrument were the SMMR sensor on the Nimbus-7 satellite, the SSM/I instrument on the Defence Meteorological Satellite Program’s (DMSP) -F8, F10, F11, F13, F14, and F15 satellites and the SSMIS instrument aboard DMSP-F16, F17 F18 and F19. Important information about satellite orbits and sensor characteristics is shown in Table 4.2. At microwave frequencies, the observed satellite radiance is expressed in terms of brightness temperature. Brightness temperatures measured by passive microwave

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

97

sensors are distributed by NOAA National Climatic Data Center (NCDC) and by National Snow and Ice Data Center (NSIDC). NSIDC distributes data in swath format (level 1b), EASE-Grid and polar stereographic projections. The collected orbital measurements for each 24-h period are mapped to a 25
25 km grid (12.5 km for high frequencies). The Nimbus-7 SMMR polar gridded brightness temperatures and sea ice concentrations (Gloersen 2006) are distributed by NSIDC through the link http://nsidc.org/data/nsidc-0007. SSM/I-SSMIS daily polar gridded brightness temperatures (Maslanik and Stroeve 2004) can be found at http://nsidc.org/data/ nsidc-0001. SMMR, SSM/I and SSMIS instruments coverage is global, except for circular sectors centred over the Poles, due to orbit inclination: these are 611 km, 311 km and 94 km in radius, respectively. The SMMR instrument operated for more than 8 years, from 26 October 1978 to 20 August 1987, transmitting data every other day; observed frequencies are listed in Table 4.2. Unlike the cross-scan instrument ESMR, the SMMR and its follow-on SSM/I and SSMIS instruments are conical scan instruments. Unfortunately, no overlap period exists between the SMMR and ESMR instruments, therefore it is difficult to compare sea ice concentrations retrieved from ESMR data with those retrieved from SMMR and later sensors. SSM/I data are available from 09 July 1987, collected daily. SSM/I frequency was changed to 19.35 GHz instead of 18 GHz for the SMMR; the orbital parameters of the DMSP satellites were also different from Nimbus-7. Comparison of brightness temperatures measured during the SSM/I-SMMR overlap period, from 09 July to 20 August 1987, when Nimbus-7 and DMSP-F8 both provided measurements, demonstrated significant differences (Cavalieri et al. 1999). A linear least-squares fit of the overlap data was obtained for each of the instrument channels (Bjorgo et al. 1997; Cavalieri et al. 1999) and the data were corrected using these linear adjustment coefficients in order to get a more reliable ice data set. The effects of transmissions from one DMSP satellite to the next were also examined and tuning was applied where necessary. SSMIS data are used from 14 December 2006. NOAA NCDC distributes passive microwave satellite data sets suitable for climate studies, called CDRs. There are two CDRs distributed by NOAA NCDC related to the mentioned instruments: SSMI(S) Brightness Temperature – CSU and SSMI(S) Brightness Temperature – RSS. The first one is produced by Colorado State University (CSU) and can be found at https://www.ncdc.noaa.gov/cdr/funda mental/ssmis-brightness-temperature-csu. The second CDR is produced by Remote Sensing Systems (RSS); the link is the following: https://www.ncdc.noaa.gov/cdr/ fundamental/ssmis-brightness-temperature-rss. Comparison of two CDRs has shown that the most notable differences between the data sets are in the methods used for intercalibration and geolocation (Brodzik and Long 2015). Observations at different frequencies and polarizations allow discrimination of different types of ice and atmospheric effects. Algorithms developed for sea ice parameters retrieval use combinations of brightness temperatures in different channels to distinguish open water, first-year (FY) ice and multiyear (MY) ice. The 19V and 37V channels are used in almost all algorithms because of the large difference in

98

E. V. Shalina et al.

the emissivity between water and ice in the 19V channel, and between FY and MY ice in the 37V channel (Svendsen et al. 1983). Different combinations of channels are also employed in the algorithms. The most widely used are: Polarization ratio PRðf, Pv , Ph Þ ¼ ½TBðf, Pv Þ  TBðf, Ph Þ=½TBðf, Pv Þ þ TBðf, Ph Þ, Gradient ratio GRðf1,f2, PÞ ¼ ½TBðf2, PÞ  TBðf1, PÞ=½TBðf2, PÞ þ TBðf1, PÞ, Polarization difference Pðf, Pv , Ph Þ ¼ TBðf, Pv Þ  TBðf, Ph Þ, where TB is brightness temperature, f – frequency and P – polarization. The polarization ratio is used in order to distinguish ice because it is small for either ice type compared to that of the ocean (Cavalieri et al. 1984). The gradient ratio allows us to distinguish between FY and MY ice types because the brightness temperature difference between the two ice types increases with increasing frequency. The advantage of using brightness temperature ratios is that they are less sensitive to physical temperature variations of the emitting layer of the ice cover. The polarization difference is used in the algorithms that include high frequencies. It is similar for all ice types and much smaller for ice than for open water, thus it is used to distinguish between these two surfaces in the near 90 GHz algorithms (Kern and Heygster 2001; Kaleschke et al. 2001). Most sea ice concentration algorithms use tie-points that describe typical emitting properties of ice and water. The tie points are introduced either as brightness temperatures or emissivities for the channels used in the algorithm. Each surface type has its own tie point. They vary from algorithm to algorithm and serve to ensure that each algorithm gives 0% sea ice concentration for the areas of open water and 100% concentration for areas of consolidated ice. In nature, brightness temperature or emissivity of the ice of a certain type or open water vary due to varying temperature and condition of the emitting layer. In order to take that variability into account most of the algorithms use different tie-points for summer and winter or use dynamical tie-points, like OSISAF (Eastwood 2011), which vary on a monthly basis, or even more often. Other sources of errors in the derived sea ice concentrations include the inability of the algorithms to discriminate among more than two radiometrically different sea ice types (three ice types in case of the NASA Team 2 algorithm), formation of the melt ponds, and weather effects. When algorithms are constructed with the purpose to retrieve concentration of two ice types, those ice types are FY and MY, since they are dominant in the Arctic. Presence (within the sensor FOV) of newly formed ice, which is very different from FY and MY in its radiometric signature, results in underestimates of the sea ice concentration by the algorithms (Grenfell et al. 1992; Ivanova et al. 2015). Another condition resulting in large errors in total ice concentration is the melt ponds on the ice surface during summer. Weather effects resulting from atmospheric water vapour, cloud liquid water, rain, and sea surface roughening

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

99

by surface winds are greatly reduced over open ocean by weather filters, however, they still may contribute to the sea ice concentration uncertainty (Andersen et al. 2006; Ivanova et al. 2015). Sea ice concentration algorithms demonstrate the best performance in winter. The accuracy of different algorithms was reported to be within 5% in winter in a high concentration ice field (Cavalieri et al. 1992; Steffen et al. 1992; Kwok 2002; Andersen et al. 2007). In summer the errors increase and the accuracy is estimated to be in the range of 15–20% (Cavalieri et al. 1992; Steffen et al. 1992; Meier and Notz 2010; Ivanova et al. 2015; Lavergne et al. 2018). An extensive inter-comparison and validation of the existing 30 algorithms for sea ice concentration calculation has been undertaken by Ivanova et al. (2014, 2015). Algorithm accuracies were evaluated for different conditions: over low and high sea ice concentrations, with thin ice present, and over areas covered by melt ponds. This study has shown that calculations based on the 19 and 37 GHz satellite data demonstrate the best performance. Algorithms that use 19V and 37V channels are NORSEX (Svendsen et al. 1983), CalVal (Ramseier 1991), Bootstrap Frequency Mode (BF, Comiso 1986) and UMass-AES (Swift et al. 1985). However, none of the algorithms was superior in all conditions, and so it was concluded that the best strategy to create an efficient algorithm for sea ice concentration calculations is to use a combination of existing algorithms. A combined algorithm has been suggested (Ivanova et al. 2015), which is created using the same scheme as OSISAF (Tonboe et al. 2016) being a composition of CalVal and Bristol (Smith 1996), with the CalVal algorithm used solely for ice concentrations below 70%, the Bristol algorithm used for concentrations higher than 90%, and both algorithms used in the concentration interval between 70% and 90% with appropriate weights. The authors also insist on using dynamic tie points because they represent naturally changing geophysical parameters of the sea ice. The sensitivity of several algorithms to melt ponds in summer is investigated in Kern et al. (2016). In this study the melt ponds distribution has been retrieved from MODIS data. As can be expected, the sea ice concentration retrieval algorithms underestimate ice concentration in the cases when melt ponds are present. For 100% sea ice with a melt pond fraction of 40% the underestimation is estimated to 14–26%. However, the underestimation disappears if the melt pond fraction on the consolidated ice reduces to 20%. Moreover, it has been shown that sea ice algorithms overestimate ice concentration if open water exists between ice floes in the field of view of the satellite. The level of overestimation depends on the sea ice concentration; it changes in the range from 14% to 26% in the case of 60% ice surface fraction and becomes stable at the level of 20% in the case when 80% of the area is covered by ice. Sea ice concentrations derived from satellite passive microwave observations can be found on a number of sites on the Internet. Different organizations use different algorithms. NSIDC (National Snow and Ice Data Center, US, http://nsidc.org/) is the oldest repository of cryosphere-related data. It began in 1976 as an analog archive and information centre, the World Data Center for Glaciology. Now it manages terabytes of remote sensing data from different satellites, satellite products, in-situ measurements from polar expeditions and other material related to polar regions. Sea

100

E. V. Shalina et al.

ice information comprises a large part of the NSDIC data records. Satellite observations of Arctic change are presented by different datasets, analytical reviews and graphical illustrations. Everyday monitoring of sea ice is made available to the scientific community and the general public by publishing of year-round scientific analysis and daily image updates of the Arctic sea ice on the NSIDC site. The list of all NSIDC passive microwave data sets can be found in Appendix 1. Some of the sea ice concentration data sets that can be used in scientific analysis of the sea ice transformations are described in brief below. Sea ice concentrations for the whole period of passive microwave satellite measurements are provided in the NSIDC data set the Sea Ice Concentrations from Nimbus-7 SMMR and DMSP SSM/I-SSMIS Passive Microwave Data (http://nsidc. org/data/nsidc-0051). The data are generated using the NASA Team algorithm (Cavalieri et al. 1996) and mapped to a 25
25 km polar stereographic grid. Similarly produced but based on near-real-time satellite brightness temperatures is the data set that can be found through the link http://nsidc.org/data/nsidc-0081. Those two data sets (nsidc-0051 and nsidc-0081) make the base for the data collection that contains up-to-date sea ice extent and concentration values and images from November 1978 to the present (http://nsidc.org/data/g02135). It is called the Sea Ice Index data set. At the product web site (http://nsidc.org/data/ seaice_index) information is organized in a way that the user can do a brief on-line analysis of sea ice transformations. On the site one can find also the previous day’s sea ice map and the monthly-averaged images, showing sea ice extent, concentration, anomalies and trends. The user can produce on-line animation, compare different images and find documentation. Comparison of sea ice concentration trends and sea ice extent trends for March and September provided by Sea Ice Index web site is shown in Fig. 4.1. The Bootstrap Sea Ice Concentrations from the Nimbus-7 SMMR and DMSP SSM/I-SSMIS data set (http://nsidc.org/data/nsidc-0079) provides sea ice concentrations generated using the Bootstrap algorithm with daily varying tie-points (Comiso 2000). Sea ice concentrations are calculated using NSIDC polar gridded brightness temperatures. The quality test of the algorithm performance includes a manual check. An Arctic Regional Ocean Observing System (Arctic ROOS, http://arctic-roos. org) is another large repository of data describing the Arctic regions. Arctic ROOS has been established by a group of 20 institutions from 9 European countries working actively with ice and ocean observations including ice forecasting, coordinated by the Nansen Environmental and Remote Sensing (NERSC), Bergen, Norway. Daily sea ice concentrations are supplied by different members of the Arctic ROOS team and analytical results of long-term observations are provided mainly by NERSC. Daily sea ice concentrations calculated by NERSC using NORSEX algorithm (Svendsen et al. 1983) from passive microwave data are presented on the “arctic-roos.org” based on satellite brightness temperatures received from NSIDC (Fig. 4.2). Sea ice concentration by EUMETSAT’s OSI SAF system is another version of daily sea ice products presented by Arctic ROOS (http://www.osi-saf.org). It is based on co-operation between several European institutions, Meteo-France being the host institute. The concentration product is based on passive microwave data

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

101

Fig. 4.1 Comparison of sea ice concentration trends (left) and sea ice extent trends (right), produced on-line for March, 1979–2018, and September, 1979–2017, at the NSIDC Sea Ice Index web site http://nsidc.org/cgi-bin/bist/bist.pl?config¼seaice_extent_trends

from SMMR, SSM/I and SSMIS. An OSI SAF sea ice concentration algorithm is a combination of two algorithms, Bristol (Smith and Barrett 1994) and the Bootstrap frequency mode (Comiso 1986) algorithms, each of them working separately over marginal or consolidated ice, and working together over intermediate ice concentrations, with dynamical tie points (Tonboe et al. 2017). The combined algorithm is run on brightness temperatures corrected using a Numerical Weather Prediction (NWP) model. Sea ice concentration data is mapped to a polar stereographic grid at 10 km spatial resolution.

102

E. V. Shalina et al.

Fig. 4.2 Daily updated time series of Arctic sea ice area derived from SSMIS data provided by NERSC from the Arctic ROOS system. (Source: http://arctic-roos.org/observations/satellite-data/ sea-ice/ice-area-and-extent-in-arctic)

At OSI SAF a multi-sensor analysis system using several sensors to analyse sea ice classes has also been developed to distinguish between FY and MY ice. The operational product uses SSMIS and scatterometer ASCAT data (see the example of a quicklook of sea ice type product in Fig. 4.3). Data is available from 2005 and can be downloaded from ftp://osisaf.met.no/archive/ice/type/. The same multi-sensor analysis system is used to produce an ice edge product. One of the latest products of EUMETSAT’s OSI SAF is the sea ice concentration climate data record (Lavergne et al. 2018) that is described in Sect. 4.1.4. The Arctic ROOS site also presents a comparison of sea ice concentration algorithms and the analysis of long-term passive microwave satellite observations. Those results will be discussed later, in Sect. 4.2. AMSR-E and AMSR2 Sea Ice Data Sets The Advanced Microwave Scanning Radiometer – Earth Observing System (AMSR-E) was a 12-channel, 6-frequency, passive-microwave radiometer system. The frequencies of the instrument are listed in Table 4.2. It measured horizontally and vertically polarized brightness temperatures, the spatial resolution of the individual measurements varied from 5.4 km at 89 GHz to 56 km at 6.9 GHz. The main advantage of AMSR-E data is that its spatial resolution doubles that of SMMR and SSM/I data. Also, AMSR-E combines into one sensor all the channels that SMMR and SSM/I had individually. The instrument was developed by the Japan Aerospace Exploration Agency (JAXA). It functioned on board the Aqua satellite from May 2, 2002 up to October 4, 2011.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

103

Fig. 4.3 Daily updated OSI-SAF sea ice type product. Spatial resolution is 10 km. (Source: http:// osisaf.met.no/p/ice/index. html#type)

AMSR2 (the successor of AMSR-E) was launched on May 18, 2012 on board GCOM-W1 Japanese satellite. The new instrument was designed with the aim to continue the valuable AMSR-E data record. Its characteristics are basically the same with some enhancements like an additional channel at 7.3 GHz. The calibrated brightness temperature data and sea ice products based on this data are available since July 2012. NSIDC archives and distributes daily, weekly, and monthly Level-1A, Level-2A, Level-2B, and Level-3 data products from AMSR-E; they are listed in Appendix 2. AMSR-E sea ice concentrations distributed by NSIDC are calculated using the NT2 algorithm (Markus and Cavalieri 2000). Data are available for the period from 01 June 2002 to 4 October 2011. The advantage of using the NT2 algorithm is that it uses the 89 GHz data that helps to resolve ambiguity between low ice concentrations and areas with strong surface effects due to snow layering. This algorithm resolved an additional type of ice – thin ice. NT2 implements a radiative transfer model to filter for weather contamination effects introduced into calculations because of using high-frequency channels. Sea ice concentration fields are provided in two spatial resolutions: 12.5 km and 25 km (http://nsidc.org/data/AE_SI12 and http://nsidc.org/data/AE_SI25). In the 12.5 km resolution data set there is also data on snow depth on the FY ice calculated using AMSR-E snow-depth-on-sea-ice algorithm described in Markus and Cavalieri (1998). Data are mapped to the same polar stereographic projection used for SMMR/SSMI/SMISS data to provide consistency and continuity with the exiting passive microwave NSIDC products. Daily AMSR-E and AMSR2 sea ice maps have also been provided by University of Bremen (UB-IUP) since the beginning of the AMSR-E period (http://www.iup. uni-bremen.de/seaice/amsr/ and https://seaice.uni-bremen.de/sea-ice-concentration/). The service is a part of the GMES (Copernicus) project Polar View and Arctic

104

E. V. Shalina et al.

Fig. 4.4 Left: Daily updated Arctic sea ice concentration produced by University of Bremen. Satellite data source: AMSR2, spatial resolution: 6.25 km. Right: subset of the ice concentration map in the Greenland Sea region. (Source: https://seaice.uni-bremen.de/sea-ice-concentration/)

ROOS service. Concentrations are calculated using the ASI algorithm (Kaleschke et al. 2001; Spreen et al. 2008), which is an enhancement of the Svendsen sea ice algorithm for frequencies near 90 GHz (Svendsen et al. 1987). The algorithm uses an empirical model to retrieve ice concentrations between 0% and 100%. Global and regional sea ice concentration data are mapped to a polar stereographic grid at 6.25 km spatial resolution (as shown in Fig. 4.4). Sea ice concentration retrieval from AMSR-E and AMSR2 is based on brightness temperatures available from JAXA free of charge (http://gcom-w1.jaxa.jp/search.html). During the ESA Climate Change Initiative (CCI) programme running from 2009 to 2018 (http://cci.esa.int/), one of the projects was dedicated to intercomparing algorithms for ice concentration in order to produce sea ice climate data sets that satisfied the GCOS requirements (CGOS 2011, 2016). These requirements specify that sea ice concentration data sets should have a spatial resolution of 10–15 km, weekly coverage, accuracy of 5% of ice area fraction and stability of 5% per decade. The CCI sea ice project has produced several ice concentration climate data records that are described in Sect. 4.1.4. MWRI Data Set Sea ice concentration monitoring is also conducted by the FengYun-3 (FY-3) satellites of the Chinese polar orbiting meteorological satellite program (Dong et al. 2009). These satellites have been in operation since 2008 and carry among others the Microwave Radiation Imager (MWRI). This is a conical

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

105

Fig. 4.5 Example of sea ice concentration maps in the Arctic (left) and Antarctic (right) on 01 January 2013 based on data from the FY-3B satellite. (Source: http://www.nsmc.org.cn/en/ NSMC/imagesDCPC/b20.jpg)

scanning radiometer operating at 10.65, 18.7, 23.8, 36.5, and 89 GHz providing brightness temperatures for estimating daily global ice concentration maps with 12.5 km resolution (Fig. 4.5). The data from the FY-3 satellites are downloaded at several receiving stations and processed and distributed at the Fengyun Satellite Data Center (http://satellite.nsmc.org.cn/PortalSite/Default.aspx).

4.1.1.2

Multi-sensor Sea Ice Products

Multi-sensor Analysed Sea Ice Extent: Northern Hemisphere (MASIE-NH) Data Set The Multi-sensor Analysed Sea Ice Extent – Northern Hemisphere (MASIE-NH) data set is produced by NSIDC in cooperation with the U.S. National Ice Center (NIC). The product provides maps of daily sea ice extent and the position of the sea ice edge for the Northern Hemisphere and 16 Arctic regions in a polar stereographic projection at 1 km and 4 km grid cell sizes. MASIE uses the most recent sea ice data from the NIC, obtained nightly. That data comes from the Interactive Multi-sensor Snow and Ice Mapping System (IMS, http://www.natice.noaa.gov/ims/) developed at NIC, which maps the extent of daily hemispheric snow and ice coverage using different sources of information. The product is created manually and uses primarily visible imagery from the POES, geostationary orbiting satellites and MODIS. Data from microwave satellites are

106

E. V. Shalina et al.

used when it is necessary to view through clouds. Additional sources of data are ground weather observations from different countries including snow and sea ice products from NIC, NCEP, NESDIS, USAF, etc. The list of all sources of information that can be used when the IMS product is created comprises 30 sensors and services. When the analysts use passive microwave satellite data, the 40% or 60% concentration outline is typically used to associate with a 7/10+ contour on the NIC ice chart and consequently the NIC ice edge situates in a more advanced position than the IMS ice edge. IMS daily data is also archived by NSIDC. IMS data are available in different resolutions. A 24-km resolution IMS snow and sea ice chart has been produced daily from February 1997 to the present. A 4-km product was implemented in 2004 and a 1 km product has been available from December 2014. The primary user of the described NIC products is the National Centers for Environmental Prediction (NCEP) where the data is used as an input to numerical weather prediction models. The quality of the IMS product depends on the availability of clear sky visible imagery, since passive microwave data (if that are used instead of visible data) are of lower resolution and the result of analysis in case of using microwave data has less information. Other sources of errors are the correctness of satellite image registration, the quality of other input data sources, and the skills of the analyst. Practical experience demonstrates that the accuracy of the IMS products is higher than the accuracy of automatically generated products based solely on satellite data. However, since the manual production method can be a source of inconsistency the data are not suitable for the long-term trends analysis. The MASIE product (https://nsidc.org/data/masie) is developed with the aim to fill a need for an intermediate type of product between operational ice charts and automatically calculated passive microwave sea ice maps. The main difference between the MASIE product and NIC ice charts is that ice charts are weekly or biweekly while MASIE products give a daily picture of ice extent that is provided in different formats so that a user can choose the most appropriate way to use the data. The difference between the MASIE and IMS products is that the MASIE product is available in a greater number of formats and it contains ice information for the whole Arctic as well as for 16 smaller Arctic regions. The MASIE-NH product comprises PNG files for browsing, GeoTIFF image files, shapefiles, Google Earth files and time series plots for 16 regions. Figure 4.6 demonstrates three of the MASIE-NH products: the daily Northern Hemisphere sea ice extent at 1 km resolution, sea ice extent in the East Siberian Sea, and a time series plot for 1 of 16 regions that shows the sea ice extent for the previous 4 weeks of a given year. MASAM2: Daily 4-km Arctic Sea Ice Concentration The next step in developing an intermediate type of product between operational ice charts and automatically calculated passive microwave sea ice maps has been done when NSIDC created the MASIE-AMSR2 (MASAM2) daily 4 km sea ice concentration product (Fetterer et al. 2015) that is based on joint use of MASIE sea ice extent and AMSR2 sea ice concentrations. In the beginning, 10 km-resolution AMSR2 data are bilinearly interpolated to the 4 km MASIE grid. Then data are jointly processed based on the assumption that the MASIE extent is more accurate. In cases where the MASIE

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

107

Fig. 4.6 Three MASIE products: (a) Daily Northern Hemisphere Sea Ice Extent on 28 Jan. 2019, area covered by ice is coloured in white, (b) Sea Ice Extent in the Kara Sea on 28 Jan. 2019 and (c) Example of a time series plot for region 5, Kara Sea, from 1 January to 28 January 2019. (Source: https://nsidc.org/data/masie)

product indicates ice, the AMSR2 sea ice concentration value is set for a grid cell if that concentration value is greater than 70%; else it is set to 70% sea ice concentration. The reason why 70% ice concentration is used as a minimum value is because the MASAM2 data product is developed for initializing an operational short-term sea ice forecast model and that restriction works well in the model’s data assimilation algorithm. An example of the MASAM2 product is shown in Fig. 4.7. Data can be downloaded from the FTP site ftp://sidads.colorado.edu/pub/DATASETS/NOAA/ G10005/.

