Earth Observation for Water Resources Management Earth Observation for Water Resources Management Current Use and Future Opportunities for the Water Sector Luis E. García Diego J. Rodríguez Marcus Wijnen Inge Pakulski Editors © 2016 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 18 17 16 15 This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http://creativecommons. org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: García, Luis E., Diego J. Rodríguez, Marcus Wijnen, and Inge Pakulski, eds. Earth Observation for Water Resources Management: Current Use and Future Opportunities for the Water Sector. Washington, DC: World Bank Group. doi:10.1596/978-1-4648-0475-5. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringe- ment rests solely with you. If you wish to re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. ISBN (paper): 978-1-4648-0475-5 ISBN (electronic): 978-1-4648-0476-2 DOI: 10.1596/978-1-4648-0475-5 Cover photo: © agsandrew/shutterstock.com. Used with permission. Further permission required for reuse. Cover design: Naylor Design, Inc. Library of Congress Cataloging-in-Publication Data has been requested. CONTENTS Forewordxiii Acknowledgmentsxv Prefacexvii About the Editors and Authors xxi Abbreviationsxxvii Executive Summary 1 The Why: Water and Earth Observations in the World Bank 1 The What: Earth Observation for Water Resources Management 3 The How: Practical Guidelines for Deciding on the Use of EO Products 5 Concluding Remarks 6 Notes7 PART I:  WATER AND EARTH OBSERVATIONS IN THE WORLD BANK 9 Aleix Serrat-Capdevila, Danielle A. García Ramírez, and Noosha Tayebi Overview 9 Chapter 1:  Key Global Water Challenges and the Role of Remote Sensing 11 Introduction 11 Water Scarcity 12 Water Quality 13 Impacts of Global Change 15 Extreme Events: Floods and Droughts 17 Conjunctive Use of Surface Water and Groundwater 20 The Food-Water-Energy Nexus 22 Green Growth and the Environment 24 Financial Issues 25 Institutional Frameworks and Governance Issues 26 Transboundary Issues 28 Notes 29 References 29   v Chapter 2:  The World Bank Group and Water 33 Introduction 33 Water Policy and Strategies 33 The Water Global Practice  34 The Water Portfolio 35 Annex 2A. Data Analysis Methodology of the Sector and   Theme Components in the Water Portfolio 37 Notes 38 References 38 Chapter 3:  The World Bank and Remote Sensing 39 Introduction  39 Internal Initiatives 39 External Initiatives 42 RS Applications in World Bank Water-Related Projects and   Analytical and Advisory Activities  43 Annex 3A. World Bank Remote Sensing Programs 46 Annex 3B. Methodology and Results of the Use of Earth Observation   Applications by Water Subsector 46 Notes 47 References 47 Chapter 4:  Key Data Needs for Good Water Management 49 Introduction 49 Key Data 49 Availability of Data 53 Operational Hydrology Today 57 Future of Operational Hydrology: Translating Data into Information 59 Annex 4A. Key Data for Water Resources Management 61 Notes 61 References 61 PART II:  EARTH OBSERVATION FOR WATER RESOURCES MANAGEMENT 63 Juan P. Guerschman, Randall J. Donohue, Tom G. Van Niel, Luigi J. Renzullo, Arnold G. Dekker, Tim J. Malthus, Tim R. McVicar, and Albert I. J. M. Van Dijk Overview63 Chapter 5:  Earth Observations and Water Issues 65 Introduction 65 EO-Related Water Resources Management in the World Bank Context 65 Field Measurement, Earth Observation, and Modeling 66 Relevant Variables Provided by EO 69 Notes 78 Reference 78 vi  |  C O N T E N T S Chapter 6:  Earth Observations for Monitoring Water Resources 79 Introduction 79 Characteristics of Sensors 79 Types of Data Obtained from Earth Observation 81 Notes 130 References 131 Chapter 7:  Assessing the Characteristics of Required and Available Earth Observation Data 145 Introduction 145 Establishing the Role of Earth Observation to Support   WRM Decision Making 147 Describing the Characteristics of EO Data Products 148 Determining the Characteristics of Minimum Required   EO Data  150 Determining the Generalized Characteristics of   EO Data Products 154 Worked Examples 163 Notes 166 References 166 PART III:  VALIDATION OF REMOTE SENSING–ESTIMATED HYDROMETEOROLOGICAL VARIABLES 167 Eleonora M. C. Demaria and Aleix Serrat-Capdevila Overview167 References168 Chapter 8:  Challenges of Remote Sensing Validation 169 Introduction 169 Methodological Approach 170 Note 171 References 171 Chapter 9:  Validation of Remote Sensing Data 173 Precipitation 173 Evapotranspiration 175 Soil Moisture 177 Snow Cover and Snow Water Equivalent 178 Surface Water Levels and Streamflows 180 Annex 9A. Summary of Scientific Literature on   Satellite Products Included in Validation 182 Notes 182 References 182 C O N T E N T S   |  vii Chapter 10:  Validation of Streamflow Outputs from Models Using Remote Sensing Inputs 185 Introduction 185 Streamflow Simulations Using Rainfall-Runoff Modeling 186 Streamflow Simulations Based on Remotely Sensed Water Levels Upstream 191 Note 191 References 192 Chapter 11:  The Bottom Line 193 Note 194 PART IV:  CONCLUDING REMARKS 195 Water and Development 195 Potential of Remote Sensing 195 Challenges 196 A Word of Caution 196 Making Decisions 196 A Word of Hope 197 Downscaling to the Local Context 197 Outlook 199 Appendix A:  Examples of Earth Observation Applications in World Bank Projects 201 Appendix B:  Examples of Water Information Product Generation Systems 205 Introduction 205 Flood Warning and Monitoring Systems 205 Soil Moisture and Drought Monitoring Systems 207 Irrigation Water Use and Crop Growth Monitoring Systems 208 Snow Extent 210 Water Resources Monitoring Systems 212 Water Resources Assessment and Scenario Studies 213 Notes 218 References 219 Index221 BOXES Box ES.1 Remote Sensing in World Bank Water-Related Projects 2 Box ES.2 Six Optical Water Quality Variables That Can Be Derived from EO Data 5 Box ES.3 Questions to Aid in Deciding Whether to Use Earth Observation for Water Resources Management 6 viii  |  C O N T E N T S Box 3.1 The Nile Basin Initiative 41 Box 5.1  Water-Related Topics and Subtopics Considered in the World Bank Context 66 Box 5.2  General Categories of Resolution and Examples of Platforms Providing This Type of Data  69 Box 7.1  Guiding Questions to Aid in the Decision Whether to Use Earth Observation for Water Resources Management  147 Box 7.2 Screening for Adequacy of Field Observations 151 Box 7.3 Examples of Accuracy Parameters 153 Box 7.4  Water Quality Variables Directly Determined by Earth Observation 161 Box 10.1 Validation of Streamflow Simulations Using Rainfall-Runoff Modeling 190 Box 10.2 Validation of Streamflow Simulations Using Remotely Sensed Water Levels Upstream  191 TABLES Table 2.1 Evolution of Key Principles over Time 35 Table 2.2 Total Water-Related Lending, by Water Subsector 37 Table 2.3 Water-Related Projects with Non-Water Sector Codes 37 Table 3.1  Number of Projects Using Remote Sensing in Water-Related Lending and Analytical and Advisory Activities, by Primary Theme  44 Table 3.2  Use of Remote Sensing in World Bank Lending and Analytical and Advisory Activities, by Water Subsector  45 Table 5.1  Relationship between Water Issues and Water Topics and Subtopics in the World Bank Water Partnership Program  70 Table 5.2 Overview of Water Issues and Relevant Variables Provided by Earth Observation 71 Table 5.3 Overview of Water Issues and Relevant Variables Provided by Earth Observation Rearranged to Focus on Spatial and Temporal Resolution  73 Table 6.1 Data Framework Comprising Domain-Characteristic Elements 80 Table 6.2 Types of Data Obtained from Earth Observation 81 Table 6.3 Overview of Main Characteristics of Some Widely Used Global Satellite-Derived Precipitation Estimates 84 Table 6.4  Overview of Sensors Most Suitable for Estimating Actual Evapotranspiration from EO Data 91 Table 6.5 Examples of Studies Using the Three General Classes of Actual ET Models 94 Table 6.6  Overview of Key Characteristics of Soil Moisture Sensors Aboard Past, Current, and Near-Future Satellite Platforms 97 Table 6.7  Overview of Sensors Most Suitable for Estimating Vegetation and Land Cover 103 Table 6.8 Examples of Global Vegetation Cover Maps 105 Table 6.9 Overview of Sensors Most Suitable for Estimating Groundwater 109 C O N T E N T S   |  ix Table 6.10  Overview of Sensors Most Suitable for Mapping Surface Water Extent and Height 113 Table 6.11 Overview of Sensors Most Suitable for Mapping Snow Extent, Snow Moisture, and Snow Water Equivalent 117 Table 6.12  Existing and Near-Future Satellite Sensor Systems of Relevance for Inland and Near-Coastal Water Quality 125 Table 7.1  Major Characteristics of Data Products and Their Type of Dependence  148 Table 7.2  Guiding Questions for Determining the Minimum Requirements of EO Data Products  151 Table 7.3  Field Data Requirements and Characteristics of EO-Based Precipitation Products  155 Table 7.4  Field Data Requirements and Characteristics of EO-Based Evapotranspiration Products  157 Table 7.5  Field Data Requirements and Characteristics of EO-Based Soil Moisture Products  158 Table 7.6  Field Data Requirements and Characteristics of EO-Based Vegetation and Vegetation Cover Products  158 Table 7.7  Field Data Requirements and Characteristics of EO-Based Groundwater Products  159 Table 7.8  Field Data Requirements and Characteristics of EO-Based Surface Water Products 160 Table 7.9  Field Data Requirements and Characteristics of EO-Based Snow Products  160 Table 7.10  Field Data Requirements and Characteristics of EO-Based Water Quality Products: Empirical Methods 161 Table 7.11  Field Data Requirements and Characteristics of EO-Based Water Quality Products: Semi-Empirical Methods 162 Table 7.12  Field Data Requirements and Characteristics of EO-Based Water Quality Products: Physics-Based Inversion Methods 162 Table 7.13  Guiding Questions for Determining the Characteristics of Required EO Data Products: Water Quality Example  164 Table 7.14  Guiding Questions for Determining the Required EO Data Product Characteristics: Efficiency of Agricultural Water Use Example 165 FIGURES Figure ES.1 Summary of Guidelines for Determining Whether to Use EO Products 7 Figure 2.1 Historical Water Lending, by Subsector, FY 2009–14 36 Figure 2.2 Historical Water Lending, by Region, FY 2009–14 36 Figure 3.1  Water-Related Lending and Analytical and Advisory Activities Using Remote Sensing, 1997–2013 43 Figure 3.2  Lending and Analytical and Advisory Activities Using Remote Sensing, by Bank Region 44 x  |  C O N T E N T S Figure 4.1  Components of the Global Terrestrial Network—Hydrology, 2013  55 Figure 4.2  Availability of Historical Monthly and Daily Discharge Data in the Global Runoff Data Center Database, 2004 and 2014  56 Figure 4.3 Seasonal Forecasts Issued by Two Regional Centers in 2013 58 Figure II.1 Schematic Showing Two Possible Ways to Read Part II 64 Figure 5.1  Conceptual Depiction of Information-Integration Paradigm Referred to as Model-Data Fusion 68 Figure 6.1 Characteristics of MODIS and Landsat TM Data Domain 80 Figure 6.2 Space-Based Precipitation Measurements from TRMM Satellite 82 Figure 6.3 Daily Rainfall Estimates for March 1, 2010, in Australia 87 Figure 6.4  Distribution of Real-Time Rain Gauges and Areas Where Satellite-Derived Precipitation is Likely to Improve Accuracy of Rainfall Estimation in Australia 87 Figure 6.5  Examples of Actual Evapotranspiration Estimates for Region in Western Queensland, Australia, during Flow Event in February 2004 89 Figure 6.6  Mapping of Actual Evapotranspiration Using High-Resolution Satellite Images for Part of Lower Gwydir Region in New South Wales, Australia 93 Figure 6.7  Comparison of Actual ET Estimates Derived from the NDTI Model with Actual ET Measurements from the Tumbarumba, NSW, Flux Tower 95 Figure 6.8 Remote Sensing–Based Soil Moisture Monitoring 96 Figure 6.9  Comparing Error Estimates for Soil Moisture Products Derived from Active and Passive Microwave Sensors Using Triple Collocation Technique 100 Figure 6.10 Combined Drought Indicator for Europe, Mid-March 2014 106 Figure 6.11  Map of AWRA-Derived Total Annual Landscape Water Yields in 2011–12 for Tasmania, Australia 107 Figure 6.12 Map of Irrigated Land Cover Types in the Krishna Basin, India 107 Figure 6.13  Example of Satellite Imagery Captured during Flood Event in Northern New South Wales, Australia 115 Figure 6.14  Schematic of the Light Interactions That Drive Optical EO Involving the Air, Water, and Substrate 123 Figure 6.15  Typical Reflectance Spectrum from Eutrophic Inland Water Body and Regions in which Different Water Quality Parameters Influence the Shape of that Spectrum 123 Figure 7.1 Guidelines for Determining Whether to Use EO Products 146 Figure III.1 Main Sources of Uncertainty in Satellite-Estimated Hydrologic Variables 168 Figure 9.1  Correlation Coefficients between Observed and Satellite-Estimated Precipitation 175 Figure 9.2  Correlation Coefficients between Observed and Satellite-Estimated Evapotranspiration 176 Figure 9.3 Errors in RS Estimations of Soil Moisture 178 C O N T E N T S   |  xi Figure 9.4 Correlation Coefficients between Observed and Satellite-Estimated Soil Moisture, by Type of Sensor  178 Figure 9.5  Correlation Coefficients and Snow Mapping Agreement between Observed and Satellite-Estimated Snow Water Equivalent and Snow Cover  180 Figure 10.1 Fractional Bias of Streamflow Simulations Forced by Rainfall Algorithms 187 Figure 10.2 RMSE of Streamflow Simulations Forced by Rainfall Algorithms 188 Figure 10.3 Relative Efficiency of Streamflow Simulations Forced by Rainfall Algorithms 189 Figure 11.1 Correlation Coefficients between Ground Observations and Satellite Estimates 194 Figure B.1  Map Showing Surface Water Extent in a Flood Event in Bolivia in March 2014 and Discharge Estimate from Passive Microwave 206 Figure B.2  Example Outputs from the Global Flood and Landslide Monitoring System 207 Figure B.3  Example Output of Selected Variables Generated by the AWAP System 209 Figure B.4  Example Output of a Crop-Coefficient (Kc) Map Produced by irriGATEWAY 210 Figure B.5 Snow Depth for the Continental United States on April 8, 2014 211 Figure B.6  Example Summary Output from AWRA for 2012 in the Murray-Darling Basin 214 Figure B.7  Example View from the Atlas of Groundwater-Dependent Ecosystems, Hosted by the Bureau of Meteorology 216 xii  |  C O N T E N T S Foreword W ater lies at the heart of economic and social To compensate for the current information gap, the development and is thus a critical factor in World Bank Group’s Water Global Practice has pulled poverty reduction. Growing economies and together knowledge on innovative technologies, such as populations require better water management to keep up viewing water from a distance, mainly through satellite with the demand for energy and food and to ensure access platforms, to help countries measure and monitor their to safe water and adequate sanitation. Twenty-first-­ water resources better. Remote sensing enables cover- century growth requires modern tools to help countries to age over large areas and spans of time without heavy field understand their water challenges, risks, and options. personnel requirements, and its accessibility, reliability, The World Economic Forum’s 2015 Global Risks report and accuracy have improved dramatically in recent years. ranks water crises as the most serious societal risk facing While both in situ and remote sensing measurements are the world, given the impacts associated with water scar- subject to specific limitations, researchers have developed city and overuse. If countries do not manage their endow- techniques that can combine or correlate data from both ments well—through improved water infrastructure and methods to benefit each other’s strengths. Understanding water management—they will not be prepared for the the potential combinations of available options has been complex challenges of the twenty-first century, will expe- a challenge for many practitioners. For this reason, Earth rience less economic growth, and may lose significant Observation for Water Resources Management: Current development gains made over the past decades. Use and Future Opportunities for the Water S ­ ector aims to While there is a broad consensus about the benefits shed light on the strengths and limitations of remote sens- of good water management, putting that knowledge into ing in order to help specialists to provide decision makers practice is usually easier said than done. To be able to with fast and reliable information. make good water decisions, countries need systematic This publication reflects experiences of more than ways to measure and monitor changes in water avail- 40 World Bank Group project leaders and more than ability. They need an accurate account of their current 20  international experts representing space agencies, resources—where, when, and how much—as well as an government organizations, and universities from Africa, illustration of the potential changes caused by seasonal, Asia, Europe, South America, and the United States. It also natural, and climate-induced variability, from rainfall integrates a report by the University of Arizona, Tucson, and runoff to evaporation and transpiration. In light of which was commissioned for this purpose, and another the harsh realities of climate change, this information is one co-funded with Australia’s national science agency, needed in larger quantities over broader areas and longer the Commonwealth Scientific and Industrial Research time periods than ever before. Organisation. We hope that the wealth of knowledge pre- Ground-based (in situ) observation networks are fun- sented in this publication will be useful for many devel- damental but, in some cases, provide infrequent or sparse opment practitioners around the world who are seeking information over small areas and at a high cost. Particu- practical answers to challenging technical questions larly in developing countries, such hydrometeorological about water and will help them to benefit from the enor- networks have deteriorated over time, at present provid- mous capabilities of the new tools that are now available. ing only limited information for managing compound problems. Undoubtedly, developing countries need inno- Jennifer Sara vative ways to get more information in an accurate, timely, Senior Director a.i. and usable format that builds on their existing infrastruc- Water Global Practice ture for monitoring water resources. The World Bank Group xiii Acknowledgments T his publication is the result of a collab- The primary authors of part I are Aleix orative effort involving the World Serrat-Capdevila (University of Arizona), Bank, the Commonwealth Scientific Danielle A. García Ramírez (World Bank), and and Industrial Research Organisation Noosha Tayebi (World Bank). The contribu- (CSIRO), and the University of Arizona. The tors are World Bank staff Thadeu Abicalil, Sara publication was edited by a core team com- Elizabeth Anthos, Timothy A.  Bouley, Anna prising World Bank consultants and staff Burzykowska, Rita Cestti, ­ ­ haudhary, Alisha C members Luis E. García, Diego J. Rodríguez, Xavier Chauvot de Beauchéne, Louise Marcus Wijnen, and Inge Pakulski. Croneborg, Bekele Debele, Erwin De Nys, The editors are especially thankful to peer Indira Ekanayake, Erick Fernández, Eric reviewers Anna Burzykowska, Bekele Debele, Foster-Moore, Maria Josefina Gabitan, Anju ­ Claire Kfouri, Xiaokai Li, and Zhongbo (Bob) Gaur, Andrew Goodland, N ­agaraja Rao Su, who provided valuable guidance and sug- Harshadeep, Valerie Hickey, Rafik Hirji, ­ gestions during the production stage. Special ­ Kremena M. Ionkova, Claire Kfouri, Johannes thanks go to professor Fernando Miralles- Kiess, Andrey V. Kushlin, Christina Leb, Qun Wilhelm and Raúl Muñoz from the University Li, Dhalia Lotayef, Katie L. McWilliams, of Maryland as well as to Matthijs Schur- ­ Hrishi Patel, Claudia Sadoff, Keiko Saito, ing and Macha Kemperman from the World Susanne Schelerling, Ahmed Shawky, Rebeca Bank, who provided valuable guidance and Soares, Pieter Waalewijn, Marcus Wishart, suggestions. and Winston Yu, as well as David Toll from the This publication involved many experts, National Aeronautics and Space Administra- researchers, and practitioners from both inside tion (NASA). and outside the World Bank. The editors and The primary authors of part II are Juan primary authors of parts I, II, and II of the P.  