Remote Sensing of Geomorphology, Volume 23 [23] 0444641777, 9780444641779

Remote Sensing of Geomorphology, Volume 23, discusses the new range of remote-sensing techniques (lidar, structure from

885 143 113MB

English Pages 300 [398] Year 2020

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Remote Sensing of Geomorphology, Volume 23 [23]
 0444641777, 9780444641779

Table of contents :
Front Cover
Remote Sensing of Geomorphology
Copyright
Contents
Contributors
Foreword
Reference
Introduction to remote sensing of geomorphology
Chapter 1: Structure from motion photogrammetric technique
1. Introduction
1.1. Brief historical summary and state of the art
1.2. Reasons for success in geomorphological surveys
2. Method
2.1. Choosing suitable settings to comply with the application at hand
2.1.1. Image quality
2.1.2. Ground sampling distance
2.1.3. Image network geometry
2.1.4. Camera parameter choice during bundle adjustment
2.1.5. Referencing: GCP weights and distribution
2.1.6. Exterior influences
2.2. Accuracy considerations in geomorphological applications
2.3. Direct geo-referencing (DG) for flexible UAV applications
2.3.1. Achievable accuracies
2.3.2. Guidelines for DG applications
3. Reconstructing processes across space
4. Reconstructing processes in time
4.1. Past and real-time reconstruction
4.2. Time-lapse imagery for 4D change detection
4.2.1. Guidelines for time-lapse SfM photogrammetry
5. Final remarks
References
Further reading
Chapter 2: Topo-bathymetric airborne LiDAR for fluvial-geomorphology analysis
1. High-resolution topography: Where is the bathymetry?
2. Synoptic fluvial bathymetry survey techniques
2.1. Topo-bathymetric lidar vs existing approaches
2.2. Topo-bathymetric airborne lidar sensors
2.3. Survey examples and typical data characteristics
3. Controls on depth penetration and surveyable rivers
3.1. Theoretical controls on the bathymetric waveform and bottom echo intensity
3.2. Results on maximum measurable depth and sensor comparison
3.3. Depth uncertainty and detail resolving capability
3.4. Surveyable rivers and survey strategy
4. Data processing
4.1. Water-surface detection, bathymetric classification, and refraction correction
4.2. FWF analysis
5. Applications in fluvial geomorphology
5.1. Multi-scale high-resolution fluvial geomorphology
5.2. Coupling with 2D-3D hydraulic modeling
5.3. Synoptic channel morphodynamics and sediment budget
6. Conclusions and remaining challenges
6.1. A priori prediction of depth penetration and river bathymetric cover
6.2. Automatic classification on massive lidar datasets
6.3. FWF analysis in the context of fluvial environments
6.4. Large-scale hydraulic modeling on topo-bathymetric data
Acknowledgments
References
Chapter 3: Ground-based remote sensing of the shallow subsurface: Geophysical methods for environmental applications
1. Introduction
2. Methods
2.1. Geo-electrical (DC resistivity) methods
2.2. EMI methods and GPR
2.3. Seismics
3. Application examples
3.1. System structure
3.1.1. The Settolo site
3.1.2. The Trecate site
3.1.3. The Aviano site
3.1.4. The Fondo Paviani site
3.1.5. The Turriaco site
3.2. Fluid dynamics monitoring
3.2.1. The Decimomannu site
3.2.2. The Trento Nord site
3.2.3. The Grugliasco site
3.2.4. The Bregonze site
3.2.5. The Bari IRSA-CNR site
4. Future challenges and conclusions
Acknowledgments
References
Further reading
Chapter 4: Topographic data from satellites
1. The importance of topography
2. Collection of topographic data from satellites
2.1. Satellite lidar
2.2. Radar
2.3. Stereo imaging
3. Global and large regional datasets
3.1. GTOPO30
3.2. SRTM
3.3. ASTER
3.4. ALOS PRISM
3.5. TanDEM-X
3.6. ArcticDEM and REMA
3.7. High Himalaya DEM
3.8. MERIT DEM
3.9. Other instruments and summary
4. Accuracy of global datasets
4.1. Common sources of error
4.2. Methods of comparison between datasets
4.3. Error estimates for specific datasets
4.3.1. SRTM accuracy
4.3.2. ASTER accuracy
4.3.3. ALOS world 3D accuracy
4.3.4. TanDEM-X DEM accuracy
4.3.5. MERIT DEM accuracy
4.3.6. ArcticDEM, REMA, and High Mountain Asia DEM accuracy
4.4. Dataset intercomparison
4.5. Summary of vertical accuracy
5. Implications of increasing resolution on geomorphic studies
5.1. Geomorphic metrics and data processing
5.2. Simple preprocessing
5.2.1. Grid resolution: Implications for curvature and slope measurements
5.3. Accuracy of channel profiles
6. Future developments
7. Conclusions
References
Chapter 5: Linking life and landscape with remote sensing
1. Introduction
2. Linking remote sensed data to life and landscapes
2.1. Erosive, depositional, and constructive processes modulated by biota
2.2. Life and landscape patterns
2.3. Measureable vegetation properties
2.4. Soils and belowground organic carbon
3. Passive remote sensing methods
3.1. Vegetation indicators from passive instruments
3.2. Coarse resolution passive sensors
3.3. Medium and fine resolution passive sensors
4. Radar
4.1. Satellite-based radar systems
5. Lidar
5.1. A primer on lidar remote sensing
5.2. Quantifying canopy structure with airborne lidar
5.2.1. Canopy height models and canopy gaps
5.2.2. Identifying individual trees
5.2.3. Mapping AGB and ACD
5.2.3.1. Area-based approaches
5.2.3.2. Individual-based approaches
5.2.3.3. Calibration and uncertainty
5.2.4. Quantifying PAI and vertical distributions of plant area density
5.3. Spaceborne lidar
5.3.1. ICESat/GLAS
5.3.2. GEDI
5.3.3. ICESat-2/ATLAS
5.4. Data fusion
6. Airborne electromagnetics
7. Conclusions
7.1. Finding the right sensor
7.2. The importance of scale
7.3. Trade-offs between resolution and spatial coverage
7.4. Future outlook
Acknowledgments
References
Chapter 6: SfM photogrammetry for GeoArchaeology
1. Remote sensing
2. SfM photogrammetry
3. SfM in geoarchaeology: Agricultural terraces in Europe
3.1. Case study: Ingram Valley (UK)
3.2. SfM workflow
3.2.1. Fieldwork
3.2.2. SfM processing
3.2.3. SfM postprocessing
3.2.4. DTM generation
3.3. Result and discussion
4. Final remarks
Acknowledgments
References
Chapter 7: Landslide analysis using laser scanners
1. Introduction
2. A short history
3. Basics of laser scanners
3.1. Lasers and safety
3.2. LiDAR devices
3.3. TOF LiDAR
3.3.1. LiDAR using phase measurements
3.3.2. LASER scanning based on triangulation
3.4. Beam characteristics
4. LiDAR uses
4.1. Issues
4.2. Different LiDAR configurations
4.3. Filtering
4.4. Georeferencing and coregistration
5. Characterization of landslides
5.1. Mapping
5.2. Rock structure characterization and rockfall sources
5.3. Volume estimation
6. Monitoring
6.1. Surface changes
6.2. Potential methods for real-time monitoring
7. Modeling based on LDTM
8. Discussion and perspectives
Acknowledgments
References
Further reading
Chapter 8: Terrestrial laser scanner applied to fluvial geomorphology
1. Challenges in using terrestrial laser scanner to understand river dynamics
2. Data acquisition
2.1. Equipment consideration
2.2. Data registration, georeferencing, and survey strategy
2.3. Boundary conditions monitoring and long-term monitoring
3. 3D point cloud postprocessing operations
3.1. Point cloud registration and preprocessing
3.2. Vegetation classification
3.3. Point-based vs raster-based analysis
3.4. Core points as a way to cope with data volume and spatial variations in point density
3.5. Metrics calculation and segmentation
3.6. 3D spatial analysis
4. Topographic change measurement and volume calculation
4.1. Source of uncertainties
4.2. Vertical change detection
4.3. 3D distance and bank erosion measurement
5. Science from point clouds in fluvial geomorphology
5.1. Grain size distribution
5.2. Sediment transport and bank erosion
5.3. Bedrock erosion
5.4. Vegetation, hydraulics, and sedimentation
6. Conclusion and outlook
Acknowledgments
References
Chapter 9: Remote sensing for the analysis of anthropogenic geomorphology: Potential responses to sediment dynamics in th ...
1. Introduction
2. Materials and methods
2.1. Relative path impact index
2.2. Connectivity index
3. Study area
3.1. Spain
3.2. Italy
4. Results
4.1. Spain
4.2. Italy
5. A holistic view of land planning
6. Conclusions
Acknowledgments
References
Further reading
Chapter 10: Using UAV and LiDAR data for gully geomorphic changes monitoring
1. Introduction
1.1. LiDAR in geosciences
1.2. Digital photogrammetry and SfM in geosciences
2. Study area: The reservoir bottom gullies from Jijia Hills (Romania)
3. Materials and methods
3.1. LiDAR data
3.2. UAV images
3.3. Structure from motion
3.3.1. SfM approach
3.3.2. Point cloud postprocessing
3.4. DEM generation
3.5. Geomorphic change detection
3.6. Geomorphological mapping
4. Results
5. Discussions
6. Conclusions
Acknowledgments
References
Further reading
Chapter 11: Zero to a trillion: Advancing Earth surface process studies with open access to high-resolution topography
1. Introduction
2. Scientific motivations for open access to topographic data
3. Broad impacts from openly available topographic data
4. OpenTopography overview and impact
5. OpenTopography partnerships
6. Lessons learned and challenges for supporting open access to topographic data
7. Outlook
Theme 1: A cloud and HPC platform for scalability and sustainability
Theme 2: Enabling community innovation
8. Conclusions
Acknowledgments
References
Chapter 12: Reproducible topographic analysis
1. Topographic analysis and (reproducible) geomorphology
2. Scientific reproducibility
2.1. Reproducibility or replicability?
2.2. Benefits of reproducible research
3. Reproducibility in the context of topographic analysis for geomorphology
3.1. Initial observations of landscape form
3.2. Paper contour map analysis
3.3. The beginning of computational topographic analysis
3.3.1. Vector representations of elevation data
3.3.2. Gridded representations of elevation data
3.4. The reproducibility of early computational topographic analysis
3.5. Modern topographic analysis
3.6. The reproducibility of modern topographic analysis
4. Barriers to reproducible topographic analysis
4.1. Topographic analysis workflows
4.2. Data
4.3. Paywalls
5. Making topographic analysis reproducible
5.1. Workflows
5.2. Data
5.3. Paywalls
5.4. Our recommendations
6. Conclusions
References
Index
Back Cover

Citation preview

Developments in Earth Surface Processes

REMOTE SENSING OF GEOMORPHOLOGY VOLUME 23

Developments in Earth Surface Processes, 23 Series Editor – J.F. Shroder, Jr. For previous volumes refer http://www.sciencedirect.com/science/bookseries/09282025

Developments in Earth Surface Processes

REMOTE SENSING OF GEOMORPHOLOGY VOLUME 23 Volume Editors

PAOLO TAROLLI Department of Land, Environment, Agriculture and Forestry, University of Padova, Legnaro (PD), Italy

SIMON M. MUDD University of Edinburgh, School of GeoSciences, Edinburgh, United Kingdom

Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands The Boulevard, Langford Lane, Kidlington, Oxford, OX5 1GB, United Kingdom 50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States First edition 2020 © 2020 Elsevier B.V. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www. elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library ISBN: 978-0-444-64177-9 ISSN: 0928-2025 For information on all Elsevier publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Candice Janco Acquisitions Editor: Amy Shapiro Editorial Project Manager: Kelsey Connors Production Project Manager: Bharatwaj Varatharajan Cover Designer: Greg Harris Typeset by SPi Global, India

Contents Contributors Foreword Introduction to remote sensing of geomorphology

3. Ground-based remote sensing of the shallow subsurface: Geophysical methods for environmental applications

ix xi

Giorgio Cassiani, Jacopo Boaga, Ilaria Barone, Maria Teresa Perri, Gian Piero Deidda, Giulio Vignoli, Claudio Strobbia, Laura Busato, Rita Deiana, Matteo Rossi, Maria Clementina Caputo, and Lorenzo De Carlo

xiii

1. Structure from motion photogrammetric technique

1 Introduction 56 2 Methods 56 3 Application examples 67 4 Future challenges and conclusions Acknowledgments 83 References 83 Further reading 89

Anette Eltner and Giulia Sofia

1 Introduction 1 2 Method 5 3 Reconstructing processes across space 12 4 Reconstructing processes in time 5 Final remarks 18 References 18 Further reading 24

80

14

4. Topographic data from satellites Simon M. Mudd

1 2 3 4 5

The importance of topography 91 Collection of topographic data from satellites Global and large regional datasets 97 Accuracy of global datasets 105 Implications of increasing resolution on geomorphic studies 113 6 Future developments 118 7 Conclusions 119 References 120

2. Topo-bathymetric airborne LiDAR for fluvial-geomorphology analysis Dimitri Lague and Baptiste Feldmann

1 High-resolution topography: Where is the bathymetry? 26 2 Synoptic fluvial bathymetry survey techniques 27 3 Controls on depth penetration and surveyable rivers 32 4 Data processing 41 5 Applications in fluvial geomorphology 44 6 Conclusions and remaining challenges 50 Acknowledgments 52 References 52

5. Linking life and landscape with remote sensing David T. Milodowski, Steven Hancock, Sonia Silvestri, and Simon M. Mudd

1 Introduction 130 2 Linking remote sensed data to life and landscapes 131

v

92

vi

Contents

3 Passive remote sensing methods 137 4 Radar 141 5 Lidar 143 6 Airborne electromagnetics 161 7 Conclusions 164 Acknowledgments 165 References 166

6. SfM photogrammetry for GeoArchaeology Sara Cucchiaro, Daniel J. Fallu, Pengzhi Zhao, Clive Waddington, David Cockcroft, Paolo Tarolli, and Antony G. Brown

1 Remote sensing 183 2 SfM photogrammetry 185 3 SfM in geoarchaeology: Agricultural terraces in Europe 187 4 Final remarks 200 Acknowledgments 200 References 201

7. Landslide analysis using laser scanners Michel Jaboyedoff and Marc-Henri Derron

1 Introduction 207 2 A short history 210 3 Basics of laser scanners 211 4 LiDAR uses 214 5 Characterization of landslides 216 6 Monitoring 219 7 Modeling based on LDTM 222 8 Discussion and perspectives 223 Acknowledgments 225 References 226 Further reading 230

8. Terrestrial laser scanner applied to fluvial geomorphology Dimitri Lague

1 Challenges in using terrestrial laser scanner to understand river dynamics 232 2 Data acquisition 233 3 3D point cloud postprocessing operations 237 4 Topographic change measurement and volume calculation 243

5 Science from point clouds in fluvial geomorphology 247 6 Conclusion and outlook 251 Acknowledgments 251 References 251

9. Remote sensing for the analysis of anthropogenic geomorphology: Potential responses to sediment dynamics in the agricultural landscapes Paolo Tarolli and Giulia Sofia

1 Introduction 255 2 Materials and methods 257 3 Study area 259 4 Results 260 5 A holistic view of land planning 264 6 Conclusions 267 Acknowledgments 267 References 267 Further reading 269

10. Using UAV and LiDAR data for gully geomorphic changes monitoring Mihai Niculița˘, Mihai Ciprian Ma˘rga˘rint, and Paolo Tarolli

1 Introduction 271 2 Study area: The reservoir bottom gullies from Jijia Hills (Romania) 274 3 Materials and methods 276 4 Results 286 5 Discussions 297 6 Conclusions 305 Acknowledgments 305 References 305 Further reading 315

11. Zero to a trillion: Advancing Earth surface process studies with open access to high-resolution topography Christopher J. Crosby, J Ramo´n Arrowsmith, and Viswanath Nandigam

1 Introduction 318 2 Scientific motivations for open access to topographic data 318

vii

Contents

3 Broad impacts from openly available topographic data 320 4 OpenTopography overview and impact 321 5 OpenTopography partnerships 328 6 Lessons learned and challenges for supporting open access to topographic data 328 7 Outlook 331 8 Conclusions 333 Acknowledgments 333 References 333

12. Reproducible topographic analysis Stuart W.D. Grieve, Fiona J. Clubb, and Simon M. Mudd

1 Topographic analysis and (reproducible) geomorphology 339 2 Scientific reproducibility 340 3 Reproducibility in the context of topographic analysis for geomorphology 344 4 Barriers to reproducible topographic analysis 354 5 Making topographic analysis reproducible 357 6 Conclusions 362 References 362

Index

369

This page intentionally left blank

Contributors J. Ramo´n Arrowsmith School of Earth and Space Exploration, Arizona State University, Tempe, AZ, United States

Anette Eltner Institute of Photogrammetry and Remote Sensing, Technische Universit€ at Dresden, Dresden, Germany

Ilaria Barone Dipartimento di Geoscienze, Universita` di Padova, Padova, Italy

Daniel J. Fallu Tromso University Museum, UiT The Artic University of Norway, Tromsø, Norway

Jacopo Boaga Dipartimento di Geoscienze, Universita` di Padova, Padova, Italy

Baptiste Feldmann Univ Rennes, CNRS, Nantes-Rennes Topo-bathymetric Lidar platform, OSUR, UMS 3343, Rennes, France

Antony G. Brown Tromso University Museum, UiT The Artic University of Norway, Tromsø, Norway; Geography and Environmental Science, University of Southampton, Southampton, United Kingdom

Stuart W.D. Grieve Queen Mary University of London, London, United Kingdom Steven Hancock University of Edinburgh, School of GeoSciences, Edinburgh, United Kingdom

Laura Busato Department of Agricultural Sciences, University of Naples Federico II, Naples, Italy Maria Clementina Caputo Italy

Michel Jaboyedoff ISTE—Institute of Earth Sciences, Risk-Group, GEOPOLIS-3793, University of Lausanne, Lausanne, Switzerland

IRSA CNR, Bari,

Dimitri Lague Univ Rennes, CNRS, Geosciences Rennes, UMR 6118, Rennes, France

Giorgio Cassiani Dipartimento di Geoscienze, Universita` di Padova, Padova, Italy

Mihai Ciprian Ma˘rga˘rint Department of Geography, Faculty of Geography and Geology, Alexandru Ioan Cuza University of Iași, Iași, Romania

Fiona J. Clubb Durham University, Durham, United Kingdom David Cockcroft Archaeological Research Services Ltd, Bakewell, DE, United Kingdom Christopher J. Crosby United States

David T. Milodowski University of Edinburgh, School of GeoSciences; University of Edinburgh, National Centre for Earth Observation, Edinburgh, United Kingdom

UNAVCO, Boulder, CO,

Sara Cucchiaro Department of Land, Environment, Agriculture and Forestry, University of Padova, Legnaro, Padova, Italy

Simon M. Mudd University of Edinburgh, School of GeoSciences, Edinburgh, United Kingdom

Lorenzo De Carlo IRSA CNR, Bari, Italy

Viswanath Nandigam San Diego Supercomputer Center—UC San Diego, La Jolla, CA, United States

Rita Deiana Dipartimento di Beni Culturali (dBC), Universita` di Padova, Padova, Italy Gian Piero Deidda Dipartimento di Ingegneria Civile, Ambientale e Architettura, Universita` di Cagliari, Cagliari, Italy

Mihai Niculița˘ Department of Geography, Faculty of Geography and Geology, Alexandru Ioan Cuza University of Iași, Iași, Romania

Marc-Henri Derron ISTE—Institute of Earth Sciences, Risk-Group, GEOPOLIS-3793, University of Lausanne, Lausanne, Switzerland

Maria Teresa Perri Dipartimento di Geoscienze, Universita` di Padova, Padova, Italy

ix

x

Contributors

Matteo Rossi Engineering Geology Lund University, Lund, Sweden

(LTH),

Sonia Silvestri University of Bologna, Department of Biological, Geological and Environmental Sciences, Bologna, Italy; Duke University, Nicholas School of the Environment, Durham, NC, United States Giulia Sofia Department of Civil & Environmental Engineering, University of Connecticut, Storrs, CT, United States Claudio Strobbia France

Realtimeseismic SA, Pau,

Paolo Tarolli Department of Land, Environment, Agriculture and Forestry, University of Padova, Legnaro, Padova, Italy Giulio Vignoli Dipartimento di Ingegneria Civile, Ambientale e Architettura, Universita` di Cagliari, Cagliari, Italy Clive Waddington Archaeological Research Services Ltd, Bakewell, DE, United Kingdom Pengzhi Zhao Earth & Life Institute, Universite Catholique de Louvain, Louvain-la-Neuve, Belgium

Foreword This latest volume on Remote Sensing of Geomorphology edited by Paolo Tarolli and Simon Mudd gives a refreshing new look at a group of topics and processes that have greatly intrigued quite a number of scientists most interested in using new technologies to investigate the landforms of our home planet. Once global overviews of regional landforms became a common endeavor more than 30 years ago (Short and Blair, 1986), the use of myriads of satellite sensing systems and new technologies and methods to assess various environmental parameters became more common. As the imagery platforms and technologies continue to improve, in fact, the methodologies developed here can also be used to assess a variety of extraterrestrial bodies as well. No doubt we will continue to use the Earth-bound term “geo” morphology to refer to the landforms on many other such bodies in space as well, even though that would be a bit of an etymological misnomer. Still, the newer methodologies discussed in this book do point to the many interesting ways of looking at near-surface and surficial landforms, and continue to break new ground. For people who are relatively new to these technologies, the rather arcane terminology, even obscure jargon, and profuse uncertain acronyms can be somewhat disheartening to those not prepared to work with the practitioners of these disciplines. Nevertheless, probably in any profoundly new area of technology such as this assessment of remote sensing of geomorphology, a certain willingness to tolerate a measure of personal

confusion while immersing oneself in the details can help facilitate understanding later. Most scientists seem to be able to wade through the technical weeds, as it were, provided that the results one achieves at the end lead to new viewpoints and useful results. This volume of papers by a number of specialists stewarded by Tarolli and Mudd can be viewed as such because they offer some introduction into new methodologies, understandings, and terminologies. Sorting out the blizzards of acronym names is just one of the ancillary benefits. Currently, most geomorphologists are at least reasonably familiar with digital elevation models (DEMs) of natural and anthropogenic topography as well as various types of scanning to obtain variable images of objects or ground surfaces. Less well understood by many scientists, for example, are the considerable variations also representing topography through the use of multiple different means such as digital terrain model (DTM), digital surface model (DSM), and triangular irregular network (TIN), and the fact that these usages differ from one country to another. Of course, once an image or a picture of any feature is obtained in a digital form capable of being measured and assessed in different ways, then the science gained can be quite impressive. Once resolutions of topographic representations become detailed enough, certain tell-tale landform structures emerge from noisy datasets and significant understandings can be obtained of process mechanics and chronologies of superposition events. In combination with the diverse new

xi

xii

Foreword

surficial age-dating techniques now available, the many varieties of remote sensing of geomorphology offer more advanced assessments of the geomorphology of any place. Structure-from-motion (SfM) photogrammetry, hyper-scale, three-dimensional landform models, high-resolution topography (HRT), laser scanning or light detection and ranging (LiDAR) point clouds of topographic data obtained from the space, aerial, or terrestrial devices, unmanned aerial vehicles (UAV) or drones, geomorphic change detection (GCD), and DEMs of differences (DoD) are all new remote-sensing methodologies applied to various fluvial, agricultural, landslide, archeological, anthropogenic, subsurface, and other geomorphic processes discussed in

this volume. These diverse new methodologies are not exhaustive coverage of new disciplines and methodologies, or of satellite platforms, but do present useful discussions that will enable readers to better understand many of the new remote-sensing technologies. John F. Shroder, Jr. Editor-in-Chief Developments in Earth Surface Processes October 25, 2019

Reference Short, N.M., Blair Jr., R.W., 1986. Geomorphology From Space: A Global Overview of Regional Landforms. NASA Scientific and Technical Information Branch, National Aeronautics and Space Administration, Washington, DC.

Introduction to remote sensing of geomorphology The Earth’s surface has fascinated scientists for centuries. For well over a hundred years, scientists have speculated about the relationship between surface topography and the processes that lead to landforms. The first topographic maps at a national scale were published by France in the late 18th century, but these did not become widespread until the late 19th century. At that time, a number of scientists began speculating on quantitative relationships between uplift, erodibility, hydrology, and sediment transport. Surveying by national agencies produced contour maps that could be used to extract data such as slope profiles or drainage areas; Ordnance Survey of the UK began producing contour maps of scale 6 in to the mile (approximately 1:10,000) in the late 1920s and early 1930s, and the USGS (US Geological Survey) began national mapping at 1:24,000 scale. Testing of hypotheses developed early in the 20th century began in earnest as workers in the 1930s, 1940s, and 1950s began using intensively measured landscapes: the Perth Amboy badlands made famous by the seminal work of Schumm (1956) were mapped at a scale of 1 in to 10 ft by Strahler and Coates in 1948. The collection of such data was labor intensive, however. This is no longer the case. Advances in the field of physics have yielded instruments that can collect vast quantities of data remotely: the volume of data at our fingertips at present is beyond the wildest dreams of late 19th century scientists. Global topographic data have gone from 1 km

resolution in 1996 to a global 90 m dataset in 2004, a global 30 m dataset in 2009, down to a global 5 m dataset in 2018. In parallel, LiDAR (light detection and ranging) technology allows very high-density topographic and point cloud data to be collected using both terrestrial and airborne instruments. Point densities greater than 20 points per square meter are now routinely collected with submeter (and sometimes subcentimeter) accuracy using airborne LiDAR instruments. The impact of satellites and terrestrial and airborne LiDAR instruments cannot be understated, but the cost of satellites means that new missions are primarily funded by national agencies, and collection of LiDAR data relies on relatively expensive instruments. In contrast, nearly every mobile phone contains a camera, and drone-mounted cameras are far less expensive than LiDAR instruments. The advances in the structure-from-motion (SfM) photogrammetric technique, where multiple images are used to create point clouds, have also advanced to the point of being routine. Today, anyone with a camera can produce dense point clouds of threedimensional objects. Remote-sensing instruments are useful for a range of applications relevant to geomorphology that go far beyond measuring topography. Radar interferometry can be used to detect very subtle (centimeter-scale) ground motions; radar and multispectral data can be used to detect vegetation. A range of

xiii

xiv

Introduction to remote sensing of geomorphology

geophysical methods can be used to detect what lies beneath the ground surface. Collectively, these remote-sensing techniques allow geomorphologists to quantify a rich array of landscape properties that can help them understand both intrinsic and extrinsic factors that shape the landscape, and can help them understand landscape history. In this book you will find chapters reviewing and exploring state-of-the-art techniques in remote sensing relevant to geomorphology. We hope that the chapters will serve as both a reference for experienced practitioners, and a guide to geomorphologists looking to use remote-sensing techniques to benefit their studies. We first have several chapters on specific techniques: Chapter 1 describes advances in the SfM photogrammetric techniques that allow generation of 3D terrain models using overlapping images acquired from different perspectives with standard compact cameras (including smartphone cameras) and georeferencing information. Chapter 2 describes the opportunities offered by the green laser; such lasers (characterized by a wavelength of λ ¼ 532 nm) can penetrate shallow water and therefore provide bathymetry of rivers, lakes, and estuaries. We also have chapters giving an overview of the remote-sensing instruments and datasets used to quantify specific landscape properties. Chapter 4 explores the instruments used and techniques available for generating topographic data from space. This chapter reviews the accuracy and availability of topographic datasets and discusses the implications for geomorphic research. Chapter 5 explores the remote-sensing techniques that can quantify features of living organisms that are likely to influence, and be influenced by, geomorphic processes; the focus is on vegetation. Chapter 3 presents a review of geophysical methods for the characterization of shallow subsurface.

The book contains chapters on specific applications of remote-sensing data as well. Chapter 8 explores the potential of terrestrial laser scanners (TLS) to solve problems in fluvial geomorphology, synthesizing examples of data acquisition, processing methods, and applications. This instrument offers an unprecedented combination of subcentimeter resolution that allows workers to capture the geometry of individual pebbles and quantify precisely the spatial variability of channel evolution. Chapter 7 explores the capability of laser scanning in the quantification of volumes, understanding mechanisms, and timing of landslide and rockfall events. The basics of LiDAR performance are reviewed and an overview of the advantages and limitations of this 3D data acquisition technique are presented. Chapter 10 combines UAV optical imagery and LiDAR data to evaluate the rate of process for four reservoir bottom gully systems between two temporal frames. Chapter 9 explores how high-resolution topography can help understand how humans are increasingly modifying the Earth’s surface. The chapter focuses on agricultural landscapes and shows how new remote-sensing technologies (e.g., airborne LiDAR), available to the public, can provide a better understanding of the interaction between anthropogenic elements, potential erosion, and associated sediment delivery. Chapter 6 shows how drones and the structure-from-motion technique can be used to quantify the history of land use and land modification in an archeological context. Geoarcheological studies have benefitted from new technological developments in remote-sensing technologies that have become an integral and important part of archeological research. In particular, structure-from-motion (SfM) photogrammetry is one of the most successful emerging techniques in high-resolution topography (HRT) and provides exceptionally fast,

Introduction to remote sensing of geomorphology

low-cost, and easy 3D survey for geoscience applications. Finally, the book includes two chapters on open data. Chapter 11 describes the efforts of Opentopography.org to archive and distribute high-resolution topographic data. The authors highlight the fact that open access to these data and a cyberinfrastructure platform that enables users to discover, manage, share, and process them increases the impact of investments on data collection and catalyzes scientific discovery. Furthermore, open and online access to data enables broad interdisciplinary use of high-resolution topography across academia and in communities such as education, public agencies, and commercial sector. Chapter 12 discusses techniques to ensure processing of remotely sensed data in geomorphic and other applications is reproducible in the chapter. They present clear definitions of the terms “replicable”

xv

and “reproducible” for geomorphic research and communicate the importance of performing reproducible analysis of remotely sensed topographic data. As editors, we are genuinely excited about the opportunities afforded by the range of remote-sensing data that is now easily accessible to geomorphologists. We hope that by offering examples of various datasets available for geomorphic investigations we can spur even more uptake of remotely sensed data than is currently the case. Paolo Tarolli Department of Land, Environment, Agriculture and Forestry, University of Padova, Legnaro (PD), Italy Simon M. Mudd University of Edinburgh, School of GeoSciences, Edinburgh, United Kingdom

This page intentionally left blank

C H A P T E R

1 Structure from motion photogrammetric technique Anette Eltnera, Giulia Sofiab a

Institute of Photogrammetry and Remote Sensing, Technische Universit€at Dresden, Dresden, Germany bDepartment of Civil and Environmental Engineering, University of Connecticut, Storrs, CT, United States

O U T L I N E 1 Introduction 1.1 Brief historical summary and state of the art 1.2 Reasons for success in geomorphological surveys

1

2 Method 2.1 Choosing suitable settings to comply with the application at hand 2.2 Accuracy considerations in geomorphological applications 2.3 Direct geo-referencing (DG) for flexible UAV applications

5

3 Reconstructing processes across space

2

4 Reconstructing processes in time 4.1 Past and real-time reconstruction 4.2 Time-lapse imagery for 4D change detection

4

6 8 9

12 14 14 16

5 Final remarks

18

References

18

Further reading

24

1 Introduction Structure from motion (SfM) photogrammetry provides hyper-scale three-dimensional (3D) landform models using overlapping images acquired from different perspectives with standard compact cameras (including smartphone cameras) and geo-referencing information. As applied to the remote sensing of geomorphology, it is not so much a single technique,

Developments in Earth Surface Processes, Volume 23 https://doi.org/10.1016/B978-0-444-64177-9.00001-1

1

© 2020 Elsevier B.V. All rights reserved.

2

1. Structure from motion photogrammetric technique

but rather a workflow employing multiple algorithms developed from computer vision, traditional photogrammetry, and more conventional survey techniques (Carrivick et al., 2016). Recent literature has provided reviews on the importance of SfM in geosciences (Carrivick et al., 2016; Eltner et al., 2016; Smith et al., 2016) or specific scientific contexts (Mancini et al., 2013; Dietrich, 2016; Entwistle et al., 2018). This contribution builds on the existing literature, to provide a showcase of the technology, relevant to the remote sensing of geomorphology.

1.1 Brief historical summary and state of the art The roots of SfM lie in two key fields: photogrammetry and computer vision. When techniques from these fields are combined with both automation and precision, the result is a comprehensive tool (Pierrot-Deseilligny and Clery, 2011) for geomorphological applications. Photogrammetry is a relatively old technique (Slama et al., 1980). In this field, the reconstruction efforts of pioneers in the 1840s initially attempted using a pair of ground cameras separated by a fixed baseline and followed by applications using cameras for estimating the shape of the terrain from ground and aerial photographs (Maybank, 1993). With the introduction of aeroplanes and space photography, the development of photogrammetry flourished, with 2D photographs used to rectify images into appropriate coordinates, or mosaicking multiple frames to estimate structures or ground elevation. In a parallel effort, the computer vision community provided the first early algorithms for 3D scene reconstructions by stereo images (Marr and Poggio, 1976) or to pioneer work on motion-based reconstruction (Ullman, 1979). The prime formalisms derived in these two communities provided the most important foundational theory for the SfM community. However, advances in SfM have been spurred mostly due to the wide range of modern applications. A search in the academic publications database Web of Sciences (WoS) for Structure from Motion (made in August 2018) delivered >3000 records since the early 1980s (Fig. 1), covering as many as 125 fields of study. Computer science and artificial intelligence is the category with the most counts of that phrase. Engineering is ranked second, remote sensing is fourth, and geosciences is currently ranked sixth. This wide range of applications of SfM results in research with different goals, hence emphasizing multiple ways of addressing SfM problems in space and time. The computer vision field features much older publications than other fields, with the first papers published in the 1980s (Bolles et al., 1987) introducing a technique for building a 3D description of a static scene from a dense sequence of images, and the latest (Zhu et al., 2018) discussing new methods for bundle adjustment (the optimization method needed to simultaneously retrieve the image pose parameters from overlapping images considering corresponding image points). Notably, the geosciences have only started producing publications incorporating SfM photogrammetry in the past decade, but with improvements in the technique moving at an incredible speed: note that a similar search in 2015 by Carrivick et al. (2016) ranked Geosciences in the ninth position. In this field, the first work was published (according to WoS) by Heimsath and Farid (2002). Here, results from three unconstrained photographs characterized hillslope topography, and yield to an estimated surface with errors of the order of 1 m. In comparison, one of the last papers published in the field at the time of the search (Smith and Warburton, 2018) illustrates that topographic data from SfM photogrammetry (with errors on the scale

FIG. 1

“Structure from motion” search in academic databases: first 25 results and number of records per discipline (as of August 2018).

4

1. Structure from motion photogrammetric technique

of 3 m) and remains the only solution for synoptic bathymetric surveys of deep, wide rivers with very low water clarity (Parsons et al., 2005; Nittrouer et al., 2008; Leyland et al., 2017). Mobilization costs are high even though decreasing sensor size and the availability of unmanned surface vehicles may significantly reduce costs in the near future. TABLE 1 Methods used for acquisition of synoptic, high-resolution topography in fluvial environments over reaches longer than 1 km. Topo

Vegetation

Shallow rivers (3 m)

Survey extent (km)

Survey Costa

Topographic airborne Lidar

++

++

2

2

>10

€€

Bathymetric SFM

++

+

+b,c

2

10 km

€€€

a

Total survey cost, meaning that the ranking of cost per km can be different depending on the total survey extent. Currently being developed (Woodget et al., 2015; Dietrich, 2017). c Limited by bottom visibility, vegetation shadow, and surface waves. d Cost depends on imagery type (multispectral, hyperspectral) and deployment platform (drone, airplane, satellite) and includes ground calibration. e Only in clear water with clear river bed. Measure type:, impossible; +, possible; ++, suitable. b

28

2. Topo-bathymetric airborne LiDAR

Although precision and spatial resolution are of centimetric range on the channel bed, MBS cannot generally measure submerged river banks unless coupled with a side-mounted, multibeam sounder. Surrounding topography and vegetation cannot be measured unless coupled with a mobile terrestrial lidar (e.g., Leyland et al., 2017). 3. Bathymetric Structure From Motion: following early attempts at stereo photogrammetry conducted through water (Westaway et al., 2000), this recent technique exploits the capacity of SFM algorithms dedicated to 3D topographic reconstruction using UAS imagery to also reconstruct the underwater 3D geometry of shallow, clear-water rivers (Woodget et al., 2015). Because SFM algorithms do not factor in the light refraction at the air/water interface, the apparent depth is shallower than the true depth. Correction methods have begun to be developed (Woodget et al., 2015; Dietrich, 2017). They show promising results with precision and accuracy on par with SFM data on dry land (0.01%–0.02% of the flying height) and the maximum depth of 1–2 m. Although this new method potentially suffers from the same limitations that spectral methods do when it comes to water clarity, surface waves, and depth limitation, it has the advantage of directly generating a continuous topo-bathymetric survey. As with UAS–SFM surveys, dense riparian vegetation presents a challenge to accurately measure ground and bathymetry (Table 1). 4. Coastal bathymetric airborne lidar has been employed by researchers for decades (e.g., Guenther, 1985; Guenther et al., 2000). These sensors use a powerful green laser (λ ¼ 532 nm) that is among the least absorbed wavelength in water and can measure depths up to 40–50 m in very clear waters and clear bottoms. Their use in the context of river surveying remains, however, limited (Hilldale et al., 2007; Bailly et al., 2010) for three reasons: (i) the mobilization cost is very high; it is greater than 1000–2000 €/km2; (ii) the laser footprint is 1–2 m in diameter, which coupled with a low point density (e.g., 0.1–1 pts/m2), results in a measurement resolution that is too coarse for small to medium-sized streams. Similarly, the topographic and vegetation records are too coarse compared to the requirement of modern topographic airborne lidar data; (iii) owing to a high power pulse of longer duration, the detection of shallow depths (100 kHz) resulting in point density greater than 5 pts/m2. All topo-bathymetric sensors need to be deployed at survey elevations above ground level (AGL) lower than 600 m for the backscattered laser beam to be focused enough, with typical AGL of 300–400 m to maximize depth penetration. At this AGL the swath is 180–250 m wide depending on the sensor.

FIG. 1 Principle of topo-bathymetric airborne lidar with the scanning geometry of the Teledyne Optech Titan DW and main capabilities. An ideal bathymetric waveform is shown on the right. Discrete echoes would correspond to the two peaks appearing in the signal corresponding to a surface echo and the bottom echo. Right: illustration of the two-point clouds (532 and 1064 nm) produced by the instrument at the confluence between the Ain and the Rh^ one (France), and the continuous topo-bathymetric, point-cloud data after automatic land-water classification, refraction correction for bathymetric points, and ground detection for the topography. The Ain bathymetry is fully covered, but the Rh^ one is only measured down to 4 m depth resulting in partial bathymetric cover.

30

2. Topo-bathymetric airborne LiDAR

The low AGL is the main difference with topographic lidar, which can be deployed at a much higher elevation. Combined with more complex post-processing of the data (see Section 4), this results in ALTB surveys having higher costs/km2 than topographic lidar but are considerably cheaper than ALB. Sensors will differ by their scanning patterns, emitted pulse energy, and backscattered signal processing, which will impact the maximum detectable depth (see Section 3.2). They will all offer the possibility of recording the entire backscattered signal for each shot, called full-waveform (FWF), for reanalysis after the flight, on top of discrete echoes that are detected onboard (Figs. 1 and 2). First-generation sensors only used a green laser (McKean et al., 2009; Fernandez-Diaz et al., 2014; Mandlburger et al., 2015; Pan et al., 2015), combined occasionally with a second survey using a topographic lidar (e.g., Pan et al., 2015). A new generation sensor now uses an additional laser beam at λ ¼ 1064 nm within the same sensor, which is essential to accurately measure water-surface elevation, automatically classify wetted areas, and to increase point-density on dry surfaces as the green laser operates as a normal topographic lidar as well (Fig. 1). These new sensors will also integrate an aerial camera to generate high-resolution orthophotos.

2.3 Survey examples and typical data characteristics This work presents new data acquired with a Titan DW instrument (dual wavelength λ ¼ 1064 nm and λ ¼ 532 nm) developed by Teledyne-Optech in 2014 (Fig. 1). It is operated since 2015 by the University of Nantes and Rennes in France with a focus on river topobathymetry and monitoring of coastal environments. It also exists in a version with a third wavelength (λ ¼ 1550 nm) operated by the National Center for Airborne Laser Mapping in the United States since 2014 (Fernandez-Diaz et al., 2016), with several datasets available for download on OpenTopography. Except for the specific green wavelength, classification of wet areas, and the correction of refraction of bathymetric echoes (see Section 4), the sensor operation and data processing are similar to traditional topographic airborne lidar. The reader interested in better understanding airborne lidar operation, data processing, and error budget is referred to dedicated books (e.g., Vosselman and Maas, 2010). We present results from two surveys of the Ain River (e.g., Lassettre et al., 2008) located in the eastern part of France (Figs. 1 and 2) that were flown for Electricite de France. The first survey in July 2015 consisted of 42 km of river length between the Poncin dam and the confluence with the Rhone. This survey required 7 h of acquisition (1.5 days of operations), resulting in 12 billion points (532 nm + 1064 nm) to process. The second survey in September 2016 added 18 km of river length survey, including 6 km that were already surveyed near the confluence with the Rh^ one. Our results and recommendations also include expertise gathered from surveying other rivers in France with various characteristics: low-energy streams with dense riparian and aquatic vegetation (Connie, Selune), intermediate-size, gravel-bed rivers with mobile bed (Vieux Rhin, Moselle), and a large sand-bedded braided section of the Loire river. When operated for topo-bathymetric surveys, the Titan DW is flown at an AGL ranging from 330 to 400 m, at a flight speed of 200 km/h, and a shot frequency of 200 kHz on each channel. This results in typical point densities of 36 pts/m2 on land (532 nm + 1064 nm) and 18pts/m2 for the bathymetry (532 nm) on a single pass. Depending on the project

FIG. 2 A 13 km extract of the 53-km long Ain topo-bathymetric survey. Data courtesy of Electricite de France. The mean point density, including line overlap, is 29 pts/m2 below water and 59 pts/m2 above.

32

2. Topo-bathymetric airborne LiDAR

specifications, an overlap of 20%–50% is imposed between flight lines, which can result in up to twice the point density. A digitizer is associated with the Titan DW to record the fullwaveform (FWF) data of the 532 nm channel at shot frequencies ranging from 100 to 200 kHz depending on the mission requirements. As fluvial corridors are generally winding and the flight-line swath is of the order of 150–200 m, complex flight plans are required to minimize acquisition time and total cost. The green laser is pitched by 7° forward to avoid strong surface reflection on water that would blind the sensor if the incidence angle is too small. The 1064 nm laser has no forward pitch to better detect the water surface. Accurate calibration of the two laser-beam geometries is essential to get geometrically consistent 532 and 1064 nm point clouds (Fig. 1). For the Titan, it proved to be stable over the 3 years of operation, with less than 1-cm mean difference of vertical and horizontal positioning measured on hard horizontal and vertical surfaces. The trajectory and attitude of the instrument are computed using a high-quality integrated inertial measurement unit and GNSS system. The trajectory is post-processed using a combination of the permanent French GNSS network (IGN RGP) and precise point positioning resulting in typical uncertainty on trajectory position ranging from 2 to 5 cm in (X, Y, Z) without any temporary GNSS base station on the ground. Given the low flight elevation in topo-bathymetry, the vertical accuracy before any ground control adjustment is better than 20 cm and is generally better than 5 cm when adjusted with ground control. The precision on flat hard surfaces, measured over a 0.5 m radius disk as the standard deviation of point distance to a mean plane, is less than 3 cm for both channels. The accuracy and precision of the bathymetric point are presented in the next section.

3 Controls on depth penetration and surveyable rivers A critical aspect of topo-bathymetric lidar is the maximum measurable depth Dmax, which remains, to date, difficult to predict a priori for a given river and sensor. In the following, we discuss the parameters controlling Dmax, we introduce a new way to evaluate some of the parameters controlling Dmax, and we synthesize previous results and our own experience to evaluate the suitability of ALTB for various types of rivers, including a recommendation on the best survey season.

3.1 Theoretical controls on the bathymetric waveform and bottom echo intensity Several factors will control the amount of light energy reflected from the bottom of a river, which ultimately controls the maximum measurable depth (e.g., Guenther, 1985; Abdallah et al., 2012). These factors can be understood by studying a bathymetric waveform (Fig. 1) that corresponds to the backscattered signal received by the instrument from which it will detect the main discrete echoes, and which can also be independently recorded, typically at a frequency of 1 MHz for further reanalysis. The actual waveform will vary with the instrument due to different impulse duration and shape, but the overall geometry will be similar. In deep water (>2 m), a bathymetric waveform has three components (Fig. 3):

3 Controls on depth penetration and surveyable rivers

33

FIG. 3 Raw waveform records from the Titan DW 532-nm channel corresponding to the cross section shown in Fig. 4 [uncorrected from light celerity difference in water (see Section 4)]. The intensity is given in a digital number (an electronic measure without unit). (A) Deepwater part in the center of the channel. (B) Shallow area near the channel banks.

1. A surface echo whose peak can be shifted in depth by several tens of centimeters from the true surface of the water and whose intensity varies due to complex optical interactions with the roughness of the water surface (e.g., waves, turbulence) and water clarity (Guenther, 1985; Mandlburger et al., 2013; Pan et al., 2015; Zhao et al., 2018). Water-surface roughness will tend to increase the reflection component of the incident laser, with a directionality dependent on roughness anisotropy (Legleiter and Fonstad, 2012). Although theoretical models exist for oceanic environments to account for the role of waves (e.g., Abdallah et al., 2012), the prediction of transmission loss and of the vertical shift of peak energy for the specific conditions of rivers remains largely unknown. Quite importantly, any vegetation cover on the surface of the water (e.g., duckweeds) or above the water (riparian vegetation) will limit the energy transmitted in the water column

34

2. Topo-bathymetric airborne LiDAR

and the bathymetric capacities. Similarly, areas of white water will backscatter most of the emitted energy due to air bubbles, with no energy left to measure the bathymetry. 2. An exponential attenuation of the signal in the water column due to a combination of absorption by water and dissolved substance, and scattering by suspended organo-detritic particles and air bubbles, including backscattering toward the sensor (Cossio et al., 2009; Abdallah et al., 2012; Legleiter and Fonstad, 2012). The scattering component is expected to be dominant in rivers for two reasons. First, at the considered wavelength, the scattering coefficient of the suspended particle should be several times larger than the absorption coefficient (Legleiter and Fonstad, 2012; Legleiter et al., 2016). Second, the narrow beam and small receiver-area configuration of topo-bathymetric sensors are very sensitive to scattering that spreads out the laser spot in the water column, thus reducing the energy received by the narrow field of view of the sensor. This is a major difference between coastal bathymetric sensors that are more sensitive to absorption (Guenther et al., 2000). The exponential decay of the energy can be characterized by the light attenuation coefficient Kd,532 nm (m1) specific to the wavelength 532 nm. For narrow aperture sensors, Kd,532 nm should be equal to the beam attenuation coefficient c(532) which is the sum of the absorption and scattering coefficients (Cossio et al., 2009). This, however, has not yet been formally demonstrated with field data. Quantitative relationships between Kd,532 nm and turbidity are not known in rivers, but Kd will increase with the concentration in suspended inorganic and organic particles. An empirical proxy for water clarity can be obtained from Secchi disk depth (SDD) measurements, in which a 30 cm black and white disk is lowered down in the water. The depth at which the disk is no longer visible is the SDD. SDD is generally used by manufacturers to specify the depth penetration of bathymetric sensors for a given bottom reflectivity. SDD measurements have been calibrated to the sunlight attenuation (i.e., downwelling irradiance) Kd in coastal environments (e.g., Devlin et al., 2008; Lee et al., 2017, 2018) and to suspended sediment concentration, however, only very limited work has been done to evaluate these relationships in rivers (Davies-Colley and Nagels, 2008), and how Kd,532 nm varies with Kd or SDD. 3. A bottom echo whose backscattered energy will be proportional to the bed reflectivity at a 532 nm wavelength. Legleiter and coauthors (2009) have measured reflectivity as low as 0.05 for wet gravel and periphyton, defined as the complex mixture of algae, biofilms, and detritus attached to the river bed. Beyond this work, the wetted bed reflectivity remains largely underconstrained in rivers. The geometry of the illuminated target is also important: a flat sandy surface behaving as a diffuse reflector is optimum, whereas aquatic vegetation acts as a complex porous reflector that may stretch the laser pulse over the aquatic canopy height and reduce the likelihood of detecting a peak. Hence, aquatic vegetation is in general very detrimental to the river bathymetric survey. If too much energy has been lost traveling in water due to turbidity or depth, the bottom echo can be smaller than the noise level of the instrument, and the bed cannot be detected. Fig. 4 shows the 3D waveform intensity field corresponding to each sample of the FWF record of the 532 nm laser (effectively an x, y, z point with an intensity) and computed with a new plugin we have developed in CloudCompare (EDF R&D, 2011). This represents the raw backscattered signal on which the sensor detects maxima in real time to produce the so-called discrete echoes (Fig. 1B), or after which a reanalysis of the FWF can be performed to detect weak echoes corresponding to deeper points. Fig. 4 shows several key characteristics of topobathymetric lidar:

3 Controls on depth penetration and surveyable rivers

35

FIG. 4

Cross section on the River Ain illustrating: (A) the backscattered energy as recorded by the full waveform records for each shot; (B) the discrete echoes recorded during the flight; (C) a detail of the topo-bathymetric transition showing the systematic offset of green surface echoes with respect to the true water-surface documented by the 1064 nm channel; and (D) the reanalysis of the full waveform after the survey to recover weaker echoes, albeit with a larger uncertainty and stronger noise for the deepest ones, related to the simplistic approach used to detect echo (simple local maxima detection).

• Bathymetric echoes have very little backscattered energy compared to dry surfaces. Fig. 4A shows the strong dissipation of energy in the water column in the middle of the river such that the deepest points have backscattered energy more than 10–20 times smaller than a dry ground echo. This makes the detection of these echoes challenging. Fig. 4B shows that discrete bathymetric echoes are detected down to 2.05 m depth in this part of the River Ain, but a very simple reanalysis of the FWF data detect weak echoes down to 3 m, albeit with additional noise and false detection within the water column (Fig. 4D).

36

2. Topo-bathymetric airborne LiDAR

• Topo-bathymetric point clouds are perfectly continuous between the dry and wetted area (Figs. 1 and 4C), an essential characteristic related to their short pulse duration. Coastal bathymetric sensors are not able to achieve this key feature of topo-bathymetric sensors (Allouis et al., 2010). This originates from the bottom echo intensity being always larger than the surface echo at very shallow depth in clear rivers ( Dmax). Previous surveys dealing with small reaches (1 km) and is in general not desirable as the operator might miss small inundated areas below the canopy. Although sensor manufacturers generally provide a solution to automatically classify surface echoes and bottom echoes in coastal or lake environments (and perform subsequent refraction correction), we are not aware of tests assessing the performance of these solutions over complex fluvial environments. To illustrate the level of automatic processing that can be achieved on large-scale surveys, we briefly show results from an algorithm of automatic bathymetric classification that we have developed working on raw 1064 and 532 nm point clouds. This approach does not use the waveform signal but rather uses the difference in water depth penetration of the 1064 and 532 nm channels and the flat nature of the water surface measured by the 1064 nm. It leverages the command line function of the open-source software CloudCompare (EDF R&D, 2011) for scripting efficient 532 nm/1064 nm point cloud comparison and 3D morphological operations (roughness, slope, point density, lowest point extraction). It can process several billion points to create a low-resolution (2-m point spacing) classification of the watersurface echoes on the 1064 nm channel and of bathymetric points in the 532 nm channel in less than 1 h on a 48 cpu server. The low-resolution water surface composed of only a few thousand points can then be checked manually for quality control and is either densified with neighboring 1064 nm echoes, interpolated if the point density is too low, or directly used in subsequent calculations. The resulting uncorrected depth map can be converted to an approximate depth map by multiplying by a factor of 0.73. This serves as a quick estimate of the bathymetric coverage used to plan additional sonar surveys in deep areas but cannot be considered a final product as the horizontal correction of refraction must be accounted for. The final water-surface 1064-nm point cloud is the key ingredient to a subsequent accurate refraction correction and high-resolution classification of the 532 and 1064-nm point clouds. In particular, 532 nm points locally above the nearest water-surface point of the 1064 nm are classified as land and subsequently classified with traditional topographic airborne lidar techniques (Vosselman and Maas, 2010). The position of 532 nm points below the nearest

4 Data processing

43

FIG. 7 Example of automatic detection of the water surface using the 1064 nm wavelength. The main channel surface is detected but also abandoned channels below a canopy, lakes, and swimming pools (SP) in the villages bordering the river. The 1064-nm water-surface point cloud is the key element for an accurate land/water classification of the 532-nm point cloud, as well as the refraction correction of bathymetric points.

1064-nm water-surface point is corrected for refraction and laser celerity change and then classified as bottom echoes, water surface, or water column echoes using a combination of depth/intensity analysis (Fig. 5) and morphological analysis. Fig. 7 shows the result of a typical water-surface classification on the lowest 13 km of the River Ain survey. Beyond accurately detecting the main river and large open water bodies such as lakes, the algorithm is able to capture abandoned channels largely screened by the tree canopy, small ponds, and swimming pools. This offers new perspectives to study the hydrological connectivity of floodplains and rivers, as well as hydrology under forest canopies.

4.2 FWF analysis Compared to point-cloud processing, FWF processing is more challenging as few software packages exist to visualize and process it, and the volume of data is typically 10 times larger than compressed point clouds. The open-source software CloudCompare (EDF R&D, 2011) can now handle FWF visualization and processing thanks to developments funded by our team (Fig. 4), but many challenges exist to overcome the use of FWF routinely. While FWF processing is not mandatory if discrete echoes are deep enough, it can potentially help in recovering weaker echoes (Fig. 3), increase accuracy, inform on water column turbidity (Pan et al., 2015; Richter et al., 2017) and bottom characteristics. Similar to topographic airborne lidar FWF processing (e.g., Mallet and Bretar, 2009), many different peak extraction techniques have been developed with no consensus yet on the best one for bathymetric waveform processing (Pan et al., 2015; Saylam et al., 2017). Preliminary attempts at retrieving Kd,532nm from single waveform shots to map spatial variations of river-water turbidity have had mixed results due to the overlap of surface echoes, water column backscatter,

44

2. Topo-bathymetric airborne LiDAR

and bottom echoes in shallow environments (Richter et al., 2017). As for point clouds, waveforms should also be corrected for refraction. FWF can be studied for a single shot, or aggregated spatially into an equivalent vertical waveform called an orthowaveform (Wu et al., 2012; Pan et al., 2016). Orthowaveforms have the benefit of transforming the waveform 3D field into a grid of waveform signal with regular horizontal spacing that facilitates the comparison with hyperspectral or multispectral imaging (Pan et al., 2016).

5 Applications in fluvial geomorphology The fluvial geomorphology community has cautiously observed the advent of ALTB and only a handful of ALTB surveys occur that cover more than a few kilometers of rivers (McKean et al., 2009). As such, ALTB has not yet moved from the realm of a promising remote sensing technique to a day-to-day survey technique for geomorphological analysis, both in the academic and private sectors. In the following section, we summarize the existing applications of ALTB and illustrate scientific opportunities. First, we summarize the main data available at hand for the fluvial geomorphologist following a successful survey using the latest generation of ALTB sensors: • A continuous topo-bathymetric survey with a point density of the order of 10–20 pts/m2, accurate to within 10 cm underwater and 5 cm above the water. Laser-spot size sets the capacity to resolve details, which for the Titan is of the order of 20–30 cm. • A water surface is known at 2 cm with the 1064-nm channel, albeit with incomplete spatial sampling. • A 3D characterization of riparian vegetation potentially with bi-spectral intensity information. • Relationships between depth and intensity that are quantitatively related to the water column optical properties and variations in bottom reflectance (Eq. 1). • An optional FWF record for each shot with elements of the records that depend on the physical characteristics of the environment (turbidity, bottom reflectance, water-surface roughness). • Optional high-resolution multispectral or hyperspectral orthoimagery (Legleiter et al., 2016). • All this acquired over several tens of kilometers. These data can be augmented with in situ measurements. Water discharge should be measured on the day of the survey in as many points as deemed necessary given the number of confluences. The optical properties of the water should be measured in various locations, with methods ranging from simple SDD measurements to more accurate inherent or apparent optical properties of the water in the 532 nm wavelength, as well as bottom reflectance (Legleiter et al., 2016). Water could also be sampled in various spots for suspended sediment concentration measurement.

5.1 Multi-scale high-resolution fluvial geomorphology The first straightforward application of topo-bathymetric lidar is the 3D analysis of channel geometry at various scales. McKean et al. (2009, 2008) have pioneered this approach and

5 Applications in fluvial geomorphology

45

have illustrated a large range of applications using EAARL data, ranging from characterizing morphological units (e.g., pools and riffle) to cross-section characteristics, longitudinal profile (e.g., slope and width variation) and the relationship between floodplains and channels (e.g., connectivity). The advent of sensors creating denser point clouds (>20 pts/m2) with very low positioning noise also offers the possibility to explore below the 1 m2 scale of EAARL data, in order to address spatial variation in bed roughness (Fig. 8). Quantifying this variation is useful when exploring the heterogeneity of channel-bed conditions with relevance to friction modeling and in-stream, meso-habitat prediction. The ability to resolve details of the channel bed must be carefully addressed by accounting for the point-cloud pattern, the bathymetric precision, the laser-spot size, and the accuracy of flight-line adjustment. As shown on sample data of the Ain (Fig. 8), sensors such as the Titan are able to resolve individual boulders and tree trunks down to approximately 20–30 cm. The ability to resolve details decreases with depth and river turbidity as diffusion increases the laser-spot size on the bed. In the case of the River Ain, bed roughness measured over 1 m is as low as 6 mm on smooth bathymetric surfaces under 2 m of water but can increase up to 10 cm in cobble-bed areas. Modern topo-bathymetric surveys can easily address morphological characteristics over about 5–6 orders of magnitude (0.2–100,000 m) both below and above water. This is unprecedented. Arguably, the full potential of ALTB for geomorphological analysis has not yet been realized. As for topographic lidar (Passalacqua et al., 2015), synoptic, high-resolution, high precision 3D topo-bathymetry requires a change in paradigm to describe rivers beyond the traditional planform or cross-sectional view inherited from coarser survey approaches. The 3D continuous nature of the river and its relationship with the floodplain and riparian vegetation must now be included.

5.2 Coupling with 2D–3D hydraulic modeling Given that ALTB surveys are performed during low flows, the water-surface slope and the flow depth extracted from these data are hardly representative of the large range of discharges experienced by the river. Applying computational models of fluid dynamics to fluvial topo-bathymetric data is thus a quasi-mandatory approach to fully harness the potential of ALTB (McKean et al., 2009, 2014; Mandlburger et al., 2015). Such models can be used to predict flow velocities, water depth, and shear stresses occurring during discharges large enough to be relevant to processes such as bedload transport, channel-bank erosion, boulder mobility, and floodplain/channel connectivity. 2D hydraulic models have been used for fluvial habitat mapping over small reaches (Crowder and Diplas, 2000; Mandlburger et al., 2015), but the challenge now lies in applying such models over very long reaches and high resolution. It is beyond the scope of this chapter to offer a complete review of CFD modeling capabilities and limits. Fig. 9 illustrates potential applications of a 2D hydraulic model resolving the vertically averaged St Venant equations without inertia applied on a 1-m topobathymetric DEM of the Ain over a 4-km-long reach (Davy et al., 2017). This model is fast and can thus be used to explore the impact of varying discharge or friction coefficient. For instance, an average friction coefficient was calibrated by minimizing the difference between the predicted water-surface elevation and the measured water-surface elevation given the knowledge of the discharge at the time of the survey. Exploring the spatial variations of the residuals can then offer insights into the quality of the model prediction.

FIG. 8 Illustration of resolution and accuracy capabilities of the latest generation topo-bathymetric sensors (Optech Titan DW). Excerpt of the 2016 Ain raw survey point cloud illustrating the capability to detect tree trunks and large boulders in the river bed as well as variable bed roughness (standard deviation of elevation measured over a 1 m sphere diameter, computed directly on the 3D point cloud with Cloudcompare). Note that the minimum roughness is 6 mm in the smoothest parts of the bed, illustrating excellent ranging noise capabilities under 2 m water depth. Topographic change between September 2016 and July 2015 measured with M3C2 algorithm at 1-m scale directly on point clouds (Lague et al., 2013), and showing the deposition of a tree trunk and several individual boulders of metric size, as well as erosion downstream of the tree trunk.

FIG. 9

Application examples of 2D high-resolution hydraulic modeling applied on large-scale (4-km) topo-bathymetric survey. A 1-m continuous topobathymetric DEM of a 4-km long reach of the Ain is used as a boundary condition to the hydrodynamic component of the EROS model that solves the vertically averaged St-Venant equations without the inertial term (Davy et al., 2017). Flow direction is from top to bottom.

48

2. Topo-bathymetric airborne LiDAR

Beyond model calibration, the inundation pattern, flow depth, velocities, and shear stress can be predicted in the river and the floodplain for a large range of discharges (Fig. 9). From this set of high-resolution inundation maps obtained for various discharges, several applications can be envisioned. Here, we illustrate a few possibilities. First, at any given channel location, the hydraulic geometry of the channel can be computed in the form of rating curves of a hydraulic variable (e.g., depth, velocity, hydraulic slope) vs discharge. Second, inundation patterns and flow velocities can be predicted for very large discharges in the context of flood-hazard prediction. Third, at any point in the floodplain, the classified vegetation cover obtained from the aboveground ALTB data can be compared to hydrological forcings such as inundation frequency or flow velocity during large flood events. Fourth, for any relevant discharge, such as the bank-full discharge, the morphological units can be mapped as a function of a combination of geometrical characteristics and flow velocities (Mandlburger et al., 2015). These are only a handful of possibilities that CFD modeling coupled with ALTB is likely to offer in the coming years for fluvial geomorphology applications.

5.3 Synoptic channel morphodynamics and sediment budget As for many high-resolution topographic applications (Passalacqua et al., 2015), repeat ALTB surveys offer the potential of exploring channel morphodynamics in a completely new way (McKean et al., 2009; Mandlburger et al., 2015). The lack of bathymetric information using airborne or ground-based lidar or photogrammetry has always resulted in large uncertainties in measuring erosion and sedimentation patterns in rivers, as well as constructing sediment budgets. Getting repeat surveys of inundated areas where most of bedload sediment transport occurs is essential for improving our understanding of how rivers evolve, and in particular the impact of extreme events over a broad range of scales. A critical aspect of repeat surveying is the level of change detection that is achievable with ALTB. Various approaches are possible to develop an error budget for airborne-lidar, pointcloud comparison (e.g., Passalacqua et al., 2015). Three sources of uncertainty should be considered: (i) the co-registration error of the two surveys that must be evaluated with fixed reference surfaces and can be assumed either uniform (Wagner et al., 2017) or spatially variable (Joerg et al., 2012); (ii) the position uncertainty that may be different on topography and bathymetry; (iii) the surface roughness which introduces uncertainty in point-cloud comparison due to differences in sampling between the surveys (Lague et al., 2013). In the simplest approach possible, one can consider a spatially uniform registration error σ reg , and a spatially variable error made of a combination of a position uncertainty Zerr and point-cloud roughness σz evaluated over the spatial scale at which vertical averaging occurs. Point-cloud roughness results both from surface roughness but also errors in flight-line adjustment in overlapping areas, ranging noise, or classification errors (Wagner et al., 2017). To evaluate the standard error of the vertical topographic difference with typical Titan topo-bathymetric surveys, we assume that the roughness component is uncorrelated between the two surveys and that the survey have the same densities, position uncertainty and point-cloud roughness. In that case, the standard error is given by.

5 Applications in fluvial geomorphology

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Z2 + σ2z,1 Z2err, 2 + σ2z,2 SE ¼ σ reg 2 + err,1 + n1 n2

49 (3)

where n1 is the number of points contained in the surface used to compute the average position and the roughness σ z,1 of the first (resp. second) survey, and n2 and σ z,2 are the equivalent values for the second survey. Here, we assume that Zerr, σ z, and n are identical during the two surveys, but these values can easily be calculated from point clouds to have a spatially variable standard error (e.g., Wagner et al., 2017). In the Ain near the confluence with the Rhone (Fig. 10), σ reg ¼ 2:58 cm (measured as the standard deviation of the M3C2 difference at 1 m between 2015 and 2016 of 10,000 points on roads). As shown in Fig. 8, point-cloud roughness measured at 1 m scale varies from 0.6 cm to more than 30-cm near large boulders, while the point density is typically 20 pts/m2. If we assume conservatively that Zerr ¼ 10 cm under water and Zerr ¼ 5 cm above the water, the combined standard error underwater varies from (Table 2) 4.38 cm in low roughness areas, to 8.31 cm in high roughness areas. Because of the high point density, the standard error is not very different under and above water for a single laser. But the doubling of point density by combining the 532 and 1064 nm above water further reduces the influence of position error and point-cloud roughness, such that the registration error is the dominant source of uncertainty in low-roughness zones. FIG. 10 Topographic change of the Ain River between September 2016 and July 2015, measured directly on topo-bathymetric point clouds after vegetation classification as the mean vertical difference averaged over a 0.5-m radius circle (M3C2 algorithm, Lague et al., 2013). The dominant signal is the lateral mobility of the Ain River, driving accretion inside bends and bank erosion outside. The vegetation cover of the 2016 survey is shown for reference. No data are available on the Rhone, as the turbidity and water levels during the 2015 survey resulted in the absence of bathymetric measurement.

50

2. Topo-bathymetric airborne LiDAR

TABLE 2 Standard errors corresponding to a vertical topographic change measured over a 0.5 m averaging radius for point densities typical of the Titan DW (20 pts/m2), and bed roughness typical of the Ain survey (see Fig. 8). Calculation using Eq. (3) and registration error of 2.58 cm Low roughness σ z 5 0.6 cm, n 5 16

High roughness σ z 5 20 cm, n 5 16

Low roughness, double ppm σ z 5 0.6 cm, n 5 32

Bathymetry, Zerr 510 cm

4.38 cm

8.31 cm

3.60 cma

Topography, Zerr 55 cm

3.13 cm

7.73 cm

2.87 cm

a

If one would double point density by imposing a 50% flight-line overlap.

Table 2 shows that the standard error is of the order of 4–8 cm with the Optech Titan. When considering the topographic change measured on the Ain near the Rhone confluence (Figs. 8 and 10), these predictions show that most of the change is largely above the standard error. In the downstream section (Fig. 10), the dominant change is related to channel lateral mobility in the sinuous parts. Finer details can be resolved, such as boulder movement (Fig. 8), deposition of tree trunks, and bed accretion/erosion on the order of a few decimeters in relation to the wake developed by roughness features. These results illustrate the ability of repeat ALTB surveys to precisely document channel morphodynamics. In this context, the high shot density of modern ALTB not only offers the capability of resolving finer details of the channel bed, but it contributes to dividing by a factor 4–5 the standard error associated with stochastic components of the error budget compared to the early ALTB systems.

6 Conclusions and remaining challenges Topo-bathymetric lidar can now be considered an operational technique to obtain synoptic, 3D high-resolution and high-accuracy surveys of rivers characterized by clear water and bed and mobile sediment with reduced aquatic vegetation cover. Depth down to 5 m can be reached in the best conditions with accuracies better than 10 cm. The continuous nature of the topography and bathymetry makes it suitable for a wide range of geomorphological applications, including erosion/sedimentation mapping with repeat surveys. As sensors have now reached a good level of maturity, four challenges need to be addressed to predict the surveyability of any given river and turn the dense lidar point clouds into usable scientific datasets for geomorphological science:

6.1 A priori prediction of depth penetration and river bathymetric cover Sensor deployment on a new river still suffers from some uncertainty on the maximum measurable depth Dmax. Although the theory predicting Dmax is well established (Eq. 2) (Guenther, 1985; Abdallah et al., 2012), many of the physical parameters related to river environments are not known. This makes the a priori prediction of the extinction depth Dmax difficult. No reason exists, however, for the range of Kd,532 nm, Rb, and Ls to be better constrained in the future for river environments through direct measurement of inherent optical properties and bottom reflectivity, discrete echoes, and FWF analysis as in Fig. 3, or the

6 Conclusions and remaining challenges

51

use of multispectral/hyperspectral imagery (Legleiter and Fonstad, 2012). This will help in narrowing down the type of rivers for which ALTB is not suitable due to high turbidity or low bed reflectance before acquiring any data.

6.2 Automatic classification on massive lidar datasets Fluvial environments can be considerably more challenging than coastal environments when it comes to the refraction correction of bathymetric echoes. Applications of ALTB over very long river corridors are still in its infancy and any kind of manual processing considerably increases the final cost of ALTB data. Automated post-processing is thus mandatory. We have developed in-house techniques to process large datasets for accurate refraction correction, but a more systematic classification of fluvial features relevant to geomorphology is needed. This includes the automated identification of banks, pools, bars, dunes, dams, boulders, bridges, riparian vegetation, terraces, or any feature of fluvial environments that can be used to enrich the 3D description of data for scientific analysis. Biogeomorphological analysis is expected to benefit significantly from advanced classification methods able to capture species diversity in relation to geomorphologic features.