108

E. V. Shalina et al.

Fig. 4.7 Example of MASAM2 product: sea ice concentration on 26 January 2019, PNG browse file. (Source: ftp://sidads. colorado.edu/pub/ DATASETS/NOAA/ G10005/)

4.1.1.3

Sea Ice from Satellite Radar Data

Besides passively sensing emissions coming from the surfaces on Earth, satellite sensors can also actively emit microwaves toward objects on the earth’s surface. These microwaves reflect from surface objects, return to the sensor, where it is measured and analysed. Satellite Side-Looking Radar (SLR) systems onboard the Okean series of satellites were used extensively in Russian ice monitoring since 1983 until 2000 (Bushuev et al. 1983; Alexandrov and Loshchilov 1993; Alexandrov et al. 2000; Johannessen et al. 2000, 2007; Asmus et al. 2002). The SLR operated at 3.2 cm with a swath width of 450 km and ground resolution of 1200
1500 m. Mosaics of SLR data were used weekly in the winter period to analyse the sea ice conditions in the Russian Arctic. The images were not used in the summer because of their poor capability for discriminating ice from water and multiyear ice from first-year ice in the melt period. The SLR images were provided by the Research Center for Operative Earth Monitoring (NTs OMZ) in Moscow (ntsomz.ru). From the 1990s Synthetic Aperture Radar (SAR) and scatterometer instruments have become increasingly important in sea ice monitoring. In particular SAR, providing high-resolution images of practically all sea ice areas in both hemispheres, is now one of the main sensors for use in both operational monitoring and research related to sea ice. During the 1990s there were three to four SAR satellites producing

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

109

Fig. 4.8 SAR satellites from 1990s to the beginning of 2020s operating in L-bad, C-band and X-band. (Source: https://www.nature.com/articles/ncomms13844/figures/2)

a limited amount of sea ice data mainly for research and demonstration purposes. In 2018 more than 12 SAR satellites are in operation, most of them have sea ice monitoring as their mission objective. Previous and present SAR satellites include Seasat (1978), ALMAZ-1 (1991), JERS-1, ERS-1/2, RADARSAT-1/2, ENVISAT, TerraSAR-X, TanDEM-X, COSMO-SkyMed, RISAT-1 (2012–2017), KOMPSAT, Sentinel-1A/B and others (Fig. 4.8). The European Remote Sensing (ERS) program, which started in 1991, represented a major milestone in satellite SAR remote sensing of sea ice, because the two satellites ERS-1 and ERS-2 have operated continuously for more than 10 years and delivered tens of thousands of SAR images of ice-covered regions around the world, including the Northern Sea Route (Johannessen 2008), see also Chap. 12 in this book. Since 1996 the Canadian RADARSAT has delivered on a commercial basis large amounts of wide-swath SAR images of Arctic sea ice areas and these images were gradually introduced into the operational ice monitoring by the Canadian Ice Service (Flett 2004; Flett and Vachon 2004) and later by other national ice centers. In 1997 mosaics composed from RADARSAT ScanSAR stripes started to be produced for the whole Arctic (Fig. 4.9). From 2003 the European ENVISAT satellite has delivered wide swath ASAR data, for sea ice observation in large areas of the Arctic (Sandven et al. 2004; Sandven and Johannessen 2006; Johannessen et al. 2007). During the decade from 2002 to 2012, ENVISAT together with RADARSAT-1 and RADARSAT-2 produced a large amount of systematic SAR data in C-band, which became the main data source for operational ice charting in many countries.

110

E. V. Shalina et al.

Fig. 4.9 Example of RADARSAT-2 ScanSAR mosaic of the Arctic in April 2008. The mosaic shows clearly the different backscatter between multiyear ice and first year ice. (Credit: NASA Images by Robert Simmon, adapted from Kwok 2010. RADARSAT image © 2010 Canadian Space Agency. https://www.nasa.gov/ multimedia/guidelines/ index.html)

RADARSAT-2, which was launched in December 2007, was a follow-on to RADARSAT-1 that terminated in April 2013. RADARSAT-2 has operated for more than 10 years and is still active, delivering SAR data to the Canadian Ice Service, numerous research projects and the Copernicus Marine Environmental and Monitoring Services which started in 2014 (http://marine.copernicus.eu/training/ education/ocean-parameters/sea-ice/). The Copernicus programme is an extensive global Earth Observation programme coordinated and funded by the European Union in collaboration with member states, space agencies and other major agencies working with monitoring and forecasting services (http://www.copernicus.eu/main/overview). A major component of Copernicus is the Sentinel satellite programme, consisting of more than ten satellites with dedicated sensors for observing atmosphere, ocean, cryosphere and terrestrial topics. For sea ice observation the Sentinel-1 programme has been developed, consisting of two dedicated SAR satellites, which also serve other ocean and land monitoring services. Sentinel-1A was launched in 2014 and Sentinal-1B in 2016, both delivering SAR data in C-band in four operational modes including interferometry and dual polarization. These satellites are dedicated to long-term operational monitoring where data are delivered in near real time to the users (https://sentinels.copernicus. eu/web/sentinel/missions/sentinel-1). For sea ice monitoring the Extra Wide swath mode (swath width 400 km) is the main mode, ensuring that the whole Arctic can be completely covered by SAR every 3 days. The SAR mosaic from 450 Sentinel-1 A/B images collected in 4 days is shown in Fig. 4.10. The free and open data policy adopted by the Copernicus programme gives open data access to all users of the Sentinel products, via a simple pre-registration, see https://sentinel.esa.int/web/sen tinel/sentinel-data-access.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

111

Fig. 4.10 The SAR mosaic from Sentinel-1 A/B images collected during 18–21 November 2017. (Source: Luo and Flett 2018)

TerraSAR-X is a commercial German SAR Earth observation satellite which was launched in June 2007. The TerraSAR-X mission is a Public Private Partnership between the German Aerospace Centre (DLR) and Europe’s aerospace manufacturer Astrium, subsidiary of the European Aeronautic Defence and Space Company (EADS). In 2010, the successor of TerraSAR-X was launched; it is called TanDEM-X and its mission is to make a pair with the first one in order to provide high resolution digital elevation measurements with good spatial coverage. The TanDEM-X satellite is almost identical to the TerraSAR-X Satellite. Both satellites fly very closely with only a few hundred meters separation, typically between 250 and 500 m. Together they comprise the first and only synthetic aperture radar interferometer in space. The primary payload is an X-band radar instrument with the working frequency of 9.6 GHz. TerraSAR-X/TanDEM-X can acquire radar data in the following main imaging modes. High-resolution SAR data from TerraSAR-X and its twin satellite facilitates detection and tracking of small ice objects such as icebergs and floes. Ice surface such as ice ridges can be identified and detailed information on ice properties, including sea ice topography, can be retrieved. That data are also used for describing melting on the ice surface and for discrimination between different types of ice. Wide ScanSAR mode provides a survey of an area of up to 400,000 km2 within a single acquisition that makes it ideal for sea ice monitoring. Staring SpotLight and Wide ScanSAR modes have recently become operational. The TerraSAR-X Staring Spotlight mode provides the highest spatial resolution (less than 1 m) presently available on a commercial spaceborne SAR system.

112

E. V. Shalina et al.

Spaceborne scatterometers have been in operation for more than 15 years, and have provided continuous synoptic microwave coverage of the whole Earth. A scatterometer transmits radar pulses and receives backscattered energy in the same way as the SAR, but is designed to obtain lower spatial resolution data from a wider extent of the observed surface. Compared to SAR, the main benefit is the wide-swath coverage that provides a daily global map of the sea ice surface at 25–50 km resolution, enabling tracking of large ice features. Sea ice extent has been observed by several spaceborne scatterometers in recent decades, with SeaWinds on QuikSCAT being the most important as it has been in operation since 1999 (e. g. Gohin and Cavanié 1995; Ezraty and Cavanié 1999; Remund and Long 1999). The maps are derived from sigma-naught data, averaged daily (see Fig. 4.11). The QuikSCAT Ku-band radar backscattering images enable the monitoring of sea-ice extent (by discriminating open water from sea ice) and classification of the arctic sea ice as MY ice and FY ice (Kwok 2004; Shalina and Johannessen 2008; Swan and Long 2012; Lindell and Long 2016a). The QuikSCAT mission ended on November 23, 2009 and the IFREMER (L’Institut Français de Recherche pour l’Exploitation de la Mer) data set (http://cersat.ifremer.fr/data/products/cata logue) covers the whole QuikSCAT operation period. Advanced Scatterometer (ASCAT) is another instrument that provides useful global data for sea ice applications. The first ASCAT was launched on the EUMETSAT MetOp-A satellite in October 2006. It became fully operational in May 2007 and continues to operate today. Another ASCAT instrument works on the MetOp-B that was launched in September 2012. Both satellites, MetOp-A and MetOp-B carry identical ASCAT instruments. The main objective of ASCAT is the measurement of wind speed and direction over the oceans, but the data is also used for observation of sea ice both in the Arctic and Antarctic (Lindell and Long 2016b). ASCAT operates at a different frequency (C-band) than SeaWinds/ QuikSCAT and has a smaller data gap near the Pole (Fig. 4.11). The EUMETSAT OSISAF uses these scatterometer data in combination with passive microwave data to produce maps of sea ice edge, sea ice types (MY and FY) and sea ice drift in both hemispheres. The data are available at http://osisaf.met.no/p/ice/index.html. Scatterometer data are provided by CERSAT, the French satellite oceanography centre in IFREMER (http://cersat.ifremer.fr/data/products/catalogue), and from the National Environmental Satellite Data and Information Service (NESDIS) at NOAA: http://manati.star.nesdis.noaa.gov/datasets/QuikSCATData.php and http:// manati.star.nesdis.noaa.gov/datasets/ASCATData.php. The Chinese National Space Administration operates the Haiyang-2A (HY-2A) satellite which carries a Ku-band scatterometer, a radar altimeter (Ku- and C-band) and a microwave radiometer imager for ocean monitoring. The HY-2A has been in operation since 2011 and provides daily scatterometer data for surface wind and sea ice monitoring in both hemispheres. Recent studies have shown that sea ice extent and sea ice types (MY/FY) can be retrieved, similar to what is obtained from QuikSCAT and ASCAT (Li et al. 2016). A comparison between the MY ice type maps from ASCAT/SSMIS (the OSISAF product) and the MY map from HY-2A/SCAT data is shown in Fig. 4.12, left and centre, while a time series of the MY ice extent and the total ice extent from January 2013 to July 2015 is shown in Fig. 4.12 right.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

113

Fig. 4.11 Scatterometer backscatter maps obtained from: QuikSCAT on 15 February 2007 (left), ASCAT on 15 February 2015 (right). Low values correspond to first-year ice (Russian coast) and strong values to multiyear ice (North of Canada and Greenland). (Source: http://cersat.ifremer.fr/ data/products/catalogue)

Since a scatterometer is a key instrument for global monitoring of weather and climate, scatterometers will be operated on future satellites under the EUMETSAT Polar System, Second Generation, 2021–2040.

4.1.1.4

Sea Ice Data from Optical and Infrared (IR) Satellite Sensors

MODIS/Terra Sea Ice Extent Measurement of sea ice parameters such as sea ice extent can be done with optical/IR sensors when cloud conditions allow observations of the earth’s surface. Optical/IR data have been available from the 1980s and are, together with passive microwave data, commonly used for sea ice observation because they are freely available and downloaded at numerous receiving stations world-wide. In the last two decades, data from the Moderate Resolution Imaging Spectroradiometer (MODIS) are among the most widely used in both sea ice monitoring and research. The daily sea ice products MOD29P1D (Terra) and MYD29P1D (Aqua) contain fields for Sea Ice by Reflectance (Fig. 4.13a), Sea Ice by Reflectance Spatial QA, Ice Surface Temperature (IST) (Fig. 4.13b), and Ice Surface Temperature Spatial QA (Riggs et al. 2006). The algorithm that retrieves sea ice from optical/IR sensors uses a modified Normalized Difference Snow Index (NDSI), which is applied with the assumption that ice is covered by snow. ISTs are calculated using split-window technique.

114

E. V. Shalina et al.

Fig. 4.12 Left: MY ice map from ASCAT/SSMIS, Center: MY ice map from HY-2A/SCAT on the same day as the left figure, Right: Time series of total Arctic ice extent from three data sets and MY ice extent from two data sets, covering the period January 2013 to July 2015. (Li et al. 2016)

Fig. 4.13 Example of MODIS sea ice data products in the Fram Strait area: (a) Sea Ice by Reflectance and (b) Ice Surface Temperature (IST) maps. (Data are from April 5, 2008, Source: https://nsidc.org/data/mod29p1d)

The data are provided in HDF-EOS format, along with corresponding metadata (Hall et al. 2006). The files are at a 1 km resolution and the data is gridded onto the EASE-Grid projection. The data is available from 24 February 2000 to the present and can be downloaded from https://nsidc.org/data/mod29p1d. Under the Copernicus programme, Sentinel-3A and -3B will be the main data source for optical and IR data covering ocean and sea ice areas, where the same sea ice products as for MODIS will be available. The optical/IR data are complementary to the SAR data from Sentinel-1. The Sentinel-3 satellites also carry a Synthetic Aperture Radar Altimeter, which provides data on sea ice freeboard and thickness. Also Sentinel-2, providing high-resolution optical images, mainly over land areas, will deliver images of sea ice in coastal areas.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

115

For more information on the Sentinel program see https://sentinels.copernicus.eu/ web/sentinel/about-sentinel-online.

4.1.2

Ice Charts from Different Sea Ice Services

4.1.2.1

Arctic and Antarctic Research Institute Sea Ice Charts

The main source of sea ice information in the Russian Federation is from the Arctic and Antarctic Research Institute (AARI), St. Petersburg, that provides centralized services for shipping and coastal activities within the Northern Sea Route, for the Barents Sea, Central Arctic Basin and Arctic seas – Greenland, Kara, Laptev, EasternSiberian, Chukchi, and for the seas with the seasonal ice cover – Baltic, White, Bering, Okhotsk, and Caspian seas. One of the main goals of the institute is to support safe maritime operations along the Northern Sea Route, including ice forecasting. Data sources used in the Arctic ice charts varied over time. In the beginning of the Arctic navigation era observations were scarce. The first airborne sea ice reconnaissance flights in the Siberian Arctic seas were carried out in the middle of the 1920s. Aerial surveys became more organized with more extensive coverage from 1933 (Borodachev and Shilnikov 2002) and AARI has produced sea ice charts starting from that year, although the first charts covered only small regions. At first, reconnaissance flights were conducted only in June, then later they were also conducted in April and May. Onboard observers used a variety of surface attributes to distinguish different types and forms of sea ice. In the beginning aerial surveys have been organized with a goal to provide information on sea ice conditions for ship navigation. But later the amount and quality of collected ice information became sufficient for facilitating sea ice research. Before 1992, ice reconnaissance flights were conducted every 10 days during the navigation season (May-October), and up to 30 day intervals during the rest of the year (Borodachev and Shilnikov 2002). Annually 30–40 aircrafts carried out 500–700 ice reconnaissance flights (Bushuev et al. 1995). After 1992 aerial surveys were only organized occasionally in order to support operational or scientific activities. With the end of regular airborne ice reconnaissance, data from meteorological satellites of the NOAA and EOS series have become the main source of information. Radar images from the Okean satellites were widely used in 1983–2000 (Alexandrov and Loshchilov 1993; Bushuev 1997). SAR data has been used occasionally in the beginning of the SAR data era and regularly in recent years. The charts from the National Ice Center are used as supplementary information for compiling weekly global ice charts and data from scatterometers are also used (Smirnov et al. 2000; Johannessen et al. 2007). Information from coastal stations and ships is used where available. Charts produced by the AARI are global ice charts for the Arctic Ocean and regional ice charts, http://www.aari.ru/. Weekly ice charts for the whole Arctic

116

E. V. Shalina et al.

Ocean are available from 1997 to the present, showing ice concentration with grades of 1–6/10th and 7–10/10th, fast ice and ice free areas in summer (Fig. 4.14a) and major sea ice types in winter maps (Fig. 4.14b). Regional ice charts for the Greenland, Barents, Kara, Laptev, East-Siberian, and Chukchi Seas are also available from 1997 onwards. These charts are more detailed and contain the following ice types in winter: new ice, nilas, young ice (grey and grey-white), first-year ice (thin, medium, and thick), and old ice (second-year and older). Additional ice information is presented in the Egg Code format (WMO-No.259 2014). In the summer, ice concentration is shown using six intervals of sea ice concentration. There is a data set archived by NSIDC (G02176, http://nsidc.org/data/g02176), based on the AARI ice charts for the period from 1933 to 2006. Each chart includes information about the partial concentration of three ice classes (multiyear, first-year, and new/young ice), the total concentration, and an indication of whether the ice is drifting or landfast (Mahoney et al. 2008). Spatial coverage of the archived charts varies significantly, having only small areas in the beginning of the presented period, before the 1950s. Most of the data set charts were compiled every 10 days during the navigation season, and compiled monthly for the rest of the year.

4.1.2.2

Canadian Ice Service Sea Ice Charts and Climatologies

The Canadian Ice Service (CIS) is a division of the Meteorological Service of Canada in the Department of Environment. The main goal of the CIS is to support safe and efficient maritime operations. The charts produced by the CIS are based on satellite data analysis, airborne sea ice reconnaissance, and forecasting. The service also derives ice climatology from collected data, does data archiving, research and development. At the CIS, radar imagery from satellites is the principle data source, confirmed by visual observations from aircrafts and helicopters. The Ice Service uses approximately 7000 Radarsat SAR images covering over a billion square kilometres annually (WMO-No.574 2000). Visual and infrared imagery collected by US polar orbiting satellites is an important additional data source for sea ice analysis. Passive microwave satellite data is used as background daily information about global ice distributions at low resolution. Important information on local ice conditions is provided by ice observers who carry out visual observations from icebreakers, aircrafts and helicopters. Daily ice charts are developed for 26 regions when there is marine activity (Sea Ice Climatic Atlas 2002). They provide timely information on ice conditions in the area showing ice concentration in tenths, stage of development and the form of the ice. The stage of development includes the following classifications: new ice, nilas, young ice, first year ice and old ice. Ice information is presented by colours according to the WMO standard and in Egg Code format. Regional ice charts for

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

117

Fig. 4.14 The AARI ice charts covering the whole Arctic: (a) a summer chart, from the period 21 to 23.08.2011, and (b) a winter chart, from 27 to 29.01.2019

118

E. V. Shalina et al.

Fig. 4.15 Regions of Canadian waters, for which ice conditions are reported by CIS

the regions shown in Fig. 4.15 are developed weekly. The delivered parameters are the same as in daily charts. Weekly ice charts are used for the strategic planning of marine operations for icebreakers and shipping companies. These charts are also used by researchers who study ice conditions. The frequency of the regional charts has increased over time. An example of a regional ice chart (for the Eastern Arctic) is shown in Fig. 4.16. Important statistical information about ice conditions in Canadian waters can be found in the 30-year Climatic Ice Atlases developed by CIS. These data sets contain statistics for the 1981–2010 years for different regions of Canada, such as the East Coast of Canada, the Great Lakes and the Northern Canadian Waters. CIS link is www.ec.gc.ca/glaces-ice/.

4.1.2.3

National Ice Center Arctic Sea Ice Charts and Climatologies

The U.S. National Ice Center (NIC) is an inter-agency sea ice analysis and forecasting centre that include the Department of Commerce/NOAA, the Department of Defense/U.S. Navy, and the Department of Homeland Security/U.S. Coast Guard. NIC has been producing Arctic and Antarctic sea ice charts since 1972. NIC ice products include: daily, weekly and seasonal products, multisensor snow and ice mapping system, seasonal ice products for selected areas, outlooks and forecasts of the Northern American Ice Service region. As for services, NIC provides Optimum Track Ship Routing (OTSR) recommendations and pre-sail ship briefings for ship routing services.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

Fig. 4.16 Eastern Arctic weekly ice chart issued by CIS on 17 November 2014

119

120

E. V. Shalina et al.

Daily NIC product delineates sea ice pack with the sea ice concentration in the interval 8–10/10th and the marginal ice zone (MIZ). The purpose of including the MIZ is that knowing its position is important for navigation planning and also is essential for sea ice forecast. The daily map is developed by an analyst based on integrated examination of real time satellite data, buoy data, meteorological data and model forecasts. Satellite data used by NIC are mainly radar, visible, and infrared imagery from a variety of sources. Microwave and scatterometer data are also used in order to have a global picture of the ice conditions. Satellite imagery constitutes over 95% of the information received and integrated into the NIC ice products (WMO-No.574 2000). Weekly ice products are sea ice concentration, stage of development and 30-day change maps, see Fig. 4.17. The latest NIC sea ice products can be downloaded from www.natice.noaa.gov/Main_Products.htm. The dataset “National Ice Center Arctic Sea Ice Charts and Climatologies in Gridded Format” can be downloaded from data.noaa.gov or from http://nsidc.org/ data/g02172. It archives operational weekly (1972–2001) and bi-weekly (2001–2007) NIC ice charts and calculated climatologies. The parameters mapped in the earliest operational charts are total concentration and fast ice coverage and for the maps after 1994 - the concentration of first-year ice, multiyear ice, and thin ice. The climatology products are monthly median, maximum, minimum, first and third quartile concentrations, as well as frequency of occurrence of ice over the entire period of record.

4.1.2.4

Met.no sea Ice Charts

The operational sea ice service from the Norwegian Meteorological Institute, met.no, produces sea ice concentration charts based on a manual interpretation of satellite data by experienced analysts. The main satellite data used are SAR images from Radarsat-2 satellite in ScanSAR wide mode and Sentinel-1 data. The area of interest is around Svalbard and the Barents Sea. In addition to SAR data, met.no uses visual and infrared data from METOP meteorological satellite, NOAA satellites and MODIS instrument. These satellites cover the observing region several times each day, however only images obtained in cloud free conditions can be used. The sea ice concentration product is updated daily around 14:00 GMT. The ice chart shows also sea surface temperature contours. Charts are available for the following regions: Svalbard, Barents Sea, North and Baltic Sea, Eastern Greenland. There is also a General View chart that covers basically half of the Arctic Ocean and all the previously mentioned areas. The latest updated ice chart is available from met.no at polarview.met.no/ clickmap.htm or wms.met.no/icechart. The example of the met.no ice chart is shown in Fig. 4.18. The same operational ice charts are also available as a Web Map Service (WMS) layer at wms.met.no/icechart/. Historical ice charts back to 1998 can be found at http://157.249.32.242/archive/. The charts in the archive are in the GIF graphical format.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

121

Fig. 4.17 Weekly sea ice products developed by NIC: (a) sea ice concentration, (b) stage of development and (c) 30-day change maps

122

E. V. Shalina et al.

Fig. 4.18 Daily ice chart from the met.no ice service. Main satellite data sources are mentioned on the map

4.1.2.5

DMI Sea Ice Charts

The Ice Service at the Centre for Ocean and Ice at the Danish Meteorological Institute (DMI) produces sea ice charts for the area around Greenland mainly to support navigation. The regions covered by the charts are: the whole area around Greenland, the Cape Farewell region, the west and east coasts of Greenland. In the area east of Greenland the ice has its origin mainly in the Arctic Ocean. It drifts southward in the East Greenland Current along the coast and eventually melts during summer time. The ice in the west region develops locally, it melts during warm season. In all parts of Greenland waters the icebergs are present all year around. However, the waters around Cape Farewell (between 59 and 62 N) are considered the most dangerous for navigation because of extremely harsh weather conditions and the presence of icebergs together with ice from the Arctic. Therefore, mapping ice conditions for the waters around Cape Farewell is very important and ice charts for that region are produced by DMI most frequently. Overview charts are issued twice a week (Mondays and Thursdays) and charts for other regions are produced according to season and navigation. The maps are either based on satellite images or reconnaissance flights from Greenland. The example ice chart is shown in Fig. 4.19. DMI charts are composed using WMO colour standards (WMO/TD-No.1215 2014).

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

123

Fig. 4.19 DMI ice chart for the east coast of Greenland

4.1.2.6

Ice Chart WMO Standards

The WMO colour standard includes two separate colour schemes (WMO/TD-No. 1215 2014). One of the two should be used on a single chart. 1. The CT Colour Code Standard is based on presenting total ice concentration (CT). It should be used when the stage of development is relatively uniform but concentration is highly variable. 2. The SoD Colour Code Standard is based on showing stage of development (SoD) of the ice cover. It is recommended for use when the concentration is relatively uniform (high) but the stage of development varies over the ice pack. The CT and SoD Colour Code Standards are presented below in Tables 4.3 and 4.4. The symbols of the Egg Code (WMO-No.259 2014) that describe ice concentration and stage of ice cover development are described below.