Guerschman, Randall J.  ­Donohue, Tom G. book would like to thank all of them for shar- Van Niel, Luigi J. Renzullo, Arnold G. Dekker, ing their experience and knowledge. Their col- Tim J. Malthus, and Tim R.  McVicar, from laboration is gratefully acknowledged. CSIRO; and Albert I. J. M. Van Dijk, from   xv CSIRO and Australian National University. Ministry of Infrastructure and the Environ- The Water Partnership Program of the Water ment); and Brian Wardlow (University of Global Practice, World Bank Group, in col- Nebraska, Lincoln). laboration with the Netherlands Space Office The primary authors of part III are ­Eleonora (NSO) and CSIRO, organized a specialist M. C. Demaria (University of Arizona) and review workshop titled “Understanding Water Aleix Serrat-Capdevila (University of Arizona). through Space,” which was held in The Hague, This part of the publication is the product of the Netherlands, from April 29 to May 2, 2014. a follow-up to the workshop “Understanding In the wake of this workshop, the following Water through Space.” The workshop partici- international experts provided the produc- pants provided valuable and inspiring ideas tion team with excellent feedback and sugges- that informed part III of this publication. tions: Wim Bastiaanssen (UNESCO, Institute This publication was made possible by the for Water Education); Richard de Jeu (Vrije financial contribution of the Water Partner- Universiteit Amsterdam); Brad Doorn (NASA); ship Program (see http://water.Worldbank Steven Greb (Group on Earth Observations and .org/water/wpp) of the Water Global Practice, Wisconsin Department of Natural Resources); World Bank Group. The production of part II Job Kleijn and ­Raimond H­ afkenscheid (Direc- was funded jointly by the World Bank’s Water torate-General for International Cooperation, Partnership Program and CSIRO. Within the Netherlands); Benjamin Koetz (­ European CSIRO, funding was provided by the Land and Space Agency); Xin Li and Wu Bingfang Water Flagship and by the Earth Observation (Chinese Academy of Sciences); Paida Man- and Informatics Future Science Platform. gara (Satellite Earth Observation and Disaster This publication was produced under the Risks); Massimo Menenti (Technical Univer- direction of Jennifer Sara (senior director a.i., sity Delft); Fernando Miralles (University of Water Global Practice), Junaid Kamal Ahmad Maryland); Mark Noort (HCP International); (former senior director, Water Global Prac- Morris Scherer-Warren (Agência Nacional de tice); William Rex (lead water resources spe- Águas); Aleix Serrat-Capdevila (University of cialist, Water Global Practice); Julia Bucknall Arizona); Soroosh Sorooshian (University of (former manager, Water Anchor); and Marie- California, Irvine); Zhongbo (Bob) Su (Fac- Chantall Uwanyiligira (practice manager, ulty of Geo-Information Science and Earth Water Global Practice). Observation, University of Twente); Michael This publication is a product of Water Part- Jasper van Loon, van der Valk (Hydrology.nl); ­ nership Program’s Global Initiative on Remote Joost Carpay, and Ruud Grim (Netherlands Sensing for Water Resources Management Space Office); Niels Vlaanderen (Netherlands (Water from Space). xvi  |  A C K N O W L E D G M E N T S Preface BACKGROUND picture of the remote sensing (RS) products available today—how they are generated, what A Global Initiative on Remote Sensing for specific water problems and situations they can Water Resources Management was launched ­ be applied to, their potential strengths and limi- in October 2013, financed by the World Bank’s tations, how better results could be obtained by Water Partnership Program of the Water Glob- using them jointly with in situ measurements, al Practice. The initiative supports Bank proj- and how they can be validated and evaluated to ect teams through (a) case studies and pilot inform the client better and enhance the Bank’s projects in selected countries to serve as the water-related operations. basis for the development of approaches and The use of remote sensing for hydrology procedures that can be replicated in other and water resources operational purposes, countries facing similar challenges; (b) target- while not new, is a fast-growing field. The term ed, short interventions of world-class experts “operational” has many definitions and may be aimed at advising and providing orientation on viewed from different perspectives. In this specific problems related to Bank investment context, however, the term refers not to the de- operations; and (c) knowledge dissemination, gree of readiness of the system to be used, but as well as advocacy and capacity-building ac- rather to the usability of products generated by tivities, in partnership with leading global and that system. In other words, the focus of atten- regional remote sensing and capacity-building tion is not the system itself or the products it organizations. generates but rather the accuracy, reliability, This publication is one product of that initia- and validity of the system products that will be tive, which seeks to improve the quality and ef- used to make a decision (or alter a past deci- fectiveness of water resources management, sion). This decision may be about the planning, planning, and project design by developing and design, and monitoring or operation of any disseminating, in collaboration with the Bank’s given institutional or physical system. It could operational staff and task team leaders, a clear pertain, for instance, to the selection of crops,   |  xvii the operation of field irrigation systems, or the presents practical guidelines for determining design of a hydraulic infrastructure such as a (a) whether the use of EO products could be reservoir. worthwhile in a specific situation and (b) how The scope of this publication is limited to the results could be improved by using them in water resources sector and, within that sector, combination with in situ measurements. to the RS estimation of key variables that form the basis of the planning, design, and operation of all water resources programs and projects— PART I. WATER AND EARTH precipitation, evapotranspiration, soil moisture, OBSERVATIONS IN THE WORLD vegetation and vegetation cover, groundwater, BANK surface water, snow cover, and water quality. The RS field is changing rapidly, and this re- Chapter 1 looks briefly at some challenges to view cannot claim to present more than a global water resources that are posed by popu- picture of the current state of the art. Never- ­ lation growth, urbanization, poverty, and other theless, this picture is a much-needed tool for problems currently facing many countries. It practitioners who have to make operational also discusses possible solutions to these chal- decisions. lenges, facilitated by the use of remote sensing, in combination with and in support of in situ measurements (this mix being especially sig- CONTENT nificant when data are scarce). Chapter 2 reviews some of the instruments In discussing the role of Earth observation that the World Bank uses to help its client (EO) in water resources management, this countries cope with these water challenges: publication goes from the general to the par- the Bank’s water policy and strategy, its Water ticular, adding more detail at each level. (The Global Practice, and the characteristics of its terms “remote sensing” and “Earth observa- lending and technical assistance portfolios. tion” are used interchangeably in this publica- Chapter 3 provides an overview of the tion.) As a framework highlighting why EO present use of RS in the Bank’s water-related needs to be considered in water-related activi- activities, including existing programs that ties, it first gives a broad overview of the major Bank staff can tap to obtain specialized assis- global challenges for water resources that exist tance for RS applications and products. Chap- today, outlines the role that remote sensing can ter 4 discusses ground measurements and RS play in tackling these challenges, and examines observations, their respective strengths and the significance of water-related projects in limitations, and the current state and future of the Bank’s portfolio and the context in which operational hydrology. remote sensing has been used to date in World Bank initiatives. To give insight into what EO can do to support operational decision making PART II. EARTH OBSERVATION in water-related projects, the publication con- FOR WATER RESOURCES tinues with a more in-depth discussion of the MANAGEMENT RS products available today—how they are generated, what specific water problems and Chapter 5 takes the results reported in part I situations they can be applied to, their poten- and summarizes the main global water issues tial strengths and limitations, and how they addressed by the World Bank—as reflected in can be validated and “ground-truthed,” to in- its portfolio—and connects them to a particu- form the client better and to enhance water-­ lar set of topics and subtopics defined by the related operations. Finally, as a how-to guide, it Water Partnership Program, which is part of xviii  |  P reface the Bank’s Water Global Practice. This facili- surface water. Chapter 10 reviews the valida- tates the identification of EO sensors to use, tion of streamflow outputs from models using often in combination with field measurement RS data as inputs. Chapter 11 summarizes the and modeling. results of this review. Chapter 6 describes eight (hydrometeor- ogical) variables of key relevance to water resources management that can be estimated PART IV. CONCLUDING REMARKS with remote sensing: precipitation, evapo- transpiration, soil moisture, vegetation and Part IV summarizes the main conclusions and land use and land cover, groundwater, surface recommendations regarding the role of water water, snow and ice cover, and water quality, in development and the great potential of RS as previously stated. It also includes a brief for improving water resources management, summary of the theoretical basis for estimat- the main challenges faced when applying it in ing these variables through Earth observation, this field, and a word of caution for its sensible a list of the current and near-future sensors use. It also reviews the main ­elements to con- that can provide such information, and, where sider when deciding whether to use RS in appropriate, a description of existing data water-related operations and briefly explains ­ products that are generated on a regular basis. how the use of RS in water resources manage- Chapter 7 provides a series of guidelines that ment could be enhanced through international project team leaders can use to decide whether cooperation, ultimately benefiting developing EO may be useful and, if so, which data sources countries. are the most suitable to consider. It also pro- Appendix A provides two examples of the vides a simple decision-making framework use of EO applications in World Bank projects. that helps to determine, for a given problem, Appendix B provides some notable examples how EO products might best be used to gener- of systems that use Earth observation, typi- ate the required information and how the EO cally in combination with ground data and data products with the most appropriate speci- modeling, to produce information on water fications should be selected. Moreover, for resources. each water resources application, information is presented about accuracy, availability, matu- rity, complexity, and reliability. AUDIENCE The audience for this book includes the client PART III. VALIDATION OF countries’ national water resources organiza- REMOTE SENSING–ESTIMATED tions, policy makers, and institutions dealing HYDROMETEOROLOGICAL with the water resources sector, as well as VARIABLES World Bank country directors, sector manag- ers, and task team leaders. This publication Part III complements part II and is structured ­ eaders will not be equally relevant to all of its r around four chapters. Chapter 8 discusses the and some may prefer to skip certain parts. For challenges inherent in the validation of RS example, policy makers in client countries or ­ estimations of hydrometeorological variables at the Bank may be interested primarily in the and explains the methodological ­ approach fol- discussion about the potential value of using lowed for the validation exercise. Chapter 9 RS for key water-related issues and its impor- reports the results of a literature review of val- tance and relevance for the Bank portfolio. idations of estimated precipitation, evapo- Task team leaders may be i ­ nterested primarily transpiration, soil moisture, snow cover, and in existing programs that they can tap to P reface   |  xix ­ btain specialized assistance for RS applica- o and how to select the most suitable EO data tions and products, which are described in products for their needs. part I. Technically oriented professionals may Everybody may be interested in perus- be especially interested in the technical expla- ing the examples and references presented nation of how RS data products relevant to throughout the publication and in the dis- water resources monitoring are derived from cussion of the validity of satellite-derived images obtained by satellite platforms, which values for key hydrometeorological vari- is provided in part II. At the operational level, ables (presented in part III). A big effort task team leaders as well as other practitio- has been made to integrate the publica- ners may be interested in the decision frame- tion’s technical and operational content in a work presented in chapter 7 to help them to coherent way, in the hope that this approach determine, for a given problem, how to use EO will offer every reader something useful in products to generate the required i ­ nformation their daily work. xx  |  P reface About the Editors and Authors EDITORS Energy of Costa Rica; Pan American Health Organization; World Health Organization; U.S. Luis E. García is a senior hydrology and water Agency for International Development (US- resources consultant in the Global Water Prac- AID); United Nations Educational, Scientific, tice of the World Bank Group. He participates and Cultural Organization (UNESCO); World in the Water Partnership Program initiatives Meteorological Organization; United Nations for the operational use of remote sensing (RS) Environment Programme (UNEP); United products in water projects and in the Water ­ Nations Development Programme; and IDB; as Expert Team. He is a civil engineer from San well as for firms from Denmark, Germany, Carlos University in Guatemala, with an MS in Guatemala, Switzerland, and the United States. hydrology and water quality from the Univer- Diego J. Rodríguez is a senior economist sity of California, Berkeley, and a PhD in hy- at the Water Global Practice of the World Bank drology and water resources planning from Group. He is the task team leader of World Bank Colorado State University, Fort Collins. He has initiatives that quantify the trade-offs of the more than 30 years of experience working in energy-water nexus (Thirsty Energy), the deci- water resources and watershed management, sion tree framework for confronting uncertain- hydrology, water quality, and applied remote ty in water resources planning, and the sensing. Prior to consulting for the Bank, he implementation of integrated urban water worked for 15 years at the Inter-American management. He is also program manager of Development Bank (IDB) in 26 Latin American the Water Partnership Program, where he pro- and Caribbean countries. He was principal wa- vides support to operational teams on the use of ter resources specialist at the IDB and team economic analysis in large water infrastructure leader for development of the IDB’s integrated investments. Prior to joining the World Bank, water resources management strategy. He has he worked at the Danish Hydraulic Institute consulted for CONAGUA (Mexico); Panama and the IDB. He has more than 20 years of Canal Authority, Ministry of Environment and ­ experience in sectoral, operational, policy, and   |  xxi strategy development in water supply, sanita- AUTHORS, PARTS I AND III tion, and water resources management. He holds a BS in economics from the University of Aleix Serrat-Capdevila is a research associate Maryland, an MA in applied economics from professor at the Department of Hydrology and Virginia Tech, and a PhD in economics (water) Water Resources, University of Arizona. He is from University of Groningen. also a member of the International Center for Marcus Wijnen is a senior water resources Integrated Water Resources Management management specialist in the Water Global (ICIWaRM), UNESCO, and of the National Practice of the World Bank Group. He is the Aeronautics and Space Administration (NASA) task team leader of the Water Expert Team, and USAID SERVIR Applied Sciences Team. which provides high-level technical support to Before obtaining his MA and PhD at the Uni- World Bank operational teams in the water versity of Arizona, he worked in Africa, South- sector. He provides a focal point on groundwa- east Asia, and Spain across public, private, and ter for the Water Global Practice and provides nongovernmental organization sectors, includ- technical support to operational teams work- ing for a year as a water and sanitation engineer ing on strategic groundwater engagements. in refugee camps and neighboring villages in Prior to joining the World Bank, he was a proj- Guinea-Conakry (West Africa). His work fo- ect leader for international consultancies and cuses on bridging the gap between scientific regional manager for Asia and the Middle East research and the transfer of new findings and at the French Geological Survey (Bureau de applications toward real-world water manage- Recherches Géologiques et Minières). For ment challenges. His main interests include more than 25 years, he has worked across the participatory planning and management ap- globe on water resources management chal- proaches; the impacts of climate change on re- lenges at the local, national, and transbound- gional water budgets and adaptation strategies; ary scale, employing a wide range of Earth how to handle uncertainty and inform human observation methodologies in the areas of sur- adaptation; water policy; and the use of satellite face water quality monitoring, hydrogeologi- precipitation products and other RS data for cal exploration, urban water use planning, and water monitoring and forecasting in poorly assessment of agricultural water use. ­ gauged basins. His main projects focus on the Inge Pakulski is an economist and senior edi- use of RS data for hydrologic applications to tor specializing in development economics and support water-related decision making, mostly environmental studies. In her early career, she in African basins. In addition to the World worked for private- and public-sector entities in Bank, he also collaborates with organizations the Netherlands in the field of development co- such as the Institute of Water Resources (U.S. operation and information technology. In re- Army Corps of Engineers), the G-WADI Pro- cent years, she has worked almost exclusively as gram (UNESCO), the Southwest Climate Sci- an editor, and mainly on development issues— ence Center (U.S. Geological Survey), the