6.3 FWF analysis in the context of fluvial environments The analysis of FWF records can benefit from previous work done on coastal environments, but should also consider the specificity of shallow fluvial environments and topobathymetric sensors. In particular, the shallow depth makes the separability of surface echoes, water-column backscatter, and bottom reflection more complex. Progress in using FWF records as an actual signal from which physical characteristics of the water surface, column, and river bed can be inverted will depend on two elements: first, the development of new signal-processing methods to improve depth accuracy, deconvolve the effect of systemimpulse response, and increase the detection of weak echoes to maximize the depth capability. Second, the acquisition of reference data sets in a variety of fluvial environments for which in situ measurements of optical properties, water turbidity, bottom reflectance, depth, and detailed analysis of water-surface characteristics are precisely measured.

6.4 Large-scale hydraulic modeling on topo-bathymetric data As ALTB datasets progressively grow in size due to larger surveys and better resolution, the need for large-scale 2D and even 3D hydraulic modeling operating at a submeter resolution increases. Hydraulic modeling should be viewed as an essential data augmentation approach in the context of fluvial ALTB. Although topo-bathymetric data alleviate the traditional problem of defining the numerical model domain, based on a limited set of channel cross sections, it remains, however, computationally too demanding to operate the current generation of 2D hydraulic models over very large scales (i.e., >50 km) at submeter resolution. ALTB data also offer new opportunities to validate model prediction based on waterelevation prediction and inject spatially explicit, friction variations using 3D information such as bed roughness and riparian vegetation to improve modeling accuracy.

52

2. Topo-bathymetric airborne LiDAR

Acknowledgments The Titan DW sensor, operated by the Nantes-Rennes Lidar Platform has been funded by the Region Pays de la Loire with funding of the RS2E-OSUNA programs and the Region Bretagne with support from the European Regional Development Fund. Patrick Launeau is greatly acknowledged for his contribution in the acquisition of the Titan DW sensor. We thank Cyril Michon, Emmanuel Gouraud, William Gentile from Geofit-Expert company, Paul Larocque and Anca Dorbinescu from Teledyne Optech, and Laurence Hubert-Moy for their contribution in the overall operation of the Titan DW sensor. We thank Electricite De France (A. Barillier, A. Clutier) for commissioning the acquisition of the Ain River survey and providing access to the data. Daniel Girardeau-Montaut is greatly acknowledged for his ongoing development of Cloudcompare which has been used both for processing and figure generation in this work.

References Abdallah, H., Baghdadi, N., Bailly, J., Pastol, Y., Fabre, F., 2012. Wa-LiD: a new LiDAR simulator for waters. IEEE Geosci. Remote Sensing Lett. 9 (4), 744–748. https://doi.org/10.1109/LGRS.2011.2180506. Allouis, T., Bailly, J.-S., Pastol, Y., Le Roux, C., 2010. Comparison of LiDAR waveform processing methods for very shallow water bathymetry using Raman, near-infrared and green signals. Earth Surf. Process. Landf. 35 (6), 640–650. https://doi.org/10.1002/esp.1959. Bailly, J.S., le Coarer, Y., Languille, P., Stigermark, C.J., Allouis, T., 2010. Geostatistical estimations of bathymetric LiDAR errors on rivers. Earth Surf. Process. Landf. 35 (10), 1199–1210. https://doi.org/10.1002/esp.1991. Cossio, T., Slatton, K.C., Carter, W., Shrestha, K., Harding, D., 2009. Predicting topographic and bathymetric measurement performance for low-SNR airborne lidar. IEEE Trans. Geoscience Remote Sens. 47 (7), 2298–2315. https://doi.org/10.1109/TGRS.2008.2011054. Crowder, D.W., Diplas, P., 2000. Using two-dimensional hydrodynamic models at scales of ecological importance. J. Hydrol. 230 (3–4), 172–191. https://doi.org/10.1016/S0022-1694(00)00177-3. Daniele, T., McKean, J., Benjankar, R.M., Wright, W., Goode, J.R., Chen, Q., et al., 2018. Mapping river bathymetries: evaluating topobathymetric LiDAR survey. Earth Surf. Process. Landf. https://doi.org/10.1002/esp.4513. Davies-Colley, R.J., Nagels, J.W., 2008. Predicting light penetration into river waters. J. Geophys. Res. Biogeo. 113 (3), 1–9. https://doi.org/10.1029/2008JG000722. Davy, P., Croissant, T., Lague, D., 2017. A precipiton method to calculate river hydrodynamics, with applications to flood prediction landscape evolution models, and braiding instabilities. J. Geophys. Res. 122, 1–22. https://doi. org/10.1002/2016JF004156. Devlin, M.J., Barry, J., Mills, D.K., Gowen, R.J., Foden, J., Sivyer, D., Tett, P., 2008. Relationships between suspended particulate material, light attenuation and Secchi depth in UK marine waters. Estuar. Coast. Shelf Sci. 79 (3), 429–439. https://doi.org/10.1016/j.ecss.2008.04.024. Dietrich, J.T., 2017. Bathymetric structure-from-motion: extracting shallow stream bathymetry from multi-view stereo photogrammetry. Earth Surf. Proc. Landforms 42, 355–364. https://doi.org/10.1002/esp.4060. EDF R&D, Telecom ParisTech, 2011. CloudCompare (version 2.9) [GPL software]. Retrieved from http://www. danielgm.net/cc/. Fernandez-Diaz, J.C., Glennie, C.L., Carter, W.E., Shrestha, R.L., Sartori, M.P., Singhania, A., et al., 2014. Early results of simultaneous terrain and shallow water bathymetry mapping using a single-wavelength airborne LiDAR sensor. IEEE J Sel. Top. Appl. Earth Obs. Remote Sens. 7 (2), 623–635. https://doi.org/10.1109/JSTARS.2013.2265255. Fernandez-Diaz, J.C., Carter, W.E., Glennie, C., Shrestha, R.L., Pan, Z., Ekhtari, N., et al., 2016. Capability assessment and performance metrics for the titan multispectral mapping lidar. Remote Sens. (Basel) 8 (11), 1–33. https://doi. org/10.3390/rs8110936. Glennie, C.L., Carter, W.E., Shrestha, R.L., Dietrich, W.E., 2013. Geodetic imaging with airborne LiDAR: the Earth’s surface revealed. Rep. Prog. Physics. Phys. Soc. (Great Britain) 76 (8), 086801. https://doi.org/10.1088/00344885/76/8/086801. Guenther, G.C., 1985. Airborne laser hydrography: system design and performance factors. NOAA Professional Paper Series 1, 385. Guenther, G.C., Cunningham, A.G., Larocque, P.E., Reid, D.J., 2000. Meeting the Accuracy Challenge in Airborne Lidar Bathymetry. National Oceanic Atmospheric Administration/NESDIS, Silver Spring, MD.

References

53

Hilldale, R.C., Raff, D., Engineer, H., 2007. Assessing the ability of airborne LiDAR to map river bathymetry. Earth Surf. Process. Landf. 33 (5), 773–783. https://doi.org/10.1002/esp.1575. otter, J., 2009. Water surface mapping from airborne laser scanH€ ofle, B., Vetter, M., Pfeifer, N., Mandlburger, G., St€ ning using signal intensity and elevation data. Earth Surf. Process. Landf. 34 (12), 1635–1649. https://doi.org/10. 1002/esp.1853. Joerg, P.C., Morsdorf, F., Zemp, M., 2012. Uncertainty assessment of multi-temporal airborne laser scanning data: a case study on an alpine glacier. Remote Sens. Environ. 127, 118–129. https://doi.org/10.1016/j.rse. 2012.08.012. Lague, D., Brodu, N., Leroux, J., 2013. Accurate 3D comparison of complex topography with terrestrial laser scanner: application to the Rangitikei canyon (N-Z). ISPRS J. Photogramm. Remote Sens. 82, 10–26. https://doi.org/10. 1016/j.isprsjprs.2013.04.009. Lassettre, N.S., Piegay, H., Dufour, S., Rollet, A.-J., 2008. Decadal changes in distribution and frequency of wood in a free meandering river, the Ain River, France. Earth Surf. Process. Landf. 33, 1098–1112. https://doi.org/10.1002/ esp.1605. Lee, Z., Shang, S., Lin, G., Liu, T., Liu, Y., Du, K., Luis, K., 2017. Secchi disk observation with spectral-selective glasses in blue and green waters. Opt. Express. 25(17)https://doi.org/10.1364/OE.25.019878. Lee, Z., Shang, S., Du, K., Wei, J., 2018. Resolving the long-standing puzzles about the observed Secchi depth relationships. Limnol. Oceanogr. 63 (6), 2321–2336. https://doi.org/10.1002/lno.10940. Legleiter, C.J., Fonstad, M.A., 2012. An introduction to the Physical Basis for Deriving River Information by Optical Remote Sensing. In: Fluvial Remote Sensing for Science and Management. Wiley-Blackwell, pp. 43–69. Legleiter, C.J., Roberts, D.A., Lawrence, R.L., 2009. Spectrally based remote sensing of river bathymetry. Earth Surf. Process. Landf. 34, 1039–1059. https://doi.org/10.1002/esp.1787. Legleiter, C.J., Overstreet, B.T., Glennie, C.L., Pan, Z., Fernandez-Diaz, J.C., Singhania, A., 2016. Evaluating the capabilities of the CASI hyperspectral imaging system and Aquarius bathymetric LiDAR for measuring channel morphology in two distinct river environments. Earth Surf. Process. Landf. 41 (3), 344–363. https://doi.org/10.1002/ esp.3794. Lejot, J., Delacourt, C., Piegay, H., Fournier, T., Tremelo, M., et al., 2007. Very high spatial resolution imagery for channel bathymetry and topography from an unmanned mapping controlled platform. Earth Surf. Process. Landf. 32 (11), 1705–1725. https://doi.org/10.1002/esp.1595. Leyland, J., Hackney, C.R., Darby, S.E., Parsons, D.R., Best, J.L., Nicholas, A.P., et al., 2017. Extreme flood-driven fluvial bank erosion and sediment loads: direct process measurements using integrated Mobile laser scanning (MLS) and hydro-acoustic techniques. Earth Surf. Process. Landf. 42 (2), 334–346. https://doi.org/10.1002/ esp.4078. Mallet, C., Bretar, F., 2009. Full-waveform topographic lidar: state-of-the-art. ISPRS J. Photogramm. Remote Sens. 64 (1), 1–16. https://doi.org/10.1016/j.isprsjprs.2008.09.007. Mandlburger, G., Pfennigbauer, M., Pfeifer, N., 2013. Analyzing near water surface penetration in laser bathymetry— a case study at the River Pielach. In: ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences, II-5/W2(November), pp. 175–180. https://doi.org/10.5194/isprsannals-II-5-W2-175-2013. Mandlburger, G., Hauer, C., Wieser, M., Pfeifer, N., 2015. Topo-bathymetric LiDAR for monitoring river morphodynamics and instream habitats—a case study at the Pielach River. Remote Sens. (Basel) 7 (5), 6160–6195. https://doi.org/10.3390/rs70506160. McKean, J., Isaak, D.J., Wright, C.W., 2008. Geomorphic controls on salmon nesting patterns described by a new, narrow-beam terrestrial–aquatic lidar. Front. Ecol. Environ. 6 (3), 125–130. https://doi.org/10.1890/070109. McKean, J., Nagel, D., Tonina, D., Bailey, P., Wright, C.W., Bohn, C., Nayegandhi, A., 2009. Remote sensing of channels and riparian zones with a narrow-beam aquatic-terrestrial LIDAR. Remote Sens. (Basel) 1 (4), 1065–1096. https://doi.org/10.3390/rs1041065. McKean, J., Tonina, D., Bohn, C., Wright, C.W., 2014. Effects of bathymetric lidar errors on flow properties predicted with a multi-dimensional hydraulic model. J. Geophys. Res. Earth 119 (3), 644–664. https://doi.org/10. 1002/2013JF002897. Nittrouer, J., Allison, M., Campanella, R., 2008. Bedform transport rates for the lowermost Mississippi River. J. Geophys. Res. 113 (F3), 1–16. https://doi.org/10.1029/2007JF000795. Pan, Z., Glennie, C., Hartzell, P., Fernandez-Diaz, J., Legleiter, C.J., Overstreet, B., 2015. Performance assessment of high resolution airborne full waveform LiDAR for Shallow River bathymetry. Remote Sens. (Basel) 7 (5), 5133–5159. https://doi.org/10.3390/rs70505133.

54

2. Topo-bathymetric airborne LiDAR

Pan, Z., Glennie, C., Fernandez-Diaz, J.C., Shrestha, R., Carter, B., Hauser, D., et al., 2016. Fusion of bathymetric LiDAR and hyperspectral imagery for shallow water bathymetry. In: International Geoscience and Remote Sensing Symposium (IGARSS), 2016–Novem, pp. 3792–3795. https://doi.org/10.1109/IGARSS.2016.7729983. Parsons, D.R., Best, J.L., Orfeo, O., Hardy, R.J., Kostaschuk, R., Lane, S.N., 2005. Morphology and flow fields of threedimensional dunes, Rio Parana´, Argentina: Results from simultaneous multibeam echo sounding and acoustic Doppler current profiling. J. Geophys. Res. 110, F04S03. https://doi.org/10.1029/2004JF000231. Passalacqua, P., Belmont, P., Staley, D.M., Simley, J.D., Arrowsmith, J.R., Bode, C.A., et al., 2015. Analyzing high resolution topography for advancing the understanding of mass and energy transfer through landscapes: A review. Earth-Sci. Rev. 148, 174–193. https://doi.org/10.1016/j.earscirev.2015.05.012. Richter, K., Maas, H.G., Westfeld, P., Weiß, R., 2017. An approach to determining turbidity and correcting for signal attenuation in airborne Lidar bathymetry. Photogrammetrie, Fernerkundung, Geoinformation 85 (1), 31–40. https://doi.org/10.1007/s41064-016-0001-0. Saylam, K., Brown, R.A., Hupp, J.R., 2017. Assessment of depth and turbidity with airborne Lidar bathymetry and multiband satellite imagery in shallow water bodies of the Alaskan north slope. Int. J. Appl. Earth Obs. Geoinf. 58, 191–200. https://doi.org/10.1016/j.jag.2017.02.012. Tarolli, P., 2014. High-resolution topography for understanding earth surface processes: opportunities and challenges. Geomorphology 216, 295–312. https://doi.org/10.1016/j.geomorph.2014.03.008. Vosselman, F., Maas, H.G., 2010. Airborne and Terrestrial Laser Scanning. CRC. Wagner, W., Lague, D., Mohrig, D., Passalacqua, P., Shaw, J., Moffett, K., 2017. Elevation change and stability on a prograding delta. Geophys. Res. Lett. 44 (4), 1786–1794. https://doi.org/10.1002/2016GL072070. Westaway, R.M., Lane, S.N., Hicks, D.M., 2000. The development of an automated correction procedure for digital photogrammetry for the study of wide, shallow, gravel-bed rivers. Earth Surf. Process. Landf. 25 (2), 209–226. Williams, R.D., Brasington, J., Vericat, D., Hicks, D.M., 2014. Hyperscale terrain modelling of braided rivers: fusing mobile terrestrial laser scanning and optical bathymetric mapping. Earth Surf. Process. Landf. 39 (2), 167–183. https://doi.org/10.1002/esp.3437. Woodget, A.S., Carbonneau, P.E., Visser, F., Maddock, I.P., 2015. Quantifying submerged fluvial topography using hyperspatial resolution UAS imagery and structure from motion photogrammetry. Earth Surf. Process. Landf. 40 (1), 47–64. https://doi.org/10.1002/esp.3613. Wu, J., Van Aardt, J.A.N., McGlinchy, J., Asner, G.P., 2012. A robust signal preprocessing chain for small-footprint waveform LiDAR. IEEE Transactions on Geoscience and Remote Sensing 50 (8), 3242–3255. https://doi.org/10. 1109/TGRS.2011.2178420. Zhao, X., Zhao, J., Zhang, H., Fengnian, Z., 2018. Remote sensing of sub-surface suspended sediment concentration by using the range Bias of green surface point of airborne LiDAR bathymetry. Remote Sens. (Basel) 10, 681. https://doi.org/10.3390/rs10050681.

C H A P T E R

3 Ground-based remote sensing of the shallow subsurface: Geophysical methods for environmental applications Giorgio Cassiania, Jacopo Boagaa, Ilaria Baronea, Maria Teresa Perria, Gian Piero Deiddab, Giulio Vignolib, Claudio Strobbiac, Laura Busatod, Rita Deianae, Matteo Rossif, Maria Clementina Caputog, Lorenzo De Carlog a

Dipartimento di Geoscienze, Universita` di Padova, Padova, Italy bDipartimento di Ingegneria Civile, Ambientale e Architettura, Universita` di Cagliari, Cagliari, Italy cRealtimeseismic SA, Pau, France dDepartment of Agricultural Sciences, University of Naples Federico II, Naples, Italy e Dipartimento di Beni Culturali (dBC), Universita` di Padova, Padova, Italy fEngineering Geology (LTH), Lund University, Lund, Sweden gIRSA CNR, Bari, Italy

O U T L I N E 1 Introduction

56

2 Methods 2.1 Geo-electrical (DC resistivity) methods 2.2 EMI methods and GPR 2.3 Seismics

56

3 Application examples 3.1 System structure

67 68

Developments in Earth Surface Processes, Volume 23 https://doi.org/10.1016/B978-0-444-64177-9.00003-5

3.2 Fluid dynamics monitoring

62 64 67

55

74

4 Future challenges and conclusions

80

Acknowledgments

83

References

83

Further reading

89

© 2020 Elsevier B.V. All rights reserved.

56

3. Ground-based remote sensing of the shallow subsurface

1 Introduction Retrieving information using remote sensing methods is not a new concept when it comes to investigating the shallow subsurface. Noninvasive methods have been developed and used for well over a century, initially for the purpose of mining exploration (e.g., Telford et al., 1990), but with further extension to other application areas. The discipline has been variously called Applied Geophysics, Exploration Geophysics, and Geophysical Prospection. The reason for using the term “geophysics” will be clearer as we discuss the relevant methods. Recently, as environmental issues have received increasing attention, the shallow subsurface has been the subject of renewed focus for noninvasive methods. The result is that a number of specifically designed methods have been developed for environmental applications. Correspondingly, it is now common to refer to such methods as “Near-Surface Geophysics,” or “Environmental Geophysics” and “Hydrogeophysics” because the focus is on environmental and/or hydrological/hydrogeological problems. The development of new terms is partly motivated by an attempt to be more precise, and by the development of a jargon where specialists recognize themselves. Here we will cast all such terms into a common framework, as the ensemble of the relevant methods and applications is, in our opinion, a single discipline that spans many applications. In the ensuing parts of this chapter, we will first describe the methodologies in their general terms in order to introduce the reader to the nuts and bolts of the methods (Section 2). Some basic understanding of physics are potentially needed here, but to ensure accessibility to a wide range of readers, we avoid equations and try to make the description as understandable as possible, directing readers interested in the details to more specialized sources. Section 3 is devoted to applications, with separate attention to different classes of problems to be solved in this area, and specifically to the issues of defining the system geometry and its fluid dynamics. For the very specialized area of detecting the possible presence and distribution of contaminants in the subsurface, we have only provided relevant references. Section 3 has no ambition to be fully comprehensive: examples are provided solely from the authors’ experience, and all material shown here is novel and yet unpublished. Some limited references to the work of the wider community are given. Section 4 covers technological advances and future challenges beyond current techniques.

2 Methods The general approach of noninvasive geophysical methods can be summarized as shown in Fig. 1. Geophysics is a discipline based purely on physical measurements that are in most cases (but not all) performed at the soil surface, and in all cases at the boundary of a domain of interest. The domain is often the soil below our feet (thus the prefix geo-), but the same methodologies can be applied to other features, for example, to man-made structures such as walls, railways, embankments, and mechanical parts. Such techniques are sometimes called Nondestructive testing or NDT. Noninvasive geophysical methods are conceptually analogous to medical imaging techniques such as X-ray, CT scanning, and magnetic resonance imaging, with the only difference being the physical processes used to obtain images.

2 Methods

57

G(x,y) = G[P(x,y,z), F = forcing conditions] Instruments that measure a physical quantity G(x,y), along the domain boundary, x,y are surface coordinates

P(x,y,z) = Physical parameter, spatially distributed in the subsoil, influencing response G(x,y)

Domain to investigate, potentially 3D with coordinates x,y,z

FIG. 1 A general conceptual framework for geophysical measurements.

The general framework can be summarized as follows: a physical instrument collects data concerning a physical quantity G. In general, G is a function of the spatial coordinates along the outer surface, often the ground surface, i.e., G(x, y). The information content of the function G(x, y) lies in its physical link to a spatially distributed physical parameter P(x, y, z). This parameter depends on soil/rock properties and its state (e.g., water content, temperature). This property is potentially time-dependent so, more generally, we can consider the state variable G(x, y, t)—e.g., soil vibration—and P(x, y, z, t). P, together with the forcing conditions F that might be required to obtain a measurable signal, is thus responsible for the physical response G. The functional relationship [P,F] ! G contains the physics of the phenomenon in question, and is, in general, the solution of a partial differential equation constrained with suitable boundary and initial conditions, where G(x, y, t) is the unknown to be determined (the state variable) and P(x, y, z, t) is the parameter modulating the partial differential equation. [P,F] ! G is generally referred to as the “forward model,” i.e., the model that allows the prediction of the system’s response, once the system’s structure P(x, y, z, t) is known. From a practical standpoint, knowledge of [P,F] ! G is essential but still insufficient to characterize the subsurface, as our aim is generally the retrieval of P(x, y, z, t) from G(x, y, t) and not vice versa. A very general schematization of geophysical activities is shown in Fig. 2: while the physics produces a signal G given P, our analysis aim is to retrieve P given the measured G (still accounting for the forcing conditions F). This latter process is named “inversion” and the conceptual inverse function [G,F] ! P is named the “inverse model.” Before inversion can be conducted, it is, however, necessary that the measured G be cleaned of any “noise” component, i.e., any physical signal that is NOT generated by the system’s particular physics, i.e., the particular [P,F] ! G we are considering: this step is named signal “processing” (Fig. 2). The inversion step in Fig. 2 is the essential component of geophysics, and not a trivial one. In fact, no matter which imaging technique we apply, there is no such thing as a closed-form inverse function [G,F] ! P. We can only retrieve P(x, y, z, t) by conducting an inversion

58

3. Ground-based remote sensing of the shallow subsurface

Survey design

Information FROM end users

Information TO end users

(Site conceptual model)

(Site conceptual model)

Physical parameter P(x, y, z)

Estimated P(x, y, z)

Signal G(x, y) Measurement

Results interpretation

Processing inversion

FIG. 2 Measurement and inversion in applied geophysics. Note that this is potentially a virtuous cycle where successive deepening of the investigation may be required.

process that consists in fitting the measured G with the predicted [P,F] ! G. This approach is equivalent to calibrating the forward model [P,F] ! G onto the field data G, effectively varying the parameter P controlling G. Such calibration can be performed on any dataset G. Unfortunately, there is no guarantee that a unique solution can be achieved within the error bounds of data from G. The core problem is that multiple spatial and temporal distributions of P(x, y, z, t) may satisfy [P,F] ! G on a given dataset, that is, solutions are not unique. Thus, in designing geophysical surveys the key goal of the practitioner is to implement a design that results in a solution of the inverse problem that is as unique as possible in the region of interest. This involves a full understanding of the practical problem to be solved: the relationship with the end user must be close, and the flow of information must go both ways. Of course, the results of geophysical surveys must be communicated to the end user in order to facilitate a valid (possibly joint) interpretation (Fig. 2). A solid understanding of the final user’s needs is critical for the geophysicist in order to design the survey. Fig. 2 also shows a link from the a priori knowledge of the end user to the geophysical survey design. On the basis of this a priori information, the geophysical survey shall be planned in order for the nonuniqueness of geophysical inversion to have minimal or no impact on the region of interest and, crucially, on the problem of interest. The final goal is always to develop a conceptual model of the site that is consistent with all available data, including the newly acquired geophysical data. All conceptual models of the site that satisfy the measured geophysical data may in fact have features in common and other alternative conceptual models (that may have been plausible at the start) may be ruled out as their geophysical response would not be consistent with the observed data. It is not uncommon that multiple forms of integrated geophysical data are collected at the same site. In this case, while each individual data inversion may lead to some nonunique results, the joint consideration of the two sets of plausible inverted models may lead to the identification of one or more conceptual models that satisfy all the available data (including non-geophysical data). This type of joint data analysis, and sometimes joint data inversion, can be extremely powerful but at the same time requires extra care in order to consider all data with respect to their actual information content, resolution, and reliability.

59

2 Methods

TABLE 1

Methods, measured quantities, physical parameters, and forcing conditions

Method

G

P

F

DC (direct current) electrical resistivity methods (often called geo-electrical methods—and in particular, electrical resistivity tomography—ERT)

Voltage

Electrical resistivity

Current intensity

Seismics

Soil vibration

Seismic velocity/ impedance

Mechanical source

Ground penetrating radar (GPR)

Electrical field

Dielectric constant (velocity/impedance)

Electrical pulse

Geo-magnetism

Magnetic field

Magnetic susceptibility/ permanent magnetization

None

Gravity

Gravitational acceleration

Density

None

Electromagnetic induction (EMI)

Secondary (timevarying) magnetic field

Electrical conductivity (1/resistivity)

Inducing magnetic field

Induced polarization (IP)

(Time-varying) Voltage

Chargeability

(Time varying) current intensity

Spontaneous potential (SP)

Voltage

Electrical resistivity

Natural current sources

All geophysical methods can be cast in the general framework described above. Of course, large differences may arise because of the different nature of the physical relationship [P,F] ! G that lies at the heart of the method. More specifically, the physical definitions listed in Table 1 apply. Each geophysical method leads to imaging the subsurface in terms of at least one (and usually, but not always, only one) physical parameter P. The information content of the data therefore lies in the values and spatial (sometimes temporal too) distribution of P(x, y, z, t). How this information can be useful for the final user depends on two key aspects of P: (a) The values of P must be informative in that they represent, albeit indirectly, a variable of direct interest for the problem at hand. Consider, for example, seismic velocity as a proxy for soil mechanical properties or electrical conductivity as indication proxy for pore water salinity. (b) The spatial distribution of P is informative, albeit not as a direct measure of physical properties, because its different values are indicative of different geological or subsurface formations, or different states of the same formation, so that the “image” of P(x, y, z, t) can be read as an “image” of the subsoil structure (or other features).

60

3. Ground-based remote sensing of the shallow subsurface

One must keep these two possible uses in mind when geophysical methods are chosen and applied to a particular problem. In more specific terms, the following criteria are should always be considered when a selection is to be made between different possible geophysical methods potentially applicable to a real-life application: (1) The parameter P of the selected geophysical method has to be informative for the goal of specific application, in terms of either (a) physical significance or (b) spatial distribution of the subsurface. (2) Once (1) is satisfied, the selected method has to have sufficient penetration to reach the required depth of investigation, and sufficient resolution to highlight the desired features at that depth. This requirement depends on, of course, both the physics of the problem ([P,F] ! G) and the survey design (including accessibility to a larger or smaller area of the bounding outer surface). In some cases, an appropriate survey design is sufficient to meet such a requirement (e.g., often for electrical resistivity tomography (ERT)), while in other cases the physics dictates the limit (e.g., for ground penetrating radar (GPR) and electromagnetic induction (EMI)). Criteria (1) and (2) must be passed by any geophysical method to be considered useful for the practical goal ad hand. Once these criteria are satisfied, a number of other factors should be taken into account for further selection among the suitable methods: cost, logistics, and environmental impacts. However, these further requirements have no impact on the choice of method if the first two requirements are not met. Shallow geophysical investigations, such as those focused on geomorphological issues, must take into account three key aspects of the subsurface (Fig. 3): – structure – (fluid)dynamics – contamination (at contaminated sites) Different geophysical methods have different sensitivities to the three subsoil components as shown in Fig. 3. While a general classification is hard to draw, Table 2 presents some indications on the relative suitability of the most common geophysical techniques for highlighting structure, dynamics, and contamination in the shallow subsoil. When applying geophysics to a site of geomorphological interest, and thus to relatively shallow depths (tens of meters at most), one should always keep in mind that structure, fluid dynamics, and (when applicable) contamination all affect the geophysical signal G. While this can be seen as a disadvantage if we are interested only in one component of the problem (e.g., structure), this fact can also be an advantage in that all components can be potentially extracted from geophysical data, provided that the individual effects on the geophysical signal are separated. For this process to fully achieve its goals, it is often necessary to couple geophysical measurements with a mechanistic model that can incorporate structure and on this basis predict dynamics and possibly also contaminant distribution. This model can therefore be calibrated against geophysical and other data, in most cases in their time-lapse changes, thus representing the site’s static and dynamic behavior to the best of the available information.

Evapo-transpiration

Evapo-transpiration

Spring

Spring

Water table Impermeable bedrock

structure FIG. 3

Water table Aquifer

Confinin

g layer

Impermeable bedrock

fluid-dynamics

Aspects of subsoil affecting shallow geophysical investigations.

Water table Aquifer

Confinin

g layer

Impermeable bedrock

contamination

Aquifer

Confinin

g layer

62

3. Ground-based remote sensing of the shallow subsurface

TABLE 2 Applicable methods and subsurface characteristics: more crosses indicate better applicability Method

Structure

Dynamics

Contamination

Seismics

+++

EM methods

++

++

+

DC resistivity methods

++

++

+

Ground penetrating radar

+++

++

+

Magnetics

++

Gravimetry

++

++

Induced polarization

+

+

++

++

++

Self-potential Nuclear magnetic resonance

+

+

++

The indication is purely qualitative. Note that in order to identify (fluid dynamics), time-lapse measurements must be conducted (with the sole exception of self-potential method).

In the following section, we sketch the specific physical and acquisition characteristics of the most common geophysical methods useful in geomorphology. Two very good reviews of methods for near-surface geophysics, and hydrogeophysics in particular (but the reviews are not limited to this topic), can be found in the books by Rubin and Hubbard (2005) and Vereecken et al. (2006).

2.1 Geo-electrical (DC resistivity) methods The development of methods based on the injection of DC electrical current in the ground, and the measurement of corresponding voltage differences between two electrodes, dates back nearly a century; for classical reviews see, for example, Keller and Frischknecht (1966) and Kelly (1977). The use of Ohm’s law allows, under simplifying conditions, the reconstruction of electrical resistivity and the distribution in the subsurface: the original method only provided one-dimensional (1D) vertical soil profiles (vertical electrical soundings or VES) or lateral resistivity 1D profiles. A major step forward was made in the early 1990s with the development of ERT that provides estimates of the spatial distribution [in two dimension (2D) or three dimension (3D)] of electrical resistivity. It can also give insight into the time evolution of electrical resistivity if repeated measurements are used. A comprehensive review of the method, including the use of time-lapse measurements, is given by Binley and Kemna (2005). ERT is probably the most widely used methodology for near-surface noninvasive characterization as: (a) it is easy to deploy on the ground; (b) there is a wide availability of software for data inversion, much of it free and open source; (c) it requires apparently little technical skill: standard acquisition sequences are often available for standard geometries, and acquisition and inversion can be performed even by untrained personnel;

2 Methods

63

(d) the spatial variability of electrical resistivity often reflects lithological contrasts as well as of the presence of water and its saline content (consider, e.g., Archie, 1942; Brovelli et al., 2005; Brovelli and Cassiani, 2010). Thus both structure and hydrological dynamics are amenable to be tackled using ERT. While the advantages of ERT are undeniable, we must stress here how the apparent ease of use of ERT is often deceiving. ERT surveys must be planned carefully, be custom-made and data quality must be checked in detail in order to remove data outliers. In addition, the method requires a firm estimate of data errors in order to control inversion. A good procedure is suggested by Binley et al. (1995) and consists in measuring the resistance of the so-called reciprocal configurations (swapping current with potential electrodes): differences between the two configurations are a good estimate of measurement errors. We also underline some limitations of ERT. In particular, the highest sensitivity of the method is close to the electrodes. This limits the distance of reliable investigation. In the most common case of surface investigations (i.e., with all electrodes placed at the soil surface), the reliable depth of investigation can be estimated to be about 1/4–1/5 of the electrode line length. Therefore, going to depths larger than 200 m requires electrode lines longer than 1 km, with serious logistic limitations, especially in urban or industrial environments. For deeper electrical investigations, it is common practice to resort to electromagnetic methods (see next section). Similarly, cross-hole ERT investigations require that the two (at least) boreholes equipped with electrodes be placed at a reciprocal distance smaller than their depth. ERT has been applied successfully to image the shallow subsurface for over 25 years. This includes surface applications, which are the most common, borehole applications (e.g., Bevc and Morrison, 1991), and laboratory investigations (Binley et al., 1996). Another key advantage of ERT is that it does not possess a spatial scale per se: unlike other methods, the resolution depends solely on electrode spacing which can go from centimeters to tens of meters. In addition, time-lapse ERT measurements allow for the imaging of time-varying processes in the shallow subsurface, with obvious links to hydrological processes. Pioneering work was conducted in the 1990s in the United States and the United Kingdom (Daily et al., 1992; Daily et al., 1995; Daily and Ramirez, 1995; LaBrecque et al., 1996; Slater et al., 1997); see Daily et al. (2004) for a review. We cannot cover the entirety of the scientific literature on ERT here, and so we refer readers to a few exemplar applications concerning contaminated sites (Cassiani et al., 2006), hyporheic zone (Crook et al., 2008), hillslope processes (Cassiani et al., 2009a, 2009b), transport in shallow aquifers (Kemna et al., 2002; Monego et al., 2010; Perri et al., 2012; Singha and Gorelick, 2005; Slater et al., 2000), and monitoring processes in hypersaline environments (Haaken et al., 2017). Some specific examples are given in Section 3. As a side note, we consider two other geo-electrical methods that differ from ERT and its antecedents in two ways—note that both methods date back to the pioneering work of early 1900s (Schlumberger, 1920): – The Mise-a`-la-masse method (MALM) MALM is a method originally developed to delineate electrically conductive ore bodies. An electrical current is passed through the body, while the resulting voltage values are

64

3. Ground-based remote sensing of the shallow subsurface

measured at the ground surface or in boreholes. The pattern of the resulting equipotential contour lines gives information on the geometry (shape, extent, dip, continuity) of the electrically conductive body. The same approach can be used for saline tracer tests. In this case, electrical current is injected into the conductive plume and its evolution is monitored with time. Recent applications to landfills and tracer tests can be found, e.g., in De Carlo et al. (2013) and Perri et al. (2018). In general, one can view MALM as a geo-electrical method in which inversion is not attempted in most cases because the geometry of the acquisition or size of the surveyed area do not allow for a sufficient data set to be acquired to image the volume of interest. Nevertheless, the information content of the data may be explored, often using some forward model and the predicted and measured voltage values are compared in a semiquantitative manner. – The Induced Polarization (IP) method IP method is a well-established geophysical exploration method (Sumner, 1976). Information concerning the subsoil are inferred from the measured voltage signals associated with polarization currents in the earth caused by a sudden change (switch off or on) in an injected current (hence the term “induced”). The method had already been adopted in the 1960s for the exploration of porphyry and massive sulfides. Acquiring time-domain IP measurements is simple using modern ERT equipment, and can be made at the same time as resistivity measurements. Inversion with imaging is also possible using the chargeability formulation of Seigel (1959). With the advancement of instrumentation, the spectral nature of the IP response, i.e., its dependence on frequency (Spectral-induced polarization or SIP), has been increasingly investigated, with the relevant inversion formulation in terms of complex electrical resistivity (e.g., Kemna et al., 2000). While IP, particularly in time domain, is often used to image the subsurface, its physical meaning is still elusive. While many mechanisms are known to exist, the relative importance of each is difficult to ascertain in practical applications (in spite of many overambitious attempts) and strong research needs are still unanswered (see Kemna et al., 2012, for a review).

2.2 EMI methods and GPR Two different classes of exploration methods are based on a more general exploitation of Maxwell’s equations of electromagnetism. In particular, once the electrical field is not stationary or quasi-stationary (DC methods), the entire realm of electromagnetic (EM) phenomena are called into play. From this broad spectrum of possible EM responses, two main phenomena may be considered under the umbrella of EMI and GPR: 1. A time-varying magnetic field generates an electrical field (and current, if conductors are present) according to Faraday’s law of EM induction. 2. The magnetic field is generated either by an electrical current, which can be described by Ampere’s or Biot-Savart’s law, or by a time-varying electrical field, which is called a displacement current and was introduced first by James C. Maxwell. In presence of an electrical conductor (e.g., the soil or the subsoil), the relative importance of the first phenomenon with respect to the second is controlled by the so-called loss factor, i.e., the ratio of electrical conductivity over the product of electrical permittivity and frequency,

2 Methods

65

which is used for exploration. If the loss factor is larger than one, then EM induction prevails, with diffusion characteristics. In this case EMI methods are possible. If the loss factor is smaller than one, then the entire suite of Maxwell’s equations applies, and the EM field behaves like a wavefield that is attenuated as an effect of electrical conductivity. This situation allows the use of what is called GPR. In a nutshell: – EMI methods measure electrical conductivity, i.e., the same as DC conductivity methods (electrical conductivity is the reciprocal of electrical resistivity), and they can be used practically in all conditions. However, the propagation of the EM field is diffusive in nature and thus poorly focused. This means that reconstruction (i.e., inversion of data) of the conductivity field is inevitably smeared. – GPR is a wave propagation method, thus it is based on reflection, refraction, waveguides, travel times, etc. It can exploit principles of geometrical optics, thus inversion is straightforward, even in tomographic terms. However, GPR does not always work. If the investigated medium is too conductive, the signal is destroyed because EM energy is converted to heat via Ohm’s law. In that event no investigation is possible. Note that in very resistive environments (such as glaciers) GPR can penetrate hundreds or thousands of meters. Near-surface applications are common for both classes of methods. Note that EMI includes a very large number of different methods (see, e.g., Telford et al., 1990), that go from very deep (e.g., Magneto-Tellurics) to very shallow (frequency-domain or FDEM) methods. For near-surface investigations, it is common to proceed with FDEM methods that are often carried out using a single coil spacing and a single frequency. In this classical approach it is only possible to produce maps of apparent electrical conductivity—i.e., at each location a single value of electrical conductivity is estimated assuming that the investigated volume is homogeneous. This can yield results both in terms of zonation and of time-lapse changes (e.g., Robinson et al., 2009; Cassiani et al., 2012). Note that EMI investigations of this type are also very useful as preliminary investigations for buried man-made structures (see, e.g., Cassiani et al., 2014). More advanced approaches, involving FDEM inversion and thus depth imaging, will be discussed below. An alternative to FDEM is given by time-domain EM (TDEM) methods, where the EM induction is triggered by a sudden cessation of current in a loop. This produces eddy currents in the subsoil that propagate similar to smoke rings to deeper regions, while getting larger and larger, and thus losing resolution. Nevertheless, it is still possible to invert the induced secondary magnetic field (generated by Ampere’s law) and produce 1D vertical profiles of electrical conductivity. The scale can go from a few meters to hundreds of meters (Nabighian and Macnae, 1991; Christiansen et al., 2006; Auken et al., 2015). Note that many other EM methods are available. The most notable one is Controlled Source Audio Magneto-Tellurics (CSAMT—Zonge and Hughes, 1991). In all cases, however, the general advantages and limitations of EMI methods apply as described above. GPR is a classical near-surface method (with the sole exception of glacier exploration, see, e.g., Parsekian et al., 2016). For a comprehensive introduction about GPR see, e.g., Annan (2005).

66

3. Ground-based remote sensing of the shallow subsurface

The applications of GPR to near-surface problems are manifold. Three aspects shall be considered: – GPR is a wave-propagation method: therefore, it is capable of “seeing” contrasts in reflection coefficients as determined primarily by electrical permittivity contrasts. Thus, it is a very effective method to characterize the structure of the near surface. This characterization can extend tentatively, down to 10 wavelengths under normal electrical conductivity values, i.e., under attenuation conditions. In practice, however, the deepest imaging depth is at most 10–20 m from the ground surface, excluding imaging through ice. A nice example is given, e.g., by Klenk et al. (2015) where the dependence of the images on changing GPR velocity is also shown as a function of changing soil moisture content. The influence of soil moisture is discussed in the following. – The velocity of the GPR EM wave depends, under normal conditions, solely on the electrical permittivity of the medium. This in turn depends, in the case of natural porous media, predominantly on the volumetric moisture content. Classical relationships have been established long ago (e.g., Topp et al., 1980); for a review see Klotzsche et al. (2018). Thus repeated GPR measurements can provide quantitative estimates of changing moisture content. This is irrespective of possible changes in soil water salinity that alter electrical conductivity in the same way as changing moisture content (a limitation for DC electrical methods such as ERT). However, GPR resolution depends on wavelength that in turn depends on source frequency. Tentatively, a typical frequency for geological investigation is 100 MHz, which corresponds to a wavelength of about 1 m. The highest frequencies (say 1 GHz) correspond to about 10 cm wavelength, with the corresponding penetration limited to about 1 m—there is no such thing as a free meal! – Because GPR energy is attenuated as a function of the electrical conductivity of the conducting medium, it is possible to estimate conductivity from attenuation. Some attempts have been made (e.g., Day-Lewis et al., 2003), even though amplitudes of GPR signals are also affected by source directivity and geometric spreading. In addition, ERT is surely a much simpler approach to measure electrical conductivity. Not surprisingly, GPR has proven very popular in near-surface applications. GPR can be used both at the surface and in boreholes. Leaving aside the simplest surface measurements for structure characterization (see an example in Section 3), GPR has been used for: – time-lapse GPR data collection from the surface to image infiltration processes (e.g., Van Overmeeren et al., 1997); – surface measurements affected by complex propagation modes in waveguides that are nevertheless very informative (Arcone, 1984; Arcone et al., 2003; Strobbia and Cassiani, 2007); – surface applications to estimate soil moisture content in agricultural contexts (e.g., Grote et al., 2003; Huisman et al., 2003); – cross-hole and hole to surface applications aimed at viewing structure and hydrological dynamics in the subsurface, sometimes in conjunction with ERT monitoring

3 Application examples

67

(Alumbaugh et al., 2002; Hubbard et al., 1997; Binley et al., 2002a, 2002b; Binley and Beven, 2003; Cassiani et al., 2004, 2008; Schmalholz et al., 2004; Cassiani and Binley, 2005; Deiana et al., 2007, 2008; Looms et al., 2008); – advanced inversion approaches to fully exploit the information content in GPR propagation (e.g., Keskinen et al., 2017).

2.3 Seismics The propagation of elastic waves in the subsurface is the basis of the most widely used geophysical exploration methods (Yilmaz, 2001). Being a wave-based method, seismic methods are focused and well amenable to imaging. Yet, for the near-surface applications as considered herein, seismic methods are often not the method of choice. The main limitation lies in the typical wavelengths of seismic wave that are often, with high frequencies, at least in the range of tens of meters. Although in some cases a combination of high frequencies and S-wave propagation may help in achieving high resolution (e.g., Deidda and Balia, 2001; Petronio et al., 2016); in general the classical seismic reflection method is unsuitable for shallow applications. Seismic refraction is an alternative approach based on Snell’s law and tracking of rays from sources to receivers (e.g., Zhang et al., 1998), and is widely applied in near-surface investigations. However, the most promising approach is probably the use of surface waves, and Rayleigh waves in particular (Aki and Richards, 2002). These waves are the solid equivalent of the waves on the surface of a fluid. Early attempts to extract information from surface waves date back to Jones (1958, 1962). A major step forward was made in the 1980s with the Spectral Analysis of Surface Waves developed by Nazarian and Stokoe II (1984) and with the historical progress in electronics and computers, which made it possible to devise multichannel techniques or Multichannel Analysis of Surface Waves (McMechan and Yedlin, 1981; Park et al., 1999). The information carried by Rayleigh waves lies in their dispersive characteristics: different frequencies travel at different speeds, and different frequencies involve different thicknesses below the ground surface, with lower frequencies involving deeper portions of the subsurface. The classical approaches produce 1D vertical profiles of shear wave velocities, the parameter that chiefly controls Rayleigh wave propagation. New techniques are being devised to produce laterally varying images of the shear wave velocities. For a complete overview, see the book by Foti et al. (2017).

3 Application examples The number of specific applications of near-surface noninvasive techniques is very large, ranging from geomorphological studies to the characterization of contaminated sites, from hydrogeological/hydrological problems to geotechnical characterization. In this section, we presented illustrative examples stemmed from our own experience. In particular, we list examples that are mainly focused on the structural characterization of the subsurface and others where the main interest is the fluid dynamics in the subsurface.

68

3. Ground-based remote sensing of the shallow subsurface

3.1 System structure 3.1.1 The Settolo site This site lies on the left bank of the Piave River, close to the city of Valdobbiadene, NE Italy. For a complete site description, see, e.g., Perri et al. (2012). The site is on a riparian aquifer used for irrigation and drinking water supply. As the water table is at an average depth of about 5–6 m b.g.l., with seasonal oscillations depending on the Piave River stage, the aquifer is very vulnerable to contamination from the surface. The area is made of alluvial sediments of the Piave River composed of gravel in a sandy matrix. With the exception of a conglomeratic moraine terrace probably of fluvio-glacial origin that stands about 40 m above the neighboring alluvial plain in the northern side of the site, the study area has no particular geomorphologic characteristics. The moraine that emerges on the sides of the alluvial plain also underlies the fluvial gravel, and forms its hydrological bedrock. The riparian aquifer geometry is relatively complex due to the presence of several buried paleo-river channels, filled with the gravelly sandy sediments. These paleo-channels are easily detectable by large-scale surface ERT surveys. Fig. 4 shows an instructive example of data taken along the same line across the main deposition and flow direction, showing the ERT results obtained using two different ERT

Electrical resistivity (Ohm m)

72 electrodes, spaced 5 m Depth (m)

1500 –20

1400

–40

1300

–60

1200 50

100

150

200

250

300

350

1100

Horizontal coordinate (m)

1000 900

48 electrodes, spaced 10 m

800

Depth (m)

700 600 500

–50

400 300 50

100

150

200

250

300

350

400

450

Horizontal coordinate (m) LINEA 2 La Linea 2 (indicata in giallo in fig. 3) è stata acquisita procedendo da NE a SO con 48 elettrodi equispaziati 10 m per un totale di 470 m. Al fine di valutare la qualità del dato aumentando la risoluzione, cioè riducendo a spaziatura interelettrodica e portandosi a profondità di investigazione ridotte, rispetto a quelle ottenute con la spaziatura 10 m, sulla stessa linea è stata acquisita una tomografia con 72 elettrodi e spaziatura interelettrodica pari a 5 m, per un totale di 355 m di sviluppo lineare ed una profondità di investigazione di poco più di 60 m rispetto al piano campagna. I risultati dell’inversione per la Linea 2, con le due differenti spaziature, sono riportati in fig. 6.

FIG. 4 Settolo site: ERT lines along the same profile. See discussion in the text.

200 100 0

3 Application examples

69

configurations: a higher resolution setup, using 72 electrodes spaced at 5 m, and a lower resolution setup using only 48 electrodes spaced at 10 m. Both these setups provide subsurface information, but the spacing of the latter, although leading to lower resolution information, allows deeper penetration of the subsurface. In both cases, a classical Wenner-Schlumberger configuration was used, offering a good balance between resolution and penetration. The instrument used was an Iris Instruments Syscal Pro resistivity meter. The inversion was performed using the Occam’s inversion freeware profiler/R2t by A. Binley (www.es. lancs.ac.uk/people/amb/Freeware). A comparison between the two ERT lines demonstrates some of the method’s characteristics. Both images show comparable anomalies: in both cases the deeper moraine is more electrically conductive than the gravel in the channels, and the contrast is large enough to provide a clear geometric image of the two formations. However, the shorter, highresolution line shows deeper anomalies as more vertically extensive—consider, e.g., the gravel body at about 250 m in the section. The same body is clearly delineated in the longer, lower resolution line, as current lines go deep enough so as to go around this resistive body and carry to the surface the relevant information (in terms of voltage differences). The shorter line also fails to detect the resistive body at 350 m that is apparent in the longer line because this anomaly is located at the right extreme of the short section, there are very few current lines crossing that area, and all of them are consequently collecting information on a much larger volume. In general, near the ends of the surveys and at depth, ERT imaging capabilities are diminished. On the other hand, a smaller electrode spacing has also clear advantages in terms of resolution, in the region where imaging is accurate. The water table appears much sharper in the shorter line than in the longer line, at a depth between 5 and 10 m (i.e., between 1 and 2 electrode spacings—note that is it probably around 6 m). Also, the shallow conductive anomaly at 130 m—a gas pipeline—is much more sharply seen by the line at 5 m spacing than by the other. 3.1.2 The Trecate site The contaminated site near Trecate (Novara, NW Italy) is the result of a dramatic incident: in 1994, a crude oil blowout took place from the TR24 well undergoing side-track drilling. Approximately 15,000 m3 of middle weight crude oil were released overland, contaminating both soil and groundwater. For a thorough discussion of the site and the state of its contamination, see, e.g., Burbery et al. (2004) and Cassiani et al. (2014). The site is characterized by a thick sequence of poorly sorted silty sands and gravels in extensive lenses, typical of braided river sediments. An artificial layer of clayey-silty material, about 1–2 m thick, placed as a liner for rice paddies, overlies most of the site. Fig. 5 shows the results of one GPR line conducted at the site using a PulseEkko Pro system with 100 MHz antennas. The survey is a typical zero-offset profile, i.e., transmitter and receiver antennas are placed next to each other and pulled together along the line. Conversion from two-way time reflection to depth is possible thanks to ancillary data collected using multiple offset GPR configurations: in particular, common depth point as described extensively also in the seismic literature. Fig. 5 shows how the GPR signal penetrates to about 7 m depth, despite the presence of fine sediments (these are potentially electrically conductive) that make the bed of the artificial rice paddies. In particular, the

70

3. Ground-based remote sensing of the shallow subsurface

Distance (m) 10

20

Tx

Rx 30

40

50

60

70

80

90

100

0 1 2

0 1 2

3 4

3 4

5 6

5 6

7 8 9 10

7 8 9 10

Depth (m)

Depth (m)

0

FIG. 5 Trecate site: a 100-MHz GPR line shows the presence of a paleo-channel created by a braided river system. See discussion in the text.

GPR line shows a paleo-channel created by a braided river system, filled with fine sediments, more electrically conductive than the surrounding gravels. The result of this geometry is that the GPR signal is more attenuated below the paleo-channel. 3.1.3 The Aviano site The industrial area in Aviano, NE Italy, has been the site of major Trichloroethilene/ Tetrachloroethilene (TCE/PCE) subsoil contamination that took place before the mid1980s. The chlorinated solvents have been detected in the deep phreatic aquifer some 12 km downstream of the site, where the water table emerges to the ground surface in a line of springs. Below the site, the phreatic surface is about 100 m deep. Thus, the migration of contaminants took place in the unsaturated zone. More specifically, the site is underlain by a thin silty clay layer (about 1 m thick), at an average depth of about 7 m below ground, that holds above a shallow perched water body. The clay layer is however not continuous everywhere, having been eroded by some paleo-channels that are encountered locally by the monitoring boreholes. These discontinuities allowed infiltration of the contaminated waters toward the deeper continuous phreatic aquifer. Therefore, the detection of these holes in the clay layers is a key aspect of site characterization and remediation. Fig. 6 shows how the clay layer can be easily detected using surface ERT lines. In both cases we used 120 electrodes spaced at 0.8 m, with a dipole–dipole skip 8 configuration (i.e., the current and potential dipoles are 8 m long) and a full reciprocal acquisition. The inversion was performed using the Occam’s inversion freeware profiler/R2t by A. Binley (www.es.lancs.ac.uk/people/ amb/Freeware). We note here that the presence of a continuous electrically conductive clay layer (ERT4 in Fig. 6) has the effect of short-circuiting the current originating from the surface, making the thickness of the layer impossible to ascertain. From Fig. 6 we can only conclude that the layer is either continuous (ERT4) or discontinuous (ERT1).

71

3 Application examples

ERT 1

Meters

East

West

Known paleo-channel Meters

Meters ERT 4

East

Meters

West

Electrical resistivity (Ohm m)

Silty clay continuous layer Meters

FIG. 6 Aviano site: two ERT lines showing, respectively, one or more discontinuities in the clay layer (ERT1 above) and its continuity (ERT4). See discussion in the text.

Fig. 7 shows the result of a cross-hole ERT survey conducted between two purposely installed boreholes roughly mid-way along line ERT4: on the left the stratigraphic log shows that the clay layer is here only about 0.5 m thick. The cross-hole ERT uses 48 electrodes, 24 in each hole at the side of the section. The electrodes are spaced at 0.8 m along each borehole. The acquisition scheme is a dipole–dipole skip 4 configuration (i.e., the current and potential dipoles are 6  0.8 ¼ 4.8 m long) and a full reciprocal acquisition. The inversion is again performed using the Occam’s inversion R2t freeware by A. Binley. Note how the electrically conductive layer is well marked, and the resistive gravels below are well visible, unlike in Fig. 6. The conductive anomaly, however, is thicker than the clay layer per se, as it incorporates also the perched aquifer above the clay layer. 3.1.4 The Fondo Paviani site The Fondo Paviani site, located near Legnago, Verona, Italy, is an embanked settlement extending over several hectares, dating back to the Late Bronze Age (Cupito` et al., 2015). The site is now largely covered by crops, and the investigation of its structure thus requires extensive nondestructive investigations. Fig. 8 shows the results of preliminary investigations at the site performed using a combination of EMI mapping and ERT lines. In particular, at this site we successfully tested the EMI inversion procedure described by Deidda et al. (2014, 2017). Data were acquired using a multifrequency conductivity meter GEM-2 by Geophex, with a total seven frequencies in the frequency range from 775 Hz and 47 kHz. The instrument

Silty clay layer

Meters

log10 (electrical resistivity) (Ohm m)

Meters FIG. 7

Aviano site: cross-hole ERT imaging. See text for a discussion.

73

3 Application examples

Z = 3.50 m 4,999,000

N UTM (m)

4,998,980

57 52 47 42 37

4,998,960

32 27 22

4,998,940

17 12 4,998,920

7

Electrical resistivity (Ωm)

N N

2 57

679,840 679,860 679,880 679,900 679,920 679,940 679,960

52

E UTM (m)

47

Meters

N

S

–3 –6 –9 –12 –15 –18

42 37 32 27 22 17

5

10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90

Meters

12 7 2

electrical resistivity (Ohm m)

FIG. 8 Fondo Paviani site: comparison between EMI inverted results at 3.5 m depth and the corresponding ERT line. See text for details and discussion.

is equipped with a differential GPS for automatic positioning, so that a 3D volume of electrical conductivity can be obtained from the 1D vertical inverted profiles. We also acquired several ERT profiles in the same area using an Iris Instrument Syscal Pro with a dipole–dipole skip 4 configuration and a full reciprocal acquisition. ERT inversion was performed using the freeware ProfileR by A. Binley (see above). The comparison shown in Fig. 8 shows how the EMI inverted data are totally comparable with the ERT results, and are extremely advantageous in terms of speed of acquisition, and consequently the spatial coverage can be much larger. Similar successful results are also described, e.g., in Boaga et al. (2018). 3.1.5 The Turriaco site The Turriaco site is located in the Friuli region (NE Italy) on the hydrographic left side of the Isonzo River. The study was designed to explore the characteristics of the Isonzo riparian zone. Here we present some results in terms of seismic characterization—note that other data, including ERT, were collected at the site. For a thorough description see Vignoli et al. (2016). From a lithological point of view, the site is a heterogeneous sequence of quaternary sediments mainly consisting of gravel with different contents of sand and silt originated

74

3. Ground-based remote sensing of the shallow subsurface

FIG. 9

Turriaco site: results of the sMOPA inversion of Rayleigh waves (right) starting from a suite of shots along the line (an example on the left). See text for details.

by Alpine erosion. The braided nature of the Isonzo River partly accounts for the heterogeneous nature of the sediments. The subsoil is an alternate sequence of facies with high (gravel and sand) and low hydraulic conductivity (clay and silt). Here we discuss some unpublished results concerning seismic acquisition and surface wave inversion. The seismic signals were generated by a sledgehammer and recorded by means of a Geode seismograph (Geometrics Inc.) using arrays of 48 4.5-Hz geophones spaced 2 m. Multiple shots were acquired along each line and processed as described in Vignoli et al. (2011). Results of the sMOPA inversion for one of the lines are shown in Fig. 9 together with a sample shot. The procedure is capable of producing a 2D profile of shear wave velocity with no need for lateral a priori constraints, which are frequently unjustified.

3.2 Fluid dynamics monitoring 3.2.1 The Decimomannu site The Decimomannu military air base is located in Southern Sardinia, Italy, a few km north of Cagliari. Some instances of soil contamination have been observed recently as a consequence of three known spills of jet fuel (JP8) from supply pipelines. Several containment and remediation activities have been in place for some years. Now, novel in situ remediation actions are being planned. Here we present the preliminary results of noninvasive monitoring of feasibility injection tests. Two different solutions were injected in sequence over a few hours. We used a fixed ERT line composed of 48 electrodes spaced at 1 m. The scheme was a dipole–dipole skip zero (i.e., with dipoles having the minimal size of 1 m) with full reciprocal acquisition. Inversion is based on the ratio inversion approach (see, e.g., Cassiani et al., 2006) using the profileR freeware by A. Binley.

75

3 Application examples

First tracer injection

Rapid tracer dilution Electrical resistivity ratio w.r.t. background (%)

Return to background conditions Second tracer injection

Tracer migration

Tracer positioning at depth

Tracer dilution

New injection of second tracer FIG. 10 Decimomannu site: time-lapse surface ERT measurements show the extent and migration of fluids injected for contaminant remediation. See text for details.

Fig. 10 shows the time-lapse results of the monitoring: the injection borehole is located at the center of the monitoring line. The entire sequence lasts for only a few hours. The results are presented as resistivity ratio with respect to background conditions, reported as a percentage. It is apparent that both injected solutions are visible in terms of electrical resistivity reduction, even though the first solution is clearly less conductive than the second. It is crucial to know where the injected solution moves in terms of invasion of the contaminated zone if an effective remediation strategy is to be implemented. 3.2.2 The Trento Nord site The site is located in the city of Trento, Trentino Province (NE Italy). Since the early 1900s the area housed important chemical industries that left a legacy of heavy contamination, particularly in terms of organic compounds. Among others techniques, in situ ozonation was

76

3. Ground-based remote sensing of the shallow subsurface

selected as the most efficient method to remediate soil and groundwater pollution, due to the high capability of ozone to oxidize organic contaminants to safe levels. In a very permeable matrix, in situ ozone hydrocarbon oxidation is very successful; in the presence of relatively low permeability soils, as is the case of the considered site, the treatment may instead be long and costly. For this reason, hydraulic fracturing by means of pressurized water has been tested in conjunction with ozonation. A preliminary phase of hydraulic fracturing testing was performed to assess the best water injection rate and pressure values. The fluid used for hydraulic fracturing was a nearly saturated NaCl brine, the arrival of which in nearby boreholes is traditionally monitored via downhole electrical conductivity meters. Cross-borehole ERT was proposed and implemented as an additional imaging technique to assess the effectiveness of brine migration following fracturing. This monitoring was proposed because ERT is known to be very sensitive to conductivity contrasts at depth. The measurement scheme was composed of a dipole–dipole skip 1 (AB-MN) and a cross-hole bipole (AM-BN) configurations, using a total of only 259 electrodes combinations. Note that while each cable had 24 electrodes, only 30 of the 48 electrodes were actually utilized, as the water level was at about 3.6 m below ground. Hereafter we present the results of the time-lapse ERT imaging during the preliminary fracturing tests (Fig. 11). Note that the injection chamber was placed in the injection hole (borehole 0 in Fig. 11)—approximately between 7.5 and 8 m, i.e., in the sand-with-gravel formation. The high injection pressure, however, clearly moved the brine upward, fracturing the silt (a desired result). This is confirmed by the time-lapse ERT data. Note that the ERT results are heavily affected in this case by the so-called borehole effect: the brine invasion of the boreholes where the electrodes are placed produces artifacts. Higher resistivity than background appears in the regions near the boreholes—a consequence of the short-circuiting within the brine-invaded boreholes themselves. Nevertheless, the results are very informative in practical terms. 3.2.3 The Grugliasco site The Grugliasco site is in the campus of the Agricultural Faculty of the University of Turin, Italy. The soil in this area is largely made of aeolian sands. The unsaturated zone has a porosity ranging between 0.35 and 0.4, high vertical permeability, and low organic content; therefore, the area represents an ideal test site for percolation studies. The water table is at about 20-m depth. A full description of the site can be found elsewhere (Cassiani, et al., 2009c; Manoli et al., 2015; Rossi et al., 2015; Raffelli et al., 2017). An irrigation experiment was performed at the site on September 28, 2004 by means of a line of sprayers, placed in the middle of the experimental area, in order to wet an area of about 3 m by 20 m completely. The soil was initially extremely dry as a consequence of an exceptionally dry summer period. We performed time-lapse surface-to-surface GPR monitoring during the experiment using a PulseEkko 100 radar system with 200 MHz antennas in wide-angle reflection and refraction (WARR) configuration, i.e., keeping the transmitter fixed and moving the receiver with offset increments equal to 10 cm over a 14-m line. A GPR WARR survey was acquired before the start of irrigation, and then roughly every 2 h over the 6-h irrigation period. Fig. 12 shows the corresponding four WARR GPR images. Here we present a simple interpretation based on classical critical refraction theory, albeit other more sophisticated approaches are possible

FIG. 11

Trento Nord site. Left: background cross-hole ERT image compared against the local lithological profile (from drilling). Right: time-lapse images of resistivity changes as % ratios with respect to background conditions. See text for details and discussion.

78

3. Ground-based remote sensing of the shallow subsurface Intercept = 13.5 ns

Intercept = 10.5 ns

0.30 m/ns

0.30 m/ns

Intercept= 22 ns

0.30 m/ns

0.075 m/ns

0.14 m/ns

0.14 m/ns

0.14 m/ns

0.14 m/ns

0.30 m/ns

0.075 m/ns

0.075 m/ns

Preirrigation WARR

0.52 m

After 2h of irrigation

0.61 m

After 4h of irrigation

0.98 m

After 5.5 h of irrigation

FIG. 12 Grugliasco site: time-lapse GPR WARR data during an infiltration experiment (above) and corresponding interpretation in terms of refraction analysis (below). See text for details and discussion.

based on guided wave analysis (Strobbia and Cassiani, 2007; Rossi et al., 2015). As a result, the infiltration speed is estimated directly from the GPR WARR data—a procedure described by van Overmeeren et al. (1997). 3.2.4 The Bregonze site The Bregonze experimental site is located in a pre-Alpine area, north of Vicenza (NorthEastern Italy)—for a full description see Vignoli et al. (2012). The area of interest is the headwater catchment drained by an ephemeral stream, characterized by roughly 1.5ha (about 200 m long and 100m wide, altitude from 375 to 395 a.m.s.l.) and very mild slopes (7.5%). From a geological point of view, the site is composed of Upper Paleocene-Holocene altered volcanic deposits, with a strong clayey component. The soil, down to about 1–1.5 m, is of a silty-clay nature with average weight fractions equal to 21% sand and 79% silt+clay, and an organic matter content around 13%. Given the very low hydraulic conductivity of the deeper subsoil, this upper soil layer controls the site’s hydrology. Therefore, monitoring of the volumetric soil moisture content is particularly important at this site in order to help calibrate full scale hydrological models (e.g., Weill et al., 2013). We used a GF Instruments CMD electro-magnetometer with different configurations. Fig. 13 shows the maps obtained using the CMD1 configuration with vertical loops, with a depth of investigation around 1.5m. Note that the mapped quantity is the apparent electrical resistivity within this 1.5 m thickness, which we calculate using an average value. Measureable changes are apparent in less than 2 months. Similar results have been observed by a number of authors (e.g., Robinson et al., 2007). 3.2.5 The Bari IRSA-CNR site This is an experimental site located in the IRSA-CNR headquarters in Bari, Apulia (Southern Italy). The site geology is characterized, from top to bottom, by a relatively thin soil layer (1.5 m) followed by a 5-m thick layer of calcarenite, a sedimentary carbonatic rock

79

3 Application examples 5,069,500

5,069,500 70

5,069,460

5,069,420 5,069,400 5,069,380 5,069,360

50 40 30

5,069,380

69

0 69

3,

3,

55

60

0

0 50 69

Easting UTM 5,069,500 70

5,069,460

5,069,460

60

5,069,420 5,069,400 5,069,380 5,069,360

60

5,069,440 50 40 30

Northing UTM

Resistivity [Wm]

5,069,440

70

5,069,480

Resistivity [Wm]

5,069,480

5,069,420 5,069,400 5,069,380 5,069,360

50 40 30

5,069,340

5,069,340 20

0 69

3,

60

0 69

3,

55

0 69

3,

50

0 45

40 3, 69

3, 69

10

0

0

5,069,300

60

0 69

3,

55

0 69

3,

50

0 45 3,

Easting UTM

10/10/2014 3,

10 69

20 5,069,320

69

22/09/2014

5,069,300 69 3, 40 0

3,

45 3, 69

3, 69

3, 69

0

10

0 40

0 60

0 69

3,

55

0 69

3,

50

0 45 3, 69

02/09/2014

5,069,300

10

5,069,500

Northing UTM

30 20

5,069,320

Easting UTM

5,069,320

40

5,069,340

0 40 3,

5,069,400

50

20

18/08/2014

5,069,300 69

5,069,420

5,069,360

5,069,340 5,069,320

60

5,069,440

Northing UTM

Resistivity [Wm]

Northing UTM

5,069,460

60

5,069,440

70

5,069,480

Resistivity [Wm]

5,069,480

Easting UTM

FIG. 13 Bregonze site: time-lapse maps of apparent electrical resistivity as obtained from EMI measurements— note the changes over time. See text for a discussion.

of marine origin of Plio-Pleistocene age. Calcarenite is a porous rock, slightly cemented, made of a granular skeleton and carbonatic cement. The calcarenite formation lies on top of a karstic fractured limestone, about 25 m thick, that constitutes the aquifer. The water table is located in this formation. At greater depths, the wells encounter a dolomitic limestone about 20 m thick. We conducted an infiltration experiment, conceptually similar to the ones described by Deiana et al. (2007, 2008). A 1.5-m deep trench was dug to allow water infiltration into the underlying calcarenite between the existing boreholes C and E, and a mildly saline aqueous solution was used as a tracer in order to be potentially visible both above and below the water table. About 20 m3 of water were injected in 4 h, from 14:30 to 18:30 on March 17, 2010. A combination of cross-hole multiple offset gather GPR, vertical radar profiles, and surface electrical resistivity tomography (ERT) was used to monitor in time-lapse mode the dynamics of the vadose zone, while the deeper part of the profile was imaged

80

3. Ground-based remote sensing of the shallow subsurface

March 17 - 18:00

Calcarenite

–6.50

0 –1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11

0.5 0.4 0.3 0.2 0.1 0

MOG GPR 0 1 2 3 4 5 6 7 8 9 Meters

Background Soil water content

Limestone

Meters below ground

Soil

0.3 0.275 0.25 0.225 0.2 0.175 0.15 0.125 0.1 0.075 0.05 0.025 0

Water content change with respect to background (from GPR)

–36.00

0 –1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11

0 –1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11

Meters below ground

Soil

Meters below ground

0.00 m –1.30

Borehole E

MOG GPR GPR MOG 0 1 2 3 4 5 6 7 8 9

March 18 - 15:00

MOG MOG GPR GPR 0 1 2 3 4 5 6 7 8 9

March 19 - 15:00 0 –1 –2 –3 –4 –5 –6 –7 –8 –9 –10 –11

Meters below ground

Borehole C

MOG GPR MOG GPR 0 1 2 3 4 5 6 7 8 9 Meters

Bari IRSA-CNR site. Left: background conditions as imaged using multiple offset gather (MOG) GPR tomography (converted to moisture content using Topp et al., 1980), compared against borehole stratigraphy. Right: time lapse changes following the trench infiltration experiment.

FIG. 14

using cross-hole time-lapse ERT. The monitoring results are potentially useful as calibration data for a variably saturated flow and transport model. Here we focus on the MOG GPR data (Fig. 14). A substantial delay is observed in the vertical migration of water, probably due to fast initial lateral spreading.