124

E. V. Shalina et al.

Table 4.3 Total concentration colour code standard

Letter C in the oval refers to ice concentration. Ct is total concentration of the ice in the area, reported in tenths. Partial concentrations (Ca, Cb, Cc) are also described in tenths and must be reported as a single digit. They are reported in order of decreasing thickness, Ca being concentration of the thickest ice and Cc describing concentration of the thinnest ice. Letter S refers to the stage of ice cover development. Sa, Sb, Sc are listed using the code shown in Table 4.5, in decreasing order of thickness. These codes correspond directly with the partial concentrations above. So and Sd are development stages (ages) of remaining ice types. So if reported is a trace of ice type thicker/older than Sa. Sd is a thinner ice type which is reported when there are four or more ice thickness types. Fa, Fb and Fc describe predominant form of ice (floe size) corresponding to Sa, Sb and Sc respectively. Table 4.6 shows the codes used to express this information.

4 Changes in Arctic Sea Ice Cover in the Twentieth and Twenty-First Centuries

4.1.3

125

Historical Long Records of Sea Ice Parameters

Climate change cannot be understood without studying long series of observations. In this section, we describe historical sea ice data available for research. The list of main sources of historical sea ice data is presented in Table 4.7. Table 4.4 Stage of development colour code standard

126

E. V. Shalina et al.

Table 4.5 Egg codes that are used to indicate stages of development Sa, Sb, Sc, Sd and So of the sea ice Description/age New ice (frazil, grease, slush, shuga) Nilas, ice rind Young Gray Gray – white First year ice Thin first year ice Medium first year ice Thick first year ice Old-survived at least one melt season Second year ice Multi-year ice Ice of land origin

Thickness 2 m >2 m

Code 1 2 3 4 5 6 7 1. 4. 7. 8 9

Table 4.6 Egg codes for forms of ice Description New ice Pancake ice Brash ice Ice cake Small ice floe Medium ice floe Big ice floe Vast ice floe Giant ice floe Fast ice Ice of land origin Undetermined or unknown

Size 0. These points determine this margin and are called support vectors (Cortes and Vapnik 1995). Thus, the good separation of the classes is achieved by a hyperplane, which has the greatest distance to the nearest point of the training samples of any class, since, in general, the error of the classifier is decreasing with the increasing of the distance. SVM performs a non-linear classification using the kernel trick. The kernel function may transform the data into a higher dimensional space to make this nonlinearly separation possible when the relation between class labels and attributes

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

277

is nonlinear. A common choice is a Gaussian kernel. The radial basis function kernel (RBF kernel) is found to work well in a wide variety of applications. The scikit-learn open source was used to implement the SVM classification method (http://scikit-learn.org/stable/index.html). Basically, SVM trains the model using low-level method and can only solve binary classification problems. In the case of multi-class classification, SVM model implements the “one-against-one” technique by fitting all binary sub-classifiers and finding the correct class by a voting mechanism. The effectiveness of SVM training depends on the selection of kernel, the kernel’s parameter (γ) and margin parameter C. The software provides a simple tool to check a grid of parameters obtaining cross-validation accuracy for each parameter setting: the parameters with the highest cross-validation accuracy are returned (Hsu et al. 2016). The SVM parameters in our case were γ ¼1 and C ¼ 1.

6.2.4.2

Support Vector Machine Based Algorithm: Sea Ice Type Classification

Sentinel-1 SAR Extra Wide swath mode scenes acquired during winter conditions were used for sea ice types classification in the Kara Sea. The EW mode wich is simular to Wide Swath Mode of Envisat and ScanSAR (SCW) of RADARSAT-2 was designed for maritime use, particularly for imaging sea ice, with a nominal swath width about 400 km and 40  40 m pixel spacing resolution. EW scenes comprise five subswaths spanning an incidence angle ranging from 19 to 47
. They were acquired in dual-polarization (HH and HV). Scenes used for elaboration of classification algorithms had been processed to Level-1 ground range detected format prior to delivery (GRDM). Thirty six scenes over the Kara and Barents seas from 2015, 2016 and 2017 acquired during winter months (from November to May) were utilized to train and test the algorithm. Development of SVM-based automated sea ice type classification algorithm included several steps: (1) image pre-processing; (2) analysis of 36 SAR images by experienced ice experts for delineation of the polygons of ice types to be classified; (3) calculation of texture features using sliding window and interpixel distance; (4) training SVM using obtained data set. Step 1 consisted of Level-1 SAR EW data pre-processing. HV band resulting from several narrower SAR beams has the visible modulation of the image intensity in range direction throughout the entire scene. Backscatter compensation at HH band is required since the signal generally decreases with incidence angle increasing. For HV band Sentinel-1 images were processed by removing thermal noise using noise estimate values provided in the image annotation data sets. Variability in backscatter with the incidence angle at HH band was compensated by angular correction technique described above. Then bands HH and HV were calibrated to obtain σ
. At the second step, the various sea ice types were selected by ice experts and the number of homogenous areas were delineated for each ice type in the images. Predefined classes obtained by manual classification of the SAR images by sea ice expert were defined as LFYI, thin FYI, medium and thick FYI, young ice and OW under various wind conditions. LFYI was identified visually based on σ
and textural

278

N. Y. Zakhvatkina et al.

Fig. 6.11 Sea ice classification of Sentinel-1 EW images, taken over the Kara Sea and eastern part of the Barents Sea on April 25, 2018: (a) mosaic of original SAR images at HH polarization; (b) sea ice types classification results

analysis since typically it has a low σ
and small texture variation. FYI medium and thick have large texture variations due to ridges and ice floe appearance. Note that visual identification of medium and thick FYI in SAR images practically is very subjective and sometimes difficult, thus these ice types were merged into one class. Young ice was visually classified as an ice contained both grey-white and grey ice, and broken ice on the ice edge as well as pancake ice (if appeared). The class ‘open water’ included the two subclasses, namely OW with high wind speed conditions and ON. At the third step, the second-order image texture features were derived from σ
at HH and HV polarizations using the GLCM method developed by Haralick et al. (1973). Feature extraction routines were implemented in Python, using the freely available MAHOTAS library (Coelho 2013). The efficient parameters of texture feature calculation (sliding window size, co-occurrence distance, etc.) were chosen based on the analysis presented above and defined as: sliding window size W ¼ 32  32; co-occurrence distance d ¼ 8; number of grey levels K ¼ 32; moving step of the sliding window ¼8. The optimal set of texture characteristics consisted of the next features: energy, contrast, variance, homogeneity, sum average, sum entropy, entropy, difference variance for HH polarization; and contrast, correlation, variance, sum average, sum variance, entropy, difference entropy, information measures of correlation for HV polarization. The number of features amounted to 16. The calculated texture features have been collocated with the manually identified classes in 36 pre-processed SAR images and the archived data sets were used as an input data for training the SVM classifier. During the final, fourth step, several test versions of SVM for ice types recognition were created and tested. SAR images used for training were employed to test prepared SVM. Based on the visual analysis of classification results, the version, where in the most cases SVM gave stable satisfactory result, was chosen. The

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

279

resulting SVM was implemented for automatic sea ice classification to divide SAR scene into the predefined classes. For illustrating developed SVM-based algorithm, an automated classification of 12 consecutive Sentinel-1 EW SAR scenes acquired on April 25, 2018 from the successive orbital passes has been performed (Fig. 6.11). Detailed examination of the image showed that it contained essentially three surface types: FYI medium, FYI thin of consolidated LFYI floes with mixture of ridges and leads, and young ice. The classified image is shown in Fig. 6.11b. The percentage of new ice which included leads with nilas and OW, refrozen leads, defined as regions with very high backscatter values (Fig. 6.11a), is higher in the north and northern-east of Novaya Zemlya that is suggested by the automated classification results. Nevertheless, the visual analysis of the results showed that the percentage of errors of ON in the north of Novaya Zemlya does not correspond to the actual quality of display. Such errors of classification are apparently caused by the residual HV thermal noise artifacts. For validation of these results it is important to bear in mind that it’s very difficult to compare qualitatively the highly-detailed ice charts with the manual ice charts. Therefore, the validation was not performed here.

6.2.4.3

Support Vector Machine Based Algorithm: Sea Ice-Open Water Discrimination

The algorithm for ice-water classification has been developed for production of ice-water maps using RADARSAT-2 SCW data. Based on the texture features SVM was trained for sea ice edge detection near MIZ in winter conditions. The algorithm processing chain included the following steps: (1) SAR data pre-processing with incidence angular correction for HH polarisation, reduction of thermal noise effect for HV polarisation, and absolute RADARSAT-2 image calibration to obtain σ
values for both channels; (2) Manual classification of SAR images by three experienced ice experts for delineation of polygons of the ice types to be classified (e.g., open water and ice with subclasses); (3) Calculation of texture features for HH and HV images; (4) Training of SVM classifier; (5) Validation of the trained automatic classifier. For the algorithm development 24 RS2 SCW scenes around Svalbard for 2011 and 2012 were used. The first step – SAR data pre-processing – has been performed using the same approach as for the other automatic algorithms described above. The second step included manual classification of SAR images into predefined classes. The main class ‘sea ice’ was chosen to include the following subclasses: (i) young ice, first-year and multiyear ice; (ii) fast ice; and (iii) broken ice on the edge (border) mixed with the ice-free areas (mostly found in the marginal ice zone). The class ‘open water’ included three subclasses: (i) open water with high wind speed conditions; (ii) open water with very high wind speed conditions; and (iii) mixture of calm open water, frazil ice, leads and nilas. For the final product the subclasses were merged into the main classes ‘sea ice’ and ‘open water’ since the similarities between the subclasses are too high for a reliable discrimination without additional data.

280

N. Y. Zakhvatkina et al.

Fig. 6.12 Sea ice-water SVM classification: (а) RS2 SCW SAR image (HH–polarization), taken over the southern part of Svalbard on 14 March 2013; (b) Collocated manual ice concentration chart provided by the Norwegian Ice Service (met.no) for the same day; (c) Results of the automatic SVM classification. (Adapted from Zakhvatkina et al. 2017)

Texture features were calculated at the third step based on the different parameter settings. Values of parameters for which the separation of normalized texture values for the classes increased were selected as optimal. They were the following: number of gray levels K ¼ 32, co-occurrence distance d ¼ 8, sliding window size W ¼ 64  64, moving step of the sliding window step ¼ 16. These parameter values were applied for calculations of all sets of texture features. Then, by means of the visual analysis, texture features which provided for the best discrimination between the ice and water classes have been selected for using in the classifier. They were: for HH channel - energy, inertia, cluster prominence, entropy, 3rd statistical moment of brightness, backscatter, and standard deviation; for HV channel- energy, correlation, homogeneity, entropy, and backscatter. At the fourth step, the SVM classifier has been traind using several pre-processed SAR images with the manually identified classes. After completing the training procedure, the resulting SVM has been applied for automatic sea ice-water classification. The final step included validation of the classification results using manual sea ice charts produced by the operational ice service at the Norwegian Meteorological Institute (MET Norway). The algorithm performance is illustrated in Fig. 6.12. The automatic SVM classification was applied to the RS2 SCW SAR image, taken over the southern part of Svalbard on 14 March 2013. This image contains several ice types, open water under different wind conditions and land. The ice-covered areas and the rough OW areas appear both bright in the HH polarisation and, therefore, practically are hardly distinguishable. Applying HV polarisation, however, provides additional information since OW areas there appear generally darker than sea ice. This is one of the major advantages of the using dual polarization.

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

6.2.5

281

Validation of Automatic Classification Algorithms

Validation of Arctic sea ice classification results is a challenging task since validation and ground truth data for large swath and high resolution SAR imagery is difficult to obtain. Verification of SAR image derived sea-ice information could be estimated by the coordinated acquisition of relevant reference data. In our study, for the validation of automatic classification algorithms we used visual inspection of SAR images, manually derived sea-ice charts from AARI and MET Norway, ice concentration maps from passive microwave radiometry. AARI ice charts (www.aari.ru) depict distribution of open water and sea ice types – nilas, young, first-year and old ice during the winter period (November – May). These manual ice charts are based on the generalization of regional ice charts, compiled from visible, infra-red and radar satellite images and reports from coastal stations and ships. They are issued every Thursday using data collected and averaged for preceding 2–5 days; the regional chats are focused on the marginal seas along the Northern Sea Route. Data are distributed in SIGRID format and classified into sea ice stages of development according to standards of the WMO. Access to archive is provided from December 2007 via http://www.aari.ru/odata/_d0015.php?lang¼1& mod¼0 or https://nsidc.org/data/G02176). Manual sea ice charts produced by the operational ice service at the Norwegian Meteorological Institute (MET Norway, http://polarview.met.no/) are freely available over the Arctic. MET Norway produces ice charts every workday using the following data sources: high resolution SAR sensors like Sentinel-1 and RADARSAT-2, low resolution microwave SSM/I and SSMIS data (DMSP satellites), MODIS images (Terra and Aqua satellites) and AVHRR data from NOAA satellites. Data are polygonized and classified into six concentration classes according to standards of the World Meteorological Organization: open water – 0–1/10, very open drift ice – 1/10–3/10, open drift ice – 4/10–6/10, close drift ice – 7/10–8/10, very close drift ice – 9/10–10/10 and fast ice – 10/10. Confusion matrixes, describing how well an algorithm can classify data, are calculated to get a quantitative measure of the classification method accuracy. In our comparison, all ice charts are assumed to represent “true” classification and the confusion matrixes are calculated for accuracy evaluation of the algorithm results.

6.2.5.1

Neural Network Based Algorithm: Sea Ice Type Classification

Algorithm of automatic Neural Network based classification of sea ice types from Envisat ASAR images has been created and tested for the winter conditions in the high Arctic. The ice types selected for classification were multiyear ice, deformed FYI, level FYI and open water/nilas. For validation of this algorithm, classification results of 12 images were compared with the ice charts issued by AARI and results of manual classification by sea ice experts. Illustration of validation results of NN automatic sea ice classification is shown in (Fig. 6.13). Two Envisat WSM SAR

282

N. Y. Zakhvatkina et al.

Fig. 6.13 Validation of NN classification of a fragment of Envisat WSM SAR images (HH– polarization), taken over the northern part of the Kara Sea: (а) NN-classification results of image on 18 January 2008 at 09:55 UTC (origional SAR image is shown in Fig. 6.9a); (b) ice chart provided by AARI for 16–18 January 2008; (c) manual classification by the sea ice expert; (d) NN clasifcation results of the image on 4 February 2008 at 11:00 UTC (origional SAR image is shown in Fig. 6.9d); (e) ice chart provided by AARI for 6 February 2008; (f) manual classification by the sea ice expert. The lower legend in (b, e) is for NN classification, the upper legend is for the ice chart. (Source: Zakhvatkina 2009)

Table 6.2 Confusion matrices for comparison of NN classification results of Envisat ASAR images with the visual expert classification

Visual expert classification* MYI LFYI* DFYI*

4 February 2008 (Fig. 6.13d,f) ON

Ice type MYI LFYI DFYI ON

64.76 0.87 1.26 12.98

6.49 86.18 23.58 34.77

25.37 10.95 73.49 14.06

3.39 1.99 1.67 38.19

NN classific. results

NN classific. results

18 January 2008 (Fig. 6.13a,c)

Ice type MYI LFYI DFYI

Vsual expert classification** MYI LFYI* DFYI *+ thick FYI and 2nd-year ice 72.15 2.46 25.35 0.71 62.08 37.19 15.61 11.06 73.29

LFYI first-year ice, where ridges cover less then 20% of ice surface, DFYI first-year ice, where ridges cover more than 20% of ice surface

images (HH polarization), taken over the northern part of the Kara Sea, were used for this illustration. This is the same images which have already been classified earlier (see Fig. 6.9). Results of NN classification of these images (Fig. 6.13a, d) were compared with the AARI ice charts, where MYI and FYI were discriminated (Fig. 6.13b, e), and results of expetr manual classification (Fig. 6.13c, f; actually,

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

283

this is coloured ice type zones A, B, C, D, E defined by expert’s visual analisys and shown in Fig. 6.9b, e). Confusion matrices calculated for comparison of NN classification with the AARI ice charts showed quite good correspondence between them (diagonal elements of confusion matrices): 79.3% for MYI, and 92.2% for FYI. The correspondence between NN and expert manual classifications has been estimated from other confusion matrices calculated for this comparison (Table 6.2). For the first image, of 18 January 2008, correspondence amounted 86% for LFYI, 73% for DFYI, 65% for MYI, and 38% for ON. This is a relatively good result because the ice structure is very complex and the expert analysis does not take into account all the small features in the image. For the second image, of 4 February 2008, correspondence amounted 72% for MYI, 62% for LFYI and 73% for DFYI in NN classification and DFYI + thick FYI and 2nd-year ice in the manual expert analysis. This is also quite good correspondence. The NN classification results were also validated using field observations near the “North Pole-35” drifting station. Airplane observations and in situ measurements showed that the ice floe where the “North Pole-35” station was located had a mixture of MYI floes of various size, thickness and configuration. In many cases MYI floes were separated by stripes of FYI with inclusions of broken MYI. The SAR signature of such mixture of MYI and FYI can be generalized to be valid for the larger areas.

6.2.5.2

Neural Network Based Algorithm: Sea Ice-Open Water Discrimination

The NN algorithm for ice-water classification described in Sect. 6.2.3 has been validated using ~300 Envisat ASAR images taken over the Fram Strait in the period from 17 January 2011 to 31 March 2012 (exept August and September). For each SAR image an error matrix between algorithm results and manual sea ice product produced by MET Norway has been calculated. To allow direct comparison, the MET Norway products were ranged into two classes: sea ice concentration values below 15% are classified as OW, and higher concentration values are classified as sea ice. Example of validation is shown in Fig. 6.14. The error matrix is represented here as an image (Fig. 6.14c) with the following three classes: no difference; sea ice error (sea ice at MET Norway chart, OW at our results); OW error (OW at MET Norway chart, sea ice at our results). The slight disagreement in the exact location of the ice-water boundaries is demonstrated by the yellow line (Fig. 6.14c). Validation of this algoritm showed that the overall accuracy was 97.5  0.77%.

284

N. Y. Zakhvatkina et al.

Fig. 6.14 Validation of automatic NN ice-water classification of the Envisat ASAR HH image taken over the Fram Strait on 9 February 2011 at 11:01 UTC (see image in Fig. 6.10a): (a) results of the NN classification with delineation of two classes: water (yellow), sea ice (dark brown); (b) ice chart of MET Norway reclassified into two classes: open water (ice concentration from 0 to 15%; yellow colour) and sea ice (ice concentration from 15% to 100%; dark brown colour); (c) difference between recalculated MET Norway chart and classification results representing error matrix as “image”: no difference is light blue, sea ice error (sea ice in MET Norway chart, OW in our results) is yellow, OW error (OW in MET Norway chart, sea ice in our results) is dark blue

6.3

Sea Ice Drift Retrieval from SAR

The high spatial resolution of order tens of meter in present SAR images makes it possible to localize and analyze dynamical features that are of great importance in sea ice studies. Furthermore, the repeated wide swath SAR coverage of the same sea ice areas, with typical one- to three-day interval as provided by the Sentinel-1

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

285

satellites, allows calculation of ice motion rather systematically in large parts of the Arctic Ocean. There are three main categories of sea ice motion estimation approaches: (i) differential methods, (ii) aerial methods and (iii) feature tracking. The first group of methods estimates a dense motion field by approximating the flow displacement for every pixel (Sun 1996; Petrou and Tian 2017). The second group uses measurement of the similarity between image blocks. Classical work by this approach has been done by Fily and Rothrock (1987) and then further developed by (Komarov and Barber 2014; Karvonen 2012). The typical resolution of the final ice drift product from wideswath data is about 5–10 km, providing useful day on sea ice dynamics on regional and basin-wide scale. Feature-based methods rely on extraction and recognition of features from sequential images with focus on optimizing feature correspondence. Since the 1990s, several feature-tracking algorithms for ice drift retrieval have been proposed (Demchev et al. 2017; Muckenhuber et al. 2016; Giles et al. 2011; Daida et al. 1990; Liu et al. 1997, McConnell et al. 1991). The feature-based technique is able to capture strong deformations, but only with sparse distribution of tracked features. Other studies aimed to combine the advantages of areal and feature-based methods (Korosov and Rampal 2017; Berg and Eriksson 2014; Kwok et al. 1990; Vesecky et al. 1988). Majority of the existing sea ice motion algorithms are based on cross-correlation between two images covering the same ice are with a time interval of a few days. Recently, an ice motion tracking system has been implemented in (Komarov and Barber 2014) for the Canadian Ice Service, which captures both the translational and rotational components of the ice motion. The method is based on combination of phase- and cross-correlation matching techniques. The method uses several resolution levels generated from the original SAR images. At each resolution level a set of control points is automatically generated based on local variances in the SAR image. Thus, various ice features (e.g. cracks and ridges) suitable for tracking are automatically identified. Such an approach makes the probability of finding matches for these control points in the second image higher compared to the regular grid approach (Kwok et al. 1990; Fily and Rothrock 1987). Prior to the ice motion tracking procedure, a Gaussian filter and a Laplace operator are sequentially applied to images at each resolution level in order to highlight the edges and other heterogeneities. Also, the Laplacian operator reduces the signal trend across the satellite track due to the backscatter dependence on the incidence angle. The ice tracking algorithm starts with the lowest resolution level to determine preliminary ice motion vectors. To derive ice feature matches, the phase-correlation (Reddy and Chatterji 1996) and cross-correlation techniques (Fily and Rothrock 1987) are combined. Such an approach makes it possible to find the translational and rotational components of ice motion from the phase-correlation technique as well as to quantitatively estimate the similarity between two sub-images. To eliminate erroneous ice motion vectors the algorithm compares ice drift vectors derived from the forward pass (tracking from the first image to the second one) and the backward pass (tracking from the second image to the first one). If the absolute value of the sum between the forward and backward drift vectors exceeds a threshold (1 pixel), then the vector is eliminated. Following this, additional filtering is performed by thresholding the cross-correlation coefficients. For the rest of the vectors, a

286

N. Y. Zakhvatkina et al.

Fig. 6.15 Sea ice motion example from sequential Sentinel-1A images acquired over Ob Bay, for 1st – 2nd December 2017 by ice tracking system at Canadian Ice Service. Red, yellow, and green colours indicate low, medium, and high levels of confidence

confidence level (low, medium, and high) is set up for each output drift vector based on its cross-correlation coefficient. The remaining vectors are further filtered by thresholding their cross-correlation coefficients. At each consecutive resolution, level the tracking algorithm is guided by the ice drift vectors found at the previous resolution level, and thereby, the ice motion field is refined. Both forward and backward passes are conducted at each resolution level to eliminate erroneous vectors. An example of ice motion results derived from sequential Sentinel-1A images acquired over the Ob Bay on December1st and 2nd, 2016 is shown in Fig. 6.15. The accuracy of the ice motion system was thoroughly examined based on the visual detection of similar ice features in sequential SAR images and ground truth GPS beacon data for various ice conditions. Very good agreement between the SAR derived vectors and ice drifting beacon trajectories located in the close proximity (less than 3 km) to the nearest SAR ice motion vectors was reported by Komarov and Barber (2014). The root-mean square error (RMSE) was 0.43 km for 36 comparison points. Ice motion can be retrieved from both co-polarization (HH) and crosspolarization (HV) channels of RADARSAT-2 ScanSAR images. Komarov and Barber (2014) demonstrated that ice motion retrieval from HV images could provide significantly more vectors compared to the co-polarization (HH) channel in situations where the HV signal is sufficiently higher than the noise equivalent sigma zero (NESZ). For example, such situations could occur over winter multiyear sea ice where high levels of HV signal are often observed. The study by Komarov and Barber (2014) showed that if the difference between HV signal and NESZ exceeds 0.003 (in linear units) than the HV channel provides a larger number of ice motion vectors compared to the HH channel. Otherwise, the HH channel is more useful for ice motion detection compared to HV. The Canadian ice motion tracking system has proven to be instrumental in many sea ice applications. Recent changes in the