water resources management, carbon emissions National Science Foundation, and the Alliance trading, public sector policies, and transport. for Global Water Adaptation. He recently re- She has edited numerous reports and studies for ceived the Commander’s Award for Civilian the World Bank, the IDB, and development Service for his “key role in helping ICIWaRM nongovernmental organizations. She is a Dutch fulfill its mission in service of UNESCO and the national and edits texts in Dutch, English, and United States (2009–2014).” Spanish. She also speaks French, Polish, and Danielle A. García Ramírez is a program an- Portuguese. She holds an MA in economics alyst at the Global Programs Unit in the Water from Erasmus University Rotterdam. Global Practice of the World Bank. She has xxii  |  A bout the E ditors and A uthors more than five years of experience working on ­ atellite-estimated precipitation for hydrolog- s water and sustainable development at the in- ic purposes in developing countries. She has a ternational level. In the past four years, she has BS in water resources engineering from the been a member of the core program manage- Universidad Nacional del Litoral in Argentina, ment team of the World Bank’s Water Partner- an MS in meteorology from the University of ship Program (WPP), a multidonor trust fund Utah, and a PhD in hydrology from the Univer- that aims to improve water resources manage- sity of ­Arizona. She specializes in improving ment and climate resilience in Bank opera- hydrologic forecasting and monitoring sys- tions. She also supports the WPP-funded tems using satellite precipitation estimates in initiatives Water from Space and Resilient flood-prone regions of South America and Water Decisions. She holds an MSc in political ­ ­ Africa where ground observations are sparse. economy from the University of Essex (U.K.) She has worked at the Centro de Cambio and specializes in quantitative methods, inter- ­ Global at the Pontificia Universidad Católica national development, and water management de Chile as a research associate, where she policy. studied the impacts of climate change on Noosha Tayebi is an RS and disaster risk Alpine basins, and as a fellow at the Northeast ­ management specialist for the World Bank’s Climate Science Center of the University of Water Global Practice. Her main expertise is in Massachusetts, where she focused on under- translating projects’ operational and informa- standing how streamflow extremes and snow tion requirements into technical specifications cover in the northeastern and upper midwest- for customized Earth observation products. ern United States are affected by projected cli- Prior to joining the World Bank, she worked matic changes. She has also worked with the for five years in research and development for UNEP to empower Kenyan women by harvest- the government of Canada on RS and geospa- ing rainwater for their water supply. tial applications using satellite data at Defence Research and Development Canada and in a private firm providing software engineering AUTHORS, PART II services to the Department of National De- fence on advanced sensor integration and data Juan Pablo Guerschman is a senior research visualization. She holds a BS in electrical engi- ­ scientist with Commonwealth Scientific and neering and an MS in systems science and Industrial Research Organisation (CSIRO) ­ operational research from the University of Land and Water. He joined CSIRO in 2005, Ottawa, Canada. ­ after receiving a PhD in agricultural sciences Eleonora Demaria is a research hydrologist- from the University of Buenos Aires meteorologist at the U.S. Department of (­Argentina). In his first years at CSIRO, his re- Agriculture’s Agricultural Research Service ­ search focused on the calibration and applica- and a research associate at the University of tion of a regional carbon cycle model and the Arizona, both in Tucson. An Earth scientist, integration of remote sensing and ground- she studies the interactions between land based observations through model-data as- s­urface and atmosphere to assess the vulnera- similation for the analysis of carbon dynamics bility of human and natural systems to weather of tropical savannas. From 2007 onward, he and climate events. Her main research inter- has been a project researcher and then re- ests are the impacts of climate change on search scientist with the Model-Data Integra- ­ extreme precipitation and streamflow events, tion Team of the Environmental Earth the role of atmospheric rivers on flooding Observation Program. He has played a leading events, and how to improve the usefulness of role in developing and evaluating methods for A bout the E ditors and A uthors   |  xxiii using satellite observations in hydrological physics from Curtin University (Perth, Western and land management applications. Between Australia) and possesses extensive experience 2009 and 2012, he led part of the research in (bio)physical modeling and RS data analysis. portfolio of the Water Information Research His research explores the role that Earth obser- and Development Alliance between the Bu- vations can play as inputs and constraints on reau of Meteorology and CSIRO Water for a biophysical models through techniques of Healthy Country Flagship dealing with RS of ­ model-data fusion and data assimilation. He land cover and landscape water and using this ­directs current research efforts to develop satel- information to inform the Australian Water lite soil moisture information products that are Resources and Assessment System. He has better suited to agricultural production model- been developing algorithms for estimating ing than satellite-derived products alone. To vegetation cover from remotely sensed data this end, he works with university collaborators across rangelands and croplands and applying on methods to downscale coarse-resolution these estimates to deliver timely information data and the observation operator for assimilat- for better management of these environments. ing data into pasture growth models. Randall J. Donohue is a research scientist Arnold G. Dekker is a research scientist with with CSIRO in the areas of ecohydrology, RS, CSIRO and director of the CSIRO Earth and environmental physics, focusing on the ­ Observation and Informatics Future Science dynamics of vegetation function under a Platform. He holds a PhD in hyperspectral re- changing climate. He has a joint PhD from mote sensing of inland water quality from Australian National University and CSIRO. ­ Vrjie Universiteit in Amsterdam. Before join- His research focuses on understanding the ing CSIRO, he developed innovative methods changing interactions between vegetation, cli- and applications using Earth observation for mate, carbon, and water, using RS as a primary inland coastal water quality detection, moni- input. He has developed frameworks for ex- toring, and management in Europe. At CSIRO, amining the role of vegetation dynamics in his scientific work focuses on physical pro- catchment hydrology and landscape carbon cesses at an aquatic ecosystem scale, suited for dynamics and for better understanding the im- resource management or for integration into pacts of elevated carbon dioxide on vegetation predictive and hindcasting models. He is an in- functioning. ternational and national leader in defining Tom G. Van Niel is a scientist with CSIRO methods for operationalizing Earth observa- interested in spatial and temporal modeling of tion of aquatic ecosystems. He has led major vegetation, water, climate, and surface radia- national and international research projects. tion and heat processes. Recently he has been He holds a dozen international positions influ- modeling evaporation over all of Australia encing science, implementation, and opera- based on meteorological data and thermal RS tionalization of Earth observation for tackling observations. He focuses on (a) improving the societal benefit areas. He holds an adjunct pro- spatial and temporal representation of surface fessorship at the University of Queensland and interactions, such as vegetation and water, is a member of the Australian Marine Science through statistical and mathematical model- ­ Association, the Australian Government Space ing; (b) spatiotemporal analysis of climate Coordination Committee, and the Australian variability; and (c) improved estimation of sur- Earth Observation Coordinating Group. He face radiation budget and water-heat balance co-represents Australia and CSIRO as part of through terrain analysis. the International Space Agencies’ Committee Luigi J. Renzullo is a senior research scientist on Earth Observation Satellites 2016 Chair with CSIRO. He received his PhD in a ­ pplied Team and in water resources–related activities xxiv  |  A bout the E ditors and A uthors of the Group on Earth Observations. He is system ­ understanding—and in the nexus of active in several United Nations task teams ­ vegetation dynamics, catchment water bal- that are considering how to use Earth observa- ance, and climate change. He has led several tion for the 2030 Sustainable Development international and national projects and cur- Goals agenda. rently serves as editor-in-chief of the Journal of Tim J. Malthus is research group leader of Hydrology and associate editor of Remote Sens- the Coastal Modelling and Sensing Group in ing of Environment. He seeks to understand the Coastal Management and Development processes and feedbacks, to rank them from Program of CSIRO’s Oceans and Atmosphere primary to tertiary importance, and to identify Business Unit. He has a BS in zoology from the those where landscape management can influ- University of Otago (Dunedin, New Zealand) ence the process (directly or i ­ndirectly). He is and more than 25 years of research experience committed to parsimonious biophysical mod- in the remote sensing of inland water quality, eling and analytical frameworks and occasion- underwater optics, field measurement, and al- ally quotes Dr. Frank Westheimer’s (1912–2007) gorithm development. He combines skills in Law: “A few months in the laboratory can save calibration, validation, and field spectroscopy a few hours in the library.” with analysis of airborne and satellite RS data Albert I. J. M. Van Dijk is professor of water to monitor environmental change and better science and management at the Fenner School inform wider environmental policies. He has a of Environment and Society in the Australian background in aquatic ecology, specifically in National University and adjunct science leader water quality of inland and coastal waters. with CSIRO Land and Water. With a PhD from Tim R. McVicar is a research scientist at Vrije Universiteit, he has expertise in catch- CSIRO with training in biophysical modeling, ment hydrology, basin water management, remote sensing, ecohydrology, and hydrocli- drought processes, water resources monitor- matology, and holds a PhD in environmental ing and forecasting, and the combination of science and remote sensing from the Austra- Earth observation, ground observation, and lian National University. He is interested in hydrological modeling. He led development of developing parsimonious modeling frame- ­ the Australian Water Resources Assessment works that capitalize on the availability of system, a large observing and modeling system long-term temporal data to understand retro- used operationally by Australia’s Bureau of spective processes, with a view to making pro- Meteorology. He has authored more than 130 spective projections—based on enhanced articles in international journals. A bout the E ditors and A uthors   |  xxv Abbreviations AAA analytical and advisory activity AASTR Advanced Along Track Scanning Thermal Radiometer ALOS Advanced Land Observation Satellite AMSR Advanced Microwave Scanning Radiometer ASACT Advanced Scatterometer ASAR Advanced Synthetic Aperture Radar ASIMUTH Applied Simulation and Integrated Modelling for the Understanding of Toxic and Harmful Algal Blooms AVHRR Advanced Very High Resolution Radiometer AWAP Australian Water Availability Project AwiFS Advanced Wide Field Sensor AWRA Australian Water Resources Assessment BEAM Basin Economic Allocation Model BGC biogeochemical global climate CC correlation coefficient CILSS Permanent Inter-State Committee for Drought Control in the Sahel CMAP Climate Prediction Center Merged Analysis of Precipitation CMORPH Climate Prediction Center MORPHing technique CNES Centre National d’Etudes Spatiales CSC Climate Services Center CSIRO Commonwealth Scientific and Industrial Research Organisation CV coefficient of variation DEM digital elevation model EMS electromagnetic spectrum ENTRO Eastern Nile Technical Regional Office EO Earth observation ERS European Remote Sensing Satellite xxvii ESA European Space Agency ET evapotranspiration EUMETSAT European Organisation for the Exploitation of Meteorological Satellites EVI enhanced vegetation index FAO Food and Agriculture Organization of the United Nations fPAR fraction of absorbed photosynthetically active radiation GCM global climate model GEMS Global Environmental Monitoring System GEO Group on Earth Observations GIS geographic information system GLDAS Global Land Data Assimilation Systems GNIP Global Network of Isotopes in Precipitation GOES geostationary operational environmental satellite GPCC Global Precipitation Climatological Center GPM Global Precipitation Measurement GRACE Gravity Recovery and Climate Experiment GRDC Global Runoff Data Center GSMaP Global Satellite Mapping of Precipitation GTN-H Global Terrestrial Network–Hydrology GW groundwater HEC-Ras Hydrologic Engineering Centers River Analysis System HH horizontal transmit and horizontal receive HIRS High Resolution Infrared Radiation Sounder HV horizontal transmit and vertical receive IBRD International Bank for Reconstruction and Development ICIWaRM International Center for Integrated Water Resources Management ICPAC Climate Prediction and Application Centre IDA International Development Association IGAD Intergovernmental Authority on Development IGRAC International Groundwater Resources Assessment Center IMERG  Integrated Multisatellite Retrievals for GPM [Global Precipitation Measurement] IPWG International Precipitation Working Group ISMN International Soil Moisture Network JAXA Japan Aerospace Exploration Agency JERS-1 Japanese Earth Resources Satellite 1 KGE Kling-Gupta efficiency LAI leaf area index LiDAR Laser Imaging, Detection, and Ranging LSWI land surface water index MERIS Medium Resolution Imaging Spectrometer MODIS Moderate Resolution Imaging Spectrometer MRE Micronet-relative efficiency NASA National Aeronautics and Space Administration NBI Nile Basin Initiative NCORE Nile Cooperation for Results Project xxviii  |  A B B R E V I A T I O N S NDMC National Drought Mitigation Center NDSI normalized difference snow index NDVI normalized difference vegetation index NIR near infrared NLDAS North American Land Data Assimilation Systems NOAA National Oceanic and Atmospheric Administration NOHRSC National Operational Hydrologic Remote Sensing Center NSE Nash-Sutcliffe efficiency NSIDC National Snow and Ice Data Center PERSIANN Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks PM Penman-Monteith PMW passive microwave POES polar-orbiting operational environmental satellite REBM resistance energy balance model RIMES Regional Integrated Multi-Hazard Early Warning System for Africa and Asia RMSE root mean squared error RS remote sensing SADC Southern Africa Development Community SAR synthetic aperture radar SEBAL Surface Energy Balance Algorithm for Land SEBS Surface Energy Balance System SERVIR Regional Visualization and Monitoring System S & I snow and ice SMAP Soil Moisture Active Passive SMMR Scanning Multichannel Microwave Radiometer SMOS Soil Moisture and Ocean Salinity Sensor SPOT Satellite for Earth Observation SPP satellite precipitation product SSM/I Special Sensor Microwave Imager SVAT soil-vegetation-atmosphere transfer SW surface water SWIR short-wave infrared TIR thermal infrared TMI TRMM Microwave Imager TMPA TRMM Multisatellite Precipitation Analysis TOA top of atmosphere TRMM Tropical Rainfall Measuring Mission UNESCO United National Educational, Scientific, and Cultural Organization USAID U.S. Agency for International Development USDA U.S. Department of Agriculture UTC Coordinated Universal Time VIIRS Visible/Infrared Imager Radiometer Suite V and LC vegetation and land cover VIS visible infrared WGMS World Glacier Monitoring Service A B B R E V I A T I O N S   |  xxix WIRADA Water Information Research and Development Alliance WISP Water Information System Platform WISP-3 Water Insight Spectrometer (with three radiometers) WOIS Water Observation and Information System WQ water quality WRM water resources management WRSI water requirement satisfaction index W3RA World-Wide Water Resources Assessment xxx  |  A bbreviations EXECUTIVE SUMMARY Water contributes to all aspects of economic staff and task team leaders as well as external and social development. Especially in devel- partners, a clear picture of the potential role oping countries, water supply, sanitation, and of Earth observation (EO)1 in addressing par- a healthy environment form the basis of suc- ticular water-related issues. This publication cessful poverty reduction and shared-growth is a product of that initiative and aims to strategies. The use of remote sensing (RS) for illustrate the why, what, and how of using EO operational purposes in hydrology and water data. resources, while not new, is a fast-growing field. The term “operational,” as used here, THE WHY: WATER AND EARTH pertains not to the readiness of RS products OBSERVATIONS IN THE themselves, but to the actual use of these prod- WORLD BANK ucts when making decisions—a decision about the planning, design, and monitoring or Development organizations confront many ­operation of any given institutional or physical challenges in a rapidly changing world. These system. It could concern, for instance, the challenges include, among others, water selection of crops, the operation of field irriga- ­ scarcity as a result of growing demands for tion systems, or the design of hydraulic infra- water, climatic variability and change, causes structure such as a reservoir. of environmental and hydrologic change A Global Initiative on Remote Sensing for other than climate, the occurrence of extreme Water Resources Management was launched events (floods and droughts), complex issues in October 2013, financed by the World related to the conjunctive use of surface Bank’s Water Partnership Program of the water and groundwater, food and energy Water Global Practice. It aims, among other dynamics, growth and environmental prob- things, to put together and disseminate, in lems, as well as governance and transbound- collaboration with the Bank’s operational ary issues. 