4 Future challenges and conclusions The use of noninvasive techniques for subsoil characterization is now well established across many application areas. Despite this, the increasing use of such techniques for shallow applications, of direct environmental interest, has posed formidable challenges over the past three decades. Many such challenges have been overcome, as shown earlier. Yet the main issue that remains is to disseminate these techniques to a wider audience that is still reluctant to apply the latest techniques. There are luckily many exceptions. For instance, the hydrological community has been an eager adopter of shallow imaging: it is very common to see papers and presentations where standard hydrologists use and comment geophysical results

4 Future challenges and conclusions

81

(hydro-geophysical in that case) as part of their own toolbox. This an extraordinary success. Other communities are more difficult to penetrate. But it is only a matter of time and commitment: geophysicists must learn to sell their results in a manner that is easy to digest for end users, and must learn from end users what their actual needs are (see Fig. 2) in order to devise the best approaches to answer their practical questions. As geophysics, and nearsurface geophysics in particular, is a frontier discipline, we (geophysicists) must learn to communicate more and better knowing that the value of the information we can provide is much higher than currently felt outside of our field. Of course, a number of advances can and should also be made. We envision quite a few areas of promising progress, in no particular order, mixing techniques and applications: – EMI inversion: the availability of EMI instruments that can collect multifrequency or multi-coil data opens a wide range of possibilities in terms of inverting (at least in vertical 1D) the EMI data and ultimately obtaining 3D volumes of electrical conductivity. Such techniques can also be used in time series. The advantage of this method is that the measured quantity is the same as that measured by ERT but it can be used as a finer resolution/detailed investigation, and can be used to complement traditional ERT data collection. See, among others, the recent application by Boaga et al. (2018) and von Hebel et al. (2018). – Advanced analysis of seismic surface waves: even though the physics have been well understood for many years, there remains much work to be done to bring surface wave (SW) investigations to a mature stage. In particular, the analysis of phase dependence versus offset, also taking into account the multimodal propagation of SWs, is the path forward to devise 2D and 3D tomographic techniques. The future of SW is brilliant, and will complement all other near-surface techniques, where classical reflection and refraction seismics still play too minor a role. The reader can follow developments along this line the pioneering work of Strobbia and Foti (2006), Vignoli and Cassiani (2010), Vignoli et al. (2011, 2012, 2016) and others to come. – Airborne EMI and TDEM in particular: the noncontact characteristics of EMI measurements paves the way to a number of exciting developments, especially when coupled with EMI inversion techniques (see above). Helicopter-based TDEM investigations are now state-of-the-art (e.g., Viezzoli et al., 2008); smaller scale applications are conceivable (e.g., using drones). – IP/SIP quantitative interpretation: while IP has been around for many years, and substantial effort has been expended to attach a physical explanation to its response, in a wide range of applications (consider, e.g., Kemna et al., 2004; Lesmes and Friedman, 2005; Binley et al., 2005; Ntarlagiannis et al., 2006; Cassiani et al., 2009a, 2009b; Kemna et al., 2012), there is still quite some way to go before one can safely interpret IP data in more than a pure “imaging” sense. Yet, IP and SIP in particular contain information that comes from aspects of the subsurface that are otherwise impossible to explore, with special reference to the properties of porous media inner interfaces, and the relevant links to permeability, grain size distribution, contamination, and other properties. Thus, we expect IP/SIP to be the object of long-term exciting research. – SP (spontaneous potential): somehow, SP has suffered the same ups and downs as IP/SIP: like IP/SIP, a true and honest quantitative interpretation of the SP technique is still difficult, despite loud claims of success in the past. Nevertheless, SP clearly carries information, and shall be further exploited (see, e.g., Naudet et al., 2003).

82

3. Ground-based remote sensing of the shallow subsurface

– Surface nuclear magnetic resonance (SNMR): while NMR is state of the art in medical imaging (as much as X-ray, CT), its use in subsurface investigations is still lagging behind, in spite of promising results (e.g., Braun et al., 2009). This is not so much due to theoretical issues, but instead because the measurable signal is very small under field conditions, so as to be overwhelmed, e.g., in urban areas. Yet, SNMR measures properties that cannot be measured otherwise, with particular emphasis on “free water,” and thus permeability. Potentially there are big advances to make, yet if the ultimate limitations will be proven to be in the physics, this might be a dead end. – Gravimetry: this is a very old technique, and often used for large-scale characterization. However, the precision of modern instruments easily extend into the microGal range. In addition, time-lapse measurements have been proven to detect changes in mass associated with water storage at a variety of scales, from the local to the regional scale (e.g., Biegert et al., 2008). Coupling with hydrological modeling is straightforward, and may open unexpected opportunities (e.g., Piccolroaz et al., 2015). – Very small scale applications, e.g., agricultural applications (plant roots): the scale of investigation of some techniques (particularly EMI and ERT) can be made small enough to investigate the subsurface in the region of practical and scientific interest for the biosphere. The concepts of the Earth’s critical zone and that of the soil–plant-atmosphere continuum have long been introduced in order to define areas of extremely high importance in a variety of natural sciences: the interface between the solid planet and its atmosphere is key to a number of vital processes, and all involve mass and energy transfer. At this scale, noninvasive techniques may prove invaluable, and major progresses are being made (e.g., Allred et al., 2008; Petersen and Al Hagrey, 2009; Cassiani et al., 2012, 2015, 2016; Consoli et al., 2017; Vanella et al., 2018; Mary et al., 2018, to mention a few). – The link to contamination: it should always be remembered that contamination can be dangerous, or even deadly, at concentrations that are so small as not to produce any physical signal (and indeed, concentrations are measured by chemical methods). Thus, it may sound overambitious to expect geophysical methods to be able to detect contamination. Yet, geophysics may identify side effects. If these effects can be disentangled from other signal sources (typically, structure and dynamics) then it is possible to relate noninvasive measurements to contaminant distributions (e.g., K€astner et al., 2012; Cassiani et al., 2014). – Data assimilation into models: modeling of processes means understanding processes. Ultimately, this is the final goal of any scientific process, including subsoil investigations. Thus major efforts shall be expended to try and blend data, and noninvasive data, with models. This can be done by ad hoc analyses (e.g., Preti et al., 2018; Robinson et al., 2007) or in a strict sense (e.g., Manoli et al., 2015; Camporese et al., 2011, 2015). Either way, this is at the heart of the scientific method, so any progress in this direction is welcome. – Last but not least, technological progresses: for instance, optical fiber measurements of seismic waves. Another possibility is directional drilling to place electrodes or optical fibers in the subsoil (Fig. 15, Busato et al., 2018). Distributed sensors (temperature, vibrations, etc.) in integrated circuits (e.g., MEMS) are also promising. The range of possibilities is immense.

References

83

FIG. 15

Vermigliana site: ERT imaging of the hyporheic zone made possible by directional drilling electrode placement below the river bed (Busato et al., 2018).

Acknowledgments We are in debt with countless people, who helped develop both the general view presented in this chapter, and the specific results we described in the case studies. As we cannot acknowledge all, admitting the risk of forgetting some, we play it safe and thank Professor Andrew M. Binley, whose example has been a reference for all of us, and whose continuous and dedicated efforts have made the discipline progress beyond expectations. He is a sincere friend for many of us.

References Aki, K., Richards, P.G., 2002. Quantitative Seismology, second ed. University Science. Allred, B.J., Daniels, J.J., Ehsani, M.R., 2008. Handbook of agricultural geophysics. CRC Press, Taylor and Francis Group, New York. Alumbaugh, D., Chang, P.Y., Paprock, L.I., Brainard, I.R., Glass, R.J., Rautman, C.A., 2002. Estimating moisture contents in the vadose zone using cross-borehole ground penetrating radar: a study of accuracy and repeatability. Water Resour. Res. 38, 1309. https://doi.org/10.1029/2001WR000754.

84

3. Ground-based remote sensing of the shallow subsurface

Annan, A.P., 2005. GPR methods for hydrogeological studies. In: Rubin, Y., Hubbard, S.S. (Eds.), Hydrogeophysics, Ser. In: vol. 50. Springer, Dordrecht, pp. 185–214. Archie, G.E., 1942. The electrical resistivity log as an aid in determining some reservoir characteristics. Trans. AIME 146, 54–62. Arcone, S.A., 1984. Field observations of electromagnetic pulse propagation in dielectric slabs. Geophysics 49, 1763–1773. Arcone, S.A., Peapples, P.R., Liu, L., 2003. Propagation of a ground-penetrating radar (GPR) pulse in a thin surface waveguide. Geophysics 68, 1922–1933. Auken, E., Christiansen, A.V., Kirkegaard, C., Fiandaca, G., Schamper, C., Behroozmand, A., Binley, A., Nielsen, E., Effersø, F., Christensen, N., Sørensen, K., Foged, N., Vignoli, G., 2015. An overview of a highly versatile forward and stable inverse algorithm for airborne, ground-based and borehole electromagnetic and electric data. Explor. Geophys. 46 (3), 223–235. https://doi.org/10.1071/EG13097. Bevc, D., Morrison, H.F., 1991. Borehole-to-surface electrical resistivity monitoring of a salt water injection experiment. Geophysics 56, 769–777. Biegert, E., Ferguson, J., Li, X., 2008. 4D gravity monitoring—Introduction. Geophysics 73. https://doi.org/ 10.1190/1.3010377. Special Section—4D gravity monitoring, WA1 (2008). Binley, A., Beven, K., 2003. Vadose zone flow model uncertainty as conditioned on geophysical data. Ground Water 41, 119–127. Binley, A.M., Cassiani, G., Middleton, R., Winship, P., 2002a. Vadose zone flow model parameterisation using crossborehole radar and resistivity imaging. J. Hydrol. 267, 147–159. Binley, A., Henry-Poulter, S., Shaw, B., 1996. Examination of solute transport in an undisturbed soil column using electrical resistance tomography. Water Resour. Res. 32, 763–769. Binley, A.M., Kemna, A., 2005. DC resistivity and induced polarization methods. In: Rubin, Y., Hubbard, S.S. (Eds.), Hydrogeophysics. Water Sci. Technol. Library, Ser. In: vol. 50. Springer, New York, pp. 129–156. Binley, A., Ramirez, A., Daily, W., 1995. Regularised image reconstruction of noisy electrical resistance tomography data. In: Beck, M.S., Hoyle, B.S., Morris, M.A., Waterfall, R.C., Williams, R.A. (Eds.), Process Tomography—1995, Proceedings of the fourth Workshop of the European Concerted Action on Process Tomography, Bergen, 6–8 April 1995, pp. 401–410. Binley, A.M., Slater, L.D., Fukes, M., Cassiani, G., 2005. The relationship between frequency dependent electrical resistivity and hydraulic properties of saturated and unsaturated sandstone. Water Resour. Res. 41. https:// doi.org/10.1029/2005WR004202. Binley, A., Winship, P., West, L.J., Pokar, M., Middleton, R., 2002b. Seasonal variation of moisture content in unsaturated sandstone inferred from borehole radar and resistivity profiles. J. Hydrol. 267, 160–172. Boaga, J., Ghinassi, M., D’Alpaos, A., Deidda, G.P., Rodriguez, G., Cassiani, G., 2018. Geophysical investigations unravel the vestiges of ancient meandering channels and their dynamics in tidal landscapes. Sci. Rep. 8, 1–8. ISSN: 2045–2322. https://doi.org/10.1038/s41598-018-20061-5. Braun, M., Kamm, J., Yaramanci, U., 2009. Simultaneous inversion of magnetic resonance sounding in terms of water content, resistivity and decay times. Near Surf. Geophys. 7, 589–598. Brovelli, A., Cassiani, G., 2010. A combination of the Hashin-Shtrikman bounds aimed at modelling electrical conductivity and permittivity of variably saturated porous media. Geophys. J. Int. 180, 225–237. https://doi.org/ 10.1111/j.1365-246X.2009.04415.x. Brovelli, A., Cassiani, G., Dalla, E., Bergamini, F., Pitea, D., Binley, A.M., 2005. Electrical properties of partially saturated sandstones: a novel computational approach with hydro-geophysical applications. Water Resour. Res.. 41. https://doi.org/10.1029/2004WR003628. Burbery, L., Cassiani, G., Andreotti, G., Ricchiuto, T., Semple, K.T., 2004. Well test and stable isotope analysis for the determination of sulphate-reducing activity in a fast aquifer contaminated by hydrocarbons. Environ. Pollut. 129 (2), 321–330. Busato, L., Boaga, J., Perri, M.T., Majone, B., Bellin, A., Cassiani, G., 2018. Hydrogeophysical characterization and monitoring of the hyporheic and riparian zones: the Vermigliana Creek case study. Sci. Total Environ. 648, 1105–1120. https://doi.org/10.1016/j.scitotenv.2018.08.179. Camporese, M., Cassiani, G., Deiana, R., Salandin, P., Binley, A., 2015. Coupled and uncoupled hydrogeophysical inversions using ensemble Kalman filter assimilation of ERT-monitored tracer test data. Water Resour. Res. 51 (5), 3277–3291. https://doi.org/10.1002/2014WR016017.

References

85

Camporese, M., Salandin, P., Cassiani, G., Deiana, R., 2011. Assessment of local hydraulic properties from electrical resistivity tomography monitoring of a three-dimensional synthetic tracer test experiment. Water Resour. Res. 47. https://doi.org/10.1029/2011WR010528. Cassiani, G., Binley, A.M., 2005. Modeling unsaturated flow in a layered formation under quasi-steady state conditions using geophysical data constraints. Adv. Water Resour. 28, 467–477. Cassiani, G., Binley, A., Kemna, A., Wehrer, M., Flores Orozco, A., Deiana, R., Boaga, J., Rossi, M., Dietrich, P., Werban, U., Zschornack, L., Godio, A., JafarGamdomi, A., Deidda, G.P., 2014. Non-invasive characterization of the Trecate (Italy) crude-oil contaminated site: links between contamination and geophysical signals. Environ. Sci. Pollut. Res. 21 (15), 8914–8931. https://doi.org/10.1007/s11356-014-2494-7. Special Issue on “New approaches for low-invasive contaminated site characterization, monitoring and modelling”. Cassiani, G., Boaga, J., Rossi, M., Fadda, G., Putti, M., Majone, B., Bellin, A., 2016. Soil-plant interaction monitoring: small scale example of an apple orchard in Trentino, North-Eastern Italy. Sci. Total Environ. 543 (Pt B), 851–861. https://doi.org/10.1016/j.scitotenv.2015.03.113. Cassiani, G., Boaga, J., Vanella, D., Perri, M.T., Consoli, S., 2015. Monitoring and modelling of soil-plant interactions: the joint use of ERT, sap flow and Eddy Covariance data to characterize the volume of an orange tree root zone. Hydrol. Earth Syst. Sci. 19, 2213–2225. https://doi.org/10.5194/hess-19-2213-2015. Cassiani, G., Bruno, V., Villa, A., Fusi, N., Binley, A.M., 2006. A saline tracer test monitored via time-lapse surface electrical resistivity tomography. J. Appl. Geophys. 59, 244–259. Cassiani, G., Godio, A., Stocco, S., Villa, A., Deiana, R., Frattini, P., Rossi, M., 2009a. Monitoring the hydrologic behaviour of steep slopes via time-lapse electrical resistivity tomography. Near Surf. Geophys. 475–486. special issue on Hydrogeophysics. Cassiani, G., Kemna, A., Villa, A., Zimmermann, E., 2009b. Spectral induced polarization for the characterization of free-phase hydrocarbon contamination in sediments with low clay content. Near Surf. Geophys. 547–562. https:// doi.org/10.3997/1873-0604.2009028. special issue on Hydrogeophysics—Methods and Processes. Cassiani, G., Ferraris, S., Giustiniani, M., Deiana, R., Strobbia, C., 2009c. Time-lapse surface-to-surface GPR measurements to monitor a controlled infiltration experiment. Boll. Geofis. Teor. Appl. 50 (2), 209–226. Cassiani, G., Fusi, N., Susanni, D., Deiana, R., 2008. Vertical radar profiles for the assessment of landfill capping effectiveness. Near Surf. Geophys. 6, 133–142. Cassiani, G., Strobbia, C., Gallotti, L., 2004. Vertical radar profiles for the characterization of the deep vadose zone. Vadose Zone J. 3, 1093–1115. Cassiani, G., Ursino, N., Deiana, R., Vignoli, G., Boaga, J., Rossi, M., Perri, M.T., Blaschek, M., Duttmann, R., Meyer, S., Ludwig, R., Soddu, A., Dietrich, P., Werban, U., 2012. Non-invasive monitoring of soil static characteristics and dynamic states: a case study highlighting vegetation effects. Vadose Zone J. 11, https:// doi.org/10.2136/2011.0195. Special Issue on SPAC—Soil-plant interactions from local to landscape scale. vzj2011.0195. Christiansen, A.V., Auken, E., Sørensen, K., 2006. The transient electromagnetic method. In: Kirsch, R. (Ed.), Groundwater Geophysics: A Tool for Hydrogeology. Springer-Verlag, Berlin, pp. 179–224. Consoli, S., Stagno, F., Vanella, D., Boaga, J., Cassiani, G., Roccuzzo, G., 2017. Partial root-drying irrigation in orange orchards: effects on water use and crop production characteristics. Eur. J. Agron. 82, 190–202. https://doi.org/ 10.1016/j.eja.2016.11.001. Crook, N., Binley, A., Knight, R., Robinson, D.A., Zarnetske, J., Haggerty, R., 2008. Electrical resistivity imaging of the architecture of sub-stream sediments. Water Resour. Res. 44. https://doi.org/10.1029/2008WR006968. Cupito`, M., Leonardi, G., Dalla Longa, E., Nicosia, C., Balista, C., Dal Corso, M., Kirleis, W., 2015. Fondo Paviani (Legnago, Verona): il central place della polity delle Valli Grandi Veronesi nella tarda Eta` del bronzo. Cronologia, aspetti culturali, evoluzione delle strutture e trasformazioni paleoambientali. In: del Veneto, P.e.P., Leonardi, G., Tine, V. (Eds.), Studi di preistoria e protostoria -2. Brevi Note, Istituto Italiano di Preistoria e Protostoria, Padova. Daily, W., Ramirez, A., 1995. Electrical-resistance tomography during in-situ trichloroethylene remediation at the Savanna River site. J. Appl. Geophys. 33, 239–249. Daily, W.A., Ramirez, A., Binley, A., LaBrecque, D., 2004. Electrical resistivity tomography. Lead. Edge 23 (5), 438–442. Daily, W.D., Ramirez, A.L., LaBrecque, D.J., Barber, W., 1995. Electrical resistance tomography experiments at the Oregon Graduate Institute. J. Appl. Geophys. 33, 227–237.

86

3. Ground-based remote sensing of the shallow subsurface

Daily, W., Ramirez, A., LaBrecque, D., Nitao, J., 1992. Electrical resistivity tomography of vadose water movement. Water Resour. Res. 28, 1429–1442. Day-Lewis, F.D., Lane, J.W., Harris Jr., J.M., Gorelick, S.M., 2003. Time-lapse imaging of saline-tracer transport in fractured rock using difference-attenuation radar tomography. Water Resour. Res. 39, 1290–1303. https://doi.org/ 10.1029/2002WR001722. De Carlo, L., Perri, M.T., Caputo, M.C., Deiana, R., Vurro, M., Cassiani, G., 2013. Characterization of the confinement of a dismissed landfill via electrical resistivity tomography and mise-a`-la-masse. J. Appl. Geophys. 98 (2013), 1–10. https://doi.org/10.1016/j.jappgeo.2013.07.010. Deiana, R., Cassiani, G., Kemna, A., Villa, A., Bruno, V., Bagliani, A., 2007. An experiment of non invasive characterization of the vadose zone via water injection and cross-hole time-lapse geophysical monitoring. Near Surf. Geophys. 5, 183–194. Deiana, R., Cassiani, G., Villa, A., Bagliani, A., Bruno, V., 2008. Model calibration of a water injection test in the vadose zone of the Po River plain using GPR cross-hole data. Vadose Zone J. 7, 215–226. https://doi.org/10.2136/ vzj2006.0137. Deidda, G.P., Balia, R., 2001. An ultrashallow SH-wave seismic reflection experiment on a subsurface ground model. Geophysics 66 (4), 1097–1104. Deidda, G.P., Diaz de Alba, P., Rodriguez, G., 2017. Identifying the magnetic permeability in multi-frequency EM data inversion. Electron. Trans. Numer. Anal. 47, 1–17. ISSN: 1068-9613. Deidda, G.P., Fenu, C., Rodriguez, G., 2014. Regularized solution of a nonlinear problem in electromagnetic sounding. Inverse Prob. 30, 1–27. ISSN: 0266-5611. https://doi.org/10.1088/0266-5611/30/12/125014. Foti, S., Lai, C.G., Rix, G.J., Strobbia, C., 2017. Surface Wave Methods for Near-Surface Site Characterization. CRC Press (9781138077737). Grote, K., Hubbard, S., Rubin, Y., 2003. Field-scale estimation of volumetric water content using ground-penetratingradar wave techniques. Water Resour. Res. 39, 1321. https://doi.org/10.1029/2003WR002045. Haaken, K., Deidda, G.P., Cassiani, G., Deiana, R., Putti, M., Paniconi, C., Scudeler, C., Kemna, A., 2017. Flow dynamics in hyper-saline aquifers: hydro-geophysical monitoring and modeling. Hydrol. Earth Syst. Sci. 21, 1439–1454. ISSN: 1027-5606. https://doi.org/10.5194/hess-21-1439-2017. Hubbard, S.S., Peterson, J.E., Majer Jr., E.L., Zawislanski, P.T., Williams, K.H., Roberts, J., Wobber, F., 1997. Estimation of permeable pathways and water content using tomographic radar data. Lead. Edge 16, 1623–1628. Huisman, J.A., Hubbard, S.S., Redman, J.D., Annan, A.P., 2003. Measuring soil water content with ground penetrating radar: a review. Vadose Zone J. 2, 477–491. Jones, R.B., 1958. In-situ measurement of the dynamic properties of the soil by vibration methods. Geotechnique 8 (1), 1–21. Jones, R.B., 1962. Surface wave technique for measuring the elastic properties and thickness of roads: theoretical development. Br. J. Appl. Phys. 13, 21–29. K€ astner, M., Braeckvelt, M., D€ oberl, G., Cassiani, G., Papini, M.P., Leven-Pfister, C., Van Ree, D., 2012. Model-Driven Soil Probing, Site Assessment and Evaluation: Guidance on Technologies. University of Rome La Sapienza Press, Rome, Italy. ISBN: 978-88-95814-72-8. Keller, G.V., Frischknecht, F.C., 1966. Electrical Methods in Geophysical Prospecting, International Series of Monographs in Electromagnetic Waves. vol. 10. Pergamon Press Inc, Oxford. 525 p. Kelly, W.E., 1977. Geoelectrical sounding for estimating hydraulic conductivity. Ground Water 15, 420–425. Kemna, A., Binley, A., Cassiani, G., Niederleithinger, E., Revil, A., Slater, L., Williams, K.H., Flores Orozco, A., ordt, A., Kruschwitz, S., Leroux, V., Titov, K., Zimmermann, E., 2012. An overview of the spectral Haegel, F.-H., H€ induced polarization method for near-surface applications. Near Surf. Geophys. 10 (6), 453–468. https://doi.org/ 10.3997/1873-0604.2012027. Kemna, A., Binley, A., Ramirez, A., Daily, W., 2000. Complex resistivity tomography for environmental applications. Chem. Eng. J. 77, 11–18. Kemna, A., Binley, A., Slater, L., 2004. Cross-borehole IP imaging for engineering and environmental applications. Geophysics 69, 97–105. Kemna, A., Vanderborght, J., Kulessa, B., Vereecken, H., 2002. Imaging and characterisation of subsurface solute transport using electrical resistivity tomography (ERT) and equivalent transport models. J. Hydrol. 267, 125–146. Keskinen, J., Klotzsche, A., Looms, M.C., Moreau, J., van der Kruk, J., Holliger, K., Stemmerik, L., Nielsen, L., 2017. Full-waveform inversion of Crosshole GPR data: implications for porosity estimation in chalk. J. Appl. Geophys. 140, 102–116. https://doi.org/10.1016/j.jappgeo.2017.01.001.

References

87

Klenk, P., Jaumann, S., Roth, K., 2015. Quantitative high-resolution observations of soil water dynamics in a complicated architecture using time-lapse ground-penetrating radar. Hydrol. Earth Syst. Sci. 19, 1125–1139. https://doi. org/10.5194/hess-19-1125-2015. Klotzsche, A., Jonard, F., Looms, M.C., van der Kruk, J., Huisman, J.A., 2018. Measuring soil water content with ground penetrating radar: a decade of progress. Vadose Zone J. 17 (1). https://doi.org/10.2136/vzj2018. 03.0052. Article Number: UNSP 180052. LaBrecque, D.J., Ramirez, A.L., Daily, W.D., Binley, A.M., Schima, S.A., 1996. ERT monitoring of environmental remediation processes. Meas. Sci. Technol. 7, 375–383. Lesmes, G., Friedman, S., 2005. Relationships between the electrical and hydrogeological properties of rocks and soils. In: Rubin, Y., Hubbard, S.S. (Eds.), Hydrogeophysics. The Netherlands, Springer, Dordrecht, pp. 87–128. Looms, M.C., Binley, A., Jensen, K.H., Nielsen, L., Hansen, T.M., 2008. Identifying unsaturated hydraulic parameters using an integrated data fusion approach on cross-borehole geophysical data. Vadose Zone J. 7, 238–248. Manoli, G., Rossi, M., Pasetto, D., Deiana, R., Ferraris, S., Cassiani, G., Putti, M., 2015. An iterative particle filter approach for coupled hydro-geophysical modeling and inversion of a controlled infiltration experiment. J. Comput. Phys. 37–51. https://doi.org/10.1016/j.jcp.2014.11.035. Mary B., Peruzzo L., Boaga J., Schmutz M., Wu Y., Hubbard S.S. and Cassiani G., Small scale characterization of vine plant root water uptake via 3D electrical resistivity tomography and Mise-a`-la-Masse method, Hydrol. Earth Syst. Sci. https://doi.org/10.5194/hess-2018-238. McMechan, G.A., Yedlin, M.J., 1981. Analysis of dispersive waves by wave field transformation. Geophysics 46, 869874. Monego, M., Cassiani, G., Deiana, R., Putti, M., Passadore, G., Altissimo, L., 2010. Tracer test in a shallow heterogeneous aquifer monitored via time-lapse surface ERT. Geophysics 75 (4), WA61–WA73. https://doi.org/ 10.1190/1.3474601. Nabighian, M.N., Macnae, J.C., 1991. Time domain electromagnetic prospecting methods. In: Nabighian, M.N. (Ed.), Electromagnetic Methods in Applied Geophysics, 02, Society of Exploration Geophysicists. vol. 1991. pp. 427–520. Naudet, V., Revil, A., Bottero, J.V., Begassat, P., 2003. Relationship between self-potential (SP) signals and redox conditions in contaminated groundwater. Geophys. Res. Lett. 30, 2091. https://doi.org/10.1029/2003GL018096. Nazarian, S., Stokoe II, K.H., 1984. In situ shear wave velocity from spectral analysis of surface waves: proceeding of the 8th conference on earthquake. Engineering 3, 31–38. Ntarlagiannis, D., Williams, K.H., Slater, L., Hubbard, S., 2006. Low-frequency electrical response to microbial induced sulfide precipitation. J. Geophys. Res. 110. https://doi.org/10.1029/2005JG000024. Park, C.B., Miller, R.D., Xia, J., 1999. Multichannel analysis of surface waves. Geophysics. 64(3). Parsekian, A.D., Bradford, J., Tsoflias, G., Arcone, S., Kulessa, B., 2016. Advancements in the measurement of the cryosphere using geophysics—introduction. Geophysics 81 (1), WAI–WAII. https://doi.org/10.1190/20151120-SPSEINTRO.1. Perri, M.T., Cassiani, G., Gervasio, I., Deiana, R., Binley, A.M., 2012. A saline tracer test monitored via both surface and cross-borehole electrical resistivity tomography: comparison of time-lapse results. J. Appl. Geophys. 79, 6–16. https://doi.org/10.1016/j.jappgeo.2011.12.011. Perri, M.T., De Vita, P., Masciale, R., Portoghese, I., Chirico, G.B., Cassiani, G., 2018. Time-lapse Mise-a´-la-Masse measurements and modelling for tracer test monitoring in a shallow aquifer. J. Hydrol. 561, 461–477. https://doi.org/ 10.1016/j.jhydrol.2017.11.013. Petersen, T., Al Hagrey, S.A., 2009. Mapping root zones of small plants using surface and borehole resistivity tomography. Lead. Edge (10), 1220–1224. Petronio, L., Boaga, J., Cassiani, G., 2016. Characterization of the Vajont landslide (North-Eastern Italy) by means of reflection and surface wave seismics. J. Appl. Geophys. 128 (2016), 58–67. https://doi.org/10.1016/ j.jappgeo.2016.03.012. Piccolroaz, S., Majone, B., Palmieri, F., Cassiani, G., Bellin, A., 2015. On the use of spatially distributed, time-lapse micro-gravity surveys to inform hydrological modeling. Water Resour. Res. 51 (9), 7270–7288. https://doi. org/10.1002/2015WR016994. Preti, F., Guastini, E., Penna, D., Dani, A., Cassiani, G., Boaga, J., Deiana, R., Romano, N., Nasta, P., Palladino, M., Errico, A., Giambastiani, Y., Trucchi, P., Tarolli, P., 2018. Conceptualization of water flow pathways in agricultural terraced landscapes. Land Degrad. Dev. 29 (3), 651–662. https://doi.org/10.1002/ldr.2764.

88

3. Ground-based remote sensing of the shallow subsurface

Raffelli, G., Previati, M., Canone, D., Gisolo, D., Bevilacqua, I., Capello, G., Biddoccu, M., Cavallo, E., Deiana, R., Cassiani, G., Ferraris, S., 2017. Local and plot scale measurements of soil moisture: an overview of different techniques applied in plain, hill and mountain experimental sites. Water 9 (9), 706. https://doi.org/10.3390/ w9090706. Robinson, D.A., Binley, A., Crook, N., Day-Lewis, F., Ferre, P.T., Grauch, V.J.S., Knight, R., Knoll, M., Lakshmi, V., Miller, R., Nyquist, J., Pellerin, L., Singha, K., Slater, L., 2007. Advancing process-based watershed hydrological research using near-surface geophysics: a vision for, and review of, electrical and magnetic geophysical methods. Hydrol. Process. 22, 3604–3635. Robinson, D.A., Lebron, I., Kocar, B., Phan, K., Sampson, M., Crook, N., Fendorf, S., 2009. Time-lapse geophysical imaging of soil moisture dynamics in tropical deltaic soils: an aid to interpreting hydrological and geochemical processes. Water Resour. Res. 45, W00D32. https://doi.org/10.1029/2008WR006984. Rossi, M., Manoli, G., Pasetto, D., Deiana, R., Ferraris, S., Strobbia, C., Putti, M., Cassiani, G., 2015. Coupled inverse modeling of a controlled irrigation experiment using multiple hydro-geophysical data. Adv. Water Resour. 82, 150–165. https://doi.org/10.1016/j.advwatres.2015.03.008. Rubin, Y., Hubbard, S.S. (Eds.), 2005. Hydrogeophysics. Springer, Dordrecht, p. 523.  sur la prospection electrique du sous-sol. Gauthier-Villars, Paris. Schlumberger, C., 1920. Etude Schmalholz, J., Stoffregen, H., Kemna, A., Yaramanci, U., 2004. Imaging of water content distributions inside a lysimeter using GPR tomography. Vadose Zone J. 3, 1106–1115. Seigel, H.O., 1959. Mathematical formulation and type curves for induced polarization. Geophysics 24 (3), 547–565. Singha, K., Gorelick, S.M., 2005. Saline tracer visualized with three-dimensional electrical resistivity tomography: field-scale spatial moment analysis. Water Resour. Res. 41. Slater, L., Binley, A., Brown, D., 1997. Electrical imaging of fractures using ground-water salinity change. Groundwater 35, 436–442. Slater, L., Binley, A.M., Daily, W., Johnson, R., 2000. Cross-hole electrical imaging of a controlled saline tracer injection. J. Appl. Geophys. 44, 85–102. Strobbia, C., Cassiani, G., 2007. Multi layer GPR guided waves in shallow soil layers for the estimation of soil water content. Geophysics 72, 17–29. https://doi.org/10.1190/1.2716374. Strobbia, C., Foti, S., 2006. Multi-offset phase analysis of surface wave data (mopa). J. Appl. Geophys. 59, 300–313. Sumner, J.S., 1976. Principles of Induced Polarisation for Geophysical Exploration. Elsevier, Amsterdam. Telford, W.M., Geldart, L.P., Sheriff, R.E., 1990. Applied Geophysics, Second ed. Cambridge University Press, pp. 645–700. Topp, G.C., Davis, J.L., Annan, A.P., 1980. Electromagnetic determina¬tion of soil water content: measurements in coaxial transmission lines. Water Resour. Res. 16, 574–582. van Overmeeren, R.A., Sariowan, S.V., Gehrels, J.C., 1997. Ground penetrating radar for determining volumetric soil water content: results of comparative measurements at two test sites. J. Hydrol. 197, 316–338. Vanella, D., Cassiani, G., Busato, L., Boaga, J., Barbagallo, S., Binley, A., Consoli, S., 2018. Use of small scale electrical resistivity tomography to identify soil-root interactions during deficit irrigation. J. Hydrol. 556, 310–324. https:// doi.org/10.1016/j.jhydrol.2017.11.025. Vereecken, H., Binley, A., Cassiani, G., Kharkhordin, I., Revil, A., Titov, K. (Eds.), 2006. Applied Hydrogeophysics. Springer-Verlag, Berlin, p. 372. Viezzoli, A., Christiansen, A.V., Auken, E., Sørensen, K., 2008. Quasi-3d modeling of airborne tem data by spatially constrained inversion. Geophysics 73, F105–F113. https://doi.org/10.1190/1.2895521. Vignoli, G., Cassiani, G., 2010. Identification of lateral discontinuities via multi-offset phase analysis of surface wave data. Geophys. Prospect. 58, 389413. Vignoli, G., Cassiani, G., Rossi, M., Deiana, R., Boaga, J., Fabbri, P., 2012. Geophysical characterization of a small pre-alpine catchment. J. Appl. Geophys. 80, 32–42. https://doi.org/10.1016/j.jappgeo.2012.01.007. Vignoli, G., Gervasio, I., Brancatelli, G., Boaga, J., Della Vedova, B., Cassiani, G., 2016. Frequency-dependent multioffset phase analysis of surface waves: an example of high resolution characterization of a riparian aquifer. Geophys. Prospect. 64 (1), 102–111. https://doi.org/10.1111/1365-2478.12256. Vignoli, G., Strobbia, C., Cassiani, G., Vermeer, P., 2011. Statistical multi-offset phase analysis (sMOPA) for surface wave processing in laterally varying media. Geophysics 76, U1–U11. https://doi.org/10.1190/1.3542076.

Further reading

89

von Hebel, C., Matveeva, M., Verweij, E., Rademske, P., Kaufmann, M.S., Brogi, C., Vereecken, H., Rascher, U., van der Kruk, J., 2018. Understanding soil and plant interaction by combining ground-based quantitative electromagnetic induction and airborne Hyperspectral data. Geophys. Res. Lett. 45 (15), 7571–7579. https://doi.org/ 10.1029/2018GL07865. Weill, S., Altissimo, M., Cassiani, G., Deiana, R., Marani, M., Putti, M., 2013. Saturated area dynamics and streamflow generation from coupled surface–subsurface simulations and field observations. Adv. Water Resour. 59, 196–208. https://doi.org/10.1016/j.advwatres.2013.06.007. € 2001. Seismic Data Analysis. Society of Exploration Geophysicists. ISBN: 1-56080-094-1. Yilmaz, O., Zhang, J., Brink, U.S., Toks€ oz, M.N., 1998. Nonlinear refraction and reflection travel time tomography. J. Geophys. Res. 103 (B12), 29743–29757. Zonge, K.L., Hughes, L.J., 1991. In: Nabighian, M.N. (Ed.), Controlled Source Audio-Frequency Magnetotellurics, Electromagnetic Methods in Applied Geophysics. Vol. 2, Society of Exploration Geophysicists, pp. 713–809.

Further reading Binley, A.M., 2018. Profiler/R2t Codes. www.es.lancs.ac.uk/people/amb/Freeware. Brovelli, A., Cassiani, G., 2011. Combined estimation of effective electrical conductivity and permittivity for soil monitoring. Water Resour. Res.. 47. https://doi.org/10.1029/2011WR010487.

This page intentionally left blank

C H A P T E R

4 Topographic data from satellites Simon M. Mudd University of Edinburgh, School of GeoSciences, Edinburgh, United Kingdom

O U T L I N E 1 The importance of topography

91

2 Collection of topographic data from satellites 2.1 Satellite lidar 2.2 Radar 2.3 Stereo imaging

92 94 95 96

3 Global and large regional datasets 3.1 GTOPO30 3.2 SRTM 3.3 ASTER 3.4 ALOS PRISM 3.5 TanDEM-X 3.6 ArcticDEM and REMA 3.7 High Himalaya DEM 3.8 MERIT DEM 3.9 Other instruments and summary

4 Accuracy of global datasets 4.1 Common sources of error 4.2 Methods of comparison between datasets 4.3 Error estimates for specific datasets 4.4 Dataset intercomparison 4.5 Summary of vertical accuracy 5 Implications of increasing resolution on geomorphic studies 5.1 Geomorphic metrics and data processing 5.2 Simple preprocessing 5.3 Accuracy of channel profiles

97 97 97 99 100 100 101 101 102 103

105 105 107 107 110 111 113 113 113 118

6 Future developments

118

7 Conclusions

119

References

120

1 The importance of topography Topographic information is fundamental to the science of geomorphology. Modern geomorphology covers topics such as sediment transport, mountain building, the effects of climate on planetary surfaces, the effect of life on Earth’s surface, and soil formation.

Developments in Earth Surface Processes, Volume 23 https://doi.org/10.1016/B978-0-444-64177-9.00004-7

91

© 2020 Elsevier B.V. All rights reserved.

92

4. Topographic data from satellites

For geomorphologists, the overall theme connecting these topics is the desire to understand how the shape of the Earth or other planetary bodies can be used to infer all the processes, and the time series of these processes, which has led to their present shape. Early pioneers in geomorphology argued for landscape formation to understand how the surface of the Earth might evolve. For example, Playfair (1802) observed that tributary rivers had gradients adjusted so that they joined trunk channels smoothly, and inferred from this that rivers eroded their own valleys. Gilbert (1877) proposed that channel gradient would scale with erosion in “more than simple ratio,” which would have a topographic outcome, and later proposed that soil creep would lead to convex hilltops (Gilbert, 1909). Early pioneers used painstaking measurements derived from both fieldwork and extraction of contour maps to explore the relationship between their observations of erosion in the field and the distribution of topographic metrics (e.g., Horton, 1945; Schumm, 1956; Hack, 1957; Morisawa, 1962). This frustrating situation, where topographic data needed to be hand curated, lasted through much of the 20th century. If a geomorphologist wanted to explore, say, the relationship between channel gradient and drainage area, they needed to extract these metrics from topographic contour maps using rulers and more prosaic equipment such as planimeters. Digital information would change this situation in the second half of the 20th century. The first digital elevation models (DEMs) emerged in the 1970s and 1980s. These allowed geomorphologists to use computers to measure features such as gradients, drainage areas, curvatures, and other topographic metrics (e.g., Peucker and Douglas, 1975; O’Callaghan and Mark, 1984; Zevenbergen and Thorne, 1987). Extraction of stream profiles and topographic metrics could be done using computers rather than with pen and paper. Instead of quantifying relationships in very small areas, such as the badland landscape in Perth Amboy, New Jersey (Schumm, 1956), the possibility of quantifying topographic metrics for entire regions became computationally possible, if only the topographic data were available. In the 1980s and 1990s, a number of countries, such as the United States, Australia, Italy, Switzerland, and the UK rushed to turn their contour maps into digital elevation data. At the same time, space agencies had become alerted to the possibility of collecting elevation data of the entire planet. This dream became a reality with the Shuttle Radar Topography Mission, when radar data, which could be used to calculate global topography, was collected over 11 days in February of 2000. The availability of satellite-derived elevation datasets has expanded rapidly in the last 3 decades. The current state of global elevation data would surely have delighted G.K. Gilbert and contemporaries. If given a time machine to transport them to the present, one could only imagine that they would spend the first year rapturously tweeting about the merits of datasets we now use routinely. This chapter aims to explore methods used to collect large topographic datasets from space, and I hope to help geomorphologists make an informed choice of topographic dataset for their applications.

2 Collection of topographic data from satellites Digital topographic data has existed for many decades, created using techniques such as photogrammetry of stereo images taken from aircraft. Why resort to launching an instrument into space, at great expense, when cameras or even lasers mounted on an aircraft would do?

2 Collection of topographic data from satellites

93

The main motivation for spaceborne instruments is simple: data collection from planetary orbits can result in spatial coverage that would take many years or even decades from airborne or terrestrial instruments. There are a number of instruments that can be mounted on orbiting platforms which produce data used to compute topography. The three dominant categories are spaceborne lidar, optical data, and radar data. Spaceborne lidar currently delivers surface elevations along linear paths, and missions are typically used to detect changes in ice surfaces and tree canopies. There may be potential for more widespread coverage of spaceborne lidar (see Section 6) but for the moment global topographic datasets are dominated by elevations derived from processing radar and optical images. Most satellite-derived topographic data is delivered as digital elevation models (DEMs) in raster format. The term DEM is a general name for topographic information; it can describe data exclusively of the bare Earth surface, but it can also describe data that includes vegetation and built structures. In some cases, the more granular terms digital terrain model (DTM), which is a bare Earth DEM, and digital surface model (DSM), which includes vegetation and built environment, are used (Fig. 1). Although the term DEM has been used frequently to describe global and regional datasets, most global datasets are derived from methods that cannot detect the ground surface through buildings and vegetation and are thus DSMs. The last 2 decades have seen the release of multiple digital surface models (DSMs) that now cover the vast majority of the terrestrial surface. For applications that require accuracy on the

FIG. 1

Two hillshades and a cartoon showing the difference between a DTM and a DSM. The hillshades are derived from minimum (DTM) and maximum (DSM) elevations within 5 m pixels of a point cloud near Klamath Oregon, USA. Data from https://opentopography.org.

94

4. Topographic data from satellites

order of meters for the bare Earth surface, such as numerical modeling of flooding (Hawker et al., 2018), we are still some years away from having high-resolution (better than 10 m) global DTMs (Schumann and Bates, 2018). However, for geomorphic applications, particularly in high-relief landscapes, the rapid increase in both the availability and resolution of DSM data has been found enormously beneficial. I begin with a review of the various methods to create DSM data from orbiting instruments.

2.1 Satellite lidar The collection of light detection and ranging (lidar) data has vastly improved our ability to measure the surface of the Earth and the vegetation canopy; indeed several chapters of this volume are dedicated to lidar data collection. Lidar is an active remote sensing method; it fires many light pulses and detects the returning photons. The return of these photons can be recorded as a full waveform, or algorithms can be applied to extract discrete points, for example, the ground surface or points in the canopy, from the waveform data (e.g., Glennie et al., 2013). Terrestrial and airborne lidar systems typically have centimeter scale accuracy (e.g., Ahokas et al., 2003; Slatton et al., 2007; Jaboyedoff et al., 2012) and given enough returns per square meter or beam sensitivity (a combination of light intensity and background noise) and this method can generate data that penetrates the plant canopy (e.g., Wulder et al., 2012). Airborne or terrestrial lidar can detect the ground surface by processing last returns (e.g., Hodgson and Bresnahan, 2004; Liu, 2008), as can spaceborne lidar in some instances (e.g., Wulder et al., 2012; Qi and Dubayah, 2016). The typical footprint of a spaceborne lidar sensor is relatively narrow, so these instruments return data along linear tracks. For example, ICESat, launched in 2003, had a footprint diameter of 65–90 m (Zwally et al., 2002) and was used to monitor changes in ice sheet elevation, although the data can also be used for remote sensing of vegetation (e.g., Lefsky et al., 2005). ICESat-2, launched September 2018, has a footprint diameter of 17 m (Markus et al., 2017). In addition to ICESat and ICESat-2, NASA’s Global Ecosystem Dynamics Investigation (GEDI) mission, launched December 2018, carries a lidar instrument intended to map global vegetation structure and has a 19–25 m footprint (Hancock et al., 2019). Because these instruments collect topographic data in ground tracks rather than full coverage of large land areas, they are not well suited for creating continuous DEMs required for geomorphic applications such as landscape-scale slope measurements, flow routing, or detection of terraces and floodplains. However, the accuracy of these instruments is high; ICESat-2 aims to have and accuracy of 5 cm (Markus et al., 2017) on gently sloping (less than one degree) terrain. The GEDI instrument aims to have canopy height accuracy of better than 1 m and should be able to detect the ground surface through canopy covers between 95% and 98% (Qi and Dubayah, 2016). Given the high vertical accuracy of this data in comparison with other data collection methods (see below), one application of this data is to constrain the accuracy of data products that continuous coverage but lower accuracy, for example, data derived from radar instruments (e.g., Koppe et al., 2015; Takaku et al., 2016; Howat et al., 2019). Table 1 shows the properties of these spaceborne lidar missions.

95

2 Collection of topographic data from satellites

TABLE 1

Selection of properties of various spaceborne lidar missions.

Mission

# beams

Along-track spacing

Beam spacing

Canopy penetration

References

IceSAT

1

200 m

NA

Very high

Zwally et al. (2002)

IceSAT-2

6

0.7 m

6 km

Low

Narine et al. (2019)

GEDI

8

60 m

600 m

High

Hancock et al. (2019)

IceSAT operated between 2003 and 2009. Both IceSAT-2 and GEDI were launched in 2018. One thing to note is that ICESat and GEDI have sufficient beam sensitivity to detect the ground surface through dense vegetation (e.g., Lefsky et al., 2005; Hancock et al., 2019), but IceSAT-2 uses photon counting detection which results in denser coverage but less canopy penetration (Narine et al., 2019).

2.2 Radar Like lidar, radar is an active remote-sensing technique that emits pulses of energy which are then returned to, and recorded by, a sensor. In fact, the theory underlying lidar measurement is derived from the more general radar equations (Mallet and Bretar, 2009). Radar instruments use a longer wavelength than lidar instruments. Topographic data is typically derived from the X and C radar bands, which range from 2.5 to 7.5 cm (Skolnik, 1981; Raney et al., 1991; Farr et al., 2007; Krieger et al., 2007), although the L band (15–30 cm wavelength) has also been used (Rosenqvist et al., 2007). In contrast, topographic lidar typically uses wavelengths in the hundreds of nanometers (Slatton et al., 2007). The resolution of radar images depends, in part, on the wavelength used, with shorter wavelengths offering higher resolution (Moreira et al., 2013). Because X and C bands have relatively shorter wavelengths, they have tended to be used for global topographic mapping, including the Shuttle Radar Topography Mission (SRTM) (Farr et al., 2007) and the TanDEMX mission (Krieger et al., 2007). One disadvantage of the X and C bands is that these wavelengths have lesser penetration of the tree canopy than longer wavelengths (e.g., Balzter, 2001). The L and P band do better at penetrating the tree canopy (e.g., Imhoff et al., 1986), and so instruments using the L band may have some advantages for detecting the bare Earth surface (Chu and Lindenschmidt, 2017). Due to the lower resolution of L-band radar compared with X and C bands, there are, at present, no globally available DEMs that use L-band radar, although the ALOS PALSAR radar data could potentially be used to create such a data product in the future (Chu and Lindenschmidt, 2017). Furthermore, data gaps in radar data can be generated from shadowing, layover, and due to difficulties in specular reflectance of water bodies (Kervyn, 2001; Reuter et al., 2007). It means that any radar dataset will require a substantial amount of data processing and void filling in order to produce a continuous DEM. This can be mitigated to some degree, but not entirely, by sophisticated acquisition planning; the recent TanDEM-X mission (see sections below) used two satellites flying in a helical pattern to reduce such data gaps (Krieger et al., 2013). Whereas topographic data using lidar is typically generated by calculating distances using time of flight of light pulses, topographic data derived from radar instruments uses interferometric techniques (Zebker and Goldstein, 1986) based on synthetic aperture radar (SAR) (Graham, 1974). This technique measures a phase shift between two radar images (Zebker and Goldstein, 1986). The use of a phase shift means that only the difference in path length

96

4. Topographic data from satellites

between pixels can be measured, and after 2π radians of a cycle, it is not possible to determine absolute distance. Therefore, algorithms for computing the absolute phase are necessary, a process called phase unwrapping (Goldstein et al., 1988). The accuracy of the resulting topographic information depends on the wavelength, satellite relative separation, and the look angle of the satellites, and scales with a quantity called the altitude of ambiguity (ea) (Massonnet and Souyris, 2008): ea ¼

Rm λ sinθ 2Bo

(1)

where Rm is the observation range (i.e., the height of the satellite), λ is the wavelength, θ is the incidence angle, and Bo is the orthogonal baseline (i.e., the distance between the two antennae or satellite images). Interferometric techniques allow large swath widths; satellite instruments can collect swaths tens of kilometers in width (Farr et al., 2007; Krieger et al., 2007). This makes continuous coverage of the Earth’s surface possible, in contrast to the discrete tracks of satellitebased lidar. It should be noted that the rapid increase in radar image resolution means that it is possible to generate topographic data using photogrammetric techniques developed for optical data (Capaldo et al., 2011); this is called radargrammetry and was used to some extent by the TanDEM-X digital elevation model (Rizzoli et al., 2017). However, SRTM and much of TanDEM-X topographic data was generated using interferometry.

2.3 Stereo imaging Photogrammetric techniques can be used to generate topographic data from optical instruments that take images at different angles (e.g., Toutin, 2001). A number of satellites and satellite constellations have collected or are collecting visible and infrared imagery that is suitable for creating topographic data. Examples include SPOT (e.g., Gugan and Dowman, 1988), Pleiades (e.g., Poli et al., 2015; Bagnardi et al., 2016; Sofia et al., 2016), WorldView (e.g., Poli et al., 2015; Hobi and Ginzler, 2012), IKONOS (e.g., Zhang and Fraser, 2008), and GeoEye (e.g., Poli et al., 2015; Barbarella et al., 2017). Commercial software is available to create topographic data with these images, but recently open source tools have emerged (Noh and Howat, 2015; Shean et al., 2016; Noh and Howat, 2017; Rita et al., 2017; Rupnik et al., 2017; Noh and Howat, 2018). Many of these images are of sufficiently high resolution (better than 1 m) so that this images can be used to create topographic data compared to lidar data (e.g., Sofia et al., 2016; Barbarella et al., 2017). There are three regional, publicly available DEMs derived from high-resolution optical images. The first of these is the High Mountain Asia DEM, which has 8-m resolution and has been compiled from GeoEye, Quickbird, and WorldView satellites (Shean, 2017). This dataset focuses on the greater Himalayan region. The ArcticDEM, which is composed of 2-m resolution data for the entirety of land north of 60 degrees north, has been compiled using primarily WorldView images, with a small number of GeoEye images (Porter et al., 2018). Finally, the Reference Elevation Model of Antarctica (REMA) is an 8-m resolution DEM of Antarctica, constructed from Worldview and GeoEye satellites (Howat et al., 2018). Lower-resolution instruments have been used to generate topographic data at a nearly global scale. The most notable, perhaps, is the Advanced Spaceborne Thermal Emission

3 Global and large regional datasets

97

and Reflection Radiometer (ASTER) (Abrams, 2000). ASTER produced hundreds of thousands of multispectral images. The highest resolution pixels from these images have 0.5-arcsecond (15-m) resolution in the visible and near infrared spectra (Abrams, 2000). These images have been used to generate global topographic information at 1 arcsecond (approximately 30 m) resolution (Abrams et al., 2010). The ASTER satellite sensor was launched in December 1999. In January 2006, the Advanced Land Observation Satellite (ALOS) was launched (Rosenqvist et al., 2007). This satellite included a number of sensors, including L-band radar (PALSAR) and the Panchromatic Remote-sensing Instrument (PRISM), which is an optical instrument with 2.5-m resolution (Rosenqvist et al., 2007). The PRISM instrument contains three independent optical system giving different views, which can then be used to create high-resolution (as high as 5 m) digital elevation models (Takaku et al., 2004).

3 Global and large regional datasets The techniques outlined in the previous section have led to an ever expanding menu of global topographic datasets. Due to improvements in instrumentation, the accuracy and resolution of these datasets have tended to increase through time. In this section, I give an brief overview of several global datasets that are widely used for geomorphic studies, and three notable datasets that cover large regions (both the Arctic and Antarctic as well as High Mountain Asia). An assessment of the relative accuracy and comparison of the strengths and weaknesses of these datasets is discussed in Section 4.

3.1 GTOPO30 The first dataset I highlight is GTOPO30 (Gesch et al., 1999), which is not derived from satellite imagery, but rather it is an amalgamation of data from multiple sources. Some of these sources are contour maps and national digital elevation data derived from aerial photogrammetry, in addition to surveying sources (Gesch and Larson, 1998). GTOPO30 (which is shortened from global topography 30 arcsecond) was not the first global topographic dataset but it had an order of magnitude higher resolution than its precursor ETOPO5 (Bamber et al., 1997). This meant that, arguably, it was the first global dataset in which large river valleys were visible (e.g., Verdin and Verdin, 1999; Wolf et al., 1999). Prior to ETOPO5 (Bamber et al., 1997) and GTOPO30 (Gesch et al., 1999), there were no truly global topographic datasets. The GTOPO30 dataset, with a resolution of approximately 1-km, was the highest resolution global dataset until 2005. This was about to radically change with the Shuttle Radar Topography Mission (SRTM).

3.2 SRTM The Shuttle Radar Topography Mission (SRTM) was revolutionary. The data were collected from both X- and C-band radar using antennae mounted on both the space shuttle Endeavor and a 60-m mast extending from the shuttle (Rabus et al., 2003). The data were collected between the 11th and 22nd of February, 2000 (Farr et al., 2007). The Jet Propulsion

98

4. Topographic data from satellites

FIG. 2 Hillshades of an area along the North Anatolian Fault in Turkey (see inset map). Top panel shows GTOPO30 30-arcsecond resolution data whereas bottom panel shows SRTM 3-arcsecond resolution data. Note that several fault traces are visible in the 3-arcsecond data that are either not visible or highly blurred in the 30-arcsecond data.

Laboratory in Pasadena, California, processed the C-band data and these data form the basis of the near global DEM that is currently available for download (Farr et al., 2007). The SRTM global DEM was first released in 2005 and covered the land surface between latitudes of 60 degrees North and 58 degrees South (Farr et al., 2007). The 2005 release was at 3-arcsecond resolution, or approximately 90-m resolution near the equator (Farr et al., 2007). The contrast between topographic features visible at 1-km and 90-m grid spacing is striking (Fig. 2). It is difficult to overstate the improvement in geomorphologists’ ability to characterize landscapes when moving from 1-km-to-90-m resolution. The SRTM 90-m data product allowed extraction or visualization of river profiles, terraces, floodplains, and many other large-scale geomorphic features over much of the Earth’s surface

3 Global and large regional datasets

99

(compare the visible features as shown in Fig. 2). At the time of its release, there were higher resolution topographic datasets for individual countries including the United States (Gesch et al., 2002) and the United Kingdom (Ordnance Survey Great Britain, 2001). However, for most countries and regions, SRTM data enabled previously impossible topographic analysis. SRTM data is unique compared with other datasets in this chapter because it represents a snapshot of the Earth’s surface in early 2000. Other datasets covered here use stacks of images collected over many years. The relatively short time span of data collection does mean that SRTM data has numerous data voids, especially in mountainous terrain (see fig. 1 in Reuter et al., 2007). Voids have been filled in SRTM version 3 and above (NASA JPL, 2013) using the method of Grohman et al. (2006) alongside data from various other sources such as the United States National Elevation Dataset (Gesch et al., 2002), the Global Multi-resolution Terrain Elevation Data (GMTED) 2010 (Danielson and Gesch, 2011), which is the successor to GTOPO30, and the ASTER global DEM (Tachikawa et al., 2011), which I describe in the following section. A separate effort by the Consortium for Spatial Information (CGIAR) produced a hole filled 90-m SRTM data product called the SRTM 4.1 product using methods described by Reuter et al. (2007) that is available via the CGIAR website. In 2015, SRTM publicly released an enhanced 1-arcsecond dataset (NASA JPL, 2013) which is now widely available on data distribution websites, replacing the 3-arcsecond global data that has been available since 2005. SRTM is currently released on data distribution websites such as OpenTopography (https://opentopography.org/) and the NASA Earthdata (https:// earthdata.nasa.gov/) as SRTM version 3. However, a newer version of this data, called NASADEM, is being processed (Crippen et al., 2016) and the provisional data is available via the United States Geological survey (https://e4ftl01.cr.usgs.gov/provisional/ MEaSUREs/NASADEM/, last access February 21, 2019).

3.3 ASTER The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) was launched in 1999. The ASTER instrument has 14 spectral bands, including three visible and near-infrared (VNIR) bands with a spatial resolution of 15 m, six short-wave-infrared (SWIR) bands with a spatial resolution of 30 m, and five thermal infrared (TIR) bands that have a spatial resolution of 90 m (Abrams, 2000). Because the instrument’s highest resolution bands are in the VNIR range, in which the ground surface can be obscured by clouds, production of the DEM involved stacking numerous images to obtain cloud-free scenes (Tachikawa et al., 2011). Thus, the ASTER global DEM (GDEM) is not synoptic, but rather represents an average of many years of data collection. Version 1 is from images collected between 2000 and 2008, whereas version 2 is from images collected between 2000 and 2011 (Tachikawa et al., 2011). When the ASTER GDEM was first released in 2009 (Abrams et al., 2010), the SRTM DEM was only available at 3-arcsecond resolution outside of the United States. The ASTER GDEM has always been available at 1-arcsecond resolution (Abrams et al., 2010) and was thus the first satellite dataset covering a significant fraction of the Earth’s surface at this resolution. In addition, the ASTER GDEM extends to 83 degrees North and South (Tachikawa et al., 2011), in contrast to SRTM which covers the terrestrial surface between

100

4. Topographic data from satellites

60 degrees North and 58 degrees South. The vertical accuracy of the ASTER GDEM is generally lower than that of SRTM (e.g., Hirt et al., 2010; Gesch et al., 2016); I discuss this in more detail in Section 4.

3.4 ALOS PRISM The Advanced Land Observing Satellite (ALOS) operated both optical and radar instruments; the instrument had an intended operation time of 5 years (Shimada et al., 2010) and operated between January 2006 and May 2011 (Tadono et al., 2014). The Japanese Aerospace Exploration Agency (JAXA) has used the Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) to generate high-resolution topographic maps at approximately 5-m (0.15arcsecond) resolution (Tadono et al., 2014). The dataset released at this resolution is called the ALOS World 3D (AW3D) dataset. Note that JAXA routinely refers to their data product as a DSM, meaning that the product includes structures and vegetation. The AW3D dataset is proprietary but it serves as the basis for publicly available lower resolution datasets. The AW3D data are derived from stacking fine resolution (0.075 arcsecond) ortho-rectified images (Tadono et al., 2015). Initial data processing was completed in 2016 (Takaku and Tadono, 2017) covering the areas between 82 degrees North and South latitude, with void filling from existing topographic datasets, primarily SRTM (Takaku et al., 2014). In parallel with the processing of the AW3D 5-m resolution dataset, JAXA has also processed a lower resolution 30-m dataset (AW3D30) which was first released in 2016 (Tadono et al., 2016). The AW3D30 has been released free of charge and like its higher resolution cousin has featured several accuracy improvements since the initial 2016 release, including reduction in void pixels due to cloud cover (Tadono et al., 2017). Version 2.1 was released in April 2018 which includes enhanced data calibration (Takaku et al., 2018) and assessment of errors by comparing with ICESat data. In order to produce the 30-m resolution DSM, the average value in a 7  7 pixel window is used and in addition a median value is also used, producing two DSMs (Tadono et al., 2016). Nearest-neighbor resampling is not used; Tadono et al. (2016) suggest this is because it would conflict with the commercial distribution of the AW3D dataset.

3.5 TanDEM-X The TanDEM-X mission was launched in 2010; this satellite was a twin of the 2007 TerraSAR-X X-band radar satellite (Krieger et al., 2007). The project design, developed as a collaboration between the German Aerospace Center (DLR) and a private company, aimed to improve upon the Shuttle Radar Topography mission by extending coverage to higher latitudes and deliver data at higher resolution (Krieger et al., 2007). The twin satellites fly in a helical pattern (Krieger et al., 2013), creating a baseline that changes with time. This configuration was selected because it was best to allow the complete mapping of the Earth at the desired accuracy; see Krieger et al. (2007, 2013). For details on the data processing, the reader is referred to Fritz et al. (2011) and Rossi et al. (2012). The resulting digital elevation model has resolution of 0.4 arcseconds (approximately 12 m at the equator). Like many recent topographic datasets, the reported accuracy metrics have

3 Global and large regional datasets

101

been generated with comparison to ICESat and other ground-based methods (Riegler et al., 2015). The data used in creating the TanDEM-X DEM have been generated using a number of scenes collected between December 2010 and 2015, with all landmasses being imaged at least twice and difficult terrain being imaged more frequently, up to seven or eight times (Rizzoli et al., 2017). The resulting dataset bears resemblance to SRTM, in that it is a radar dataset, but also to ASTER and ALOS AW3D in that the dataset is a composite of numerous scenes.

3.6 ArcticDEM and REMA The ArcticDEM (Porter et al., 2018) and the Reference Elevation Model of Antarctica (REMA) (Howat et al., 2019) aimed to fill data gaps at high latitudes. These were derived from optical satellites, primarily the WorldView-1, WorldView-2, and WorldView-3 satellites, with some contributions from GeoEye-1 and other satellites. Because vegetation is minimal at these high latitudes, very high-resolution DEMs can be produced. The ArcticDEM has mosaiced DEMs at 2-m resolution (Fig. 3), whereas REMA is mosaiced to 8-m resolution. These DEMs are also available as time-sliced strips, in support of studies into changing snow and ice thicknesses, and mosaiced data as well as slices are available from the Polar Geospatial Center (PGC; https://www.pgc.umn.edu/). Both ArcticDEM and REMA have been assembled using the Surface Extraction from TIN-based Searchspace Minimization stereophotogrammetry software (SETSM) (Noh and Howat, 2015, 2017, 2018). The original algorithm (Noh and Howat, 2015) was tested against 20 sites with lidar elevation points from Operation Icebridge (Studinger et al., 2010) using a coregistration method that compares the closest point in the two datasets (Noh and Howat, 2015). These testing DEMs included sites in both high- and low-relief terrain and found RMSE of 3.8 and 2.0 m in the horizontal and vertical direction, respectively (Noh and Howat, 2015). Coregistration using simple translation reduced this error to 0.2 m (Noh and Howat, 2015). These two datasets use images starting in 2007 and continuing to the present. Strip data is delivered as discrete time slices (see https://www.pgc.umn.edu/) but due to clouds and other factors strips do not contain continuous data, whereas mosaics represent numerous images to generate near continuous topography.

3.7 High Himalaya DEM Similar to the ArcticDEM and REMA, a research-led effort has computed high-resolution topographic data from the Himalayan mountain belt using optical imagery. The DEMs were released as the High Mountain Asia topographic dataset in 2017 (Shean, 2017). The dataset includes both time-stamped topographic data and more spatially complete mosaics and is available from the National Snow and Ice Data Center (NSIDC; https://nsidc.org/ data/highmountainasia). The data was collected using WorldView-1, WorldView-2, WorldView-3, GeoEye-1, GeoEye-2, and Quickbird images collected between 2002 and 2016 (Shean, 2017). Data was processed using an open source method (Shean et al., 2016) that has been tested in sites in Antarctica and Greenland against various datasets, including airborne lidar and is estimated to have, after coregistration, errors of between 0.1 and 0.5 m

102

4. Topographic data from satellites

FIG. 3

Area surrounding the Zhupanovsky Volcano, Kamchatka, Russia. Data is projected into UTM zone 57, referenced to the WGS84 ellipsoid.

(Shean et al., 2016), similar to the method of Noh and Howat (2015). The dataset contains many data gaps, but where data is available the 8-m resolution is better than more continuous, publicly available datasets such as ALOS AW3D30 and SRTM 1 arcsecond data.

3.8 MERIT DEM Another recent entry into the list of freely available, near-global topographic datasets is the MERIT DEM (Yamazaki et al., 2017). This DEM is derived from a combination of SRTM version 3 data and for high-latitudes ALOS AW3D30 data; the dataset covers the globe between 82 degrees North and South latitude. This dataset is of the same resolution as the early SRTM

3 Global and large regional datasets

103

FIG. 4 Comparison of the MERIT DEM and SRTM version 3 DEM at the confluence of the Ubangi and Congo Rivers, Democratic Republic of the Congo (labeled DRC on inset map). Because SRTM is a DSM, many midchannel islands and adjacent floodplains are recorded as lying several meters above the level of the river, whereas in the MERIT DEM, this bias is removed.

data: 3 arcseconds. The novelty of this data, however, is that it uses multiple processing steps to remove errors, and in addition attempts to remove vegetation (Yamazaki et al., 2017). The MERIT DEM is thus the first large-scale DEM that moves toward a true DTM rather than a DSM and can resolve more realistic floodplain topography (Fig. 4). All other datasets described in this section use either radar or optical data as their sources including DSMs.

3.9 Other instruments and summary Several other notable large DEM datasets are available, although for reasons explained below I do not focus on these for the remaining chapter. The Indian Space Agency has

104

4. Topographic data from satellites

TABLE 2 A selection of global and large regional topographic datasets. Dataset name

Method

Resolution

References

GTOPO30

Various

30 arcsecond

Gesch et al. (1999)

SRTM

C band radar

3 (2004) and 1 arcsecond (2015)

Farr et al. (2007)

ASTER

Multispectral optical

1 arcsecond

Tachikawa et al. (2011)

ALOS AW3D

Multispectral optical

0.15 arcsecond

Tadono et al. (2014)

ALOS AW3D30

Multispectral optical

1 arcsecond

Tadono et al. (2016)

TanDEM-X DEM

X band radar

0.4 arcsecond

Krieger et al. (2013)

ArcticDEM

Optical

2m

Porter et al. (2018)

REMA

Optical

8m

Howat et al. (2018)

High Himalaya DEM

Optical

8m

Shean (2017)

MERIT DEM

Various

3 arcsecond

Yamazaki et al. (2017)

All but the ALOS AW3D and TanDEM-X DEM datasets are available free of charge to scientists through various web portals. The TanDEM-X DEM has seen widespread use in the scientific community as data is released at no cost via a proposal process. The AW3D dataset is commercial but serves as the basis for the AW3D30, which is available free of charge.

produced a 30-m resolution DEM over the Indian subcontinent using the CartoSAT satellites, which are a constellation of optical satellites (Muralikrishnan et al., 2013). The accuracy of these data is similar to ASTER, and less accurate than SRTM at the same resolution (Jain et al., 2018), although in some locations the CartoDEM outperforms these datasets (e.g., Muralikrishnan et al., 2013). The CartoDEM has less spatial coverage than SRTM, however. Another global-scale DEM product is available through the Alaska Satellite Facility (ASF); it is a 12.5-m DEM used to terrain correct ALOS PALSAR radar tiles (ASF Engineering, 2015). This DEM has not been tested for vertical accuracy and for most of the world is a resampled version of SRTM which then uses software from ASF to smooth errors (ASF Engineering, 2015). The data product is referenced to the Earth ellipsoid rather than the geoid. The geoid is the theoretical sea level at any point on the Earth’s surface considering gravitational variation and can vary from the ellipsoid by several hundred meters (Snyder, 1987). Topographic data is almost always delivered with elevations referenced to the geoid, and to compare the ALOS PALSAR data product to other data products such as SRTM, a conversion from the ellipsoid is required. Because this dataset is derived primarily from SRTM and has not been tested for vertical accuracy, I do not include it in further discussion. Table 2 lists the primary data sources and DEM products explored in the remaining chapter. Some users of satellite datasets might use these for detection of land surface changes (e.g., Purinton and Bookhagen, 2018) and so Table 3 lists the dates of data collection of these large regional and global datasets.

105

4 Accuracy of global datasets

TABLE 3 The time evolution of both data capture and availability of global and large regional datasets. Dataset name

Year(s) of data collection

Year of initial release

GTOPO30

Various

1996

SRTM

2000

2004

ASTER

2000–10

2009

ALOS AW3D

2006–11

2016

ALOS AW3D30

2006–11

2016

TanDEM-X (GDEM)

2011–16

2016

ArcticDEM

2007–ongoing

2018

REMA

2007–ongoing

2018

High Himalaya DEM

2002–16

2017

4 Accuracy of global datasets With so many global-scale topographic datasets now publicly available, one may struggle to identify the dataset most suited to a particular study. A variety of factors may influence a researcher’s decision, including the size of topographic data, the ease of obtaining the data, the time at which the data was collected, and the accuracy of the data. A number of studies have investigated the relative accuracy of various satellite data products listed above. Here I attempt to summarize the findings of these studies.