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

287

exchange of sea ice between the Arctic Ocean and the Canadian Arctic Archipelago using this system were quantified in (Wohlleben et al. 2012; Howell et al. 2013a, b). Ice motion data provided by the system was used to study hazardous ice features such as ice islands and thick MYI in the Beaufort Sea (Barber et al. 2014). The system has been used for verification of numerical sea ice prediction models at Environment and Climate Change Canada (ECCC). Another example of operational ice tracking system using pattern matching technique has been developed by Danish Technical University (DTU) (Pedersen et al. 2015). The system provides ice displacement vectors on a regular grid, which is convenient for many practical applications such as ice deformation calculations. This algorithm does not use multiple resolutions, but runs on 300 m resolution SAR data. The search area is 40 km in radius, but this may also depend on the time difference between the two images. The algorithm uses normalized cross correlation to measure the similarity between image patches and provides ice displacement vectors on a regular grid at every 10 km on the daily basis. The inherent nature of radar backscatter due to multiplicative speckle noise reduces the quality of the SAR images. This has negative impact on feature extraction and image matching (Dellinger et al. 2015; Wang et al. 2012). For successful retrieval of sea ice parameters from SAR images, it is essential to take the speckle noise into account. Standard approaches for feature detection and description, such as SURF (Bay et al. 2008) and Scale Feature Invariant Transform (Lowe 2004), can give reliable results for optical images, but not for SAR images. That is because these methods assume that the noise in the SAR image has Gaussian distribution and therefore uses Gaussian scale space representation for feature extraction. Gaussian blurring does not preserve important image details and blurs to the same extent both details and noise (Perona and Malik 1990). To mitigate this problem (Demchev et al. 2017) proposed to use Accelerated-KAZE (A-KAZE) features (Alcantarilla et al. 2013) for ice drift retrieval from SAR data. This approach exploits the advantages from nonlinear scale space representations for feature extraction. The different statof-the-art feature techniques from Computer Vison such as SIFT and Oriented FAST and Rotated BRIEF (ORB) (Rublee et al. 2011) have been compared. In order to evaluate the computational and tracking efficiency of each of the compared methods,

Fig. 6.16 Computation time and tracking performance for A-KAZE, SIFT and ORB features. (Source: Demchev 2018)

288

N. Y. Zakhvatkina et al.

Fig. 6.17 Sea ice displacement field obtained by feature-based ice drift algorithm from a pair of Sentinel-1A images for 19–20 February 2016 over Lincoln Sea

a timing evaluation analysis have been performed considering ice feature tracking for a pair of sequential SAR images of resolution 5.686  5.686 pixels. A comparison between the A-KAZE, SIFT and ORB algorithms, was performed for cases where 1000, 2000 and 10,000 features were detected from the same pair of SAR images. The number of successfully tracked features is also shown (Fig. 6.16). It should be noted that the run-time of the algorithms greatly differs depending on the number of detected features and image resolution. Nevertheless, the running time ratio between the different algorithms remained roughly the same. A-KAZE was approximately three times faster than SIFT, while ORB was about ten times more time-efficient than SIFT. At the same time, there is a quite different picture in terms of successfully tracked features. A-KAZE provided about two to three times more correspondences than SIFT and ORB under the same conditions. SIFT however efficiently tracked more features than ORB. Therefore, ORB could be considered as the most time-efficient method and seem to be reliable in real-time applications with hard constraints on execution time, but its performance on feature tracking on SAR images is limited. SIFT demonstrated good applicability for ice drift tracking when filtering by anisotropic diffusion as an initial step is applied, while A-KAZE exhibits the best compromise between tracking performance and execution time compared to

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

289

Fig. 6.18 Tuned ORB (first column, 6920 vectors), SIFT (second column, 1585 vectors) and SURF has been obtained (third column, 518 vectors) (a) drift vectors, (b) number of vectors per grid cell and (c) root mean square distance in km. (Source: Muckenhuber et al. 2016)

SIFT and ORB as shown in Fig. 6.16. Example of ice drift speed field obtained by the A-KAZE-based algorithm shown in Fig. 6.17. A study by Muckenhuber et al. (2016) analyze the applicability of the ORB features for ice drift retrieval. Here, the ORB is adopted and tuned from the test cases with Sentinel-1 SAR images representative of sea ice conditions between Greenland and Severnaya Zemlya during winter/spring. The performance of the algorithm is compared to two other feature tracking algorithms (SIFT and SURF) (Fig. 6.18). Applied on a test image pair acquired over Fram Strait, the tuned ORB algorithm produces the highest number of vectors (6920, SIFT: 1585 and SURF: 518) while being computationally most efficient (66 s, SIFT: 182 s and SURF: using a 2.7 GHz processor with 8 GB memory). For validation purpose, 314 manually drawn vectors have been compared with the closest calculated vectors, and the resulting root mean square error of ice drift is 563 m. All test image pairs show significantly better performance of the HV channel. On average, around four times more vectors have been found using HV polarization.

290

N. Y. Zakhvatkina et al.

Fig. 6.19 Example of ice drift field for the period between 2nd and 4th March, 2016 retrieved using the combined ice drift algorithm. (Source: Korosov and Rampal 2017)

Several studies have demonstrated use of combined algorithms. The first operational ice tracking algorithm was developed at the Alaska SAR Facility (Kwok et al. 1990). The system combined several procedures and algorithms for locating image pairs and for tracking sea ice in both the central pack and marginal ice zones using pattern and Ψ -s feature matching, respectively. Typically, the algorithm retrieved ice displacement vectors at every 5–10 km. Following the same idea, Berg and Eriksson (Berg and Eriksson 2014) proposed an advanced segmentation technique prerequisite for feature tracking. Recently, the promising results has been obtained using another hybrid algorithm by Korosov and Rampal (2017). The main advantage of the combination is that the feature tracking rapidly provides the first guess estimate of ice drift in a few unevenly distributed key points, and pattern matching accurately provides drift vectors on a regular or irregular grid. Based on thorough sensitivity analysis of the algorithm, optimal sets of parameters are suggested for retrieval of sea ice drift on various spatial and temporal scales. Due to its features, the method is suitable for validation purposes and assimilation into ice models. An example of the ice drift field over the Wandel Sea from a pair of Sentinel-1A images, shown in Fig. 6.19. Sea ice drift retrieval from satellite data still unsolved, challenging and one of the crucial task related sea ice study. Past, present and future satellite missions, particularly SAR, with their great capabilities is opening-up new horizons for sea ice retrieval and analysis. Probably, the new methods will combine existing approaches include pattern or feature matching and innovative techniques that takes the physical properties of the SAR imaging into account.

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

6.4

291

Summary

With the increasing interest to study sea in Arctic and the large amount of SAR data provided by the Sentinel-1 satellites, it is necessary to develop automatic retrieval algorithms for sea ice variables. A number of algorithms have been proposed and tested for monitoring of the state of the sea ice cover of the Arctic. These algorithms have all their advantages and disadvantages related to applicability in different regions, seasons and specific ice conditions. The operational sea ice charting services use satellite SAR images as the major data source, but the ice chart production is still done manually by ice analysts. With the large amount of SAR data available every day, the manual ice analysis is very labor intensive. It is therefore necessary to develop automated ice classification and ice drift algorithms for SAR images. Currently several automatic algorithms for the classification of SAR images of sea ice based on NN and SVM have been proposed. Sea ice cover is highly heterogeneous and the differences between the different ice types are less pronounced in SAR images, thus, using only backscatter of sea ice would lead to misclassification of SAR images. To support the identification of different ice types, the use of texture features is a good solution. In that case, a set of texture features is analysed to select the effective number of features with preferred computational parameters. In the future, automatic algorithms for the classification of sea ice may become an integral part of an automated integrated system for processing of satellite information intended for monitoring of the polar regions, including land areas and the Arctic Ocean.

References Abreu, R. D. (2000). RADAR sea ice signatures: An operational primer. In Proceedings of Workshop on mapping and archiving of sea ice data – the expanding role of radar, Ottawa, Canada, May 2–4 2000. JCOMM Tech. Rep. № 7, 85–94. Absolute calibration of ASAR Level 1 products. (2004). Issue 1, revision 5, https://earth.esa.int/c/ document_library/get_file?folderId¼38042&name¼DLFE-661.pdf Agnew, T., & Howell, S. (2003). The use of operational ice charts for evaluating passive microwave ice concentration data. Atmosphere-Ocean, 41, 317–331. Albregtsen, F. (2008). Statistical texture measures computed from gray level coocurrence matrices. Image Processing Laboratory Department of Informatics, University of Oslo. Alcantarilla, P. F., Nuevo J., Bartoli A. (2013). Fast explicit diffusion for accelerated features in nonlinear scale spaces. British Machine Vision Conference (BMVC). Aldenhoff, W., Heuzé, C., & Eriksson, L. E. B. (2018). Comparison of ice/water classification in Fram Strait from C- and L-band SAR imagery. Annals of Glaciology, 59, 112–123. https://doi. org/10.1017/aog.2018.7. Alexandrov, V. Y., & Piotrovskaya, N. Y. (2008a). Assessment of the backscatter for various sea ice types from ENVISAT ASAR images. Earth Observation from Space, 4, 3–11. in Russian. Alexandrov, V. Y., & Piotrovskaya, N. Y. (2008b). Digital processing of ENVISAT ASAR images of the sea ice. Problems of Arctic and Antarctic, 1(78), 90–94. in Russian.

292

N. Y. Zakhvatkina et al.

Alexandrov, V. Y., Sandven, S., Kloster, K., Bobylev, L. P., & Zaitsev, L. V. (2004). Comparison of sea ice signatures in OKEAN and RADARSAT radar images for the northeastern Barents Sea. Canadian Journal of Remote Sensing, 30(6), 882–892. Arkett, M., Flett, D., Roger, De A. (2006). Sea ice type and open water discrimination for operational ice monitoring with RADARSAT-2. In: Proc. of 2006 IEEE International Symposium on Geoscience and Remote Sensing. Arkett, M., Flett, D., Roger, De A. (2007). C-band multiple polarization SAR for ice monitoring – what can it do for the Canadian ice service. In: Proc. of ENVISAT Symposium 2007, Montreux, Switzerland, 23–27 April 2007 (ESA SP-636). Askne, J., Carlstrom, A., Dierking, W., Ulander, L. (1994). ERS-1 SAR backscatter modeling and interpretation of sea ice signatures. In: Proc. of IGARSS, Pasadena, California, Aug. 8–12, 162–164. Atkinson, P. M., & Tatnall, A. R. L. (1997). Neural networks in remote sensing. International Journal of Remote Sensing, 18(4), 699–709. Balss, U., Breit, H., & Fritz, T. (2010). Noise-related radiometric correction in the TerraSAR-X multimode SAR processor. IEEE Transactions on Geoscience and Remote Sensing, 48(2), 741–750. Baraldi, A., & Parmiggiani, F. (1995). An investigation of the textural characteristics associated with gray level cooccurrens matrix statistical parameters. IEEE Transactions on Geoscience and Remote Sensing, 33(2), 293–304. Barber, D. G., McCullough, G., Babb, D., Komarov, A. S., Candlish, L. M., Lukovich, J. V., Asplin, M., Prinsenberg, S., Dmitrenko, I., & Rysgaard, S. (2014). Climate change and ice hazards in the Beaufort Sea. Elementa: Science of the Anthropocene, 2, 000025. https://doi.org/ 10.12952/journal.elementa.000025. Bay, H., Ess, A., Tuytelaars, T., & Van Gool, L. (2008). Speeded-up robust features (SURF). Computer Vision and Image Understanding, 110(3), 346–359. Beaven, S. G., Gogineni, S. P., & Shanablen, M. (1994). Radar backscatter signatures of thin ice in the Central Arctic. International Journal of Remote Sensing, 15(5), 1149–1154. Belchansky, G. I., Duglas, D. S., & Mordvintsev, I. N. (1998). Identification of sea ice types using radar data from “OKEAN-01” satellites and Look-Up Tables. Earth Observation from Space (Issledovanie Zemli iz Kosmosa), 6, 52–65. In Russian. Benediktsson, A., Swain, P. H., & Ersoy, O. K. (1990). Neural network approaches versus statistical methods in classification of multisource remote sensing data. IEEE Transactions on Geoscience and Remote Sensing, 28(4), 540–552. Berg, A., & Eriksson, L. E. (2012). SAR algorithm for sea ice concentration evaluation for the Baltic sea. IEEE Geoscience and Remote Sensing Letters, 9(5), 938–942. https://doi.org/10. 1109/lgrs.2012.2186280. Berg, A., & Eriksson, L. E. B. (2014). Investigation of a hybrid algorithm for sea ice drift measurements using synthetic aperture radar images. IEEE Transactions on Geoscience and Remote Sensing, 52(8), 5023–5033. Bogdanov, A. V., Sandven, S., Johannessen, O. M., Alexandrov, V. Y., & Bobylev, L. P. (2005). Multisensor approach to automated classification of sea ice image data. IEEE Transactions on Geoscience and Remote Sensing, 43, 1648–1664. Borodachev, V. E. (1998). Ice of Kara Sea. St. Petersburg, Hydrometeoizdat, 182 p. (in Russian). Brekke, C., Holt, B., Jones, C., & Skrunes, S. (2014). Discrimination of oil spills from newly formed sea ice by synthetic aperture radar. Remote Sensing of Environment, 145, 1–14. Bushuev, A. V., Bychenkov Yu. D., Loschilov V.S., Masanov A. D. (1983). Investigation of ice cover with help of side looking radar (SLR). Hydrometeoizdat, Leningrad, 120 pp. (in Russian). Bushuev, A. V., Volkov, N. A., & Grischenko, V. D. (1995). Sea ice observations and studies, construction of the automated ice information system. Problems of Arctic and Antarctic, 70, 104–119. in Russian. Clausi, D. A. (2002). An analysis of co-occurrence texture statistics as a function of grey level quantization. Canadian Journal of Remote Sensing, 28, 45–62.

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

293

Clausi, D. A., & Yue, B. (2004). Comparing cooccurrence probabilities and Markov random fields for texture analysis of SAR sea ice imagery. IEEE Transactions on Geoscience and Remote Sensing, 42(1), 215–228. https://doi.org/10.1109/TGRS.2003.817218. Clausi, D. A., Qin, A. K., Chowdhury, M. S., Yu, P., & Maillard, P. (2010). MAGIC: MAp-guided ice classification system. Canadian Journal of Remote Sensing, 36., suppl. 1, S13–S25. Coelho, L. P. (2013). MAHOTAS: Open source software for scriptable computer vision. Journal of Open Research Software, 1(1). Cornford, D., Nabney, I. T., & Ramage, G. (2001). Improved neural network scatterometer forward models. Journal of Geophysical Research, 106(C10), 22331–22338. Cortes, C., & Vapnik, V. (1995). Support-vector networks. Machine Learning, 20, 273–297. Daida, J., Samadani, R., & Vesecky, J. F. (1990). Object-oriented feature tracking algorithms for sar images of the marginal ice zone. IEEE Transactions on Geoscience and Remote Sensing, 28(4), 573–589. Dellinger, F., Delon, J., Gousseau, Y., Michel, J., & Tupin, F. (2015). Sar-sift: A sift-like algorithm for sar images. IEEE Transactions on Geoscience and Remote Sensing, 53(1), 453–466. Demchev, D. (2018). Retrieval, analysis and monitoring methods for sea ice and iceberg drift in the Arctic using radar data. PhD thesis (submitted), St. Petersburg State University, Dissertation in Oceanology, Russia, 244 p. Demchev, D., Volkov, V., Kazakov, E., Alcantarilla, P. F., Sandven, S., & Khmeleva, V. (2017). Sea ice drift tracking using A-KAZE features. IEEE Transactions on Geoscience and Remote Sensing, 55(9), 5174–5184. Deng, H., & Clausi, D. A. (2005). Unsupervised segmentation of synthetic aperture Radar sea ice imagery using a novel Markov random field model. IEEE Transactions on Geoscience and Remote Sensing, 43(3), 528–538. Dierking, W. (2010). Mapping of different sea ice regimes using images from Sentinel-1 and ALOS synthetic aperture radar. IEEE Transactions on Geoscience and Remote Sensing, 48, 1045. Dierking, W. (2013). Sea ice monitoring by synthetic aperture radar. Oceanography, 26, 100–111. Dierking, W., & 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), 957–970. Dierking, W., Skriver, H., Gudmandsen, P. (2003). SAR polarimetry for sea ice classification. Proc. of the POLinSAR Workshop. ESA, 14–16 Jan. 2003, Frascati, Italy, Special Publication: ESA SP– 529, pp. 109–118. Dierking, W., Skriver, H., Gudmandsen, P. (2004). On the improvement of sea ice classification by means of radar polarimetry. Proc. of the 23rd European Association of Remote Sensing Laboratories (EARSeL) Symposium, June 2–5 2004, Ghent, Belgium, pp. 203–209. Dokken, S. T., Markus, P. W. T., Askne, J., & Bjork, G. (2002). ERS SAR characterization of coastal polynyas in the Arctic and comparison with SSM/I and numerical model investigations. Remote Sensing of Environment, 80, 321–335. Drinkwater, M. R., Crawford, J. P., Cavalieri, D. J. (1991). Multi-frequency, multi-polarization SAR and radiometer sea ice classification. Proceeding of International Geoscience and Remote Sensing Symposium (IGARSS’91). New York, USA, 3–6 Jun 1991, Vol. I., pp. 107–111. Drinkwater, M. R., Kwok, R., Rignot, E., Israelsson, H., Onstott, R. G., Winebrenner, D. P. (1992). Potential applications of polarimetry to the classification of sea ice. In F. D. Carsey (Ed.), Microwave remote sensing of sea ice (Geophysical monograph 68) (pp. 419–430). Washington, DC: American Geophysical Union. Fetterer, F. M., Gineris, D., & Kwok, R. (1994). Sea ice type maps from Alaska synthetic aperture radar facility imagery: An assessment. Journal of Geophysical Research, 99(C11), 22,443–22,458. Fily, M., & Rothrock, D. (1987). Sea ice tracking by nested correlations. IEEE Transactions on Geoscience and Remote Sensing, (5), 570–580.

294

N. Y. Zakhvatkina et al.

Freeman, A. (1992). Radiometric calibration of SAR image data. XVII Congress for Photogrammetry and Remote Sensing (ISPRS). http://www.isprs.org/proceedings/XXIX/congress/part1/ 212_XXIX-part1.pdf. Geldsetzer, T., & van der Sanden, J. J. (2013). Identification of polarimetric and nonpolarimetric C-band SAR parameters for application in the monitoring of lake ice freeze-up. Canadian Journal of Remote Sensing, 39(3), 263–275. Geldsetzer, T., & Yackel, J. J. (2009). Sea ice type and open water discrimination using dual co-polarized C-band SAR. Canadian Journal of Remote Sensing, 35, 73–84. Giles, A., Massom, R., Heil, P., & Hyland, G. (2011). Semi-automated feature tracking of east Antarctic sea ice from Envisat ASAR imagery. Remote Sensing of Environment, 115(9), 2267–2276. Gill, R. S. (2001). Operational detection of sea ice edges and icebergs using SAR. Canadian Journal of Remote Sensing, 27(5), 411–432. Gill, R. S. (2003). SAR ice classification using fuzzy screening method. In Proc. of Workshop on Applications of SAR Polarimetry and Polarimetric Interferometry (POLinSAR), 14–16 January 2003, Frascati, Italy. Gill, J. P. S., & Yackel, J. J. (2012). Evaluation of C-band SAR polarimetric parameters for discriminating of first-year sea ice types. Canadian Journal of Remote Sensing, 38, 306–323. Greidanus, H., & Santamaria, C. (2014). First analyses of sentinel-1 images for maritime surveillance. JRC Science and policy reports, 56 pp. Haarpaintner, J., & Solbø, S. (2007). Automatic ice-ocean discrimination in SAR imagery, Norut IT-report, Tech. Rep. Hara, Y., Atkins, R. G., Shin, R. T., Kong, J. A., Yueh, S. H., & Kwok, R. (1995). Application of neural networks for sea ice classification in polarimetric SAR images. IEEE Transactions on Geoscience and Remote Sensing, 33(3), 740–748. Haralick, R. M., Shanmugam, K. S., & Dinstein, I. (1973). Textural features for image classification. IEEE Transactions on Systems, Man, and Cybernetics, 3, 610–621. Haverkamp, D., Soh, L. K., Tsatsoulis, C. 1993. A dynamic local thresholding technique for sea ice classification. In the 1993 IEEE International Geoscience and Remote Sensing Symposium (IGARRS’93), August 18–21, Tokyo, Japan. pp. 638–640. Haykin, S. (2006). Neural networks: A comprehensive foundation (2nd ed.). Englewood Cliffs: Prentice-Hall. Heerman, P. D., & Khazenie, N. (1992). Classification of multispectral remote sensing data using a back-propagation neural network. IEEE Transactions on Geoscience and Remote Sensing, 30 (1), 81–88. Heinrichs, J. F., Cavalieri, D. J., & Markus, T. (2006). Assessment of the AMSR-E ice concentration product at the ice edge using RADARSAT-1 and MODIS imagery. IEEE Transactions on Geoscience and Remote Sensing, 44, 3070–3079. Holmes, Q. A., Nuesch, D. R., & Shuchman, R. A. (1984). Textural analysis and real-time classification of sea ice types using digital SAR data. IEEE Transactions on Geoscience and Remote Sensing, 22(2), 113–120. Hong, D., & Yang, C. S. (2018). Automatic discrimination approach of sea ice in the Arctic Ocean using Sentinel-1 Extra Wide Swath dual-polarized SAR data. International Journal of Remote Sensing, 39(13), 4469–4483. Howell, S. E. L., Wohlleben, T., Dabboor, M., Dersken, C., Komarov, A. S., & Pizzolato, L. (2013a). Recent changes in the exchange of sea ice between the Arctic Ocean and the Canadian Arctic Archipelago. Journal of Geophysical Research, 118(7), 3595–3607. Howell, S. E. L., Wohlleben, T., Komarov, A. S., Pizzolato, L., & Derksen, C. (2013b). Recent extreme light sea ice years in the Canadian Arctic Archipelago: 2011 and 2012 eclipse 1998 and 2007. The Cryosphere, 7, 1753–1768. https://doi.org/10.5194/tcd-7-1313-2013. Hsu, C. -W., Chang, C. -C., & Lin, C. -J. (2016). A practical guide to support vector classification. Initial version: 2003. Last updated: May 19, 2016. http://www.csie.ntu.edu.tw/~cjlin/papers/ guide/guide.pdf.