1 The successful tackling of issues such as these observations has been in decline globally since lays the foundation for sustainable development the 1980s. Among the many reasons for this and poverty reduction strategies that organiza- decline is that, particularly in developing tions such as the World Bank help ­ client coun- regions, real-time, ground-based measure- tries to develop. A review of the Bank’s ments have been marked by relative scarcity, water-related projects shows that, over the last poor accessibility, deficient quality control, and five years, the share of these kinds of projects in lack of availability and sharing options. the total portfolio has almost ­doubled—reaching Remote sensing plays an increasingly impor- about 18 percent. Nearly 800 projects with tant role in providing the information needed water-related themes were approved between to confront key water challenges. In poorly fiscal year 2002 and fiscal year 2012. Of these, gauged basins, at time intervals of several days, the majority dealt with water supply and sanita- real-time satellite estimates of precipitation tion or with flood protection. Projects on irriga- and derived streamflow forecasts can help tion and drainage and on hydropower ranked managers to allocate water among users and to second and third, respectively. operate reservoirs more efficiently. In large riv- These and other water-related projects, at ers, data on river and lake surface elevation can some point and in one way or another, undoubt- be used to estimate flow in the upper parts of edly needed data on precipitation, temperature, the basin and to predict flow downstream. Soil evapotranspiration, normalized difference veg- moisture observations may give insight into etation index, streamflow, soil moisture, wind how much irrigation is needed, as well as help speed, groundwater recharge, groundwater to forecast and monitor drought conditions. level, surface water level, snow or ice cover, Water managers in snow-dominated areas can snow or ice water equivalent, pumping and use estimates of snow cover and snow water groundwater change, land subsidence evalua- equivalent to assess how much water is in stor- tions, water surface elevation, and water qual- age and determine what watersheds it is ity. Traditionally, ground observations have stored in. provided these kinds of data. However, the Remote sensing also enables the monitoring number of ground hydrometeorological of many parameters of surface water quality to assess the repercussions of river basin manage- ment policies, land use practices, and non- BOX ES.1 point-source pollution as well as the likelihood of algal blooms and other threats to the quality Remote Sensing in World Bank Water-Related Projects of water supply systems. In World Bank water-related projects, the following sectors and themes In situations involving the food-water- have been the highest users of remote sensing: energy nexus, governance and adaptive man- • Flood protection and general water, sanitation, and flood protec- agement, or transboundary settings, remote tion (more than 50 percent of projects using RS) sensing may help decision makers to adjust • Irrigation and drainage (25 percent) past policies or facilitate early warnings by • Climate change related (12 percent) providing information from parts of a basin • Natural disaster management (17 percent) lying outside a nation’s borders. In collaboration with space agencies in In projects with a large water resources management component (55 ­ percent), remote sensing has been used primarily in lending operations Europe, Japan, and the United States, the (46 percent) and less frequently in advisory and analytical support (9 per- World Bank has increasingly been using RS cent of projects using RS). data, as summarized in box ES.1. Note: Sectors and themes are not mutually exclusive. Despite the growing demand for RS data, the percentage of projects in the portfolio that have 2  |  E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T used these technologies is still low. There is Because of its fine-scale spatial and temporal great potential for their use in operations variability, monitoring precipitation in large related to climate variability and change, agri- areas challenges field-based measurement cultural systems, and water systems planning ­ networks. Gauge density is not the only factor and management. Actual or planned uses of RS affecting when and where satellite data are products vary from the evaluation of a project’s expected to improve the rainfall estimation; impact on agricultural water management, other factors include the type of topography agricultural water-saving measures, and sup- and rainfall. However, in large parts of the port services to the provision of input for mod- globe, rain gauge networks are sparse, and ern, basin-wide water resources information available evidence increasingly suggests that systems; feasibility studies; basin planning, satellite-derived precipitation, together with monitoring, and forecasting; transboundary weather model reanalysis estimates, can pro- options for mitigating flood risks; investment vide highly valuable rainfall estimates and planning and basin decision support systems; narrow the information gap. ­ and institutional or community planning Evapotranspiration involves two processes— frameworks for addressing environmental and evaporation and transpiration—that occur social issues. simultaneously and are therefore difficult to The huge potential of RS applications has distinguish one from the other. Evaporation is created the need for an easily accessible com- the change from a liquid to a gas. It may occur pilation of available products and their suit- from the Earth’s surface (for example, the soil, a ability for various water resource management water body, or other type of surface), through needs as well as for guidelines to support deci- plant leaves (transpiration), and from rainfall on sions regarding when and how to use them the surface of the leaves (interception). Actual more effectively for operational purposes. evapotranspiration is difficult to measure, let alone estimate accurately, both spatially and temporally over large areas. This is not the case THE WHAT: EARTH OBSERVATION of potential evapotranspiration, which can be FOR WATER RESOURCES readily calculated using commonly measured MANAGEMENT hydrometeorological variables. Actual evapotranspiration can be estimated Some key variables are usually involved in through three methods: (a) empirical ­ methods, these activities and, given the present state-of- (b) energy balance methods, and (c) the the-art technology, may be estimated using Penman-Monteith method. Each has specific remote sensing. These variables are precipita- strengths and weaknesses. These methods tion, evapotranspiration, soil moisture, vegeta- form the starting point of numerous EO-based tion and vegetation cover, groundwater, level implementation models. While it is unlikely and extent of surface water,2 snow cover, and that any single approach will be best suited to optical water quality. This publication first estimating actual evapotranspiration in every identifies the minimum spatial and temporal situation, a common issue is having a system in resolution requirements for these types of vari- place for robust and repeatable assessment of ables and subsequently links them to relevant the models. While EO-based models of actual water activities. evapotranspiration may still be used for places Precipitation is the process by which water where no ground measurements exist, if their returns from the atmosphere to the Earth’s reliability (probability of errors) cannot be surface in liquid form (rain), in solid form assessed, their suitability for management (snow or hail), or in combined form (sleet). purposes may not be known with certainty. E X E C U T I V E S U M M A R Y   |  3 Soil moisture is defined as the amount of distinguishing between structurally distinct water in the uppermost layers of the soil col- types of vegetation. umn, where the definition of “uppermost” Groundwater is the water contained in the varies with sensing technology or modeling saturated zone—the subsurface volume below application and can vary from the top 1 centi- the water table—where water fills the cracks meter to the first 1 meter of soil or more. The and pores of rock, sediment, and soil. Ground- monitoring of soil moisture has advanced water is a critical source of water for human considerably over the last decade, with inno- consumption and agriculture, especially where vative ground- and satellite-based technolo- surface water is scarce or polluted. It also mod- gies for monitoring large areas. Global erates streamflow, producing the longer-term monitoring of soil moisture is only achievable base flow component of total flows, which with satellite Earth observation in conjunc- decouples flows somewhat from the variability tion with field-based soil moisture monitor- inherent in the climatic drivers of streamflow. ing networks. Satellite soil moisture sensing As groundwater lies below the land surface, technology is based on either radiometric there are currently no techniques for using measurements of emissions from the soil Earth observation to determine the groundwa- (passive microwave approach) or radar tech- ter level directly, so the use of remote sensing nology that transmits a pulse of electromag- here is inferential and has limitations. The netic radiation to the Earth’s surface and main indirect techniques used are satellite measures the backscattered signal (active gravity field mapping (gravimetry) and radar approach). Objective assessments comparing interferometry. the accuracy of satellite estimates of soil Surface water, as treated here, refers to nat- moisture with the accuracy of surface mea- ural or man-made reservoirs, very wide rivers, surements are necessary to gain the user com- and water accumulation caused by flooding, munity’s acceptance of the products and often which can range from small overbank floods involve evaluations against field-based soil near water streams to very large floods cover- moisture measurements. ing hundreds of square kilometers. Measuring Vegetation is the collective term for the cov- surface water elevation using EO technology erage of plants across land areas, vegetation can provide estimates of changes in total water attributes, and processes related to the prop- volume in reservoirs and wetlands and also be erties and functioning of those plants when used to estimate river discharge, although this considered at the landscape scale. Vegetation is currently only possible in wide rivers (that is, plays an important role in the hydrologic rivers several hundreds of meters wide). A cycle—partitioning precipitation between large number of algorithms exist for mapping evaporation and runoff. A large percentage of surface water areas. One major disadvantage of terrestrial evaporation is transpired by vegeta- using optical imagery is that the images are tion. It can be characterized quantitatively, subject to cloud contamination. Radar and pas- using measures of height, canopy and stem sive microwave imagery is not affected by density, leaf area, and the like, or qualitatively, clouds or water vapor and therefore can pro- as classes of vegetation or types of cover, such vide useful information on surface water under as forest, croplands, and tundra. Classes of clouds. vegetation cover are identified using combina- Snow cover exists where snow accumula- tions of remotely sensed variables—often com- tion is sufficient for the land surface to have a bined with ancillary data such as climate and reasonably continuous layer of snow. As snow land use maps—and field observations. RS esti- contains freshwater, meltwater from snow mates of vegetation height and biomass help in cover is an important source of water for 4  |  E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T consumption, irrigation, and power generation BOX ES.2 in many parts of the globe. In the visible wave- lengths, snow is generally highly reflective Six Optical Water Quality Variables That Can Be (that is, characterized by a high albedo), which Derived from EO Data makes it relatively easy to detect, as it contrasts Directly assessed: with the surrounding landscape. The areal extent of snow cover can be detected using • Chlorophyll optical, near infrared, and microwave sensors • Cyanobacterial pigments or a combination of these. Active and passive • Colored dissolved organic matter microwave sensors are the primary means of • Total suspended matter detecting snow depth, snow water equivalent, Indirectly assessed: and snow wetness. • Vertical attenuation of light coefficient Water quality refers to the physical, chemi- • Turbidity/Secchi disk transparency cal, and biological content of water and may vary geographically and seasonally, irrespec- tive of the presence of specific pollution sources. Earth observation can only directly assess water quality parameters—including can be obtained through Earth observation are many chemical compounds, such as nutri- classified according to their potential useful- ents—if they have a direct expression in the ness. In addition, the most appropriate spatial optical response of the water body. Only a sub- and temporal resolutions for each variable are set of these variables, often referred to as opti- listed. cal water quality variables, can be assessed directly through Earth observation (box ES.2). In some cases, nonoptical products may be THE HOW: PRACTICAL GUIDELINES estimated through inference, proxy relation- FOR DECIDING ON THE USE OF EO ships, or data assimilation with remotely PRODUCTS sensed optical properties of products such as nitrogen, phosphate, organic and inorganic For many potential applications, the use of EO micropollutants, and dissolved oxygen. How- data products will immediately and obviously ever, these relationships are stochastic, may be useful for improving water resource man- not be causal, and may have a limited range of agement and water monitoring. Yet in other validity. By making use of the combined infor- cases, guidance may be needed to decide mation in directly measurable optical proper- whether Earth observation could be useful ties, it is possible to derive information about and, if so, which data sources would be the eutrophication, environmental flows, and car- most suitable to consider. For those cases, a bon and primary productivity. simple decision framework is included to help Detailed information about the various sen- to determine, for a given issue, (a) the optimum sors is summarized in tables in the main text. use of EO products, that is, those that generate These tables provide an overview of EO capa- the required information, and (b) the optimum bilities regarding estimation of the eight vari- selection method, that is, the one that ensures ables of interest previously mentioned. The the EO data products with the most appropri- satellite sensors are described in terms of their ate specifications are chosen. For the applica- spectral, radiometric, and temporal character- tion to each specific area of water resources istics. For each pertinent water resource man- management, the issues of accuracy, availabil- agement activity, the relevant variables that ity, maturity, complexity, and reliability should E X E C U T I V E S U M M A R Y   |  5 v BOX ES.3 few include validation of the results of those tools. Still, some reports on validation efforts Questions to Aid in Deciding Whether to Use Earth can be found in the literature. Observation for Water Resources Management It is presently believed that the combined use of EO and field data generally provides the 1.  Define the nature of the water resource management problem. best information outcomes, based on a review • What questions need to be answered? of the literature on the validation of EO-­ • What policies or regulations drive these questions? estimated hydrometeorological variables and • Who are the stakeholders and beneficiaries of a solution to the the following considerations: (a) overall satel- problem? lite estimations are well correlated with 2. Explore the capacity of sustaining and maintaining decision support ground observations; (b) despite these strong and monitoring programs. correlations, satellite estimates are relatively • Is local capability available and adequate? uncertain; (c) despite the uncertainties inher- • Is training needed? ent in in situ measurements, it is believed that • Are local and international resources required? the measurements of an EO data product will 3.  Define the status of existing data and observation networks. rarely be as accurate as those of an equivalent • What metering is currently available? field measurement; and (d) despite the gener- • What is the condition of the data networks? ally lower accuracy, EO products still are an • Are there any impediments to sharing, collating, and archiving the important alternative data source because EO data (for example, transboundary issues)? imagery can provide information with greater • What, if anything, has been done in the past to address the issues spatial extent, spatial density, and temporal at hand? frequency than most field-based (point-based) • Has any monitoring or modeling been conducted? observation networks. 4.  Evaluate the adequacy of field observations. • Are the observations well defined? • Are the spatial density, frequency, continuity, and period of inter- CONCLUDING REMARKS est detailed? • Are observations accurate and available? Good water resources management and plan- ning are essential to sustain economic and human development. Especially in developing be duly considered. The main questions to ask nations, there is a need to bridge the gap between are presented in box ES.3. existing technologies and operational applica- Information that helps to answer these tions in support of the planning, design, questions—as a basis for determining whether operation, and management of water resources. ­ specific EO products meet the data require- There is great potential for space-based Earth ments of the water resources management observation to enhance the capability to moni- problem under consideration—is provided in tor the Earth’s vital water resources, especially tables in the main text. The step-by-step pro- in data-sparse regions of the globe. Despite this cedure is shown in the simplified diagram in potential, EO data products are currently figure ES.1. underused in water resources management. The validity or ground truth3 of EO prod- Practitioners, especially in developing ucts is also an important characteristic to be countries, would benefit from efforts to bridge taken into account when considering their use. the gap between scientific-academic and real- Yet, while numerous reports and publications world uses of RS technology. Factors to con- on hydrologic applications of remote sensing sider are the cost of implementation, financial discuss available tools (products and models), support, technical orientation, and definition 6  |  E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T of clear procedures and criteria to assess the Figure ES.1 Summary of Guidelines for Determining Whether to Use EO usability of RS products for decision making Products and planning conditioned by uncertainty Decision Tree Questions/Rationale (error estimates), accuracy (characterization of errors), precision (spatial and temporal res- Define the nature of the WRM problem to be solved? WRM problem Institutions? Relevant olution), timeliness, and validity of the data. A stakeholders? good understanding of the potential and limi- tations of in situ measurements and EO- derived data can inform the design of special Conditions of data network, Define the status of tools for specific purposes. Thus communica- existing observation data sharing possibility, existing tion between scientists, researchers, and prac- networks monitoring and models, etc.? titioners should be a two-way street. NOTES Do you need Adequacy of field No to use Earth observation data? observations? 