4.1 Common sources of error DEMs derived from spaceborne instruments suffer from several sources of error (Fisher and Tate, 2006), leading to digital surfaces that contain artifacts (Fig. 5). Yamazaki et al. (2017) separated these into four categories, partly defined by their spatial scale. First, speckle noise refers to random noise at the pixel scale which is common in satellite-derived datasets (e.g., Maire et al., 2003; Rodrı´guez et al., 2006; Stevenson et al., 2010). On a longer wavelength, errors such as those derived from uncertainty in the positioning of the satellite, the exact angle of the satellite at the time of data collection, and in the case of radar instruments, phase errors (Rodrı´guez et al., 2006) can lead to longer wavelength errors; Yamazaki et al. (2017) identifies both 500 m and 50 km stripe noise and long wavelength (>50 km) absolute biases. Finally, because satellite-based instruments in general cannot penetrate the tree canopy, there is a systematic positive bias in the elevation compared to bare earth topography (O’Loughlin et al., 2016). There is some optimism that secondary datasets can be used to subtract canopy height from digital surface model (DSM) data delivered by radar instruments (e.g., Kugler et al., 2014; Schlund et al., 2019), but these efforts have not yet resulted in global scale DEMs that are canopy free, with the notable exception of the MERIT DEM (Yamazaki et al., 2017).

106

4. Topographic data from satellites

FIG. 5 Comparison of hillshades from the High Mountain Asia DEM, the ALOS AW3D30 DEM, and the SRTM 1 arcsecond DEM. Note the spurious roughness elements in SRTM.

The measures of error quantify the differences between the DEM and the comparison dataset, which could be averaged kinematic global positioning system points, or averaged lidar data, or other datasets at the same resolution. The most popular error metrics are the root-mean-square error (RMSE), and the absolute 90% linear error, which is the absolute deviation of the data from the comparison dataset at the 90% quantile (e.g., Wessel et al., 2018). The 90% linear error is sometimes abbreviated LE90 in accuracy reporting.

4 Accuracy of global datasets

107

4.2 Methods of comparison between datasets The accuracy of satellite-derived topographic data can be measured in a variety of ways. These can be broken into two broad categories: comparison with remotely sensed data and comparison with ground control points. There are several options for comparing remotely sensed data. Comparative datasets could include other satellite-derived data products or data from ground or airborne instruments. In addition, a number of the datasets described in the previous sections are achieved by stacking several layers of topographic data, acquired at different times. The relative stability of elevations between two images or derived DEMs can be compared. The advantage of methods that use remote sensing, particularly with freely available topographic data, is that they do not require expensive fieldwork. In addition, regional coverage can be achieved if two raster datasets are compared on a pixel by pixel basis. This method of determining topographic data accuracy precludes direct comparison between a topographic dataset and the real surface of the Earth, relying instead on statistical comparison between topographic datasets. A different remote-sensing approach to DEM validation that offers a more direct comparison of the true Earth surface is to use high-precision elevation products such as ICESat (Zwally et al., 2002). For flat areas, the accuracy of the ICESat altimeter is approximately 14 cm (Shuman et al., 2006). The linear tracks of ICESat data can be curated to remove water bodies and clouds, leading to a dataset that approximates GPS Ground Control Points (GCPs) collected with GPS instruments (e.g., Carabajal et al., 2010). Similar datasets can be created from Land, Vegetation, and Ice Sensor (LVIS) data (Blair et al., 1999), which has on the order of 10 cm accuracy (Brunt et al., 2017), and slightly better accuracy for ICESat 2 (Markus et al., 2017). Alternatively, satellite-derived data products can be compared to high-precision survey data. This can include stationary geodetic networks or high accuracy surveys based on global positioning (GPS). The United States National Geodetic Survey (https://www.ngs.noaa.gov/), for example, maintains many thousands of ground control points. GPS points have been collected for supporting specific satellite missions or for regional studies of elevation accuracy (e.g., Rodrı´guez et al., 2005; Purinton and Bookhagen, 2017; Wessel et al., 2018). The latter requires a great deal of fieldwork, and thus such direct comparison is only available for a relatively small area on the surface of the Earth. The datasets described in Section 3 were originally published with error assessments. A number of satellite-derived datasets compare their results with existing digital elevation data that has been obtained using other methods (e.g., aerial photogrammetry, surveying, and digitization of preexisting contour maps). One example is the United States National Elevation Dataset (Gesch et al., 2002), which has evolved through time to take on new sources, but was compared with SRTM (Smith and Sandwell, 2003) and ASTER data (Tachikawa et al., 2011) prior to widespread adoption of lidar data into the NED.

4.3 Error estimates for specific datasets 4.3.1 SRTM accuracy The SRTM dataset was assessed using hundreds of thousands of ground control points obtained by Kinematic Global Positioning (KGPS) (Rodrı´guez et al., 2005).

108

4. Topographic data from satellites

These were distributed globally across KGPS “tracks”; there were 6 such tracks in North America, 5 in South America, 4 in Africa, 11 in Eurasia, 4 in Australia, and 2 in New Zealand (Rodrı´guez et al., 2005). Data points along these tracks were averaged across SRTM pixel sizes, resulting in just over 2 million ground control points (Rodrı´guez et al., 2005, 2006). This effort reported that across all continents, the 90% absolute linear errors for SRTM were between 5 and 9 m, and 90% geolocation errors between 9 and 12 m (Rodrı´guez et al., 2006). Over ice and snow these errors can be somewhat larger as the C-band radar can penetrate these materials; Rignot et al. (2001) found firn penetration in Greenland of 92 and 12 m on the exposed ice surface. They also found penetration of temperate ice for the C band at Brady Glacier, Alaska at 42 m. As I explain below, this perhaps gives a more optimistic view of the errors than is perhaps warranted, as the greater 10% errors may be concentrated in mountainous regions, where many geomorphologists conduct their research. The new NASADEM product, derived from reprocessed SRTM data, has yet to undergo full-error analysis. It is, however, expected to have better accuracy than SRTM version 3 as steps have been taken to remove spikes, correct vertical tilt errors from the original data collection, and contain improved void filling routines (Crippen et al., 2016). Preliminary assessment of the NASADEM over the United States suggests the RMSE errors of 2.3 m (Simard et al., 2016). For SRTM version 3, Rodrı´guez et al. (2006) did not report RMSE error but Gesch et al. (2016) suggest that it is 4.15 m, meaning that the NASADEM should be expected to be more accurate than SRTM version 3. 4.3.2 ASTER accuracy Initial versions of the ASTER dataset were validated against a combination of ground control points measured by differential GPS (DGPS), known mapped elevations, and compared with other datasets (Hirano et al., 2003). Later efforts compared ASTER DEMs with the United States National Elevation Dataset (Gesch et al., 2002) on a pixel-by-pixel basis (Tachikawa et al., 2011), and with geodetic control point dataset from the National Geodetic Survey (NGS) that includes more than 13,000 ground control points (GCPs) distributed throughout the contiguous United States (Gesch et al., 2016). The study of Gesch et al. (2016) found that the root-mean-square error of the ASTER version 3 data product is 8.52 m and the 95% linear error is 16.7 m. SRTM was also tested by Gesch et al. (2016) with RMSE and 95% linear errors of 4.15 and 8.13 m, respectively; the SRTM version was not reported but given the date of publication it can be assumed these data are for SRTM version 3. 4.3.3 ALOS world 3D accuracy For the ALOS World 3D dataset, similar metrics were used. Takaku et al. (2016) used ICESat data points averaged over the 64-m footprint of the ICESat instrument. The AW3D dataset has 5-m pixel spacing and found the 90% linear error was on the order of 5 m (see their Table 3). Lidar and various ground control points from GPS datasets were for smaller numbers of sites; one result was that the 90% linear errors were strongly correlated with topographic slope; the 90% linear error was less than 4 m for slopes up to 20 degrees, but rising to 5.28 m for pixels with slopes between 20 and 30 degrees, and rising again to 9.20 m for topographic slopes over 30 degrees. Given that the AW3D30 dataset is derived from averaging the 5-m dataset, errors for the 30-m dataset should be slightly greater. According to the AW3D30 website (https:// www.eorc.jaxa.jp/ALOS/en/aw3d30/index.htm,lastaccess21Feb2019), version 2.1 of the

4 Accuracy of global datasets

109

dataset was released in April 2018 with higher accuracy than previous versions, although the ALOS World 3D team have not published specific accuracy metrics of this new dataset. 4.3.4 TanDEM-X DEM accuracy Topographic data generated from the TanDEM-X system was first tested by comparing the elevations derived from radar images taken at different times; Rizzoli et al. (2012) identified problems in reproducibility of elevation in high mountain terrain and suggested methods to improve accuracy for final data release. At that time, point-to-point errors were distributed about zero with 90% of errors falling within approximately 1 m in test sites. Later data products derived from the TanDEM-X mission have been tested against globally distributed lidar datasets, the ICESat mission, and ground control points (GCPs) from the CompassData archive (Koppe et al., 2015). CompassData is a private company but a truncated GCP archive can, at the time of writing, be downloaded from their website (this includes over 45 thousand ground control points). This initial study suggested differences between the TanDEM-X DEM and reference datasets of between 1 and 3.5 m (Koppe et al., 2015). It should be noted that reference areas in this study did not include data in high mountain areas; see Fig. 2 of Koppe et al. (2015). The TanDEM-X DEM has also been compared to ICESat data (Rizzoli et al., 2017). ICESat data includes quality assessments of each data point (Schutz et al., 2005) and algorithms were developed to only take the most reliable ICESat data points (Gonzalez et al., 2010). For the assessment carried out by Rizzoli et al. (2017), the planet was divided into one degree by one degree geocells between 60 degrees North and South latitude, which coarsened to 1 degree by 2 degree between 60 and 80 degrees latitude, and 1 degree by 4 degrees between 80 and 90 degrees latitude. The height accuracy with 90% error in all geocells was 0.88 m (Rizzoli et al., 2017). The TanDEM-X DEM has also been compared with a large dataset of kinematic global positioning system data, collected specifically for that purpose (Wessel et al., 2018). This included over 14 million averaged GPS tracks across all continents except Antarctica. Wessel et al. (2018) found that the TanDEM-X data had an RMSE error less than 1.4 m and 90% linear errors on the order of 2 m (ranging from 1.52 m in Asia to 2.11 m in Africa). The errors were sensitive to land cover; 90% linear errors were greater under forest cover, with LE90 errors under deciduous and evergreen forests of 3.84 and 4.37 m, respectively. Over glaciers, partial penetration of the radar signals can lead to biases in the ice surface elevation; Dehecq et al. (2016) found in the Alps these average on the order of 4 m at at 4000 m above sea level. It should be noted that the TanDEM-X DEM aims to have less than 2 m of 90% linear error for slopes of less than 20 degrees, but the design specification rises to 4 m for slopes of greater than 20 degrees (Huber et al., 2009). Similar sensitivity to topographic gradient was found by Singh et al. (2016) for both ASTER and SRTM data in the high Himalaya, with 90% linear errors rising from just under 20 m for gently sloping terrain (less than 5 degrees) and for both data to 40 and 82 m for ASTER and SRTM, respectively, at slopes between 30 and 45 degrees. 4.3.5 MERIT DEM accuracy The MERIT DEM (Yamazaki et al., 2017) is based on SRTM data but has undergone error removal; its accuracy was tested against a dataset of ICESat lowest returns (Harding and Carabajal, 2005). The errors were tested using only flat areas (slopes less than 10 degrees);

110

4. Topographic data from satellites

in these areas, the MERIT DEM has 90% linear error of 5 m for all flat areas and 9 m for flat areas with forests. This compares with 90% linear errors for the SRTM 3-arcsecond product of 7 m for all flat areas and 17 m for flat and forested areas (Yamazaki et al., 2017). Fig. 2 in Yamazaki et al. (2017) shows that MERIT DEM errors are concentrated in mountains, whereas SRTM data has large errors both in mountains but also in densely forested regions such as the Amazon basin, the central African rainforest, and south east Asia. Hirt (2018) used a maximum slope approach to screen both SRTM and MERIT DEMs for artifacts (e.g., spikes, pits and line defects) and found that the SRTM 4 product had more than 1300 artifacts whereas the MERIT DEM had 108. 4.3.6 ArcticDEM, REMA, and High Mountain Asia DEM accuracy The accuracy of regional datasets such as ArcticDEM, REMA, and the High Mountain Asia DEM has received less attention than accuracy of global datasets. In a conference abstract, Candela et al. (2017) suggested that the ArcticDEM has an average accuracy of 0.1  0.07 m when compared to ICESat data. REMA has 90% linear error of 1.78, 1.74, and 1.25 m when compared with the Airborne Topographic Mapper (ATM), the Land, Vegetation, and Ice Sensor (LVIS), and the ICECAP laser altimeter systems, and a 90% linear error of 1 m in comparison to ICESat-1 (Howat et al., 2019). The High Mountain Asia DEM does not report accuracy but the algorithm used to generate the DEM was found to have 90% linear error of 3.91 and 2.51 m at two sites in Greenland when compared with GLAS (a satellite altimeter), ATM, and LVIS (airborne lidar instruments) (Shean et al., 2016). Shean et al. (2016) also suggests that the error scales linearly with topographic slope. This finding is consistent with that of M€ uller et al. (2014); they found systematic increases in error with increasing gradient in DEMs generated using photogrammetry.

4.4 Dataset intercomparison Numerous studies have carried out intercomparison of global and regional datasets. I focus on more recent studies because the most widely used datasets have undergone several iterations. Readers of this chapter are unlikely to download old versions of, for example, SRTM or ASTER. Grohmann (2018) compared SRTM, ASTER, TanDEM-X, and AW3D30 DEMs in Brazil, using the TanDEM-X DEM as the reference DEM. Multiple regions were studied, and the LE90 differences between SRTM and TanDEM-X ranged from 3.59 to 13.54 m. The LE90 range for ASTER 6.88 to 24.37 m whereas for the ALOS AW3D30, LE90 ranged from 3.36 to 14.41 m. Grohmann (2018) also found higher slopes in the TanDEM-X data, which should be expected for higher-resolution data (e.g., Zhang and Montgomery, 1994; Zhang et al., 1999; Walker and Willgoose, 1999; Grohmann, 2015; Grieve et al., 2016). Satge et al. (2016) used ICESat GCPs to evaluate absolute errors of SRTM v3 (which is also confusingly called SRTM G1 because it is the global 1 arcsecond dataset) and ASTER GDEM v2 and quantified error by topographic gradients. On flat areas (up to 2 degrees slope), the RMSE errors of SRTM and ASTER were 4.1 and 8.1 m, respectively, but for gradients greater than 20 degrees these rose to 14.9 and 18.2 m, respectively.

4 Accuracy of global datasets

111

Purinton and Bookhagen (2017) used differential GPS (dGPS) to compare SRTM, ASTER, TanDEM-X, and ALOS W3D data. The mean error between dGPS measurements was less than 2.5 m across all datasets but the standard deviation varied significantly: 2.42 m in the case of TanDEM-X, 2.81 m for AW3D30, 3.33 for SRTM v3, and 9.48 for ASTER version 2 (Purinton and Bookhagen, 2017). Rexer and Hirt (2014) found mean errors of 8.5 m for ASTER version 2 and of 6 ms for SRTM version 3 over Australia. Additional assessments have included comparisons between global DEM datasets and a global dataset of airport runways (Becek et al., 2016). Becek (2008) suggested comparing runways against satellite-derived DEMs because of various advantages; the centerlines of runways are in the public domain, they are reported with centimeter accuracy, and the linear, flat nature of runways allows comparisons to evaluate internal noise in the satellite-derived DEM. This effort suggested mean differences between runway data and the TanDEM-X DEM data of 0.79 m (Becek et al., 2016), compared to mean differences of 3.6 m for ASTER version 2 and 1.7 m for the SRTM 3 arcsecond dataset (Becek et al., 2016). The standard deviations were 7.3, 2.2, and 0.8 m for ASTER, SRTM, and TanDEM-X DEM, respectively (Becek et al., 2016). Studies in smaller regions broadly agree with these larger intercomparison studies. Mukherjee et al. (2013) found that errors between GPS ground control points collected in the Indian Himalaya and DEMs were greater for the ASTER DEM (version 2) than for SRTM (CGIAR version 4.1) and uncertainties were greater for steeper slopes. Florinsky et al. (2018) found the ALOS AW3D30 outperformed both ASTER v2 and SRTM v1 based on a geodetic survey in Russia. Yakubu et al. (2019) compared SRTM 4.1 and ASTER version 2 in Ghana using leveling data and found RMSE errors of 6.5 and 9.1 m for the two datasets, respectively. Alganci et al. (2018) compared ALOS AW3D30, SRTM v3, and ASTER v2 data with a highresolution DEM from stereo aerial images over Turkey supplied by the General Command of Mapping, a government agency. This study found that accuracy differed over different land classes; across all but the building land classes, the ALOS AW3D30 DEM outperformed SRTM version 3. Yap et al. (2018) compared the ALOS AW3D30, SRTM v3, and ASTER version 2 against geodetic measurements in Cameroon. They found that ASTER was the least accurate with similar accuracy between SRTM and ALOS AW3D30 except in areas with a slope of less than 2 degrees, where the ALOS AW3D30 was more accurate. ASTER was the least accurate among the three. There are many other examples of data comparisons of specific regions that in general find similar conclusions to those reported above (e.g., Hirt et al., 2010; Suwandana et al., 2012; Amans et al., 2013; Jing et al., 2014; Santillan and Makinano-Santillan, 2016; Hu et al., 2017).

4.5 Summary of vertical accuracy From the numerous studies of vertical accuracy for the SRTM, ASTER, ALOS World 3D-30 and TanDEM-X datasets, it is clear that the ASTER DEM has the lowest vertical accuracy. The TanDEM-X dataset has greater vertical accuracy than the SRTM and the AW3D30 datasets. It should be noted, however, that Schwanghart and Scherler (2017) suggest that in very high-relief terrain, the TanDEM-X dataset may have worse vertical accuracy than both SRTM and AW3D30 datasets. This finding certainly should be investigated further.

112

4. Topographic data from satellites

The vertical accuracy of both SRTM and AW3D30 appear to be similar. However, the AW3D30 dataset seems to be slightly more accurate, particularly in low-relief areas (e.g., Alganci et al., 2018; Florinsky et al., 2018; Yap et al., 2018). In addition, the AW3D30 dataset appears to outperform even the TanDEM-X dataset when extracting river profiles (Schwanghart and Scherler, 2017; Boulton and Stokes, 2018). Finally, the vertical accuracy of all these DEMs degrades as topographic gradients increase (Fig. 6).

River profiles extracted from SRTM 1 arcsecond and ALOS AW3D30 DEMs shown with orange and blue dots, respectively. The site is in the San Bernardino Mountains in East Twin Creek. The dots are the center points of the channel pixels in the 30-m DEMs. These are shown superimposed over a lidar hillshade from https:// opentopography.org. The bold line in the box plots shows the median offset, and the box encloses the data between the first and third quartile of the offsets. The AW3D30’s median error relative to the lidar data is approximately half that of the SRTM data. River profiles from SRTM do not trace channel center lines as well as AW3D30 in this location.

FIG. 6

5 Implications of increasing resolution on geomorphic studies

113

5 Implications of increasing resolution on geomorphic studies Because topographic data is typically delivered on a square grid, increasing the resolution by a factor of two will increase the file size by a factor of four. A 30-m DEM will take up nine times the storage space as a 90-m DEM. A small DEM that is 1.08 km on a side will contain 144 pixels at 90-m grid spacing (e.g., the MERIT DEM grid spacing) but 8100 pixels at 12-m grid spacing (e.g., TanDEM-X grid spacing). One can see how storage size can increase rapidly. Most global DEMs are delivered as 2-byte integer data (e.g., SRTM, AW3D30). That is, every pixel in a DEM takes up 2 bytes of storage. However, integer data limits the gradients that can be resolved in the DEM; if the data resolution is 30 m then the topographic gradients can only increase in increments of 0.0333, but in humid landscapes rivers may have gradients less than 0.01 for drainage areas of as little as 30 km2 in soft lithology (e.g., Hack, 1957; Morisawa, 1962). With higher accuracy DEMs, data is beginning to be delivered as 4 byte floating point numbers, which will double the size of the file at the same spatial resolution. This seems a small price to pay in order to better resolve topographic gradients in channels, floodplains, and ridgetops. Increasing the number of pixels also affects the time of computation of any topographic data processing. A new generation of flow-routing algorithms such as FASTSCAPE (Braun and Willett, 2013) or the algorithms of Barnes (2017) aim to make the time of computation for flow routing and landscape evolution modeling scale with the number of pixels. However, many implementations of flow accumulation in GIS software use a sorting algorithm, wherein the computation time scales with n log(n) where n is the number of pixels (Knuth, 1998). With this increased burden on data storage and computation it is reasonable for a geomorphologist to ask; what do I get in exchange for increased file size and a long wait for my topographic analysis to finish? From a simple esthetic point of view, increased resolution means one can see features in high-resolution data that are simply not visible in lower resolution data (Fig. 7). The implications for geomorphic research go well beyond simple esthetics, however.

5.1 Geomorphic metrics and data processing Most geomorphic studies go well beyond simply looking at topography or hillshade images and attempt to quantify derived metrics from the landscape in order to understand underlying processes, tectonic forcing, natural hazards, hydrologic and sediment responses to landscape forcing, and many other applications of geomorphic analysis. The processing of topographic data will influence the results and therefore the interpretations.

5.2 Simple preprocessing Almost any analysis of topographic data will involve some degree of preprocessing or data manipulation. There are a few issues that seem trivial but could affect a geomorphic analysis. Firstly, almost all topographic data is provided in geographic reference systems, where coordinates are referenced to a sphere (in latitude and longitude). Geomorphologists are frequently interested in gradients, lengths and areas, meaning that these geographic coordinate

114

4. Topographic data from satellites

FIG. 7 Hillshade images of a site in California (see inset map). Panel on top right is 90 m SRTM v3 data over 1 km GTOPO30 data. Other three panels are the area indicated in the top left panel by the arrow.

systems need to be converted to projected coordinate systems (where the surface of a sphere or ellipsoid is projected onto a plane). A projected coordinate system can preserve areas or angles but not both (Snyder, 1987). When reprojecting, the grid of the projected coordinate system will not match the grid of the geographic coordinate system, which necessitates resampling the DEM. In most software used for projections and transformations, the default method is the nearest neighbor method. This should be avoided as it results in a spurious striping of the DEM that does not exist in the original data. A bilinear or cubic resampling method should be used instead (Fig. 8). 5.2.1 Grid resolution: Implications for curvature and slope measurements A number of studies have demonstrated that the accuracy of DEMs derived from satellite degrades in high-relief terrain (e.g., Mukherjee et al., 2013; Satge et al., 2016; Shean et al., 2016; Singh et al., 2016; Takaku et al., 2016). Derived topographic metrics such as gradient or curvature are also sensitive to grid resolution. If coarse resolution data is used, for example, gradient tends to be underestimated (e.g., Gao, 1997; Warren et al., 2004; Vaze et al., 2010). Topographic slope and curvature are calculated in two dimensions on DEMs, but the degradation of these metrics with increasing grid resolution can be illustrated with one-dimensional examples. Grieve et al. (2016) explored how finite difference

5 Implications of increasing resolution on geomorphic studies

115

FIG. 8 Hillshade images showing the artifacts introduced by projecting data and using a nearest neighbor sampling scheme. Large river confluence in middle of image is the confluence of the Green and Colorado Rivers in Utah, USA.

approximations of slope and curvature can be recast as data filters. Actual topography is continuous, but DEMs represent samples of this continuous data, and we can interpret this through the prism of spectral analysis. In spectral analysis, data is approximated as the linear sum of sine waves that have different frequencies. If we calculate slope or curvature using finite difference, we are applying a filter, meaning that we transform input data into output data by using a series of linear functions (i.e., multiplying the input data with a series of weights). Because the filter is working on sample data, it will not perfectly represent the continuous data and there are measures to describe how close the filter will represent the original data. The ratio between the filtered amplitude and the original amplitude is called the gain. In addition, the ratio between the continuous gain and the discrete gain is called the fidelity. The gain (G) is a function of both the sample spacing (Δx) and the wavelength of feature of interest (say, a ridge or a valley). Landscapes may contain many different wavelengths (Perron et al., 2008) from large-scale tectonic structures, to ridge and valley topography, to the spacing of first-order drainage basins. Both the gain and the fidelity are not casted as functions of the wavelength but rather of the wavenumber (Jenkins and Watts, 1968), which is

116

4. Topographic data from satellites

calculated as ω ¼ 2π/L[L1] where ω is the wavenumber with wavelength L [L]. Higher wavenumbers correspond to shorter wavelengths. We can look at the gain and fidelity for resulting from a one-dimensional curvature (Cx) difference operation: Cx ¼

ζ ðxΔxÞ  2ζ x + ζ ðx + ΔxÞ ðΔxÞ2

,

(2)

where ζ [L] is the elevation of the land surface, x [L] is a location in space, Cx is the curvature at location x, and Δx [L] is the grid interval. The subscripts denote the discrete locations where elevation is evaluated. Grieve et al. (2016) calculated the gain and fidelity equations for Eq. (2) based on Jenkins and Watts (1968), where the gain is Gðω; ΔxÞ ¼

2 ðΔxÞ

2

½1  2cosðωΔxÞ + cos2 ðωΔxÞ1=2

(3)

and the fidelity is Fðω; ΔxÞ ¼

2 ðΔxÞ2 ω2

½1  2cosðωΔxÞ + cos2 ðωΔxÞ1=2 :

(4)

The gain function, G, is greater for high frequencies, meaning that landscape features that have high frequency (i.e., short wavelengths) such as ridgecrests, treethrow mounds, or local roughness involve relatively large values of curvature compared to longer wavelength features such as ridge-valley topography or geologic folds. The fidelity function (Eq. 4) shows how information is lost for different wavelengths and grid resolutions. Grieve et al. (2016) showed how the fidelity relates to the wavelength of the landscape feature being analyzed. To achieve a fidelity, F, of 0.9 requires that L/Δx is equal to approximately 6 grid points per wavelength. A fidelity F ¼ 0.95 requires 8 points per wavelength, and F ¼ 0.99 requires 18. The minimum wavelength of a feature that can be resolved is defined by the Nyquist wavenumber, which is defined as Δx/L ¼ 1/2 (e.g., Jenkins and Watts, 1968; Pipaud et al., 2015), but if a landscape feature (e.g., a ridgetop or valley bottom) has a short wavelength relative to the grid spacing, then information will be lost (e.g., Pipaud et al., 2015; Grieve et al., 2016). If we use a DEM that has 12 m grid spacing (like, for example, the TanDEM-X DEM), then we can only capture the curvature of a ridgeline that had a wavelength of 36–48 m (one does not need the entire wave to capture the peak of the waveform). This means landscapes with sharp ridges (e.g., Roering et al., 2007; Hurst et al., 2013) will underestimate the magnitude of the curvature. Many studies have explored how topographic gradients are underestimated by lowresolution data (e.g., Gao, 1997; Warren et al., 2004; Vaze et al., 2010; Pipaud et al., 2015), and Grieve et al. (2016) calculated the fidelity of a simple one-dimensional gradient operation: FS ðω; ΔxÞ ¼

1 ½sinðωΔxÞ: Δxω

(5)

Eq. (5) shows that increasing grid spacing (i.e., greater values of Δx) reduces the ability to resolve local gradients; instead it resolves regional gradients. To achieve a fidelity FS ¼ 0.9, for example, requires L/Δx or approximately 8 grid points per wavelength. A fidelity FS ¼ 0.95 requires 11 points per wavelength, and FS ¼ 0.99 requires 18. If we compare this with the

5 Implications of increasing resolution on geomorphic studies

117

FIG. 9 Comparison of curvature (left column) and gradient (right column) for 12- and 30-m resolution topographic data. The main drainage in the center of the figure is Wounded Knee Creek, South Dakota, USA.

results for curvature, we see that accurate measurements of gradient require even higher grid resolution for most values of fidelity. If we then look at real landscapes, the loss of information as the grid resolution coarsens becomes obvious. Fig. 9 shows that areas with high curvature values (both negative and positive) and high gradient values are not resolved as grid values coarsen.

118

4. Topographic data from satellites

Different DEM data sources have different errors, and so this can also affect secondary metrics. For example, Pipaud et al. (2015) found that the ASTER DEM produced higher mean gradients than TanDEM-X and SRTM 1 arcsecond data but this was due to noise in the ASTER data. However, if errors of the topographic data for two DEMs are similar, derived metrics such as curvature and gradient will be less accurate, in coarser data. This inaccuracy will be greater in steeper channels and hillslopes, and sharper ridges and valleys.

5.3 Accuracy of channel profiles Many geomorphic applications focus specifically on channels, and the relative accuracy of channel profiles extracted from large satellite-derived DEMs does not necessarily mirror the errors either from absolute accuracy of the DEM or from accuracy related to the grid resolution. Although the number of studies on this topic is somewhat limited, two studies (Schwanghart and Scherler, 2017; Boulton and Stokes, 2018) have compared SRTM, ALOS AW3D30, and TanDEM-X and found the ALOS AW3D30 to outperform other DEMs. Despite the fact that SRTM and ALOS AW3D30 have similar reported vertical accuracy, the TanDEM-X DEM has better reported vertical accuracy. Visual inspection in very steep mountain areas suggests that the ALOS AW3D30 DEM can perform well compared with other DEMs.

6 Future developments Enormous progress has been made over the last 2 decades in terms of both the resolution, accuracy, and availability of global topographic data. However, there is still much room for improvement. The hydrology community has highlighted how the errors of current global DEMs, and in particular their inherent bias due to the fact they include trees and buildings, make global flood modeling extremely challenging (e.g., Sampson et al., 2016; Schumann and Bates, 2018; Winsemius et al., 2019). The MERIT DEM (Yamazaki et al., 2017) is the first attempt at a global DTM, but for many geomorphic applications its 3-arcsecond resolution is too coarse. There have been some studies suggesting that high resolution, continuous coverage but lower accuracy datasets could be fused with high accuracy, coarse resolution data to produce a better bare-Earth data. For example, both Lee et al. (2018) and Qi and Dubayah (2016) demonstrated that the TanDEM-X and GEDI could be combined to quantify the elevation of the ground surface below dense forest canopy. There is also evidence that L-band and P-band radar (wavelength 0.6–1.2 m) may be used to reliably detect both the tree canopy and ground surface (Tebaldini and Rocca, 2012). The European Space Agency’s BIOMASS mission uses a P-band radar with the aim of estimating biomass from backscatter data at 200-m resolution, but will spend some time in tomographic mode meaning that it will collect global tree canopy height and surface elevations (Le Toan et al., 2011). This data could be fused with existing DSM datasets to create higher-resolution global DTMs. In addition to creation of higher-resolution global DTMs, other developments would benefit geomorphic and hydrologic research. Schumann and Bates (2018) highlight the need for open access data and better vertical accuracy. The desired characteristics of the next generation of topographic data are similar for both the geomorphology and hydrology communities.

7 Conclusions

119

Desired characteristics of the next generation of satellite derived topographic datasets 1. 2. 3. 4.

Open access, fully open license Higher accuracy than present, particularly in steep terrain A digital terrain model (DTM) rather than a digital surface model (DSM) 5-m resolution or better

The criteria listed above limit options. Making the data open almost certainly requires that the data collection effort is funded by governments and not as a commercial enterprise. In addition, the inability to see through the tree canopy precludes optical satellites as well as X- and C-band radar satellites. Although Schumann (2014) has advocated creation of a global dataset by merging airborne data that is collected by governments and charities, this may lead to a large variation in data quality. Another option is a satellite lidar instrument capable of swath mapping with the ability to penetrate the tree canopy. These criteria have been clear for some time; in 2007, the National Research Council of the United States (NRC) published recommendations for key future Earth observation missions and highlighted the need for a global 5-m resolution topographic map at subdecimeter accuracy (National Research Council, 2007). In that report, one of the 15 Earth-Observing missions recommended was the “Land surface topography for landslide hazards and water runoff” (LIST) mission, which would be conducted using a satellite mounted laser altimeter, with an estimated cost of US$300 million (National Research Council, 2007). The NRC report triggered further investigation by NASA, which added further technical specifications on the LIST mission: (i) a 5-km swath to make global data collection tractable; (ii) ability to detect the surface through thin clouds; (iii) night and day operation; (iv) large dynamic range to account for reflectance; (v) ability to detect returns though dense vegetation; (vi) a pulse rate that allows ranging through clouds; and (vii) very high efficiency to make the mission cost effective (Yu et al., 2010). Further work specified that, in order to attain global coverage, LIST requires on the order of 1000 profiling beams, which compares to 6 for IceSAT-2 (Krainak et al., 2012). Krainak et al. (2012) identify the main challenge is the development of efficient detectors and lasers. Although LIST has not been approved for launch, no competing space agency has scheduled a similar mission and so, in this author’s opinion, advocating the LIST mission should be among the priorities for both the geomorphology and hydrology communities.

7 Conclusions Just before the turn of the century, geomorphologists had limited access to digital elevation models and had to do with national datasets derived from a range of methods and paper maps. The only global-scale digital elevation model had a grid spacing of 30 arcseconds, which translates to roughly 1-km resolution. We now have commercial data products at 5- and 12-m resolutions that cover much of the globe. Freely available data spans the entire

120

4. Topographic data from satellites

planet at 30 m resolution. This has enabled a wide range of geomorphic applications and studies and global DEMs underpin much geomorphic work performed today, particularly in remote regions. Due to information degradation at coarser grid scales, however, there is still much room for improvement. A freely available, global scale 5 m data product that represents the Earth’s surface, rather than the top of the tree canopy and buildings, would be extremely beneficial for studying of hydrology, sediment transport, and hazards. The data available to modern geomorphologists, derived from satellite instruments makes today an extremely exciting time to be a geomorphologist, and many theories proposed in the late 1800s may now be tested in ways that could not have been dreamed by early pioneers of landscape analysis.

References Abrams, M., 2000. The advanced spaceborne thermal emission and reflection radiometer (ASTER): data products for the high spatial resolution imager on NASA’s Terra platform. Int. J. Remote Sens. 21 (5), 847–859. https://doi.org/ 10.1080/014311600210326. Abrams, M., Bailey, B., Tsu, H., Hato, M., 2010. The ASTER global DEM. Photogramm. Eng. Remote. Sens. 76 (4), 344–348. Ahokas, E., Kaartinen, H., Hyypp€a, J., 2003. A quality assessment of airborne laser scanner data. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 34 (Pt. 3/W13), 1–7. Alganci, U., Besol, B., Sertel, E., 2018. Accuracy assessment of different digital surface models. ISPRS In. J. Geo Inf. 7 (3), 114. https://doi.org/10.3390/ijgi7030114. Amans, O.C., Beiping, W., Ziggah, Y.Y., 2013. Assessing vertical accuracy of SRTM Ver 4.1 and ASTER GDEM Ver 2 using differential GPS measurements—case study in Ondo State, Nigeria. Int. J. Sci. Eng. Res. 4 (12), 1–9. ASF Engineering, 2015. ASF radiometric terrain corrected products: algorithm theoretical basis document. https:// media.asf.alaska.edu/uploads/RTC/rtc_atbd_v1.2_final.pdf. Bagnardi, M., Gonzalez, P.J., Hooper, A., 2016. High-resolution digital elevation model from tri-stereo Pleiades-1 satellite imagery for lava flow volume estimates at Fogo Volcano. Geophys. Res. Lett. 43 (12), 6267–6275. https://doi. org/10.1002/2016GL069457. Balzter, H., 2001. Forest mapping and monitoring with interferometric synthetic aperture radar (InSAR). Prog. Phys. Geogr. 25 (2), 159–177. https://doi.org/10.1177/030913330102500201. Bamber, J., Muller, J.-P., Manganayake, A., 1997. A global 5 arc minute digital elevation model derived from the geodetic phase of ERS-1 (DEM). In: Proc. of the 3rd ERS Symposium—Space at the Service of the Environment, Florence, vol. 414. p. 1779. Barbarella, M., Fiani, M., Zollo, C., 2017. Assessment of DEM derived from very high-resolution stereo satellite imagery for geomorphometric analysis. Eur. J. Remote Sens. 50 (1), 534–549. https://doi.org/10.1080/22797254. 2017.1372084. Barnes, R., 2017. Parallel non-divergent flow accumulation for trillion cell digital elevation models on desktops or clusters. Environ. Model. Softw. 92, 202–212. https://doi.org/10.1016/j.envsoft.2017.02.022. Becek, K., 2008. Investigating error structure of shuttle radar topography mission elevation data product. Geophys. Res. Lett. 35 (15), L15403. https://doi.org/10.1029/2008GL034592. Becek, K., Koppe, W., Kutoglu, S.H., 2016. Evaluation of vertical accuracy of the WorldDEM(tm) using the runway method. Remote Sens. 8 (11), 934. https://doi.org/10.3390/rs8110934. Blair, J.B., Rabine, D.L., Hofton, M.A., 1999. The laser vegetation imaging sensor: a medium-altitude, digitisationonly, airborne laser altimeter for mapping vegetation and topography. ISPRS J. Photogramm. Remote Sens. 54 (2), 115–122. https://doi.org/10.1016/S0924-2716(99)00002-7. Boulton, S.J., Stokes, M., 2018. Which DEM is best for analyzing fluvial landscape development in mountainous terrains? Geomorphology 310, 168–187. https://doi.org/10.1016/j.geomorph.2018.03.002. Braun, J., Willett, S.D., 2013. A very efficient O(n), implicit and parallel method to solve the stream power equation governing fluvial incision and landscape evolution. Geomorphology 180–181, 170–179. https://doi.org/10.1016/j. geomorph.2012.10.008.

References

121

Brunt, K.M., Hawley, R.L., Lutz, E.R., Studinger, M., Sonntag, J.G., Hofton, M.A., Andrews, L.C., Neumann, T.A., 2017. Assessment of NASA airborne laser altimetry data using ground-based GPS data near summit station, Greenland. Cryosphere 11 (2), 681–692. https://doi.org/10.5194/tc-11-681-2017. Candela, S.G., Howat, I., Noh, M.J., Porter, C.C., Morin, P.J., 2017. ArcticDEM validation and accuracy assessment. In: AGU Fall Meeting Abstracts, vol. 51. p. C51A-0322. Capaldo, P., Crespi, M., Fratarcangeli, F., Nascetti, A., Pieralice, F., 2011. High-resolution SAR radargrammetry: a first application with COSMO-SkyMed SpotLight imagery. IEEE Geosci. Remote Sens. Lett. 8 (6), 1100–1104. https:// doi.org/10.1109/LGRS.2011.2157803. Carabajal, C.C., Harding, D.J., Suchdeo, V.P., 2010. Icesat lidar and global digital elevation models: applications to desdyni. In: 2010 IEEE International Geoscience and Remote Sensing Symposium, pp. 1907–1910. Chu, T., Lindenschmidt, K.-E., 2017. Comparison and validation of digital elevation models derived from InSAR for a flat Inland delta in the high latitudes of Northern Canada. Can. J. Remote. Sens. 43 (2), 109–123. https://doi.org/ 10.1080/07038992.2017.1286936. Crippen, R., Buckley, S., Agram, P., Belz, E., Gurrola, E., Hensley, S., Kobrick, M., Lavalle, M., Martin, J., Neumann, M., Nguyen, Q., Rosen, P., Shimada, J., Simard, M., Tung, W., 2016. Nasadem global elevation model: methods and progress. ISPRS Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 41B4, 125–128. https://doi.org/ 10.5194/isprs-archives-XLI-B4-125-2016. Danielson, J.J., Gesch, D.B., 2011. Global multi-resolution terrain elevation data 2010 (GMTED2010). U.S. Geological Survey USGS Numbered Series 2011-1073. Dehecq, A., Millan, R., Berthier, E., Gourmelen, N., Trouve, E., Vionnet, V., 2016. Elevation changes inferred from TanDEM-X data over the Mont-Blanc area: impact of the X-band interferometric bias. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 9 (8), 3870–3882. https://doi.org/10.1109/JSTARS.2016.2581482. Farr, T.G., Rosen, P.A., Caro, E., Crippen, R., Duren, R., Hensley, S., Kobrick, M., Paller, M., Rodriguez, E., Roth, L., Seal, D., Shaffer, S., Shimada, J., Umland, J., Werner, M., Oskin, M., Burbank, D., Alsdorf, D., 2007. The shuttle radar topography mission. Rev. Geophys. 45 (2), RG2004. https://doi.org/10.1029/2005RG000183. Fisher, P.E., Tate, N.J., 2006. Causes and consequences of error in digital elevation models. Prog. Phys. Geogr. 30 (4), 467–489. https://doi.org/10.1191/0309133306pp492ra. Florinsky, I.V., Skrypitsyna, T.N., Luschikova, O.S., 2018. Comparative accuracy of the AW3d30 DSM, ASTER GDEM, and SRTM1 DEM: a case study on the Zaoksky testing ground, Central European Russia. Remote Sens. Lett. 9 (7), 706–714. https://doi.org/10.1080/2150704X.2018.1468098. Fritz, T., Rossi, C., Yague-Martinez, N., Rodriguez-Gonzalez, F., Lachaise, M., Breit, H., 2011. Interferometric processing of TanDEM-X data. In: 2011 IEEE International Geoscience and Remote Sensing Symposium, pp. 2428–2431. Gao, J., 1997. Resolution and accuracy of terrain representation by grid DEMs at a micro-scale. Int. J. Geograph. Inf. Sci. 11 (2), 199–212. https://doi.org/10.1080/136588197242464. Gesch, D.B., Larson, K.S., 1998. Techniques for development of global 1-kilometer digital elevation models. In: Proceedings, Pecora Thirteen Symposium, August 20–22, 1996 (CD-ROM). Sioux Falls, South Dakota. Gesch, D.B., Verdin, K.L., Greenlee, S.K., 1999. New land surface digital elevation model covers the Earth. Eos 80 (6), 69. https://doi.org/10.1029/99EO00050. Gesch, D., Oimoen, M., Greenlee, S., Nelson, C., Steuck, M., Tyler, D., 2002. The national elevation dataset. Photogramm. Eng. Remote. Sens. 68 (1), 5–11. Gesch, D., Oimoen, M., Danielson, J., Meyer, D., 2016. Validation of the ASTER global digital elevation model version 3 over the conterminous United States. In: ISPRS—International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XLI-B4. Copernicus GmbH, pp. 143–148. Gilbert, G.K., 1877. Geology of the Henry Mountains. Government Printing Office, Washington, DC USGS Unnumbered Series. Gilbert, G.K., 1909. The convexity of hilltops. J. Geol. 17 (4), 344–350. Glennie, C.L., Carter, W.E., Shrestha, R.L., Dietrich, W.E., 2013. Geodetic imaging with airborne LiDAR: the Earth’s surface revealed. Rep. Prog. Phys. 76 (8), 086801. https://doi.org/10.1088/0034-4885/76/8/086801. Goldstein, R.M., Zebker, H.A., Werner, C.L., 1988. Satellite radar interferometry: two-dimensional phase unwrapping. Radio Sci. 23 (4), 713–720. https://doi.org/10.1029/RS023i004p00713. Gonzalez, J.H., Bachmann, M., Scheiber, R., Krieger, G., 2010. Definition of ICESat selection criteria for their use as height references for TanDEM-X. IEEE Trans. Geosci. Remote Sens. 48 (6), 2750–2757. https://doi.org/10.1109/ TGRS.2010.2041355.

122

4. Topographic data from satellites

Graham, L.C., 1974. Synthetic interferometer radar for topographic mapping. Proc. IEEE 62 (6), 763–768. https://doi. org/10.1109/PROC.1974.9516. Grieve, S.W.D., Mudd, S.M., Milodowski, D.T., Clubb, F.J., Furbish, D.J., 2016. How does grid-resolution modulate the topographic expression of geomorphic processes? Earth Surf. Dyn. 4 (3), 627–653. https://doi.org/10.5194/esurf4-627-2016. Grohman, G., Kroenung, G., Strebeck, J., 2006. Filling SRTM voids: the delta surface fill method. Photogramm. Eng. Remote. Sens. 72 (3), 213–216. Grohmann, C.H., 2015. Effects of spatial resolution on slope and aspect derivation for regional-scale analysis. Comput. Geosci. 77, 111–117. https://doi.org/10.1016/j.cageo.2015.02.003. Grohmann, C.H., 2018. Evaluation of TanDEM-X DEMs on selected Brazilian sites: comparison with SRTM, ASTER GDEM and ALOS AW3d30. Remote Sens. Environ. 212, 121–133. https://doi.org/10.1016/j.rse.2018.04.043. Gugan, D., Dowman, I., 1988. Topographic mapping from spot imagery. Photogramm. Eng. Remote. Sens. 54 (10), 1409–1414. Hack, J., 1957. Studies of longitudinal profiles in Virginia and Maryland. United States Government Printing Office, Washington, DC U.S. Geological Survey Professional Paper 294-B. Hancock, S., Armston, J., Hofton, M., Sun, X., Tang, H., Duncanson, L.I., Kellner, J.R., Dubayah, R., 2019. The GEDI simulator: a large-footprint waveform lidar simulator for calibration and validation of spaceborne missions. Earth and Space Science 6 (2), 294–310. https://doi.org/10.1029/2018EA000506. Harding, D.J., Carabajal, C.C., 2005. ICESat waveform measurements of within-footprint topographic relief and vegetation vertical structure. Geophys. Res. Lett.. 32(21)https://doi.org/10.1029/2005GL023471. Hawker, L., Bates, P., Neal, J., Rougier, J., 2018. Perspectives on digital elevation model (DEM) simulation for flood modeling in the absence of a high-accuracy open access global DEM. Front. Earth Sci. 6, 233. https://doi.org/ 10.3389/feart.2018.00233. Hirano, A., Welch, R., Lang, H., 2003. Mapping from ASTER stereo image data: DEM validation and accuracy assessment. ISPRS J. Photogramm. Remote Sens. 57 (5), 356–370. https://doi.org/10.1016/S0924-2716(02)00164-8. Hirt, C., 2018. Artefact detection in global digital elevation models (DEMs): the maximum slope approach and its application for complete screening of the SRTM v4.1 and MERIT DEMs. Remote Sens. Environ. 207, 27–41. https://doi.org/10.1016/j.rse.2017.12.037. Hirt, C., Filmer, M.S., Featherstone, W.E., 2010. Comparison and validation of the recent freely available ASTERGDEM ver1, SRTM ver4.1 and GEODATA DEM-9s ver3 digital elevation models over Australia. Aust. J. Earth Sci. 57 (3), 337–347. https://doi.org/10.1080/08120091003677553. Hobi, M.L., Ginzler, C., 2012. Accuracy assessment of digital surface models based on WorldView-2 and ADS80 stereo remote sensing data. Sensors 12 (5), 6347–6368. https://doi.org/10.3390/s120506347. Hodgson, M.E., Bresnahan, P., 2004. Accuracy of airborne lidar-derived elevation. https://doi.org/10.14358/ PERS.70.3.331. Horton, R.E., 1945. Erosional development of streams and their drainage basins; hydrophysical approach to quantitative hydrology. GSA Bull. 56 (3), 275–370. https://doi.org/10.1130/0016-7606(1945)56[275:EDOSAT]2.0.CO;2. Howat, I., Morin, P., Porter, C., Noh, M.-J., 2018. The Reference Elevation Model of Antarctica. Harvard Dataverse. https://doi.org/10.7910/DVN/SAIK8B. Howat, I.M., Porter, C., Smith, B.E., Noh, M.-J., Morin, P., 2019. The reference elevation model of Antarctica. Cryosphere 13 (2), 665–674. https://doi.org/10.5194/tc-13-665-2019. Hu, Z., Peng, J., Hou, Y., Shan, J., 2017. Evaluation of recently released open global digital elevation models of Hubei, China. Remote Sens. 9 (3), 262. https://doi.org/10.3390/rs9030262. Huber, M., Wessel, B., Kosmann, D., Felbier, A., Schwieger, V., Habermeyer, M., Wendleder, A., Roth, A., 2009. Ensuring globally the TanDEM-X height accuracy: analysis of the reference data sets ICESat, SRTM and KGPS-tracks. In: 2009 IEEE International Geoscience and Remote Sensing Symposium, vol. 2. pp. II-769–II-772. Hurst, M.D., Mudd, S.M., Attal, M., Hilley, G., 2013. Hillslopes record the growth and decay of landscapes. Science 341 (6148), 868–871. https://doi.org/10.1126/science.1241791. Imhoff, M., Story, M., Vermillion, C., Khan, F., Polcyn, F., 1986. Forest canopy characterization and vegetation penetration assessment with space-borne radar. IEEE Trans. Geosci. Remote Sens. GE-24 (4), 535–542. https://doi. org/10.1109/TGRS.1986.289668. Jaboyedoff, M., Oppikofer, T., Abellan, A., Derron, M.-H., Loye, A., Metzger, R., Pedrazzini, A., 2012. Use of LIDAR in landslide investigations: a review. Nat. Hazards 61 (1), 5–28. https://doi.org/10.1007/s11069-010-9634-2.

References

123

Jain, A.O., Thaker, T., Chaurasia, A., Patel, P., Singh, A.K., 2018. Vertical accuracy evaluation of SRTM-GL1, GDEMV2, AW3d30 and CartoDEM-V3.1 of 30-m resolution with dual frequency GNSS for lower Tapi Basin India. Geocarto Int. 33 (11), 1237–1256. https://doi.org/10.1080/10106049.2017.1343392. Jenkins, G.M., Watts, D.G., 1968. Spectral Analysis and It Applications. Holden-Day, San Francisco. Jing, C., Shortridge, A., Lin, S., Wu, J., 2014. Comparison and validation of SRTM and ASTER GDEM for a subtropical landscape in Southeastern China. Int. J. Digital Earth 7 (12), 969–992. https://doi.org/10.1080/17538947. 2013.807307. Kervyn, F., 2001. Modelling topography with SAR interferometry: illustrations of a favourable and less favourable environment. Comput. Geosci. 27 (9), 1039–1050. https://doi.org/10.1016/S0098-3004(00)00158-8. Knuth, D.E., 1998. The Art of Computer Programming, second ed. Sorting and Searching, vol. 3. Addison Wesley Longman Publishing Co., Inc., Redwood City, CA, USA Koppe, W., Henrichs, L., Hummel, P., 2015. Assessment of WorldDEMTM global elevation model using different references. In: 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 5296–5299. Krainak, M.A., Abshire, J.B., Camp, J., Chen, J.R., Coyle, B., Li, S.X., Numata, K., Riris, H., Stephen, M.A., Stysley, P., Yang, G., Yu, A.W., 2012. Laser transceivers for future NASA missions. In: Proc. SPIE 8381, Laser Technology for Defense and Security VIII, 83810Y, vol. 8381. . Krieger, G., Moreira, A., Fiedler, H., Hajnsek, I., Werner, M., Younis, M., Zink, M., 2007. TanDEM-X: a satellite formation for high-resolution SAR interferometry. IEEE Transactions on Geoscience and Remote Sensing 45 (11), 3317–3341. https://doi.org/10.1109/TGRS.2007.900693. Krieger, G., Zink, M., Bachmann, M., Brautigam, B., Schulze, D., Martone, M., Rizzoli, P., Steinbrecher, U., Walter Antony, J., De Zan, F., Hajnsek, I., Papathanassiou, K., Kugler, F., Rodriguez Cassola, M., Younis, M., Baumgartner, S., Lopez-Dekker, P., Prats, P., Moreira, A., 2013. TanDEM-X: a radar interferometer with two formation-flying satellites. Acta Astronaut. 89, 83–98. https://doi.org/10.1016/j.actaastro.2013.03.008. Kugler, F., Schulze, D., Hajnsek, I., Pretzsch, H., Papathanassiou, K.P., 2014. TanDEM-X Pol-InSAR performance for forest height estimation. IEEE Trans. Geosci. Remote Sens. 52 (10), 6404–6422. https://doi.org/10.1109/ TGRS.2013.2296533. Le Toan, T., Quegan, S., Davidson, M.W.J., Balzter, H., Paillou, P., Papathanassiou, K., Plummer, S., Rocca, F., Saatchi, S., Shugart, H., Ulander, L., 2011. The BIOMASS mission: mapping global forest biomass to better understand the terrestrial carbon cycle. Remote Sens. Environ. 115 (11), 2850–2860. https://doi.org/10.1016/j. rse.2011.03.020. Lee, S., Fatoyinbo, T., Qi, W., Hancock, S., Armston, J., Dubayah, R., 2018. Gedi and Tandem-X fusion for 3D forest structure parameter retrieval. In: IGARSS 2018–2018 IEEE International Geoscience and Remote Sensing Symposium, pp. 380–382. Lefsky, M.A., Harding, D.J., Keller, M., Cohen, W.B., Carabajal, C.C., Espirito-Santo, F.D.B., Hunter, M.O., Oliveira, R.D., 2005. Estimates of forest canopy height and aboveground biomass using ICESat. Geophys. Res. Lett. 32(22). https://doi.org/10.1029/2005GL023971. Liu, X., 2008. Airborne LiDAR for DEM generation: some critical issues. Prog. Phys. Geogr. 32 (1), 31–49. https://doi. org/10.1177/0309133308089496. Maire, C., Datcu, M., Audenino, P., 2003. SAR DEM filtering by mean of Bayesian and multi-scale, nonstationary methods. In: IGARSS 2003. 2003 IEEE International Geoscience and Remote Sensing Symposium. Proceedings (IEEE Cat. No.03CH37477), vol. 6. pp. 3916–3918. Mallet, C., Bretar, F., 2009. Full-waveform topographic lidar: state-of-the-art. ISPRS J. Photogramm. Remote Sens. 64 (1), 1–16. https://doi.org/10.1016/j.isprsjprs.2008.09.007. Markus, T., Neumann, T., Martino, A., Abdalati, W., Brunt, K., Csatho, B., Farrell, S., Fricker, H., Gardner, A., Harding, D., Jasinski, M., Kwok, R., Magruder, L., Lubin, D., Luthcke, S., Morison, J., Nelson, R., Neuenschwander, A., Palm, S., Popescu, S., Shum, C., Schutz, B.E., Smith, B., Yang, Y., Zwally, J., 2017. The ice, cloud, and land elevation satellite-2 (ICESat-2): science requirements, concept, and implementation. Remote Sens. Environ. 190, 260–273. https://doi.org/10.1016/j.rse.2016.12.029. Massonnet, D., Souyris, J.-C., 2008. Imaging With Synthetic Aperture Radar. CRC Press. Moreira, A., Prats-Iraola, P., Younis, M., Krieger, G., Hajnsek, I., Papathanassiou, K.P., 2013. A tutorial on synthetic aperture radar. IEEE Geosci. Remote Sens. Mag. 1 (1), 6–43. https://doi.org/10.1109/MGRS.2013.2248301. Morisawa, M.E., 1962. Quantitative geomorphology of some watersheds in the appalachian plateau. GSA Bull. 73 (9), 1025–1046. https://doi.org/10.1130/0016-7606(1962)73[1025:QGOSWI]2.0.CO;2.

124

4. Topographic data from satellites

Mukherjee, S., Joshi, P.K., Mukherjee, S., Ghosh, A., Garg, R.D., Mukhopadhyay, A., 2013. Evaluation of vertical accuracy of open source digital elevation model (DEM). Int. J. Appl. Earth Obs. Geoinf. 21, 205–217. https://doi.org/ 10.1016/j.jag.2012.09.004. M€ uller, J., G€ artner-Roer, I., Thee, P., Ginzler, C., 2014. Accuracy assessment of airborne photogrammetrically derived high-resolution digital elevation models in a high mountain environment. ISPRS J. Photogramm. Remote Sens. 98, 58–69. https://doi.org/10.1016/j.isprsjprs.2014.09.015. Muralikrishnan, S., Pillai, A., Narender, B., Reddy, S., Venkataraman, V.R., Dadhwal, V.K., 2013. Validation of Indian National DEM from Cartosat-1 data. J. Indian Soc. Remote Sens. 41 (1), 1–13. https://doi.org/10.1007/s12524-0120212-9. Narine, L.L., Popescu, S., Neuenschwander, A., Zhou, T., Srinivasan, S., Harbeck, K., 2019. Estimating aboveground biomass and forest canopy cover with simulated ICESat-2 data. Remote Sens. Environ. 224, 1–11. https://doi.org/ 10.1016/j.rse.2019.01.037. NASA, J.P.L., 2013. NASA Shuttle Radar Topography Mission Global 1 arc second [Data set]. NASA EOSDIS Land Processes DAAC. https://doi.org/10.5067/MEaSUREs/SRTM/SRTMGL1.003. National Research Council, 2007. Earth Science and Applications From Space: National Imperatives for the Next Decade and Beyond. The National Academies Press, Washington, DC. https://doi.org/10.17226/11820. Noh, M.-J., Howat, I.M., 2015. Automated stereo-photogrammetric DEM generation at high latitudes: surface extraction with TIN-based search-space minimization (SETSM) validation and demonstration over glaciated regions. GIsci. Remote Sens. 52 (2), 198–217. https://doi.org/10.1080/15481603.2015.1008621. Noh, M.-J., Howat, I.M., 2017. The surface extraction from TIN based search-space minimization (SETSM) algorithm. ISPRS J. Photogramm. Remote Sens. 129, 55–76. https://doi.org/10.1016/j.isprsjprs.2017.04.019. Noh, M.-J., Howat, I.M., 2018. Automatic relative RPC image model bias compensation through hierarchical image matching for improving DEM quality. ISPRS J. Photogramm. Remote Sens. 136, 120–133. https://doi.org/10.1016/ j.isprsjprs.2017.12.008. O’Callaghan, J.F., Mark, D.M., 1984. The extraction of drainage networks from digital elevation data. Comput. Vis. Graph. Image Process. 28 (3), 323–344. https://doi.org/10.1016/S0734-189X(84)80011-0. O’Loughlin, F.E., Paiva, R.C.D., Durand, M., Alsdorf, D.E., Bates, P.D., 2016. A multi-sensor approach towards a global vegetation corrected SRTM DEM product. Remote Sens. Environ. 182, 49–59. https://doi.org/10.1016/j. rse.2016.04.018. Ordnance Survey Great Britain, 2001. Land-Form PANORAMA User Guide v3.0. Perron, J.T., Kirchner, J.W., Dietrich, W.E., 2008. Spectral signatures of characteristic spatial scales and nonfractal structure in landscapes. J. Geophys. Res. Earth Surf. 113(F4). https://doi.org/10.1029/2007JF000866. Peucker, T.K., Douglas, D.H., 1975. Detection of surface-specific points by local parallel processing of discrete Terrain elevation data. Comput. Graph. Image Process. 4 (4), 375–387. https://doi.org/10.1016/0146-664X(75)90005-2. Pipaud, I., Loibl, D., Lehmkuhl, F., 2015. Evaluation of TanDEM-X elevation data for geomorphological mapping and interpretation in high mountain environments—a case study from SE Tibet, China. Geomorphology 246, 232–254. https://doi.org/10.1016/j.geomorph.2015.06.025. Playfair, J., 1802. Illustrations of the Huttonian Theory of the Earth. Neill and Co, Edinburgh. http://archive.org/ details/NHM104643. Poli, D., Remondino, F., Angiuli, E., Agugiaro, G., 2015. Radiometric and geometric evaluation of GeoEye-1, WorldView-2 and Pleiades-1A stereo images for 3D information extraction. ISPRS J. Photogramm. Remote Sens. 100, 35–47. https://doi.org/10.1016/j.isprsjprs.2014.04.007. Porter, C., Morin, P., Howat, I., Noh, M.-J., Bates, B., Peterman, K., Keesey, S., Schlenk, M., Gardiner, J., Tomko, K., Willis, M., Kelleher, C., Cloutier, M., Husby, E., Foga, S., Nakamura, H., Platson, M., Wethington, M., Williamson, C., Bauer, G., Enos, J., Arnold, G., Kramer, W., Becker, P., Doshi, A., D’Souza, C., Cummens, P., Laurier, F., Bojesen, M., 2018. ArcticDEM. Harvard Dataverse. https://doi.org/10.7910/DVN/OHHUKH. https://dataverse.harvard.edu/dataset.xhtml?persistentId¼doi:10.7910/DVN/OHHUKH. Purinton, B., Bookhagen, B., 2017. Validation of digital elevation models (DEMs) and comparison of geomorphic metrics on the southern central Andean Plateau. Earth Surf. Dyn. 5 (2), 211–237. https://doi.org/10.5194/esurf-5-2112017. Purinton, B., Bookhagen, B., 2018. Measuring decadal vertical land-level changes from SRTM-C (2000) and TanDEMX (2015) in the south-central Andes. Earth Surf. Dyn. 6 (4), 971–987. https://doi.org/10.5194/esurf-6-971-2018. Qi, W., Dubayah, R.O., 2016. Combining Tandem-X InSAR and simulated GEDI lidar observations for forest structure mapping. Remote Sens. Environ. 187, 253–266. https://doi.org/10.1016/j.rse.2016.10.018.

References

125

Rabus, B., Eineder, M., Roth, A., Bamler, R., 2003. The shuttle radar topography mission—a new class of digital elevation models acquired by spaceborne radar. ISPRS J. Photogramm. Remote Sens. 57 (4), 241–262. https://doi. org/10.1016/S0924-2716(02)00124-7. Raney, R.K., Luscombe, A.P., Langham, E.J., Ahmed, S., 1991. RADARSAT (SAR imaging). Proc. IEEE 79 (6), 839–849. https://doi.org/10.1109/5.90162. Reuter, H.I., Nelson, A., Jarvis, A., 2007. An evaluation of void-filling interpolation methods for SRTM data. Int. J. Geograph. Inf. Sci. 21 (9), 983–1008. https://doi.org/10.1080/13658810601169899. Rexer, M., Hirt, C., 2014. Comparison of free high resolution digital elevation data sets (ASTER GDEM2, SRTM v2.1/ v4.1) and validation against accurate heights from the Australian National Gravity Database. Aust. J. Earth Sci. 61 (2), 213–226. https://doi.org/10.1080/08120099.2014.884983. Riegler, G., Hennig, S.D., Weber, M., 2015. Worlddem—a novel global foundation layer. In: Stilla, U., Heipke, C. (Eds.), In: Pia15+hrigi15–Joint ISPRS Conference, Vol. I, vol. 40–43. Copernicus Gesellschaft Mbh, Gottingen, pp. 183–187. Rignot, E., Echelmeyer, K., Krabill, W., 2001. Penetration depth of interferometric synthetic-aperture radar signals in snow and ice. Geophys. Res. Lett. 28 (18), 3501–3504. https://doi.org/10.1029/2000GL012484. Rita, M.D., Nascetti, A., Crespi, M., 2017. Open source tool for DSMs generation from high resolution optical satellite imagery: development and testing of an OSSIM plug-in. Int. J. Remote Sens. 38 (7), 1788–1808. https://doi.org/ 10.1080/01431161.2017.1288305. autigam, B., Kraus, T., Martone, M., Krieger, G., 2012. Relative height error analysis of TanDEM-X elRizzoli, P., Br€ evation data. ISPRS J. Photogramm. Remote Sens. 73, 30–38. https://doi.org/10.1016/j.isprsjprs.2012.06.004. Rizzoli, P., Martone, M., Gonzalez, C., Wecklich, C., Borla Tridon, D., Br€autigam, B., Bachmann, M., Schulze, D., Fritz, T., Huber, M., Wessel, B., Krieger, G., Zink, M., Moreira, A., 2017. Generation and performance assessment of the global TanDEM-X digital elevation model. ISPRS J. Photogramm. Remote Sens. 132, 119–139. https://doi. org/10.1016/j.isprsjprs.2017.08.008. Rodrı´guez, E., Morris, C.S., Belz, J.E., Chapin, E.C., Martin, J.M., Daffer, W., Hensley, S., 2005. An assessment of the SRTM topographic products. In: Technical Report JPL D-31639, Jet Propulsion Laboratory, Pasadena, California. Rodrı´guez, E., Morris, C.S., Belz, J.E., 2006. A global assessment of the SRTM performance. Photogramm. Eng. Remote. Sens. 72 (3), 249–260. https://doi.org/10.14358/PERS.72.3.249. Roering, J.J., Perron, J.T., Kirchner, J.W., 2007. Functional relationships between denudation and hillslope form and relief. Earth Planet. Sci. Lett. 264 (1), 245–258. https://doi.org/10.1016/j.epsl.2007.09.035. Rosenqvist, A., Shimada, M., Ito, N., Watanabe, M., 2007. ALOS PALSAR: a Pathfinder mission for global-scale monitoring of the environment. IEEE Trans. Geosci. Remote Sens. 45 (11), 3307–3316. https://doi.org/10.1109/ TGRS.2007.901027. Rossi, C., Rodriguez Gonzalez, F., Fritz, T., Yague-Martinez, N., Eineder, M., 2012. TanDEM-X calibrated Raw DEM generation. ISPRS J. Photogramm. Remote Sens. 73, 12–20. https://doi.org/10.1016/j.isprsjprs.2012.05.014. Rupnik, E., Daakir, M., Pierrot Deseilligny, M., 2017. MicMac—a free, open-source solution for photogrammetry. Open Geospatial Data Softw. Standards 2 (1), 14. https://doi.org/10.1186/s40965-017-0027-2. Sampson, C.C., Smith, A.M., Bates, P.D., Neal, J.C., Trigg, M.A., 2016. Perspectives on open access high resolution digital elevation models to produce global flood hazard layers. Front. Earth Sci. 3. https://doi.org/10.3389/ feart.2015.00085. Santillan, J.R., Makinano-Santillan, M., 2016. Vertical accuracy assessment of 30-M resolution ALOS, ASTER, and SRTM Global DEMs over Northeastern Mindanao, Philippines. ISPRS Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 41B4, 149–156. https://doi.org/10.5194/isprs-archives-XLI-B4-149-2016. Satge, F., Denezine, M., Pillco, R., Timouk, F., Pinel, S., Molina, J., Garnier, J., Seyler, F., Bonnet, M.-P., 2016. Absolute and relative height-pixel accuracy of SRTM-GL1 over the South American Andean Plateau. ISPRS J. Photogramm. Remote Sens. 121, 157–166. https://doi.org/10.1016/j.isprsjprs.2016.09.003. Schlund, M., Baron, D., Magdon, P., Erasmi, S., 2019. Canopy penetration depth estimation with TanDEM-X and its compensation in temperate forests. ISPRS J. Photogramm. Remote Sens. 147, 232–241. https://doi.org/10.1016/j. isprsjprs.2018.11.021. Schumann, G.J.P., 2014. Technology: fight floods on a global scale. Nature 507, 169. https://doi.org/10.1038/507169e. Schumann, G.J.P., Bates, P.D., 2018. The need for a high-accuracy, open-access global DEM. Front. Earth Sci. 6. https://doi.org/10.3389/feart.2018.00225. Schumm, S.A., 1956. Evolution of drainage systems and slopes in badlands at Perth Amboy, New Jersey. GSA Bull. 67 (5), 597–646. https://doi.org/10.1130/0016-7606(1956)67[597:EODSAS]2.0.CO;2.