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

295

Isleifson, D., Hwang, B., Barber, D. G., Scharien, R. K., & Shafai, L. (2010). C-band polarimetric backscattering signatures of newly formed sea ice during fall freeze-up. IEEE Transactions on Geoscience and Remote Sensing, 48(8), 3256–3267. Jefferies, B. (2012). RADARSAT-2—new ice information products. Proc. of the 13th Meeting— International Ice Charting Working Group, 15–19 October 2012, Tromso, Norway. Johannessen, O. M., Alexandrov, V., Frolov, I., Bobylev, L., Sandven, S., Miles, M., Pettersson, L., Kloster, K., Smirnov, V., Mironov, Y., & Babich, N. (2007). Remote sensing of sea ice in the Northern sea route: Studies and applications. Chichester: Springer-Praxis. Johansson, A. M., Brekke, C., Spreen, G., & King, J. A. (2018). X-, C-, and L-band SAR signatures of newly formed sea ice in Arctic leads during winter and spring. Remote Sensing of Environment, 204, 162–180. Kaleschke, L., & Heygster, G. (2004). Towards multi-sensor microwave remote sensing of frost flowers on sea ice. Annals of Glaciology, 39(1), 219–222. Karvonen, J. A. (2004). Baltic Sea ice SAR segmentation and classification using modified pulsecoupled neural networks. IEEE Transactions on Geoscience and Remote Sensing, 42(7), 1566–1574. Karvonen, J. (2010). C-band sea ice SAR classification based on segmentwise edge features. In P. Imperatore & D. Riccio (Eds.), Geoscience and remote sensing new achievements (pp. 129–146). IntechOpen. https://doi.org/10.5772/9093. [Online] Available: https://www. intechopen.com/books/geoscience-and-remote-sensing-new-achievements/c-band-sea-ice-sarclassification-based-on-segmentwise-edge-features Karvonen, J. (2012). Operational SAR-based sea ice drift monitoring over the Baltic Sea. Ocean Science, 8(4), 473–483. Karvonen, J., Simila, M., & Makynen, M. (2005). Open water detection from Baltic Sea ice Radarsat-1 SAR imagery. IEEE Transactions on Geoscience and Remote Sensing, 2, 275–179. Komarov, A. S., & Barber, D. G. (2014). Sea ice motion tracking from sequential dual-polarization RADARSAT-2 images. IEEE Transactions on Geoscience and Remote Sensing, 52(1), 121–136. Korosov, A., & Park, J. -W. (2016) Very high resolution classification of Sentinel-1A data using segmentation and texture features. Proc. of European Space Agency Living Planet Symposium, 9–13 May 2016, Prague, Czech Republic, ESA-SP Vol. 740, pp.44. Korosov, A., & Rampal, P. (2017). A combination of feature tracking and pattern matching with optimal parametrization for sea ice drift retrieval from SAR data. Remote Sensing, 9, 258. https://doi.org/10.3390/rs9030258. Korosov, A., Hansen, M. W., Yamakava, A. (2015a). Nansat – scientist friendly toolbox for processing satellite data. World Ocean Scientific Congress, 2–8 February, 2015, Cochin, India. Korosov, A., Zakhvatkina, N., Muckenhuber, S. (2015b) Ice/water classification of sentinel-1 images. Geophysical Research Abstracts. Vol. 17, EGU General Assembly 2015. EGU201512487-1. Kwok, R., & Cunningham, G. F. (1994). Backscatter characteristics of the winter ice cover in the Beaufort Sea. Journal of Geophysical Research, 99(C4), 7787–7802. Kwok, R., Curlander, J. C., McConnell, R., & Pang, S. S. (1990). An ice-motion tracking system at the Alaska SAR facility. IEEE Journal of Oceanic Engineering, 15(1), 44–54. Kwok, R., Cunningham, G. F., & Holt, B. (1992). An approach to identification of sea ice types from spaceborne SAR data. In F. D. Carsey (Ed.), Microwave remote sensing of sea ice (Geophysical monograph 68) (pp. 355–360). Washington, DC: American Geophysical Union. Kwok, R., Nghiem, S. V., Martin, S., Winebrenner, D. P., Gow, A. J., Perovich, D. K., Swift, C. T., Barber, D. G., Golden, K. M., & Knapp, E. J. (1998). Laboratory measurements of sea ice: Connections to microwave remote sensing. IEEE Transactions on Geoscience and Remote Sensing, 36(5), 1716–1730. Leigh, S., Zhijie, W., & Clausi, D. A. (2014). Automated ice–water classification using dual polarization SAR satellite imagery. IEEE Transactions on Geoscience and Remote Sensing, 52, 5529–5539.

296

N. Y. Zakhvatkina et al.

Levrini, G., & Attema, E. (2001). ENVISAT calibration and validation approach. European Space Agency. http://envisat.esa.int/m-s/envisat_mission_2001/CalVal/1Intro.html Liu, A. K., Martin, S., & Kwok, R. (1997). Tracking of ice edges and ice floes by wavelet analysis of SAR images. Journal of Atmospheric and Oceanic Technology, 14(5), 1187–1198. Liu, H., Guo, H., & Zhang, L. (2015). SVM-based sea ice classification using textural features and concentration from RADARSAT-2 Dual-Pol ScanSAR data. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 8(4), 1601–1613. Liu, J., Scott, K. A., Gawish, A., & Fieguth, P. (2016). Automatic detection of the ice edge in SAR imagery using curvelet transform and active contour. Remote Sensing, 8(6), 480. https://doi.org/ 10.3390/rs8060480. Lowe, D. G. (2004). Distinctive image features from scale-invariant keypoints. International Journal of Computer Vision, 60(2), 91–110. Lundhaug, M., Johannessen, O. M., & Esbensen, K. H. (2002). ERS SAR studies of sea ice signatures in the Pechora Sea and Kara Sea region. Canadian Journal of Remote Sensing, 28 (2), 1–14. Maillard, P., Clausi, D. A., & Deng, H. (2005). Map-guided sea ice segmentation and classification using SAR imagery and a MRF segmentation scheme. IEEE Transactions on Geoscience and Remote Sensing, 43, 2940–2951. Mäkynen, M., & Karvonen, J. (2017). Incidence angle dependence of first-year sea ice backscattering coefficient in Sentinel-1 SAR imagery over the Kara Sea. IEEE Transactions on Geoscience and Remote Sensing, 55, 11. Mäkynen, M. P., Manninen, A. T., Similä, M. H., Karvonen, J. A., & Hallikainen, M. T. (2002). Incidence angle dependence of the statistical properties of C-band HH-polarization backscattering signatures of the Baltic Sea ice. IEEE Transactions on Geoscience and Remote Sensing, 40, 2593–2605. MANICE. (2005). Manual of standard procedures for observing and reporting ice conditions, Canadian Ice Service – Environment Canada, ISBN 0-660-62858-9, Catalogue No. En56175/2005. McConnell, R., Kwok, R., Curlander, J. C., Kober, W., & Pang, S. S. (1991). Y-S correlation and dynamic time warping: Two methods for tracking ice floes in sar images. IEEE Transactions on Geoscience and Remote Sensing, 29(6), 1004–1012. Meier, W. N., Fetterer, F., Stewart, S. J., & Helfrich, S. (2015). How do sea-ice concentrations from operational data compare with passive microwave estimates? Implications for improved model evaluations and forecasting. Annals of Glaciology, 69, 332–340. Mejia, C., Thiria, S., Tran, N., Crépon, M., & Badran, F. (1998). Determination of the geophysical model function of the ERS-1 scatterometer by the use of neural networks. Journal of Geophysical Research, 103(C6), 12853–12868. Melling, H. (1998). Detection of features in first-year pack ice by synthetic aperture radar (SAR). International Journal of Remote Sensing, 19(6), 1223–1249. Moen, M.-A. N., Anfinsen, S. N., Doulgeris, A. P., Renner, A. H. H., & Gerland, S. (2015). Assessing polarimetric SAR sea-ice classifications using consecutive day images. Annals of Glaciology, 56(69), 285–294. Moen, M. -A. N., Doulgeris, A. P., Anfinsen, S. N., Renner, A. H. H., Hughes, N., Gerland, S., & Eltoft, T. (2013). Comparison of feature based segmentation of full polarimetric SAR satellite sea ice images with manually drawn ice charts. The Cryosphere, 7(6), 1693–1705. Muckenhuber, S., Korosov, A. A., & Sandven, S. (2016). Open-source feature-tracking algorithm for sea ice drift retrieval from Sentinel-1 SAR imagery. The Cryosphere, 10(2), 913–925. Nghiem, S. V., & Bertoia, C. (2001). Multi-polarization C-band SAR signatures of Arctic sea ice. Focus issue on ice and icebergs. Canadian Journal of Remote Sensing, 2(5), 387–402. Nghiem, S. V., Martin, S., Perovich, D. K., Kwok, R. M., Drucker, R., & Gow, A. J. (1997). A laboratory study of the effect of frost flowers on C-band radar backscatter from sea ice. Journal of Geophysical Research, 102(C2), 3357–3370.

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

297

Ochilov, S., & Clausi, D. A. (2012). Operational SAR sea-ice image classification. IEEE Transactions on Geoscience and Remote Sensing, 50, 4397–4408. Onstott, R. G. (1992). SAR and Scatterometer signatures of sea ice. In F. D. Carsey (Ed.), Microwave remote sensing of sea ice (Geophysical monograph 68) (pp. 73–102). Washington, DC: American Geophysical Union. Onstott, R. G., & Shuchman, R. A. (2004). SAR measurements of sea ice. In C. R. Jackson & J. A. Apel (Eds.), SAR marine User’s manual (pp. 81–116). Washington, DC: NOAA. Park, J. W., Korosov, A. A., Babiker, M., Sandven, S., & Won, J. (2018). Efficient thermal noise removal for Sentinel-1 TOPSAR cross-polarization channel. IEEE Transactions on Geoscience and Remote Sensing, 56(3), 1555–1565. https://doi.org/10.1109/TGRS.2017.2765248. Partington, K. C., Flach, J. D., Barber, D., Isleifson, D., Meadows, P. J., & Verlaan, P. (2010). Dualpolarization C-band radar observations of sea ice in the Amundsen Gulf. IEEE Transactions on Geoscience and Remote Sensing, 48(6), 2685–2691. Pedersen, L. T., Saldo, R., Fenger-Nielsen, R. (2015). Sentinel-1 results: Sea ice operational monitoring. Proc. of 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), https://doi.org/10.1109/igarss.2015.7326403. Perona, P., & Malik, J. (1990). Scale-space and edge detection using anisotropic diffusion. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(7), 734 629–734 639. Petrou, Z. I., & Tian, Y. L. (2017) High-resolution sea ice motion estimation with optical flow using satellite spectroradiometer data. IEEE Transactions on Geoscience and Remote Sensing, 55, 1339–1350. Pratt, W. (1982). Digital image processing (translation from English). Moscow: Mir, vol. 2, 1982 (in Russian). RADARSAT-2 product description. (2011). MacDonald Dettwiler and Associates (MDA), Tech. Rep. Issue 1/8, http://gs.mdacorporation.com/includes/documents/RN-SP-52-1238RS2ProductDescription1-8_15APR2011.pdf. Last access: 1 July 2014. 15 April 2011. Ramsay, B., Manore, M., Weir, L., Wilson, K., & Bradley, D. (1998). Use of RADARSAT data in the Canadian ice service. Canadian Journal of Remote Sensing, 24(1), 36–42. Reddy, S., & Chatterji, B. (1996). An FFT-based technique for translation, rotation, and scaleinvariant image registration. IEEE Transactions on Image Processing, 5(8), 1266–1271. Ressel, R., Frost, A., & Lehner, S. (2015). A neural network-based classification for sea ice types on X-band SAR images. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 8(7), 3672–3680. Ressel, R., Singha, S., Lehner, S., Rosel, A., & Spreen, G. (2016). Investigation into different Polarimetric features for sea ice classification using X-band synthetic aperture radar. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 9, 3131–3143. Romeiser, J., Horstmann Caruso, M. J., & Graber, H. C. (2013). A descalloping post-processor for ScanSAR images of ocean scenes. IEEE Transactions on Geoscience and Remote Sensing, 51, 3259–3272. Rublee, E., Rabaud, V., Konolige, K., Bradski, G. (2011). ORB: An efficient alternative to SIFT or SURF. In: 2011 International conference on computer vision, IEEE, pp. 2564–2571. Sandven, S., Johannessen, O. M., Miles, M. W., Pettersson, L. H., & Kloster, K. (1999). Barents Sea seasonal ice zone features and processes from ERS 1 synthetic aperture radar: Seasona ice zone experiment 1992. Journal of Geophysical Research, 104(C7), 15843–15857. Sandven, S., Dalen, O., Lundhaug, M., Kloster, K., Alexandrov, V. Y., & Zaitsev, L. V. (2001). Sea ice investigations in the Laptev Sea area in late summer using SAR data. Canadian Journal of Remote Sensing, 27(5), 502–516. Scheuchl, B., Flett, D., Caves, R., & Cumming, I. (2004). Potential of RADARSAT-2 data for operational sea ice monitoring. Canadian Journal of Remote Sensing, 30, 448–461. Schmidt, K., Tous Ramon, N., & Schwerdt, M. (2018). Radiometric accuracy and stability of Sentinel-1A determined using point targets. International Journal of Microwave and Wireless Technologies, 10, 538–546. https://doi.org/10.1017/S1759078718000016.

298

N. Y. Zakhvatkina et al.

Sentinel-1 Product Specification. (2016). Document S1-RS-MDA-52-7441, Issue 3/3, 2016. https:// sentinel.esa.int/documents/247904/349449/ Shokr, M. E. (1991). Evaluation of second-order texture parameters for sea ice classification from radar images. Journal of Geophysical Research, 96, 10625–10640. Shokr, M. E. (2009). Compilation of a radar backscatter database of sea ice types and open water using operational analysis of heterogeneous ice regimes. Canadian Journal of Remote Sensing, 35, 369–384. Soh, L. K., & Tsatsoulis, C. (1999a). Texture analysis of SAR sea ice imagery using gray level co-occurrence matrices. IEEE Transactions on Geoscience and Remote Sensing, 37, 780–795. Soh, L. K., & Tsatsoulis, C. (1999b). Unsupervised segmentation of ERS and RADARSAT sea ice images using multiresolution peak detection and aggregated population equalization. International Journal of Remote Sensing, 20(15–16), 3087–3109. Soh, L. K., & Tsatsoulis, C. (2002). ARCTOS: A knowledge engineering software tool for images. International Journal of Human-Computer Studies, 57(6), 469–496. Soh, L. K., Tsatsoulis, C., Gineris, D., & Bertoia, C. (2004). ARKTOS: An intelligent system for SAR sea ice image classification. IEEE Transactions on Geoscience and Remote Sensing, 42, 229–248. Sun, Y. (1996). Automatic ice motion retrieval from ERS-1 SAR images using the optical flow method. International Journal of Remote Sensing, 17(11), 2059–2087. Thermal Denoising of Products Generated by Sentinel-1 IPF. (2017). Issue 1.1. https://sentinel.esa. int/documents/247904/2142675/Thermal-Denoising-of-Products-Generated-by-Sentinel-1-IPF. Thomsen, B. B., Nghiem, S., & Kwok, R. (1998). Polarimetric C-band SAR observations of sea ice in the Greenland Sea. Proc. of 1998 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 6–10 July, Seattle, WA, USA, Cat. No. 98CH36174, vol. 5, pp. 2502–2504. https://doi.org/10.1109/IGARSS.1998.702259. Topouzelis, K., Singha, S., & Kitsiou, D. (2016). Incidence angle normalization of Wide Swath SAR data for oceanographic applications. Open Geosciences, 8(1), 450–464. https://doi.org/10. 1515/geo-2016-0029. Tso, B., & Mather, P. M. (2001). Classification methods for remotely sensed data. London: Taylor and Francis. 2001. Ulaby, F. T., Moore, R. K., & Fung, A. K. (1981). Microwave remote sensing: Active and passive. Reading: Addison-Wesley. Vesecky, J. F., Samadani, R., Smith, M. P., Daida, J. M., & Bracewell, R. N. (1988). Observation of sea-ice dynamics using synthetic aperture radar images: Automated analysis. IEEE Transactions on Geoscience and Remote Sensing, 26(1), 38–48. Volkov, V. A., Zakhvatkina, N., Mushta, A., Vesman, A., Kazakov, E., Shakhin, D. (2018). Automated monitoring of the sea ice in the Kara Sea for practical applications. Proc. of SCAR-IASC Open Science Conference. 19–23 June 2018, Davos, Switzerland, pp. 43. Wackerman, C. C. (1992). Digital SAR Image. In F. D. Carsey (Ed.), Microwave remote sensing of sea ice (Geophysical monograph 68) (pp. 105–110). Washington, DC: American Geophysical Union. Wackerman, C. C., Miller D. L. (1996). An automated algorithm for sea ice classification in the marginal ice zone using ERS-1 synthetic aperture radar imagery. ERIM, Ann Arbor, MI, Tech. Rep. Wadhams, P. (2003). How does Arctic sea ice form and decay? [Online]. Available: http://www. arctic.noaa.gov/essay_wadhams.html. Wang, S., You, H., & Fu, K. (2012). BFSIFT: A novel method to find feature matches for sar image registration. IEEE Geoscience and Remote Sensing Letters, 9(4), 649–653. Weeks, W. F., & Ackley, S. F. (1986). Growth structure and properties of sea ice. In N. Untersteiner (Ed.), The geophysics of sea ice (NATO ASI Series, Series B: Physics) (Vol. 146., Chap. 1, pp. 9–153). New York: Plenum Press. Winebrenner, D. P., Bredow, J., Fung, A. K., Drinkwater, M. R., Nghiem, S., Gow, A. J., Perovich, D. K., Grenfell, T. C., Han, H. C., Kong, J. A., Lee, J. K., Mudallar, S., Onstott, R. G., Tsang, L.,

6 SAR Sea Ice Type Classification and Drift Retrieval in the Arctic

299

& West, R. D. (1992). Microwave sea ice signature modelling. In F. D. Carsey (Ed.), Microwave remote sensing of sea ice (Geophysical monograph 68) (pp. 137–175). Washington, DC: American Geophysical Union. Winebrenner, D. P., Holt, B., & Nelson, E. D. (1996). Observation of autumn freeze-up in the Beaufort and Chukchi seas using the ERS-1 synthetic aperture radar. Journal of Geophysical Research, 101(C7), 16401–16419. Wohlleben, T., Howell, S., Agnew, T., & Komarov, A. (2012). Sea ice motion and flux within the Prince Gustaf Adolf Sea, Queen Elizabeth Islands, Canada during 2010. Atmosphere-Ocean. https://doi.org/10.1080/07055900.2012.750232. World Meteorological Organization. (1989). WMO sea ice nomenclature (p. 259). Geneva: WMO Rep. Yu, Q., & Clausi, D. A. (2008). IRGS: Image segmentation using edge penalties and region growing. IEEE Transactions on Pattern Analysis, 30, 2126–2139. Yu, Q., Moloney, C., & Williams, F. M. (2002). SAR sea-ice texture classification using discrete wavelet transform based methods. Proc. of 2002 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), vol. 5, pp. 3041–3043. Zakhvatkina, N. (2009). Identification of Sea Ice in the Arctic using SAR data. PhD thesis, Arctic and Antarctic Research Institute, Dissertation for the degree of candidate of physical and mathematical sciences in Oceanology, Russia, 169 p. (in Russian). Zakhvatkina, N., Alexandrov, V., Johannessen, O. M., Sandven, S., & Frolov, I. (2013). Classification of sea ice types in ENVISAT synthetic aperture radar images. IEEE Transactions on Geoscience and Remote Sensing, 51, 2587–2600. Zakhvatkina, N., Korosov, A., Muckenhuber, S., Sandven, S., & Babiker, M. (2017). Operational algorithm for ice water classification on dual-polarized RADARSAT-2 images. The Cryosphere, 11(1), 33–46. https://doi.org/10.5194/tc-11-33-2017.

Chapter 7

Sea Ice Drift in the Arctic Vladimir A. Volkov, Alexandra Mushta, and Denis Demchev

The role of dynamic factors, in particular the pattern and intensity of ice drift, in the formation of the ice regime and changes in the area of ice cover in the Arctic Ocean is examined (Gudkovich 1961; Belyakov et al. 1984a, b). Earlier it was shown that only 20% of the variability of the ice cover is determined by thermal factors (Makshtas et al. 2003). The relationship between large-scale variability of the ice drift field in the Arctic Ocean and changes in the sea ice extent occurring during the last decades is considered based on a unique data set of daily ice drift fields obtained by analysis of satellite remote sensing observations. Analysis of vector fields of sea ice drift is performed using a vector-algebraic method, which makes it possible to significantly compress the initial information and describe the vector fields by a limited set of vectors and scalar parameters. A combined analysis of variability of the drift fields in relation to changes in the type of atmospheric circulation was performed in the framework of classification of large-scale atmospheric processes in the Arctic (Dydina 1964, 1982; Atlas of the Arctic 1985; Volkov et al. 2012, 2016).

7.1

Contribution of Sea Ice Drift to the Spatial Distribution of Ice Cover

The problem of global warming and its consequences on the decline of sea ice extent in the Arctic Ocean has been widely discussed (IPCC 2013). The Arctic ice area has decreased in the last decades to record low values primarily due to the increase in V. A. Volkov (*) · A. Mushta Nansen International Environmental and Remote Sensing Centre, Saint Petersburg, Russia e-mail: [email protected]; [email protected] D. Demchev Arctic and Antarctic Research Institute, Saint Petersburg, Russia Nansen International Environmental and Remote Sensing Centre, Saint Petersburg, Russia e-mail: [email protected] © Springer Nature Switzerland AG 2020 O. M. Johannessen et al. (eds.), Sea Ice in the Arctic, Springer Polar Sciences, https://doi.org/10.1007/978-3-030-21301-5_7

301

302

V. A. Volkov et al.

Green House Gasses, in particular CO2. However, if one takes into account the fact that warming and cooling events were repeatedly observed for centuries, other factors also exist which influence the climate in the Arctic. Sea surface temperature is one of the important factors, and has in general increased during the last 100 years (IPCC 2013) apart from the early cooling in the Arctic from 1940 to 1970 (Johannessen et al. 2007). The sea ice extent in the Arctic Basin (AB) experienced significant fluctuations in the beginning of the twentieth century: a high sea ice extent was observed in the early twentieth century, however from 1910 to 1940 it decreased due to the Early Warming, thereafter increasing to 1970 due to the Early Cooling before dramatically decreasing to the present time (Johannessen et al. 2007). Sea ice extent fluctuations are strongly influenced by the atmospheric circulation which can be characterized by the Arctic Oscillation index (AO) in the Arctic region (Zakharov 1996; Frolov et al. 2007, 2010; Vinje 2001, see Chaps. 4, 5, and 10 for more details). The ice cover in the Arctic Ocean is non-uniform and the different temporal patterns of its variability in different parts of the Arctic Ocean should be taken into account (Belyakov et al. 1984; Frolov et al. 2007; Volkov et al. 2002). Thus, the analysis of variability of the ice cover should not only deal with the ice extent but also the geographical distribution of the sea ice in the Arctic Basin and, very importantly, the dynamics of the sea ice expressed by its drift. The wind regime and ocean currents, especially in the winter and spring period, significantly influence the sea ice distribution, formation of zones of ridging, cracks and fractures, in which new ice and young ice forms. Therefore, the typical features of young ice distribution can contain information on the structure of surface currents (Borodachev 1998; Volkov et al. 2002). Hence not only thermal but also dynamic processes in the ocean are reflected in the spatial structure of the ice cover. If the ice cover of the Arctic Seas were immobile then the ice thickness distribution by the end of winter would have been determined only by thermal factors: the air temperature, water heat content, and conditions of heat exchange between the atmosphere and the ocean. The ice motion significantly changes the distribution of ice cover thickness and its entire meso- and macro-structure (Frolov et al. 2007; Borodachev 1998; Volkov et al. 2002). Regions of divergent flow can be formed as a result of ice motion. When these regions of divergent flow are persistent we see polynyas form, where the productivity of ice formation is very high during winter. Water exchange and freshwater runoff can also significantly influence the ice regime. Strong freshening of the surface layer changes the growth of the ice cover, while a weakening of ice export through the Fram Strait creates more favourable conditions for the formation of multiyear ice, increasing the time of presence of ice floes in the Arctic Basin. Furthermore, the ice growth is directly dependent on the number of degree-days of frost, i.e., on the severity of winter. The large-scale patterns of ice and current circulation in the Arctic Ocean are the anticyclonic Beaufort Gyre (BG) in the Canadian Basin in part of the Amerasian Basin, and the Transarctic current (ТАC) in the Eurasian part. In addition to the Transarctic Current and the anticyclonic Beaufort Gyre – there are also temporal cycles, with different frequencies in the Eurasian and Amerasian Basins (Belyakov et al. 1984; Frolov et al. 2007). In general the variability of atmospheric pressure and wind above the ocean, expressed by the AO, leads to the

7 Sea Ice Drift in the Arctic

303

formation of different types of ice drift and current circulation which influences the variability of the Beaufort Gyre and the Transarctic current (Gudkovich 1961; Proshutinsky and Johnson 1997). For example, a weakened BG, occupying a relatively large area and a decreased ice export with the TAC passing out of the Arctic Ocean through the Fram Strait, causes the residence time of sea ice in the high latitudes of the AB to increase, which creates conditions for growth in ice thickness and increase in ice area. Based on the above, one can show that the variability of the ice drift field in the Arctic Ocean is primarily caused by the atmospheric circulation pattern expressed by the AO index which in turn is also influenced by the increase in Green House Gasses in the atmosphere (Kuzmina et al. 2005). In addition, the changing dynamics of the ice drift fields redistribute sea ice in the AB. In this Chapter we have analysed the multiyear changes of the drift fields based on space observations and furthermore assessed how the ice drift variability is influencing the changes of the ice cover in recent decades.