1. The terms “remote sensing” and “Earth observation” are used interchangeably in this ­ publication. Do not Yes 2. Surface water, as used in this publication, refers use EO to water that is on the Earth’s surface, such as in a stream, river, lake, reservoir, wetland, or Can EO flooded area. potentially provide the EO potential use? No required data? 3. In remote sensing, ground truth refers to ­ information collected on location. Ground truth Yes allows image data to be related to real features and materials on the ground. The collection of EO product suitability? ­ ground-truth data enables calibration of RS data Spatial resolution? and aids in the interpretation and analysis of Temporal resolution what is being sensed. (revisit frequency)? Record length? Determine minimum required data characteristics In situ data requirements? Reliability? Accuracy? Maturity? Complexity? Can the EO Use EO product meet the data No Yes product(s) requirements? Note: WRM = water resources management; EO = Earth observation. E X E C U T I V E S U M M A R Y   |  7 PART I Water and Earth Observations in the World Bank Aleix Serrat-Capdevila, Danielle A. García Ramírez, and Noosha Tayebi OVERVIEW Water is a key driver of economic and social develop- for development. Water thus is a fundamental input ment as well as of poverty reduction. Growing econ- for sustainable economic growth. omies and populations need more water resources to While the World Bank has maintained its mission sustain economic activity and provide greater access of reducing global poverty through economic devel- to drinking water and improved sanitation, generate opment, the nature of the work it undertakes to fulfill renewable energy, or increase sustainable food pro- that mission has changed over the decades of its exis- duction. While water can be catalytic for economic tence. The scope of the problems that the Bank has to growth and development, too much or too little grapple with has expanded significantly, particularly water can also be a constraint if countries are unable in the realm of sustainable development, and today to prepare for climate-related hazards. includes challenges related to climate change, As water is present in most parts of the economy, resource depletion, natural disasters, and urbaniza- better water management is critical to helping peo- tion. These challenges require the collection and pro- ple, economies, and ecosystems to thrive, reduce cessing of data of much larger orders of magnitude poverty, and sustain prosperity. Better water than the poverty challenge alone requires. Under resources management also requires institutional these circumstances, the Bank has started tapping capacity and enabling environments in which stake- into remote sensing techniques as a means of acquir- holders participate in finding integrated solutions ing the extensive data needed to advance its goals.   9 CHAPTER 1 Key Global Water Challenges and the Role of Remote Sensing INTRODUCTION challenges have been grouped under represen- tative themes deemed especially relevant to Over the past century, water resources man- reducing poverty and promoting shared agement has increased in complexity as tech- growth—the overarching issues deserving nological advances of the Industrial Revolution attention. They are the scarcity and quality of have allowed humans to intervene in and water; the impacts of climate change and vari- modify the hydrologic cycle in unprecedented ability as well as changes not related to climate; ways. In many cases, this has been done with- the management of floods and droughts; the out acknowledging the environmental and management of the conjunctive use of surface social costs, without a long-term vision for water and groundwater; the complex links planning and management, and without any between water, energy, and food production; regulation or oversight. In addition, a good the need for alternative models of economic physical understanding of the impact of growth (such as green growth); financial chal- human intervention has often been lacking. lenges in the provision of water and sanitation Population and economic growth, as well as services; the need for better governance; and changes in land use and global dynamics, have improved cooperation in the management of pushed the use of water resources beyond the transboundary waters. limits of long-term sustainability in many This chapter summarizes the key chal- regions of the world. Several water resources lenges that undermine the optimal design, problems and challenges have taken or are use, and management of water and the socio- taking center stage in the sustainable manage- ecological systems that provide hydrologic ment arena. services beneficial to humans. These chal- The analysis of the World Bank project lenges are all related to the rational use of portfolio has identified several water-related water resources and the goals of environmen- challenges to sustainable development. These tal sustainability, economic efficiency, and   11 social equity. Each section deals with a spe- subsidies being given to groundwater-irri- cific challenge, reviewing the role of remote gated agriculture); (e) lack of institutional sensing (RS) observations in addressing that frameworks and enforcement and low pri- challenge and discussing possible approaches ority and visibility in political agendas; (f ) and ways to address it. lack of information and proper education regarding the impacts of unsustainable use and poor management practices; or (g) a WATER SCARCITY combination of all of the above. Effective management of demand can often sig- Water scarcity is a human-centered concept nificantly reduce water scarcity through resulting when overall demand for water improved technology and improved effi- exceeds supply. It often is a problem not only of ciency of water systems. At the same time, supply, but also of demand; it affects the Sahel very efficient water use and overallocation and Bangladesh, but also the Colorado River of existing water resources may lead to basin. Water scarcity arises when water supply brittle or nonresilient systems. and demand are out of balance: Water scarcity is sometimes so severe that there • Limited supply. The supply of water can be is not even enough water to satisfy basic human low due to either low hydrologic availabil- and animal needs, as is the case in some parts of ity (in arid and semiarid areas) or the lack the Sahel, where it often results in widespread of integrated systems to buffer seasonal hunger. In places where famine is tied directly or interannual variability (lack of access to local food production through the availability to water in reservoirs or aquifers or lack of rainfall and water in small water bodies of sustainable and reliable mechanisms to (springs, holes, watercourses), real-time moni- extract it). The overreliance on a single toring of rainfall and hydrology is essential to source of water may increase the vulner- prepare mitigation measures for the most ability of supply systems. In the Indus and exposed and vulnerable populations. Ganges basins, overreliance on a single Population growth, irrigation, and urban- source can be caused by the lack of proper ization are by far the most significant stress- management of conjunctive use, leading ors on water management. Urban centers in to a decline in aquifer levels, and by the developing nations are among the fastest- struggle of poor rural households to access growing areas (Africa has an annual growth water when pumping costs rise or deeper rate of 3.9 percent—the highest rate of urban wells are needed. The notion of water qual- population growth in the world). Growing ity should also be considered, as polluted demands for water, changing land use cover water is not a usable resource. and availability of water resources, as well as • Unlimited demand. In most instances, deteriorating water quality (due to new eco- water scarcity is due to (a) low agricultural nomic activities and poor sanitation) are efficiencies, as well as municipal, indus- becoming major challenges for sustainable trial, and conveyance inefficiencies; (b) no water resources management and, hence, for or lack of enforcement of limits on water sustainable economic growth. allocations or regulations on groundwater A good physical understanding of resource pumping or the allocation of resources (to dynamics through time and space and good prevent overallocation); (c) lack of proper near-real-time monitoring of hydrologic pricing of costs associated with water use; balances are essential to the proper manage- (d) subsidies on energy costs (the main ment of water as a scarce resource. This 12  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k understanding forms the basis for addressing • In large rivers, altimeter data of river sur- all of the water resource challenges described faces can be used to estimate flow in the in this chapter. Coping with climate and water upper parts of the basin and to predict variability requires good monitoring and downstream flows, issue flood warnings, accounting tools, to help users to prioritize and manage water allocation and operations and curb demand ahead of time in dry years (see Hossain et al. 2014). and to consume more wisely in wet years. A • Soil moisture observations may provide better quantification of hydrologic fluxes and insights into how much irrigation is needed storages will help to enhance water security— as well as help to correct missed events or and thus the health and well-being of false detections of satellite precipitation ­ populations—by providing more reliable products and to assess flood risk. information on water availability and use. Remote ­ sensing—also called Earth observa- • ET estimates may help water managers tion (EO)—allows the measurement of many to understand the dynamics of ground- hydrometeorological and environmental vari- water pumping in agricultural areas and ables as well as the identification of many the impact of water policies implemented types of land use cover. Some of the relevant and changes made in energy subsidies for variables are precipitation, soil moisture, ter- pumping. More generally, they can be used restrial water storage (including vadose zone in combination with ground data to under- water and groundwater), evapotranspiration stand efficiencies in water use, aimed at (ET), normalized difference vegetation index decreasing the rate of nonbeneficial and (NDVI), surface water elevation, and some increasing the productivity of beneficial water quality variables. Good monitoring of evapotranspiration (Wu et al. 2013). hydrometeorological fluxes and their parti- • The Satellite Water Monitoring and Flow tioning will enable integrated, flexible water Forecasting System for the Yellow River—a management approaches that take advantage Sino-Dutch cooperation project—is a good of diverse water resources and potential feed- example of integrated efforts using RS data backs between system components. (from hourly visual and thermal infrared Remote sensing can provide spatially dis- bands) to support river basin manage- tributed and timely observations that may ment, including energy and water balance, allow better forecasting and more efficient use drought monitoring and early warning, of water. For instance, and flow and flood forecasting (Rosema • Operational managers in snow-dominated et al. 2008). areas can use estimates of snow cover1 and snow water equivalent to know how much WATER QUALITY water is in storage and in what watersheds it is stored. Water quality issues are broad and differ • In poorly gauged basins with times of con- widely, depending on the specific context centration of many days, real-time satel- and setting. Water quality issues in surface lite estimates of precipitation and derived and groundwater can originate from (a) the streamflow forecasts can help manag- lack of or poor-quality water supply and san- ers to allocate water among users and to itation services, (b) land use practices, operate reservoirs more efficiently, taking (c) industrial activities, and (d) management into account how much water the river is issues involving natural contaminants. The expected to bring in the following days. deteriorating quality of rivers and aquifers C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  13 worldwide can also result from a combina- origin, it is human induced. Poor or deficient tion of these f­ actors. management can contaminate large ground- In many regions of the world, urban and water reservoirs by drawing in naturally occur- periurban aquifers are polluted due to the com- ring pollutants, including salt, fluoride, arsenic, bination of inadequate treatment of ­ sanitation and radioactivity. and wastewater and a range of e ­ conomic activi- Pharmaceuticals originating in wastewa- ties. Such contamination ­ exacerbates the lack ter are becoming one of the main challenges of access to clean water, which is itself a direct in developed countries. However, this issue is cause of poor health and malnutrition (diar- not receiving much attention in developing rhea is one of the main causes of malnutrition). countries. Those contaminants are difficult to Lack of access to drinking water also has an detect and expensive to quantify; moreover, impact on education (as children have to help their effect is not easily neutralized through their parents collect water) and poverty. Thus special treatment. There is still only limited water supply and sanitation are essential com- understanding of their impact on human and ponents of integrated approaches to reducing environmental health, although significant malnutrition and poverty. hormonal changes in some aquatic species Land use practices such as deforestation have been observed, among other effects. and farming can severely affect the quality of In addition to monitoring fluxes and stor- surface water and groundwater. While the age levels of water, remote sensing also offers disruption of vegetation cover and soil prac- the possibility of monitoring many parameters tices can have an impact on total dissolved of water quality. This makes it possible to fol- solids and turbidity, the use of fertilizers for low water quality in time and space across vast agriculture can lead to eutrophication and regions, significantly complementing costly hypoxia in surface water systems as well as to and limited field-point measurements. The severely polluted aquifers. The use of chemi- water quality variables that can be assessed cal pesticides can also cause imbalances in with remote sensing are temperature, chloro- natural ­trophic chains. phyll (an indicator of phytoplankton biomass, Industrial activities are a common source of trophic, and nutrient status and the most contamination, especially in countries that do widely used index of water quality and nutri- not have well-established regulatory and ent status globally), cyano-phycocyanin and enforcement mechanisms. Polluting activities cyano-phycoerythrin (indicators of cyanobac- may include mining and metallurgy, process- terial biomass, which are common in harmful ing and manufacturing industries, and the like. and toxic algal blooms), colored dissolved Water quality issues can also stem from organic matter (the optically measurable com- overexploitation and lack of management of ponent of dissolved organic matter in the water bodies with different natural character- water column, sometimes used as an indicator istics, sometimes bringing about saline intru- of organic matter and aquatic carbon), and sion in coastal aquifers or upconing of saline total suspended matter and non-algal particu- groundwater below freshwater aquifers. These late matter (important for assessing the ­quality water quality issues, in turn, lead to water scar- of drinking water and controlling the light in city and may compromise human health. For aquatic environments). instance, they are responsible for the naturally Direct applications of remote sensing for occurring high concentrations of arsenic in the management include the following: groundwater of floodplains and deltas or other “geogenic” elements such as fluoride or ura- • Monitoring water quality to assess the nium. While this kind of pollution has a natural impacts of river basin management policies 14  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k and land use practices on the environment monthly and annual levels (Anagnostopoulos and surface water. The spatial dimension et al. 2010), and consequent hydrologic simula- of this monitoring capability is important. tions alone are not fit for planning and design purposes and likewise unsuitable for forecasts • Monitoring the likelihood of algal blooms and reservoir operations (Kundzewicz and and other water quality threats to water Stakhiv 2010; Stakhiv 2010). However, GCMs supply systems. are designed and intended to project future Dekker and Hestir (2012) provide a good over- changes in climate and not to predict the view of the state of the art, reporting that the weather, two different tasks whose boundaries main impediment to making RS monitoring are often blurred in debates regarding GCM operational is the lack of bio-optical informa- accuracy. tion for parameterizing and validating remotely Should an attempt be made to use GCM sensed information on water quality. results to derive new flood frequencies at an hourly rate for specific basins? In the South- western United States, several studies have IMPACTS OF GLOBAL CHANGE found that, while average precipitation will decrease, precipitation extremes will increase While it is important to understand the under- (Domínguez et  al. 2012; Emori and Brown lying causes and future projections of climate 2005; Meehl et  al. 2007). While GCMs and variability and change, many of the impacts of regional climate models may provide useful climate change on hydrology and climate projections of changes in the frequency and ­ variability may be difficult to distinguish from magnitude of extreme events, the uncertainties ­ normal climate variability. The signal of may be too large to inform design investments. anthropogenic climate change is expected to At the same time, climate impact studies increase gradually, as is its impact, but its dis- have a more straightforward benefit for man- tribution across the globe will be uneven and agement and planning that is related to changes will depend on latitude and geography. in averages (for which climate model projec- Global climate models (GCMs), also known tions seem to be less uncertain) and the long- as general circulation models, project changes term availability of water. Consequently, if in precipitation and temperature. Beyond a changes in meteorological variables affect the research exercise, in what ways can impact mean states of hydrologic variables, efforts studies of climate variability and change should be directed at quantifying these average inform management and planning decisions? changes (using GCMs) and the envelope of What is the take-home message for decision uncertainty that contains them (using tradi- makers? Leaving aside the debate on how tional and new approaches), so that water GCM data should be used, studies of the managers and decision makers can adapt to impact of climate change on water resources changes in water availability (Serrat-Capdevila can potentially be applied to two realms: (a) and Mishra 2012). planning and design efforts to cope with While precipitation projections are uncertain extreme events in the upper end of the spec- and sometimes contradictory, there is little dis- trum (taking into account their return peri- agreement over the fact that the Earth’s atmo- ods) and (b) management aspects depending sphere is warming and that temperatures will on average water availability over the next continue to rise. One can easily expect that in decades. glacier-dominated regimes, flows could increase Even when aggregated over spatial scales, in the short term, which is associated with a GCM results contain a lot of uncertainty at the period of progressive glacier melting. In the C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  15 longer term, flows are expected to decrease floods, mean flows, and droughts is widely rec- drastically and become highly variable, having ognized. Dissecting the various contributions lost the regulating influence of a snowpack and to hydrologic change from natural climatic thus being subject to the vagaries of liquid drivers, as well as from land use change, vege- precipitation. tation