126

4. Topographic data from satellites

Schutz, B.E., Zwally, H.J., Shuman, C.A., Hancock, D., DiMarzio, J.P., 2005. Overview of the ICESat mission. Geophys. Res. Lett. 32 (21), L21S01. https://doi.org/10.1029/2005GL024009. Schwanghart, W., Scherler, D., 2017. Bumps in river profiles: uncertainty assessment and smoothing using quantile regression techniques. Earth Surf. Dyn. 5 (4), 821–839. https://doi.org/10.5194/esurf-5-821-2017. Shean, D., 2017. High mountain Asia 8-meter DEM mosaics derived from optical imagery, version 1. https://doi.org/ 10.5067/kxovq9l172s2. Shean, D.E., Alexandrov, O., Moratto, Z.M., Smith, B.E., Joughin, I.R., Porter, C., Morin, P., 2016. An automated, open-source pipeline for mass production of digital elevation models (DEMs) from very-high-resolution commercial stereo satellite imagery. ISPRS J. Photogramm. Remote Sens. 116, 101–117. https://doi.org/10.1016/j. isprsjprs.2016.03.012. Shimada, M., Tadono, T., Rosenqvist, A., 2010. Advanced land observing satellite (ALOS) and monitoring global environmental change. Proc. IEEE 98 (5), 780–799. https://doi.org/10.1109/JPROC.2009.2033724. Shuman, C.A., Zwally, H.J., Schutz, B.E., Brenner, A.C., DiMarzio, J.P., Suchdeo, V.P., Fricker, H.A., 2006. ICESat Antarctic elevation data: preliminary precision and accuracy assessment. Geophys. Res. Lett. 33(7). https://doi.org/ 10.1029/2005GL025227. Simard, M., Neumann, M., Buckley, S., 2016. Validation of the new SRTM digital elevation model (NASADEM) with ICESAT/GLAS over the United States. In: 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 3227–3229. Singh, M.K., Gupta, R.D., Snehmani, Bhardwaj, A., Ganju, A., 2016. Scenario-based validation of moderate resolution DEMs freely available for complex Himalayan Terrain. Pure Appl. Geophys. 173 (2), 463–485. https://doi.org/ 10.1007/s00024-015-1119-5. Skolnik, M.I., 1981. Introduction to Radar Systems, second ed. McGraw-Hill, Singapore. Slatton, K.C., Carter, W.E., Shrestha, R.L., Dietrich, W., 2007. Airborne laser swath mapping: achieving the resolution and accuracy required for geosurficial research. Geophys. Res. Lett. 34(23). https://doi.org/ 10.1029/2007GL031939. Smith, B., Sandwell, D., 2003. Accuracy and resolution of shuttle radar topography mission data. Geophys. Res. Lett. 30 (9), 1467. https://doi.org/10.1029/2002GL016643. Snyder, J.P., 1987. Map projections: A working manual. U.S. Government Printing Office, Washington, DC. http:// pubs.er.usgs.gov/publication/pp1395. USGS Numbered Series 1395. Sofia, G., Bailly, J., Chehata, N., Tarolli, P., Levavasseur, F., 2016. Comparison of Pleiades and LiDAR digital elevation models for terraces detection in farmlands. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 9 (4), 1567–1576. https://doi.org/10.1109/JSTARS.2016.2516900. Stevenson, J.A., Sun, X., Mitchell, N.C., 2010. Despeckling SRTM and other topographic data with a denoising algorithm. Geomorphology 114 (3), 238–252. https://doi.org/10.1016/j.geomorph.2009.07.006. Studinger, M., Koenig, L., Martin, S., Sonntag, J., 2010. Operation icebridge: using instrumented aircraft to bridge the observational gap between icesat and icesat-2. In: 2010 IEEE International Geoscience and Remote Sensing Symposium, pp. 1918–1919. Suwandana, E., Kawamura, K., Sakuno, Y., Kustiyanto, E., Raharjo, B., 2012. Evaluation of ASTER GDEM2 in comparison with GDEM1, SRTM DEM and topographic-map-derived DEM using inundation area analysis and RTK-dGPS data. Remote Sens. 4 (8), 2419–2431. https://doi.org/10.3390/rs4082419. Tachikawa, T., Hato, M., Kaku, M., Iwasaki, A., 2011. Characteristics of ASTER GDEM version 2. In: 2011 IEEE International Geoscience and Remote Sensing Symposium, pp. 3657–3660. Tadono, T., Ishida, H., Oda, F., Naito, S., Minakawa, K., Iwamoto, H., 2014. Precise global DEM generation by ALOS PRISM. ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci. 4, 71–76. https://doi.org/10.5194/isprsannals-II4-71-2014. Tadono, T., Takaku, J., Tsutsui, K., Oda, F., Nagai, H., 2015. Status of “ALOS world 3D (AW3D)” global DSM generation. In: 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 3822–3825. Tadono, T., Nagai, H., Ishida, H., Oda, F., Naito, S., Minakawa, K., Iwamoto, H., 2016. Generation of the 30 M-Mesh global digital surface model by ALOS PRISM. ISPRS Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 41B4, 157–162. https://doi.org/10.5194/isprs-archives-XLI-B4-157-2016. Tadono, T., Takaku, J., Ohgushi, F., Doutsu, M., Kobayashi, K., 2017. Updates of “AW3D30” 30 M-MESH global digital surface model dataset. In: 2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 5656–5657.

References

127

Takaku, J., Tadono, T., 2017. Quality updates of “AW3D” global DSM generated from ALOS PRISM. In: 2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 5666–5669. Takaku, J., Futamura, N., Iijima, T., Tadono, T., Shimada, M., Shibasaki, R., 2004. High resolution DEM generation from ALOS PRISM data—simulation and evaluation. In: IGARSS 2004. 2004 IEEE International Geoscience and Remote Sensing Symposium., vol. 7. pp. 4548–4551. Takaku, J., Tadono, T., Tsutsui, K., 2014. Algorithm development of high resolution global DSM generation by ALOS prism. In: 2014 IEEE Geoscience and Remote Sensing Symposium, pp. 4784–4787. Takaku, J., Tadono, T., Tsutsui, K., Ichikawa, M., 2016. Validation of “AW3D” global DSM generated from ALOS PRISM. ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci. 4, 25–31. https://doi.org/10.5194/isprs-annalsIII-4-25-2016. Takaku, J., Tadono, T., Tsutsui, K., Ichikawa, M., 2018. Quality improvements of “AW3D” global DSM derived from ALOS PRISM. In: IGARSS 2018—2018 IEEE International Geoscience and Remote Sensing Symposium, pp. 1612–1615. Tebaldini, S., Rocca, F., 2012. Multibaseline polarimetric SAR tomography of a boreal forest at P- and L-bands. IEEE Trans. Geosci. Remote Sens. 50 (1), 232–246. https://doi.org/10.1109/TGRS.2011.2159614. Toutin, T., 2001. Elevation modelling from satellite visible and infrared (VIR) data. Int. J. Remote Sens. 22 (6), 1097–1125. https://doi.org/10.1080/01431160117862. Vaze, J., Teng, J., Spencer, G., 2010. Impact of DEM accuracy and resolution on topographic indices. Environ. Model. Softw. 25 (10), 1086–1098. https://doi.org/10.1016/j.envsoft.2010.03.014. Verdin, K.L., Verdin, J.P., 1999. A topological system for delineation and codification of the Earth’s river basins. J. Hydrol. 218 (1), 1–12. https://doi.org/10.1016/S0022-1694(99)00011-6. Walker, J.P., Willgoose, G.R., 1999. On the effect of digital elevation model accuracy on hydrology and geomorphology. Water Resour. Res. 35 (7), 2259–2268. https://doi.org/10.1029/1999WR900034. Warren, S.D., Hohmann, M.G., Auerswald, K., Mitasova, H., 2004. An evaluation of methods to determine slope using digital elevation data. Catena 58 (3), 215–233. https://doi.org/10.1016/j.catena.2004.05.001. Wessel, B., Huber, M., Wohlfart, C., Marschalk, U., Kosmann, D., Roth, A., 2018. Accuracy assessment of the global TanDEM-X digital elevation model with GPS data. ISPRS J. Photogramm. Remote Sens. 139, 171–182. https://doi. org/10.1016/j.isprsjprs.2018.02.017. Winsemius, H.C., Ward, P.J., Gayton, I., ten Veldhuis, M.-C., Meijer, D.H., Iliffe, M., 2019. Commentary: the need for a high-accuracy, open-access global DEM. Front. Earth Sci. 7. https://doi.org/10.3389/feart.2019.00033. Wolf, A.T., Natharius, J.A., Danielson, J.J., Ward, B.S., Pender, J.K., 1999. International river basins of the world. Int. J. Water Resour. Dev. 15 (4), 387–427. https://doi.org/10.1080/07900629948682. Wulder, M.A., White, J.C., Nelson, R.F., Naesset, E., Orka, H.O., Coops, N.C., Hilker, T., Bater, C.W., Gobakken, T., 2012. Lidar sampling for large-area forest characterization: a review. Remote Sens. Environ. 121, 196–209. https:// doi.org/10.1016/j.rse.2012.02.001. Yakubu, C.I., Ayer, J., Laari, P.B., Amponsah, T.Y., Hancock, C.M., 2019. A mutual assessment of the uncertainties of digital elevation models using the triple collocation technique. Int. J. Remote Sens. 40, 1–14. https://doi.org/ 10.1080/01431161.2019.1579388. Yamazaki, D., Ikeshima, D., Tawatari, R., Yamaguchi, T., O’Loughlin, F., Neal, J.C., Sampson, C.C., Kanae, S., Bates, P.D., 2017. A high-accuracy map of global terrain elevations. Geophys. Res. Lett. 44 (11), 5844–5853. https://doi.org/10.1002/2017GL072874. Yap, L., Kande, L.H., Nouayou, R., Kamguia, J., Ngouh, N.A., Makuate, M.B., 2018. Vertical accuracy evaluation of freely available latest high-resolution (30 m) global digital elevation models over Cameroon (Central Africa) with GPS/leveling ground control points. Int. J. Digital Earth 12, 1–25. https://doi.org/10.1080/17538947.2018.1458163. Yu, A.W., Krainak, M.A., Abshire, J.B., Cavanaugh, J., Valett, S., Ramos-Izquierdo, L., 2010. Airborne lidar simulator for the lidar surface topography (LIST) mission. In: 25th International Laser Radar Conference; 5–9 July 2010; St. Petersburg, Russia. Zebker, H., Goldstein, R., 1986. Topographic mapping from interferometric synthetic aperture radar observations. J. Geophys. Res. Solid Earth Planets 91 (B5), 4993–4999. https://doi.org/10.1029/JB091iB05p04993. Zevenbergen, L.W., Thorne, C.R., 1987. Quantitative analysis of land surface topography. Earth Surf. Process. Landf. 12 (1), 47–56. https://doi.org/10.1002/esp.3290120107. Zhang, C., Fraser, C., 2008. Generation of digital surface model from high resolution satellite imagery. (Pt. B1) In: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. pp. 784–785. Beijing.

128

4. Topographic data from satellites

Zhang, W., Montgomery, D.R., 1994. Digital elevation model grid size, landscape representation, and hydrologic simulations. Water Resour. Res. 30 (4), 1019–1028. https://doi.org/10.1029/93WR03553. Zhang, X., Drake, N.A., Wainwright, J., Mulligan, M., 1999. Comparison of slope estimates from low resolution DEMs: scaling issues and a fractal method for their solution. Earth Surf. Process. Landf. 24 (9), 763–779. https://doi.org/ 10.1002/(SICI)1096-9837(199908)24:9¡763::AID-ESP9¿3.0.CO;2-J. Zwally, H.J., Schutz, B., Abdalati, W., Abshire, J., Bentley, C., Brenner, A., Bufton, J., Dezio, J., Hancock, D., Harding, D., Herring, T., Minster, B., Quinn, K., Palm, S., Spinhirne, J., Thomas, R., 2002. ICESat’s laser measurements of polar ice, atmosphere, ocean, and land. J. Geodyn. 34 (3–4), 405–445. https://doi.org/10.1016/S0264-3707 (02)00042-X.

C H A P T E R

5 Linking life and landscape with remote sensing David T. Milodowskia,b, Steven Hancocka, Sonia Silvestric,d, Simon M. Mudda a

University of Edinburgh, School of GeoSciences, Edinburgh, United Kingdom bUniversity of Edinburgh, National Centre for Earth Observation, Edinburgh, United Kingdom cUniversity of Bologna, Department of Biological, Geological and Environmental Sciences, Bologna, Italy dDuke University, Nicholas School of the Environment, Durham, NC, United States

O U T L I N E 1 Introduction 2 Linking remote sensed data to life and landscapes 2.1 Erosive, depositional, and constructive processes modulated by biota 2.2 Life and landscape patterns 2.3 Measureable vegetation properties 2.4 Soils and belowground organic carbon

130 131

131 132 133 137

3 Passive remote sensing methods 3.1 Vegetation indicators from passive instruments 3.2 Coarse resolution passive sensors 3.3 Medium and fine resolution passive sensors

137

4 Radar 4.1 Satellite-based radar systems

141 143

Developments in Earth Surface Processes, Volume 23 https://doi.org/10.1016/B978-0-444-64177-9.00005-9

138 139 140

5 Lidar 5.1 A primer on lidar remote sensing 5.2 Quantifying canopy structure with airborne lidar 5.3 Spaceborne lidar 5.4 Data fusion

143

6 Airborne electromagnetics

161

7 Conclusions 7.1 Finding the right sensor 7.2 The importance of scale 7.3 Trade-offs between resolution and spatial coverage 7.4 Future outlook

164 164 165

Acknowledgments

165

References

166

129

143 145 154 161

165 165

© 2020 Elsevier B.V. All rights reserved.

130

5. Linking life and landscape with remote sensing

1 Introduction Of all the known planetary bodies, Earth is unique in supporting life, at least as far as we know. Plants colonized the terrestrial surface of the Earth somewhere between 420 and 380 million years before present (Kenrick and Crane, 1997), and other forms of terrestrial life followed. Living organisms can affect our planet’s surface in a multitude of ways. Microbes and fungi can affect chemical weathering (e.g., Barker et al., 1998; Banfield et al., 1999), plants can both add cohesion to soils (e.g., Schmidt et al., 2001; Pollen and Simon, 2005), protect them from runoff-driven erosion (e.g., Gyssels et al., 2005; Dunne et al., 2010), and also encourage sediment movement through root growth and tree toppling (e.g., Gabet et al., 2003), to give just a few examples. Many animals also affect the surface of the planet. For example, beavers create dams (e.g., Butler and Malanson, 2005); gophers, moles, and wombats burrow through the soil (e.g., Gabet, 2000; Yoo et al., 2005; Wilkinson et al., 2009); and, of course, humans move millions of tons of soil and Earth in an effort to grow food, build infrastructure, and create shelter (e.g., Wilkinson and McElroy, 2007). In addition, the physical, hydrological, and geochemical characteristics of landscapes impose a template on which terrestrial ecosystems develop, placing major controls on the availability of moisture and nutrients, local microclimate and disturbance regimes (e.g., Porder et al., 2004; Detto et al., 2013; Swetnam et al., 2017; Jucker et al., 2018a). Landscapes should therefore be considered as coupled ecogeomorphic systems. If we wish to understand how life affects the evolution of Earth’s surface, one approach is to try to quantify the properties of living things that are most likely to control the evolution of landscapes. We may start by identifying the different ways organisms might affect the landscapes. Corenblit et al. (2011) separated landform modification by organisms into two broad categories: autogenic, where organisms modify the landscape through their own tissues, either living or dead, and allogenic, where organisms modify nonliving materials and their fluxes through either mechanical or chemical means. The former broadly contains all bioconstruction that directly builds landscapes (e.g., formation of reefs or raised peat domes) (e.g., Clymo, 1984; Toomey et al., 2016), whereas the latter includes processes such as erosion directly caused by organisms (e.g., Mauri et al., 2019), bioconstruction that alters the transport properties of nonliving sediments (e.g., Noffke et al., 2001), stabilization of nonliving materials via various mechanisms that add cohesion or resistance (e.g., Kemper et al., 1989; Fang et al., 2013), and bioturbation, where living organisms perturb nonliving sediments (e.g., Wilkinson et al., 2009). Bioturbation commonly leads to bioerosion on sloping surfaces (e.g., Gabet et al., 2003). A number of review papers have summarized these processes and listed both organisms that serve as geomorphic agents, as well as the different environments in which these organisms are active (Gurnell et al., 2002; Naylor et al., 2002; Corenblit et al., 2011; Gurnell, 2013; Fei et al., 2014; Pawlik et al., 2016; Brantley et al., 2017). Our aim here is not to cover the entirety of biogeomorphology, but rather to give some examples of how remote sensing can be used to quantify features of organisms that can affect geomorphic processes. Many forms of life, although important for landscape evolution, are difficult to detect using remote sensing methods. We therefore direct our focus on vegetation, which has the convenient property that it has aboveground components that can be seen

2 Linking remote sensed data to life and landscapes

131

using remote sensing instruments. The structure of this chapter is as follows: (i) we provide an overview of landscape-vegetation feedbacks to illustrate the importance of vegetation, and thus highlight why one should consider remote sensing of vegetation within geomorphological studies; (ii) we introduce a series of remote sensing technologies, from spaceborne to aerial and ground-based platforms, with a focus on light detection and ranging (lidar) due to its increasing ubiquity in geomorphologic analyses and the unparalleled level of detail with which both the vegetation and underlying topography can be mapped; (iii) we conclude the chapter by introducing a new approach that makes use of geophysical methods from airborne platforms to explore the characteristics of the soil below the terrain surface with implications for linking Earth surface processes with belowground organic matter and carbon content.

2 Linking remote sensed data to life and landscapes Why should a geomorphologist bother quantifying vegetation? As geomorphologists, we characteristically consider vegetation as an annoyance: it obscures the morphology of the landscape and the topographic signatures of the processes that sculpt them. Indeed one of the revolutionary aspects of lidar remote sensing is the ability to “strip” the vegetation from the landscape to provide a high-resolution model of the underlying topography. In the introduction we eluded to a multitude of mechanistic pathways through which vegetation directly and indirectly influences geomorphic processes. Before delving into the more technical aspects of remote sensing of vegetation, we expand upon this review to highlight why we should consider quantifying vegetation as geomorphologists, and provide some examples of the types of characteristics that we may wish to measure.

2.1 Erosive, depositional, and constructive processes modulated by biota The importance of vegetation in controlling erosion has been recognized for over a century (e.g., Gilbert, 1877; Ellison, 1945; Langbein and Schumm, 1958), and widely used models of soil erosion dating to the 1970s such as the Universal Soil Loss Equation (USLE; Wischmeier et al., 1971) or the revised USLE (RUSLE, Renard et al., 1997) include coefficients for land cover type and fraction of land cover. Many soil erosion models have emerged over the last few decades, but they all contain some description of land cover (e.g., de Vente et al., 2013). Some of these models include explicit vegetation properties such as leaf area index (LAI) (e.g., de Roo et al., 1996; Pelletier, 2012), with other models using plant characteristics such as proxy data for biomass (Kirkby and Neale, 1987), or proxy measures of vegetation abundance (Haboudane et al., 2002). Other models try to quantify landscape evolution by using sediment flux laws (cf. Dietrich et al., 2003) that involve a biotic component (e.g., Collins et al., 2004; Istanbulluoglu and Bras, 2005; Saco et al., 2007), which in some cases may even drive landscape evolution of the system (Ingram, 1982; Clymo, 1984). In ecosystems with abundant belowground biomass, vegetation can directly contribute to the construction of the landscape, providing a large portion of the soil volume, as, for example, in salt marshes or peatlands. In fact, the soil organic matter in

132

5. Linking life and landscape with remote sensing

marshes can reach 30%–40% by weight of the soil (Roner et al., 2016), whereas in peatlands this value is easily greater than 90% (Hergoualc’h and Verchot, 2011). Such organic content partially consists of live plant tissues but its large majority is made by the dead organic matter accumulated over time and remained undecomposed due to anoxic conditions. In these environments, geomorphological shapes and patterns reflect the interactions between the underlying terrain form, vegetation, climate, and hydrology (Rydin and Jeglum, 2013). Various metrics involving vegetation and sometimes animals (gophers appear to hold a particular appeal to geomorphologists, e.g., Yoo et al., 2005; Gabet et al., 2014) appear in models that describe the co-evolution of biota and landscapes: examples include the fraction of vegetation cover and biomass, both of which are amenable to remote sensing through a number of different platforms operating at a range of spatial scales. In fluvial geomorphology, plants play a major role in stabilizing channel banks (e.g., Millar, 2000; Pollen and Simon, 2005) and are thus instrumental in controlling both meandering (e.g., Perucca et al., 2007) and the emergence and nature of channel braiding (e.g., Nanson and Knighton, 1996). Many properties of vegetation may affect fluvial processes. Typical factors influencing flow and channel behavior are vegetation density (e.g., Bennett et al., 2002). Other properties important for the interaction between fluvial processes and vegetation include the rigidity of plant stems controls how they respond to water flowing over and around them (Edmaier et al., 2011; Nepf, 2012), and this rigidity depends on the plant species (Nepf, 2012). The identification of plant species from remote sensing data has as rich a literature as that on quantifying biomass from remotely sensed data, and numerous reviews are available (e.g., Kerr and Ostrovsky, 2003; Yu et al., 2006). Multiple instruments have been used, from imagery analysis (e.g., Jang and Shimizu, 2007), hyperspectral data (e.g., Silvestri et al., 2003), and lidar (e.g., Bertoldi et al., 2011). Many hillslope processes, such as the rates of bioturbation and slope failure, are linked to vegetation (e.g., Gabet et al., 2003; Marston, 2010; Osterkamp et al., 2011; Pawlik, 2013; Amundson et al., 2015). One of the key factors in shallow slope failure is the cohesion provided by plant roots (e.g., Schmidt et al., 2001; Reubens et al., 2007; Stokes et al., 2009). Different plant species have roots with different tensile strength (e.g., Cohen et al., 2011; Hales, 2018), so species identification can be important in linking landsliding with vegetation properties. Root density is linked to aboveground biomass (AGB) (e.g., Cairns et al., 1997). In addition, the diameter of the root ball in trees is thought to be roughly proportional to the crown diameter (e.g., Gilman, 1989; Danjon et al., 2005). The crown diameter of trees may be detected using lidar data (see Section 5). Less attention has been paid to the influence of geomorphic processes on the density and species distribution of organisms, but some examples exist. For example, several studies have documented how increased rates of sedimentation can stimulate plant growth in coastal settings (e.g., Deng et al., 2008; Baustian and Mendelssohn, 2015), whereas in upland landscapes, accelerated rates of erosion can reduce biomass densities (e.g., Ravi et al., 2010; Milodowski et al., 2015). Such linkages highlight the potential for eco-geomorphic feedbacks that potentially influence the long-term evolution of landscapes.

2.2 Life and landscape patterns Planetary surfaces are rich in repeating patterns, from drainage networks on Titan (e.g., Black et al., 2012) to dunes on Mars (e.g., Bridges et al., 2012). On Earth, the influence of life

2 Linking remote sensed data to life and landscapes

133

on these patterns is not always obvious (Dietrich and Perron, 2006), but certain landscapes have proven to be a rich hunting ground for life-landscape feedbacks. One example is striped or radial patterns that emerge in arid landscapes (von Hardenberg et al., 2001), banded vegetation patterns in salt marshes (e.g., Koppel et al., 2005), string patterns of alternating peat ridges and pools in raised boreal peatlands (Foster et al., 1983), or strange hummocks called mima mounds (Gabet et al., 2014) that are present in temperate regions (Fig. 1). Observations of these patterns originally involved optical data in the visible spectrum, and such data can be used to extract landscape patterns (e.g., Barbier et al., 2006). If one has both patience and time, one can digitize these features by hand based on, say, satellite imagery (e.g., Deblauwe et al., 2008). Multiple algorithms exist to produce vegetation and land cover maps (e.g., Gong et al., 2012), including segmentation algorithms (e.g., Shi and Malik, 2000; Dronova et al., 2012), supervised classification (e.g., Muchoney et al., 2000), and maximum entropy methods (see, e.g., Fourcade et al., 2014) among other methods. Various theories about the formation of these features have been presented (e.g., Rietkerk and Koppel, 2008; Tarnita et al., 2017). Frequently, numerical models of these patterns involve simulation of biomass density, water uptake by plants, and response of living organisms to water and nutrient stresses (e.g., Ingram, 1982; Saco et al., 2007; Borgogno et al., 2009; Meron, 2012; Saco and Heras, 2013; Gabet et al., 2014). Quantifying these properties from remote sensing data has occupied both ecologists and remote sensing scientists for decades and a rich literature exists on these topics.

2.3 Measureable vegetation properties For detection of both patterns emerging from feedbacks between life and landscape, and in order to assess the influence of biota on geomorphic processes, we need data that can be ingested by either models or different analytical techniques. The remainder of this chapter is focused on the kinds of remote sensing data that are available to geomorphologists that wish to use properties of living things in their models and analyses. We focus on vegetation, as it is most amenable to capture by remote sensing instruments. Our review briefly explores detection of vegetation through spaceborne instruments. We then move on in greater detail to high-resolution imaging of vegetation using airborne instruments. For geomorphic applications, we commonly wish to characterize properties of plants at the scale of meters to tens of meters, which remains a challenge from spaceborne platforms. The gold standard, in our opinion, for characterizing the plant canopy is airborne lidar, as it allows spatially continuous measures of the complete canopy structure at centimeter-scale precision. But before we proceed with details on vegetation characterization with lidar, we first take a whirlwind tour of other data products that can be used to quantify biomass, species distributions, and other properties of terrestrial vegetation. Here, we focus on remote sensing datasets that are likely to be of most use to geomorphologists. As we have seen in the previous sections, a few key parameters occur that have been used in both models fusing geomorphology and vegetation, as well as analyses that link vegetation to geomorphic processes. Fractional cover of vegetation plays a role in many erosion studies, as does categorization of vegetation type. Another important metric that is used in modeling studies is the biomass of vegetation. As we have seen in slope stability studies, belowground biomass

134

5. Linking life and landscape with remote sensing

FIG. 1 Three vegetated landscapes. (A) Fairy circles in Namibia at 24.4172° S, 15.6675° W. (B) A mosaic of patterns in the Scheldt estuary c. Mima mounds near the San Andreas Fault in the Carrizo Plain, California.

2 Linking remote sensed data to life and landscapes

135

is crucial to understanding root cohesion but it is often inferred from AGB. Aboveground biomass, in turn, is often derived from canopy height or other quantities. We briefly review widely available datasets categorized by the instruments used to collect the data. Numerous authors have demonstrated that remote sensing can quantify a number of vegetation properties, also known as biophysical variables, at landscape to global scales (Exbrayat et al., 2019). These include estimates of continuous biophysical variables, including canopy cover, crown cover, leaf (or plant) area index, vertical foliage profiles, height, biomass, basal area, stand density, and species distribution, in addition to categorical land cover classifications. Formal definitions of these variables, with discussions of the maturity of their retrieval and accuracy assessments, are maintained by the CEOS Land Product Validation working groups (e.g., Herold et al., 2006). Brief definitions are given below: Land cover classifications are perhaps the most widely used remote sensing of vegetation product and often underpin remote sensing retrievals of other biophysical variables (Go´mez et al., 2016). In a land cover classification, the land surface type (e.g., open water or bare soil) or plant functional type (e.g., grassland or broadleaf forest) category of each point on the ground is defined. These determine the presence or absence of a certain type of vegetation in a given area, or the fractional coverage of each land cover type. A wide range of land cover classification maps are available, derived from different remote sensing instruments using different methods and with different land cover class categories. Land cover classification data are a key input into many soil erosion models (see review in de Vente et al., 2013), including the factorial scoring model (FSM; Verstraeten et al., 2003), the SPAtially distributed scoring model (SPADS; de Vente et al., 2008), SSY (Cerdan et al., 2010), WATSEM-SEDEM (Van Oost et al., 2000), the Pan-European Soil Erosion Risk Assessment (PESERA; Kirkby et al., 2008), and the Soil and Water Assessment Tool (SWAT; Arnold et al., 1998), and land cover changes have been attributed to large-scale changes in sediment yields (e.g., Deng et al., 2012; Wang et al., 2016). Canopy cover is the fraction of vegetation vs sky that can be seen when looking up through the canopy (or the fraction of vegetation vs ground that can be seen looking down), and is equal to one minus the gap fraction (e.g., Korhonen and Morsdorf, 2014). It is normally either defined looking vertically with an orthogonal projection, or over a hemispherical view, and care should be taken to ensure that the version a dataset uses is clearly defined (Tang et al., 2019b). The former is what most remote sensing instruments measure whereas the latter is what is traditionally measured by ground-based methods. The canopy cover plays a role in determining erosion from processes such as rainsplash (e.g., Dunne et al., 2010), wind erosion (e.g., Li et al., 2007), and water erosion (e.g., Dura´n Zuazo and Rodrı´guez Pleguezuelo, 2008) and is used in soil erosion models such as PESERA (Kirkby et al., 2008), the Limburg soil erosion model (LISEM; De Roo et al., 1996), and in modified version of the revised universal soil loss equation (e.g., Zhou et al., 2008). Vegetation cover is also a key parameter in some landscape evolution models that connect vegetation cover to changing channel erodibility (e.g., Istanbulluoglu and Bras, 2005). Crown cover is the fraction of ground covered by vegetation crown extents. This is similar to canopy cover, but ignores the small gaps within vegetation crowns. Some soil erosion studies € make a distinction between canopy cover and crown cover (e.g., Ozhan et al., 2005), but this is rare.

136

5. Linking life and landscape with remote sensing

Leaf area index (LAI) is defined as half the total surface area of photosynthetically active material (the leaf area) per unit ground area (Fernandes et al., 2014) and is expressed in m2/m2. The separation of photosynthetically active and nonphotosynthetically active material is nontrivial, so an alternative and easier measure is plant area index (PAI), which is half the total surface area of all plant material per unit ground area (Woodgate et al., 2016; Vincent et al., 2017). These are a two-dimensional measurement but can be extended to three dimensions by defining the leaf area per unit volume, the leaf area density (LAD), expressed in m2/m3. This can either be as a function of height, forming a foliage profile (Tang et al., 2012), or in a set of cubes, known as a voxel grid (Hosoi and Omasa, 2006). The LAI is used as an input into some soil erosion models such as the model developed by Pelletier (2012) as well as LISEM (De Roo et al., 1996). The amount of leaves within a canopy places direct controls on the rates of interception and evapotranspiration, thus recognition exists that LAI is related to the uptake of water by plants, and therefore affects runoff and sediment transport (e.g., Caylor et al., 2005). LAI has also been used to quantify forest characteristics for landslide modeling (e.g., Band et al., 2012). Height, although simple conceptually, poses some difficulty for remote sensing instruments because one needs information about both the elevation of the top of the plant canopy and the ground. Lidar has been used for this purpose (e.g., Simard et al., 2011); whereas passive optical instrument struggle as they can only “see” the top of the canopy. Some radar wavelengths are also able to penetrate the plant canopy (e.g., Bergen et al., 2009), in particular the longer wavelengths (L or P bands) which return data not just from the plant canopy but also from branches and stems (Mcdonald et al., 1990). Geomorphic models tend not to use canopy height as a functional metric, but canopy height is sometimes used to quantify biomass (see below) and is used in models of tree toppling that can result in sediment transport (e.g., Constantine et al., 2012). Biomass is the dry mass of living material. Remote sensing generally estimates the aboveground biomass density (AGBD), which is the dry mass of aboveground plant material per unit ground area (generally expressed in Mg/ha). Numerous techniques have been used to take data from remote sensing instruments and convert them to biomass. Some techniques use relationships between physical characteristics of plants such as canopy height and biomass derived from ground measurements (e.g., Feldpausch et al., 2012), whereas others use data compilations of the relationship between biomass and measurements from instruments, such as radar backscatter (e.g., Mitchard et al., 2009; Saatchi et al., 2011). Some numerical models of landscape evolution and pattern formation use biomass density as a key parameter (e.g., Kirwan and Murray, 2007; Saco et al., 2007); others link biomass to parameters that affect hydrodynamics (e.g., Mudd et al., 2010; Camporeale et al., 2013), whereas other models contain an empirical coefficient linked to vegetation density, which is broadly defined but most closely maps to biomass density (e.g., Collins et al., 2004). Other properties that are correlated with biomass are basal area, which is the total area covered by stems and trunks of plants, and stand density is the number of stems or trunks per unit area. Species distribution Remote sensing of different plant species has a rich history and many different methods have been used to differentiate one species from another (e.g., Kerr and Ostrovsky, 2003). Different plant species play a role in geomorphic processes, and so some geomorphic investigations have used remote sensing data of the spatial distribution of

3 Passive remote sensing methods

137

different species. For example, in salt marsh ecosystem, different species have different hydrodynamic and sediment trapping properties (e.g., Nikora, 2010; Nepf, 2012), and they are indicative of different marsh surface elevations (Silvestri et al., 2003; Moffett et al., 2010). In steep forested landscapes, different tree species provide different levels of root stabilization (e.g., Bischetti et al., 2007; Hales et al., 2009). Because vegetation can be highly heterogeneous, detecting plant species with lower resolutions (e.g., pixels >30 m) is very challenging, but a number of sensors, primarily multi- and hyperspectral passive sensors, have been successfully used to map the spatial distribution of species over local areas (e.g., Gillespie et al., 2008). Belowground properties For obvious reasons, mapping biophysical parameters relating to the belowground properties of vegetation, such as root biomass and rooting depth, is inherently challenging for remote sensing techniques. Estimating spatial distributions of belowground properties of vegetation has traditionally relied on assumed scaling relationships with aboveground properties, for example, linking above/belowground biomass (Hwang et al., 2015), or through assimilating remotely sensed data into a process-based models (Knorr et al., 2010; Smallman et al., 2017).

2.4 Soils and belowground organic carbon The large majority of terrestrial carbon is stored belowground (Kayler et al., 2017). More importantly, above- and belowground biomass are not reliable indicators of the distribution of the soil organic carbon stored in soils (Scharlemann et al., 2014), which is in fact mainly stored in permafrost regions and boreal moist soils, where the AGB is limited, in contrast with tropical forested ecosystems. While allometric relationships can be quite successful in estimating belowground biomass using AGB as a proxy, new methods are needed to estimate the organic matter and the total carbon pool stored in soils. Recent approaches using geophysical sensors carried by aircrafts and helicopters appear very promising methods for exploring soils directly below the terrain surface, where the organic matter accumulates and is preserved by wet soil conditions, as, for example, in peatlands (Silvestri et al., 2019a,b) or is preserved by very low temperatures as in permafrost (Pastick et al., 2013). This receives our attention in the latter part of the chapter.

3 Passive remote sensing methods Passive optical sensors are the most widely available and longest heritage instruments. These sensors passively detect and record incoming electromagnetic energy, as opposed to active sensors that emit energy and record how much of this emitted energy is reflected. Different kinds of sensors record different portions of the electromagnetic spectrum. Your eyes detect visible light, which has wavelengths from 380 to 700 nm. The humble digital camera will record specific wavelengths within the range red (635–700 nm), green (520–560 nm), and blue (450–490 nm) spectral ranges, and can easily be modified to record within the infrared range (900 nm), and therefore quantify normalized difference vegetation index (NDVI) (see Section 3.1). In a remote sensing image, these wavelengths will each receive a band: i.e., an RGB image will have three bands, each indicating the intensity of light in the three wavelengths.

138

5. Linking life and landscape with remote sensing

Most remote sensing instruments record a far greater range of electromagnetic wavelengths than those in the visible spectrum. These have been broadly categorized in “multispectral” and “hyperspectral” instruments. The difference between these is simply the number of bands and their spectral resolution. Multispectral instruments can have up to a few tens of bands. For example, the Landsat 8 instrument has 11 bands ranging from 430–450 nm (in the blue wavelength) to 11.5–12.51 μm in the infrared wavelength (Roy et al., 2014). The MODIS instrument (Barnes et al., 1998) has 36 bands over a similar wavelength range (from a 620–670 nm band to a 14–14 μm band). On the other hand, hyperspectral instruments can have hundreds of very narrow bands.

3.1 Vegetation indicators from passive instruments Remote sensing scientists have spent the last few decades exploring how different wavelengths correspond to different features of interest (e.g., vegetation, atmospheric properties, properties of water), and the interested reader can find many books on the subject. Here, we focus on the most widely used indicators of vegetation. We begin with perhaps the most widely used vegetation indicator: the normalized difference vegetation index, or NDVI. This index came about because of the observation that energy in the red wavelength is absorbed by chlorophyll whereas near-infrared energy is strongly reflected by leaves (e.g., Justice et al., 1985). As leaf area increases, the reflectance of red wavelengths decreases, because a larger amount of energy at these wavelengths is absorbed for photosynthesis. Conversely, the infrared reflectance increases. Thus, early workers proposed that an index composed of the ratio of radiance from these wavelengths could be used as a proxy measure for biophysical variables (e.g., canopy cover or LAI; Tucker, 1979), although in practice this signal saturates as vegetation density increases (becoming insensitive once LAI reaches 4, or when canopy cover reaches 70%–80%). The ratio between these wavelengths became known as the NDVI and is now widely calculated with data derived from multiple satellites and instruments (Brown et al., 2006). For instruments with both near-infrared and red bands, NDVI is calculated using NDVI ¼ (ρnir  ρir)/(ρnir + ρir) where ρ is the amplitude of the spectral band and the subscripts nir and ir indicate near-infrared and infrared, respectively. NDVI depends on the sun and viewer angle through the bidirectional reflectance distribution function (BRDF), which is a function that describes how light is reflected from opaque surfaces. It also depends on the sensor wavebands. This means that NDVI from, say, MODIS is not directly comparable to AVHRR, for example (different bands), and NDVI from different AVHRR and Landsat missions can be different (different orbits causing different sun angles, and degrading sensors over time). These discrepancies have resulted in efforts to overcome these differences in order to produce harmonized products. For example, the GIMMS dataset attempted to correct the entire AVHRR archive, covering multiple satellites, to produce a harmonized dataset (Tucker et al., 2005), whereas MODIS uses a BRDF correction to allow comparison between sensors (MODIS Terra and MODIS Aqua) and time of year (Schaaf et al., 2002). Because the calculation of the NDVI is affected by the type and accuracy of the atmospheric correction (e.g., Gao, 1996) and canopy background (e.g., Liu and Huete, 1995), alternative

3 Passive remote sensing methods

139

metrics have been proposed that aim to minimize these errors. A modification of NDVI, called the enhanced vegetation index (EVI), includes adjustments for atmospheric conditions and a soil adjustment, is also routinely collected, although evidence exists that this is more sensitive to topographic effects than NDVI (e.g., Matsushita et al., 2007) and it also requires additional bands that are not available on all passive optical sensors, such as AVHRR. Since the development of NDVI, wide ranges of two-band vegetation indices have been developed (e.g., Gong et al., 2003), but NDVI is still the most widely used vegetation index. Early attempts to quantify LAI simply related measured LAI values to NDVI values derived from remote sensing, the relationship of which differs between vegetation types (e.g., Myneni et al., 1997; Steltzer and Welker, 2006). More recent LAI data products have involved implementing process models that simulate how radiation is reflected for different scenarios of LAI and leaf and soil reflectances (e.g., Knyazikhin et al., 1998). These algorithms underlie data products from widely used instruments such as MODIS (e.g., Myneni et al., 2002). For more information on the methods to calculate LAI from remote sensing instruments, see the review by Zheng and Moskal (2009). Another commonly reported metric is the fractional vegetation cover (FVC). Early attempts at producing global FVC maps involved processing of NDVI data, where the fractional vegetation cover was calculated (Gutman and Ignatov, 1998; Zeng et al., 2000) as FVC ¼ ðNDVI  NDVIsÞ=ðNDVIv  NDVIsÞ

(1)

Authors have noted that calculation of fractional vegetation coverage can be skewed by changes to NDVI values with different pixel sizes, and so have proposed to use scaleinvariant methods to determine fractional vegetation coverage (e.g., Jiang et al., 2006).

3.2 Coarse resolution passive sensors Cameras mounted on aircraft have been around for many years, but our discussion focuses on satellites as these allow very large regions to be mapped. We consider first instruments that record at coarse resolution (we define this as 250 m per pixel or greater) as these instruments were the first that allowed true global vegetation data products. We then move on to finer resolution instruments. We begin by noting that the instrument that demonstrated the utility of passive instruments, and laid the groundwork for all future missions, was the Landsat series of satellites. The first was launched in 1972 by NASA (originally its name was ERTS-1 satellite) and carried two instruments: a camera system named Return Beam Vidicon (RBV), and the multispectral scanner system (MSS). An interesting historical note is that the MSS was one of the very earliest digital cameras (predating Kodak’s 1975 camera) and was a one pixel camera, using a complex scanning mechanism to build up an image. The superiority of the MSS instrument became clear very soon, with its four bands (green, red, and two infrared bands) and 80 m pixel spacing (e.g., Maul and Gordon, 1975). However, the Landsat data archive was not made freely available until 2008 (Woodcock et al., 2008), and as a consequence widespread data processing leading to scientific publications came from a later satellite. Beginning in the mid-1970s, the National Oceanic and Atmospheric Administration (NOAA) mounted a radiometer called the advanced very high resolution radiometer

140

5. Linking life and landscape with remote sensing

(AVHRR) on polar-orbiting meteorological satellites (Cracknell, 1997). This satellite had a number of channels, sampling at a resolution of approximately 1 km, in the wavelengths 0.58–12.50 μm (Cracknell, 1997). Notably, AVHRR had only a single visible band and therefore could not generate RGB images. Nevertheless, what was intended to be an instrument devoted to making meteorological observations became a widely used source of information for global vegetation (Cracknell, 2001). Note that NDVI cannot be calculated using raw data because it is affected by the interference of the atmosphere (see earlier). Because of this different global datasets use a variety of atmospheric corrections. For AVHRR data alone, there are several global datasets with different corrections for sources of error such as aerosols or water vapor: details can be found in Beck et al. (2011). Increasing numbers of satellites collect data from which NDVI and other similar indices can be derived. The moderate resolution imaging spectroradiometer (MODIS) instrument includes 36 spectral bands, many of which can be used for vegetation indices (Barnes et al., 1998; Justice et al., 1998). MODIS data products are widely and freely available from NASA (https://modis.gsfc.nasa.gov/data/, retrieved 22-May-2019) and other data distribution systems at resolutions of 1 km (and coarser) to 250 m; the nearly daily repeats of MODIS satellites are turned into 16 day composites to address problems with cloud cover as well reduce data volumes (e.g., Yengoh et al., 2015). The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) instrument, which was primarily designed to monitor oceanic productivity, has been used to produce a monthly 4-km NDVI dataset between 1997 and 2010 (e.g., Yengoh et al., 2015). Both early SPOT satellites (e.g., SPOT 3) (Saint, 1996) and ENVISAT (Rast et al., 1999) instruments have been used to generate 1-km resolution 10-day datasets (e.g., Yengoh et al., 2015). Large data platforms such as the Google Earth Engine (Gorelick et al., 2017) continue to add global NDVI datasets; a recent addition is data derived from the Visible/Infrared Imager/Radiometer Suite (VIIRS) satellite (Vargas et al., 2013), which has a 375-m resolution at nadir, and 1-day repeat cycle, with products released at intervals of 1, 8 and 16 days. The VIIRS instrument is intended as a replacement for AVHRR and MODIS (e.g., Skakun et al., 2017) and has somewhat better constrained red and near-infrared bands (0.6–0.68 μm and 0.85–0.88 μm, respectively; Vargas et al., 2013).

3.3 Medium and fine resolution passive sensors As mentioned previously, the AVHRR instrument was not the first multispectral imaging instrument in space: the MSS sensor mounted on the LANDSAT platform, first launched in 1972, predated AVHRR by several years and was also higher resolution (e.g., Maul and Gordon, 1975). LANDSAT has involved eight different satellites: Landsat 1–3 carried the MSS instrument, while a higher resolution instrument, the thematic mapper TM (at 30-m pixel size), was introduced on Landsat 4 (e.g., Markham et al., 2004). The TM instrument had seven bands (blue, green, red, near-infrared, two mid-infrared, and thermal infrared) and its characteristics have been maintained over the years, from Landsat 4 until the latest satellite of the series, Landsat 8, which was put in orbit on February 11, 2013 and is still active. Landsat 7 did include ETM+, which is slightly different from the thematic mapper on the previous satellite, and two bands (a deep blue coastal/aerosol band and a shortwave-infrared

4 Radar

141

cirrus band) were added to the thematic mapper of Landsat 8, called the operational land imager (OLI). Landsat 8 therefore includes a total of 11 bands, spanning ultra-blue to thermal wavelengths (Roy et al., 2014). The consistency in maintaining the same spectral bands over decades provide highly homogeneous data that allow one to go back in time and study the evolution of earth surfaces for almost 50 years. The tremendous archive of images, collected every 16–18 days over the entire globe for decades, independently of meteorological conditions or visibility, makes the LANDSAT mission unique and probably the most valuable Earth-observing satellite mission of all. Free higher resolution data are also available. One of the most popular missions is certainly the one that the European Space Agency started in 2014 with the Sentinel satellites. Of these, Sentinel-2 includes two polar-orbiting satellites hosting multispectral instruments with 13 spectral bands from visible to mid-infrared wavelengths with variable spatial resolution (from 10 m for visible and near-infrared up to 60 m). Sentinel-2A was launched on June 23, 2015 while Sentinel-2B was launched on March 7, 2017. The third add to the mission is Sentinel-3 that was launched on February 16, 2016 and hosts several instruments including several radiometers, including an Ocean and Land Color Instrument (OLCI). Some of the missions mentioned above have followed from commercial missions. Sentinel 2, for example, is descended from the SPOT satellites (Satellite Pour l’Observation de la Terre). The SPOT missions, consisting of seven generations of satellites first launched in 1986, provides data through the French National Centre for Space Studies, CNES. The SPOT mission started in 1986 providing 20 m resolution multispectral data, and recent SPOT 6 and 7 have data resolution of up to 8 m in the multispectral bands (https://earth.esa.int/web/ eoportal/satellite-missions/s/spot-6-7, retrieved 06-June-2019). In addition, many very high resolution commercial instruments (e.g., finer than 1 m pixels) include both red and nearinfrared bands and can therefore be used for vegetation metrics such as NDVI; examples include Quickbird, GeoEye, IKONOS, WorldView-2, and Pleiades (e.g., Corbane et al., 2014).

4 Radar Radar is an active remote sensing technique that sends out pulses of microwave radiation. This radiation is then reflected by objects, and some of the radiation is reflected back to the instrument. This is called backscatter. Radar instruments record the amount of radiation “scattered” by objects (sometimes called “scatterers”). The amount of radiation scattered back to the receiving instrument varies as a function of the object being measured, and this feature of radar instruments has been used to map various properties of plant canopies. In the early 1990s, a number of authors began to demonstrate that various backscattering data (e.g., backscattering coefficients, ratio of coefficients between radar wavelengths) from synthetic aperture radar correlated with biomass densities measured from ground plots (e.g., Dobson et al., 1992; Le Toan et al., 1992; Ranson and Sun, 1994). Thirty years later, there are a wide range of radar instruments mounted on satellites, and numerous authors have used these instruments to map biomass and other vegetation properties on a global scale (for a review, see Sinha et al., 2015). Radar is typically classified into a number of bands, which correspond to different wavelengths of radiation. In general, radar backscatter should respond strongly to components of

142

5. Linking life and landscape with remote sensing

the canopy that are a similar size to the wavelength of the radar band (e.g., Imhoff, 1995). The X and C bands have short wavelengths, sensitive to leaves and small branches, and therefore saturate quickly, struggling to penetrate dense vegetation canopies. In contrast, longer wavelength bands, such as the L (20 cm wavelength) and P (60 cm wavelength) bands are sensitive to larger branches and trunks, and penetrate into more dense canopies. Consequently these longer wavelength bands have been used to quantify canopy structure and biomass in forested regions (e.g., Le Toan et al., 1992; Mitchard et al., 2009), although L band also saturates in dense, high biomass forest canopies, such as those found across the humid tropics (e.g., Joshi et al., 2017). The saturation point of the different radar wavelengths is uncertain (e.g., Woodhouse et al., 2012). Radar data have some advantages over optical data: (i) critically, in contrast to optical systems, radar is able to collect information through cloud cover, and unlike passive optical systems, radar works at night (e.g., Sinha et al., 2015); (ii) radar backscatter intensity, and other radar backscatter properties, such as coherence, are correlated with key vegetation characteristics, such as density, biomass and height, although typically requires empirical calibrations that are ecosystem specific (e.g., Le Toan et al., 2011). Measurements of canopy height are possible with polarimetric SAR interferometry, with the reported accuracy achieved with X-band interferometry approaching those achieved by lidar in some cases (Papathanassiou and Cloude, 2001; Balzter et al., 2007). New techniques, such as the random volume over ground method, promise to require less empirical calibration than previous techniques (Kugler et al., 2015; Qi et al., 2019). Radar remote sensing of vegetation does, however, have some complications: (i) radar wavelengths interact strongly with water, thus the backscattered signal can vary significantly if there are variations in the moisture status of the soil or vegetation (e.g., Karam et al., 1992; Bindlish and Barros, 2001; Park et al., 2019); (ii) in complex, and particularly in steep, topography, inversion of the backscattered signal is especially challenging due to the impact of strongly varying angles of incidence on the amplitude of the reflected signal, in addition to the issue of having to unwrap the across-track resolution from the topography (e.g., Luckman, 1998; Sun et al., 2002); (iii) radar backscatter is noisy, often characterized by a salt-and-pepper “speckle” texture, which requires spatial filtering and therefore a loss of spatial resolution (Woodhouse, 2006). Nevertheless, despite the fact that most of the current satellite radar systems were not specifically designed for biomass estimation or other applications quantifying vegetation properties, existing instruments have been extensively employed in a number of settings to quantify aboveground biomass and biomass change (e.g., Mitchard et al., 2009; Englhart et al., 2011; Hamdan et al., 2011; Solberg et al., 2013; Vafaei et al., 2018). In general, however, global or large regional (e.g., pantropical) biomass maps have been created with some combination of lidar data that is sometimes used to calibrate spatially continuous radar data (e.g., Asner et al., 2010; Lefsky, 2010; Saatchi et al., 2011; Baccini et al., 2012; Avitabile et al., 2016). The European Space Agency’s BIOMASS mission with a proposed launch date of 2022 is aimed at using creating a continuous map of biomass using P-band polarimetric SAR at between 100- and 200-m spatial resolution (Le Toan et al., 2011). The P-band was chosen for this mission because of its long wavelength as sensitivity of radar data to biomass increases with wavelength (e.g., Dobson et al., 1992; Le Toan et al., 1992). Below is a summary of satellite radar instruments.

5 Lidar

143

4.1 Satellite-based radar systems X-band: TerraSAR-X and TanDEM-X (Breit et al., 2010), Cosmo-Skymed (Covello et al., 2010). Due to the short wavelength, the backscatter over vegetation saturates at very low biomass, but X-band interferometry has been shown to be sensitive to vegetation parameters even for high biomass forest (Qi et al., 2019). C-band: Sentinel 1 a/b (Torres et al., 2012), Radarsat 1/2 (Raney et al., 1991), ERS1/2 (Mohr and Madsen, 2001). These are also short wavelength systems and typically have the same limitations as X-band for vegetation. Sentinel-1 includes two polar-orbiting satellites hosting SAR instruments, one launched on April 3, 2014 and the other launched on April 25, 2016; S-band: NovaSAR (Bird et al., 2013; launched 2018), NISAR (Rosen et al., 2017; planned launch 2021). At the time of writing, the full potential of S-band has still to be fully evaluated. Greater availability of NovaSAR data may increase interest in the use of this wave band. L-band: PALSAR 1/2 (Rosenqvist et al., 2007), SAOCOM (an instrument lunged by the Agentinian Space Agency), NISAR (Rosen et al., 2017; planned launch 2021). These longer wavelength (23 cm) systems saturate at higher vegetation biomass, since it provides some penetration through the upper canopy and obtains more signal from the larger elements within the forest canopy. GlobBiomass has provided global estimates of AGB using PALSAR data. P-band: So far, no P-band system has flown in space. The ESA-BIOMASS mission (Le Toan et al., 2011; Quegan et al., 2019) planned for 2022 will be the first, with the objective of providing repeated annual maps of AGB over the course of its duration. The longer wavelength of P-band radar has the greatest ability to penetrate forest canopies and so will prevent saturation even in the densest forests. However, this wavelength is also used by space object tracking radar systems so cannot be turned on within their line of sight. This prevents its use over North America, Central America, and Europe.

5 Lidar We now turn our attention to lidar. Lidar has revolutionized the way we can quantify the structure of the plant canopy and the landscapes upon which they grow, by mapping the three-dimensional structure of both the canopy and underlying topography with exquisite detail. This makes lidar—and airborne lidar in particular—the ideal technology with which to explore the linkages that bind life and landscape at process-relevant length scales, and thus receives particular attention in this review.

5.1 A primer on lidar remote sensing Lidar exploits laser technology to directly sample the three-dimensional structure of vegetation and underlying topography (Lefsky et al., 2002). The basic premise of lidar remote sensing is that the distance between the sensor and the target object, which in this case is the vegetation or ground surface, can be precisely determined based on the time taken from the emission of each laser pulse and the reflection of this energy back to the sensor. As each pulse propagates through the canopy towards the ground, energy is scattered by intersecting

144

5. Linking life and landscape with remote sensing

vegetation and ultimately the ground surface. This backscattered waveform therefore actively samples the three-dimensional canopy architecture and underlying topography, and can be precisely georeferenced in three-dimensional space using accurate GPS technology. Energy reflected back to the sensor from leaves and branches characterize the vertical stratification of material within the footprint of the beam; pulses that penetrate through to be reflected from the ground surface provide a means with which to construct “bare earth” digital elevation models (DEMs) and accurately determine the absolute height aboveground of the intervening vegetation. Lidar wavelengths are typically in the range of 532–1550 nm: much shorter than radar (e.g., Eitel et al., 2016). Lidar possesses two unique advantages over other remote sensing methods: (i) it has been shown to be sensitive to the very densest vegetation on Earth (Drake et al., 2003), unlike alternatives based on passive optical and shorter wavelength radar (L-, C-, and X-band); and (ii) it provides a relatively direct measurement of the three-dimensional positioning of vegetation, requiring no empirical models to predict basic biophysical variables, such as canopy height. Lidar sensors can be employed on the ground (terrestrial laser scanning, TLS; e.g., Brodu and Lague, 2012), on aircraft and UAVs (airborne laser scanning, ALS; Lefsky et al., 1999; Wallace et al., 2012; Asner and Mascaro, 2014; Wieser et al., 2017; Kellner et al., 2019), or on satellites (spaceborne laser scanning; Harding and Carabjal, 2005; Saatchi et al., 2011; Simard et al., 2011; Hancock et al., 2019). In general, as the relative distance between the sensor and the ground increases, the resolution at which vegetation and ground are resolved decreases, due to beam divergence, attenuation, and pulse frequency. Thus, TLS and UAV-ALS provide exceptionally high resolution, with 1000s of pulses per square meter and very small beam footprints, to several tens of pulses per square meter for larger airborne platforms with repeat overflights (footprints typically 101–100 m) to more disparate sampling from spaceborne instruments, with pulses separated by 102–103 m and footprints of 101–102 m. The trade-off, of course, is with the spatial scale over which observations can be made, with airborne platforms providing capacity to survey at landscape-scale, and spaceborne platforms at continental-scale. Sensors may differ in terms of the power (and therefore penetrating capacity), range of scan angles and beam divergence (determining the footprint of the sensor), leading to variation in the precision with which canopy and topographic features are resolved. In practice, analysis of lidar data can be undertaken using the full backscattered waveform (full-waveform lidar) or, more commonly, the waveform is processed into discrete point “returns,” each representing the amplitude peaks in the returned waveform, generating dense, georeferenced clouds of data points, often referred to as point clouds (Fig. 2). The process of discretization of the point cloud leads to a reduction in the information content, particularly in the case of “diffuse” targets (i.e., vegetation), which generate broad peaks in the returned waveform that are not fully captured in the resultant point cloud (Armston et al., 2013; Hancock et al., 2017). Nevertheless, discrete return lidar is currently the more widely utilized of the two, in part due to the additional technical and computational hurdles associated with full-waveform processing. The processing full-waveform lidar, and subsequent preprocessing the discrete point cloud to separate ground returns from vegetation (aboveground) returns, is typically undertaken during a preprocessing stage following the survey and are considered to be beyond the scope of this chapter. These topics have, however, been covered extensively elsewhere (e.g., see: Slatton et al., 2007; Mallet and Bretar, 2009; Glennie et al., 2013).

145

5 Lidar

A 850

Point classifications

Elevation (m)

825 800 775 750 725 700 0

50

100

150

200

250

300

350

400

B 850

Height aboveground

Elevation (m)

825 800 775 750 725 700 0

50

100

150

200

250

300

350

400

Distance from start of transect (m)

FIG. 2

Point clouds for a 10-m wide transect through a mixed conifer forest mantling the slopes of a tributary to the Little North Fork Feather River, California: (A) point classifications (green: vegetation; blue: ground); (B) height of lidar returns aboveground. Transect runs from S to N; coordinates at transect start: (647405, 4396800; UTM 10N).

Finally, due to its current prevalence, the emphasis in this chapter is on small footprint, discrete return lidar, rather than the full-waveform equivalent, although some of the references listed utilize the latter. For more details on full-waveform lidar, a good place to start is the review by Mallet and Bretar (2009).

5.2 Quantifying canopy structure with airborne lidar 5.2.1 Canopy height models and canopy gaps The most basic canopy metric is simply the top-of-canopy height. Once the point cloud has been classified into ground, vegetation and noise returns (e.g., Axelsson, 2000), this is a relatively trivial procedure as long as the point density is sufficiently high (4 pts./m2 in forests, Leitold et al., 2015). Firstly, one calculates the height aboveground for each return, as opposed to absolute elevation, simply by subtracting the elevations from the corresponding DEM (Fig. 3). The topography-corrected point cloud heights can then be rasterized at the desired resolution simply by finding the highest lidar return within each pixel. The resultant surface model is termed the canopy height model (CHM; Fig. 3). CHMs underlie many other

146

5. Linking life and landscape with remote sensing

Digital surface model

Digital elevation model

Canopy height model

Distance (m)

400 300 200 100 0 0

100

600

200

300

400

700 800 Elevation A.S.L. (m)

900

0

600

300 100 200 Distance (m)

400

0

700 800 Elevation A.S.L. (m)

900

0

100

200

300

400

20 40 60 Height aboveground (m)

FIG. 3 Lidar-derived models for a tributary catchment feeding the Little North Fork Feather River, California. From left to right: top of canopy (digital surface model), underlying topography (digital elevation model), and the height of the vegetation aboveground (canopy height model).

canopy structural metrics; for example, they are employed ubiquitously in methods to map aboveground carbon density (ACD; see below). Once a CHM has been produced, it is straightforward to then explore basic aspects of the canopy structure. One important aspect of canopy structure that bears relevance to both ecology and geomorphology is the canopy gap fraction. There are, in fact, two commonly used metrics: the canopy cover, which is the ratio of canopy to all hits, and the crown cover, which is the ratio of first return canopy hits to all first returns. Canopy gaps are spaces within the canopy for which there is no overhead vegetation cover. Their distribution within forest canopies reflects the combination of tree growth and mortality, the balance of which shifts through the ecological succession of forest ecosystems (Coomes et al., 2011). They may also be generated by a range of disturbance processes, including fire (e.g., Spies et al., 1990), disease (e.g., Rizzo et al., 2000), wind throw (e.g., Mitchell, 2012), landslides (e.g., Guariguata, 1990), and of course by human interventions such as logging (e.g., Asner et al., 2004). Gaps in the canopy have an important ecological function, allowing light to penetrate deeper into the canopy, stimulating growth of the understory and ultimately facilitating regeneration (Brokaw, 1985; Schnitzer and Carson, 2001). Canopy gaps also impact on geomorphic processes; for example, reduced root density in these locations may enhance susceptibility to landslides (Roering et al., 2003). In the field, gaps are traditionally measured at a specified reference level (Brokaw, 1982), although the exact reference level chosen may vary depending on the application. This protocol can be readily tailored to, and indeed expanded with, airborne lidar surveys: using the CHM, one can simply discretize individual gaps at multiple reference heights, therefore allowing detailed quantitative analyses of gap-size distributions that can be related to the dynamics of ecosystem succession and disturbance regimes across landscapes (Kellner et al., 2009, 2011).

5 Lidar

147

5.2.2 Identifying individual trees For lidar surveys with high pulse densities (ideally greater than 10 pulses m2), it may be possible to move beyond simply mapping the bulk forest characteristics such as canopy height, and instead extract point clouds for individual trees within the forest, permitting analysis of ecosystem structure to be undertaken at the individual crown level (Ferraz et al., 2016) (Fig. 4). A multitude of methods have been developed to segment point clouds into individual crowns, but most obey the same basic framework, which is first to identify the tree-tops are based on local maxima, and subsequently segment crowns delineated based either on regiongrowing algorithms (Li et al., 2012; Dalponte and Coomes, 2016) or watershed analyses based on inverted CHMs (Chen et al., 2006; Solberg et al., 2006; Duncanson et al., 2014). In order to map understory trees, more complex approaches are required. For example, Duncanson et al. (2014) developed an iterative procedure, whereby for each crown initially segmented,

FIG. 4 Segmentation of tree crowns from a discrete return lidar survey flown over a region of tropical forest on Barro Colorado Island (BCI), in the Panama Canal. Taken from Ferraz, A., Saatchi, S., Mallet, C., Meyer, V., 2016. Lidar detection of individual tree size in tropical forests. Remote Sens. Environ. 183, 318–333. https://doi.org/10.1016/j.rse.2016.05. 028 (web archive link).

148

5. Linking life and landscape with remote sensing

overstory points are split from the understory based on minima in the vertical distribution of points, which are then segmented again into individual understory crowns, and so on until there are no further understory layers to segment. Ultimately, the performance of these segmentation algorithms depends significantly on the complexity of the individual tree crowns and the canopies they together comprise; broadleaf crowns are generally more challenging to delineate than conifers, while understory trees, in particular, are inherently more difficult to extract than their upper canopy counterparts (Solberg et al., 2006; Kaartinen et al., 2012; Duncanson et al., 2014). Consequently, in many settings, accurate representation of canopies at the scale of individual trees remains elusive. Nevertheless, using repeat lidar surveys, Duncanson and Dubayah (2018) have demonstrated that it is possible to generate process-level insight into the dynamics of forest turnover and productivity based on just a subset of the trees. Furthermore, with the emergence of UAV-based lidar platforms, able to fly at low elevations with sampling densities close to those achieved with terrestrial lidar systems, accurate segmentation of trees from the point cloud will likely become increasingly feasible, potentially even to the resolution of individual trunks (Wieser et al., 2017). 5.2.3 Mapping AGB and ACD How much carbon is stored within vegetation? How is it distributed across landscapes? Are the ecological dynamics of forest ecosystems leading to net sequestration or release of carbon from the biosphere? The aboveground biomass (AGB, often expressed as AGBD, which is the AGB per unit area or ACD, which the mass of carbon per unit area stored in aboveground vegetation; ACD  AGB2) is another fundamental attribute of forest ecosystems. Standing biomass reflects the balance between primary production, the allocation of this production to different plant tissues, and the rate at which these components of the ecosystem turn over. High AGB may result from either high productivity, or slow turnover rates (or both), and vice versa. The availability of light, moisture and nutrients, and the exposure to potential disturbance processes, such as wind, fire, and landslides, vary across landscapes. Consequently, productivity, allocation patterns, and turnover rates may also shift across landscapes, resulting in spatial variations in AGB. By mapping forest structural attributes with high precision, airborne lidar surveys have become ubiquitous for quantifying these spatial variations in ACD (e.g., Lefsky et al., 1999, 2002; Drake et al., 2003; Naesset and Gobakken, 2008; Wulder et al., 2012; Asner and Mascaro, 2014), and therefore open a window into the process-level linkages that control ACD variations across landscapes. Most estimates of tree biomass, and therefore AGB and ACD, are founded on the basis that biomass scales systematically with tree size, such that there exist functional allometric relationships between biomass and the physical aspects of tree form, such as height and trunk diameter (West et al., 1999; Chave et al., 2005; Niklas, 2006; Mitchard et al., 2014). In the case of lidar-based AGB mapping, the vast majority of AGB-scaling relationships are calibrated against estimates derived from field inventory plots (e.g., Lefsky et al., 1999; Asner and Mascaro, 2014), rather than against directly measured AGB from harvested, stand-level plots that provide a direct quantification of the AGB that is “perceived” by the lidar sensor (Colgan et al., 2013). Ultimately, the uncertainty in lidar-derived AGB is fundamentally dependent on the errors inherent to the plot inventory collection, and particularly on the quality and uncertainty of the allometric models used to estimate biomass (Clark and Kellner, 2012), which are

5 Lidar

149

typically strongly heteroscedastic; the uncertainty increases with the size of the tree. Secondly, AGB is not only determined by the size of the tree, but also the wood density (e.g., Baker et al., 2004). Wood density varies between species, and is often tied functionally to the competitive life-strategy of a given tree (e.g., Swenson and Enquist, 2007). Fast-growing species that prioritize growth over longevity will usually have lower wood densities than slower growing species that favor long woody residence times. “Fast” life strategies are often utilized by pioneer species and those adapted to resource-rich, highly competitive environments, while the latter are characteristic of later stage successional species and resourcelimited environments (Chave et al., 2009). Accounting for wood density variations based on remote sensing is a challenge, as this trait is not directly measured with the lidar scanner, and often relies on regional estimates (Chave et al., 2005; Asner et al., 2012). This may lead to significant, systematic errors in AGB if not taken into consideration (Mitchard et al., 2014). These aspects will be discussed further in the following sections. 5.2.3.1 AREA-BASED APPROACHES

The most widespread—and easiest—approach to mapping AGB and ACD is to use relatively simple area-based scaling approaches. These methods follow directly on from allometric expectations that tree—and therefore stand—level biomass scales with the height of the canopy and/or other canopy attributes. A plethora of potential metrics can be derived from the CHM and related to AGB and ACD. Here we will focus on the method proposed by Asner and others (Asner et al., 2012; Asner and Mascaro, 2014; Fig. 5), based on the mean top of canopy height, TCH, developed as a generic model applicable to a wide range of tropical forest ecosystems, as this is both widely used and rooted in allometric theory. Furthermore, methods based on TCH, are less sensitive to instrument- and survey-specific factors, such as beam divergence, power, and sensor sensitivity, all of which impact the point cloud representation of the vertical canopy profile (Naesset, 2009), although large variations in these factors may well still lead to potential bias (see Roussel et al., 2017). Based on the field-based allometric model proposed by Chave et al. (2005), Asner et al. (2012) proposed the following aggregated stand allometric equation to calibrate against field inventory estimates: ACD ¼ aTCHb1 BAb2 ρb3

(2)

where BA represents plot basal area, ρ represents basal area weighted wood density, a is a scalar, and the exponents b1, b2, and b3 are allometric scaling coefficients. This model is calibrated using log–log linear regression, with the multiplication of the necessary correction factor to account for the bias associated with the back-transformation of model error from log–log space (Baskerville, 1972):  (3) CF ¼ exp σ 2 =2 where σ is the root mean square error from the regression model (in log–log space). Direct measurements of BA and ρ would require detailed field inventory, potentially prohibiting landscape-level applications; however, BA and height are themselves linked via allometric relationships (Chave et al., 2005), while ρ can be estimated based on regional estimates from the literature (Chave et al., 2009; Zanne et al., 2009) or local knowledge of dominant species.

150

5. Linking life and landscape with remote sensing

350

Hawaii Wet

N. Madagascar Wet

N. Peru Moist

S. Peru Mont. Wet

Panama Dry

Reg. RMSE: 43.6 BIAS: 19 Uni. RMSE: 36.1 BIAS: 46

Reg. RMSE: 42.4 BIAS: 0.3 Uni. RMSE: 42.9 BIAS: 4.1

Reg. RMSE: 18.1 BIAS: 0.6 Uni. RMSE: 19.9 BIAS: 8.9

Reg. RMSE: 23.3 BIAS: 2.5 Uni. RMSE: 31 BIAS: 20

Reg. RMSE: 3.3 BIAS: 2.5 Uni. RMSE: 5.4 BIAS: 20

Haiwaii Moist

S. Madagascar Moist

S. Peru Moist

Panama Moist

Colombia Moist

Reg. RMSE: 37.6 BIAS: –0.3 Uni. RMSE: 36 BIAS: 0.1

Reg. RMSE: 26.2 BIAS: –0.7 Uni. RMSE: 37.5 BIAS: 18.3

Reg. RMSE: 26 BIAS: 0 Uni. RMSE: 30.6 BIAS: –12

Reg. RMSE: 16.2 BIAS: 0 Uni. RMSE: 16.8 BIAS: –13

Reg. RMSE: 16.8 BIAS: 0 Uni. RMSE: 16.6 BIAS: –0.5

Haiwaii Dry

S. Madagascar Dry

S. Peru Submon. Wet

Panama Wet

Reg. RMSE: 15.1 BIAS: 0 Uni. RMSE: 24 BIAS: 1.5

Reg. RMSE: 5.3 BIAS: 0 Uni. RMSE: 7.1 BIAS: 3

Reg. RMSE: 13.2 BIAS: 0.2 Uni. RMSE: 25 BIAS: 18.3

Reg. RMSE: 9.9 BIAS: 0 Uni. RMSE: 17.6 BIAS: 13.6

250 150

EACDfield (Mg C ha–1)

50 0

350 250 150 50 0

350 250 150 50 0

0 10 20 30 40

0 10 20 30 40

0 10 20 30 40

0 10 20 30 40

LiDAR TCH (m)

FIG. 5 Performance of top of canopy height (TCH)-based aboveground carbon density (ACD) models compared to plot estimates (EACDfield) for 14 different tropical forest sites. Taken from Asner, G.P., Mascaro, J., 2014. Mapping tropical forest carbon: calibrating plot estimates to a simple lidar metric. Remote Sens. Environ. 140, 614–624. https://doi.org/10. 1016/j.rse.2013.09.023 (web archive link).

In their general model, Asner and Mascaro (2014) proposed that both BA and ρ could be estimated based on simple relationships with TCH (Fig. 5): BA ¼ α1 TCH

(4)

ρ ¼ α1 TCHb4

(5)

Calibrating these against a series of over 800 field inventory plots, spanning 14 distinct tropical ecoregions they were able to explain 85% of the variance in ACD, compared to 92.3% of the variance explained using direct plot-level basal area and weighted wood density estimates. Without accounting for covariation of BA and TCH as above, the quality of the regression models deteriorated substantially—simple power law fits against TCH resulted in R2 of 0.53.