7.2

Datasets of Sea Ice Drift in the Arctic

During the last decades of the twentieth century it has become possible to obtain detailed data of the sea ice drift due to the development of satellite remote sensing. There are now several databases which describe large-scale sea ice drift fields in the Arctic Ocean that are accessible via the Internet and are available for analysis. The IFREMER (Laboratoire d’Oceanographie Spatiale) and Pathfinder (Fowler 2003; Fowler et al. 2013) databases of sea ice drift fields were used in this study. Both databases cover the entire Arctic Ocean where sea ice existed, although the IFREMER data only covers the winter period while the Pathfinder data also covers the summer. In the period 1991–2001, datasets from IFREMER are based on observations from satellite passive microwave radiometers SSM/I with a 3-day interval in a 62
62 km regular grid. Higher resolution data are available from the AMSR/E1,2 sensors covering the period from 2002 to the present, where the ice drift vectors are available on a regular grid of 32.5 km. Estimates of the accuracy of such vectors fields (Laboratoire d’Oceanographie Spatiale), show that this dataset allows one to adequately estimate the ice dynamics in the Arctic Basin since the average accuracy of the velocity module is 2.60 +/ 0.08 cm/s and the average accuracy of the drift direction is 0.9 +/ 1.5 , both compared with the drifting buoy observations, International Arctic Buoy Program (IABP, http://iabp.apl.washington.edu). Another version of the ice drift field dataset, the Pathfinder, contains the results of integrated analysis of visible and infrared data by different instruments (AVHRR, SMMR, SSMI) from different satellites: Aqua, POES, NIMBUS-7, DMSP F8–F17 from 1979 through to the present, including data from the International Arctic Buoy Program (IABP, http://iabp.apl.washington.edu). The spatial coverage of drift data for the northern hemisphere is from 48.4 to 90 N, interpolated to a regular grid of 25 km. The following analysis of the variability of ice drift fields was based on the use of both databases. It should be emphasized that this is a very extensive set of vector data

304

V. A. Volkov et al.

covering almost 40 years with high spatial and temporal resolution. The problem with all these data sets is the need to analyse large amounts of vector observations, which requires the application of special methods. A concept of a vector-algebraic approach suggested by (Belyshev et al. 1981, 1983; Rozhkov 2009) and adapted for the analysis of sea ice drift fields by (Volkov et al. 2012) was used for “compression” of these vectors data sets for further analyses.

7.3

Methodology of Analysis of Sea Ice Drift Time Series and Fields

The ice drift, as well as currents and wind, is often considered as a vector process and is described by a set of two-dimensional vectors with a specific temporal step. In our case, a vector-algebraic method was chosen as the most justified mathematically, see below for an explanation. It should be noted that, at the present time, this methodology has the status of being a standard method in the framework of the Russian Unified Information System on the World Ocean – ESIMO (Unified state. . . n.d.; Belyshev et al. 1981, 1983; Methodological letter. . . 1984; Volkov et al. 2012). To describe the vector drift series as a random two-dimensional process in the vector-algebraic approach, a concept of the Euclidean vector with the vector module V, the direction φ, and with the Cartesian projections VX and VY is used. The parallelogram law, vector and tensor multiplication and coordinate transformation rules when turning the initial coordinate system is defined, including the calculation of a series of scalar invariant parameters, which also take into account vector change both in absolute value and direction. Thus, unlike the two-component vector representation which makes it difficult to analyse the process as a whole, when the vector process is described separately for each of the orthogonal components with the vector-algebraic approach used here the process as a whole is described by a set of scalar parameters. The simplest graphical representation of the main invariant parameters is presented in Fig. 7.1. It is convenient to use the linear invariant I1, which is an analogue to the MSD – a quantitative characteristic of the total variability of the vector process, regardless of whether the module or direction of the vectors of currents changes, for intercomparison of dispersions of analysed processes. Dashed lines in the figure denote examples of the drift vectors of the initial series. For a more detailed description of vector series, derivatives of invariant parameters can be used, for example:  x ¼ λ2 λ1 , characterizing elongation of the dispersion ellipse; v ¼ I 1 =m , which is an analogue of the variation coefficient (at v  1 stable process, at v > 1– unstable). Thus, the vector-algebraic method allows us to adequately describe vector series using a set of invariant scalar parameters and construct the fields of these parameters

7 Sea Ice Drift in the Arctic

305

Fig. 7.1 Graphical interpretation of invariant parameters of the vector ! series, m! : Vector of V mathematical expectation— an average vector for a given realisation (absolute value and direction); λ1, λ2: The mean square difference (MSD) of the orthogonal components of the current velocity along the axes of maximum and minimum variability of the vector value—elements of the MSD ellipse; αo: The direction of the major axis of the MSD ellipse

in order to compare time separated fields, and describe the degree of interrelation between wind fields and ice drift. The vector-algebraic method was used to calculate invariant scalar statistical characteristics for the time series at the nodes of the regular grid, then the averaged drift fields were constructed, which were then classified to the corresponding types of atmospheric circulation over the north polar region (Dydina 1964, 1982; Dmitriyev 1994; Atlas of the Arctic 1985; Volkov et al. 2016). The classification is based on the localization of surface atmospheric pressure (baric) centres over the Arctic including several types of synoptic processes of 3–5 days duration each, divided into six main classes (Fig. 7.2). The change in the vorticity of the large-scale wind field in the AB creates conditions that are either favorable for the growth or melting of ice. However, a stronger Transarctic Current increases the outflow of ice from the Arctic Basin through the Fram Strait.

7.4

Spatial Structure and Variability of Sea Ice Drift Fields in the Arctic Ocean

Using the long-term series of ice drift vectors derived from satellite observations has allowed us to develop a new classification of large-scale ice drift fields and illustrate the successive process of transition from one circulation type to another. Furthermore, we are able to compare these changes with the variability of the types of atmospheric circulation according to a classification system of atmospheric processes (Atlas of the Arctic 1985; Dydina 1964, 1982) and to relate these changes to variations in ice conditions in the Arctic Ocean over the past decades.

306

V. A. Volkov et al.

Fig. 7.2 The six main classes of a structure of atmospheric surface baric centres over the Arctic

The first reliable information about the large-scale structure of the circulation of water and ice in the Arctic Ocean, based on oceanographic measurements throughout the ocean in the years 1955–1956, was published by (Timofeev 1960) proving the existence of an extensive anticyclonic circulation in the Canadian Basin, named the Beaufort Gyre (BG), as well as the well-known Transarctic Current (TAC) discovered by Nansen’s Fram Drift from 1893 to 1896 (Nansen 1902). The variability of large-scale ice circulation in the Arctic characterized by a combination of two main patterns – TAC and BG, was first proposed by (Gudkovich 1961) who identified these two types of ice circulation to be associated with different atmosphere circulation patterns (Fig. 7.3). Type А is characterized by intensification of the Arctic polar anticyclone with a high AO index in winter (Fig. 7.2, case A) and subsequent development of a large area of anticyclonic ice circulation centred in the Beaufort Sea (Fig. 7.3) The central flow of TAC is intensified and displaced towards the Eurasian mainland. This contributes to ice export from the Laptev, East-Siberian and Chukchi Seas northward. Type B is characterized by weakening of the Arctic polar anticyclone with a low AO index (Fig. 7.2, case B) and a subsequent decrease in the area of the anticyclonic circulation. The central flow of TAC is displaced towards the Beaufort Sea and this contributes to a cyclonic ice and water circulation in the East-Siberian and Chukchi Seas. The northward ice export from the Laptev Sea is weakened.

7 Sea Ice Drift in the Arctic

307

!

Fig. 7.3 Examples of calculated ice drif types. The arrows indicate the average vectors ( m! , V Fig. 7.1), colors – the variance of the velocity modulus (I1 according to Fig. 7.1). On the left – type A, on the example of winter 2009; on the right – type B, on the example of winter 1993

These two types are well reproduced in the analysed series of the drift fields, constructed on the basis of more than 30 years of data (Fig. 7.3). A wind driven ice drift and barotropic ocean model for the period 1946–1993 showed similar types of sea ice circulation in the Arctic Ocean, depending on the nature of the general atmospheric circulation (Proshutinsky and Johnson 1997). It should be emphasized that the first of the three types of ice circulation identified by us (Fig. 7.4a) is consistent with both types identified by Gudkovich. In our version of the classification, the types of Gudkovich are taken as subclasses of our first type. The most common type was the “classical” structure of large-scale surface circulation of ice, characterized by a pronounced BG (Fig. 7.4a) which occupies most of the Amerasian sub-basin and creates conditions for the growth of ice due to its long stay in the high-latitude zone of the B.Th. The ТАС branches out from the BG in the northern part of the Chukchi Sea and transports ice and water to the Fram Strait. According to the calculated maps, such a scheme was intense in 1979–1982, 1984, 1986, 1992, 1997, 2001, 2003, and 2012–2015. The years following those when split drift occurred in the Arctic are connected with low values of the AO. The second most common structure of the sea ice drift field from our analyses, not previously noted in studies of the Arctic ocean, was that the TAC zone is intensified, expanded, and directed toward the Laptev Sea with a significant decrease in the diameter and intensity of the BG, including its displacement to the shores of the Canadian Arctic Archipelago (Fig. 7.4b). The increase in the TAC causes ice removal from the Arctic Ocean through the Fram Strait. This type of drift usually does not continue for a long time, it was identified in 1987, 1994, 2002, and was most clearly expressed in 2006, 2007, and 2009.

308

V. A. Volkov et al.

Fig. 7.4 Examples of different types of ice circulation in the Arctic Basin in the winter period, occurred in different years (on the basis of pathfinder database) (a) Intense BG drift, (b) Intense TAC drift, (c)“Split” drift. It was found that in any year there are BG and TC, but with varying ! degrees of intensity and “isolation” from each other. The arrows indicate the average vectors (m! , V Fig. 7.1), colors – the variance of the velocity modulus (I1 in Fig. 7.1)

The third scheme indicated the lowest degree of BG and TAC intensity (Fig. 7.4c). In this type the main ice flow originates in the Laptev Sea and blocks the ice stream from the Chukchi Sea into the TAC, reducing the dynamic interaction between sub-basins. This contributes to the accumulation of ice in the Amerasian sub-basin, and at the same time causes less ice in the Eurasian sub-Basin. This transitional type of ice drift pattern contributes to the so-called “ice opposition”, when the ice conditions in the western and eastern sectors of the Russian Arctic in the summer season are significantly different (Sokolov 1962). Such a structure can be clearly identified in 1983, 1988–1990, 1998–1999, 2004, and 2010–2011.

7 Sea Ice Drift in the Arctic

309

Fig. 7.5 Examples of mutual location of TAC (black line) and BG (circle) in different years (1980, 1983, 1994), during winter period

Beside these types there are also transient structures where the TAC trajectories and location of the BG centre change from year-to-year. The BG centre is usually in the area of the Canadian Basin, but in some years it moves to the southwest or southeast and exits closer to the East-Siberian Sea. In 2002, the BG centre was displaced almost to the pole. The TAC, depending on the year, can begin in the EastSiberian Sea, in the Laptev Sea or can even be displaced to the Kara Sea (Fig. 7.5). The main regions of ice export from the Arctic Basin depend directly on the pattern of general large-scale ice drift, the current circulations and their variations. The ice export direction depends both on the location of the TAC and the degree of development of the BG as well as the location of its centre. The main export flow is through the Fram Strait, but there are also other routes for ice outflow. For example, when the central flow of the TAC is closer to Eurasia, we observe ice drift directed to the straits between Spitsbergen -Franz-Josef Land – Severnaya Zemlya Archipelagoes. When the TAC Center is passing close to the pole, part of the ice flows to the straits of the Canadian Arctic Archipelago. When the BG is weakening with a developed southern periphery in the Beaufort Sea, the ice transport is increased in the direction of the Bering Strait. The average large scale drift speeds in the winter season change significantly from year-to-year. The drift speed in the TAC is 1–3 cm/s on average. Two maximums are identified in the period under consideration, when the speeds reached 6–7 cm/s (1991–1992) and 5 cm/s (2000–2001). The drift speeds in the BG are lower (mainly about 1 cm/s), the largest speeds were observed in 1979–1982 (up to 5 cm/s) and in 2002–2004 (about 3 cm/s), but speeds were low in 1985–1986 and 1992–1997. Changes in the extent of the BG and the location of its centre are related to the change in intensity of the TAC, and the location of its central flow. The maximum TAC intensity is observed along the “East-Siberian Sea – Fram Strait” direction when the BG is absent (1994, 2005). When the BG is intensified the central flow of the TAC is along the direction “Laptev Sea – Fram Strait”, and when the BG is weakened (1985, 1986, 1992–1995 and 1997) – along the direction of “Laptev Sea – Greenland”. The average drift speed in the Fram Strait is 15 cm/s, and in the straits between Spitsbergen – Franz-Josef Land – Severnaya Zemlya – 8 cm/s. The maximum speed

310

V. A. Volkov et al.

in the Fram Strait was observed in the winter of 2004–2005 – with speeds up to 30 cm/s, when the central flow of the TAC passed near the pole. The minimum speed in the Fram Strait was observed during the period 1995–1998 (about 10 cm/s). We have shown that the intensity and location of the TAC and the BG significantly change from year-to-year. The differences in the average winter drift speed at the interannual scale in the TAC can be up to double or more, but in the BG up to fivefold and more, influencing the redistribution of ice of different ages within the Arctic Ocean including the ice export from the Arctic Basin. The total variability of average drift speed in time (dispersion) is characterized by the dispersion tensor invariant I1 (according to Fig. 7.1). The average total variability values, and regions with the maximum of I1 in the wintertime, as well as the average drift speed change significntly from year-to-year. The main regions with the largest variability are the region north of Cape Barrow, in the straits between Novaya Zemlya and Spitsbergen, part of the TAC to the north of Spitsbergen, and the Fram Strait. When the BG and TAC are well developed the largest variability is observed in the area of the Fram Strait and when poorly developed – in the area of Cape Barrow, where a positive anomaly is practically always observed both in the distribution of average drift speed and total dispersion. During the observation period from 1991 through to the present the I1 values for winter in the TAC when compared with BG are always higher by a value of about 2 cm/s, varying between 9 and 14 cm/s in the TAC and from 8 up to 13 cm/s in the BG. However the analysis of interannual changes of I1 shows that in 2002 there was a modification of the drift field variability. Before this time one observed consistent gradual changes of I1, while in 2002 I1 reached its maximum value – up to 4 cm/s. From 2003 the amplitude of fluctuations in I1 increased, both in the TAC and in the BG indicating an increase in the dynamic instability of the ice drift field in the Arctic Basin. Calculating the correlation between the mean annual I1 in the different regions with the I for the whole Arctic Basin (AB) one can estimate the degree of uniformity of the ice drift field variability. It turned out that the relationship between the fluctuations in the TAC and BG is rather weak (the correlation coefficient is 0.35), while the relationship between the AB-TAC and the AB-BG are higher, with correlation coefficients of 0.51 and 0.86. This indicates that the ice drift field variability is primarly determined by the extent of development of the BG. The most favorable conditions for an increase in the sea ice extent of the Arctic Basin are formed during group B (Fig. 7.2b) when one observes an absence of strong advection of warm air masses from lower latitudes, dominance of air flows with an east component, and minimum cloudiness with a high AO index. In the years with the maximum sea ice extent in the Arctic Basin, positive anomalies of atmospheric pressure and frequencies of occurrence of anticyclones over its entire region were observed. As a rule, increases in the time of ice presence in the zone with less cloudiness should facilitate a growth in ice thickness during the winter time.

7 Sea Ice Drift in the Arctic

311

Thus, the processes of type B form the conditions for ice cover growth – blocking the passage of cyclones and heat transport to the Arctic and stabilizing the anticyclonic structure of the ice drift field. The other groups of synoptic processes which can influence the ice cover decay are groups А and K (Fig. 7.2) which are characterized by the development of cyclonic activity over much of the Arctic Basin in particular group К – in the western and eastern parts of the Arctic Ocean destroying the anticyclonicity in the drift fields.

7.5

Summary

The intensity and position of the main large-scale drift fields of sea ice (TAC and BG), affecting the conditions of redistribution of ice in the Arctic Basin and its removal through the Fram Strait, variations from year to year under the influence of changes in the structure of atmospheric circulation matches with changes in the AO index. Within the framework of the proposed classification of surface atmospheric pressure fields (synoptic processes), it was shown that different types of atmospheric circulation form three main typical structures of the drift fields, which defines the conditions for melting, ice growth, and its removal from the Arctic Basin, mainly through the Fram Strait. The predominant influence on the formation of ice growth, conditions in the BG, occur under Group B, where the BG is stabilizeed and blocks the entry of warm cyclones into high latitudes. At the same time, ice removal through the Fram Strait is weakened. With the prevalence of atmospheric circulation structures of type A and K, conditions are created for more rapid melting of the ice cover. Acknowledgements The authors are grateful to Prof. Valentine A. Rozkov (St. Petersburg State University) for valuable advice given in the process of manuscript preparation (Chap. 7), as well as to members of the Oceanographic committee of the Russian Geographical Society for an efficient discussion during a seminar.

References Atlas of the Arctic. (1985). Section 7. Types of synoptic processes and anomalies of the weather regime. Мoscow: Main Administration of Geodesy and Cartography under the USSR Council of Ministers. 89 p. (In Russian). Belyakov, L. N., Volkov, V. A., Gazova, L. A., & Ponomarev, V. I. (1984a). Study of the interannual variability of the ocean water and ice circulation in the Arctic Ocean using a diagnostic model. Problems of the Arctic and Antarctic, 58, 45–54. In Russian.

312

V. A. Volkov et al.

Belyakov, L. N., Volkov, V. А., Ponomarev, V. I., & Chernyshev, А. F. (1984b). Peculiarities of interannual variability of water circulation of the Arctic Basin. Doklady Akademii Nauk, 276(4), 946–949. In Russian. Belyshev, А. P., Klevantsov, Y. P., & Rozhkov, V. А. (1981). On illusions and reality in the methods of analysis of sea currents. Proceedings of the State Oceanographic Institute, 157, 3–19. (In Russian). Belyshev, А. P., Klevantsov, Y. P., & Rozhkov, V. А. (1983). Probabilistic analysis of sea currents. Leningrad: Gidrometeoizdat. 1983. 264 p. (In Russian). Borodachev, V. Е. (1998). Ice of the Kara sea. St. Petersburg: Gidrometeoizdat. 182 p. (In Russian). Dmitriyev, А. А. (1994). Variability of the atmospheric processes of the Arctic and its account in long-range forecasts. Leningrad: Gidrometeoizdat. 207 p. (In Russian). Dydina, L. А. (1964). Macro-circulation method of weather forecasts for 3–10 days for the Arctic. Leningrad: Gidrometeoizdat. 391 p. (In Russian). Dydina, L. А. (1982). Peculiarity of development of synoptic processes in the Arctic and their use in medium-range forecasts. Leningrad: Gidrometeoizdat. 224 p. (In Russian). Fowler, C. (2003). Polar pathfinder daily 25 km EASE-grid sea ice motion vectors. Boulder: National Snow and Ice Data Center. Digital media. Fowler, C., Maslanik, J., Emery, W., & Tschudi, M. (2013). Polar pathfinder daily 25 km EASEgrid sea ice motion vectors. Version 2. (indicate subset used). Boulder: National Snow and Ice Data Center. Frolov, I. Е., Karklin, V. P., & Gudkovich, Z. M. (2005). Intra-secular climate changes, ice cover area of the Eurasian Arctic, Seas and their possible causes. Meteorol Hydrol, 6, 5–14. In Russian. Frolov, I. Е., Gudkovich, Z. M., Karklin, V. P., Kovalev Ye, G., & Smolyanitsky, V. M. (2007). Scientific studies in the Arctic. V. 2. Climatic changes of the ice cover of the Eurasian shelf seas. St. Petersburg: Nauka. 136 p. (In Russian). Frolov, I. Е., Gudkovich, Z. M., Karklin, V. P., & Smolyanitsky, V. M. (2010). Changes of climate of the Arctic and the Antarctic – result of action of natural causes. Problemy Arktiki i Antarktiki, 2(85), 52–61. In Russian. Gudkovich, Z. M. (1961). Relation of the ice drift in the Arctic Basin to ice conditions in the soviet Arctic seas. Proceedings of Oceanographic Commission of AN SSSR, 11, 13–20. In Russian. IABP (E-resource). (n.d.). – URL: http://iabp.apl.washington.edu. Accessed date: 16.01 2019. IPCC 2013. (2013). IPCC: Climate change 2013. The physical science basis. In T. F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex, & P. M. Midgley (Eds.), Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change. Cambridge, UK/New York., 1535 p.: Cambridge University Press. https://doi.org/10.1017/CBO9781107415324. Johannessen, O. M., Yu, A. V., Ye, F. I., Sandven, S., Pettersson, L. H., Bobylev, L. P., Kloster, K., Smirnov, V. G., Mironov, Y. U., & Babich, N. G. (2007). Remote sensing of sea ice in the Northern sea route. Studies and applications. Chichester: Springer-Praxis. 472 p. Laboratoire d’oceanographie spatiale: Cersat, Ifremer (E-resource). (n.d.). URL: http://cersat. ifremer.fr/data/. Accessed date: 16.12.2018. Makshtas, A. P., Shoutilin, S. V., & Andréas, E. L. (2003). Possible dynamic and thermal causes for the recent decrease in sea ice in the Arctic Basin. Journal of Geophysical Research, 108, C7. https://doi.org/10.1029/2001JC000878. Methodological letter on the probabilistic analysis of vector time series of the velocities of currents and wind. (1984). Leningrad; Gidrometeoizdat, 1984, 62 p. (In Russian). Nansen, F. (1902). The Norwegian polar expedition 1893–1896. Scientific Results, 3, 346–351. Proshutinsky, A. Y., & Johnson, M. A. (1997). Two circulation regimes of the wind-driven arctic ocean. Journal of Geophysical Research, 15(102), 12493–12514. Rozhkov, V. A. (2009). Methods and tools for statistical processing and analysis of the situation in the oceans on the example of hydrometeorology. Obninsk: VNIIGMI. World Data Center. 416 p. (In Russian).

7 Sea Ice Drift in the Arctic

313

Sokolov, A. L. (1962). Ice drift in the Arctic basin and changes of the ice conditions along the Northern sea route. Arctic and Antarctic Research, 11, 46–42. In Russian. Timofeev, V. T. (1960). The water masses of the Arctic Basin. Leningrad: Gidrometeoizdat. 191 p. (In Russian). Unified state information system on the situation in the world ocean (ESIMO (E-resource). (n.d.). URL: http://portal.esimo.ru/portal. Accession date: 16.12 2018. Vinje, T. (2001). Anomalies and trends of sea ice extents and atmospheric circulation in the Nordic seas during the period 1864–1998. Journal of Climate, 14(3), 255–267. Volkov, V. A., Johannessen, O. M., Borodachev, V. E., Voinov, G. N., Pettersson, L. H., Bobylev, L. P., & Kouraev, A. V. (2002). Polar seas oceanography: An integrated case study of the Kara sea. Chichester: Springer/Praxis. 495 p. Volkov, V. A., Ivanov, N. E., & Demchev, D. M. (2012). Application of a vectoral-algebraic method for investigation of spatial-temporal variability of sea ice drift and validation of model calculation in the Arctic Ocean. Journal of Operational Oceanography, 5(2), 61–70. Volkov, V. A., Mushta, A. V., Demchev, D. M., Korzhikov, A. Y., & Sandven, S. (2016). Relation of large-scale variations of the sea ice drift fields in the Arctic Ocean with climatic changes of total ice concentrations during last decades. Problems of the Arctic and Antarctic, 2, 50–63. In Russian. Zakharov, V. F. (1996). Sea ice in climatic system. Leningrad: Gidrometeoizdat. 213 p. (In Russian).