cover change, and anthropogenic The change in hydrologic regimes can be increases in atmospheric concentrations of caused by changes in climate (global or local), carbon dioxide, is very challenging. changes in land cover and use, and direct Nonclimatic drivers of change in socioeco- human intervention in the hydrologic cycle logical systems such as the global economy, (dams, pumping). While climate change combined with natural climatic variability, impacts on hydrology have a low signal-to- can change social vulnerabilities and power noise ratio, changes in land cover and use can relationships. Edwards (2006) argues that severely affect the partitioning of rainfall into globalization has accentuated, rather than the different components of the hydrologic reduced, national and regional differences. cycle. Changes in the amount and partitioning Sub-Saharan Africa, in particular, has become of precipitation into evapotranspiration, infil- increasingly marginalized in terms of benefit- tration, and runoff are the main source of ing from the global economy (O’Brien and changes in hydrologic regimes. Leichenko 2000). Population growth, rural to Land cover change usually has an immediate urban migration, and land and ecosystem deg- impact on hydrologic responses. Villarini et al. radation and deforestation in developing (2009) show that changes in land use–related regions are all sources of change in hydrologic cover can have a significant influence on the and water demand. hydrologic response of a basin. During the Given the extent of human-induced global urbanization of their study basin, the 1,000- environmental change, current climate projec- year flood became the 10-year flood in a period tions, and the expected impacts on hydrology, of 50 years. This illustrates the fact that water resources, glaciers, and snow and land nonclimatic, anthropogenic changes often cover, Earth observations are sorely needed to stress water management more in the short monitor the dynamics of change. Land use term, with regard to the design of flood param- often changes in response to land and water eters. Nevertheless, changes in climate and management practices, which, in turn, are climate-induced vegetation are also likely to be influenced by global economic forces. a significant factor in regional water balances. Monitoring hydrometeorological and envi- Salas et al. (2012) broadly review scientific ronmental variables will help to document the efforts to characterize natural and anthropo- effects of global change. Observing, identify- genic sources of change and provide a picture ing, documenting, and understanding the of the combination of processes affecting the dynamics of change should be the foundation water cycle. In addition to the ones mentioned for the design and implementation of adapta- above, these include volcanic explosions and tion measures. As trends and changes in vari- large forest fires, which both influence the bal- ables caused by global changes will likely ance and composition of atmospheric energy manifest themselves unevenly in space and as well as ground cover. The El Niño Southern time, remote sensing is essential to comple- Oscillation, the Pacific Decadal Oscillation, the ment limited ground observations. Atlantic Multidecadal Oscillation, the Arctic Good monitoring allows (a) identifying Oscillation, and others influence climate at changes in meteorological variables and interannual and multidecadal intervals, and attributing causes, (b) assessing the impacts their effect on the magnitude and frequency of on hydrologic variables and dynamics, and 16  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k (c)  understanding observed changes in water utilities. Using advanced classification regional water budgets and water resources techniques, they created a multitemporal availability. (1984–2010) view of land cover change Changes in seasonal and annual snow cover, along the rapidly growing Tucson-Phoenix evolution in the length of glaciers, as well as urban corridor. These classifications created changes in vegetation, cloud cover, rainfall multitemporal maps of changing urban resi- ­ rates, soil moisture content, evapotranspira- dential, urban commercial-industrial, agri- tion, and other hydrologic variables of interest cultural, roads, bare ground, natural desert can best be quantified and spatially monitored cover, riparian, and water areas. These data with Earth observation. The combination of were subsequently integrated into an ongoing RS observations and climate impact studies analysis of urban and water policy and water using land surface and hydrologic models allocation within the region, making it easier offers a vantage point for informing adaptation to evaluate the correlation of water availabil- to climate and global change. ity and use, socioeconomic drivers, and direc- One of the main challenges in the analysis of tion and magnitude of changes in land use or climatic variability and trends is to reconcile cover. modern RS observations with long historical ground records. Most historical data sets of gauge precipitation that span a significant time EXTREME EVENTS: FLOODS period are not updated continually with near- AND DROUGHTS real-time observations. An example in this context is the Climate Research Unit time- Floods and droughts entail great economic series gauge precipitation data set, which pro- costs and loss of lives worldwide every year. vides a century-long record of monthly The magnitude and frequency of extreme precipitation from 1901 to 2009. While such events are expected to increase with intensifi- data sets can be used for climatologic analysis cation of the hydrologic cycle due to global of the historical period, they usually cannot be warming (IPCC 2007, 2014)2 and its regional used for near-real-time drought monitoring, as hydrologic impacts (Domínguez et  al. 2012; they are not updated continually. However, Serrat-Capdevila et al. 2013). Coping with vari- near-real-time, quasi-global precipitation ability requires different approaches to accom- products span a decade and a half at most. This modate events from different sides of the mismatch between long-term historical data spectrum. sets and near-real-time observations poses a A report by the American Water Resources challenge for assessing the impacts of climate Association, Proactive Flood and Drought Man- variability and change as well as for spatial agement (Dennis 2013), presents some lessons monitoring of drought. learned regarding how to manage extreme Remote sensing can be combined with events. The report recommends developing ground observations and socioeconomic anal- management strategies based on existing ysis of water use to provide insights that are hydrologic observations, data, and continued useful for planning, policy, and management; monitoring, taking into account the spatial Hartfield et al. (2014) provides a good illustra- analysis of exposure and vulnerability to floods tion. In a collaborative initiative with scholars and droughts. “Soft” ecosystem-based solu- and water management practitioners, they tions are also recommended, promoting eco- analyze the dynamics of water supply and system services and functions as part of sanitation infrastructure and urban growth comprehensive approaches, both for flood- using RS observations and information from plain reconnection (flood attenuation) and for C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  17 enhanced recharge purposes (drought risk Several approaches use RS data to monitor reduction through conjunctive use). the physical dimensions of drought, as illus- The report highlights the importance of trated by the following examples: “planning for the unexpected” by anticipating extreme events of magnitudes not yet seen. • The Surface Hydrology Group at Prince- While this recommendation is difficult to ton University operates the experimental translate into design investments, special Africa Drought Monitor (Sheffield et  al. methodologies such as the decision scaling 2014). It provides Africa-wide maps of pre- approach have been developed (Brown et  al. cipitation, temperature, wind speed, and 2012) and reported (García et al. 2014) to pro- hydrologic variables, as simulated by the vide a cost-efficient way of adapting to chang- variable infiltration capacity model over ing risks. These methodologies also highlight the entire continent at a 0.25° resolution.4 the need to involve all stakeholders—politi- The HyDros Lab from the University of cians, decision makers, the private sector, Oklahoma also provides global maps of agencies, and competing interest groups—in near-real-time streamflow and soil mois- an equitable and thoughtful process, to ensure ture estimates from their Coupled Rout- a coordinated and comprehensive, multiscale ing and Excess STorage model (Wang et al. approach. The shared vision planning 2011). 5 Both of these applications use the approach originated with the necessity to multisatellite precipitation analysis prod- cope with drought and was developed by plan- uct of the Tropical Rainfall Measuring ning practitioners who had to address water Mission as input on precipitation. scarcity issues and planning challenges in • Going from the global to the basin scale, the their professional capacity. More about the Hydrology and Water Resources Depart- shared vision planning approach can be found ment of the University of Arizona, in part- in Cardwell, Langsdale, and Stephenson nership with the National Aeronautics and (2009). Space Administration (NASA) SERVIR Pro- With intensification of the hydrologic cycle, gram and the International Center for Inte- extreme events are expected to become more grated Water Resources Management of the intense and more frequent, although estimates United Nations Educational, Scientific, and of future intensities and frequencies are Cultural Organization (UNESCO), has been fraught with large uncertainties. Diversifying developing experimental monitors and 7- to resources and approaches, building flexibility 10-day streamflow forecasts in watersheds into the system, as well as conserving unused of four international African basins: the buffers3 may be the key to ensuring resilience Mara (Kenya, Tanzania), the Upper Zam- to extreme events and to minimizing their eco- bezi (Angola, Namibia, ­Zambia), the Tekeze nomic costs. Full allocation and maximal use of ­ enegal (Guinea, (Eritrea, Ethiopia), and the S water resources, without considering buffers Mali, Mauritania, Senegal). These efforts and redundancies in the system, could, in the represent a multimodel and multiproduct short term, lead to optimal but “brittle” solu- approach using state-of-the-art climate tions in the event of shocks to the system. projections to develop streamflow forecasts Monitoring current trends and hydrologic con- and assess climate change impacts on water ditions is essential to choosing appropriate availability for the current century.6 management actions in time and to being able to share key information with users and stake- • Satellite rainfall estimates can also be used holders. RS measurements can be very useful to derive a grid cell-level water requirement in this context. satisfaction index (WRSI)—the percentage 18  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k ratio of actual crop evapotranspiration for the application to be beneficial. Flood over a reference-crop evapotranspiration warning and alert systems focus on the magni- (non-water limited). The WRSI can be a tude of peak flows. Accurately forecasting peak good indicator of yield reduction due to flows at an acceptable level of precision water limitation. The Famine Early Warn- requires rainfall estimates of a relatively high ing System Network of the U.S. Agency spatial and temporal resolution (Li et al. 2008). for International Development (USAID) As the extent of damage and loss of life depend combines monitoring of rainfall, using on the performance of a flood warning system, the rainfall estimation algorithm version the levels of acceptable uncertainty are much 2 (RFE 2.0) of the National Oceanic and lower than in other applications, such as reser- Atmospheric Administration (NOAA), voir operations (which are mostly concerned and crop production, using the WRSI, with water volumes). with socioeconomic variables (for exam- A forecast system has two types of predic- ple, prices) and a livelihoods approach to tion errors: type I (missed predictions) and understand the strategies that people use type II (false alerts). While type I errors have a to meet their basic needs. This provides short-term impact (flood damage or loss of life), insights into which population groups are type II errors reduce the credibility of the fore- most vulnerable to food insecurity, how cast. The fraction of the target population that long they remain vulnerable, and what the will respond to a flood warning or alert depends best mitigation approaches are (Verdin on a system’s past performance. Thus if an RS et al. 2005). system generates too many false alarms, an alarm from a correct prediction would likely be • Earth observation can be used to moni- ignored, with potentially catastrophic effects. tor small water holes, which are especially RS products can be used in different appli- relevant to rural livelihoods, pastoralists cations aimed at informing flood warnings. For and their herds, and wildlife migrations. A example, NASA-funded project uses a water-balance approach to model water levels of pools in • In the face of significant uncertainties in closed basins.7 The European Space A­ gency’s globally available near-real-time satellite TIGER program uses Landsat visual imagery rainfall products, the reliability of satellite- at 30-meter resolution to monitor changes in based forecasts of rainfall-runoff floods the size of “small” water bodies. may vary with the setting, product, and sea- • Estimates of the NDVI can also be used to son and may not always be, at present, suffi- derive the vegetation health index, which cient for real-world flood warning and alert Yan et al. (2014) find to be a more accurate systems (Serrat-Capdevila, Valdes, and detector of agricultural droughts for irrigated Stakhiv 2013). The need for information areas than the standard precipitation index. on changes in rainfall-runoff to provide streamflow forecasts adds an additional The use of satellite estimates for flood fore- layer of uncertainty and can magnify errors casting applications and flood alert systems is in estimating the peak magnitude of the perhaps the most complex, with regard to the flood (Nikopoulos et al. 2010). The Hydro- rainfall-runoff transformations involved (mag- Estimator of Central America Flash Flood nifying the rainfall errors in peak flow), the Guidance produces a flash flood threat hydrodynamic modeling that will determine index using the “excess amount of rainfall flood levels at specific locations, and the level for a three-hour period over what is needed of precision and accuracy that will be needed to cause bank-full flows in small streams.”8 C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  19 • Soil moisture estimates can also be useful Meteorological Organization include the flood for predicting floods, as they provide infor- forecasting Distributed Model Intercompari- mation on the “wetness” of a basin and son Project, the development of a framework thus the partitioning of rainwater between for assessing the efficiency of flood forecasting infiltration and runoff, depending on the services, and the establishment of regional saturation level of the headwaters of the flash flood guidance systems using integrated watershed. In addition, soil moisture can observations and model outputs. In a collab- help to correct for errors produced using orative effort of the World Meteorological satellite precipitation products such as Organization, NOAA, USAID, and the Hydro- false detections and missed events. logic Research Centre, integrated satellite observations, in situ observations, and models • An effective and innovative approach to are used to implement flash flood guidance flood prediction in large rivers is the use of systems in streams in many regions and trans- surface altimeter measurements. Tennes- boundary basins. Several satellite-based flood see Technological University, in collabora- prediction and monitoring systems are tion with the Institute of Water Modeling reported to be almost operational (Lawford (Bangladesh), the NASA-USAID SERVIR 2014), such as the Integrated Flood Analysis Program, and the International Centre System, provided by UNESCO’s International for Integrated Mountain Development Centre for Water Hazard and Risk Manage- (Nepal), developed an 8-day flood forecast- ment, the Global Flood Alert System, provided ing system using river surface altimeter by the International Flood Network, and measurements from the Jason-2 satellite Global Flood and Landslide Monitoring, pro- in the Ganges-Brahmaputra-Meghna sys- vided by NASA and the Goddard Space Flight tem, which drains the Himalayas through Center (Lawford 2014). Bangladesh (Hossain et al. 2014). Measure- For these systems to be truly useful for ments of river surface levels upstream and water managers and decision makers, future a hydrodynamic model (HEC-Ras) are research and applications will have to consider used to predict how the observed water the reliability of flood warning systems in addi- levels upstream will propagate to areas tion to the estimated uncertainty in streamflow downstream. Forecast validation efforts forecasts. Reflecting these complexities, Cen- have shown a root mean square error of tral America Flash Flood Guidance labels its 0.7 meter, with lead times up to 10 days at flash flood threat index as “experimental” and the India-Bangladesh border (with errors “not for operational use.” Because of the short ranging up to 0.5 meter and exception- time of concentration and the latency of rain- ally even over 1 meter). Given the much fall products, the window for early warning larger changes in river levels and the fact action is limited in small watersheds. that there is currently no alternative for an 8-day lead time forecast in Bangladesh, these errors are considered acceptable CONJUNCTIVE USE OF SURFACE given that only RS data are used for the WATER AND GROUNDWATER forecasts. Efforts are under way to improve the forecasts by using additional altimeter The adoption of pumps and rural electrifica- measurements from other satellite sensors tion in the mid-twentieth century made it (F. Hossain, personal communication).9 ­ possible to tap vast groundwater resources, Regarding the role of international organiza- opening up a domain that institutions had not tions, ongoing activities related to the World governed before. This led to a challenging issue 20  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k regarding the management of surface water on the water supply, something usually based and groundwater, which has not yet been fully on models. Studies of capture through model- resolved. The social organization—that is, the ing are documented in the Colorado River institutions involved in managing the conjunc- delta (Maddock, Serrat-Capdevila, and Valdes tive use of surface water and groundwater—is 2010) and the San Pedro basin in Arizona- still seeking ways to regulate the use of this Sonora (Leake, Pool, and Leenhouts 2008). technological innovation. Capture maps display the impact that pumping Efficient conjunctive use of surface water in any given area would have on a specific and groundwater provides an opportunity to water body, such as a nearby river. By showing handle variable surface flows and work toward in a spatially explicit manner the degree to sustainability in regions with severely overex- which pumping will intercept water