5 Lidar

151

The use of the above model, in which ACD is calculated essentially based only on TCH and an estimate of wood density appropriate for the setting assumes that these relationships hold within the spatial extent of the lidar survey. However, it is perfectly feasible that changes in environmental conditions may change over sufficiently short length scales—for example, over geological boundaries, or across geomorphic gradients (Ferry et al., 2010; Quesada et al., 2012; Jucker et al., 2018a)—such that both wood density and the stocking coefficient may vary to the extent that simple height-based models are not sufficient to capture spatial variations in ACD. Recent studies emerging from studies in the complex tropical forests of Malaysian Borneo (Coomes et al., 2017; Jucker et al., 2018b), and elsewhere (Duncanson et al., 2015), have demonstrated that TCH is sometimes a relatively poor predictor of both BA and ρ ( Jucker et al., 2018b.). Coomes et al. (2017) found that the correspondence between field- and lidar-based ACD estimates was improved using a power-law relationship between BA and the gap fraction measured at 20 m, rather than the TCH-based stocking coefficient. Canopy cover and TCH are correlated. Jucker et al. (2018b) estimated the expected canopy cover for a given TCH based on logistic regression and then used the residual canopy cover, alongside TCH, within their submodel for predicting basal area. Accounting for basal area in this way resulted in an improvement in the calibrated relationship, reducing the RMSE by over 18% ( Jucker et al., 2018b), although clearly the modified relationship is bespoke to the ecosystem in which it was calibrated. In general, accounting for spatial variations in wood density is challenging from a lidar perspective, but this limitation could be overcome by fusing lidar surveys with high-resolution hyperspectral imaging, from which spectral signatures of ρ variation can be derived ( Jucker et al., 2018a). Moreover, these studies highlight the importance of testing alternative model structures against the available inventory data, ensuring that the lidar-based ACD model is calibrated against test sites that span the ecological range of the landscape. 5.2.3.2 INDIVIDUAL-BASED APPROACHES

The alternative to using area-based methods to map ACD is to map ACD at a tree-by-tree level (Duncanson et al., 2015; Dalponte and Coomes, 2016; Coomes et al., 2017), using allometric relationships that relate the biomass of a given tree to its height and crown width (Duncanson et al., 2015; Jucker et al., 2018b). Theoretically, if one could precisely map the crowns of every tree within a landscape, it should be possible to significantly outperform area-averaged models. However, in practice, the performance of individual-based ACD models depends to a large extent on the characteristics of the forest. In general, individual-based models tend to work best in relatively sparse forests with simple characteristic crown geometries that are more accurately segmented using automated algorithms (Duncanson et al., 2015). Conversely, while missing understory trees can be accounted for using empirical correction factors (Dalponte and Coomes, 2016), in dense, closed canopy forests, segmentation algorithms struggle to separate the canopy crowns accurately, thus degrading individual-based ACD estimates. In these cases, individual-based ACD models may be no better—or even worse—than the simpler area-based approaches, despite the additional computational expense.

152

5. Linking life and landscape with remote sensing

5.2.3.3 CALIBRATION AND UNCERTAINTY

Sources of uncertainty in lidar-based ACD maps may creep in through a number of sources, from errors in the original allometric models used to construct the field inventory, to errors associated with the field-lidar comparison. Firstly, it is important to note that the inventory-based calibration data are themselves model estimates based on allometric equations. The application of these allometric models to estimate tree biomass in different species or regions presents a major source of uncertainty, with potential for systematic bias ( Jenkins et al., 2003; Chave et al., 2004; Yanai et al., 2010), which will propagate through the calibration of lidar-based ACD estimates. Allometric models are typically constructed via harvesting of individual trees at different stages in their growth, with trunk diameter, then tree height accounting for the most variance in biomass between samples (e.g., Chave et al., 2005). Due to the labor intensity involved, often the samples upon which allometric models are limited in terms of size, particularly in the case of large trees in most cases there are no site-specific—and frequently no species-specific—models, necessitating the use of regional-national levels allometric models based on regions or species groups ( Jenkins et al., 2003; Chave et al., 2005). Different sets of allometric equations can lead to significant differences in the plot-based AGB, with allometric equations that include height as well as trunk diameter reducing both bias and error relative to those that rely on trunk diameter alone (Feldpausch et al., 2012; Zhao et al., 2012; Chen, 2015). Finally, as a result of heteroscedasticity, uncertainties at the plot level are dominated by the largest trees, which often dominate the biomass of their plots (Chave et al., 2003). Note that heteroscedasticity does not appear to be as much of an issue for TLS based biomass estimates (Calders et al., 2015). As a result, this technology is increasingly likely to replace standard diameter-based field inventory estimates of biomass at the plot scale, as its application becomes more widespread. Added on top of the model error associated with the allometric equations is the measurement error in the field inventory. Additional sources of error in the calibration of lidar metrics arise as a consequence of the discordance between the ACD estimated within the plot, and that observed by the lidar, specifically: (i) GPS positional error, both in the location of the inventory plot and the geo-referencing of the lidar point cloud, leading to misalignment of the field plots and corresponding maps of TCH (Asner, 2009; Frazer et al., 2011). (ii) Temporal differences between lidar and field surveys (Babcock et al., 2016). This becomes increasingly problematic in dynamic forests, such as early successional systems and forests subject to (or recovering from) disturbance in the interval separating the field and airborne campaigns. (iii) A mismatch between the way in which biomass is allocated in space between the field plots (stem localized) and the corresponding lidar point cloud (area distributed). This leads to differential inclusion of overlapping canopy for trees close to the edge and may account for 50% of the overall uncertainty in the calibration (Mascaro et al., 2011). Errors of this type tend to be greater for canopies hosting larger tree crowns, due to greater potential for overlap across the plot boundaries. In general, errors of type (i) and type (iii) decrease as the inventory size used in the calibration process increases, as they are predominantly edge effects, and—assuming sensible

5 Lidar

153

plot geometries—the fraction of the overall plot area proximal to a plot edge decreases as plot area increases (Frazer et al., 2011; Mascaro et al., 2011). Many of the above errors are strongly non-Gaussian, heteroscedastic, and may vary substantially as the structure of the canopy changes across landscapes. Propagating errors through the calibration process is therefore not amenable to analytical solutions, and instead requires Monte Carlo approaches (Gonzalez et al., 2010; Yanai et al., 2010; Rejou-Mechain et al., 2017), in which the uncertainty associated with each step of the calibration process is estimated based on thousands of simulations of measurement and allometric errors. The variance in the ensemble of calibrations therefore represents the uncertainty in the calibrated lidar-ACD model. Pixel-wise uncertainties may be surprisingly large, but diminish as they are aggregated over larger spatial scales (Gonzalez et al., 2010). Finally, it is important to note that many of these sources of uncertainty are not specific to lidar remote sensing, but would equally apply to other remote sensing technologies. 5.2.4 Quantifying PAI and vertical distributions of plant area density The abundance and density of vegetation in the canopy are additional fundamental functional characteristics of forest ecosystems. The vertical and lateral distribution of leaves, and their traits, control microclimate, light availability, and canopy biogeochemical fluxes ( Jarvis and McNaughton, 1986; Ellsworth and Reich, 1993; Parker et al., 2004; Binkley et al., 2013; von Arx et al., 2013). To a large extent, canopy structure also determines the range of environmental niches within a landscape, and their spatial connectivity, therefore is fundamentally important with respect to biodiversity (Gray et al., 2007; Clawges et al., 2008; Coomes et al., 2009; Struebig et al., 2013; McLean et al., 2016). The quantity of leaves within a canopy is typically characterized by the LAI (units: m2 m2), defined as the total one-sided leaf area per unit ground surface area (Asner et al., 2003). The vertical distribution of this leaf area within the canopy defines the leaf area density (LAD, units: m2 m2 m1) distribution. Importantly, airborne lidar does not provide a distinction between leaves and other parts of the canopy, such as twigs, branches, and the trunk, so it is more accurately quantifying the PAI and plant area distribution (PAD) (Woodgate et al., 2016; Vincent et al., 2017). PAI and PAD are intimately related to the canopy gap fraction (MacArthur and Horn, 1969; Ni-Meister et al., 2001), and can be estimated with lidar data based on a siumple onedimensional Beer-Lambert approximation of light propagation (e.g., Harding et al. 2001; Stark et al., 2012; Tang et al., 2012). Consider a canopy, assumed to be horizontally uniform, with a vertical distribution of PAD described by the function, P(z), where z is the depth into the canopy from its top. The probability of a gap fraction above a specified canopy depth is represented by the function Φ(z). Moving an increment, dz, into the canopy, the penetrating radiation is intercepted at a rate that is proportional to the P(z) and the propagation distance. The rate of change of Φ(z) is proportional to P(z), as described by the following equation: d∅ðzÞ ¼ κPðzÞ∅ðzÞ dz

(6)

where κ is a correction factor that accounts for canopy characteristics that modulate this relationship, such as the leaf angle distribution. To constrain the quantity of leaves within

154

5. Linking life and landscape with remote sensing

a canopy layer, i, bounded by zi-1 and zi Eq. (6) is rearranged to give P(z) as a function of the remaining variables, and then integrated with respect to z between these limits, yielding:   1 ln ∅ðzi Þ =∅ðzi1 Þ (7) Pi ¼ κΔz The PAI is simply calculated as the vertical integral of the PAD distribution, or equivalently, by evaluating Eq. (7) over the full canopy thickness. If inverting a full-waveform ALS signal, we can substitute Φ(z ¼x) as the fraction of the returned radiation penetrating to a depth z ¼ x (Harding et al., 2001; Tang et al., 2012, 2014; Armston et al., 2013). Exploiting the full-waveform ALS in this way carries the advantages that (i) no information on foliage distributions is lost in the discretization of the waveform, and (ii) that it is possible to estimate the ground and canopy reflectances, and consequently generate a self-calibrated model for the total gap fraction and κ (Armston et al., 2013). Full-waveform ALS therefore gives more accurate estimates of canopy cover, LAI, PAI, and vertical PAD profiles than discrete return methods. In contrast, when employing discrete return lidar, it is necessary to approximate Φ(z¼ x) as the fraction of pulses that penetrate to a depth z ¼ x before intercepting vegetation, after aggregating the point cloud over a specified neighbourhood, i.e., n(z  x)/n (e.g., Stark et al., 2012; Hopkinson et al., 2013; Schneider et al., 2017). One could consider the distribution of first returns, which is close to the field sampling approach proposed by MacArthur and Horn (1969). Alternatively, Φ(z¼ x) could be estimated based on all available returns, weighted by the number of returns associated with their associated lidar pulse (e.g., Armston et al., 2013). Whether using the full waveforms or discrete returns, the above model is a simplification of the heterogeneity in real canopies, where the clumping of foliage at the twig, branch, crown, and stand level (depending on the footprint/aggregation window) reduces the effective intercepting PAI (e.g., Chen et al., 1991). Nevertheless, this method has been validated against both direct harvested columns and terrestrial laser scans at two sites in the Brazilian Amazon (Stark et al., 2012), with the resultant canopy structures also consistent with expectations based on tree demographics at these sites (Stark et al., 2015), and in Panama with fullwaveform lidar (Tang et al., 2012). Similar approaches using waveform lidar in temperate forests have also yielded sensible canopy structures (Harding et al., 2001), suggesting that the above assumptions are likely to provide a good first-order approximation of canopy plant area distributions. The PAI is simply calculated as the vertical integral of the PAD distribution, or equivalently, by evaluating Eq. (7) over the full canopy thickness.

5.3 Spaceborne lidar Airborne laser scanning is arguably the most accurate method for measuring landscapescale vegetation, but the costs associated make it unsuitable for larger scale studies. For that, satellites are required. There have been three operational spaceborne lidar missions designed to measure the Earth’s land surface. NASA’s Ice, Cloud, and Land Elevation Satellite (ICESat)/Geoscience Laser Altimeter System (GLAS) operated between 2003 and 2009, NASA’s ICESat-2/ATLAS was launched in September 2018 for a nominal 3-year mission, and NASA’s Global Ecosystem Dynamics Investigation (GEDI) was launched in December 2018 for a nominal mission length of 2 years after its on-orbit checkout the following April. In order to achieve high enough laser pulse energies to get a usable signal-to-noise ratio (SNR)

5 Lidar

155

at the altitude of a satellite (>400 km), spaceborne lidar’s lasers pulse at much lower rates than airborne systems (100 Hz compared to 300 kHz). When combined with the much faster speeds of a satellite compared to an aircraft (17,000 mph compared to 300 mph) spaceborne lidars are unable to achieve the measurement densities of airborne lidar needed to produce continuous maps of vegetation parameters. The primary remote sensing objective varies between these different missions: the ICESat missions’ primary goal is to map the elevations of the polar ice caps; GEDI, in contrast, is specifically designed to map forest characteristics. Consequently, each sensor employs different approaches to reach their land surface measurement goals, resulting in canopy properties of mixed that of varying quality and resolution. 5.3.1 ICESat/GLAS The Ice, Cloud, and Land Elevation Satellite (ICESat) flew between the January 13, 2003 and October 11, 2009 and carried the Geoscience Laser Altimeter System (GLAS) lidar instrument. It was primarily designed to measure ice cap elevations for estimating ice volume (Zwally et al., 2002). To achieve high SNRs, GLAS pulsed one of its three 1064 nm laser at a time at 40 Hz, leaving 200 m between footprints along a track and several kilometers between tracks at the equator. This sparse sampling was further worsened by instrument constraints only allowing it to be operated for 2 or 3 months a year, with the SNR markedly degrading over the instrument’s lifetime. Each laser shot illuminated an approximately circular area on the ground with a diameter of between 90 and 65 m and a nominal pulse energy of 100 mJ. The reflected energy was recorded by a full-waveform detector, sampling the returned intensity every 1 ns (15 cm range resolution). An example waveform over vegetation is shown in Fig. 6. The ability of ICESat/GLAS to estimate bare Earth elevation and tree height is apparent. ICESat/GLAS has been extensively used for measuring vegetation (Harding and Carabjal, 2005). To measure vegetation with full-waveform lidar, first background noise must be removed, the ground return identified and the remaining waveform classified as vegetation. Several methods to remove background noise and separate ground and canopy portions of the waveform. The most widely used is Gaussian decomposition (Hofton et al., 2000), where a noise threshold is set based upon the statistics of the background noise (mean noise level plus a number of standard deviations). Then a number of Gaussians are fit to the signal using nonlinear optimization and the lowest Gaussian assumed to be the ground return. The distance between the start of the signal above background noise and the center of the ground return gives tree height (Los et al., 2012). ICESat waveforms had six Gaussians fit to each waveform, and this decomposed waveform was provided as a data product (Brenner et al., 2012). Alternative methods to identify features, while not as widely used as Gaussian decomposition, are reviewed in Hancock et al. (2015). Any topographic variation across the footprint stretches out the returns in the waveform (Fig. 7). If there is no clear inflection point between the ground and canopy portions of the waveform, the two cannot be reliably separated. ICESat’s large footprint made this a particular problem (Harding and Carabajal, 2005; Duncanson et al., 2010), particularly in the presence of understorey vegetation (Hancock et al., 2012). For this reason, most users did not trust ICESat to measure vegetation over slopes greater than 5° or 12° (Lefsky, 2010; Simard et al., 2011; Los et al., 2012) without having access to additional topographic information, such as an independent DEM (Rosette et al., 2010).

156

5. Linking life and landscape with remote sensing

585

Range (m)

590

Laser footprint

Tree top

Canopy return Canopy height

595 Ground 600

605

Multiple scattering

Radiant flux FIG. 6 An example waveform produced by a simulator for a 30-m footprint lidar incident over a forest on flat topography. The ground return is visible at a range of 600-m and the tree tops at a range of 586-m (a canopy height of 16-m). Taken from Hancock, S., Lewis, P., Foster, M., Disney, M., Muller, J.-P., 2012. Measuring forests with dual wavelength lidar: a simulation study over topography. Agric. For. Meteorol. 161, 123–133. https://doi.org/10.1016/ j.agrformet.2012.03.014.

FIG. 7 An illustration of topographic variation “blurring” the ground signal in the reflected waveform, based on a simulated 30 m diameter full-waveform lidar footprint over a Sitka spruce forest on a 30 degree slope. The canopy and ground portion are blurred, with no clear inflection point. Taken from Hancock, S., Lewis, P., Foster, M., Disney, M., Muller, J.-P., 2012. Measuring forests with dual wavelength lidar: a simulation study over topography. Agric. For. Meteorol. 161, 123–133. https://doi.org/10.1016/j.agrformet.2012.03.014 (web archive link).

ICESat/GLAS data have been used to generate three global canopy height maps so far (Lefsky, 2010; Simard et al., 2011; Los et al., 2012). Note that a large footprint (>5 m), waveform lidar gives the height of the highest point in the canopy aboveground rather than the height of a single tree (canopy height rather than tree height) and so is sometimes compared to the traditional forestry measure of Lorey’s height (Lefsky, 2010). Due to the sparse sampling density, these were produced at either 0.5° resolution using only ICESat data (Los et al., 2012), or at 500 m or 1 km resolution by combining with ancillary datasets and statistical models (Lefsky, 2010; Simard et al., 2011).

5 Lidar

157

These maps used different approaches to remove errors due to slopes. Simard et al. (2011) rejected all ICESat footprints over slopes greater than 5o, where slope was estimated by topographic data derived from the Shuttle Radar Topography Mission (SRTM). This assumes SRTM elevation, which would be at a point within the canopy, roughly followed the ground elevation. Los et al. (2012) rejected all ICESat footprints over SRTM slopes greater than 12°. Lefsky (2010) used an empirical slope to height correction presented in Lefsky et al. (2007). These maps show significant differences between them, particularly in terms of maximum tree height in a given area. The full-waveform of ICESat/GLAS can also be used to estimate PAI, LAI, and the foliage profile (Tang et al., 2014). These require the ground portion of the waveform to be identified, as for canopy height estimates. Once the canopy portion of the waveform has been isolated, the equations presented in Section 5.2.4 can be applied to produce a foliage profile. These equations assume that all vegetation is randomly distributed and so while they have been shown to be accurate in closed canopies, with no gaps between crowns, (Tang et al., 2012), open canopies, with gaps between crowns, may suffer from overestimates of plant area density at lower heights (Ni-Meister et al., 2001). A global canopy cover map has been produced, which is sensitive to canopy cover variations even in dense tropical forests (>80% cover) where other satellite products, such as MODIS, saturate (Tang et al., 2019a,b). It has been shown that lidar waveform metrics are well correlated with forest biomass (Drake et al., 2003) and can be used to predict stand level biomass using the area approaches described in Section 5.2.3. A number of studies have used ICESat/GLAS data to predict biomass, both on its own and through fusion with other datasets (Saatchi et al., 2011; Baccini et al., 2012; Avitabile et al., 2016). These maps show disagreement (Mitchard et al., 2013) and future missions, such as NASA’s GEDI (described in Section 5.3.1.) and ESA’s BIOMASS (Le Toan et al., 2011) aim to address this uncertainty. All the raw ICESat/GLAS data is freely available from the national Snow and Ice data center (web ref https://nsidc.org/data/icesat, last accessed 12-August-2019) and several of the derived global maps are available (Saatchi et al., 2011; Simard et al., 2011; Baccini et al., 2012; Los et al., 2012; Avitabile et al., 2016). 5.3.2 GEDI NASA’s GEDI is the first spaceborne lidar specifically designed to measure vegetation (Dubayah et al., 2020). It was launched on December 5, 2018 and mounted on the Japanese Experimental Module Exposed Facility (JEM-EF), on the International Space Station (ISS) for an expected 2 year mission. This limits its coverage to 51.6° S to 51.6° N. It is a 1064 nm, full-waveform lidar with 15 cm sampling, much like ICESat, but uses a 22-m diameter footprint and fires three lasers simultaneously at 242 Hz, with one laser split into two beams. All four beams are then dithered (a two position scanning system) to produce eight ground tracks. The smaller footprint allows measurements on much steeper slopes (GEDI is expected to have low errors until beyond 30°) and the additional laser beams and dithering give a 4.2-km wide swath with 600-m across-track footprint spacing and 60-m along-track footprint spacing (Fig. 8). The higher pulse rate results in less energy per pulse, with 10 mJ in the two unsplit beams (known as “power” beams) and around 4.2 mJ in the split beams

158

5. Linking life and landscape with remote sensing

3 lasers 4 beams 8 ground tracks

Coverage laser

Full power laser

Full power laser

Along-track direction

60 m 600 m 4.2 km

FIG. 8

Illustration of GEDI sampling pattern showing the two power and coverage beams dithered to produce eight ground tracks of footprints, and the along and across-track spacing. Note that the along-track and across-track distances are not to scale. At any one instant, four laser pulses from the 3 lasers hit the ground, with a total eight tracks produced by dithering the four beams across-track. Taken from Dubayah, R., et al., 2020, The Global Ecosystem Dynamics Investigation: high-resolution laser ranging of the Earth’s forests and topography. Sci. Remote Sens. 1, 100002.

(known as “coverage beams”). It is expected that the power beams will be able to detect the ground through “99.5% canopy cover by night and 98% by day while the coverage beams will be able to detect the ground through 96% by night and 92% by day” (Hancock et al., 2019). Each waveform looks similar to that in Fig. 8 (with less blurring over topography due to the smaller footprint). Its greater sampling density will allow the production of 1-km resolution data products without the need to fuse with other datasets. The GEDI mission has just begun at the time of writing, with the first data released in January 2020 (Fig. 9; Dubayah et al., 2020). Data products, publically available from a NASA data center, will include bare Earth elevation, canopy height, canopy cover, LAI, foliage profile, and AGB density, both as footprint level products (sparse samples of 22-m diameter circles) and as 1-km gridded products. For the biomass product, hybrid-inference statistical theory is used to estimate the mean and standard error of the gridded value from GEDI data alone (Patterson et al., 2019) and it can also be combined with other datasets to attempt higher resolution estimates (Saarela et al., 2018). GEDI has a mission requirement to estimate the biomass of at least 80% of 1-km forested pixels within its orbital bounds with a standard error of less than 20%. 5.3.3 ICESat-2/ATLAS Launched in September 2018 for a nominal 3 year mission, NASA’s Ice, Cloud, and Land Elevation Satellite-2 (ICESat-2), carrying the Advanced Topographic Laser Altimeter System (ATLAS) lidar, is designed to continue the measurement of ice-cap volume begun by ICESat.

220

180

20

Lowest mode

240

0

Height (m)

200

Mean noise

Elevation (m)

Highest return 40

Lowest mode

Lowest return

160

–20

Elevation (m)

200

250

300

350

400

450

DN

210

180

150

0

1000

2000

3000

4000

5000

Distance along transect (m) FIG. 9 GEDI data from one track acquired over Brazil. The data has been processed to level 2A, whereby the highest and lowest returns, and the center of each mode between these points, have been identified. Each vertical green bar represents the position of the energy quantiles (0–100%) for individual waveforms along-transect, with darker shades representing higher amplitudes in the returned signal. The ground elevation is determined based on the elevation of the lowest mode in the waveform. An example waveform is shown in the inset figure, labeled with the highest and lowest returns (horizontal, grey, dashed lines), and the center of the lowest mode (horizontal red line). Taken from Dubayah, R. et al., 2020, The Global Ecosystem Dynamics Investigation: High-resolution laser ranging of the Earth’s forests and topography, Sci. Remote Sens. 1, 100002.

160

5. Linking life and landscape with remote sensing

FIG. 10 Example of a simulated ICESat-2 photon cloud for a 200-m transect over a sparse (50% canopy cover) conifer forest in Sonoma County, California. Brown points have been classified as ground returns, green as canopy and blue as noise. The band of noise photons (blue) represents the 100 m window in which photons are accepted. Note that this section of data would be condensed to two measures of tree height and canopy cover, in order to cope with the low SNR.

ICESat-2 uses a photon-counting system rather than the full-waveform system employed by ICESat/GLAS and GEDI. Photon-counting systems use detectors which are so sensitive, they can detect a single photon. This allows much lower energy laser pulses to be used. An example of a simulated ICESat photon cloud is shown in Fig. 10. ICESat-2’s laser pulses at 10 kHz with only 1.2 mJ per pulse, illuminating a 14-m diameter circle per shot. The high pulse rate allows footprints to be separated by only 70-cm alongtrack, giving a continuous transect of returned photons. Beams are split into six ground tracks, grouped into a power and a coverage beam. Each pair is separated by 3 km to give a 6 km swath. In order to avoid a long dead-time (minimum separation between returned photons for the detector to reset), ICESat-2 uses a 532 nm laser (1064-nm laser frequency doubled to 532 nm). This has a low reflectance over vegetation, so while there are 5–10 signal photons returned per shot over ice, over vegetation there are only 1–3 signal photons returned per shot (Neuenschwander et al., 2019). Photon-counting systems are particularly susceptible to background noise. Combined with the low number of signal photons returned, ICESat-2 has a much lower SNR than ICESat or GEDI. Detecting the ground and vegetation in these low SNR conditions requires advanced signal processing techniques (Neuenschwander and Magruder, 2016). They rely on the fact that the signal photons will be more tightly clumped than noise photons, as all signal photons must arise from real objects, while noise photons will be randomly spread through space and time (Tang et al., 2016). Thus photons in areas with low return density can be assumed to be noise while those in high-density areas can be assumed to be signal. Simulation results suggest that the power beams will be able to detect the ground through vegetation by night (Narine et al., 2019), but the low SNR means that vegetation property variables can only be estimated as a single value over a 100 m transect. Planned vegetation products are canopy height, and roughness, all as a single value over a 100 m transect, and a classified photon point cloud which will allow a bare Earth elevation

6 Airborne electromagnetics

161

transect to be retrieved (Neuenschwander et al., 2019). Work to assess the accuracy of canopy height estimates with real ICESat-2 data, and whether it can provide other biophysical parameters, such as biomass, is ongoing. As only the power beams are expected to be suitable for measuring vegetation, ICESat-2 will have sparser coverage than GEDI for the same time period (though much more dense than ICESat), although three lasers are fitted with only one fired at a time. This should allow operation for much longer than the nominal 3 year mission life.

5.4 Data fusion The above sections have presented the unique advantages of lidar, allowing near direct estimates of canopy height and cover. However, the high energy requirements of a lidar mean that it will never be able to approach the same spatial and temporal coverage as passive optical or radar systems. To address this lack of lidar coverage, the accurate measures of lidar can be combined with continuous (in space and time) passive optical or radar data to produce an optimally fused product (Wulder et al., 2012; Asner et al., 2018). Some of the global biomass maps described above used this approach, fusing ICESat with SAR or passive optical data (Saatchi et al., 2011; Simard et al., 2011). More recently Qi et al. (2019) used simulated GEDI data to calibrate a semi-physical radar scattering model to produce a continuous 12-m resolution tree height from single pass SAR interferometry. Saarela et al. (2018) developed a rigorous statistical framework to allow sparse, accurate data (such as GEDI) to estimate biomass from Landsat data, with errors correctly propagated.

6 Airborne electromagnetics We have seen how lidar can be successfully applied to retrieve the AGB, and the ACD and how, thanks to allometric relations, these values can be used to estimate also belowground biomass and carbon. However, this gives an incomplete picture of the total carbon pool stored in the world ecosystems. It is, in fact, estimated that the amount of carbon stored in soils (topsoils plus subsoils) is at least three times the carbon stored in the vegetation phytomass (i.e., sum of the above and the belowground biomass), with soils storing almost 1500 Pg of carbon vs the 500 Pg stored in vegetation (Scharlemann et al., 2014). However, the uncertainty of these estimates is very high, with values for the global soil carbon that, for other authors, may reach 2500 Pg ( Jansson et al., 2010; Batjes, 2014). It is interesting to note that the large majority of the terrestrial carbon is stored in wet and moist soils, with large amounts of carbon stored in permafrost regions (Scharlemann et al., 2014). Considering these estimates and their associated uncertainties, it is notable to observe how the scientific community has put an enormous effort in accurately determining the carbon stored aboveground with innovative methods, lidar for example, almost neglecting the search for new methodologies for the precise quantification of belowground organic matter and carbon at the regional to the global scale. This is probably due to the challenges posed by the belowground exploration. We know that carbon and biomass found belowground can be explored using field methods, based on direct measurements and coring or on ground-based geophysical

162

5. Linking life and landscape with remote sensing

methods. Despite the high accuracy that these methods may provide locally, they are highly inefficient if a regional assessment is required. The difficulties related with field campaigns are also related with extrapolating the observations performed over small areas to larger domains. This is a risky operation because any interpolation/extrapolation of point measurements needs high-density field observations and a well-characterized covariance model (Keaney et al., 2013). Remote sensing technologies would provide the necessary ability in surveying large and difficult-to-access areas; however, challenges related to the type of sensors that can penetrate the soil and their vertical resolution must be tackled. Light does not penetrate under the soil surface and satellite radar has been found to penetrate the near surface only in very dry soil conditions (Lasaponara and Masini, 2013). A solution may come from airborne geophysical methods, which have been mainly applied in the past to detect deep geological structures and mineral reservoirs. Recent applications of geophysical methods from airborne platforms to near surface targets have been found extremely accurate, especially for modeling land subsidence (Smith and Knight, 2019) and mapping saltwater intrusion in coastal areas (Viezzoli et al., 2010). Can these methods be applied to monitor belowground organic matter and carbon stocks in forested wet and moist environments? Are they also applicable to permafrost regions? Recent applications of transient airborne electromagnetics (AEM) to the quantification of the organic matter and belowground carbon content in peatlands (Silvestri et al., 2019a,b) have been found very successful. In general, among the several airborne technologies that apply geophysical methods (i.e., airborne magnetics, airborne gravity, airborne gravity gradiometery, AEM), the most dramatic advances in the last two decades have been made by the AEM technology (Thomson et al., 2007). Within this broad category, instruments using the frequency domain systems have certainly improved, but the development of the time domain (transient) systems has been truly transformative, providing data for new applications to the characterization of processes occurring near the soil surface. Before proceeding with the description of results on belowground carbon storage in peatlands, we briefly describe the transient AEM technology. These instruments include systems that are carried along flight lines at a height between 30 and 40 m above the ground by helicopters or fixed wing aircraft (Thomson et al., 2007). The instrument has a transmitter where a stationary alternate current flows, generating a primary magnetic field that penetrates underground. When the current is turned off abruptly, the magnetic field decays inducing eddy currents that circulate in the near surface and propagate down into the soil in loops. The eddy currents decrease over time and hence create a secondary magnetic field that decays and can be measured by a receiver. The resistivity (or conductivity) associated with different materials underground is extracted through a numerical inversion of the raw data. The inversion is applied to the data collected along each flight line, however given a survey of several flight lines the result allows a 3D reconstruction of the resistivity of the materials stored underground. Near surface applications strongly depends on the ability of the instruments to measure the signal at very early times (i.e., starting to measure just a few μs after the transmitter is turned off and the intensity of the current goes to zero) (Schamper et al., 2014). This allows to resolve the layers very close to the soil surface. For the study of peatlands, AEM technology is used to detect the contrast between the electrical resistivity of peat and the underlying substrate, hence allowing the detection of

6 Airborne electromagnetics

163

FIG. 11 The figure shows one of the several bogs located in Norway (this one is located near Brøttum) (Silvestri et al., 2019a). The dots on the map are the AEM returns collected along the flight lines, colored according to the peat thickness. The blue line across the map of the bog corresponds to the section shown at the bottom of the figure, where lidar data (the point cloud above the surface) are combined with the AEM results (the peat bottom surface) in a 3D visualization. (Coordinate reference system: WGS 84 UTM 32N. Base map produced from a Sentinel-2 image).

the bottom surface of the peatland. This is fundamental information to estimate the volume of the peat, together with the quantification of its areal extent and its topographic elevation. Once the organic carbon per soil volume is known and based on the volume of peat stocked within the peatland, the total organic carbon content can be calculated. Fig. 11 shows the 3D model of a Norwegian bog (Silvestri et al., 2019a) where the peat volume combined with field measurements and lab analyses allowed estimation of the carbon stock. The section is created combining the returns from lidar, downloaded from the Norwegian National Mapping Authority website (https://hoydedata.no/LaserInnsyn/), with the peat bottom surface extracted from AEM data. Such combination allows a 3D visualization of the peatland, comprising both aboveground and belowground data. Mapping permafrost is another challenging objective that can be achieved using AEM (e.g., Pastick et al., 2013). Given a material, the electrical resistivity may be used to detect its frozen, more resistive portion from its unfrozen part. Combining AEM data with Landsat multispectral and thermal data and digital elevation models, Pastick et al. (2013) estimated the activelayer thickness (ALT) of a large permafrost area in Alaska.

164

5. Linking life and landscape with remote sensing

Even though the number of studies currently available on the volume quantification of peat or permafrost ALT is limited, the use of airborne geophysical methods for belowground carbon assessment is extremely promising and may stimulate further research in this important field.

7 Conclusions Geomorphologists have speculated about the connection between geomorphology and living organisms, and vegetation in particular, for many decades. Over the last several decades, numerous techniques have developed to quantify vegetation on the Earth’s surface, and a number of widely available datasets now exist that create a multitude of opportunities for geomorphologists to explore how vegetation patterns, biomass, and other functional traits affect landscape evolution. Each technique mentioned in the sections above has its own advantages and disadvantages, and there are significant differences in the potential resolution, temporal frequency, and spatial scale at which observations can be made. There are therefore a number of critical aspects that should be considered when considering which products to employ to address problems at the interface between geomorphology and ecology:

7.1 Finding the right sensor Different sensors provide different information on the underlying vegetation, and a given sensor may be more or less appropriate in different ecosystems. Passive optical data is widely available, providing both high temporal resolution data stretching back to the 1970s and, more recently, high spatial resolution data. But passive optical measurements saturate over moderately dense canopies (LAI of 4; canopy cover of 70%–80%). Lidar provides a near direct measure of canopy height and cover that does not saturate in even the densest tropical forests, but energy requirements mean its spatial and temporal coverage will always be much lower than passive optical and radar. Radar provides comparable coverage to passive optical, all weather capability, and longer wavelengths bands (e.g., P-band, and particularly L-band) are sensitive to denser forests than passive optical. On the other hand, radar data is a less direct measure of vegetation structure than lidar, requiring either some form of empirical relationship or a semiphysical model, and most radar bands suffer from saturation in dense forests; empirical relationships are not always readily portable from one setting to another but could be generated with other, physically derived remote sensing data, e.g., Qi et al. (2019). Understanding the limitations of each sensor, and how these impact on the various vegetation characteristics that can be quantified, is therefore essential for understanding which research questions can be adequately addressed with different technology. Moreover, lidar remote sensing is generally the most powerful and information-rich source of information on forest canopies, although the spatial resolution is strongly dependent on the platform, and can be fused readily with other remote sensing data to facilitate upscaling. This is convenient, as lidar data is becoming increasingly ubiquitous in geomorphological studies due to its unparalleled ability to map the morphology of the ground surface.

7 Conclusions

165

7.2 The importance of scale Moreover, it is widely acknowledged that DEM resolution places fundamental constraints on the insights that can be gained on geomorphic processes (e.g., Grieve et al., 2016). This principle holds when employing remote sensing to link the characteristics of vegetation to geomorphic processes and topographic form. Process-relevant length-scales that underpin bio-geomorphic coupling are typically on the order of meters (subhillslope). Thus, while coarse products (10–100 m resolution) are undoubtedly useful for some applications, there is inevitably information loss moving to these resolutions and—as is the case in many geomorphological problems—the impacts of vegetation are unlikely to scale in a linear fashion.

7.3 Trade-offs between resolution and spatial coverage As is the case with many aspects of remote sensing, remote sensing of vegetation typically involves a trade-off between the resolution at which vegetation is resolved and the spatial extents over which surveys can be made. This is clearly manifest in the shifts in resolution when comparing platforms from terrestrial to UAV to aircraft-mounted to satellite platforms, with spatial resolution generally become coarser as the distance between the sensor and target increases. Even with airborne lidar surveys, there are trade-offs between maximizing the spatial coverage and increasing the number of overlapping flight lines, thus providing finer resolution information on the canopy and topography.

7.4 Future outlook Remote sensing of vegetation continues to expand. UAV platforms continue to become increasingly available for research purposes, permitting very high-resolution surveys, and making multi-temporal data-collection increasingly feasible (Kellner et al., 2019). Lidar continues to expand in coverage and data quality, and is increasingly utilizing multiple wavelengths (e.g., Eitel et al., 2016), or being paired with additional remote sensing techniques, such as hyperspectral imagery (e.g., Paz-Kagan et al., 2017; Jucker et al., 2018a), to provide novel insights into landscape-level variations in forest characteristics. Upcoming spaceborne remote sensing missions, in particular GEDI and BIOMASS, will offer state-of-the-art, continent-wide maps of vegetation characteristics, and in the case of BIOMASS, several consecutive years of methodologically consistent maps of AGB change. This explosion of new technology and data availability is opening up more and more opportunities to address ecological and geomorphological problems.

Acknowledgments This presented research on peatlands and airborne geophysical methods is part of the project CReScenDo (Combining Remote Sensing Technologies for Peatland Detection and Characterization) that has received funding from the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 747809. We thank Iain Woodhouse for reviewing the section on Radar.

166

5. Linking life and landscape with remote sensing

References Amundson, R., Heimsath, A., Owen, J., Yoo, K., Dietrich, W.E., 2015. Hillslope soils and vegetation. Geomorphology 234, 122–132. https://doi.org/10.1016/j.geomorph.2014.12.031. Armston, J., Disney, M., Lewis, P., Scarth, P., Phinn, S., Lucas, R., Bunting, P., Goodwin, N., 2013. Direct retrieval of canopy gap probability using airborne waveform lidar. Remote Sens. Environ. 134, 24–38. https://doi.org/ 10.1016/j.rse.2013.02.021. Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic modeling and assessment Part I: model development1. J. Am. Water Resour. As. 34, 73–89. https://doi.org/10.1111/j.1752-1688.1998.tb05961.x. Asner, G.P., 2009. Tropical forest carbon assessment: integrating satellite and airborne mapping approaches. Environ. Res. Lett. 4, 034009. https://doi.org/10.1088/1748-9326/4/3/034009. Asner, G.P., Brodrick, P.G., Philipson, C., Vaughn, N.R., Martin, R.E., Knapp, D.E., Heckler, J., Evans, L.J., Jucker, T., Goossens, B., Stark, D.J., Reynolds, G., Ong, R., Renneboog, N., Kugan, F., Coomes, D.A., 2018. Mapped aboveground carbon stocks to advance forest conservation and recovery in Malaysian Borneo. Biol. Conserv. 217, 289–310. https://doi.org/10.1016/j.biocon.2017.10.020. Asner, G.P., Keller, M., Silva, J.N.M., 2004. Spatial and temporal dynamics of forest canopy gaps following selective logging in the eastern Amazon. Glob. Chang. Biol. 10, 765–783. https://doi.org/10.1111/j.1529-8817.2003.00756.x. Asner, G.P., Mascaro, J., 2014. Mapping tropical forest carbon: calibrating plot estimates to a simple LiDAR metric. Remote Sens. Environ. 140, 614–624. https://doi.org/10.1016/j.rse.2013.09.023. Asner, G.P., Mascaro, J., Muller-Landau, H.C., Vieilledent, G., Vaudry, R., Rasamoelina, M., Hall, J.S., Breugel, M.v., 2012. A universal airborne LiDAR approach for tropical forest carbon mapping. Oecologia 168, 1147–1160. https:// doi.org/10.1007/s00442-011-2165-z. Asner, G.P., Powell, G.V.N., Mascaro, J., Knapp, D.E., Clark, J.K., Jacobson, J., Kennedy-Bowdoin, T., Balaji, A., Paez-Acosta, G., Victoria, E., Secada, L., Valqui, M., Hughes, R.F., 2010. High-resolution forest carbon stocks and emissions in the Amazon. PNAS 107, 16738–16742. https://doi.org/10.1073/pnas.1004875107. Asner, G.P., Scurlock, J.M.O., Hicke, J.A., 2003. Global synthesis of leaf area index observations: implications for ecological and remote sensing studies. Glob. Ecol. Biogeogr. 12, 191–205. https://doi.org/10.1046/j.1466822x.2003.00026.x. Avitabile, V., Herold, M., Heuvelink, G.B.M., Lewis, S.L., Phillips, O.L., Asner, G.P., Armston, J., Ashton, P.S., Banin, L., Bayol, N., Berry, N.J., Boeckx, P., Jong, B.H.J.d., DeVries, B., Girardin, C.A.J., Kearsley, E., Lindsell, J.A., Lopez-Gonzalez, G., Lucas, R., Malhi, Y., Morel, A., Mitchard, E.T.A., Nagy, L., Qie, L., Quinones, M.J., Ryan, C.M., Ferry, S.J.W., Sunderland, T., Laurin, G.V., Gatti, R.C., Valentini, R., Verbeeck, H., Wijaya, A., Willcock, S., 2016. An integrated pan-tropical biomass map using multiple reference datasets. Glob. Chang. Biol. 22, 1406–1420. https://doi.org/10.1111/gcb.13139. Axelsson, P., 2000. DEM generation from laser scanner data using adaptive TIN models. In: International Archive of Photogrammetry and Remote Sensing XXXIII (Part B4). pp. 110–117. Babcock, C., Finley, A.O., Cook, B.D., Weiskittel, A., Woodall, C.W., 2016. Modeling forest biomass and growth: coupling long-term inventory and LiDAR data. Remote Sens. Environ. 182, 1–12. https://doi.org/10.1016/j. rse.2016.04.014. Baccini, A., Goetz, S.J., Walker, W.S., Laporte, N.T., Sun, M., Sulla-Menashe, D., Hackler, J., Beck, P.S.A., Dubayah, R., Friedl, M.A., Samanta, S., Houghton, R.A., 2012. Estimated carbon dioxide emissions from tropical deforestation improved by carbon-density maps. Nat. Clim. Chang. 2, 182–185. https://doi.org/10.1038/nclimate1354. Baker, T.R., Phillips, O.L., Malhi, Y., Almeida, S., Arroyo, L., Fiore, A.D., Erwin, T., Killeen, T.J., Laurance, S.G., Laurance, W.F., Lewis, S.L., Lloyd, J., Monteagudo, A., Neill, D.A., Patino, S., Pitman, N.C.A., Silva, J.N.M., Martinez, R.V., 2004. Variation in wood density determines spatial patterns in Amazonian forest biomass. Glob. Chang. Biol. 10, 545–562. https://doi.org/10.1111/j.1365-2486.2004.00751.x. Balzter, H., Rowland, C.S., Saich, P., 2007. Forest canopy height and carbon estimation at Monks Wood National Nature Reserve, UK, using dual-wavelength SAR interferometry. Remote Sens. Environ. 108, 224–239. Band, L.E., Hwang, T., Hales, T.C., Vose, J., Ford, C., 2012. Ecosystem processes at the watershed scale: mapping and modeling ecohydrological controls of landslides. Geomorphology 137, 159–167. https://doi.org/10.1016/j. geomorph.2011.06.025. Banfield, J.F., Barker, W.W., Welch, S.A., Taunton, A., 1999. Biological impact on mineral dissolution: application of the lichen model to understanding mineral weathering in the rhizosphere. PNAS 96, 3404–3411. https://doi.org/ 10.1073/pnas.96.7.3404.

References

167

Barbier, N., Couteron, P., Lejoly, J., Deblauwe, V., Lejeune, O., 2006. Self-organized vegetation patterning as a fingerprint of climate and human impact on semi-arid ecosystems. J. Ecol. 94, 537–547. https://doi.org/10.1111/j.13652745.2006.01126.x. Barker, W.W., Welch, S.A., Chu, S., Banfield, J.F., 1998. Experimental observations of the effects of bacteria on aluminosilicate weathering. Am. Mineral. 83, 1551–1563. https://doi.org/10.2138/am-1998-11-1243. Barnes, W.L., Pagano, T.S., Salomonson, V.V., 1998. Prelaunch characteristics of the moderate resolution imaging spectroradiometer (MODIS) on EOS-AM1. IEEE Trans. Geosci. Remote Sens. 36, 1088–1100. https://doi.org/ 10.1109/36.700993. Baskerville, G.L., 1972. Use of logarithmic regression in the estimation of plant biomass. Can. J. For. Res. 2, 49–53. https://doi.org/10.1139/x72-009. Batjes, N.H., 2014. Total carbon and nitrogen in the soils of the world. Eur. J. Soil Sci. 65, 10–21. https://doi.org/ 10.1111/ejss.12114_2. Baustian, J.J., Mendelssohn, I.A., 2015. Hurricane-induced sedimentation improves marsh resilience and vegetation vigor under high rates of relative sea level rise. Wetlands 35, 795–802. https://doi.org/10.1007/ s13157-015-0670-2. Beck, H.E., McVicar, T.R., van Dijk, A.I.J.M., Schellekens, J., de Jeu, R.A.M., Bruijnzeel, L.A., 2011. Global evaluation of four AVHRR–NDVI data sets: intercomparison and assessment against Landsat imagery. Remote Sens. Environ. 115, 2547–2563. https://doi.org/10.1016/j.rse.2011.05.012. Bennett, S.J., Pirim, T., Barkdoll, B.D., 2002. Using simulated emergent vegetation to alter stream flow direction within a straight experimental channel. Geomorphology 44, 115–126. https://doi.org/10.1016/S0169-555X(01) 00148-9. Bergen, K.M., Goetz, S.J., Dubayah, R.O., Henebry, G.M., Hunsaker, C.T., Imhoff, M.L., Nelson, R.F., Parker, G.G., Radeloff, V.C., 2009. Remote sensing of vegetation 3-D structure for biodiversity and habitat: Review and implications for lidar and radar spaceborne missions. J. Geophys. Res. Biogeo. 114. https://doi.org/10. 1029/2008JG000883. Bertoldi, W., Gurnell, A.M., Drake, N.A., 2011. The topographic signature of vegetation development along a braided river: results of a combined analysis of airborne lidar, color air photographs, and ground measurements. Water Resour. Res. 47. https://doi.org/10.1029/2010WR010319. Bindlish, R., Barros, A.P., 2001. Parameterization of vegetation backscatter in radar-based, soil moisture estimation. Remote Sens. Environ. 76, 130–137. https://doi.org/10.1016/S0034-4257(00)00200-5. Binkley, D., Campoe, O.C., Gspaltl, M., Forrester, D.I., 2013. Light absorption and use efficiency in forests: why patterns differ for trees and stands. For. Ecol. Manage. 288, 5–13. https://doi.org/10.1016/j.foreco.2011.11.002. Bird, R., Whittaker, P., Stern, B., Angli, N., Cohen, M., Guida, R., 2013. NovaSAR-S: a low cost approach to SAR applications. In: Conference Proceedings of 2013 Asia-Pacific Conference on Synthetic Aperture Radar (APSAR). Presented at the Conference Proceedings of 2013 Asia-Pacific Conference on Synthetic Aperture Radar (APSAR), pp. 84–87. Bischetti, G.B., Chiaradia, E.A., Simonato, T., Speziali, B., Vitali, B., Vullo, P., Zocco, A., 2007. Root strength and root area ratio of forest species in Lombardy (Northern Italy). In: Stokes, A., Spanos, I., Norris, J.E., Cammeraat, E. (Eds.), Eco-and Ground Bio-Engineering: The Use of Vegetation to Improve Slope Stability, Developments in Plant and Soil Sciences. Springer, Netherlands, pp. 31–41. Black, B.A., Perron, J.T., Burr, D.M., Drummond, S.A., 2012. Estimating erosional exhumation on Titan from drainage network morphology. J. Geophys. Res. Planets. 117. https://doi.org/10.1029/2012je004085. Borgogno, F., D’Odorico, P., Laio, F., Ridolfi, L., 2009. Mathematical models of vegetation pattern formation in ecohydrology. Rev. Geophys. 47. https://doi.org/10.1029/2007rg000256. Brantley, S.L., Eissenstat, D.M., Marshall, J.A., Godsey, S.E., Balogh-Brunstad, Z., Karwan, D.L., Papuga, S.A., Roering, J., Dawson, T.E., Evaristo, J., Chadwick, O., McDonnell, J.J., Weathers, K.C., 2017. Reviews and syntheses: on the roles trees play in building and plumbing the critical zone. Biogeosciences 14, 5115–5142. https://doi.org/ 10.5194/bg-14-5115-2017. Breit, H., Fritz, T., Balss, U., Lachaise, M., Niedermeier, A., Vonavka, M., 2010. TerraSAR-X SAR processing and products. IEEE Trans. Geosci. Remote Sens. 48, 727–740. https://doi.org/10.1109/TGRS.2009.2035497. Brenner, A.C., Zwally, H.J., Bentley, C.R., Csatho, B.M., Harding, D.J., Hofton, M.A., Minster, J.-B., Roberts, L., Saba, J.L., Thomas, R.H., 2012. The Algorithm Theoretical Basis Document for the Derivation of Range and Range Distributions from Laser Pulse Waveform Analysis for Surface Elevations, Roughness, Slope, and Vegetation Heights. https://ntrs.nasa.gov/search.jsp?R¼20120016646.

168

5. Linking life and landscape with remote sensing

Bridges, N.T., Ayoub, F., Avouac, J.-P., Leprince, S., Lucas, A., Mattson, S., 2012. Earth-like sand fluxes on Mars. Nature 485, 339–342. https://doi.org/10.1038/nature11022. Brodu, N., Lague, D., 2012. 3D terrestrial lidar data classification of complex natural scenes using a multi-scale dimensionality criterion: applications in geomorphology. ISPRS J. Photogramm. Remote Sens. 68, 121–134. https://doi.org/10.1016/j.isprsjprs.2012.01.006. Brokaw, N.V.L., 1985. Gap-phase regeneration in a tropical forest. Ecology 66, 682–687. https://doi.org/10.2307/ 1940529. Brokaw, N.V.L., 1982. The definition of treefall gap and its effect on measures of forest dynamics. Biotropica 14, 158–160. https://doi.org/10.2307/2387750. Brown, M.E., Pinzon, J.E., Didan, K., Morisette, J.T., Tucker, C.J., 2006. Evaluation of the consistency of long-term NDVI time series derived from AVHRR, SPOT-vegetation, SeaWiFS, MODIS, and Landsat ETM+ sensors. IEEE Trans. Geosci. Remote Sens. 44, 1787–1793. https://doi.org/10.1109/TGRS.2005.860205. Butler, D.R., George, P.M., 2005. The geomorphic influences of beaver dams and failures of beaver dams. Geomorphology 71 (1–2), 48–60. https://doi.org/10.1016/j.geomorph.2004.08.016. Cairns, M.A., Brown, S., Helmer, E.H., Baumgardner, G.A., 1997. Root biomass allocation in the world’s upland forests. Oecologia 111, 1–11. Calders, K., Newnham, G., Burt, A., Murphy, S., Raumonen, P., Herold, M., Culvenor, D., Avitabile, V., Disney, M., Armston, J., Kaasalainen, M., 2015. Nondestructive estimates of above-ground biomass using terrestrial laser scanning. Methods Ecol. Evol. 6, 198–208. https://doi.org/10.1111/2041-210X.12301. Camporeale, C., Perucca, E., Ridolfi, L., Gurnell, A.M., 2013. Modeling the interactions between river morphodynamics and riparian vegetation. Rev. Geophys. 51, 379–414. https://doi.org/10.1002/rog.20014. Caylor, K.K., Manfreda, S., Rodriguez-Iturbe, I., 2005. On the coupled geomorphological and ecohydrological organization of river basins. Adv. Water Resour. 28, 69–86. https://doi.org/10.1016/j.advwatres.2004.08.013. Cerdan, O., Govers, G., Le Bissonnais, Y., Van Oost, K., Poesen, J., Saby, N., Gobin, A., Vacca, A., Quinton, J., Auerswald, K., Klik, A., Kwaad, F.J.P.M., Raclot, D., Ionita, I., Rejman, J., Rousseva, S., Muxart, T., Roxo, M.J., Dostal, T., 2010. Rates and spatial variations of soil erosion in Europe: a study based on erosion plot data. Geomorphology 122, 167–177. https://doi.org/10.1016/j.geomorph.2010.06.011. olster, H., Fromard, F., Higuchi, N., Chave, J., Andalo, C., Brown, S., Cairns, M.A., Chambers, J.Q., Eamus, D., F€ Kira, T., Lescure, J.-P., Nelson, B.W., Ogawa, H., Puig, H., Riera, B., Yamakura, T., 2005. Tree allometry and improved estimation of carbon stocks and balance in tropical forests. Oecologia 145, 87–99. https://doi.org/10.1007/ s00442-005-0100-x. Chave, J., Condit, R., Aguilar, S., Hernandez, A., Lao, S., Perez, R., 2004. Error propagation and scaling for tropical forest biomass estimates. Philos. Trans. R. Soc. Lond. B Biol. Sci. 359, 409–420. https://doi.org/10.1098/ rstb.2003.1425. Chave, J., Condit, R., Lao, S., Caspersen, J.P., Foster, R.B., Hubbell, S.P., 2003. Spatial and temporal variation of biomass in a tropical forest: results from a large census plot in Panama. J. Ecol. 91, 240–252. https://doi.org/10.1046/ j.1365-2745.2003.00757.x. Chave, J., Coomes, D., Jansen, S., Lewis, S.L., Swenson, N.G., Zanne, A.E., 2009. Towards a worldwide wood economics spectrum. Ecol. Lett. 12, 351–366. https://doi.org/10.1111/j.1461-0248.2009.01285.x. Chen, Q., 2015. Modeling aboveground tree woody biomass using national-scale allometric methods and airborne lidar. ISPRS J. Photogramm. Remote Sens. 106, 95–106. https://doi.org/10.1016/j.isprsjprs.2015.05.007. Chen, J.M., Black, T.A., Adams, R.S., 1991. Evaluation of hemispherical photography for determining plant area index and geometry of a forest stand. Agric. For. Meteorol. 56, 129–143. Chen, Q., Baldocchi, D., Gong, P., Kelly, M., 2006. Isolating individual trees in a Savanna woodland using small footprint lidar data. Photogramm. Eng. Remote Sens. 72, 923–932. https://doi.org/10.14358/pers.72.8.923. Clark, D.B., Kellner, J.R., 2012. Tropical forest biomass estimation and the fallacy of misplaced concreteness. J. Veg. Sci. 23, 1191–1196. https://doi.org/10.1111/j.1654-1103.2012.01471.x. Clawges, R., Vierling, K., Vierling, L., Rowell, E., 2008. The use of airborne lidar to assess avian species diversity, density, and occurrence in a pine/aspen forest. Remote Sens. Environ. 112, 2064–2073. https://doi.org/ 10.1016/j.rse.2007.08.023. Clymo, R.S., 1984. The limits to peat bog growth. Philosophical Philos. Trans. R. Soc., B 303, 605–654. https://doi.org/ 10.1098/rstb.1984.0002. Cohen, D., Schwarz, M., Or, D., 2011. An analytical fiber bundle model for pullout mechanics of root bundles. J. Geophys. Res. 116, F03010. https://doi.org/10.1029/2010jf001886.

References

169

Colgan, M.S., Asner, G.P., Swemmer, T., 2013. Harvesting tree biomass at the stand level to assess the accuracy of field and airborne biomass estimation in savannas. Ecol. Appl. 23, 1170–1184. https://doi.org/10.1890/120922.1. Collins, D.B.G., Bras, R.L., Tucker, G.E., 2004. Modeling the effects of vegetation-erosion coupling on landscape evolution. J. Geophys. Res. Earth. 109. https://doi.org/10.1029/2003JF000028. Constantine, J.A., Schelhaas, M.-J., Gabet, E., Mudd, S.M., 2012. Limits of windthrow-driven hillslope sediment flux due to varying storm frequency and intensity. Geomorphology 175, 66–73. https://doi.org/10.1016/j. geomorph.2012.06.022. Coomes, D.A., Kunstler, G., Canham, C.D., Wright, E., 2009. A greater range of shade-tolerance niches in nutrient-rich forests: an explanation for positive richness–productivity relationships? J. Ecol. 97, 705–717. https://doi.org/ 10.1111/j.1365-2745.2009.01507.x. Coomes, D.A., Dalponte, M., Jucker, T., Asner, G.P., Banin, L.F., Burslem, D.F.R.P., Lewis, S.L., Nilus, R., Phillips, O.L., Phua, M.-H., Qie, L., 2017. Area-based vs tree-centric approaches to mapping forest carbon in Southeast Asian forests from airborne laser scanning data. Remote Sens. Environ. 194, 77–88. https://doi.org/10.1016/j. rse.2017.03.017. Coomes, D.A., Holdaway, R.J., Kobe, R.K., Lines, E.R., Allen, R.B., 2011. A general integrative framework for modelling woody biomass production and carbon sequestration rates in forests. J. Ecol. 100, 42–64. https://doi.org/ 10.1111/j.1365-2745.2011.01920.x. uttler, F., Alleaume, S., Ienco, D., Teisseire, M., 2014. Monitoring the phenology of mediterranean natCorbane, C., G€ ural habitats with multispectral sensors—An analysis based on multiseasonal field spectra 2014. In: IEEE Geoscience and Remote Sensing Symposium. Quebec City, QC, pp. 3934–3937. Corenblit, D., Baas, A.C.W., Bornette, G., Darrozes, J., Delmotte, S., Francis, R.A., Gurnell, A.M., Julien, F., Naiman, R.J., Steiger, J., 2011. Feedbacks between geomorphology and biota controlling Earth surface processes and landforms: a review of foundation concepts and current understandings. Earth Sci. Rev. 106, 307–331. https:// doi.org/10.1016/j.earscirev.2011.03.002. Covello, F., Battazza, F., Coletta, A., Lopinto, E., Fiorentino, C., Pietranera, L., Valentini, G., Zoffoli, S., 2010. COSMO-SkyMed an existing opportunity for observing the Earth. J. Geodyn. 49, 171–180. https://doi.org/ 10.1016/j.jog.2010.01.001. WEGENER 2008 - Proceedings of the 14th General Assembly of Wegener. Cracknell, A.P., 2001. The exciting and totally unanticipated success of the AVHRR in applications for which it was never intended. Adv. Space Res. 28, 233–240. https://doi.org/10.1016/s0273-1177(01)00349-0. Cracknell, A.P., 1997. Advanced Very High Resolution Radiometer AVHRR. CRC Press. Dalponte, M., Coomes, D.A., 2016. Tree-centric mapping of forest carbon density from airborne laser scanning and hyperspectral data. Methods Ecol. Evol. 7, 1236–1245. https://doi.org/10.1111/2041-210x.12575. Danjon, F., Fourcaud, T., Bert, D., 2005. Root architecture and wind-firmness of mature Pinus pinaster. New Phytol. 168, 387–400. https://doi.org/10.1111/j.1469-8137.2005.01497.x. De Roo, A.P.J., Wesseling, C.G., Ritsema, C.J., 1996. LISEM: a single-event physically based hydrological and soil erosion model for drainage basins. I: theory, input and output. Hydrol. Process. 10, 1107–1117. de Vente, J., Poesen, J., Verstraeten, G., Govers, G., Vanmaercke, M., Van Rompaey, A., Arabkhedri, M., BoixFayos, C., 2013. Predicting soil erosion and sediment yield at regional scales: where do we stand? Earth Sci. Rev. 127, 16–29. https://doi.org/10.1016/j.earscirev.2013.08.014. de Vente, J., Poesen, J., Verstraeten, G., Van Rompaey, A., Govers, G., 2008. Spatially distributed modelling of soil erosion and sediment yield at regional scales in Spain. Global Planet. Change 60, 393–415. https://doi.org/ 10.1016/j.gloplacha.2007.05.002. Deblauwe, V., Barbier, N., Couteron, P., Lejeune, O., Bogaert, J., 2008. The global biogeography of semi-arid periodic vegetation patterns. Glob. Ecol. Biogeogr. 17, 715–723. https://doi.org/10.1111/j.1466-8238.2008.00413.x. Deng, Z., An, S., Zhao, C., Chen, L., Zhou, C., Zhi, Y., Li, H., 2008. Sediment burial stimulates the growth and propagule production of Spartina alterniflora Loisel. Estuar. Coast. Shelf Sci. 76, 818–826. Deng, L., Shangguan, Z., Li, R., 2012. Effects of the grain-for-green program on soil erosion in China. Int. J. Sediment Res. 27, 120–127. https://doi.org/10.1016/s1001-6279(12)60021-3. Detto, M., Muller-Landau, H.C., Mascaro, J., Asner, G.P., 2013. Hydrological networks and associated topographic variation as templates for the spatial organization of tropical forest vegetation. PLoS One. 8. https://doi.org/ 10.1371/journal.pone.0076296. Dietrich, W.E., Bellugi, D.G., Sklar, L.S., Stock, J.D., Heimsath, A.M., Roering, J.J., 2003. Geomorphic transport laws for predicting landscape form and dynamics. Geophys. Monogr.-Am. Geophys. UNION 135, 103–132.

170

5. Linking life and landscape with remote sensing

Dietrich, W.E., Perron, J.T., 2006. The search for a topographic signature of life. Nature 439, 411–418. https://doi.org/ 10.1038/nature04452. Dobson, M.C., Ulaby, F.T., LeToan, T., Beaudoin, A., Kasischke, E.S., Christensen, N., 1992. Dependence of radar backscatter on coniferous forest biomass. IEEE Trans. Geosci. Remote Sens. 30, 412–415. https://doi.org/ 10.1109/36.134090. Drake, J.B., Knox, R.G., Dubayah, R.O., Clark, D.B., Condit, R., Blair, J.B., Hofton, M., 2003. Above-ground biomass estimation in closed canopy Neotropical forests using lidar remote sensing: factors affecting the generality of relationships. Glob. Ecol. Biogeogr. 12, 147–159. https://doi.org/10.1046/j.1466-822x.2003.00010.x. Dronova, I., Gong, P., Clinton, N.E., Wang, L., Fu, W., Qi, S., Liu, Y., 2012. Landscape analysis of wetland plant functional types: the effects of image segmentation scale, vegetation classes and classification methods. Remote Sens. Environ. 127, 357–369. https://doi.org/10.1016/j.rse.2012.09.018. Dubayah, R., Blair, J.B., Goetz, S., Fatoyinbo, L., Hansen, M., Healey, S., Hofton, M., Hurtt, G., Kellner, J., Luthcke, S., Armston, J., Tang, H., Duncanson, L., Hancock, S., Jantz, P., Marselis, S., Patterson, P.L., Qi, W., Silva, C., 2020. The Global Ecosystem Dynamics Investigation: high-resolution laser ranging of the Earth’s forests and topography. Sci. Remote. Sens. 1, 100002. ISSN 2666–0172. https://doi.org/10.1016/j.srs.2020.100002. Duncanson, L., Dubayah, R., 2018. Monitoring individual tree-based change with airborne lidar. Ecol. Evol. 8, 5079–5089. https://doi.org/10.1002/ece3.4075. Duncanson, L.I., Cook, B.D., Hurtt, G.C., Dubayah, R.O., 2014. An efficient, multi-layered crown delineation algorithm for mapping individual tree structure across multiple ecosystems. Remote Sens. Environ. 154, 378–386. https://doi.org/10.1016/j.rse.2013.07.044. Duncanson, L.I., Dubayah, R.O., Cook, B.D., Rosette, J., Parker, G., 2015. The importance of spatial detail: assessing the utility of individual crown information and scaling approaches for lidar-based biomass density estimation. Remote Sens. Environ. 168, 102–112. https://doi.org/10.1016/j.rse.2015.06.021. Duncanson, L.I., Niemann, K.O., Wulder, M.A., 2010. Estimating forest canopy height and terrain relief from GLAS waveform metrics. Remote Sens. Environ. 114, 138–154. https://doi.org/10.1016/j.rse.2009.08.018. Dunne, T., Malmon, D.V., Mudd, S.M., 2010. A rain splash transport equation assimilating field and laboratory measurements. J. Geophys. Res. Earth 115, F01001. https://doi.org/10.1029/2009JF001302. Dura´n Zuazo, V.H., Rodrı´guez Pleguezuelo, C.R., 2008. Soil-erosion and runoff prevention by plant covers. A review. Agron. Sustain. Dev. 28, 65–86. https://doi.org/10.1051/agro:2007062. Edmaier, K., Burlando, P., Perona, P., 2011. Mechanisms of vegetation uprooting by flow in alluvial non-cohesive sediment. Hydrol. Earth Syst. Sci. 15, 1615–1627. https://doi.org/10.5194/hess-15-1615-2011. ofle, B., Vierling, L.A., Abella´n, A., Asner, G.P., Deems, J.S., Glennie, C.L., Joerg, P.C., LeWinter, A.L., Eitel, J.U.H., H€ Magney, T.S., Mandlburger, G., Morton, D.C., M€ uller, J., Vierling, K.T., 2016. Beyond 3-D: the new spectrum of lidar applications for earth and ecological sciences. Remote Sens. Environ. 186, 372–392. https://doi.org/10.1016/ j.rse.2016.08.018. Ellison, W.D., 1945. Some effects of raindrops and surface-flow on soil erosion and infiltration. Trans. Am. Geophys. Union 26, 415. https://doi.org/10.1029/tr026i003p00415. Ellsworth, D.S., Reich, P.B., 1993. Canopy structure and vertical patterns of photosynthesis and related leaf traits in a deciduous forest. Oecologia 96, 169–178. https://doi.org/10.1007/bf00317729. Englhart, S., Keuck, V., Siegert, F., 2011. Aboveground biomass retrieval in tropical forests—the potential of combined X- and L-band SAR data use. Remote Sens. Environ. 115, 1260–1271. https://doi.org/10.1016/j.rse.2011.01.008. Exbrayat, J.-F., Bloom, A.A., Carvalhais, N., Fischer, R., Huth, A., Mac Bean, N., Williams, M., 2019. Understanding the land carbon cycle with space data: current status and prospects. Surv. Geophys. 40, 735–755. https://doi.org/ 10.1007/s10712-019-09506-2. Fang, H., Shang, Q., Chen, M., He, G., 2013. Changes in the critical erosion velocity for sediment colonized by biofilm. Sedimentology 61, 648–659. https://doi.org/10.1111/sed.12065. Fei, S., Phillips, J., Shouse, M., 2014. Biogeomorphic impacts of invasive species. Annu. Rev. Ecol. Evol. Syst. 45, 69–87. https://doi.org/10.1146/annurev-ecolsys-120213-091928. Feldpausch, T.R., Lloyd, J., Lewis, S.L., Brienen, R.J.W., Gloor, M., Mendoza, A.M., Lopez-Gonzalez, G., Banin, L., Salim, K.A., Affum-Baffoe, K., Alexiades, M., Almeida, S., Amaral, I., Andrade, A., Araga˜o, L.E.O.C., Murakami, A.A., Arets, E.J.M.M., Arroyo, L., Aymard C, G.A., Baker, T.R., Ba´nki, O.S., Berry, N.J., Cardozo, N., Chave, J., Comiskey, J.A., Alvarez, E., Oliveira, A.d., Fiore, A.D., Djagbletey, G., Domingues, T.F., Erwin, T.L., Fearnside, P.M., Franc¸a, M.B., Freitas, M.A., Higuchi, N., Iida, Y., Jimenez, E., Kassim, A.R.,

References

171

Killeen, T.J., Laurance, W.F., Lovett, J.C., Malhi, Y., Marimon, B.S., Marimon-Junior, B.H., Lenza, E., Marshall, A.R., Mendoza, C., Metcalfe, D.J., Mitchard, E.T.A., Neill, D.A., Nelson, B.W., Nilus, R., Nogueira, E.M., Parada, A., Peh, K.S.-H., Cruz, A.P., Pen˜uela, M.C., Pitman, N.C.A., Prieto, A., Quesada, C.A., Ramı´rez, F., Ramı´rez-Angulo, H., Reitsma, J.M., Rudas, A., Saiz, G., Saloma˜o, R.P., Schwarz, M., Silva, N., Silva-Espejo, J.E., Silveira, M., Sonke, B., Stropp, J., Taedoumg, H.E., Tan, S., Steege, H.t., Terborgh, J., TorelloRaventos, M., Heijden, G.M.F.v.d., Va´squez, R., Vilanova, E., Vos, V.A., White, L., Willcock, S., Woell, H., Phillips, O.L., 2012. Tree height integrated into pantropical forest biomass estimates. Biogeosciences 9, 3381–3403. https://doi.org/10.5194/bg-9-3381-2012. Fernandes, R., Plummer, S., Nightingale, J., Baret, F., Camacho, F., Fang, H., Garrigues, S., 2014. Global leaf area index product validation good practices. In: CEOS Working Group on Calibration and Validation-Land Product Validation Sub-Group. Version 2. CEOS. Ferraz, A., Saatchi, S., Mallet, C., Meyer, V., 2016. Lidar detection of individual tree size in tropical forests. Remote Sens. Environ. 183, 318–333. https://doi.org/10.1016/j.rse.2016.05.028. Ferry, B., Morneau, F., Bontemps, J.-D., Blanc, L., Freycon, V., 2010. Higher treefall rates on slopes and waterlogged soils result in lower stand biomass and productivity in a tropical rain forest. J. Ecol. 98, 106–116. https://doi.org/ 10.1111/j.1365-2745.2009.01604.x. Foster, D.R., King, G.A., Glaser, P.H., Wright, H.E., 1983. Origin of string patterns in boreal peatlands. Nature 306, 256–258. https://doi.org/10.1038/306256a0. odder, D., Secondi, J., 2014. Mapping species distributions with MAXENT using a geoFourcade, Y., Engler, J.O., R€ graphically biased sample of presence data: a performance assessment of methods for correcting sampling bias. PLoS One. 9. https://doi.org/10.1371/journal.pone.0097122. Frazer, G.W., Magnussen, S., Wulder, M.A., Niemann, K.O., 2011. Simulated impact of sample plot size and co-registration error on the accuracy and uncertainty of LiDAR-derived estimates of forest stand biomass. Remote Sens. Environ. 115, 636–649. https://doi.org/10.1016/j.rse.2010.10.008. Gabet, E.J., 2000. Gopher bioturbation: field evidence for non-linear hillslope diffusion. Earth Surf. Process. Landf. 25, 1419–1428. https://doi.org/10.1002/1096-9837(200012)25:133.0.co;2-1. Gabet, E.J., Perron, J.T., Johnson, D.L., 2014. Biotic origin for Mima mounds supported by numerical modeling. Geomorphology 206, 58–66. https://doi.org/10.1016/j.geomorph.2013.09.018. Gabet, E.J., Reichman, O.J., Seabloom, E.W., 2003. The effects of bioturbation on soil processes and sediment transport. Annu. Rev. Earth Planet. Sci. 31, 249–273. https://doi.org/10.1146/annurev.earth.31.100901.141314. Gao, B., 1996. NDWI—a normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sens. Environ. 58, 257–266. https://doi.org/10.1016/s0034-4257(96)00067-3. Gilbert, G.K., 1877. Report on the Geology of the Henry Mountains. Department of the Interior, US Geographical and Geological Survey of the Rocky Mountain Region, Washington D.C. Gillespie, T.W., Foody, G.M., Rocchini, D., Giorgi, A.P., Saatchi, S., 2008. Measuring and modelling biodiversity from space. Prog. Phys. Geog. 32, 203–221. https://doi.org/10.1177/0309133308093606. Gilman, E.F., 1989. Predicting root spread from trunk diameter and branch spread. Arboric. J. 13, 25–32. https://doi. org/10.1080/03071375.1989.9756398. Glennie, C.L., Carter, W.E., Shrestha, R.L., Dietrich, W.E., 2013. Geodetic imaging with airborne LiDAR: the Earth’s surface revealed. Rep. Prog. Phys. 76. https://doi.org/10.1088/0034-4885/76/8/086801. Go´mez, C., White, J.C., Wulder, M.A., 2016. Optical remotely sensed time series data for land cover classification: a review. ISPRS J. Photogramm. Remote Sens. 116, 55–72. https://doi.org/10.1016/j.isprsjprs.2016.03.008. Gong, P., Pu, R., Biging, G.S., Larrieu, M.R., 2003. Estimation of forest leaf area index using vegetation indices derived from Hyperion hyperspectral data. IEEE Trans. Geosci. Remote Sens. 41, 1355–1362. Gong, P., Wang, J., Yu, L., Zhao, Y., Zhao, Y., Liang, L., Niu, Z., Huang, X., Fu, H., Liu, S., Li, C., Li, X., Fu, W., Liu, C., Xu, Y., Wang, X., Cheng, Q., Hu, L., Yao, W., Zhang, H., Zhu, P., Zhao, Z., Zhang, H., Zheng, Y., Ji, L., Zhang, Y., Chen, H., Yan, A., Guo, J., Yu, L., Wang, L., Liu, X., Shi, T., Zhu, M., Chen, Y., Yang, G., Tang, P., Xu, B., Giri, C., Clinton, N., Zhu, Z., Chen, J., Chen, J., 2012. Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM$\mathplus$ data. Int. J. Remote Sens. 34, 2607–2654. https://doi.org/ 10.1080/01431161.2012.748992. Gonzalez, P., Asner, G.P., Battles, J.J., Lefsky, M.A., Waring, K.M., Palace, M., 2010. Forest carbon densities and uncertainties from Lidar, QuickBird, and field measurements in California. Remote Sens. Environ. 114, 1561–1575. https://doi.org/10.1016/j.rse.2010.02.011.