Chapter 8

Sea Ice Modelling Matti Leppäranta, Valentin P. Meleshko, Petteri Uotila, and Tatiana Pavlova

This chapter presents the geophysical background of mesoscale to large-scale sea ice modelling, discusses the structure of existing advanced models, and shows results from model simulations on future sea ice conditions in the Arctic Ocean. The first and second section introduce the theory of material description of sea ice, sea ice thermodynamics, and the fundamental equations of sea ice dynamics – conservation laws of ice, momentum and heat, and sea ice rheology. Thermodynamics is a one-dimensional problem while dynamics when integrated across sea ice thickness becomes a two-dimensional problem, and they are linked together by the ice conservation law. Forcing of sea ice dynamics and thermodynamics is via atmospheric and oceanic boundary layers, where parameterizations of exchange of momentum, heat, moisture, and salinity are of key importance. Sea ice properties relevant to climate are primarily sea ice extent, concentration and thickness. Mesoscale and large-scale sea ice models treat sea ice as a continuum by thickness distribution. These models consist of momentum equation, conservation laws of heat, salt and ice, and ice rheology. The main models used for climate investigations are the Los Alamos model CICE (Community Ice CodE) and LIM (The Louvain-la-Neuve Sea Ice Model), which are reviewed including their parameterizations. Moreover, the performance of these two main models are briefly assessed against observational reference data based on hindcasts and projections. Aspects of data assimilation are discussed where modelling is closely linked to sea ice remote sensing data.

M. Leppäranta (*) · P. Uotila Institute for Atmospheric and Earth System Research (INAR), University of Helsinki, Helsinki, Finland e-mail: matti.lepparanta@helsinki.fi; petteri.uotila@helsinki.fi V. P. Meleshko · T. Pavlova Voeikov Main Geophysical Observatory, Saint Petersburg, Russia e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2020 O. M. Johannessen et al. (eds.), Sea Ice in the Arctic, Springer Polar Sciences, https://doi.org/10.1007/978-3-030-21301-5_8

315

316

M. Leppäranta et al.

Specific attention is given to model performance in sea ice simulation using runs from CMIP3 (Coupled Model Intercomparison Project, Phase 3) and CMIP5 (Coupled Model Intercomparison Project, Phase 5). Sea ice extent and thickness are considered as the most important climate variables, which are needed for validation of current climate models. To validate model performance, comprehensive data sets are also required. Several important characteristics were considered and compared with observations: sea ice cover for seasons with maximum and minimum sea ice extent, sea ice trend in September, sea ice annual cycle, linear trend of sea ice area and volume, sea ice thickness distribution. A brief review is also provided for sea ice assimilation system and its current performance capability.

8.1 8.1.1

Sea Ice Geophysics Sea Ice Fields

A “sea ice landscape” consists of leads and ice floes with ridges, hummocks and other morphological characteristics. Mechanical behavior of sea ice fields depends largely on ice concentration, the relative area of ice in a given spot, and ice thickness (e.g., Rothrock 1975b; Hibler 1986; Leppäranta 2011). Ice types have been defined originating from practical shipping activities in ice-covered waters (WMO 1970–2017). They are based on the appearance, i.e. on how the ice looks to an observer on a ship or in an aircraft (Fig. 8.1). The formation mechanism, aging and deformation influence the appearance, which therefore contains information of the ice thickness, for which direct measurements are limited. Sea ice fields undergo continuous changing due to thermal and mechanical processes. It is one of the fundamental tasks of sea ice geophysics to understand and model the evolution of the ice conditions in a basin-wide scale.

Fig. 8.1 Drift ice material shown in a satellite image and in a photograph taken from a ship. The satellite image is over the Barents Sea; the length of the island Novaya Zemlya on the right is about 1000 km. (Satellite image © NASA, Visible Earth Team)

8 Sea Ice Modelling

317

Sea ice is examined over a wide range of scales. Microscale includes individual grains and ice impurities extending to 0.1 m. In the local scale, 0.1–10 m, sea ice is a solid polycrystalline continuum with sub-structure classified according to the formation mechanisms. Ice floe scale extends from 10 m to 10 km, including individual floes and ice forms such as rubble, pressure ridges and landfast ice. On larger scales, the sea ice medium is called drift ice, and, as in dynamical oceanography, mesoscale refers to 50–500 km and large scale beyond that. The internal time scales of drift ice are 1–2 h due to inertia and 10–100 days due to advection and deformation. Forced periodic variations are caused by shallow water waves, Earth’s rotation, and tides. Sea ice grows by different thermodynamic mechanisms, which show up in a layered vertical structure. The main forms of ice are congelation ice, snow-ice and frazil ice (Eicken and Lange 1989). Congelation ice crystals grow down from the bottom of ice as normally ice grows in lakes. It is the dominant form of ice in the Arctic Ocean. Frazil ice is produced in open water areas – leads and polynyas – when turbulence, surface waves and currents prevent the formation of a thin, solid ice cover. Snow-ice grows from slush, mixture of snow and liquid water, which can be seawater, melt water or liquid precipitation. It is common in the seasonal sea ice zone where snow accumulation is large. There are two more specific forms of ice: anchor ice and platelet ice. Anchor ice forms in the sea bottom in turbulent conditions in shallow regions, and it may rise to surface with bottom sediments due to its buoyancy. This is observed, e.g., in the Siberian shelf. Platelet ice forms in Antarctic waters next to floating ice shelves, which act as a condition to the start-up (Langhorne et al. 2015). Upwelling glacial meltwater becomes supercooled and forms large platelet crystals onto the bottom of sea ice. The thickness of sea ice is defined as the distance between the upper and lower ice surfaces. It is an irregular field because of fracturing, new ice growth in leads, and ridging and hummocking. Thermal growth may reach 2 m in 1 year, which is called first-year ice, and 3–5 m in a few years, which is called multiyear ice. Thickness of ice provides the two-dimensional ice strength and the ice volume. In ice engineering the thickness is a key parameter in estimating ice forces on structures (e.g., Palmer and Croasdale 2012). Ice floats on sea surface since its density is about 10% less than the density of seawater. The freeboard is, according to Archimedes’ law, h0 ¼


 ρ ρ 1  i h f  s hs ρw ρw

ð8:1Þ

where ρi, ρs and ρw are the densities of ice, snow and water, respectively, and hf and hs are the ice floe and snow thicknesses. This equation is the basis to estimate sea ice thickness from freeboard or draft measurements. The potential energy of floating ice is (Rothrock 1975a) Ep ¼

1 ρi ð ρw  ρi Þ 2 gh f 2 ρw

ð8:2Þ

318

M. Leppäranta et al.

where g is the acceleration due to gravity. Potential energy production is important in ice dynamics since it is correlated to the work done to build deformed ice. The horizontal structure of drift ice fields is well revealed by optical satellite images (Fig. 8.1). The elementary particles are ice floes, described by their thickness and characteristic diameter d. The floe size varies over a wide range and may reach tens of kilometers in the Central Arctic Ocean. The behavior of a drift ice field depends on its horizontal size L and the thickness and size of ice floes. The ratio hf/L tells of the stability of the ice cover in that for large hf/L forces do not build up enough to break ice cover. A drift ice material particle is a set of ice floes, with its size D it contains  (D/d)2 floes. For the continuum approach as normally taken, the number of floes in the particle should be large or D/d ≿ 10. Also, the particle size should be much less than Q the scale of changes or the gradient scale Λ ¼ ∇Q where Q is an ice field property. Summarizing, d > 0 for A  1; (3) Compressive strength > shear strength; (4) Tensile strength is small; and (5) No memory. Plastic rheology laws were introduced for drift ice in the 1970s in the AIDJEX (Arctic Ice Dynamics Joint Experiment) program with yield strength increasing with ice thickness (Coon et al. 1974; Pritchard 1975; Hibler 1979). Scaling and fracture of sea ice was analyzed in a monograph of Weiss (2013). To solve the plastic flow in numerical modelling, stresses below the yield level must be solved. This involves small deformations. In the original plastic drift ice rheology, the small deformation regime was taken care by an elastic model (Coon et al. 1974; Pritchard 1975). A few years later, Hibler (1979) introduced a viscousplastic rheology, which is more feasible for long-term sea ice simulations. These rheologies have served as the basis of later plastic models, where the main concern has been the shape of the yield curve and the speed of the numerical solution. A computationally effective elastic-viscous-plastic model was developed by Hunke and Dukewicz (1997) with a technical elastic term added to the Hibler (1979)

8 Sea Ice Modelling

329

viscous-plastic model. This model has become widely used. Anisotropic plastic rheologies have also been introduced (Coon et al. 1998). Isotropic plastic drift ice rheologies have the minimum of three physical parameters: compressive strength, shear strength and stress level dependence of ice state. The ratio between compressive and shear strength provides the shape of the yield curve, while the stress level defines the curve size. A common form of the compressive strength is P ¼ Pn hn exp½Cð1  AÞ

ð8:17Þ

where Pn and n are the compressive strength parameters, and C is the strength reduction constant for opening. We can allow the power to vary in the range ½  n  2, depending on the mode of deformation (see Coon et al. 1974). Due to the high sensitivity of strength to compactness, we have C >> 1. In the original version of this formula (Hibler 1979), n ¼ 1 and C ¼ 20 but also the choice n ¼ 2 has been later employed. Two-dimensional plastic yielding is specified with a yield curve F(σ 1, σ 2) ¼ 0 where σ1 and σ2 are the principal stresses (Fig. 8.8). Drucker’s postulate (Drucker 1950) see also Coon et al. 1974) for stable materials states that the yield curve serves as the plastic potential, and consequently the failure strain is directed perpendicular to the yield curve, known as the normal, or associated flow rule (e.g., Davis and Selvadurai 2012). When the ice fails, the plastic flow is obtained from the equation of motion. Drift ice is strain hardening in compression, and therefore pressure ice formation may proceed only to a certain limit. Another approach was recently proposed considering drift ice as an elastic-brittle medium (Girard et al. 2011). This approach assumes that ice deformation is based on multiscale fracturing and friction and gives a more realistic structure for ice

Fig. 8.8 (a) Drift ice yield curves: Coulomb or triangular (Coon et al. 1974), teardrop (Coon et al. 1974), and elliptic (Hibler 1979); (b) Compressive strength of drift ice as a function of compactness for (i) Floe collision models (Shen et al. 1986), (ii) Floe contact models based on Harr (1977), and (iii) Standard version of Hibler (1979) sea ice rheology

330

M. Leppäranta et al.

deformation vs. length scale (Schulson 2004; Weiss and Dansereau 2016). The model was later expanded into a Maxwell elastic-brittle rheology by Dansereau et al. (2015). The elastic-rheology allows a more detailed and physically based description of fracturing of ice cover, while the classical plastic modelling provides a feasible solution to sea ice dynamics with minimum number of parameters. Sea ice cover is forced by the tangential air–ice and water–ice stresses (e.g., Andreas 1998; McPhee 2008) and the surface pressure gradient. These are given, respectively, by τ a ¼ ρa C a U a ð cos θa þ sin θa kÞU a

ð8:18aÞ

τ w ¼ ρw Cw jU w  ujð cos θw þ sin θw kÞðUw  uÞ

ð8:18bÞ

G ¼ ρi gh∇ξ

ð8:18cÞ

where Ca and Cw are air and water drag coefficients, θa and θw are the boundary layer angles in air and water, Ua and Uw are the surface wind and current velocities, and ξ is the water level elevation. Representative values of the parameters are for geostrophic flow, in neutral stratification, Ca ¼ 2.5 103 (Andreas 1998), Cw ¼ 5.5 103 (McPhee 2008) and θa ¼ θw ¼ 25 . In a stratified fluid, the drag parameters depend also on the stability of the stratification (Andreas 1998; McPhee 2008). The lateral boundary of a sea ice field can be solid, i.e. shoreline, where the ice velocity is zero, or free, where the ice does not support normal stresses. The boundary conditions are often replaced by a simplified form: open water is defined as ‘ice with zero thickness’ and then the open boundary question is removed. Landfast ice is often taken as a solid boundary condition that can be seriously biased in the ice melting season (see Yang et al. 2015). Table 8.1 shows a magnitude analysis of the terms of the equation of motion based on the typical scales. Wind stress is the main driving force, largely compensated by ice–water stress and the internal friction. Coriolis acceleration is important for thick ice. In free drift, internal friction is zero and the wind stress is compensated by ice–water stress. The inertia can become significant for very rapid changes in the forcing. Advective acceleration is very small and will remain smaller than the ice– Table 8.1 Scaling of the equation of motion of drift ice. The representative elementary scales are: ice thickness H ¼ 1 m, ice velocity U ¼ 10 cm s1, ice strength P ¼ 50 kPa, wind velocity Ua ¼ 10 m s1, water velocity Uw ¼ 0, surface slope ∇ξ ¼ 106, time T ¼ 1 day, and horizontal length L ¼ 100 km Term Local acceleration Advective acceleration Coriolis term Internal friction Air stress Water stress Pressure gradient

Scale ρHU/T ρHU2/L ρHfU PH/L ρaCaUag2 ρwCwU2 ρHg∇ξ

Value (Pa) 0.001 0.0001 0.01 0–0.5 0.25 0.05 0.01

Comments 0.025 for rapid changes (T ¼ 1 h) Long-term effects may be significant 0.05 for H ¼ 3 m and U ¼ 20 cm s1 0 – open ice field, 0.5 – compact ice field Mostly significant 0.2 for U ¼ 20 cm s1 Mostly less than Coriolis term

8 Sea Ice Modelling Fig. 8.9 (a) The free drift solution as the vector sum of wind-driven ice drift and geostrophic current (when the Coriolis acceleration is significant). (b) Illustration of the solution of ice drift in the presence of internal friction in the marginal ice zone

331 wind

a wind-driven ice drift (2% and 30 degrees)

ice velocity

current velocity

b

water stress term as long as H/L < Cw, a condition that is valid except for extremely exceptional cases. There are three principal time scales: local acceleration T I ¼ CUw H , advection T D ¼ UL , and the inertial period f1. These time scales are well separated, TI d 3 h–1 day

Comments Geostrophic wind Geostrophic wind Geostrophic current Geostrophic current At h ¼ 1 m Half of the compressive strength Change of strength by e1 due to ΔA Power law, power between 1/2 and 2 Viscous–plastic transition Needed for low level ice state Separation is straightforward d ~ floe size or scale of fractures Stability requirement

floes. Hopkins and Hibler (1991) and Hopkins (1994) examined the sea ice ridging process with a discrete particle model. However, granular models have not overcome the continuum approach, and all sea ice climate models are based on the continuum approximation. The parameters of sea ice dynamics models can be put into four groups: I Free drift parameters (drag forces), II Rheology parameters, III Ice state redistribution parameters, and IV Numerical solution parameters. As an example, the Hibler (1979) sea ice model parameterization is shown in Table 8.2, which was based on field data and model experiments. The free drift wind factor has been about 2%, as supported by observations. This case represents a minimum set of parameters for feasible full models. In more complicated cases, ice state has more levels and then rheology and ice state redistribution need higher level parameterization. Drift ice models have mostly taken a viscous approach, largely due to computational reasons. In short-term modelling, the time scale of ice cover evolution is 1 h– 10 days. Because the inertial time scale of sea ice is quite small (less than 1 h), the initial ice velocity can be taken as zero. For coastal zone problems, a proper treatment of the boundary configuration is critical. The original viscous-plastic drift ice model of Hibler (1979) included an inconsistency of having non-zero stress for an ice field at rest, with σ ¼ ½PI for ε_ ¼ 0. This inconsistency was later removed (Hibler 2001). Also, in the case of low, long-term forcing, the viscous case has an unrealistic feature in leading tocontinuous  creep. The length of an ice beam would be changed by a factor of exp ε_ t due to creep, and this would account for 2.5% in 1 month for a creep of 108 s1. The viscous strain-rates should be much smaller than allowed in the model but it is not exactly known how much. Normally the ice is mostly dynamically active that keeps the creep within the noise.

8 Sea Ice Modelling

337

The viscous-plastic, VP model is quite demanding in terms of computational time. To improve this, an elastic term 1 ∂σ ¼ ε_ E ∂t

ð8:21Þ

where E is the Young’s modulus, was added to the rheology by Hunke and Dukewitz (1997). The elastic term is rather technical than physical, but as a result the plastic sea ice flow can be solved very efficiently. The model, known as the elastic-viscousplastic or EVP model, has become the most common sea ice dynamics modelling approach. Although the ice drift problem offers quite interesting basic research possibilities, the principal science motivation has come from the sea ice introducing a particular air–sea interface. The exchange of momentum, heat and matter between the atmosphere and the ocean goes through drift ice fields in high latitudes, and this interface experiences transport as well as opening and closing due to the ice drift. This is crucially important to the regional weather and to the global climate. The ice extent, largely influenced by the ice drift, has a key role in the cryospheric albedo effect. Also, the ice with itself transports latent heat and fresh water, and ice melting gives a considerable heat sink and fresh water flux into the oceanic surface layer. In the ecology of polar seas, the location of the ice edge with the ice melting processes is a fundamental boundary condition for the summer productivity. A more recent research line for sea ice dynamics is in paleoclimatology and paleoceanography (Bischof 2000). Data archive of drift ice and icebergs exists in marine sediments, and via its influence on ocean circulation the drift ice has been an active agent in the global climate history. In the practical world three major questions are connected with sea ice dynamics. First, sea ice models have been applied for tactical navigation to provide short-term forecasts of the ice conditions. Secondly, ice forcing on ships and fixed structures is affected by the dynamical behavior of the ice (e.g., Sanderson 1988). Third, the question of pollutant transport by drifting sea ice has become an important issue (Pfirman et al. 1995). In particular, risk assessment for oil spills and oil combating require proper oil transport and dispersion models for ice-covered seas (e.g., Ovsienko et al. 1999).

8.2.3

Mesoscale Sea Ice Models

Mesoscale (50–500 km) models have been used for regional investigations and process studies in the polar oceans. Their structure is essentially as in large-scale models but the regionality often means a stronger role for the boundary conditions. In ice forecasting applications, initial conditions are also a major issue. The objective of short-term modelling may be basic research of drift ice dynamics, including coupled ice–ocean–atmosphere modelling, simulations to examine the influence of

338

M. Leppäranta et al.

ice dynamics on planned marine operations and ice forecasting. The deformation length scale of compact drift ice is around 100 km, and therefore internal friction has a major role in fully ice-covered basins. Therefore, the mobility of Hudson Bay ice is weak, and it has been observed in the Bothnian Bay, Baltic Sea, that at half meter thickness compact ice cover becomes stationary (Leppäranta 2011). Oil spills are difficult problems in ice conditions. Good oil combating methods do not exist, and it is difficult to keep track where the oil is going. The oil may penetrate into the ice sheet and drift with the ice or drift on the surface of openings and beneath sea ice. A simple modelling approach is an oil advection model with the ice and surface current with random diffusion superposed using, e.g., a Monte Carlo method (e.g., Venkatesh et al. 1990). An advanced, physical model treats oil as a viscous medium with density and viscosity dependent on the type of the oil (Ovsienko et al. 1999). They state that for ice compactness more than 30% oil practically drifts with the ice, and that in slush or brash the thickness of oil film can be much larger than in open cold water. As the concentration increases, at 80% level the oil is trapped between ice floes and at above 95% the oil is forced beneath the ice. In long-term modelling the time-scale is 1 month – 100 years. The approach is dynamic-thermodynamic, sometimes dynamics is even neglected. The initial conditions are arbitrary in very long time-scales but relevant for the ice state in monthly problems. The objective may be (i) Basic research of drift ice geophysics, (ii) Ice climatology investigations, and (iii) Coupled ice–ocean-atmosphere climate modelling. First only thermodynamic models were available for the times of freezing and ice break-up and for the evolution of ice thickness. But it became easily clear that realistic ice dynamics are needed for the ice transport and, in particular, for opening and closing of leads. Large amount of heat is transmitted through leads from the ocean to the atmosphere. By freezing and melting the ice has a major influence on the hydrographic structure of the ocean, and also there the motion of ice has an important role since the ice melts on a different region from where it forms. The ice climate problem received increasing attention in the 1990s in regional ice climate modelling (Haapala and Leppäranta 1996). These models are forced by synoptic weather conditions for calibration with ice charts for the validation. The models are capable to simulate the evolution of the whole ice season from one summer to the next one. Calibrated models have been then used for ice season scenarios based on the existing atmospheric climate scenarios. Here we present Baltic Sea ice modelling as an example (see Sects. 8.3 and 8.4 and Chap. 9 for the case of the Arctic Ocean). The Baltic Sea is a brackish-water basin in northern Europe, with area of 0.4 million km2 (Leppäranta and Myrberg 2009). Ice occurs in the basin annually reaching a maximum annual thickness of 0.5–1 m. Short-term ice forecasting was commenced in the operational ice service in Finland in the 1970s for winter shipping (Leppäranta 1981). In the coastal and archipelago areas there is landfast ice, which is stable and smooth for most of the winter, supported by islands and grounded ice ridges on shoals. The landfast ice zone extends to the depths of 5–15 m depending on the thickness of ice. Ice ridges are the most difficult obstacles in winter shipping and they cause the largest forces against marine structures in the Baltic Sea. The winter shipping is assisted by 20–25

8 Sea Ice Modelling

339

Fig. 8.11 Calibration of a Baltic Sea ice climate model: (a) Ice chart showing reality, (b) Ice thickness (cm) and open water surface temperature from the model ( C) from the model, (c) Mean thickness of deformed ice (cm). The model initial time was May 1, 1983, and the comparison here is for March 22, 1984 (Haapala and Leppäranta 1996)

icebreakers, however the cargo transportation has still suffered from delays. A revised viscous-plastic three-level ice state (compactness, mean thickness, and deformed ice volume) model was taken into the operational ice forecasting in 1992 (Leppäranta and Zhang 1992). This model has been examined in detail for the dynamics in different basins, in particular for scaling and the influence of coastal geometry and islands (e.g., Wang et al. 2003). It has worked well down to bay of 15 km size with thin ice moving under strong wind. Calibration of the Baltic Sea ice climate model for a normal winter is shown in Fig. 8.11 (Haapala and Leppäranta 1996). The initial time was May 1st. The figure shows comparison between the model and observed ice condition in March when the ice extent was at largest. There are small discrepancies in that the surface temperature in the south is too high in the model, and further north the model shows an open water region but ice chart shows there thin ice. Elsewhere the mean thickness came out quite well. Similar results were obtained for comparisons in mild and severe winters.

340

M. Leppäranta et al.

Table 8.3 Sea ice parameters useful in climate modeling Quantity Ice concentration and thickness 2 Ice temperature Ice salinity Ice drift velocity Snow thickness and temperature

8.2.4

No. of variables 2 1 1 2 2

Strong coupling Atmosphere and ocean Atmosphere and ocean Ocean Ocean Atmosphere

Sea Ice Properties Relevant to Climate

Sea ice forms a specific atmosphere–ocean interface and becomes thereby a major issue in climate modelling. The role of sea ice in this problem consist of its influence on the air–sea transfer of momentum, heat and matter and the salt/fresh water fluxes between ice and sea water. Sea ice drift is important in this play since it transports, deforms ice fields with opening and closing of leads, and modifies sea ice boundaries in the ocean surface. To model the sea ice interface properly, the necessary dependent ice quantities are listed in Table 8.3. Altogether there are 8 ice variables to be solved in each lateral grid cell. They are obtained using the conservation laws of momentum, ice, snow, heat and salt. The ice layer is strongly coupled to the ocean and the thermodynamics is strongly coupled also to the atmosphere (see Chapter 3.2 for more details). In the following Sects. 8.3 and 8.4, different sea ice data sets are referred with discussing the outcome of various models. These data sets are described in more detail in earlier chapters in this book, sea ice area and extent in Chap. 4 and sea ice thickness in Chap. 5.

8.3 8.3.1

Large-Scale Sea Ice Models Background

In Sect. 8.1.1, the spatial scales for sea-ice dynamics were defined. Accordingly, in regional seas, such as the Baltic Sea, sea-ice dynamics are mesoscale, while at hemispheric scales, such as the Arctic Ocean and the Southern Ocean, sea-ice dynamics also occur on the largest possible scale on Earth. Similar numerical code is applied to simulate both mesoscale and large-scale sea-ice dynamics and the code is capable to resolve both scales simultaneously. Mesoscale models have a higher spatial resolution and regional configuration in comparison to large-scale models which are often global and address long time scales, from seasonal to climatic. We start by investigating large-scale sea-ice models in the following section, as this provides a good general view of state-of-the-art sea-ice modelling in terms of physics parameterization.