that ploited aquifers, balancing water withdrawals would otherwise contribute to baseflows in the with managed recharge. It requires local river, capture maps can help managers to research on recharge processes and the poten- choose pumping locations that minimize the tial for managed aquifer recharge,10 storage, impact of groundwater abstractions or to and recovery. Research is also needed on determine feasible pumping rates for a partic- locally adapted policies and approaches to ular fixed location. managing the system and achieving a sustain- Taking a river basin perspective, investments able pumping yield, supported by a conjunc- in managing conjunctive use should aim to tive use strategy and implementation planning. reshape infrastructure at all scales to promote The robustness of integrated conjunctive use groundwater recharge and to build manage- water management plans can then be tested ment capacity (monitoring systems, institutional with a range of climate variability and climate adaptations, best practices, and greater incen- change scenarios to account for uncertainty. tive compatibility) around improved groundwa- Examples exist in Western and Southern ter governance frameworks and p ­ articipatory India, where, faced with dropping ground- approaches (Shah, Darghouth, and Dinar 2006; water levels and aquifer mining, local com- Wijnen et al. 2012). Aquifer m ­ anagement orga- munities and governments are building nizations are needed, perhaps embedded or water harvesting and recharge structures working closely with basin c ­ ouncils and gather- aimed at increasing groundwater recharge ing representatives from all relevant decision- during the rainy season, when surface water making and user ­organizations. Governance- and is abundant. These measures not only repre- incentive-based approaches to overcome policy sent a successful flood mitigation strategy, challenges are summarized later in this chapter. but also improve water security for human To achieve sustainable management in con- and agricultural consumption, protecting junctive use of water, several objectives must be and stabilizing rural livelihoods against pursued simultaneously: introducing managed drought and decreasing groundwater levels aquifer recharge, improving efficiency in water (Shah 2003). use, and adopting water demand–curbing In conjunctive use management, the con- measures to conserve water. The need for inte- ­ cept of capture is important: pumping water grated approaches is illustrated by case studies from an aquifer system that is hydraulically in Peru’s Pacific Coast valleys such as Ica, where connected to a surface water system will even- agribusiness export companies use high-­ tually deplete the surface water system. Such efficiency drip irrigation—farming a larger area depletion is a form of surface water capture. It than before with the same amount of water and is essential that water managers understand eliminating return flows to the aquifer—without the concept of capture and estimate its effects reducing their demand for water from the C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  21 acquifer (Garduño and Foster 2010). This issue example, kilograms of grain) produced per is common in many other settings such as the cubic meter of water consumed to grow Hai basin in China (Wijnen et al. 2012), where a that crop. monitoring system was developed to under- • The impacts of irrigation policies, energy stand the consumptive use of water using subsidies, and other policies on agricul- ­satellite-based ET estimates. It is also important tural water use can also be monitored to consider water quality issues in conjunctive through ET observations. use management schemes, in order to minimize potential threats, such as recharging contami- • Estimates of soil moisture and evapotrans- nant loads over time via infiltration from agri- piration can be used to inform both farm- culture, poor onsite sanitation, upconing of ers and irrigation managers of the state saline groundwater into the freshwater aquifer, of their fields when water conservation is or other management-induced water quality urgent. While satellite-derived ET esti- issues (arsenic, fluoride, and radioactive mates make it possible to calculate the net contamination). loss of water to the atmosphere, remotely Good monitoring of aquifer dynamics and sensed soil moisture provides information water use patterns will facilitate the develop- about the water content of the upper 1 to ment of solutions for sustainable management. 5 centimeters of soil. By combining groundwater monitoring with • Finally, while RS measurements of changes capacity building of local communities (for in the gravity field—due to changes in ter- example, participatory monitoring approaches restrial water content—have a very large in India), farmers and other users can learn footprint (the Gravity Recovery and Cli- how to manage their groundwater resources. mate Experiment, GRACE, has a 300-kilo- The sharing of collective measurements can meter footprint), regional aquifer-level also serve as a platform for transparency, out- estimates may be used to constrain reach, and discussion. regional groundwater models. In this way, Conjunctive water use management would GRACE data on aquifer levels can be rec- benefit from RS applications such as the onciled with local (ground) well measure- following: ments to provide insight into what areas • Estimation of evapotranspiration can help would benefit most from managed aquifer to quantify irrigated extensions and the recharge. consumptive use of water, making it easier • In large river systems with significant agri- to maintain an accurate inventory of the culture, such as the Indo-Gangetic plains, number of water users and the volume of altimeter data on river levels upstream can water used. Similarly, it can contribute to help managers and operators to maximize an understanding of water use patterns in artificial recharge efforts and optimize time and space. Evapotranspiration can the allocation of water among uses and provide information about irrigated and rechargeable aquifers. nonirrigated areas. In the United States, it is monitored so that insurance claims filed for failed harvests purportedly caused by lack THE FOOD-WATER-ENERGY NEXUS of access to irrigation water may be verified. • Estimates of evapotranspiration can be Integrated assessments of resource use are often used to quantify crop water productiv- lacking, and analysis frameworks are rarely ity—the amount of marketable crops (for multidisciplinary. How can well-informed, 22  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k cross-sector, and integrated decisions be made water availability, energy demand, as well as when academic research on integrated assess- production, which will ripple through the sys- ments is still in progress, relatively recent, or tem in ways that are difficult to predict even lacking? Food requires water for agricul- (adapted from Rodríguez, van den Berg, and ture and energy for growth and transportation. McMahon 2012). At the regional and local lev- Water requires energy to be accessed, treated, els, the ensuing changes in resource use, eco- conveyed, and delivered. In addition, energy nomic activities, and land use will be shaped by prices significantly affect the cost of building global economic forces (energy prices, market and maintaining water infrastructure, partly demands), water availability, and climate through the production and delivery cost of change impacts. O’Brien and Leichenko (2000) inputs (Rodríguez, van den Berg, and McMahon dub the combined effects of globalization and 2012). In some cases, such as reverse osmosis local manifestations of climate change impacts desalination, energy prices directly influence “double exposure.” production costs because of the high energy At a higher level, integrated assessments are requirements of operation. essential to understanding coupled dynamics. Water is often used to generate energy, and The analysis of social metabolism takes a mul- it is expected that renewable energy sources tidisciplinary look at how society combines (biofuels) will demand even more water. water, energy, and other resources to produce Extractive fossil fuel activities in the energy goods or promote social well-being, as well as sector can severely compromise water quality. how it grows and maintains itself. Such inte- Biofuel production competes with food pro- grative, quantitative assessments are useful for duction for water and space, raising the price comparing competing future scenarios and of staple foods. Subsidies in the energy sector comprehensive multisector plans for a region. meant to lower the cost of food production Near-real-time monitoring of hydrometeo- often undermine the sustainability of ground- rological variables with remote sensing can water use. In contrast, rising global energy help to inform short-term decisions on water prices (due to economic and population for food security and crop production, as well growth) and continued water scarcity will lead as for hydropower generation. The Famine to higher food prices, which may push more Early Warning System described earlier illus- people below the poverty line. Taking into trates the use of a satellite precipitation prod- account current practices and technology and uct to monitor food security and so does the a global population of 9 billion by 2050, Hanjra CropWatch System in China (appendix B)—a and Qureshi (2010) estimate a current water global crop monitoring system using a wide gap for food production of 3,300 cubic kilome- array of RS data for applications such as crop ters per year. condition monitoring, drought monitoring, Without returning to the Malthusian versus crop acreage estimation, crop yield estimation, Cornucopian debate regarding the role that grain production estimation, and cropping technology and markets may play in solving index monitoring (Wu et al. 2014). the sustainability challenge of our time, it is The Food Early Solutions for Africa Micro- clear that investments targeting improvements Insurance Project of the Netherlands Ministry in irrigation infrastructure and water produc- of Development Cooperation uses visual and tivity can help to meet the demand for water thermal infrared Meteosat imagery to monitor for food production (Falkenmark and Molden water balance, focusing on precipitation and 2008). In addition, projected increases in tem- relative evapotranspiration. Having found that perature and changes in precipitation patterns, the water balance (precipitation minus relative due to climate change, will lead to changes in evapotranspiration) fits well with reported C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  23 discharges and that the relative evapotranspira- become the second-largest economy in the tion is more closely related and proportional to world, it is estimated that the costs of environ- reported crop yields in pilot areas, two mental impacts represent about 9 percent of Meteosat-based insurance indexes have been its gross domestic product. This fact threatens proposed: the “dekad relative evapotranspira- both its competitiveness and its welfare. tion” (an agricultural drought index) and the Green growth is an approach to economic “dekad cold cloud duration” (an excessive pre- growth that incorporates the following ele- cipitation index). Numerous pilot projects with ments: sustainable natural resources man- insurance companies in Africa have shown that agement; more resilient communities, based these two indexes provide an excellent alterna- on the adoption of eco-friendly practices tive to precipitation-based approaches, per- (permaculture, soil and water conservation) forming “as good as or even better” (Rosema and designs (comprehensive planning); et al. 2014). investments in environmentally integrated Remote sensing of evapotranspiration can infrastructure, green technologies, and inno- provide valuable information on the impacts vation; and the gradual introduction of new that changes in energy subsidies or prices have pricing schemes for resource use that fully on water use and could even inform allocation account for externalities. decisions in situations of energy scarcity due to Investing in environmentally sustainable competing needs for water. growth is good for long-term economic pros- An issue that could be explored further is pects because it increases natural capital the use of satellite-derived measures of evapo- (through  ­better ­management of scarce resour transpiration and soil moisture to identify ces); raises labor productivity by  improving regional soil and water management practices health; increases physical capital by better that have certain desirable characteristics or managing natural risks (ecosystems provide build on positive feedbacks, thereby promoting regulatory and protective services, including sustainable farming livelihoods. flood protection, coastal storm protection, and infiltration and soil aquifer storage); increases the efficiency of resource use; stimulates the GREEN GROWTH AND THE economy in the short term (through green ENVIRONMENT investments); accelerates the development and dissemination of innovation; and creates Models of economic growth have often ignored knowledge spillovers. the role of natural capital as a factor of produc- Infrastructure is a central issue in develop- tion and limited the assessment of economic ing countries, first because their infrastruc- growth to physical capital (infrastructure, ture needs are acute and second because machinery, buildings, hardware), labor (popu- infrastructure policies are central to support- lation, education, health), and productivity ing green growth and alleviating water scar- (technology, efficiency). Yet unsustainable city. Since infrastructure decisions have a management of the environment ultimately high potential for “regret” (as they are long- results in the destruction and full depreciation lived), the current infrastructure gap offers of natural capital, with negative repercussions developing countries the opportunity to for output in the short or medium term. If nat- “build right.” ural capital is considered a factor of produc- Remote sensing can help to improve the tion, environmental policies are a beneficial quality and effectiveness of water infrastruc- investment. While China has grown at an ture planning and project design by comple- annual rate of 10 percent in recent years and menting information provided by in situ 24  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k measurements with available satellite-derived Nevertheless, assistance fell 3 percent across products. Remote sensing can be particularly all water sectors in 2011, the largest drop valuable considering the importance of moni- since 1997 (Rodríguez, van den Berg, and toring the dynamics of a complex and fast- ­McMahon 2012).11 changing world and evaluating the impact of Estimates of investment needs vary widely— human interventions, new infrastructure, pol- from US$103 billion per year for all developing icies, and regulations on the environment: countries until 2015 (Yepes 2008) to US$22 bil- lion per year just for Africa, to ensure that the • Monitoring changes in land use cover after continent can reach the Millennium Develop- implementing green growth projects or ment Goal by the year 2020 (Foster and adopting certain policies as well as moni- Briceño-Garmendia 2010). Funding sources for toring changes in hydrologic variables developing countries include (a) public contri- (such as soil moisture and evapotranspira- butions in the form of official development tion in the wake of soil conservation inter- assistance (grants, low-interest loans, technical ventions, for example) can provide insights assistance from donors and international finan- into the extent to which project objectives cial institutions) and contributions from local have been met. governments funded by tax revenues; (b) pri- • Similarly, remote sensing can provide vate contributions, which have been halved valuable information for spatial planning during the last two decades (as the financial cri- purposes. sis lowered the tolerance for risky investments and a paradigm shift occurred toward more investment money but smaller investments); FINANCIAL ISSUES and (c) household contributions (toilets, septic tanks), which are not well documented but are Water infrastructure that ensures a reliable estimated to contribute one-third of total water water supply and well-functioning sanitation sector investments in Sub-Saharan Africa. and irrigation services is a cornerstone of Various factors hamper the effectiveness of sustainable development and poverty allevia- investments, such as the fluctuation of over- tion, providing food and livelihood security seas development assistance from year to year, in addition to healthy living conditions. The inadequate execution of or failure to execute infrastructure gap is the difference between budgeted funds by local governments, decen- current levels of spending in water-related tralization of responsibilities coupled with infrastructure and service provision and the inadequate follow-up of decentralized funding spending levels required to meet the devel- and capabilities, and strong urban-rural dis- opment targets. This gap has widened since parities in the focus of investments. In addi- the 1990s, because of the financial crisis, pop- tion, reported risks and ratings of water ulation growth, and deficits in the opera- provisioning utilities often do not reflect the tional budget for water provisioning services sustainability of the water supply or future in most developing countries. International hydrologic variability. For example, utility rat- financial institutions have attempted to offset ings in the United States are based on the vol- these deficits through development assis- ume of water sales (short-term revenues), even tance to the water sector. The World Bank if groundwater is currently being mined Group committed more than US$100 billion beyond sustainable levels. Climate risks and in 2009 to maintain and expand existing uncertainty should be factored into long-term infrastructure in countries that had cut their adaptation plans, allowing for adequate financ- service budgets during previous crises. ing and pricing of services. C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  25 The main financial issues in the water sec- of data require some funding. In general, a tor relate to infrastructure for water provi- good observational system should ensure sioning and sanitation. To help to close the proper monitoring and reporting of impacts on funding gap, Rodríguez, van den Berg, and water resources and the environment from McMahon (2012) propose a “reform cycle” specific investments, measures, policies, and with five circular and iterative components, initiatives. This should be reflected in better, incorporating the needs of all stakeholders. more sustainable management and provide a They suggest the following five components: credibility asset for securing investment from (a) service providers deliver services more public and private entities. Moreover, invest- efficiently (by reducing nonrevenue water ser- ments in capacity building (for in-house exper- vices, improving billing and collections, and tise of government agencies) usually provide a carefully choosing technology for water ser- high return on investment. vices); (b) pricing of water is based on sound cost-recovery models (by covering the full financial cost of the services provided to guar- INSTITUTIONAL FRAMEWORKS antee their sustainability, providing incentives AND GOVERNANCE ISSUES to use water more efficiently, giving financial compensation for ecosystem services, and As new technologies progressively shape the introducing tariff