172

5. Linking life and landscape with remote sensing

Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., Moore, R., 2017. Google earth engine: planetaryscale geospatial analysis for everyone. Remote Sens. Environ. 202, 18–27. https://doi.org/10.1016/j. rse.2017.06.031. Gray, M.A., Baldauf, S.L., Mayhew, P.J., Hill, J.K., 2007. The response of avian feeding guilds to tropical forest disturbance. Conserv. Biol. 21, 133–141. https://doi.org/10.1111/j.1523-1739.2006.00557.x. Grieve, S.W.D., Mudd, S.M., Milodowski, D.T., Clubb, F.J., Furbish, D.J., 2016. How does grid-resolution modulate the topographic expression of geomorphic processes? Earth Surf. Dyn. 4, 627–653. https://doi.org/10.5194/esurf-4627-2016. Guariguata, M.R., 1990. Landslide disturbance and forest regeneration in the Upper Luquillo mountains of Puerto Rico. J. Ecol. 78, 814. https://doi.org/10.2307/2260901. Gurnell, A., 2013. Plants as river system engineers. Earth Surf. Process. Landf. 39, 4–25. https://doi.org/10.1002/ esp.3397. Gurnell, A.M., Piegay, H., Swanson, F.J., Gregory, S.V., 2002. Large wood and fluvial processes. Freshw. Biol. 47, 601–619. https://doi.org/10.1046/j.1365-2427.2002.00916.x. Gutman, G., Ignatov, A., 1998. The derivation of the green vegetation fraction from NOAA/AVHRR data for use in numerical weather prediction models. Int. J. Remote Sens. 19, 1533–1543. https://doi.org/10.1080/ 014311698215333. Gyssels, G., Poesen, J., Bochet, E., Li, Y., 2005. Impact of plant roots on the resistance of soils to erosion by water: a review. Prog. Phys. Geog. 29, 189–217. https://doi.org/10.1191/0309133305pp443ra. Haboudane, D., Bonn, F., Royer, A., Sommer, S., Mehl, W., 2002. Land degradation and erosion risk mapping by fusion of spectrally-based information and digital geomorphometric attributes. Int. J. Remote Sens. 23, 3795–3820. https://doi.org/10.1080/01431160110104638. Hales, T.C., 2018. Modelling biome-scale root reinforcement and slope stability. Earth Surf. Process. Landf. 43, 2157–2166. https://doi.org/10.1002/esp.4381. Hales, T.C., Ford, C.R., Hwang, T., Vose, J.M., Band, L.E., 2009. Topographic and ecologic controls on root reinforcement. J. Geophys. Res. 114, F03013. https://doi.org/10.1029/2008jf001168. Hamdan, O., Aziz, H.K., Rahman, K.A., 2011. Remotely sensed l-band sar data for tropical forest biomass estimation. J. Trop. For. Sci. 23, 318–327. Hancock, S., Anderson, K., Disney, M., Gaston, K.J., 2017. Measurement of fine-spatial-resolution 3D vegetation structure with airborne waveform lidar: calibration and validation with voxelised terrestrial lidar. Remote Sens. Environ. 188, 37–50. https://doi.org/10.1016/j.rse.2016.10.041. Hancock, S., Armston, J., Hofton, M., Sun, X., Tang, H., Duncanson, L.I., Kellner, J.R., Dubayah, R., 2019. The GEDI simulator: a large-footprint waveform lidar simulator for calibration and validation of spaceborne missions. Earth Space Sci. 6, 294–310. https://doi.org/10.1029/2018ea000506. Hancock, S., Armston, J., Li, Z., Gaulton, R., Lewis, P., Disney, M., Danson, F.M., Strahler, A., Schaaf, C., Anderson, K., Gaston, K.J., 2015. Waveform lidar over vegetation: an evaluation of inversion methods for estimating return energy. Remote Sens. Environ. 164, 208–224. https://doi.org/10.1016/j.rse.2015.04.013. Hancock, S., Lewis, P., Foster, M., Disney, M., Muller, J.-P., 2012. Measuring forests with dual wavelength lidar: a simulation study over topography. Agric. For. Meteorol. 161, 123–133. https://doi.org/10.1016/j.agrformet. 2012.03.014. Harding, D.J., Lefsky, M.A., Parker, G.G., Blair, J.B., 2001. Laser altimeter canopy height profiles: methods and validation for closed-canopy, broadleaf forests. Remote Sens. Environ. 76, 283–297. https://doi.org/10.1016/s00344257(00)00210-8. Harding, D.J., Carabajal, C.C., 2005. ICESat waveform measurements of within-footprint topographic relief and vegetation vertical structure. Geophys. Res. Lett. 32, L21S10. https://doi.org/10.1029/2005GL023471. Hergoualc’h, K., Verchot, L.V., 2011. Stocks and fluxes of carbon associated with land use change in Southeast Asian tropical peatlands: A review. Global Biogeochem. Cycles 25, GB2001. https://doi.org/10.1029/2009gb003718. Herold, M., Woodcock, C.E., Gregorio, A.d., Mayaux, P., Belward, A.S., Latham, J., Schmullius, C.C., 2006. A joint initiative for harmonization and validation of land cover datasets. IEEE Trans. Geosci. Remote Sens. 44, 1719–1727. https://doi.org/10.1109/tgrs.2006.871219. Hofton, M.A., Minster, J.B., Blair, J.B., 2000. Decomposition of laser altimeter waveforms. IEEE Trans. Geosci. Remote Sens. 38, 1989–1996. Hopkinson, C., Lovell, J., Chasmer, L., Jupp, D., Kljun, N., van Gorsel, E., 2013. Integrating terrestrial and airborne lidar to calibrate a 3D canopy model of effective leaf area index. Remote Sens. Environ. 136, 301–314. https://doi. org/10.1016/j.rse.2013.05.012.

References

173

Hosoi, F., Omasa, K., 2006. Voxel-based 3-D modeling of individual trees for estimating leaf area density using highresolution portable scanning lidar. IEEE Trans. Geosci. Remote Sens. 44, 3610–3618. https://doi.org/10.1109/ tgrs.2006.881743. Hwang, T., Band, L.E., Hales, T.C., Miniat, C.F., Vose, J.M., Bolstad, P.V., Miles, B., Price, K., 2015. Simulating vegetation controls on hurricane-induced shallow landslides with a distributed ecohydrological model. J. Geophys. Res. Biogeo. 120, 361–378. https://doi.org/10.1002/2014jg002824. Imhoff, M.L., 1995. A theoretical analysis of the effect of forest structure on synthetic aperture radar backscatter and the remote sensing of biomass. IEEE Trans. Geosci. Remote Sens. 33, 341–351. https://doi.org/10.1109/ tgrs.1995.8746015. Ingram, H.A.P., 1982. Size and shape in raised mire ecosystems: a geophysical model. Nature 297, 300–303. https:// doi.org/10.1038/297300a0. Istanbulluoglu, E., Bras, R.L., 2005. Vegetation-modulated landscape evolution: effects of vegetation on landscape processes, drainage density, and topography. J. Geophys. Res. 110. https://doi.org/10.1029/2004jf000249. Jang, C.-L., Shimizu, Y., 2007. Vegetation effects on the morphological behavior of alluvial channels. J. Hydraul. Res. 45, 763–772. https://doi.org/10.1080/00221686.2007.9521814. Jansson, C., Wullschleger, S.D., Kalluri, U.C., Tuskan, G.A., 2010. Phytosequestration: carbon biosequestration by plants and the prospects of genetic engineering. Bio Science 60, 685–696. https://doi.org/10.1525/bio.2010.60.9.6. Jarvis, P.G., McNaughton, K.G., 1986. Stomatal control of transpiration: scaling up from leaf to region. In: MacFadyen, A., Ford, E.D. (Eds.), Advances in Ecological Research. Academic Press, pp. 1–49. https://doi. org/10.1016/S0065-2504(08)60119-1. Jenkins, J.C., Chojnacky, D.C., Heath, L.S., Birdsey, R.A., 2003. National-scale biomass estimators for United States tree species. For. Sci. 49, 12–35. https://doi.org/10.1093/forestscience/49.1.12. Jiang, Z., Huete, A.R., Chen, J., Chen, Y., Li, J., Yan, G., Zhang, X., 2006. Analysis of NDVI and scaled difference vegetation index retrievals of vegetation fraction. Remote Sens. Environ. 101, 366–378. https://doi.org/10.1016/j. rse.2006.01.003. Joshi, N., Mitchard, E.T.A., Brolly, M., Schumacher, J., Ferna´ndez-Landa, A., Johannsen, V.K., Marchamalo, M., Fensholt, R., 2017. Understanding ‘saturation’ of radar signals over forests. Sci. Rep. 7, 3505. https://doi.org/ 10.1038/s41598-017-03469-3. Jucker, T., Bongalov, B., Burslem, D.F.R.P., Nilus, R., Dalponte, M., Lewis, S.L., Phillips, O.L., Qie, L., Coomes, D.A., 2018a. Topography shapes the structure, composition and function of tropical forest landscapes. Ecol. Lett. 21, 989–1000. https://doi.org/10.1111/ele.12964. Jucker, T., Asner, G.P., Dalponte, M., Brodrick, P.G., Philipson, C.D., Vaughn, N.R., Teh, Y.A., Brelsford, C., Burslem, D.F.R.P., Deere, N.J., Ewers, R.M., Kvasnica, J., Lewis, S.L., Malhi, Y., Milne, S., Nilus, R., Pfeifer, M., Phillips, O.L., Qie, L., Renneboog, N., Reynolds, G., Riutta, T., Struebig, M.J., Sva´tek, M., Turner, E.C., Coomes, D.A., 2018b. Estimating aboveground carbon density and its uncertainty in Borneo’s structurally complex tropical forests using airborne laser scanning. Biogeosciences 15, 3811–3830. https://doi.org/10.5194/ bg-15-3811-2018. Justice, C.O., Townshend, J.R.G., Holben, B.N., Tucker, C.J., 1985. Analysis of the phenology of global vegetation using meteorological satellite data. Int. J. Remote Sens. 6, 1271–1318. https://doi.org/10.1080/ 01431168508948281. Justice, C.O., Vermote, E., Townshend, J.R.G., Defries, R., Roy, D.P., Hall, D.K., Salomonson, V.V., Privette, J.L., Riggs, G., Strahler, A., Lucht, W., Myneni, R.B., Knyazikhin, Y., Running, S.W., Nemani, R.R., Wan, Z., Huete, A.R., Leeuwen, W.v., Wolfe, R.E., Giglio, L., Muller, J., Lewis, P., Barnsley, M.J., 1998. The moderate resolution imaging spectroradiometer (MODIS): land remote sensing for global change research. IEEE Trans. Geosci. Remote Sens. 36, 1228–1249. https://doi.org/10.1109/36.701075. Kaartinen, H., Hyypp€a, J., Yu, X., Vastaranta, M., Hyypp€a, H., Kukko, A., Holopainen, M., Heipke, C., Hirschmugl, M., Morsdorf, F., esset, E.N., Pitk€anen, J., Popescu, S., Solberg, S., Wolf, B.M., Wu, J.-C., 2012. An international comparison of individual tree detection and extraction using airborne laser scanning. Remote Sens. (Basel) 4, 950–974. https://doi.org/10.3390/rs4040950. Karam, M., Fung, A., Lang, R., Chauhan, N., 1992. A microwave-scattering model for layered vegetation. IEEE Trans. Geosci. Remote Sens. 30, 767–784. https://doi.org/10.1109/36.158872. Kayler, Z., Janowiak, M., Swanston, C., 2017. Global Carbon, U.S. Department of Agriculture, Forest Service, Climate Change Resource Center. https://www.fs.usda.gov/ccrc/topics/global-carbon. Keaney, A., McKinley, J., Graham, C., Robinson, M., Ruffell, A., 2013. Spatial statistics to estimate peat thickness using airborne radiometric data. Spat. Stat. 5, 3–24. https://doi.org/10.1016/j.spasta.2013.05.003.

174

5. Linking life and landscape with remote sensing

Kellner, J.R., Armston, J., Birrer, M., Cushman, K.C., Duncanson, L., Eck, C., Falleger, C., Imbach, B., Kra´l, K., Kru˚cek, M., Trochta, J., Vrsˇka, T., Zgraggen, C., 2019. New opportunities for forest remote sensing through ultra-high-density drone lidar. Surv. Geophys. 40, 959–977. https://doi.org/10.1007/s10712-019-09529-9. Kellner, J.R., Asner, G.P., Vitousek, P.M., Tweiten, M.A., Hotchkiss, S., Chadwick, O.A., 2011. Dependence of forest structure and dynamics on substrate age and ecosystem development. Ecosystems 14, 1156–1167. https://doi.org/ 10.1007/s10021-011-9472-4. Kellner, J.R., Clark, D.B., Hubbell, S.P., 2009. Pervasive canopy dynamics produce short-term stability in a tropical rain forest landscape. Ecol. Lett. 12, 155–164. https://doi.org/10.1111/j.1461-0248.2008.01274.x. Kemper, W.D., Bullock, M.S., Dexter, A.R., 1989. Soil cohesion changes. In: Larson, W.E., Blake, G.R., Allmaras, R.R., Voorhees, W.B., Gupta, S.C. (Eds.), Mechanics and Related Processes in Structured Agricultural Soils. NATO ASI SeriesSpringer, Netherlands, Dordrecht, pp. 81–95. https://doi.org/10.1007/978-94-009-2421-5_8. Kenrick, P., Crane, P.R., 1997. The origin and early evolution of plants on land. Nature 389, 33–39. https://doi.org/ 10.1038/37918. Kerr, J.T., Ostrovsky, M., 2003. From space to species: ecological applications for remote sensing. Trends Ecol. Evol. 18, 299–305. https://doi.org/10.1016/s0169-5347(03)00071-5. Kirkby, M.J., Irvine, B.J., Jones, R.J.A., Govers, G., 2008. The PESERA coarse scale erosion model for Europe. I. - Model rationale and implementation. Eur. J. Soil Sci. 59, 1293–1306. https://doi.org/10.1111/j.1365-2389.2008.01072.x. Kirkby, M.J., Neale, J.C., 1987. A soil erosion model incorporating seasonal factors. In: International Geomorphology. John Wiley, Chichester, pp. 189–210. Kirwan, M.L., Murray, A.B., 2007. A coupled geomorphic and ecological model of tidal marsh evolution. Proc. Natl. Acad. Sci. 104, 6118–6122. https://doi.org/10.1073/pnas.0700958104. Knorr, W., Kaminski, T., Scholze, M., Gobron, N., Pinty, B., Giering, R., Mathieu, P.-P., 2010. Carbon cycle data assimilation with a generic phenology model. J. Geophys. Res. 115, G04017. https://doi.org/10.1029/ 2009jg001119. Knyazikhin, Y., Martonchik, J.V., Myneni, R.B., Diner, D.J., Running, S.W., 1998. Synergistic algorithm for estimating vegetation canopy leaf area index and fraction of absorbed photosynthetically active radiation from MODIS and MISR data. J. Geophys. Res. Atmos. 103, 32257–32275. https://doi.org/10.1029/98jd02462. Koppel, J.v.d., Wal, D.v.d., Bakker, J.P., Herman, P.M.J., 2005. Self-organization and vegetation collapse in salt marsh ecosystems. Am. Nat. 165, E1–E12. https://doi.org/10.1086/426602. Korhonen, L., Morsdorf, F., 2014. Estimation of canopy cover, gap fraction and leaf area index with airborne laser scanning. In: Maltamo, M., Næsset, E., Vauhkonen, J. (Eds.), Forestry Applications of Airborne Laser Scanning: Concepts and Case Studies, Managing Forest Ecosystems. Springer, Netherlands, Dordrecht, pp. 397–417. https:// doi.org/10.1007/978-94-017-8663-8_20. Kugler, F., Lee, S., Hajnsek, I., Papathanassiou, K.P., 2015. Forest height estimation by means of Pol-InSAR data inversion: the role of the vertical wavenumber. IEEE Trans. Geosci. Remote Sens. 53, 5294–5311. https://doi.org/ 10.1109/TGRS.2015.2420996. Langbein, W.B., Schumm, S.A., 1958. Yield of sediment in relation to mean annual precipitation. Eos Trans. Am. Geophys. Union 39, 1076–1084. Lasaponara, R., Masini, N., 2013. Satellite synthetic aperture radar in archaeology and cultural landscape: an overview. Archaeol. Prospect. 20, 71–78. https://doi.org/10.1002/arp.1452. Le Toan, T., Beaudoin, A., Riom, J., Guyon, D., 1992. Relating forest biomass to SAR data. IEEE Trans. Geosci. Remote Sens. 30, 403–411. https://doi.org/10.1109/36.134089. Le Toan, T., Quegan, S., Davidson, M.W.J., Balzter, H., Paillou, P., Papathanassiou, K., Plummer, S., Rocca, F., Saatchi, S., Shugart, H., Ulander, L., 2011. The BIOMASS mission: mapping global forest biomass to better understand the terrestrial carbon cycle. Remote Sens. Environ. 115, 2850–2860. https://doi.org/10.1016/j. rse.2011.03.020. Lefsky, M.A., 2010. A global forest canopy height map from the moderate resolution imaging spectroradiometer and the geoscience laser altimeter system. Geophys. Res. Lett. 37, L15401. https://doi.org/10.1029/ 2010gl043622. Lefsky, M.A., Cohen, W.B., Harding, D.J., Parker, G.G., Acker, S.A., Gower, S.T., 2002. Lidar remote sensing of above-ground biomass in three biomes. Glob. Ecol. Biogeogr. 11, 393–399. https://doi.org/10.1046/j.1466822x.2002.00303.x. Lefsky, M.A., Harding, D., Cohen, W.B., Parker, G., Shugart, H.H., 1999. Surface lidar remote sensing of basal area and biomass in deciduous forests of Eastern Maryland, USA. Remote Sens. Environ. 67, 83–98. https://doi.org/ 10.1016/s0034-4257(98)00071-6.

References

175

Lefsky, M.A., Keller, M., Pang, Y., Camargo de, P.B., Hunter, M.O., 2007. Revised method for forest canopy height estimation from Geoscience Laser Altimeter System waveforms. J. Appl. Remote Sens. 1, 013537. Leitold, V., Keller, M., Morton, D.C., Cook, B.D., Shimabukuro, Y.E., 2015. Airborne lidar-based estimates of tropical forest structure in complex terrain: opportunities and trade-offs for REDD$\mathplus$. Carbon Balance Manag. 10. https://doi.org/10.1186/s13021-015-0013-x. Li, J., Okin, G.S., Alvarez, L., Epstein, H., 2007. Quantitative effects of vegetation cover on wind erosion and soil nutrient loss in a desert grassland of southern New Mexico, USA. Biogeochemistry 85, 317–332. https://doi.org/ 10.1007/s10533-007-9142-y. Li, W., Guo, Q., Jakubowski, M.K., Kelly, M., 2012. A new method for segmenting individual trees from the lidar point cloud. Photogramm. Eng. Remote Sens. 78, 75–84. https://doi.org/10.14358/pers.78.1.75. Liu, H.Q., Huete, A., 1995. A feedback based modification of the NDVI to minimize canopy background and atmospheric noise. IEEE Trans. Geosci. Remote Sens. 33, 457–465. https://doi.org/10.1109/36.377946. Los, S.O., Rosette, J.A.B., Kljun, N., North, P.R.J., Chasmer, L., Sua´rez, J.C., Hopkinson, C., Hill, R.A., van Gorsel, E., Mahoney, C., Berni, J.A.J., 2012. Vegetation height and cover fraction between 60° S and 60° N from ICESat GLAS data. Geosci. Model Dev. 5, 413–432. https://doi.org/10.5194/gmd-5-413-2012. Luckman, A.J., 1998. The effects of topography on mechanisms of radar backscatter from coniferous forest and upland pasture. IEEE Trans. Geosci. Remote Sens. 36, 1830–1834. https://doi.org/10.1109/36.718651. MacArthur, R.H., Horn, H.S., 1969. Foliage profile by vertical measurements. Ecology 50, 802–804. https://doi.org/ 10.2307/1933693. Mallet, C., Bretar, F., 2009. Full-waveform topographic lidar: state-of-the-art. ISPRS J. Photogramm. Remote Sens. 64, 1–16. https://doi.org/10.1016/j.isprsjprs.2008.09.007. Markham, B.L., Storey, J.C., Williams, D.L., Irons, J.R., 2004. Landsat sensor performance: history and current status. IEEE Trans. Geosci. Remote Sens. 42, 2691–2694. Marston, R.A., 2010. Geomorphology and vegetation on hillslopes: interactions, dependencies, and feedback loops. Geomorphology 116, 206–217. https://doi.org/10.1016/j.geomorph.2009.09.028. Mascaro, J., Detto, M., Asner, G.P., Muller-Landau, H.C., 2011. Evaluating uncertainty in mapping forest carbon with airborne LiDAR. Remote Sens. Environ. 115, 3770–3774. https://doi.org/10.1016/j.rse.2011.07.019. Matsushita, B., Yang, W., Chen, J., Onda, Y., Qiu, G., 2007. Sensitivity of the enhanced vegetation index (EVI) and normalized difference vegetation index (NDVI) to topographic effects: a case study in high-density cypress forest. Sensors 7, 2636–2651. https://doi.org/10.3390/s7112636. Maul, G.A., Gordon, H.R., 1975. On the use of the earth resources technology satellite (LANDSAT-1) in optical oceanography. Remote Sens. Environ. 4, 95–128. https://doi.org/10.1016/0034-4257(75)90008-5. Mauri, L., Sallustio, L., Tarolli, P., 2019. The geomorphologic forcing of wild boars. Earth Surf. Process. Landf. 44, 2085–2094. https://doi.org/10.1002/esp.4623. Mcdonald, K.C., Dobson, M.C., Ulaby, F.T., 1990. Using mimics to model L-band multiangle and multitemporal backscatter from a walnut orchard. IEEE Trans. Geosci. Remote Sens. 28, 477–491. https://doi.org/10.1109/ tgrs.1990.572925. McLean, K.A., Trainor, A.M., Asner, G.P., Crofoot, M.C., Hopkins, M.E., Campbell, C.J., Martin, R.E., Knapp, D.E., Jansen, P.A., 2016. Movement patterns of three arboreal primates in a Neotropical moist forest explained by LiDAR-estimated canopy structure. Landsc. Ecol. 31, 1849–1862. https://doi.org/10.1007/s10980-016-0367-9. Meron, E., 2012. Pattern-formation approach to modelling spatially extended ecosystems. Ecol. Model. 234, 70–82. https://doi.org/10.1016/j.ecolmodel.2011.05.035. Millar, R.G., 2000. Influence of bank vegetation on alluvial channel patterns. Water Resour. Res. 36, 1109–1118. Milodowski, D.T., Mudd, S.M., Mitchard, E.T.A., 2015. Erosion rates as a potential bottom-up control of forest structural characteristics in the Sierra Nevada Mountains. Ecology 96, 31–38. https://doi.org/10.1890/14-0649.1. Mitchard, E.T., Saatchi, S.S., Baccini, A., Asner, G.P., Goetz, S.J., Harris, N.L., Brown, S., 2013. Uncertainty in the spatial distribution of tropical forest biomass: a comparison of pan-tropical maps. Carbon Balance Manag. 8, 10. https://doi.org/10.1186/1750-0680-8-10. Mitchard, E.T.A., Feldpausch, T.R., Brienen, R.J.W., Lopez-Gonzalez, G., Monteagudo, A., Baker, T.R., Lewis, S.L., Lloyd, J., Quesada, C.A., Gloor, M., Steege, H.t., Meir, P., Alvarez, E., Araujo-Murakami, A., Araga˜o, L.E.O.C., Arroyo, L., Aymard, G., Banki, O., Bonal, D., Brown, S., Brown, F.I., Cero´n, C.E., Moscoso, V.C., Chave, J., Comiskey, J.A., Cornejo, F., Medina, M.C., Costa, L.D., Costa, F.R.C., Fiore, A.D., Domingues, T.F., Erwin, T.L., Frederickson, T., Higuchi, N., Coronado, E.N.H., Killeen, T.J., Laurance, W.F., Levis, C., Magnusson, W.E., Marimon, B.S., Junior, B.H.M., Polo, I.M., Mishra, P., Nascimento, M.T., Neill, D., Vargas, M.P.N., Palacios, W.A., Parada, A., Molina, G.P., Pen˜a-Claros, M., Pitman, N., Peres, C.A., Poorter, L., Prieto, A.,

176

5. Linking life and landscape with remote sensing

Ramirez-Angulo, H., Correa, Z.R., Roopsind, A., Roucoux, K.H., Rudas, A., Saloma˜o, R.P., Schietti, J., Silveira, M., Souza, P.F.d., Steininger, M.K., Stropp, J., Terborgh, J., Thomas, R., Toledo, M., Torres-Lezama, A., Andel, T.R.v., Heijden, G.M.F.v.d., Vieira, I.C.G., Vieira, S., Vilanova-Torre, E., Vos, V.A., Wang, O., Zartman, C.E., Malhi, Y., Phillips, O.L., 2014. Markedly divergent estimates of Amazon forest carbon density from ground plots and satellites. Glob. Ecol. Biogeogr. 23, 935–946. https://doi.org/10.1111/geb.12168. Mitchard, E.T.A., Saatchi, S.S., Woodhouse, I.H., Nangendo, G., Ribeiro, N.S., Williams, M., Ryan, C.M., Lewis, S.L., Feldpausch, T.R., Meir, P., 2009. Using satellite radar backscatter to predict above-ground woody biomass: a consistent relationship across four different African landscapes. Geophys. Res. Lett. 36, L23401. https://doi. org/10.1029/2009gl040692. Mitchell, S.J., 2012. Wind as a natural disturbance agent in forests: a synthesis. Forestry 86, 147–157. https://doi.org/ 10.1093/forestry/cps058. Moffett, K.B., Robinson, D.A., Gorelick, S.M., 2010. Relationship of salt marsh vegetation zonation to spatial patterns in soil moisture, salinity, and topography. Ecosystems 13, 1287–1302. https://doi.org/10.1007/s10021-010-9385-7. Mohr, J.J., Madsen, S.N., 2001. Geometric calibration of ERS satellite SAR images. IEEE Trans. Geosci. Remote Sens. 39, 842–850. https://doi.org/10.1109/36.917909. Muchoney, D., Borak, J., Chi, H., Friedl, M., Gopal, S., Hodges, J., Morrow, N., Strahler, A., 2000. Application of the MODIS global supervised classification model to vegetation and land cover mapping of Central America. Int. J. Remote Sens. 21, 1115–1138. https://doi.org/10.1080/014311600210100. Mudd, S.M., D’Alpaos, A., Morris, J.T., 2010. How does vegetation affect sedimentation on tidal marshes? Investigating particle capture and hydrodynamic controls on biologically mediated sedimentation. J. Geophys. Res. 115, F03029. https://doi.org/10.1029/2009jf001566. Myneni, R.B., Hoffman, S., Knyazikhin, Y., Privette, J.L., Glassy, J., Tian, Y., Wang, Y., Song, X., Zhang, Y., Smith, G.R., Lotsch, A., Friedl, M., Morisette, J.T., Votava, P., Nemani, R.R., Running, S.W., 2002. Global products of vegetation leaf area and fraction absorbed PAR from year one of MODIS data. Remote Sens. Environ. 83, 214–231. https://doi. org/10.1016/s0034-4257(02)00074-3. Myneni, R.B., Ramakrishna, R., Nemani, R., Running, S.W., 1997. Estimation of global leaf area index and absorbed par using radiative transfer models. IEEE Trans. Geosci. Remote Sens. 35, 1380–1393. https://doi.org/ 10.1109/36.649788. Naesset, E., 2009. Effects of different sensors, flying altitudes, and pulse repetition frequencies on forest canopy metrics and biophysical stand properties derived from small-footprint airborne laser data. Remote Sens. Environ. 113, 148–159. https://doi.org/10.1016/j.rse.2008.09.001. Naesset, E., Gobakken, T., 2008. Estimation of above- and below-ground biomass across regions of the boreal forest zone using airborne laser. Remote Sens. Environ. 112, 3079–3090. https://doi.org/10.1016/j.rse.2008.03.004. Nanson, G.C., Knighton, A.D., 1996. Anabranching rivers: their cause, character and classification. Earth Surf. Process. Landf. 21, 217–239. https://doi.org/10.1002/(sici)1096-9837(199603)21:33.0.co;2-u. Narine, L.L., Popescu, S., Neuenschwander, A., Zhou, T., Srinivasan, S., Harbeck, K., 2019. Estimating aboveground biomass and forest canopy cover with simulated ICESat-2 data. Remote Sens. Environ. 224, 1–11. https://doi.org/ 10.1016/j.rse.2019.01.037. Naylor, L.A., Viles, H.A., Carter, N.E.A., 2002. Biogeomorphology revisited: looking towards the future. Geomorphology 47, 3–14. https://doi.org/10.1016/s0169-555x(02)00137-x. Nepf, H.M., 2012. Hydrodynamics of vegetated channels. J. Hydraul. Res. 50, 262–279. https://doi.org/ 10.1080/00221686.2012.696559. Neuenschwander, A.L., Magruder, L.A., 2016. The potential impact of vertical sampling uncertainty on ICESat2/ATLAS Terrain and canopy height retrievals for multiple ecosystems. Remote Sens. 8, 1039. https://doi. org/10.3390/rs8121039. Neuenschwander, A., Popescu, S., Nelson, R., Harding, D., Pitts, K., Robbins, J., Pederson, D., Sheridan, R., 2019. ICE, CLOUD, and Land Elevation Satellite (ICESat-2) Algorithm Theoretical Basis Document (ATBD) for Land - Vegetation Along-track products (ATL08) (ATBD Document No. Release 001). NASA. Niklas, K.J., 2006. A phyletic perspective on the allometry of plant biomass-partitioning patterns and functionally equivalent organ-categories. New Phytol. 171, 27–40. https://doi.org/10.1111/j.1469-8137.2006.01760.x. Nikora, V., 2010. Hydrodynamics of aquatic ecosystems: an interface between ecology, biomechanics and environmental fluid mechanics. River Res. Appl. 26, 367–384. https://doi.org/10.1002/rra.1291. Ni-Meister, W., Jupp, D.L.B., Dubayah, R., 2001. Modeling lidar waveforms in heterogeneous and discrete canopies. IEEE Trans. Geosci. Remote Sens. 39, 1943–1958. https://doi.org/10.1109/36.951085.

References

177

Noffke, N., Gerdes, G., Klenke, T., Krumbein, W.E., 2001. Microbially induced sedimentary structures: a new category within the classification of primary sedimentary structures. J. Sediment. Res. 71, 649–656. https://doi.org/ 10.1306/2dc4095d-0e47-11d7-8643000102c1865d. Osterkamp, W.R., Hupp, C.R., Stoffel, M., 2011. The interactions between vegetation and erosion: new directions for research at the interface of ecology and geomorphology. Earth Surf. Process. Landf. 37, 23–36. https://doi.org/ 10.1002/esp.2173. € € N., Hızal, A., G€ okbulak, F., Serengil, Y., 2005. Cover and management factors for the Ozhan, S., Balcı, A.N., Ozyuvaci, Universal Soil-Loss Equation for forest ecosystems in the Marmara region, Turkey. For. Ecol. Manage. 214, 118–123. https://doi.org/10.1016/j.foreco.2005.03.050. Papathanassiou, K.P., Cloude, S.R., 2001. Single-baseline polarimetric SAR interferometry. IEEE Trans. Geosci. Remote Sens. 39 (11), 2352–2363. Park, S.-E., Jung, Y.T., Cho, J.-H., Moon, H., Han, S.-H., 2019. Theoretical evaluation of water cloud model vegetation parameters. Remote Sens. (Basel) 11, 894. https://doi.org/10.3390/rs11080894. Parker, G.G., Harmon, M.E., Lefsky, M.A., Chen, J., Pelt, R.V., Weis, S.B., Thomas, S.C., Winner, W.E., Shaw, D.C., Frankling, J.F., 2004. Three-dimensional structure of an old-growth Pseudotsuga-Tsuga canopy and its implications for radiation balance, microclimate, and gas exchange. Ecosystems 7, 440–453. https://doi.org/10.1007/ s10021-004-0136-5. Pastick, N.J., Jorgenson, M.T., Wylie, B.K., Minsley, B.J., Ji, L., Walvoord, M.A., Smith, B.D., Abraham, J.D., Rose, J.R., 2013. Extending airborne electromagnetic surveys for regional active layer and permafrost mapping with remote sensing and ancillary data, Yukon Flats Ecoregion, Central Alaska. Permafr. Periglac. Process. 24, 184–199. https:// doi.org/10.1002/ppp.1775. Patterson, P.L., Healey, S.P., ahl, G.S., Saarela, S., Holm, S., Andersen, H.-E., Dubayah, R.O., Duncanson, L., Hancock, S., Armston, J., Kellner, J.R., Cohen, W.B., Yang, Z., 2019. Statistical properties of hybrid estimators proposed for GEDI—NASA’s global ecosystem dynamics investigation. Environ. Res. Lett. 14. https://doi.org/ 10.1088/1748-9326/ab18df. Pawlik, L., 2013. The role of trees in the geomorphic system of forested hillslopes—a review. Earth Sci. Rev. 126, 250–265. https://doi.org/10.1016/j.earscirev.2013.08.007. Pawlik, L., Phillips, J.D., Sˇamonil, P., 2016. Roots, rock, and regolith: biomechanical and biochemical weathering by trees and its impact on hillslopes—a critical literature review. Earth Sci. Rev. 159, 142–159. https://doi.org/ 10.1016/j.earscirev.2016.06.002. Paz-Kagan, T., Brodrick, P.G., Vaughn, N.R., Das, A.J., Stephenson, N.L., Nydick, K.R., Asner, G.P., 2017. What mediates tree mortality during drought in the southern Sierra Nevada? Ecol. Appl. 27, 2443–2457. https://doi.org/ 10.1002/eap.1620. Pelletier, J.D., 2012. A spatially distributed model for the long-term suspended sediment discharge and delivery ratio of drainage basins. J. Geophys. Res. Earth 117, F02028. https://doi.org/10.1029/2011jf002129. Perucca, E., Camporeale, C., Ridolfi, L., 2007. Significance of the riparian vegetation dynamics on meandering river morphodynamics. Water Resour. Res. 43, W03430. https://doi.org/10.1029/2006wr005234. Pollen, N., Simon, A., 2005. Estimating the mechanical effects of riparian vegetation on stream bank stability using a fiber bundle model. Water Resour. Res. 41, W07025. https://doi.org/10.1029/2004wr003801. Porder, S., Paytan, A., Vitousek, P.M., 2004. Erosion and landscape development affect plant nutrient status in the Hawaiian Islands. Oecologia 142, 440–449. https://doi.org/10.1007/s00442-004-1743-8. Qi, W., Lee, S.-K., Hancock, S., Luthcke, S., Tang, H., Armston, J., Dubayah, R., 2019. Improved forest height estimation by fusion of simulated GEDI Lidar data and TanDEM-X InSAR data. Remote Sens. Environ. 221, 621–634. https://doi.org/10.1016/j.rse.2018.11.035. Quegan, S., Toan, T.L., Chave, J., Dall, J., Exbrayat, J.-F., Minh, D.H.T., Lomas, M., D’Alessandro, M.M., Paillou, P., Papathanassiou, K., Rocca, F., Saatchi, S., Scipal, K., Shugart, H., Smallman, T.L., Soja, M.J., Tebaldini, S., Ulander, L., Villard, L., Williams, M., 2019. The European Space Agency BIOMASS mission: measuring forest above-ground biomass from space. Remote Sens. Environ. 227, 44–60. https://doi.org/10.1016/j.rse.2019.03.032. Quesada, C.A., Phillips, O.L., Schwarz, M., Czimczik, C.I., Baker, T.R., Patin˜o, S., Fyllas, N.M., Hodnett, M.G., Herrera, R., Almeida, S., Da´vila, E.A., Arneth, A., Arroyo, L., Chao, K.J., Dezzeo, N., Erwin, T., di Fiore, A., Higuchi, N., Coronado, E.H., Jimenez, E.M., Killeen, T., Lezama, A.T., Lloyd, G., Lo´pez-Gonza´lez, G., Luiza˜o, F.J., Malhi, Y., Monteagudo, A., Neill, D.A., Vargas, P.N., Paiva, R., Peacock, J., Pen˜uela, M.C., Cruz, A.P., Pitman, N., Filho, N.P., Prieto, A., Ramı´rez, H., Rudas, A., Saloma˜o, R., Santos, A.J.B., Schmerler, J., Silva, N., Silveira, M., Va´squez, R., Vieira, I., Terborgh, J., Lloyd, J., 2012. Basin-wide variations in Amazon forest

178

5. Linking life and landscape with remote sensing

structure and function are mediated by both soils and climate. Biogeosciences 9, 2203–2246. https://doi.org/ 10.5194/bg-9-2203-2012. Raney, R.K., Luscombe, A.P., Langham, E.J., Ahmed, S., 1991. RADARSAT (SAR imaging). Proc. IEEE 79, 839–849. https://doi.org/10.1109/5.90162. Ranson, K.J., Sun, G., 1994. Mapping biomass of a northern forest using multifrequency SAR data. IEEE Trans. Geosci. Remote Sens. 32, 388–396. https://doi.org/10.1109/36.295053. Rast, M., Bezy, J.L., Bruzzi, S., 1999. The ESA medium resolution imaging spectrometer MERIS a review of the instrument and its mission. Int. J. Remote Sens. 20, 1681–1702. https://doi.org/10.1080/014311699212416. Ravi, S., Breshears, D.D., Huxman, T.E., D’Odorico, P., 2010. Land degradation in drylands: interactions among hydrologic–aeolian erosion and vegetation dynamics. Geomorphology 116, 236–245. https://doi.org/10.1016/j. geomorph.2009.11.023. Rejou-Mechain, M., Tanguy, A., Piponiot, C., Chave, J., Herault, B., 2017. Biomass: an r package for estimating aboveground biomass and its uncertainty in tropical forests. Methods Ecol. Evol. 8, 1163–1167. https://doi.org/ 10.1111/2041-210x.12753. Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., Yoder, D.C., 1997. Predicting soil erosion by water: a guide to conservation planning with the revised universal soil loss equation (RUSLE). In: USDA Agriculture Handbook (Washington). USDA, Washington, DC, p. 703. Reubens, B., Poesen, J., Danjon, F., Geudens, G., Muys, B., 2007. The role of fine and coarse roots in shallow slope stability and soil erosion control with a focus on root system architecture: a review. Trees 21, 385–402. https:// doi.org/10.1007/s00468-007-0132-4. Rietkerk, M., Koppel, J.v.d., 2008. Regular pattern formation in real ecosystems. Trends Ecol. Evol. 23, 169–175. https://doi.org/10.1016/j.tree.2007.10.013. Rizzo, D.M., Slaughter, G.W., Parmeter, Jr, J.R., 2000. Enlargement of canopy gaps associated with a fungal pathogen in Yosemite Valley, California. Can. J. For. Res. 30, 1501–1510. https://doi.org/10.1139/x00-073. Roering, J.J., Schmidt, K.M., Stock, J.D., Dietrich, W.E., Montgomery, D.R., 2003. Shallow landsliding, root reinforcement, and the spatial distribution of trees in the Oregon Coast Range. Can. Geotech. J. 40, 237–253. https://doi. org/10.1139/t02-113. Roner, M., D’Alpaos, A., Ghinassi, M., Marani, M., Silvestri, S., Franceschinis, E., Realdon, N., 2016. Spatial variation of salt-marsh organic and inorganic deposition and organic carbon accumulation: inferences from the Venice lagoon, Italy. Adv. Water Resour. 93, 276–287. https://doi.org/10.1016/j.advwatres.2015.11.011. Rosen, P., Hensley, S., Shaffer, S., Edelstein, W., Kim, Y., Kumar, R., Misra, T., Bhan, R., Sagi, R., 2017. The NASA-ISRO SAR (NISAR) mission dual-band radar instrument preliminary design. In: 2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS). Presented at the 2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), pp. 3832–3835. https://doi.org/10.1109/IGARSS.2017.8127836. Rosenqvist, A., Shimada, M., Ito, N., Watanabe, M., 2007. ALOS PALSAR: a pathfinder mission for global-scale monitoring of the environment. IEEE Trans. Geosci. Remote Sens. 45, 3307–3316. https://doi.org/10.1109/ TGRS.2007.901027. Rosette, J.A.B., North, P.R.J., Sua´rez, J.C., Los, S.O., 2010. Uncertainty within satellite LiDAR estimations of vegetation and topography. Int. J. Remote Sens. 31, 1325–1342. https://doi.org/10.1080/01431160903380631. Roussel, J.-R., Caspersen, J., Beland, M., Thomas, S., Achim, A., 2017. Removing bias from LiDAR-based estimates of canopy height: accounting for the effects of pulse density and footprint size. Remote Sens. Environ. 198, 1–16. https://doi.org/10.1016/j.rse.2017.05.032. Roy, D.P., Wulder, M.A., Loveland, T.R., Woodcock, C.E., Allen, R.G., Anderson, M.C., Helder, D., Irons, J.R., Johnson, D.M., Kennedy, R., Scambos, T.A., Schaaf, C.B., Schott, J.R., Sheng, Y., Vermote, E.F., Belward, A.S., Bindschadler, R., Cohen, W.B., Gao, F., Hipple, J.D., Hostert, P., Huntington, J., Justice, C.O., Kilic, A., Kovalskyy, V., Lee, Z.P., Lymburner, L., Masek, J.G., McCorkel, J., Shuai, Y., Trezza, R., Vogelmann, J., Wynne, R.H., Zhu, Z., 2014. Landsat-8: science and product vision for terrestrial global change research. Remote Sens. Environ. 145, 154–172. https://doi.org/10.1016/j.rse.2014.02.001. Rydin, H.a., Jeglum, J.K., 2013. Peatland habitats. In: The Biology of Peatlands. Oxford University Press, pp. 1–20. https://doi.org/10.1093/acprof:osobl/9780199602995.003.0001. Saarela, S., Holm, S., Healey, S., Andersen, H.-E., Petersson, H., Prentius, W., Patterson, P., esset, E.N., Gregoire, T., ahl, G.S., 2018. Generalized hierarchical model-based estimation for aboveground biomass assessment using GEDI and Landsat data. Remote Sens. (Basel) 10, 1832. https://doi.org/10.3390/rs10111832.

References

179

Saatchi, S.S., Harris, N.L., Brown, S., Lefsky, M., Mitchard, E.T.A., Salas, W., Zutta, B.R., Buermann, W., Lewis, S.L., Hagen, S., Petrova, S., White, L., Silman, M., Morel, A., 2011. Benchmark map of forest carbon stocks in tropical regions across three continents. PNAS 108, 9899–9904. https://doi.org/10.1073/pnas. 1019576108. Saco, P.M., Heras, M.M.l., 2013. Ecogeomorphic coevolution of semiarid hillslopes: emergence of banded and striped vegetation patterns through interaction of biotic and abiotic processes. Water Resour. Res. 49, 115–126. https://doi. org/10.1029/2012wr012001. Saco, P.M., Willgoose, G.R., Hancock, G.R., 2007. Eco-geomorphology of banded vegetation patterns in arid and semiarid regions. Hydrol. Earth Syst. Sci. 11, 1717–1730. https://doi.org/10.5194/hess-11-1717-2007. Saint, G., 1996. SPOT 4 VEGETATION system: association with high resolution data for multiscale studies. Adv. Space Res. 17, 107–110. https://doi.org/10.1016/0273-1177(95)00455-N. Calibration and Applications of Satellite Sensors for Environmental Monitoring. Schaaf, C.B., Gao, F., Strahler, A.H., Lucht, W., Li, X., Tsang, T., Strugnell, N.C., Zhang, X., Jin, Y., Muller, J.-P., Lewis, P., Barnsley, M., Hobson, P., Disney, M., Roberts, G., Dunderdale, M., Doll, C., d’Entremont, R.P., Hu, B., Liang, S., Privette, J.L., Roy, D., 2002. First operational BRDF, albedo nadir reflectance products from MODIS. Remote Sens. Environ. 83, 135–148. https://doi.org/10.1016/S0034-4257(02)00091-3. The Moderate Resolution Imaging Spectroradiometer (MODIS): a new generation of Land Surface Monitoring. Schamper, C., Jørgensen, F., Auken, E., Effersø, F., 2014. Assessment of near-surface mapping capabilities by airborne transient electromagnetic data—an extensive comparison to conventional borehole data. Geophysics 79, B187–B199. https://doi.org/10.1190/geo2013-0256.1. Scharlemann, J.P., Tanner, E.V., Hiederer, R., Kapos, V., 2014. Global soil carbon: understanding and managing the largest terrestrial carbon pool. Carbon Manag. 5, 81–91. https://doi.org/10.4155/cmt.13.77. Schmidt, K.M., Roering, J.J., Stock, J.D., Dietrich, W.E., Montgomery, D.R., Schaub, T., 2001. The variability of root cohesion as an influence on shallow landslide susceptibility in the Oregon Coast Range. Can. Geotech. J. 38, 995–1024. https://doi.org/10.1139/t01-031. Schneider, F.D., Morsdorf, F., Schmid, B., Petchey, O.L., Hueni, A., Schimel, D.S., Schaepman, M.E., 2017. Mapping functional diversity from remotely sensed morphological and physiological forest traits. Nat. Commun. 8. https://doi.org/10.1038/s41467-017-01530-3. Schnitzer, S.A., Carson, W.P., 2001. Treefall gaps and the maintenance of species diversity in a tropical forest. Ecology 82, 913. https://doi.org/10.2307/2679891. Shi, J., Malik, J., 2000. Normalized cuts and image segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 22, 888–905. https://doi.org/10.1109/34.868688. utz, H., Pfaffhuber, A.A., Viezzoli, A., 2019a. Peatland volume Silvestri, S., Christensen, C.W., Lysdahl, A.O.K., Ansch€ mapping over resistive substrates with airborne electromagnetic technology. Geophys. Res. Lett. 46, 6459–6468. https://doi.org/10.1029/2019gl083025. Silvestri, S., Knight, R., Viezzoli, A., Richardson, C.J., Anshari, G.Z., Dewar, N., Flanagan, N., Comas, X., 2019b. Quantification of peat thickness and stored carbon at the landscape scale in tropical peatlands: a comparison of airborne geophysics and an empirical topographic method. J. Geophys. Res. Earth Surf. 124, 3107–3123. https://doi.org/ 10.1029/2019JF005273. Silvestri, S., Marani, M., Marani, A., 2003. Hyperspectral remote sensing of salt marsh vegetation, morphology and soil topography. Phys. Chem. Earth 28, 15–25. https://doi.org/10.1016/s1474-7065(03)00004-4. Simard, M., Pinto, N., Fisher, J.B., Baccini, A., 2011. Mapping forest canopy height globally with spaceborne lidar. J. Geophys. Res. 116. https://doi.org/10.1029/2011jg001708. Sinha, S., Jeganathan, C., Sharma, L.K., Nathawat, M.S., 2015. A review of radar remote sensing for biomass estimation. Int. J. Environ. Sci. Technol. 12, 1779–1792. https://doi.org/10.1007/s13762-015-0750-0. Skakun, S., Justice, C.O., Vermote, E., Roger, J.-C., 2017. Transitioning from MODIS to VIIRS: an analysis of interconsistency of NDVI data sets for agricultural monitoring. Int. J. Remote Sens. 39, 971–992. https://doi.org/ 10.1080/01431161.2017.1395970. Slatton, K.C., Carter, W.E., Shrestha, R.L., Dietrich, W., 2007. Airborne laser swath mapping: achieving the resolution and accuracy required for geosurficial research. Geophys. Res. Lett. 34, L23S10. https://doi.org/ 10.1029/2007gl031939. Smallman, T.L., Exbrayat, J.-F., Mencuccini, M., Bloom, A.A., Williams, M., 2017. Assimilation of repeated woody biomass observations constrains decadal ecosystem carbon cycle uncertainty in aggrading forests. J. Geophys. Res. Biogeo. 122, 528–545. https://doi.org/10.1002/2016jg003520.

180

5. Linking life and landscape with remote sensing

Smith, R., Knight, R., 2019. Modeling land subsidence using InSAR and airborne electromagnetic data. Water Resour. Res. 55, 2801–2819. https://doi.org/10.1029/2018wr024185. Solberg, S., Astrup, R., Breidenbach, J., Nilsen, B., Weydahl, D., 2013. Monitoring spruce volume and biomass with InSAR data from TanDEM-X. Remote Sens. Environ. 139, 60–67. https://doi.org/10.1016/j.rse.2013.07.036. Solberg, S., Naesset, E., Bollandsas, O.M., 2006. Single tree segmentation using airborne laser scanner data in a structurally heterogeneous spruce forest. Photogramm. Eng. Remote Sens. 72, 1369–1378. https://doi.org/10.14358/ pers.72.12.1369. Spies, T.A., Franklin, J.F., Klopsch, M., 1990. Canopy gaps in Douglas-fir forests of the Cascade Mountains. Can. J. For. Res. 20, 649–658. https://doi.org/10.1139/x90-087. Stark, S.C., Enquist, B.J., Saleska, S.R., Leitold, V., Schietti, J., Longo, M., Alves, L.F., Camargo, P.B., Oliveira, R.C., 2015. Linking canopy leaf area and light environments with tree size distributions to explain Amazon forest demography. Ecol. Lett. 18, 636–645. https://doi.org/10.1111/ele.12440. Stark, S.C., Leitold, V., Wu, J.L., Hunter, M.O., Castilho, C.V.d., Costa, F.R.C., McMahon, S.M., Parker, G.G., Shimabukuro, M.T., Lefsky, M.A., Keller, M., Alves, L.F., Schietti, J., Shimabukuro, Y.E., Branda˜o, D.O., Woodcock, T.K., Higuchi, N., Camargo, P.B.d., Oliveira, R.C.d., Saleska, S.R., 2012. Amazon forest carbon dynamics predicted by profiles of canopy leaf area and light environment. Ecol. Lett. 15, 1406–1414. https://doi.org/ 10.1111/j.1461-0248.2012.01864.x. Steltzer, H., Welker, J.M., 2006. Modelling the effect of photosynthetic vegetation properties of the NDVI-LAI relationship. Ecology 87, 2765–2772. https://doi.org/10.1890/0012-9658(2006)87[2765:mteopv]2.0.co;2. Stokes, A., Atger, C., Bengough, A.G., Fourcaud, T., Sidle, R.C., 2009. Desirable plant root traits for protecting natural and engineered slopes against landslides. Plant and Soil 324, 1–30. https://doi.org/10.1007/s11104-009-0159-y. Struebig, M.J., Turner, A., Giles, E., Lasmana, F., Tollington, S., Bernard, H., Bell, D., 2013. Quantifying the biodiversity value of repeatedly logged rainforests. In: Global Change in Multispecies Systems: Part 3. Elsevier, pp. 183–224. https://doi.org/10.1016/b978-0-12-417199-2.00003-3. Sun, G., Ranson, K.J., Kharuk, V.I., 2002. Radiometric slope correction for forest biomass estimation from SAR data in the Western Sayani Mountains, Siberia. Remote Sens. Environ. 79, 279–287. https://doi.org/10.1016/s0034-4257 (01)00279-6. Swenson, N.G., Enquist, B.J., 2007. Ecological and evolutionary determinants of a key plant functional trait: wood density and its community-wide variation across latitude and elevation. Am. J. Bot. 94, 451–459. https://doi. org/10.3732/ajb.94.3.451. Swetnam, T.L., Brooks, P.D., Barnard, H.R., Harpold, A.A., Gallo, E.L., 2017. Topographically driven differences in energy and water constrain climatic control on forest carbon sequestration. Ecosphere. 8, e01797. https://doi.org/ 10.1002/ecs2.1797. Tang, H., Armston, J., Hancock, S., Marselis, S., Goetz, S., Dubayah, R., 2019a. Characterizing global forest canopy cover distribution using spaceborne lidar. Remote Sens. Environ. 231, 111262. https://doi.org/10.1016/ j.rse.2019.111262. Tang, H., Dubayah, R., Brolly, M., Ganguly, S., Zhang, G., 2014. Large-scale retrieval of leaf area index and vertical foliage profile from the spaceborne waveform lidar (GLAS/ICESat). Remote Sens. Environ. 154, 8–18. https://doi. org/10.1016/j.rse.2014.08.007. Tang, H., Dubayah, R., Swatantran, A., Hofton, M., Sheldon, S., Clark, D.B., Blair, B., 2012. Retrieval of vertical LAI profiles over tropical rain forests using waveform lidar at La Selva, Costa Rica. Remote Sens. Environ. 124, 242–250. https://doi.org/10.1016/j.rse.2012.05.005. Tang, H., Song, X.-P., Zhao, F.A., Strahler, A.H., Schaaf, C.L., Goetz, S., Huang, C., Hansen, M.C., Dubayah, R., 2019b. Definition and measurement of tree cover: a comparative analysis of field-, lidar- and landsat-based tree cover estimations in the Sierra national forests, USA. Agric. For. Meteorol. 268, 258–268. https://doi.org/10.1016/ j.agrformet.2019.01.024. Tang, H., Swatantran, A., Barrett, T., DeCola, P., Dubayah, R., 2016. Voxel-based spatial filtering method for canopy height retrieval from airborne single-photon lidar. Remote Sens. (Basel) 8, 771. https://doi.org/10.3390/ rs8090771. Tarnita, C.E., Bonachela, J.A., Sheffer, E., Guyton, J.A., Coverdale, T.C., Long, R.A., Pringle, R.M., 2017. A theoretical foundation for multi-scale regular vegetation patterns. Nature 541, 398–401. https://doi.org/10.1038/ nature20801. Thomson, S., Fountain, D., Watts, T., Airborne geophysics—evolution and revolution: Fifth Decennial International Conference on Mineral Exploration, Toronto, Ontario, Decennial Mineral Exploration Conferences, Proceedings, 2007, 19–37.

References

181

Toomey, M.R., Ashton, A.D., Raymo, M.E., Perron, J.T., 2016. Late Cenozoic sea level and the rise of modern rimmed atolls. Palaeogeogr. Palaeoclimatol. Palaeoecol. 451, 73–83. https://doi.org/10.1016/j.palaeo.2016.03.018. Torres, R., Snoeij, P., Geudtner, D., Bibby, D., Davidson, M., Attema, E., Potin, P., Rommen, B., Floury, N., Brown, M., Traver, I.N., Deghaye, P., Duesmann, B., Rosich, B., Miranda, N., Bruno, C., L’Abbate, M., Croci, R., Pietropaolo, A., Huchler, M., Rostan, F., 2012. GMES Sentinel-1 mission. Remote Sens. Environ. 120, 9–24. https://doi.org/10.1016/j.rse.2011.05.028. The Sentinel Missions - New Opportunities for Science. Tucker, C.J., 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens. Environ. 8, 127–150. https://doi.org/10.1016/0034-4257(79)90013-0. Tucker, C.J., Pinzon, J.E., Brown, M.E., Slayback, D.A., Pak, E.W., Mahoney, R., Vermote, E.F., Saleous, N.E., 2005. An extended AVHRR 8-km NDVI dataset compatible with MODIS and SPOT vegetation NDVI data. Int. J. Remote Sens. 26, 4485–4498. https://doi.org/10.1080/01431160500168686. Vafaei, S., Soosani, J., Adeli, K., Fadaei, H., Naghavi, H., Pham, T., Bui, D.T., 2018. Improving accuracy estimation of forest aboveground biomass based on incorporation of ALOS-2 PALSAR-2 and sentinel-2A imagery and machine learning: a case study of the Hyrcanian forest area (Iran). Remote Sens. (Basel) 10, 172. https://doi.org/10.3390/ rs10020172. Van Oost, K., Govers, G., Desmet, P., 2000. Evaluating the effects of changes in landscape structure on soil erosion by water and tillage. Landsc. Ecol. 15, 577–589. https://doi.org/10.1023/a:1008198215674. Vargas, M., Miura, T., Shabanov, N., Kato, A., 2013. An initial assessment of Suomi NPP VIIRS vegetation index EDR. J. Geophys. Res. Atmos. 118 (12), 301–12,316. https://doi.org/10.1002/2013jd020439. Verstraeten, G., Poesen, J., Vente, J.d., Koninckx, X., 2003. Sediment yield variability in Spain: a quantitative and semiqualitative analysis using reservoir sedimentation rates. Geomorphology 50, 327–348. https://doi.org/ 10.1016/s0169-555x(02)00220-9. Viezzoli, A., Tosi, L., Teatini, P., Silvestri, S., 2010. Surface water-groundwater exchange in transitional coastal environments by airborne electromagnetics: the Venice Lagoon example. Geophys. Res. Lett 37, L01402. https://doi. org/10.1029/2009gl041572. Vincent, G., Antin, C., Laurans, M., Heurtebize, J., Durrieu, S., Lavalley, C., Dauzat, J., 2017. Mapping plant area index of tropical evergreen forest by airborne laser scanning. A cross-validation study using LAI2200 optical sensor. Remote Sens. Environ. 198, 254–266. https://doi.org/10.1016/j.rse.2017.05.034. von Arx, G., Pannatier, E.G., Thimonier, A., Rebetez, M., 2013. Microclimate in forests with varying leaf area index and soil moisture: potential implications for seedling establishment in a changing climate. J. Ecol. 101, 1201–1213. https://doi.org/10.1111/1365-2745.12121. von Hardenberg, J., Meron, E., Shachak, M., Zarmi, Y., 2001. Diversity of vegetation patterns and desertification. Phys. Rev. Lett. 87. https://doi.org/10.1103/physrevlett.87.198101. Wallace, L., Lucieer, A., Watson, C., Turner, D., 2012. Development of a UAV-LiDAR system with application to forest inventory. Remote Sens. (Basel) 4, 1519–1543. https://doi.org/10.3390/rs4061519. Wang, X., Zhao, X., Zhang, Z., Yi, L., Zuo, L., Wen, Q., Liu, F., Xu, J., Hu, S., Liu, B., 2016. Assessment of soil erosion change and its relationships with land use/cover change in China from the end of the 1980s to 2010. Catena 137, 256–268. https://doi.org/10.1016/j.catena.2015.10.004. West, G.B., Brown, J.H., Enquist, B.J., 1999. A general model for the structure and allometry of plant vascular systems. Nature 400, 664–667. https://doi.org/10.1038/23251. Wieser, M., Mandlburger, G., Hollaus, M., Otepka, J., Glira, P., Pfeifer, N., 2017. A case study of UAS borne laser scanning for measurement of tree stem diameter. Remote Sens. (Basel) 9, 1154. https://doi.org/10.3390/rs9111154. Wilkinson, B.H., McElroy, B.J., 2007. The impact of humans on continental erosion and sedimentation. Geol. Soc. Am. Bull. 119, 140–156. https://doi.org/10.1130/b25899.1. Wilkinson, M.T., Richards, P.J., Humphreys, G.S., 2009. Breaking ground: pedological, geological, and ecological implications of soil bioturbation. Earth Sci. Rev. 97, 257–272. https://doi.org/10.1016/j.earscirev.2009.09.005. Wischmeier, W.H., Johnson, C.B., Cross, B.V., 1971. A soil erodibility nomograph for farmland and construction sites. J. Soil Water Conserv. 26, 189–193. Woodcock, C.E., Allen, R., Anderson, M., Belward, A., Bindschadler, R., Cohen, W., Gao, F., Goward, S.N., Helder, D., Helmer, E., Nemani, R., Oreopoulos, L., Schott, J., Thenkabail, P.S., Vermote, E.F., Vogelmann, J., Wulder, M.A., Wynne, R., 2008. Free access to Landsat imagery. Science 320. https://doi.org/10.1126/science.320.5879.1011a 1011a–1011a. Woodgate, W., Armston, J.D., Disney, M., Jones, S.D., Suarez, L., Hill, M.J., Wilkes, P., Soto-Berelov, M., 2016. Quantifying the impact of woody material on leaf area index estimation from hemispherical photography using 3D canopy simulations. Agric. For. Meteorol. 226–227, 1–12. https://doi.org/10.1016/j.agrformet.2016.05.009.

182

5. Linking life and landscape with remote sensing

Woodhouse, I.H., 2006. Introduction to Microwave Remote Sensing. CRC Press, Boca Raton, Florida. Woodhouse, I.H., Mitchard, E.T.A., Brolly, M., Maniatis, D., Ryan, C.M., 2012. Radar backscatter is not a ‘direct measure’ of forest biomass. Nat. Clim. Chang. 2, 556–557. https://doi.org/10.1038/nclimate1601. Wulder, M.A., White, J.C., Nelson, R.F., esset, E.N., Ørka, H.O., Coops, N.C., Hilker, T., Bater, C.W., Gobakken, T., 2012. Lidar sampling for large-area forest characterization: a review. Remote Sens. Environ. 121, 196–209. https:// doi.org/10.1016/j.rse.2012.02.001. Yanai, R.D., Battles, J.J., Richardson, A.D., Blodgett, C.A., Wood, D.M., Rastetter, E.B., 2010. Estimating uncertainty in ecosystem budget calculations. Ecosystems 13, 239–248. https://doi.org/10.1007/s10021-010-9315-8. Yengoh, G.T., Dent, D., Olsson, L., Tengberg, A.E., Tucker, C.J., 2015. Main global NDVI datasets, databases, and software. In: Use of the Normalized Difference Vegetation Index (NDVI) to Assess Land Degradation at Multiple Scales. Springer International Publishing, pp. 45–50. https://doi.org/10.1007/978-3-319-24112-8_8. Yoo, K., Amundson, R., Heimsath, A.M., Dietrich, W.E., 2005. Process-based model linking pocket gopher (Thomomys bottae) activity to sediment transport and soil thickness. Geology 33, 917. https://doi.org/ 10.1130/g21831.1. Yu, Q., Gong, P., Clinton, N., Biging, G., Kelly, M., Schirokauer, D., 2006. Object-based detailed vegetation classification. with airborne high spatial resolution remote sensing imagery. Photogramm. Eng. Remote Sens. 72, 799–811. https://doi.org/10.14358/PERS.72.7.799. Zanne, A.E., Lopez-Gonzalez, G., Coomes, D.A., Ilic, J., Jansen, S., Lewis, S.L., Miller, R.B., Swenson, N.G., Wiemann, M.C., Chave, J., 2009. Global wood density database. Online: http://datadryad.org/resource/ doi:10.5061/dryad.234. Zeng, X., Dickinson, R.E., Walker, A., Shaikh, M., DeFries, R.S., Qi, J., 2000. Derivation and evaluation of global 1-km fractional vegetation cover data for land modeling. J. Appl. Meteorol. 39, 826–839. https://doi.org/10.1175/15200450(2000)0392.0.co;2. Zhao, F., Guo, Q., Kelly, M., 2012. Allometric equation choice impacts lidar-based forest biomass estimates: a case study from the Sierra National Forest, CA. Agric. For. Meteorol. 165, 64–72. https://doi.org/10.1016/j. agrformet.2012.05.019. Zheng, G., Moskal, L.M., 2009. Retrieving leaf area index (LAI) using remote sensing: Theories, methods and sensors. Sensors 9, 2719–2745. https://doi.org/10.3390/s90402719. Zhou, P., Luukkanen, O., Tokola, T., Nieminen, J., 2008. Effect of vegetation cover on soil erosion in a mountainous watershed. Catena 75, 319–325. https://doi.org/10.1016/j.catena.2008.07.010. Zwally, H.J., Schutz, B., Abdalati, W., Abshire, J., Bentley, C., Brenner, A., Bufton, J., Dezio, J., Hancock, D., Harding, D., Herring, T., Minster, B., Quinn, K., Palm, S., Spinhirne, J., Thomas, R., 2002. ICESat’s laser measurements of polar ice, atmosphere, ocean, and land. J. Geodyn. 34, 405–445. https://doi.org/10.1016/S0264-3707(02) 00042-X.

C H A P T E R

6 SfM photogrammetry for GeoArchaeology Sara Cucchiaroa, Daniel J. Fallub, Pengzhi Zhaoc, Clive Waddingtond, David Cockcroftd, Paolo Tarollia, Antony G. Brownb,e a

Department of Land, Environment, Agriculture and Forestry, University of Padova, Legnaro, Padova, Italy bTromso University Museum, UiT The Artic University of Norway, Tromsø, Norway cEarth & Life Institute, Universite Catholique de Louvain, Louvain-la-Neuve, Belgium d Archaeological Research Services Ltd, Bakewell, DE, United Kingdom eGeography and Environmental Science, University of Southampton, Southampton, United Kingdom

O U T L I N E 1 Remote sensing

183

2 SfM photogrammetry

185

3 SfM in geoarchaeology: Agricultural terraces in Europe 187 3.1 Case study: Ingram Valley (UK) 188

3.2 SfM workflow 3.3 Result and discussion

191 196

4 Final remarks

200

Acknowledgments

200

References

201

1 Remote sensing The use of remote-sensing (RS) data, from imaging to scanning, has now become an integral and routine part of geoarchaeological studies. Even in the early days of aerialphotographic imagery, it was realized that this technology could, under different light and ground conditions, reveal significant subsurface information, particularly in arable lands

Developments in Earth Surface Processes, Volume 23 https://doi.org/10.1016/B978-0-444-64177-9.00006-0

183

© 2020 Elsevier B.V. All rights reserved.