8 Sea Ice Modelling

341

Sea ice is an important part of Earth’s climate system because it effectively regulates the amount of energy being transferred between the atmosphere and polar oceans (Vaughan et al. 2013). Analogously, an advanced knowledge of mechanisms redistributing the sea ice is required for an accurate description of the state and evolution of climate. Our current understanding on sea-ice related climate dynamics is incorporated in complex climate models (Griffies 2004). To better understand the sea-ice related climate dynamics as a step towards the development of more realistic earth system models, coupled global ocean-ice models can be used (Griffies 2009). A realistic simulation of the evolution of sea ice in an ocean-ice model is a difficult task. This is because even relatively small inconsistencies in atmospheric and oceanic energy and freshwater fluxes can result in large deviations in the distribution of sea ice (Dethloff et al. 2001; Uotila et al. 2012). As a result, the amount of sea ice may dramatically increase or melt, and consequently change the upper ocean stratification by modifying its salinity. Although the sea-ice distribution may relatively quickly return back to a realistic looking state, for example as a response to observation based atmospheric forcing, the upper ocean stratification may remain unrealistic much longer, even persistently. In addition to the model binding issues, an ocean-ice model is exposed to limitations inherent to imperfect, prescribed atmospheric forcing. Namely, the ocean cannot affect the prescribed atmosphere in an ocean-ice model (Griffies et al. 2009). This absence of interaction impacts the atmosphere-ocean processes, such as sea-ice edge movements and thermohaline circulation, so that representation is likely to differ in the fully coupled configuration compared to the ocean sea-ice one sharing the same ocean-ice configuration. For example, one might expect that the Atlantic thermohaline circulation is stronger in the ocean-ice model than in the fully coupled configuration, where the lower atmosphere warms by the oceanic heat transmitted by the air-ocean sensible heat flux, which is proportional to the oceanatmosphere temperature difference. These limitations have to be kept in mind when interpreting results from an ocean-ice model. Despite these shortcomings, there are many good reasons to use an ocean-ice model instead of a fully coupled climate model. First, fully coupled models are computationally very expensive, particularly for sensitivity experiments, which require a large number of decadal or even centennial simulations in order to determine oceanic responses. Second, ocean-ice model simulations are isolated from biases that arise when coupling to a potentially inaccurate atmospheric model. Third, when forced with observed atmospheric states, an ocean-ice model simulation time series can directly be compared with available ocean and sea-ice observations to assess the level of realism of these simulations. A large body of literature has already been published presenting ocean-ice model assessments, which provide important benchmark results when carrying out qualitative model assessments. Papers of the CORE-II virtual special issue of the Ocean Modelling Journal, such as Danabasoglu et al. (2014, 2016), Downes et al. (2015), Farneti et al. (2015), Griffies et al. (2014), and Wang et al. 2016a, b), in the ORA-IP special issue of the Climate Modelling Journal, such as Chevallier et al. (2017) are

342

M. Leppäranta et al.

particularly relevant for polar assessments. In the majority of currently used ocean climate models the grid configuration does not resolve ocean eddies. In these coarseresolution ocean-ice models, the eddy transport of momentum and heat needs to be parameterized. Many recent simulations also share the use of CORE bulk formulae and some forcing files, such as the river runoff, with the CORE-II experiments. In the polar context, the most important CORE-II papers include Downes et al. (2015), where the Antarctic sea ice and Southern Ocean water masses are analyzed, and Wang et al. (2016a, b) who investigated the Arctic sea ice and the Arctic Ocean freshwater. Recently, Chevallier et al. (2017) and Uotila et al. (2018) analyzed the Arctic and Antarctic sea ice in a set of ocean reanalyses to assess how the assimilation of observations in ORA-IP affects the sea-ice characteristics. As CORE-II results, Chevallier et al. (2017) and Uotila et al. (2018) results are a useful benchmark when assessing the skill of ocean-ice model performance. The specific need for sea-ice model assessment arises from the permanent evolution of formulations and parameterizations implemented by ocean-ice model developers. After a major release of new ocean-ice model version, a detailed diagnostics of its performance is of primary importance in order to find out is of the user community to test the new code. Related to this, some configurations are used in major projects and require rather urgent testing. For example, the climate modelling community is preparing to contribute to CMIP6 with model simulations. Before that, the modelling teams have to decide upon their CMIP6 model configurations. It is therefore valuable for them to obtain the supportive information on the performance of their model, even in the ocean-ice mode without the coupling to the atmosphere.

8.3.2

Current Main Sea-Ice Models

Two sea-ice models, LIM and CICE, are perhaps more widely used for large-scale and meso-scale sea-ice modelling than others. Moreover, they are typically used within coupled climate modelling frameworks, coupled at least with an ocean model and often also with an atmospheric model. Their success is largely due to the inclusive community development approach and freely distributed open source code, including documentation and test cases, which makes their use particularly easy. Other important models exist, such as by Winton (2000), Zhang and Rothrock (2001), Salas y Melia (2005) and Rampal et al. (2015), but as LIM and CICE share many characteristics in terms of numerics implementation and physical parameterizations with each other and with other models, their description and assessment provide a good introduction to today’s sea-ice modelling.

8 Sea Ice Modelling

8.3.2.1

343

Los Alamos National Laboratory Community Ice CodE, CICE

The Los Alamos National Laboratory Community Ice CodE (CICE) sea-ice model represents the state-of-the-art in sea-ice modelling (e.g. Flocco et al. 2012; Holland and Kwok 2012). CICE uses an elastic-viscous-plastic dynamics scheme (EVP; Hunke and Dukowicz 1997) for the internal ice stress, an incremental linear remapping for the ice advection term, and computes the ice thickness redistribution through ridging and rafting schemes by assuming an exponential redistribution function. The recent release of the LANL Community Ice CodE is version 5 (CICE5; Hunke and Lipscomb 2010; Hunke et al. 2015). The CICE5 sea-ice salinity profile is prognostic, changes in time, while the snow on ice is assumed to be fresh. Further, CICE5 features explicit parameterizations for melt ponds (Flocco et al. 2012), elastic-anisotropic-plastic rheology (Tsamados et al. 2013) and an option for sea-ice biogeochemical modelling and passive tracers (Jeffrey et al. 2011). In terms of the atmosphere-ocean momentum exchange, an advanced form drag parameterization at the ice-atmosphere boundary is implemented (Tsamados et al. 2014). CICE has been designed to couple with the ocean and atmosphere components of climate models. For large-scale simulations, a typical time step of the sea-ice model is 1 h and the momentum equation is solved by using an iterative scheme within each time step. For example, CICE in the Australian ACCESS coupled system for CMIP5, was coupled with the ocean model at every time step while the atmospheric forcing was updated every 6 h by the data atmospheric model via the OASIS coupler code (Valcke 2006). CICE sea ice was divided into five thickness categories of ice and open water and had four vertical ice layers and one snow layer in each category. CICE in ACCESS for CMIP5 run on the ocean grid which gives enhanced resolution of 10 km in the Arctic due to the orthogonal curvilinear tripolar grid (Bi and Marsland 2010; Uotila et al. 2012). Due to this Arctic grid refinement, meso-scale could be resolved, at least to some extent, even in this global model configuration (Fig. 8.12). The simulated sea-ice distribution is impacted by many predefined internal sea-ice model parameters (Hunke 2010; Uotila et al. 2012). In Table 8.4, typical values of the most significant thermodynamic and dynamic parameters are listed. These values follow the ones used in experiments described by Uotila et al. (2013). In these experiments, the short-wave parameterization scheme chosen was the default NCAR Community Climate System Model version 3 (CCSM3) scheme, where visible and infrared albedos are prescribed. The wavelength of 700 nm separated the visible and infrared bands. Additionally, ice and snow albedos on both spectral bands depended on the sea-ice or snow surface temperature and thickness. When ice became thinner than 0.1 m, the ice albedo decreased smoothly, following the arctangent function, toward the open ocean value of 0.06. When the surface temperature rose within the temperature ranges (Table 8.4) to 0  C, then the albedo linearly decreased by the albedo change parameter (Table 8.4). The visible snow albedo decreased by 0.1 and the infrared snow albedo decreased by 0.15 when

344

M. Leppäranta et al.

Fig. 8.12 A curvilinear tripolar grid configuration demonstrating higher horizontal resolution in the Arctic than in the lower latitudes. (From Madec and The NEMO Team 2015) Table 8.4 Adapted from Bi et al. (2013). Values of important CICE sea-ice model parameters to adjust the sea-ice distribution in the Australian ACCESS1.3 CMIP5 experiments Name Ocean-ice turning angle Ridging parameter value Cold deep snow albedo Bare ice albedo Melting deep snow albedo Temperature range to determine snow melting in albedo calculation Temperature range to determine bare ice melting in albedo calculation Albedo change to determine bare ice melting in albedo calculation Ice-ocean drag Minimum ice-ocean friction velocity Ice conductivity option Surface roughness of ice Ice-ocean heat exchange coefficient

Value 16 3 m½ 0.84 0.68 0.72 0.5 0.25 0.075 0.00536 0.0005 m s1 Bubbly 0.0005 m 0.004

the snow surface temperature rose from the temperature ranges to 0  C. The snow patchiness parameter impacted on how the albedo was averaged over a grid cell weighted by the ratio of ice and snow-covered portions as shown by Uotila et al. (2012). Essentially a higher snow patch value decreased the average albedo.

8 Sea Ice Modelling

345

In addition to the variables related to the short-wave radiation scheme, the value of the surface roughness of sea-ice affected the ice-atmosphere momentum and energy exchange and was a predefined constant in the model (see Table 8.4). Moreover, the ice-ocean energy and momentum exchange included predefined variables, such as the ice-ocean stress drag coefficient, stress turning angle and heat exchange coefficient (Table 8.4). Additionally, the minimum value of the ice-ocean friction velocity was predefined. The sensitivity of the sea-ice distribution to the parameters related to the oceanice heat exchange has been shown to be comparable or higher than its sensitivity to the parameters used to adjust the short-wave radiation scheme (Hunke 2010; Uotila et al. 2012). In terms of internal ice properties, the ice conductivity needed to be selected as well as the number of vertical layers in ice. The sea-ice deformation was affected by the values of the e-folding scale of ridged ice and the maximum thickness of rafted ice. This ridging parameter affected the sea-ice distribution to the same extent than albedo values (Uotila et al. 2012). In summary, finding the best possible values, within the realistic range, for these parameters is essential for a successful sea-ice simulation.

8.3.2.2

Louvain-la-Neuve Sea Ice Model, LIM

NEMO is a modelling framework designed for climate research and operational oceanography (Madec and The NEMO Team 2015). LIM2 has been the standard sea-ice model of the NEMO framework for almost a decade. In June 2015, a new and more sophisticated LIM version, LIM3.6 (denoted LIM3 for brevity), became available as the reference sea-ice model, coupled to the OPA ocean component, in the NEMOv3.6 framework. Notably, as LIM, also CICE is a commonly used sea-ice model in the NEMO framework. The ocean component OPA is a finite difference, hydrostatic, primitive equation ocean general circulation model. The vertical coordinate system is based on z levels with partial cell thicknesses allowed at the sea floor. The vertical mixing of tracers and momentum uses the TKE scheme (Gaspar et al. 1990; Blanke and Delecluse 1993). A quadratic bottom friction boundary condition is applied together with an advective and diffusive bottom boundary layer for temperature and salinity tracers (Beckmann and Döscher 1997). The model uses a non-linear variable volume scheme for the free surface, and an energy-enstrophy conserving scheme for momentum advection. A no-slip boundary condition is applied on the momentum equations with the horizontal Laplacian momentum diffusion. The tracer equations in OPA use the TVD advection scheme by Zalesak (1979) with the Laplacian diffusion along isoneutral surfaces. LIM2 is a simple mono-category dynamic-thermodynamic sea-ice model. Detailed descriptions of the model and the ocean-sea ice coupling are provided in Fichefet and Morales Maqueda (1997, 1999), and Timmermann et al. (2005). The thermodynamic component is the three-layer (one layer of snow and two layers of ice) formulation of Semtner (1976) to account for sensible heat storage and vertical

346

M. Leppäranta et al.

heat conduction within snow and ice. The effect of sub-grid-scale snow and ice thickness distribution is implicitly parameterized with the effective thermal conductivity. The storage of latent heat in ice, resulted from the trapping of shortwave radiation by brine pockets (salt water pockets in ice), is taken into account in a rather simplistic way. Vertical and lateral ice growth/decay rates are obtained from prognostic energy budgets at the upper and lower snow/ice interfaces and at lateral interfaces in leads. The surface snow/ice albedo depends on the state of the surface (frozen or melting), the snow depth, and ice thickness. When the load of snow is large enough to depress the snow-ice interface under the water level, seawater is supposed to infiltrate the entirety of the submerged snow and to freeze there, forming snow-ice. The model includes a lead parameterization. As in CICE; ice dynamics are simulated by assuming that sea ice behaves as a two-dimensional Elasto-ViscousPlastic (EVP) continuum in dynamical interaction with atmosphere and ocean (Hunke and Dukowicz 1997). The momentum transport equations are written in curvilinear, orthogonal coordinates and are numerically solved in the Arakawa C-grid following the formulation by Bouillon et al. (2009, 2013). The ice advective terms are computed using the second-order moment-conserving scheme of Prather (1986). In summary, the most relevant physical sea-ice processes, such as brine pockets, lateral melting, effective heat conduction due to unresolved sub-grid-scale ice thickness variations, surface albedo, penetration of radiation through the ice, snow ice formation, are implemented in LIM2. The new LIM3 code implements many sea-ice physics improvements compared to the previous LIM2 code. LIM3 advances reside mainly in the introduction of multiple ice categories to represent the sub-grid scale ice thickness distribution (Thorndike et al. 1975), and an explicit description of the multi-layer halo-thermodynamic component including the brine dynamics, prognostic equations for the sea-ice salinity, and their impact on ice thermal properties, such as the sea-ice conductivity, and ocean-ice salt exchanges when ice forms or melts. A detailed description of the LIM3 physics is given by Rousset et al. (2015). The LIM3 model features an adjustable vertical resolution with N ice layers, although, as in LIM2, two layers of ice and one of snow as these are recommended values according to the default NEMOv3.6 configuration. While the storage of latent heat in brine is highly parameterized in LIM2 using the heat reservoir, it is explicitly represented in LIM3, using a vertically varying salinity profile. Temporal salinity variations in LIM3 resolved using parameterizations of brine entrapment and drainage processes based on a simplification of the brine drainage model of Vancoppenolle et al. (2007). In both sea-ice models, ice is “levitating” (following the convention of Campin et al. 2008) over the ocean, and the growth and melt of ice impact the ocean mass and the salinity, but do not affect the pressure experienced by the ocean surface. LIM3 does include an explicit sub-grid-scale ice thickness distribution that enables to resolve the intense growth and melt of thin ice, including the stronger atmosphere-ocean heat fluxes through thin ice, as well as the redistribution of thinner ice onto thicker ice due to ridging and rafting. As with the number of ice layers, we follow the default NEMOv3.6 configuration and use five ice thickness categories.

8 Sea Ice Modelling

347

The behavior of an earlier version of LIM3 compared to LIM2 was investigated by Massonnet et al. (2011). They found that the multiple ice categories increase the seasonal to inter-annual variability of the sea-ice extent, that the LIM3 elasticviscous-plastic rheology enhances the response of ice to wind stress, and that the Arakawa C-grid formulation and the air-sea ice drag coefficient affect the simulated ice export through Fram Strait and the ice accumulation along the Canadian Arctic Archipelago. Compared to LIM2, these results indicate enhanced LIM3 sea-ice simulation capability due to physics improvements. Due to the more sophisticated physics, LIM3 can be expected to be better suitable for meso-scale sea-ice modelling than LIM2. Differences between LIM2 and LIM3 sea-ice physics provide an opportunity to explore the impacts of these differences for the simulated sea ice. Such analysis provides information on how important LIM3 different physics improvements are when enhancing the realism of the sea-ice simulation. From this, perspective, we compare simulated sea ice between two otherwise identical simulations except one carried out using LIM2 and another one with LIM3. These simulations are described in detail by Uotila et al. (2017).

8.3.3

Impact of Sea-Ice Physics Parameterizations

The configuration of the experiments diagnosed in the following paragraphs include significant improvements compared to earlier versions of NEMO-LIM. A major improvement is that LIM3 now entirely conserves the energy and mass due to the corrected numerics (Rousset et al. 2015). Another significant change is the new ocean component available in NEMO3.6_STABLE, released in July 2015. We also have 75 vertical ocean levels. In terms of the prescribed atmospheric states, we use the ones based on DRAKKAR forcing version 5.2 which is derived from the ECMWF reanalysis products and are found to perform better at high-latitudes than the NCEP/NCAR one (Bromwich et al. 2011). Then, we employ currently widely used CORE bulk formulae (Large and Yeager 2004; Griffies et al. 2009; Danabasoglu et al. 2014) at the ocean surface to calculate the energy, momentum and freshwater fluxes.

8.3.3.1

Sea-Ice Concentration and Extent

First, we discuss the geographical distribution of ice concentration, its mean seasonal cycle and interannual variations. Sea-ice concentration represents the fraction of ocean area covered by sea ice and is computed as the areal coverage in a given grid cell. Sea-ice extent here is defined as the total area of grid cells with sea-ice concentration of 15% or more. In September, the geographical distribution of LIM3 sea-ice concentration presents high values in the Canadian Arctic Archipelago with a realistic latitudinal decrease toward the Eurasian Arctic (Fig. 8.13a). LIM3 tends to generally

348

M. Leppäranta et al.

a

b

c

d

e

f

Fig. 8.13 Geographical distribution of Arctic sea-ice concentration averaged for September (a–c) and for March 2003–2012 (d–f). (a, d) show values simulated by LIM3, while (b, e) show LIM3 difference with Meier et al. (2013) passive microwave observations and (c, f) with LIM2. Sea-ice concentration differences are computed only where both values are present. Only areas where the sea-ice concentration is greater than 15% are plotted. In (a, d), thick black lines show the observed sea-ice edge as the 15% sea-ice concentration isopleth

underestimate the ice concentration, by ~20% in the central Arctic to ~50% in the East Siberian Sea, while the Laptev and Kara Seas are almost ice-free (Fig. 8.13). This negative summer sea-ice concentration bias is linked to an underestimation of sea-ice thickness in those areas both in winter and summer (see below). By contrast, too-large ice concentration is found in the Beaufort Sea (Fig. 8.13b). Clearly, the representation of ice concentration in the two models significantly differs in summer: LIM2 produces higher sea-ice concentration compared to LIM3 everywhere in the Arctic Ocean, and the difference increases radially from the Canadian Arctic Archipelago toward the Eurasian Arctic (Fig. 8.13c). While LIM2 is closer to observations in the Eurasian basin, it cannot reproduce the seasonal cycle of ice area in the Beaufort and East Siberian Seas toward the Bering Strait, where the sea-ice pack stays, unrealistically, rather uniform with a sharp transition to the ice edge and a too small open ocean fraction. The LIM3 March sea-ice edge is relatively well simulated (Fig. 8.13d) and the geographical distribution of ice concentration (Fig. 8.13e) is in close agreement with satellite estimates. LIM3 differs from observations over most of the Arctic ice pack just by a few per cent. Larger differences (reproduced in LIM2 as well) are located in the Atlantic marginal ice zone, where LIM3 overestimates its width and sea-ice

8 Sea Ice Modelling

a

349

b

Fig. 8.14 (a) Simulated (colored lines) and observed (black lines and grey shadings; NSIDC Fetterer et al. 2002) mean seasonal cycle of monthly mean sea-ice extent over the period 2003–2012 for the northern hemisphere (NH). The sea-ice extent is calculated as the area with sea-ice concentration 15% or more. Dashed lines and grey shadings denote the minimum and maximum annual monthly extents during the same period. In (b), annual maximum, mean and minimum time series of simulated and observed sea-ice extents in the NH over the period of 1958–2012 are presented. Red lines denote the standard NEMO-LIM3 simulation, green lines denote the standard LIM2 simulation, and blue lines denote the LIM3 single-category sea-ice simulation

concentration up to 40% along the East coast of Greenland, and in the Labrador and Barents Seas. Overall, LIM2 reproduces differences of similar magnitude, but it tends to uniformly overestimate observations and has a higher sea-ice concentration bias than LIM3 (Fig. 8.13f). Fourteen global ocean-ice models analyzed by Wang et al. (2016a) systematically produced too large sea-ice extents in March. This seems not the case here, which possibly indicates the effect of different atmospheric forcing set. Models analyzed by Wang et al. (2016a) were forced with COREII inter-annual data, which are NCEP based, while our simulations were forced with the DFS 5.2 data, which are ERA based. Mean seasonal cycles of the modelled sea-ice extents are shown in Fig. 8.14a, b together with the NSIDC observations, all averaged over the years 2003–2012. The LIM3 sea-ice extent closely follows the observed data and represents a clear improvement compared to LIM2, particularly in summer (Fig. 8.14a). The maximum (minimum) LIM3 sea-ice extent reaches 15.6 (5.2)  106 km2 in March (September). These values are fairly close to the NSIDC observed ones, which are 15.0 (5.1)  106 km2. The respective LIM2 values are 16.3 (8.3)  106 km2 and clearly too high. LIM2 does not manage to melt enough ice and systematically overestimates the NH sea-ice extent. The largest difference is found in summer (up to 40% from both LIM3 and observations), while it is reduced to 5% (10%) from LIM3 (observations) in winter. On the contrary, the LIM3 multi-category sea-ice thickness distribution allows for larger rates of melting due to its thin ice categories

350

M. Leppäranta et al.

compared to the mono-category LIM2 and enhances the seasonal cycle of sea-ice extent bringing it closer to observations. Associated with the better mean seasonal cycle, the inter-annual time series of sea-ice extent is improved in LIM3 compared to LIM2 (Fig. 8.14b). The correlation between the simulated and NSIDC observed monthly mean sea-ice extent anomalies is 0.87 for both LIM3 and LIM2, when all months are taken into account. For March only, the LIM3-NSIDC correlations become somewhat higher (0.90) compared to LIM2-NSIDC one (0.86). In both LIM models, the simulated inter-annual variability agrees well with the NSIDC inter-annual variability being larger for the September time series than for the March time series. Although the spatial distribution of LIM3 sea-ice concentration differs from the observed one, both the maximum and minimum sea-ice extent are well reproduced by LIM3, as shown by the time series in Fig. 8.14b that closely follow the NSIDC data in 1979–2012. Moreover, LIM3 realistically captures most of the summer minimum records, with the 2007 minimum extent of 3.4  106 km2, compared to the observed 4.3  106 km2. As for the seasonal cycle, LIM2 systematically overestimates yearly minimum, maximum and mean sea-ice extents during the whole period of integration, for example the 2007 minimum is overestimated by 50%. The difference between the two LIM models is clear from the first few simulated years, when the LIM2 sea-ice loss from the initial state is too small to bring the mean sea-ice extent closer to observed values. The two LIM models show comparable negative sea-ice extent trends in March, which are less negative than satellite observed trends (Table 8.5). These linear trends were calculated for the 1979–2012 period. In September, the LIM3 trend is close to the observed one, while the LIM2 negative trend is too small. As concluded by Wang et al. (2016a), LIM2, which overestimates the Arctic sea-ice thickness, has too low September trend, while LIM3, which has thinner ice, produces a realistic September trend. Although these 30-year sea-ice extent trends look reasonable for LIM3, there is some peculiar inter-decadal variability that deserves a closer look. Contrary to the observed, the LIM3 NH annual mean sea-ice extent increases in the mid-1990s and then stays a relatively constant before a sharp decline in the 2000s (Fig. 8.14b). Reasons for this unrealistic looking sea-ice extent behavior are Table 8.5 September and March hemispheric sea-ice extent anomaly trends 1979–2012 as 106 km2 per decade. Trends are statistically significant at the 95% level (p-value