184

6. SfM photogrammetry for GeoArchaeology

through so-called “crop marks” (Barber, 2011). In addition, site recording (or planning) was routinely augmented by high-resolution, oblique photography from extendable poles, or photographic towers (Fussell, 1982). This offered some three-dimensional (3D) capability from stereo pairs, but this was limited and digital photogrammetry has only really advanced with the advent of digital single-lens reflex (DSLR) cameras and sufficient computing power (Doyon et al., 2019). The next major RS development in geoarchaeological studies was the use of wavelengths at the edge or outside the visible part of the electromagnetic spectrum, particularly near infrared (NI) and infrared (IR). NI has proved particularly valuable for demarcating field systems, including infields from outfields, and settlement plans through differences in vegetation and soil properties (Verhoeven et al., 2009; Verhoeven, 2012). Examples include Bronze Age fields systems on Bodmin Moor, UK (Johnson et al., 2008), and the mapping of the Roman town of Altinum on the Po Plain during a severe drought in 2007 (Ninfo et al., 2009). Although it was realized that satellite remote sensing could be valuable for archeology back in the early days of its availability (Lasaponara and Masini, 2011), the low spatial resolution of early data limited its use in geoarchaeology to large-scale systems, such as irrigation networks and tells in semiarid regions (Kouchoukos, 2001; Parcak, 2007). However, from the availability of data from the Landsat TM satellite (which had a spatial precision of 30 m), and SPOT satellite (with resolution down to 10 m) onwards, more geoarchaeological applications have emerged. Examples include the mapping of Roman centuriation (Romano and Tolba, 1996) and the landscape around Stonehenge in England (Fowler, 1995). Even higher spatial resolution with Quickbird satellite multispectral imagery has allowed the use of both NIR and more complex indices such as the Normalized Difference Vegetation Index (NDVI) for the mapping of medieval crop marks in southern Italy (Lasaponara and Masini, 2007). The advantage of NDVI is it can detect crop marks through the vigor of crops or other vegetation. A related method is the Tasseled cap transformation, which can be used to estimate soil depth in plowed fields (Brown et al., 1990). The advent in the 1990s of airborne scanners was a revolution in the use of RS data in geoarcheology. Active methods, such as light detection and ranging (LiDAR), have now become almost a standard in archeology (Brown, 2008; Evans et al., 2013; H€ammerle and H€ ofle, 2018; Beach et al., 2019; Penny et al., 2019; Tarolli et al., 2019) and can provide invaluable information in three ways; first, because of the ability of LiDAR to penetrate vegetation including woodland; second, because of the reflection of subsurface conditions through micro-topography; and third, because of the potential information value of additional data, such as intensity of the return signal. One of the first demonstrations of the ability of LiDAR to penetrate woodland was the discovery of field boundaries under ancient woodland in the Forest of Dean, UK (Hoyle, 2008), which was quickly followed by other National Parks in the United States and elsewhere including the United States (USGS, 2011; New Forest, 2016; South Downs National Park, 2019). Combining LiDAR data with that from aerial photographs and geomorphological mapping to drive geoarchaeological evaluation and prospection programs in advance of development, particularly for large quarries, was pioneered in northern England as part of the Till-Tweed project (Passmore and Waddington, 2009, 2012) and which gave rise to the endorsement of this approach in English planning guidance (MHEF, 2008). LiDAR has been used in the archeological evaluation of large developments such as the high-speed rail projects (Georges-Leroy et al., 2013). High-resolution

2 SfM photogrammetry

185

topography can both reflect human activities (Cao et al., 2020) and/or natural features such as paleochannels that are sediment traps ideal for geoarchaeological studies. Indeed, this has been formalized into a protocol for the evaluation of the geoarchaeological potential of areas of gravel extraction that commonly border floodplains in Europe (Carey et al., 2006, 2017). In these studies, the intensity of LiDAR return is used to map wetter areas that normally correspond to deeper soils, fine and organic sediments, and negative features. The most advanced scanning currently is the use of airborne multispectral and hyperspectral scanners that again can be used for crop marks (Aqdus et al., 2008), classical city plans (Cavalli et al., 2007), and even shallow marine features and survey (Guyot et al., 2019). In many ways, the development of ground-based systems has mirrored that of airborne remote sensing, except that developments in civil engineering and geological monitoring were also important. Early long-range, distance-laser scanners were used in the early 2000s to monitor cliff failures (Rosser et al., 2005; Lim et al., 2010), river-bed morphology (Brasington et al., 2012), debris flow (Blasone et al., 2014), rockfalls (Williams et al., 2018), and glacial environments (Whitworth et al., 2006). The earliest and invaluable archeological applications of terrestrial laser scanners (TLS) were in cave mapping, which allowed the modeling of cave geometry and the creation of exact replica caves (Gonza´lez-Aguilera et al., 2009), and the recording of complex ancient classical world structures (Brutto et al., 2017). TLS has unrivaled utility in the scanning of inaccessible archeology, such as intertidal archeology and it can be used to model processes associated with the archeological features such as tidal mill basin volume (Lobb et al., in press). Due to both its accuracy and speed, TLS is also highly suited to the monitoring of erosion that can threaten archeological sites such as coastal prehistoric sites around the North Sea (Lobb and Brown, 2016). A development— terrestrial hyperspectral scanning—has been used to record excavation stratigraphy from a Neolithic site in northern Sweden (Linderholm et al., 2019). Both high-resolution aerial photography and TLS are particularly suitable for mapping cultivation terraces and lynchets (cultivation ridges on slopes), which due to their scale (1–5 m in typical riser height), are not normally recorded on topographic maps. This has been done for historic period agricultural terraces in Catalonia (Kinnaird et al., 2017) and is applied here to prehistoric terraces. Now, new high-resolution survey techniques are available and they allow us to undertake low-cost and very detailed surveys in the field of geoarchaeology. One of the most successful emerging techniques in high-resolution topographic (HRT) survey is structure-from-motion (SfM) photogrammetry (Westoby et al., 2012), which was born from the evolution of classical photogrammetry but exploits the advantages of digital photography and computer vision.

2 SfM photogrammetry Nowadays, SfM photogrammetry paired with multiview stereo (MVS), hereafter together referred to as SfM, represents a powerful and successful tool to produce high-quality 3D surfaces for geoscience applications. In the literature, several researches have used this technology to carry out different kinds of analysis and studies on: structural geology (e.g., Bemis et al., 2014); debris-flow dynamics (Cucchiaro et al., 2019); surveying submerged surfaces (e.g., Woodget et al., 2015; Dietrich, 2017); soil erosion (Glendell et al., 2017); design of drainage network (Pijl et al., 2019) or agricultural terraces 3D reconstruction (Pijl et al., 2020);

186

6. SfM photogrammetry for GeoArchaeology

gullies and badlands (e.g., Smith and Vericat, 2015; St€ ocker et al., 2015; Koci et al., 2017); fluvial morphology (e.g., Javernick et al., 2014; Marteau et al., 2017) and aquatic applications (Carrivick and Smith, 2018); glaciers (e.g., Immerzeel et al., 2014; Piermattei et al., 2015; Mallalieu et al., 2017); monitoring on landslide displacement (e.g., Stumpf et al., 2015; Turner et al., 2015; Clapuyt et al., 2017; Eker et al., 2018); coastal recession (e.g., James and Robson, 2012; Westoby et al., 2012); open-pit mining areas (Chen et al., 2015; Xiang et al., 2018); and extraction of biophysical forest or plants parameters and monitoring (e.g., Zarco-Tejada et al., 2014; Malambo et al., 2018; Iglhaut et al., 2019). Moreover, studies are shifting from proof-ofconcepts in topographic survey to genuine applications including quantification of bathymetric surveys, underwater archeology, grain-size mapping, restoration monitoring, habitat classification, geomorphological change detection, and sediment transport path delineation (Carrivick and Smith, 2018). In short time, SfM has had a transformative effect on geoscience research providing exceptionally fast, low-cost, and easy 3D survey (Fonstad et al., 2013), with point accuracies comparable to other HRT survey methods [e.g., TLS, LiDAR, and global navigation satellite systems (GNSS); Tarolli, 2014]. Clapuyt et al. (2016) showed that the accuracies obtained with SfM were of the same order of magnitude as those obtained with more traditional HRT survey methods for a broad range of landforms and landscapes. SfM has proved to be extremely versatile and useful in different environments, where traditional techniques had high costs. For example, in complex and rugged environment, the use of methods such as TLS is limited by access constraints (e.g., for large instruments) and the power requirements in remote areas (Westoby et al., 2012). The use of LiDAR for surveys of small extension has still relatively high costs, requires specific processing and sometimes does not reach the required accuracy and the point density in complex terrains (Victoriano et al., 2018), whereas SfM images acquisition is several orders of magnitude cheaper. Furthermore, the issues of cost and time constraints for some methods can make it difficult to conduct repeated surveys, i.e., multitemporal surveys needed to properly characterize geomorphic processes. The increasing use of a SfM is linked to the development of user-friendly SfM software (Cucchiaro et al., 2018b) and the use of the unmanned aerial vehicles (UAV) that have evolved greatly in the last decade in electronic sophistication, ease-of-use, and reduced cost. Now, different kind of UAVs meet different requirements in the SfM surveys (Carrivick et al., 2016). Moreover, SfM allows the choice of a wide range of other acquisition platforms (Table 1) based on the features of the surveyed area: pixel resolution, spatial coverage, image quality, and cost effectiveness (Smith et al., 2015). The SfM technique also offers the possibility of integrating images taken from different acquisition platforms if certain working methods are respected. For example, an integrated approach combining ground-based and aerial images can help overcome site-specific disadvantages (e.g., ground-based images are not able to guarantee areal coverage, whereas aerial photos may show a poor representation of vertical surfaces, being influenced by the vegetation). However, to carry out the data fusion between aerial and ground photos, it is important to use the same camera with the same focal length to minimize the integration problems in the photogrammetric models (Cucchiaro et al., 2018a). This approach also benefits from the acquisition of data from two different observation directions (i.e., nadir for UAV images and oblique for terrestrial images; St€ ocker et al., 2015). In general, the choice of the sensor, the flight height, and the focal length are fundamental aspects to be considered (O’Connor et al., 2017).

3 SfM in geoarchaeology: Agricultural terraces in Europe

TABLE 1

187

SfM platforms types and their features

SfM platforms

Main features

Survey scale

Fixed-wing aircraft

Long-range capability, highly efficient in terms of energy use, demands a takeoff and landing strip (not be feasible in remote and/or rugged terrain)

Large areas

Dual rotor systems (e.g., Heli)

Restricted battery life, highly flexible systems for almost any terrain, not suitable in blustery conditions

Medium range

Multicopters

High flexibility in complex topography and stability in most weather conditions, but limited range and flight time

Medium scale

Kites, lighter-thanair balloons

Full control over the frequency and target of image acquisition, not suitable in windy conditions, limited by a moderate maximum operation height

Medium scale

Gyrocopter

Wide swath imagery, flying not possible in adverse weather

Large areas

Hand-held poles

Fine spatial resolution imagery, complete control over image acquisition

Detail scale

Ground-based (handheld)

Detail-scale 3D reconstruction, especially of the steep or subvertical surfaces, limited spatial coverage

Fine spatial scale

The application of SfM photogrammetry technique also requires the appropriate software to postprocess photos and a network of ground control points (GCPs) to scale and georeference the SfM results. GCPs are fundamental for the accuracy and repeatability of the survey ( James et al., 2017a,b). The great versatility of SfM is now offering an optimal platform for archeology (Howland et al., 2014; Mertes et al., 2014; Prins et al., 2014; Bojakowski et al., 2015; Landeschi et al., 2016; Pierdicca et al., 2016) that benefits from fresh technological developments to record the 3D structures. Indeed, the traditional protocols based on hand-drawn plans and sections no longer come up to the standards of precision achieved by the new methods in recording the archeological structures more accurately (Lo´pez et al., 2016). The results of SfM photogrammetry can be processed further to create 3D models and scaled plans for the study of the physical and functional characteristics of surveyed objects and, in geoarchaeology research where it can record both topographies and sections.

3 SfM in geoarchaeology: Agricultural terraces in Europe Agricultural terraces are not just archeological features but were fundamental to the success of European agriculture in hilly terrains, and were until recently, part of a sustainable agricultural and social system. TerrACE archeological research project (ERC-2017-ADG: 787790, 2018–2023; https://www.terrace.no/) is a 5-year European Research Council grant funded by the European Union. The goals of the TerrACE Project are to create a methodological step change in the understanding of terraces by applying new scientific methodology to agricultural terraces across Europe, by bringing together landscape archeology, geomorphology, and paleoecology. The techniques address several themes including: the mapping and recording of terraces and lynchets in as finer detail as is possible, dating terrace systems, and understanding their original and later purposes and uses. The improve mapping of terrace

188

6. SfM photogrammetry for GeoArchaeology

landscapes can be reached thought HRT techniques (Sofia et al., 2014), also using automatic extraction algorithms (Tarolli et al., 2014). HRT can be used to identify agricultural terraced walls, spatial heterogeneity, and multi-temporal measures of terrace degradation through topographic attributes. These approaches start from the availability of large-scale topographic LiDAR datasets that allow construction of a high-resolution ( 1 m) digital terrain models (DTMs) from the bare ground data, by filtering vegetation from raw LiDAR data. These allow the mapping of terraces in areas where photointerpretation is not possible, such as through woodland, and in areas where no previous information is available; for example, vegetated terraced sites in remote zones. The LiDAR data can be used for a first and rapid assessment of the location of terraces particularly in abandoned systems that might require management and renovation planning. Moreover, the proposed procedure is an efficient approach that overcomes classic difficulties associated with working on large scales, approaching private owners, and accessing terraced areas for conducting ground surveys over large areas. Once terraced sites have been labeled and identified, the SfM technique (using UAV) can be used to carry out higher-resolution surveys and DTMs ( 0.25–0.10 m) useful to analyze in detail the topographic features (scaled plans, profiles, and sections) and attributes of terraces systems. Instead, in the areas where the LiDAR data are not available or sufficiently accurate in terms of resolution, the SfM technique offers the possibility, as mentioned above, to carry out very detailed surveys to detect terraced areas through a specific workflow in which multiple acquisition platforms can be used to overcome the limits related to the SfM survey scale and vegetated zones.

3.1 Case study: Ingram Valley (UK) The TerrACE project is examining a sample of terrace systems that represents nearly all of Europe’s climatic zones in six study areas: Ingram Valley and other sites in the United Kingdom (maritime temperate; Frodsham and Waddington, 2004), Leikanger and Sognefjorden in Norway (cool maritime; Skrede, 2005), Pays de Herve, Belgium (continental temperate; Van Oost et al., 2000), Valla d’Arene and St Victoire in the French Alps (humid Mediterranean; Walsh and Mocci, 2003), Cinque Terre Ligurian Hills, and globally important agricultural heritage systems (GIAHS) Soave traditional vineyards in Italy (Mediterranean; Tarolli et al., 2014), and Stymphalos and sites in eastern Crete (dry Mediterranean; Walsh et al., 2017). The study presented here is from the first study case in the Ingram Valley in the Cheviot Hills of NE England within the Northumberland National Park (Fig. 1). The site is located immediately adjacent to Plantation Camp enclosure on the east slope of the hillside below Brough Law Hillfort, approximately 1 km west of Ingram village in the upper Breamish valley. The park is known for its upland multi-period archeological landscapes (Frodsham and Waddington, 2004) and the features on Ingram Farm are a scheduled monument because they are a fine example of this multilayered or palimpsest landscape (Lotherington and Waddington, 2019). Features include cairnfields, settlements, hillfort/enclosures, field systems, and agricultural terraces. It is one of the largest scheduled monuments in England (5.7 km2). This study focuses on the Plantation Camp agricultural terraces, which have received previous archeological attention. Two trenches were excavated in 1997/1998 and

3 SfM in geoarchaeology: Agricultural terraces in Europe

189

FIG. 1 Location of study area. Ingram Valley—Northumberland National Park (UK). The photo (taken in May 2019) shows the Ingram Valley looking north-west with medieval ridge and furrow in the foreground, the prehistoric agricultural terraces to the right of the plantation cloaked in brown vegetation, and the river Breamish further to the right.

a longer trench in 1999 by Waddington (Frodsham and Waddington, 2004). The archeological sequence comprises the cultivation terraces as the earliest component that are currently radiocarbon dated as commencing in the Early Bronze Age c. 1800–1500 BC, which are in turn overlain by a trackway that leads to a late Iron Age or Roman Iron Age enclosure (Plantation Camp). Further up the hillside on the crown of the hill is the well-preserved remains of a stone-walled hillfort known as Brough Law, which has been radiocarbon dated to the first few centuries BC in the late Iron Age. The next phase of activity is evidenced by a large expanse of broad ridge and furrow cultivation remains of Anglo-Saxon origin that overly the lowest part of the prehistoric cultivation terraces. A postmedieval, stone-walled enclosure and outfield boundary system overlies the ridge and furrow. Prehistoric cultivation terraces are rare in the United Kingdom and so the detailed survey and excavation undertaken as part of this project is of national importance. In all, there are seven terraces covering a small area of about 9000 m2 (Fig. 2A). Important aims of the work include determining the form and construction of the terraces, which initially appeared indeterminate in form between true bench-type terraces with wall risers and lynchets. The case presented here is particularly interesting and challenging as in the Ingram landscape a palimpsest of terraces occurs from the prehistoric to the postmedieval period with very thick vegetation cover in the form of bracken. We also aim, eventually, to be able to tie the subsurface and chronostratigraphic models together in four-dimensional (4D) agricultural terrace heritage models. Satellite imagery from Google Earth vaguely shows the prehistoric agricultural terracing running along the contour, with the much later better-preserved medieval ridge and furrow (Fig. 2C) showing clearly running across the slope. It is also just visible on open source LiDAR data provided by the UK Environment Agency (Data Service Platform; https://environment.data.gov.uk/). This LiDAR data cover the whole Ingram valley (Fig. 2B), however, the DTMs derived from LiDAR survey have a resolution of 2 m (Fig. 2B), which is not enough to identify and map in detail all the terraces and lynchets in

0

A

25

50

100

m

0

50

100

200

m

B

C

FIG. 2 The Ingram terrace site: (A) orthophoto of terraces site in 2007. (B) DTM of Ingram Valley at 2 m resolution provided by the UK Environment Agency. (C) Screenshot of satellite imagery from Google Earth of Ingram terraces site with the prehistoric agricultural terraces, Plantation Camp enclosures, and the medieval ridge and furrow marked.

191

3 SfM in geoarchaeology: Agricultural terraces in Europe

the study area (some of them have heights below 1 m). For this reason, a SfM survey was carried out to realize higher-resolution topographic data of the Ingram terrace area, together with the excavation and sampling from a new geoarchaeological excavation trench (65 m  1 m) that encompassed the length of the prehistoric agricultural terrace sequence. The HRT survey facilitated the analysis of geomorphological features, the topographic recording, and measurement of the various archeological remains, as well as the recording of the excavation, based on the high-resolution data from the DTM.

3.2 SfM workflow 3.2.1 Fieldwork In SfM surveys, the choice of the appropriate SfM platform is a key aspect. After a detailed analysis of the field site, we decided to integrate ground-based and UAV (nadir and oblique) images because this area is very challenging to survey on the ground, given the huge level of bracken infestation across the lower slopes of the hillside covering the medieval ridge and furrow and the agricultural terraces (Fig. 4A). The aerial survey gave us the possibility of covering a large area in a short time, and therefore, we chose to survey a wider zone (around 40 ha; Fig. 3) than just the terrace area, while the ground-based photos captured the fine and otherwise hidden details. In particular, we analyzed the area from the Brough Law

N W

E S

0

50

100

200

m

FIG. 3 SfM survey and GCPs network in the Ingram study area.

192

6. SfM photogrammetry for GeoArchaeology

Hillfort (situated overlooking the Plantation Camp terraces, as well as the much of the rest of the nearby Breamish Valley) to the Breamish river to study the long-term evolution of this tract of landscape in finer detail than was hitherto possible. By surveying up to the river, this allowed the morphology of the valley side to be compared with that of the valley floor and the opportunity to determine whether past agricultural remains could be detected on the flood plain, as well as any evidence for surviving palaeo-environmental deposits in features such as infilled palaeo-channels. Since the study area was large, it encompassed considerable variation in slope morphology (Fig. 4B), complex topography, and vegetation cover (Fig. 4C). The study area was, therefore, divided into different SfM zones (Fig. 3) that were surveyed through planned and manual UAV flights tougher with ground-based photos in May 2019. Nadir and oblique UAV images were collected with a DJI Zenmuse X4S camera (20 M pixels, focal length 8.8 mm, 1-in. CMOS Sensor) mounted on a professional UAV (DJI Matrice210v2; Fig. 4D) that has high flexibility and stability in most weather conditions and needs only a small space for takeoff and landing. In zones with uniform altitude (a.s.l.), the UAV flight control unit (coupled to a GNSS) was used to plan the UAV flight strips using software that adjusts the height and speed of flight accordingly, and the image overlap (optimal overlap is 80% in flight direction and a flight strip overlap of 60%). The flight altitudes were in the range of 25–45 m to ensure high resolution and a sufficiently large overlap (image footprint with a mean ground sampling distance of 0.006–0.011 m). In areas with important slope change, the manual flight mode was used with a time-lapse function of the camera that allowed the capture an image at 3 s intervals, sufficient to guarantee the overlap in sequential photographs, which is essential for the image matching algorithms used in SfM (Eltner et al., 2016). Ground-based and UAV images (nadir and oblique photos very close to the ground) were taken in vegetated areas (Fig. 4C), over the terrace complex and along the trench excavation (Fig. 4E and F) using the same Zenmuse X4S camera to maximize the resolution of the SfM survey. For the ground-based surveys, the photographs were taken using an adequate average depth distance from the object, based on a mean baseline of 3 m between adjacent camera positions, to avoid large jumps in scale. Before image acquisition, the GCPs (Figs. 3 and 4B) were distributed throughout the study area so that GCPs could be visible in as many images as possible and easily distinguishable from the surrounding landscape (Smith et al., 2015). Indeed, the number, location, and distribution of GCPs are a fundamental aspect and were based on the features of the studied area, extension, and desired resolution (Cucchiaro et al., 2018a). A Leica ATX1230 GG GNSS allowed us to survey n ¼ 137 GCP (Fig. 3) with a planimetric positional accuracy ranging from 0.02 to 0.03 m and vertical uncertainties ranging from 0.03 to 0.04 m in real-time kinematic (RTK) mode. All the points coordinates were referred to the British National Grid (EPSG: 27700) reference system. 3.2.2 SfM processing Processing of SfM datasets is not limited by the SfM method or by the camera platform but by computing power, which with modern computers and GPU processing, for example, is becoming much less of a limitation than with early geoscience usage of SfM (Carrivick and Smith, 2018). Thus, large-scale processing works, like this need powerful computers and SfM photogrammetry software. The image dataset (no. of photos 3782) was processed

3 SfM in geoarchaeology: Agricultural terraces in Europe

193

FIG. 4 Pictures from the Ingram field survey: (A) the geoarchaeology excavation trench cut over the terrace complex, (B) example of GCP used in the SfM survey, (C) the circular-shaped Plantation Camp enclosure now cloaked in vegetation with trees in its center, (D) DJI Matrice210v2 used in the UAV SfM survey, (E) detailed view of the excavation trench with GCPs in place, and (F) detail of excavation trench during the SfM survey. In all, 80 GCPs were placed inside and along the trench.

194

6. SfM photogrammetry for GeoArchaeology

with an 2xIntel Xeon Bronze 3106 CPU @ 1.70 Ghz and 256 GB RAM, 2xNVIDIA GeForce RTX 2080 Ti, through Agisoft Photoscan Pro v 1.4.5 (Manual Agisoft Lens, 2010) dividing the photos in the different SfM surveys (Fig. 3). Agisoft Photoscan (hereafter Photoscan) combines computer vision routines of SfM and MVS algorithms to extract the 3D point clouds from the images, creating 3D models of the scene and, additionally, orthomosaics. The first preliminary step is masking (Fig. 5A) unwanted objects (e.g., water, vegetation, and clouds in ground-based images) in the photos uploaded in the software. Then, five main steps were followed: (i) camera calibration using Agisoft Lens, an automatic lens calibration routine that uses LCD screen as a calibration target and supports estimation of the camera calibration matrix of DJI Zenmuse X4S, including nonlinear distortion coefficients. This precalibration step was useful to estimate camera parameters that were used in the next process, i.e., (ii) alignment where ground-based and UAV photos were directly fused to the alignment process in Photoscan to avoid subsequent data fusion problems at level of point clouds (Cucchiaro et al., 2018a). During the alignment step, common features in the set of images were identified and matched, the internal camera parameters and relative orientation of the camera at the time of image acquisition were estimated, and construction of the image network took place (Carrivick et al., 2016; Piermattei et al., 2016). This first alignment (“low accuracy” in Agisoft Photoscan) allowed the removal of unwanted (e.g., vegetation; Fig. 5B) or outliers data (i.e., points that are clearly located off the surface or have anomalous large image residuals), and deleting the photos that the software does not align for different reasons. (iii) Scaling and georeferencing of the 3D sparse point cloud using a seven-parameter, linear-similarity transformation based on the XYZ coordinates of GCPs (Smith et al., 2015), evaluating the level of GCPs uncertainty before including these data to avoid adversely affecting data accuracy ( James et al., 2017a). The location and manual marking of GCPs (Fig. 5A) on at least two photographs helped to remove deformations such as the “dome effect” ( James and Robson, 2014), and to refine the camera calibration parameters (Fonstad et al., 2013; Eltner et al., 2016). Some of the GCPs (1/3) were used as control points (CPs) in the different Agisoft Photoscan projects to provide an independent measure of accuracy (the difference between the real coordinates in this point and the modeled values; i.e., residuals). With GCPs, the alignment (“high accuracy” in Agisoft Photoscan) was rerun to improve the image alignment in light of this information. (iv) Camera optimization: refined the camera and tie-point locations (homologous points that link different images), and the camera calibration parameters of each image, through the bundle adjustment algorithm (least-squares network optimization; Granshaw, 1980) that improved their values during the camera alignment step by incorporating GCPs and removing obvious outliers and incorrect matches from the sparse point cloud. Moreover, the optimization process was done through appropriate weighting of observations of tie and control point images in bundle adjustment to enhance a real error characterization ( James et al., 2017a). (v) 3D high-density point clouds and orthomosaics: involved the implementation of MVS image matching algorithms that increased the point density by several orders of magnitude (Woodget et al., 2015), operating at the individual pixel scale to build dense clouds (Fig. 5B; Piermattei et al., 2015), and orthomosaics. Then, mesh (Fig. 5C), tiled models (Fig. 5D), and orthomosaics were generated and exported from Photoscan, being the resolution of these in agreement with the point cloud density and the resolution of the photos.

3 SfM in geoarchaeology: Agricultural terraces in Europe

195

FIG. 5 Examples of SfM processing steps and outputs. (A) Photo of Ingram terrace area where the vegetated parts were masked and GCPs were manually located in Agisoft Photoscan. (B) Point cloud of vegetated area (Fig. 4C). (C) Examples of the point cloud in Ingram area (terrace complex on the left and Brough Law Hillfort at top right). (D) The mesh at 0.25 m resolution viewing the site from the north-east looking up toward Plantation Camp terraces and Brough Law from the across the valley floor. (E) Tiled model of the whole Ingram SfM survey. (F) Example of CSF filter application to extract the ground points in much vegetated zone (Fig. 4C).

196

6. SfM photogrammetry for GeoArchaeology

3.2.3 SfM postprocessing The dense SfM point cloud had to be postprocessed to minimize potential sources of error and noise in the topographic data because SfM technology presented frequent problems linked to photogrammetric workflow that could lead to numerous outliers and corrupt subsequent analysis (Smith et al., 2015; Carrivick et al., 2016) if the SfM process was not correctly applied. The first dense cloud editing was performed by means of the CloudCompare software (Omnia Version 2.10.2; http://www.danielgm.net) through a manual filtering, the cloth simulation filter (CSF; Zhang et al., 2016) and the “SOR filter tool.” The manual filter was used to delete unwanted objects in the point cloud (e.g., isolated trees and shrubs; Fig. 5E), while the CSF filter (Fig. 5F) extracted the ground points in much vegetated and complex areas (Fig. 4C). Then, the SOR filter was used to remove outliers through the computation of the average distance of each point to its neighbors (it rejects the points that are farther than the average distance plus a defined number of times the standard deviation). After the checking of possible alignment problems (displacements or differences in altitude between adjacent SfM surveys link to GNSS survey errors; Cucchiaro et al., 2019), the point cloud of different SfM surveys (Fig. 3) was merged together in CloudCompare software generating a huge point cloud (1,091,540,500 points with a mean density of 2700 points/m2) for the whole Ingram area. 3.2.4 DTM generation The point cloud was decimated in order to reduce the processing constraints and the extremely high density of the 3D cloud. The geostatistical Topography Point Cloud Analysis Toolkit (ToPCAT) implemented in the Geomorphic Change Detection software for ArcGIS, (Wheaton et al., 2010; available in http://gcd6help.joewheaton.org/) was used to decimate the point cloud. This tool (used in several studies: e.g., Javernick et al., 2014; Vericat et al., 2014; Marteau et al., 2017) allows an intelligent decimation by decomposing the point cloud into a set of nonoverlapping grid cells and calculates statistics for the observations in each grid (e.g., minimum, mean, maximum elevation). Following the work by Brasington et al. (2012), the minimum elevation within each grid cell was considered the ground elevation and a grid cell of 0.10 m was selected to regularize the dataset. The point cloud obtained by ToPCAT (37,180,100 points with a mean density of 100 points/m2) was used to calculate a triangular irregular network (TIN) that was converted to rasters obtaining two DTMs.

3.3 Result and discussion The SfM workflow allowed the generation of a DTM at 0.25 m (Fig. 6A) for the whole Ingram area, whereas a higher-resolution DTM (0.10 m; Fig. 6B) was carried out for the terrace complex so as to achieve a very detailed reconstruction of the topographic features of archeological and geomorphological interest applicable to the TerrACE project. Compared to the DTM at 2 m resolution (Fig. 2B), the DTM at 0.25 m of Ingram Valley provided a significantly enhanced level of detail including much greater clarity of the prehistoric terrace system, the Plantation Camp enclosures, Brough Law Hillfort and the medieval ridge and furrow, and the overlying postmedieval stone-walled boundaries (Fig. 6A). Prior to this high-resolution SfM survey, the prehistoric terraces were virtually invisible on existing

FIG. 6

(A) Shaded relief map of the SfM DTM at 25 cm on the DTM at 2 m resolution (Fig. 2B) for the Ingram Valley. The Brough Law Hillfort is to the left, the prehistoric agricultural terraces are central and to the immediate right of the Plantation Camp enclosures, and the medieval ridge and furrow remains are to the right and are clearly visible. The postmedieval straight stone-walled boundaries overly both the prehistoric agricultural terraces as well as the medieval ridge and furrow. (B) Shaded relief map of the SfM DTM at 10 cm where it is possible to identify the seven prehistoric agricultural terraces, trackway above them and the medieval ridge and furrow despite the bracken infestation which cloaks the prehistoric agricultural terrace complex.

198

6. SfM photogrammetry for GeoArchaeology

remote-sensing data and hence why they were initially recognized from ground-level survey and not from aerial photographs. Moreover, the higher-resolution DTM (0.10 m) shows the terraces (Fig. 6B), Brough Law Hillfort, Plantation Camp, and the ridge and furrow feature very clearly despite the severe problem of bracken infestation that severely obscures these and many more archeological sites across the Cheviot Hills and Northumberland National Park. It also provided an accurate and high level of detail of the archeological features and soil and sediment stratigraphy along the excavation trench (Fig. 6B). This high-resolution modeling has helped significantly in creating an accurate record of what is an awkward archeological trench to record due to the range of elevation along its length and the complexity and subtle color changes in the sediment stratigraphy observable in section. Furthermore, the output of the SfM workflow as point clouds allowed for the extraction of profiles, sections, scaled plans (Fig. 7A), and orthomosaics (Fig. 7B) of the terrace complex and the excavation trench (Fig. 7C). These tools, adding a clear visual dimension to the drawn section, can make the archeological work and measurements easier, faster, more accurate, as well as also allowing for more accurate and repeat interpretation. Indeed, these data can be useful to extract metric of archeological and geomorphological features that are to be included in the Ingram archeological report (Archaeological Research Services, n.d.). This HRT study has provided a level of detail that had not been hitherto been achievable on this nationally important site and has overcome many of the problems encountered when attempting to survey complex archeological palimpsests obscured by dense vegetation and situated on steep, nonuniform slopes. An additional terrace was identified that had not been recognized before due to the HRT study bringing out a level of detail that had not been previously observable. This has stretched the surviving extent of the terrace complex, as well as showing a direct relationship with the ridge and furrow cultivation remains that can be seen to directly overly it. The trackway leading to Plantation Camp had been questioned by some archeologists, but now the clarity of the HRT study shows it very clearly and leading directly to Plantation Camp and the top of the terrace complex (Fig. 7C). The methodology described in this study has shown it to be a rapid, cost-effective, and highly accurate technique for surveying archeological sites at both a landscape and localized scale and adding new and more accurate information in nationally important UK landscapes and beyond. The assessment of the GNSS and SfM surveys errors for the Ingram study area (Table 2) shows that the quality of SfM surveys was adequate for investigating topographic features of the terrace area and recording and analyzing the excavation trench structure. The SfM survey results highlighted the benefits of the acquisition of data from two different observation directions and platforms (UAV and ground based). This helped to: (i) avoid gaps in data; (ii) increase the individual point precision, point clouds density (St€ ocker et al., 2015; Cucchiaro et al., 2018a), the robustness of topographic mapping, and the high-resolution detail; and (iii) reduce error in estimated camera parameters, thus minimizing systematic DTM deformation errors or large-area distortions ( James et al., 2017a). Indeed, the ground-based photos provided a more accurate representation of complex surfaces for detail scale, 3D reconstruction, especially when steep or subvertical surfaces, such as the vertical walls of terraces, are surveyed (Cucchiaro et al., 2018a). This integrated approach preserved fine-grained topographic detail, permitted accurate survey of highly vegetated areas (Fonstad et al., 2013), while also allowing for the capture of large spatial datasets. The remarkable results of the SfM surveys at Ingram were also achieved through the careful distribution of

FIG. 7 Useful SfM outputs for archeological work. (A) Point clouds, scaled plans, profiles, and sections of the geoarchaeology excavation trench. (B) Detailed orthomosaic (5 cm) of the study area made through SfM technique. (C) DTM at 0.1 m resolution looking down vertically over the prehistoric agricultural terraces (n. 1), the Plantation Camp enclosures to the left (n. 2), the trackway (n. 3), the medieval ridge and furrow to the right (n. 4), and postmedieval boundaries (n. 5). TABLE 2 Characteristics of the GPS and SfM surveys for the Ingram study area and in particular for the trench zone Number of images processed

Number of GCPs (as control, [as check])a

GNSS positional accuracy of GCPs (Easting-Northing— Height; m)

GCPs image precision (pixel—m)b

Tie point image precision (pixel—m)b

CPs image precision (m)b

All Ingram area

3782

137 [40]

0.03–0.04

1.014–0.075

0.903–0.172

0.078

Trench area

570

80 [27]

0.03–0.04

2.130–0.046

0.873–0.152

0.048

SfM survey

a

1/3 of the GCPs were used as CP. Measures provided by Photoscan software. GCPs image precision reflects the precision in image space that GCP observations were made to, whereas tie-points precision is the equivalent measure for the tie points.

b

200

6. SfM photogrammetry for GeoArchaeology

GCPs across the study area. This influenced the final quality of the process of georeferencing, mitigated systematic errors (Vericat et al., 2009; James and Robson, 2012; Koci et al., 2017; James et al., 2017b), and helped the merging between the different SfM surveys that had common GCPs. Indeed, the GCPs network was fundamental in this SfM survey because it allowed us to register and merge together very detailed and high-resolution surveys that otherwise would not be possible to manage due to the huge number of images acquired for a large study area such as that at Ingram. The alignment process was fundamental to increase the quality of the whole point cloud (Cucchiaro et al., 2019). The limited ability to process very heavy SfM data (in terms of gigabytes) for wide study areas is perhaps the potential weakness in this approach. However, a robust SfM workflow and technological developments can certainly help to increase the performance of this technique. The present work highlights how the precision in SfM surveys could only be guaranteed through a careful planning of appropriate survey, accurate data post-processing, and an uncertainty assessment, identifying and minimizing the potential sources of error in SfM topographic data.

4 Final remarks The SfM photogrammetry technique has provided a number of advances for geoarchaeological studies, but it can produce datasets containing large errors, if not correctly applied, especially in wide and complex topographic zones, and in terrains dominated by vegetation. As shown by the case study discussed in this chapter, SfM technique carried out low cost (and time) HRT for large areas, showing the different dimensions, orientations, and distribution of cultivation-related and settlement features. This technique allowed rapid, accurate survey of complex archeological features at a landscape scale that are otherwise almost unsurveyable due to dense vegetation cover—in this case bracken infestation, thereby revealing new archeological remains, as well as confirming physical relationships, and thus chronostratigraphic relationships within and between component monuments. Moreover, SfM can be effective in the estimation of metrics and geomorphological features of cultivation terraces such as riser height and slopes from high-resolution DTMs. SfM produced archeological recording of excavation trenches by integrating ground-based and UAV survey, which can add a 3D element to traditional section mosaics and allows integrated archiving of surface and subsurface data. Indeed, this photogrammetric technique extracted 3D models, profiles, sections, scaled plans, and orthomosaic of trench excavations, simplifying and speeding the archeologist’s field and postexcavation work.

Acknowledgments The research is funded by Advanced ERC Grant TerrACE: Terrace Archaeology and Culture in Europe (ERC-2017ADG: 787790, 2018–2023, https://www.terrace.no/). The authors acknowledge help and support from the Northumberland National Park and landowners. A special thanks to all those involved with the project, particularly all of the volunteers who put in a tremendous amount of effort during the excavations. We would also like to thank Lee McFarlane, the Historic England Inspector of Ancient Monuments, Chris Jones, Historic Environment Officer at Northumberland National Park, who supported and advised on the archaeological works, and Ross Wilson from Ingram Farm.

References

201

References AgiSoft LLC, 2010. Agisoft Lens User-Manual. Version 1.2.0. http://www.agisoft.com/downloads/user-manuals. Accessed on 22 June 2019. Aqdus, S.A., Drummond, J., Hanson, W.S., 2008. Discovering archaeological cropmarks: a hyperspectral approach. Int. Arch. Photogramm. Remote. Sens. Spat. Inf. Sci. 37, 361–366. Archaeological Research Services, n.d. in preparation. Survey and Excavation at Plantation Camp Agricultural Terraces, Ingram, Northumberland. Archaeological Research Services Report No. Barber, M., 2011. A History of Aerial Photography and Archaeology: Mata Hari’s Glass Eye and Other Stories. Historic England, London. 304 p. Beach, T., Beach, S.L., Krause, S., Guderjan, T., Valdez Jr., F., Fernandez-Diaz, J.C., Eshleman, S., Doyle, C., 2019. Ancient Maya wetland fields revealed under tropical forest canopy from laser scanning and multiproxy evidence. Proc. Natl. Acad. Sci. U. S. A. 116, 21469–21477. Bemis, S.P., Micklethwaite, S., Turner, D., James, M.R., Akciz, S., Thiele, S.T., Bangash, H.A., 2014. Ground-based and UAV-based photogrammetry: a multi-scale, high-resolution mapping tool for structural geology and paleoseismology. J. Struct. Geol. 69, 163–178. Blasone, G., Cavalli, M., Marchi, L., Cazorzi, F., 2014. Monitoring sediment source areas in a debris-flow catchment using terrestrial laser scanning. Catena 123, 23–36. Bojakowski, P., Bojakowski, K.C., Naughton, P., 2015. A comparison between structure from motion and direct survey methodologies on the Warwick. J. Marit. Archaeol. 10 (2), 159–180. Brasington, J., Vericat, D., Rychkov, I., 2012. Modeling riverbed morphology, roughness, and surface sedimentology using high-resolution terrestrial laser scanning. Water Resour. Res. 48, 1–18. Brown, A.G., 2008. Geoarchaeology, the four dimensional (4D) fluvial matrix and climatic causality. Geomorphology 101, 278–297. Brown, A.G., Schneider, H., Rice, R.J., Milton, E.J., 1990. Remote sensing soil erosion: airborne thematic mapper data on soil variation in Mediterranean arable land in Southern Spain. In: Procs. of the NERC Symposium on Airborne Remote Sensing 1990. British Geological Survey, Keyworth, Nottingham, pp. 7–18. Brutto, M.L., Sciortinoa, R., Garraffaa, A., 2017. RPAS and TLS techniques for archaeological survey: the case study of the archaeological site of Eraclea Minoa (Italy). Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 52, 433–438. Cao, W., Sofia, G., Tarolli, P., 2020. Geomorphometric characterisation of natural and anthropogenic land covers. Prog. Earth Planet. Sci. 7, 2. Carey, C.J., Brown, A.G., Challis, K.C., Howard, A., Cooper, L., 2006. Predictive modelling of multi-period geoarchaeological resources at a river confluence. J. Archaeol. Prospect. 13, 241–250. Carey, C., Howard, A.J., Jackson, R., Brown, A.G., 2017. Utilizing multi-period geoarchaeological predictive models as a framework for archaeological investigation in river valleys: an integrated case study from the Lugg, Valley, Herefordshire, UK. J. Archaeol. Sci. Rep. 11, 658–673. Carrivick, J.L., Smith, M.W., 2018. Fluvial and aquatic applications of structure from motion photogrammetry and unmanned aerial vehicle/drone technology. Wiley Interdiscip. Rev. Water. 6 (1). ISSN 2049-1948. Carrivick, J.L., Smith, M.W., Quincey, D.J., 2016. Structure From Motion in the Geosciences. New Analytical Methods in Earth and Environmental Science. Wiley-Blackwell. Cavalli, R., Colosi, F., Palombo, A., Pignatti, S., Poscolieri, M., 2007. Remote hyperspectral imagery as a support to archaeological prospection. J. Cult. Herit. 8, 272–283. Chen, J., Li, K., Chang, K.J., Sofia, G., Tarolli, P., 2015. Open-pit mining geomorphic feature characterisation. Int. J. Appl. Earth Obs. Geoinf. 42, 76–86. Clapuyt, F., Vanacker, V., Van Oost, K., 2016. Reproducibility of UAV-based earth topography reconstructions based on structure-from-motion algorithms. Geomorphology 260, 4–15. Clapuyt, F., Vanacker, V., Schlunegger, F., Van Oost, K., 2017. Unravelling earth flow dynamics with 3-D time series derived from UAV-SfM models. Earth Surf. Dyn. 5, 791–806. Cucchiaro, S., Cavalli, M., Vericat, D., Crema, S., Llena, M., Beinat, A., Marchi, L., Cazorzi, F., 2018a. Monitoring topographic changes through 4D-structure-from-motion photogrammetry: application to a debris-flow channel. Environ. Earth Sci. 2018 (77), 632. Cucchiaro, S., Maset, E., Fusiello, A., Cazorzi, F., 2018b. 4D-SfM photogrammetry for monitoring sediment dynamics in a debris-flow catchment: software testing and results comparison. Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. XLII-2, 281–288.

202

6. SfM photogrammetry for GeoArchaeology

Cucchiaro, S., Cavalli, M., Vericat, D., Crema, S., Llena, M., Beinat, A., Marchi, L., Cazorzi, F., 2019. Geomorphic effectiveness of check dams in a debris-flow catchment using multi-temporal topographic surveys. Catena 174, 73–83. Dietrich, J.T., 2017. Bathymetric structure-from-motion: extracting shallow stream bathymetry from multi-view stereo photogrammetry. Earth Surf. Process. Landf. 42, 355–364. Doyon, W., Adams, M.D., Simmons, B., 2019. Photogrammetry is the New Archaeological Photography: 3D Modelling at Abydos March 14, 2019. Abydos Archaeology. Eker, R., Aydın, A., H€ ubl, J., 2018. Unmanned aerial vehicle (UAV)-based monitoring of a landslide: Gallenzerkogel landslide (Ybbs-Lower Austria) case study. Environ. Monit. Assess.. 190. Eltner, A., Kaiser, A., Castillo, C., Rock, G., Neugirg, F., Abella´n, A., 2016. Image-based surface reconstruction in geomorphometry-merits, limits and developments. Earth Surf. Dyn. 4, 359–389. Evans, D.H., et al., 2013. Uncovering archaeological landscapes at Angkor using lidar. Proc. Natl. Acad. Sci. U. S. A. 110, 12595–12600. Fonstad, M.A., Dietrich, J.T., Courville, B.C., Jensen, J.L., Carbonneau, P.E., 2013. Topographic structure frommotion: a new development in photogrammetric measurement. Earth Surf. Process. Landf. 38, 421–430. Fowler, M.J.F., 1995. Detection of archaeological features on multispectral satellite imagery. AARGnews 10, 7–14. Frodsham, P., Waddington, C., 2004. The Breamish Valley archaeology project 1994–2002. In: Frodsham, P. (Ed.), Archaeology in Northumberland National Park. Council for British Archaeology, pp. 171–189. Research Report 136, 2004. Fussell, A., 1982. Terrestrial photogrammetry in archaeology. World Archaeol. 14, 157–172. Georges-Leroy, M., Dambrine, E., Dupouey, J.-L., Etienne, D., 2013. Lidar helps to decipher land-use history in Lorrain, France. In: Understanding Landscapes, From land Discovery to their Spatial Organisation, Conference: Proceedings of the XVI World Congress of the International Union of Prehistoric and Protohistoric Sciences at Florianopolis, Brazil, pp. 115–122. Glendell, M., McShane, G., Farrow, L., James, M.R., Quinton, J., Anderson, K., Evans, M., Benaud, P., Rawlins, B., Morgan, D., Jones, L., Kirkham, M., DeBell, L., Quine, T.A., Lark, M., Rickson, J., Brazier, R.E., 2017. Testing the utility of structure-from-motion photogrammetry reconstructions using small unmanned aerial vehicles and ground photography to estimate the extent of upland soil erosion. Earth Surf. Process. Landf. 42, 1860–1871. Gonza´lez-Aguilera, D., Mun˜oz-Nieto, A., Go´mez-Lahoz, J., Herrero-Pascual, J., Gutierrez-Alonso, G., 2009. 3D digital surveying and modelling of cave geometry: application to paleolithic rock art. Sensors 9, 1108–1127. Granshaw, S.I., 1980. Bundle adjustment methods in engineering photogrammetry. Photogramm. Rec. 10 (56), 181–207. Guyot, A., Lennon, M., Thomas, N., Gueguen, S., Petit, T., Lorho, T., Cassen, S., Hubert-Moy, L., 2019. Airborne hyperspectral imaging for submerged archaeological mapping in shallow water environments. Remote Sens. (Basel) 11, 2237–2257. H€ ammerle, M., H€ ofle, B., 2018. Introduction to LiDAR in geoarchaeology from a technological perspective. In: Siart, C., Forbriger, M., Bubenzer, O. (Eds.), Digital Geoarchaeology. Natural Science in Archaeology. Springer, Cham. Howland, M.D., Kuester, F., Levy, T.E., 2014. Structure from motion: twenty-first century field recording with 3D technology. Near Eastern Archaeol. 77 (3), 187–191. Hoyle, J., 2008. The Forest of Dean, Gloucestershire Lidar Survey of Selected Areas of Woodland and the Aggregates Resource Area. Archaeology Service, Gloucestershire County Council. Iglhaut, J., Cabo, C., Puliti, S., Piermattei, L., O’Connor, J., Rosette, J., 2019. Structure from motion photogrammetry in forestry: a review. Curr. For. Rep. 5, 155–168. Immerzeel, W.W., Kraaijenbrink, P.D.A., Shea, J.M., Shrestha, A.B., Pellicciotti, F., Bierkens, M.F.P., de Jong, S.M., 2014. High-resolution monitoring of Himalayan glacier dynamics using unmanned aerial vehicles. Remote Sens. Environ. 150, 93–103. James, M.R., Robson, S., 2012. Straightforward reconstruction of 3D surfaces and topography with a camera: accuracy and geoscience application. J. Geophys. Res. 117, F03017. James, M.R., Robson, S., 2014. Mitigating systematic error in topographic models derived from UAV and groundbased image networks. Earth Surf. Process. Landf. 39, 1413–1420. James, M.R., Robson, S., D’Oleire-Oltmanns, S., Niethammer, U., 2017a. Optimising UAV topographic surveys processed with structurefrom-motion: ground control quality, quantity and bundle adjustment. Geomorphology 280, 51–66.

References

203

James, M.R., Robson, S., Smith, M.W., 2017b. 3-D uncertainty-based topographic change detection with structurefrom-motion photogrammetry: precision maps for ground control and directly georeferenced surveys. Earth Surf. Process. Landf. 42, 1769–1788. Javernick, L., Brasington, J., Caruso, B., 2014. Modeling the topography of shallow braided rivers using structurefrom-motion photogrammetry. Geomorphology 213, 166–182. Johnson, N., Bonney, D., Rose, P., 2008. Bodmin Moor An Archaeological Survey Volume 1: The Human Landscape to c 1800. English Heritage, London. Kinnaird, T., Bolo`s, J., Turner, A., Turner, S., 2017. Optically-stimulated luminescence profiling and dating of historic agricultural terraces in Catalonia (Spain). J. Archaeol. Sci. 78, 66–77. Koci, J., Jarihani, B., Leon, J.X., Sidle, R., Wilkinson, S., Bartley, R., 2017. Assessment of UAV and ground-based structure from motion with multi-view stereo photogrammetry in a gullied savanna catchment. ISPRS Int. J. Geo-Inf. 6, 328. Kouchoukos, N., 2001. Satellite images and near eastern landscapes. Near Eastern Archaeol. 64 (1/2), 80–91. Landeschi, G., Nilsson, B., Dell’Unto, N., 2016. Assessing the damage of an archaeological site: new contributions from the combination of image-based 3D modelling techniques and GIS. J. Archaeol. Sci. Rep. 10, 431–440. Lasaponara, R., Masini, N., 2007. Detection of archaeological crop marks by using satellite QuickBird multispectral imagery. J. Archaeol. Sci. 34, 214–221. Lasaponara, R., Masini, N., 2011. Satellite remote sensing in archaeology: past, present and future perspectives. J. Archaeol. Sci. 38, 1995–2002. Lim, M., Rosser, N.J., Allison, R.J., Petley, D.N., 2010. Erosional processes in the hard rock coastal cliffs at Staithes, North Yorkshire. Geomorphology 114, 12–21. Linderholm, J., Geladi, P., Goretta, N., Bendoula, R., Gobrecht, A., 2019. Near infrared and hyperspectral studies of archaeological stratigraphy and statistical considerations. Geoarchaeology 34, 311–321. Lobb, M., Brown, A.G., 2016. Terrestrial laser scanning and coastal erosion at Low Hauxley. In: Waddington, C., Bonsall, C. (Eds.), Archaeology and Environment on the North Sea Littoral: A Case Study from Low Hauxley. Oxbow Books, Oxford, pp. 291–301. Lobb, M., Brown, A.G., Leyland, J., Bernard, V., Daire, M.Y., Langoue¨t, L., An estuarine tide-scape of production: terrestrial laser scanning (TLS) of fixed fishing structures and a tidal mill in the Leguer Estuary, in press, Brittany, France. Antiquity Lo´pez, J.A.B., Jimenez, G.A., Romero, M.S., Garcı´a, E.A., Martı´n, S.F., Medina, A.L., Guerrero, J.A.E., 2016. 3D modelling in archaeology: the application of structure from motion methods to the study of the megalithic necropolis of Panoria (Granada, Spain). J. Archaeol. Sci. Rep. 10, 495–506. Lotherington, R., Waddington, C., 2019. Plantation Camp Agricultural Terraces, Northumberland Written Scheme of Investigation for an Archaeological Evaluation. Archaeological Research Services Ltd. 19 p. Malambo, L., Popescu, S.C., Murray, S.C., Putman, E., Pugh, N.A., Horne, D.W., Richardson, G., Sheridan, R., Rooney, W.L., Avant, R., Vidrine, M., McCutchen, B., Baltensperger, D., Bishop, M., 2018. Multitemporal fieldbased plant height estimation using 3D point clouds generated from small unmanned aerial systems highresolution imagery. Int. J. Appl. Earth Obs. Geoinf. 64, 31–42. Mallalieu, J., Carrivick, J.L., Quincey, D.J., Smith, M.W., James, W.H.M., 2017. An integrated structure-from-motion and time-lapse technique for quantifying ice-margin dynamics. J. Glaciol. 63, 937–949. Marteau, B., Vericat, D., Gibbins, C., Batalla, R.J., Green, D.R., 2017. Application of structure-from-motion photogrammetry to river restoration. Earth Surf. Process. Landf. 42, 503–515. Mertes, J., Thomsen, T., Gulley, J., 2014. Evaluation of structure from motion software to create 3d models of late nineteenth century great lakes shipwrecks using archived diver-acquired video surveys. J. Marit. Archaeol. 9 (2), 173–189. MHEF (Minerals and Historic Environment Forum), 2008. Mineral Extraction and Archaeology: A Practice Guide. English Heritage on behalf of the Minerals and Historic Environment Forum, London. New Forest, 2016. Laser Mapping Uncovers Hidden Secrets in the New Forest. http://www.hlsnewforest.org.uk/ 2016/01/25/laser-mapping-uncovers-hidden-secrets-of-the-new-forest/. Accessed on October 2019. Ninfo, A., Fontana, A., Mozzi, P., Ferrarese, F., 2009. The map of Altinum, ancestor of Venice. Science 325, 577. O’Connor, J., Smith, M.J., James, M.R., 2017. Cameras and settings for aerial surveys in the geosciences: optimising image data. Prog. Phys. Geogr. 41, 325–344.

204

6. SfM photogrammetry for GeoArchaeology

Parcak, S., 2007. Satellite remote sensing methods for monitoring archaeological tells in the Middle East. J. Field Archaeol. 32 (1), 65–81. Passmore, D.G., Waddington, C., 2009. Managing archaeological landscapes in Northumberland. In: Till-Tweed Studies. vol. 1. Oxbow Books and English Heritage, Oxford. Passmore, D.G., Waddington, C., 2012. Archaeology and environment in Northumberland. In: Till-Tweed Studies. vol. 2. Oxbow Books and English Heritage, Oxford. Penny, D., Hall, T., Evans, D., Polkinghorne, M., 2019. Geoarchaeological evidence from Angkor, Cambodia, reveals a gradual decline rather than a catastrophic 15th-century collapse. Proc. Natl. Acad. Sci. U. S. A. 116 (11), 4871–4876. Pierdicca, R., Frontoni, E., Malinverni, E.S., Colosi, F., Orazi, R., 2016. Virtual reconstruction of archaeological heritage using a combination of photogrammetric techniques: Huaca Arco Iris, Chan Chan, Peru. Digit. Appl. Archaeol. Cult. Herit. 3, 80–90. Piermattei, L., Carturan, L., Guarnieri, A., 2015. Use of terrestrial photogrammetry based on structure-from-motion for mass balance estimation of a small glacier in the Italian Alps. Earth Surf. Process. Landf. 40, 1791–1802. Piermattei, L., Karel, W., Vettore, A., Pfeifer, N., 2016. Panorama image sets for terrestrial photogrammetric surveys. ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci. III-5, 159–166. Pijl, A., Tosoni, M., Roder, G., Sofia, G., Tarolli, P., 2019. Design of terrace drainage networks using UAV-based highresolution topographic data. Water 11, 814. Pijl, A., Bailly, J.-B., Feurer, D., El Maaoui, M.A., Boussema, M.R., Tarolli, P., 2020. TERRA: terrain extraction from elevation Rasters through repetitive anisotropic filtering. Int. J. Appl. Earth Obs. Geoinf.. 84. Prins, A.B., Adams, M.J., Homsher, R.S., Ashley, M., 2014. Digital archaeological fieldwork and the Jezreel Valley regional project, Israel. Near Eastern Archaeol. 77 (3), 192–197. Romano, D.G., Tolba, O., 1996. Remote sensing and GIS in the study of Roman centuriation in the Corinthia, Greece. In: Kamermans, H., Fennema, K. (Eds.), Interfacing the Past: Computer Applications and Quantitative Methods in Archaeology CAA95. In: Analecta Praehistorica Leidensia 28, University of Leiden, Leiden, pp. 457–463. 2 vols. Rosser, N.J., Petley, D.N., Lim, M., Dunning, S.A., Allison, R.J., 2005. Terrestrial laser scanning for monitoring the process of hard rock coastal cliff erosion. Q. J. Eng. Geol. Hydrogeol. 38 (4), 363–375. Skrede, M.A., 2005. ‘Utmark’ the outfields and industry and idealogy in the Iron age. In: Holm, et al., (Ed.), In: Archaeological Series International, vol. 1. University of Bergen, pp. 31–41. Smith, M., Vericat, D., 2015. From experimental plots to experimental landscapes: topography, erosion and deposition in sub-humid badlands from structure-from-motion photogrammetry. Earth Surf. Process. Landf. 40, 1656–1671. Smith, M.W., Carrivick, J.L., Quincey, D.J., 2015. Structure from motion photogrammetry in physical geography. Prog. Phys. Geog. 40, 247–275. Sofia, G., Marinello, F., Tarolli, P., 2014. A new landscape metric for the identification of terraced sites: the slope local length of auto-correlation (SLLAC). ISPRS J. Photogramm. Remote Sens. 96, 123–133. South Downs National Park, 2019. Secrets of High Woods. https://www.southdowns.gov.uk/discover/heritage/ secrets-of-the-high-woods/. Accessed on October 2019. St€ ocker, C., Eltner, A., Karrasch, P., 2015. Measuring gullies by synergetic application of UAV and close-range photogrammetry—a case study from Andalusia, Spain. Catena 132, 1–11. Stumpf, A., Malet, J.P., Allemand, P., Pierrot-Deseilligny, M., Skupinski, G., 2015. Ground-based multi-view photogrammetry for the monitoring of landslide deformation and erosion. Geomorphology 231, 130–145. Tarolli, P., 2014. High-resolution topography for understanding earth surface processes: opportunities and challenges. Geomorphology 216, 295–312. Tarolli, P., Preti, F., Romano, N., 2014. Terraced landscapes: from an old best practice to a potential hazard for soil degradation due to land abandonment. Anthropocene 6, 10–25. Tarolli, P., Cao, W., Sofia, G., Evans, D., Ellis, E.C., 2019. From features to fingerprints: a general diagnostic framework for anthropogenic geomorphology. Prog. Phys. Geogr. Earth Environ. 43, 95–128. Turner, D., Lucieer, A., de Jong, S.M., 2015. Time series analysis of landslide dynamics using an Unmanned Aerial Vehicle (UAV). Remote Sens. 7, 1736–1757. USGS (United States Geological Survey), 2011. Lidar Vegetation Mapping in National Parks: Gulf Coast Network. USGS Fact Sheet. https://pubs.usgs.gov/fs/2011/3137/pdf/fs2011-3137.pdf. Accessed on October 2019. Van Oost, K., et al., 2000. Evaluating the effects of changes in landscape structure on soil erosion by water and tillage. Landsc. Ecol. 15, 579–591. Verhoeven, G.J., 2012. Near-infrared aerial crop mark archaeology: from its historical use to current digital implementations. J. Archaeol. Method Theory 19, 132–160.

References

205

Verhoeven, G.J., Smet, P.F., Poelman, D., Vermeulen, F., 2009. Spectral characterization of a digital still camera’s NIR modification to enhance archaeological observation. IEEE Trans. Geosci. Remote Sens. 47 (10), 3456–3468. Vericat, D., Brasington, J., Wheaton, J., Cowie, M., 2009. Accuracy assessment of aerial photographs acquired using lighter-than-air blimps: low-cost tools for mapping river corridors. River Res. Appl. 28, 985–1000. Vericat, D., Smith, M., Brasington, J., 2014. Patterns of topographic change in sub-humid badlands determined by high-resolution multi-temporal topographic surveys. Catena 120, 164–176. Victoriano, A., Brasington, J., Guinau, M., Furdada, G., Cabre, M., Moysset, M., 2018. Geomorphic impact and assessment of flexible barriers using multi-temporal LiDAR data: the Portaine mountain catchment (Pyrenees). Eng. Geol. 237, 168–180. Walsh, K., Mocci, F., 2003. Fame and marginality: the archealogy of Montagne Sainte Victorie (Provence, France). Am. J. Arch. 107, 45–69. Walsh, K., Brown, A.G., Gourley, B., Scaife, R., 2017. Archaeology, hydrogeology and geomythology in the Stymphalos Valley. J. Archaeol. Sci. Rep. 15, 446–458. Westoby, M.J., Brasington, J., Glasser, N.F., 2012. ’Structure-from-Motion’ photogrammetry: a low-cost, effective tool for geoscience applications. Geomorphology 179, 300–314. Wheaton, J.M., Brasington, J., Darby, S.E., Sear, D.A., 2010. Accounting for uncertainty in DEMs from repeat topographic surveys: improved sediment budgets. Earth Surf. Process. Landf. 35, 136–156. Whitworth, M., Giles, D., Anderson, I., Clewitt, M., 2006. Terrestrial laser scanning for applied geoscience studies in the urban environment. In: IAEG2006. Paper number 252: 1–9. Williams, J.G., Rosser, N.J., Hardy, R.J., Brain, M.J., Afana, A.A., 2018. Optimising 4-D surface change detection: an approach for capturing rockfall magnitude-frequency. Earth Surf. Dyn. 6, 101–119. Woodget, A.S., Carbonneau, P.E., Visser, F., Maddock, I.P., 2015. Quantifying submerged fluvial topography using hyperspatial resolution UAS imagery and structure from motion photogrammetry. Earth Surf. Process. Landf. 40, 47–64. Xiang, J., Chen, J., Sofia, G., Tian, Y., Tarolli, P., 2018. Open-pit mine geomorphic changes analysis using multitemporal UAV survey. Environ. Earth Sci. 77, 220. Zarco-Tejada, P.J., Diaz-Varela, R., Angileri, V., Loudjani, P., 2014. Tree height quantification using very highresolution imagery acquired from an unmanned aerial vehicle (UAV) and automatic 3D photo-reconstruction methods. Eur. J. Agron. 55, 89–99. Zhang, W., Qi, J., Wan, P., Wang, H., Xie, D., Wang, X., Yan, G., 2016. An easy-to-use airborne LiDAR data filtering method based on cloth simulation. Remote Sens. (Basel) 8 (6), 501.

This page intentionally left blank

C H A P T E R

7 Landslide analysis using laser scanners Michel Jaboyedoff, Marc-Henri Derron ISTE—Institute of Earth Sciences, Risk-Group, GEOPOLIS-3793, University of Lausanne, Lausanne, Switzerland

O U T L I N E 1 Introduction

207

2 A short history

210

3 Basics of laser scanners 3.1 Lasers and safety 3.2 LiDAR devices 3.3 TOF LiDAR 3.4 Beam characteristics

211 211 212 213 214

4 LiDAR uses 4.1 Issues 4.2 Different LiDAR configurations 4.3 Filtering 4.4 Georeferencing and coregistration

214 214 215 215 216

5 Characterization of landslides 5.1 Mapping

216 216

5.2 Rock structure characterization and rockfall sources 5.3 Volume estimation

217 218

6 Monitoring 6.1 Surface changes 6.2 Potential methods for real-time monitoring

219 219

7 Modeling based on LDTM

222

8 Discussion and perspectives

223

Acknowledgments

225

References

226

Further reading

230

220

1 Introduction The advent of light detection and ranging (LiDAR) (see Table 1 for acronyms) has revolutionized the study of landslides and geomorphology because it provides an extraordinarily fine resolution topography (Carter et al., 2001). Electronic components and computers at affordable prices have made this technique widely available since the beginning of the

Developments in Earth Surface Processes, Volume 23 https://doi.org/10.1016/B978-0-444-64177-9.00007-2

207

© 2020 Elsevier B.V. All rights reserved.

208

7. Landslide analysis using laser scanners

TABLE 1

Main acronyms used in the text.

Acronym

Meaning

ALS

Airborne laser scanner

DEM

Digital elevation model

DoD

DEM of difference

DTM

Digital terrain model

EDM

Electronic distance meter

EWS

Early warning system

FOV

Field of view

GNSS

Global Navigation Satellite System

GIS

Geographic information system

GPS

Global positioning system

ICP

Iterative closest point method

InSAR

Interferometric synthetic aperture radar

IMU

Inertial measurement unit

LASER

Light amplification by stimulated emission of radiation

LiDAR

Light detection and ranging

LDTM

LiDAR DTM

LOS

Line of sight

PC

Point cloud

SfM

Structure from motion

TIN

Triangular irregular network

TLS

Terrestrial laser scanner

TOF

Time of flight

UAV

Unmanned aerial vehicle

21st century. The number of geoscientific publications is increasing at a nearly exponential rate since around 1990 demonstrating its impact on geosciences (Abellan et al., 2016). There are several types of LiDAR and several configurations on the market such as terrestrial LiDAR or terrestrial LASER scanning (TLS), handy portable and mobile LiDAR, LiDAR mounted on drones, and airborne LASER scanners (ALS), of which the LiDAR instruments mounted on unmanned aerial vehicles (UAVs) is the latest advancement in LiDAR technology. We use the term drone for UAV. These devices produce data in the form of point clouds (PC), which are composed of three-dimensional (3D) Cartesian coordinates X, Y, Z, and the intensity of the reflected signal. Other associated sensors may provide other attributes such as RGB colors. These 3D points are obtained by processing the returned signal, i.e., the part of the emitted pulse backscattered toward the device. Several peaks of varying intensities are

1 Introduction

209

FIG. 1 Principles of LiDAR measurement using the example of a TLS. We can observe the different effects of the beam divergence on the illuminated area and multiple echoes. Modified from Jaboyedoff, M., Oppikofer, T., Abellan, A., Derron, M.-H., Loye, A., Metzger, R., Pedrazzini, A., 2012. Use of LiDAR in landslide investigations: a review. Nat. Hazards, 61, 5–28. DOI: 10.1007/s11069-010-9634-2.

typically extracted from the returned signal (which is in the form of a full wave). These peaks correspond to echoes from different reflectors such as the ground or vegetation (Fig. 1). In landslide research, the last pulse, i.e., in principle, the one corresponding to the ground, is typically used. For this purpose, the signal is automatically treated to record only the last pulse, or the last pulse is extracted during postprocessing. A PC can be projected on a horizontal grid to obtain a digital elevation model (DEM) by interpolating the PC points. If only the last pulses are kept and if they are cleaned from nonground points, then the DEM is a digital terrain model (DTM). In that case, DEM or DTM can be considered to be a 2.5D PC because no overhang can be represented. Due to their resolution and precision, LiDAR techniques (Shan and Toth, 2018; Vosselman and Maas, 2010) represent a major advance for landslide sciences in recent decades (SafeLand Deliverable 4.1., 2010; Jaboyedoff et al., 2012, 2018; Abella´n et al., 2014; Pradhan, 2017). Landslide mapping and characterization have greatly improved with the use hillshade or slope angle maps of LiDAR DTMs (Chigira et al., 2004; Glenn et al., 2006; Ardizzone et al., 2007). It has enabled extraction and characterization of the micromorphology of landslides (McKean and Roering, 2004). One of the main advances is the ability to characterize structures of rock faces that are not easily accessible and to provide joint characteristics (Kemeny and Post, 2003; Oppikofer et al., 2008; Sturzenegger and Stead, 2009). The high accuracy and resolution of LiDAR data provide an easy way to quantify volumes that are deposited by debris flows (Bull et al., 2010; Bremer and Sass, 2012) or removed by rockfall (Rosser et al., 2007; Carrea et al., 2014). These data also enable quantification of changes in topography from erosion, mass movements (Collins and Sitar, 2008; Michoud et al., 2015), and rockfalls (Lim et al., 2010).

210

7. Landslide analysis using laser scanners

The velocity or/and direction of slope movements can be extracted from the comparison of PCs (Teza et al., 2007; Travelletti et al., 2014; Fey et al., 2015). These data can provide displacement vector fields, which has the potential to lead to near real-time continuous monitoring using LiDAR (Kromer et al., 2015, 2017) to forecast failure. The study of several rockfall deformations allowed for the characterization of deformations before failure (Oppikofer et al., 2008; Abellan et al., 2010; Royan et al., 2014). The modeling of landslides has benefitted from the input of LiDAR data. It has improved rockfall modeling by adding more detail to the terrain roughness (Agliardi and Crosta, 2003), and provided input data for slope stability analysis (Brideau et al., 2011). Analysis of the susceptibility of hillslopes to shallow landslides (Dietrich et al., 2001) was also greatly improved by using high-resolution DTM from LiDAR data. For rockfalls, Matasci et al. (2018) developed a 3D susceptibility assessment based on PCs. Additional examples of applications are given below, following a review of the evolution of LiDAR in recent decades. The basics of LiDAR techniques are described because they give insight into the capabilities and drawbacks of LiDAR technologies.

2 A short history The principle of LASERs was established by Einstein (1917). The first functional LASER was designed by Theodor Maiman from the Hughes Research Laboratory in 1960. The first experiment of distance measurement by a range finder or “LiDAR” was performed a few months after the construction of the LASER (Brooker, 2009). In the 1970s, LASER techniques were used to measure distance precisely using EDM (electronic distance measuring); the first stand-alone devices coupled with a theodolite (Petrie and Toth, 2018a) often required a reflector, which is not the case for present-day LiDAR. The first topographic survey of the Earth’s surface was performed by a LASER altimeter deployed on an airplane, providing measurements with accuracy better than 30 cm at 300 m elevation (Miller, 1965). In 1984, a topographic survey along a profile was performed by NASA using an airborne oceanic LiDAR (Krabill et al., 1984). The major issue at this time was the positioning. The appearance of global positioning systems (GPS) allowed workers to acquire topographic transects over the Greenland ice sheet with an accuracy of