reforms); (c) governments way in which humans interact with the envi- improve public expenditure (by clarifying who ronment, new socioeconomic structures and pays for what costs, strengthening the commit- arrangements emerge and evolve in response ment to the water sector, and subsidizing users to the need to manage and regulate those inter- and service providers cautiously); (d) all stake- actions. The observed management disconnect holders develop sound sector governance (by between surface water and groundwater is a achieving political stability, the rule of law, gov- good example of how institutional frameworks ernment effectiveness, regulatory quality, pub- have struggled to keep pace with technological lic accountability, and a clear definition of the advances. Since the adoption of groundwater mandate of the main actors—policy, manage- pumping, the institutional organization ment, infrastructure development, service regarding management of the conjunctive use provision, financing, and regulation); and of surface water and groundwater has yet to be (e)  governments and donors leverage resolved. Society still needs to regulate the use resources to attract private investment (which of this technological innovation in an inte- demands solvent utilities, sound governance grated manner. Transparent, adaptive manage- structures, and local capacity to plan and exe- ment techniques are ideal under current cute budgets). circumstances, as they allow people the flexi- Good monitoring capability can facilitate bility to react to changes. However, success- the communication with users and stakehold- fully integrating existing and new policies is ers, for instance, when making the case for not easy. compensating ecosystem services and cover- Attempts to introduce sustainable policies ing the costs of environmental degradation are often hampered by the trade-off between rather than leaving the tab for future genera- the short-term benefits of (over)exploitation tions. Better governance also requires more and the long-term benefits of environmentally transparent information and monitoring. Very sound policies, the unequal distribution of large amounts of RS data that may be useful in power and influence among various user this context are freely available and open to the groups, the relatively short terms of political public; only the processing and interpretation office (four to five years), and the political 26  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k disadvantage inherent in defending long-term they need to make those decisions, hydromete- issues (requiring regulation, demand manage- orological agencies should explain how their ment, and water use and service charges) ver- monitoring, forecasts, and assessments can be sus addressing short-term needs with limited operationalized in the decision-making pro- budgets and capacity. cess. The climate adaptation strategies and Flexibility is an important aspect of a good, plans to be developed and their exact charac- adaptive management practice. Institutions teristics will depend on past and current obser- should be able to change past policies based on vations of water and environmental systems. their observed impacts on the system. In this These strategies and plans will have to recon- feedback loop linking the latest observations cile development goals with specific interven- with the next decision-making steps, close col- tions. Sustainable societies are those that laboration is vital between those who monitor, reinvest in knowledge and understanding; study, and interpret the behavior of the system capacity building and training should always be and those who ultimately make the decisions. a central focus of water security (Serrat- Traditionally, these two groups have worked Capdevila and Mishra 2012). for different institutions, and communication Understanding the dynamics of power in a between them has not necessarily been fluid. governance system—and the interactions That is why an adaptive management mecha- between political and economic processes that nism is needed that will foster the development shape such dynamics—is essential to the design of new organisms and institutional strategies and implementation of development strategies capable of putting new knowledge to practical and policies (World Bank 2009). Political use. For management to be truly adaptive, both economy can be defined in practical terms as the policies and the institutions must be flexi- “the way in which different stakeholders influ- ble (Serrat-Capdevila et al. 2009, 2014). ence policy, governance, and resource alloca- Especially with Earth observations, there is tion and thereby influence outcomes” (Wijnen insufficient capacity to close the feedback loop et  al. 2012). In the management of common- between system monitoring, modeling and sci- pool resources where abstractions by one user entific analysis, stakeholder participation, and benefit the individual but diminish the pool decision making. Democratic societies are available to others, monitoring and informa- striving toward open and transparent water tion transparency, to a large extent made pos- governance systems, supported by participa- sible by Earth observation, are essential to tory mechanisms. The best approaches are enabling both top-down (government control, those that manage to integrate structured pub- privatization) and bottom-up (collective man- lic participation, planning and management agement of common-pool resources) manage- processes, and strong scientific input, thereby ment approaches (Hardin 1968; Ostrom contributing to science-based decision mak- et al. 1999). A good understanding of the power ing. Thus both public and private institutions, relations between users, user groups, agencies as well as the general public, should be from multiple sectors, politicians, and the involved. National hydrometeorological agen- ­ voting public should steer the design of policy cies should, among other things, be responsi- and governance approaches that are likely to ble for analyzing and interpreting up-to-date work best in a specific setting. observational records that are linked directly Remote sensing can support governance to water management decision-making needs. and institutional frameworks. For example, While decision makers should spell out their criteria for making specific climate- and water- • RS data can shed light on issues of data trans- sensitive decisions as well as the information parency and information control, preventing C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  27 the pursuit of hidden agendas and informing consideration to the access to and allocation other political economy challenges. of water resources. The challenges of transboundary basins are • RS data may help financiers and donors to often conditioned by a lack of data sharing and determine whether specific decisions are thus a lack of reliable information for all the based on sound science and information or riparian countries regarding the hydrologic guided by other interests. Even if the lat- state of the basin as a whole, beyond its bor- ter is true, building capabilities to use and ders. In addition, the parties involved may dis- interpret RS data is likely to be beneficial pute the veracity of the data and information in the long term, as the political economy shared. context may evolve (see the example in Subramanian, Brown, and Wolf (2012) Wijnen et al. 2012). review five case studies of transboundary col- laboration from the perspective of country TRANSBOUNDARY ISSUES decision makers, providing a better under- standing of the political economy of coopera- Transboundary basins cover more than half of tion. They classify perceived risk in five the world’s land surface, and their management categories and propose seven risk-reduction can give rise to conflicts. A good characteriza- approaches, the first of which is to expand tion of a resource is the basis of international information, knowledge, and skills. This agreements on its use as a shared resource, approach should also involve observation and more specifically, whether the sharing of sur- analysis to meet gaps in knowledge, as well as face water or groundwater is addressed. Trans- training and capacity building. boundary agreements on surface water usually Remote sensing is a tool for indirectly mea- revolve around the delivery or release of suring hydrologic states in a (transboundary) streamflows at particular border locations basin beyond a nation’s borders and for verify- (where watercourses cross borders) over a ing shared information. For example, period of time. International coordination of • Remote sensing can help all parties to transboundary groundwater management is understand the resource dynamics (rivers more recent and perhaps more complex. Inter- and aquifer systems) because data shar- national water treaties after World War II ing and proper monitoring are the basis began to include uses not related to navigation, for successful collaborative agreements such as flood control, hydropower develop- and management efforts. Satellite-derived ment, water quality management, and water products can address the need of trans- allocation. Historically, the most challenging boundary agreements for periodic moni- element of a deal has been getting countries to toring and data sharing. agree on the allocation of water quantities between the appropriate co-basins. • Remote sensing can also be used to predict In addition to transboundary water issues flows because it provides information from involving sovereign nations, open conflicts, parts of the basin that lie outside a nation’s disputed territories, and other geopolitical borders. The Institute of Water Model- issues often have a significant water dimen- ing’s flood forecasts in Bangladesh using sion. Examples of this are the conflict altimeter data is a perfect example. Altim- between the West Bank and Gaza and Israel eter measurements from satellite Jason-2 and the conflict between Sudan and South provide surface water levels from river Sudan. Any serious attempt to resolve con- reaches in India that are 600 kilometers flicts in these contexts must give due upstream from the Bangladeshi borders. 28  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k This allows extending the lead time from 10. Managed aquifer recharge involves building infra- a scarce three days (if observations were structure or modifying the landscape to enhance groundwater recharge. done at the border, as in the past) to eight 11. Rodríguez, van den Berg, and McMahon (2012) days (Hossain et al. 2014). Many other RS describe the current state of water financing and applications can also be useful in trans- propose a set of approaches to improve efficiency boundary settings. in financing and reach development targets. NOTES REFERENCES 1. Throughout this publication, the terms Anagnostopoulos, G. G., D. Koutsoyiannis, A. Christo- “snow cover” and “snow depth” are used fides, A. Efstratiadis, and N. Mamassis. 2010. “A interchangeably. Comparison of Local and Aggregated Climate 2. Although the Intergovernmental Panel on Model Outputs with Observed Data.” Hydrological Climate Change seemed less sure about this Sciences Journal 55 (7): 1094–10. in 2014 than in 2007, the overall meaning still Brown, C., Y. Ghile, M. Laverty, and K. Li. 2012. “Deci- holds true: “Recent detection of increasing sion Scaling: Linking Bottom-Up Vulnerability trends in extreme precipitation and discharge Analysis with Climate Projections in the Water in some catchments implies greater risks of Sector.” Water Resources Research 48 (9): W09537. flooding at regional scale (medium confidence)” Cardwell, H., S. Langsdale, and K. Stephenson. 2009. (IPCC 2014, 8), and “It is very likely that heat “The Shared Vision Planning Primer: How to waves will occur more often and last longer and Incorporate Computer Aided Dispute Resolu- that extreme precipitation events will become tion in Water Resources Planning.” Shared Vision more intense and frequent in many regions” Planning, Institute for Water Resources, U.S. (IPCC 2014, 10). Army Corps of Engineers. http://www.sharedvi- 3. Buffers can help to cope with variability and sionplanning.us/resReference.cfm. change by building redundancy in systems. Dekker, A. G., and E. L. Hestir. 2012. Evaluating the For example, in times of drought, unused Feasibility of Systematic Inland Water Quality water resources can be tapped that are not Monitoring with Satellite Remote Sensing. Water fully utilized during periods of no drought; the for a Healthy Country National Research Flag- consumption of water for some uses can be ship. Victoria: Commonwealth Scientific and decreased if needed or can be shut off for a cer- Industrial Research Organisation. tain period of time; and the users of reclaimed Dennis, L. 2013. Proactive Flood and Drought Manage- water can be changed in times of drought. In ment: A Selection of Applied Strategies and Lessons addition, because droughts are likely to have Learned from around the United States, edited by different impacts on different types of water B. Bateman, W. Wright, and D. Duke. Middleburg, resources, the ability to switch the system from VA: American Water Resources Association Policy one set of tapped resources to another builds Committee. resilience. In the case of floods, the existence Domínguez, F., E. Rivera, D. P. Lettenmaier, and C. L. of floodable areas, storage pools, soil and water Castro. 2012. “Changes in Winter Precipitation conservation practices, as well as recharge infra- Extremes for the Western United States under a structures will help to slow down water flows, Warmer Climate as Simulated by Regional Cli- increase recharge, and lower the magnitude of mate Models.” Paper submitted to Proceedings of flood peaks. the National Academy of Sciences of the United 4. This interactive application can be found at States of America. http://hydrology.princeton.edu/monitor/. Edwards, P. 2006. “Examining Inequality: Who Really 5. See http://hydro/ou.edu/. Benefits from Global Growth?” World Develop- 6. The near-real-time forecasts can be found at ment 34 (10): 1667–95. http://www.swaat.arizona.edu. Emori, S., and S. Brown. 2005. “Dynamic and Ther- 7. See http://watermon.tamu.edu/. modynamic Changes in Mean and Extreme Pre- 8. See http://www.hrc-lab.org/right_nav_widgets/ cipitation under Changed Climate.” Geophysical realtime_caffg/index.php. Research Letters 32 (17): L17706. 9. The forecasts are generated and publicly dis- Falkenmark, M., and D. Molden. 2008. “Wake Up to played on the Institute of Water Modeling website Realities of River Basin Closure.” Water Resources (http://apps.iwmbd.com/satfor/#). Development 24 (2): 201–15. C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  29 Foster, V. and C. Briceño-Garmendia, eds. 2010. Report 2008-5207, U.S. Geological Survey, Africa’s Infrastructure: A Time for Transformation. Washington, DC. Washington, DC: World Bank. Li, L., Y. Hong, J. Wang, R. F. Adler, F. S. Policelli, S. García, L. E., J. H. Matthews, D. J. Rodríguez, M. Habib, D. Irwin, T. Korme, and L. Okello. 2008. Wijnen, K. N. DiFrancesco, and P. Ray. 2014. “Evaluation of the Real-Time TRMM-Based Beyond Downscaling: A Bottom-Up Approach to Multi-Satellite Precipitation Analysis for an Climate Adaptation for Water Resources Manage- Operational Flood Prediction System in Nzoia ment. AGWA Report 01. Washington, DC: World Basin, Lake Victoria, Africa.” Natural Hazards 50 Bank Group. (1): 109–23. Garduño, H., and S. Foster. 2010. Sustainable Ground- Maddock, T. III, A. Serrat-Capdevila, and J. B. Valdes. water Irrigation: Approaches to Reconciling 2010. “Modeling Surface Water Depletions Due to Demand with Resources. Strategic Overview 4. Groundwater Pumping in a Transboundary Basin.” Washington, DC: World Bank. Paper prepared for the International Conference Hanjra, M. A., and M. E. Qureshi. 2010. “Global Water “Transboundary Aquifers: Challenges and New Crisis and Future Food Security in an Era of Cli- Directions (ISARM2010),” Paris, December 6–8. mate Change.” Food Policy 35 (5): 365–77. Meehl, G. A., C. Tebaldi, H. Teng, and T. C. Peterson. Hardin, G. 1968. “The Tragedy of the Commons.” 2007. “Current and Future U.S. Weather Extremes Science 162 (3859): 1243–48. ­ and El Niño.” Geophysical Research Letters 34 Hartfield, K., G. Schneier-Madanes, S. Marsh, and E. (20): L20704. Curley. 2014. “Remote Sensing of Urban Change: Nikopoulos, E. I., E. N. Anagnostou, F. Hossain, M. Water and Urban Growth in the Sun Corridor Gebremichael, and M. Borga. 2010. “Understand- of the U.S. Southwest.” SWAN Report Working ing the Scale Relationships of Uncertainty Propa- Paper, Sustainable Water Action Network, Law- gation of Satellite Rainfall through a Distributed rence, KS. Hydrologic Model.” Journal of Hydrometeorology Hossain, F., A. H. M. Siddique-E-Akbor, W. Yigzaw, 11 (2): 520–32. S. Shah-Newaz, M. Hossain, L. C. Mazumder, T. O’Brien, K. L., and R. M. Leichenko. 2000. “Double Ahmed, C. K. Shum, H. Lee, S. Biancamaria, F. J. Exposure: Assessing the Impacts of Climate Turk, and A. Limaye. 2014. “Crossing the Valley Change within the Context of Economic Glo- of Death: Lessons Learned from Implementing balization.” Global Environmental Change 10 (3): an Operational Satellite-Based Flood Forecasting 221–32. System.” Bulletin of the American Meteorological Ostrom, E., J. Burger, C. B. Field, R. B. Norgaard, and D. Society 95 (8): 1201–07. Policansky. 1999. “Revisiting the Commons: Local IPCC (Intergovernmental Panel on Climate Change). Lessons, Global Challenges.” Science 284 (5412): 2007. Climate Change 2007: The Physical Science 278–82. Basis; Contribution of Working Group I to the Rodríguez, D. J., C. van den Berg, and A. McMahon. Fourth Assessment Report of the Intergovernmen- 2012. “Investing in Water Infrastructure: Capi- tal Panel on Climate Change, edited by S. Solomon, tal, Operations, and Maintenance.” Water Paper D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. 74051, World Bank, Washington, DC. Averyt, M. Tignor, and H. Miller. Cambridge, U.K.: Rosema, A., M. De Weirdt, S. Foppes, R. Venneker, S. Cambridge University Press. Maskey, Y. Gu, W. Zhao, C. Wang, X. Liu, S. Rao, ———. 2014. “Climate Change 2014 Synthesis Report: D. Dai, Y. Zhang, L. Wen, D. Chen, Y. Di, S. Qiu, Q. Summary for Policymakers.” IPCC, Geneva. Wang, L. Zhang, J. Liu, L. Liu, L. Xie, R. Zhang, Kundzewicz, Z. W., and E. Z. Stakhiv. 2010. “Are Cli- J. Yang, Y. Zhang, M. Luo, B. Hou, L. Zhao, L. mate Models ‘Ready for Prime Time’ in Water Zhu, X. Chen, T. Yang, H. Shang, S. Ren, F. Sun, Y. Resources Management Applications, or Is More Sun, F. Zheng, Y. Xue, Z. Yuan, H. Pang, C. Lu, G. Research Needed?” Hydrological Sciences Journal Liu, X. Guo, D. Du, X. He, X. Tu, W. Sun, B. Bink, 55 (7): 1085–89. and X. Wu. 2008. Satellite Monitoring and Flow Lawford, R. 2014. The GEOSS Water Strategy: From Forecasting System for the Yellow River Basin. Observations to Decisions. Geneva: Group on Scientific final report of ORET Project 02/09- Earth Observations. CN00069. Delft: EARS, December. Leake, S. A., D. R. Pool, and J. M. Leenhouts. 2008. Rosema, A., J. van Huystee, S. Foppes, J. van der “Simulated Effects of Ground-Water Withdrawals Woerd, E. Klaassen, J. Barendse, M. van Assel- and Artificial Recharge on Discharge to Streams, donk, M. Dubreuil, S. Régent, S. Weber, A. Karaa, Springs, and Riparian Vegetation in the Sierra G. Reusche, R. Goslinga, M. Mbaka, F. Gosselink, Vista Subwatershed of the Upper San Pedro R. Leftley, J. Kyokunda, J. Kakweza, R. Lynch, Basin, Southeastern: U.S. Geological Survey Sci- and K. Stigter. 2014. FESA Micro-Insurance: entific Investigations.” Scientific Investigation Crop Insurance Reaching Every Farmer in Africa. 30  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k Scientific Final Report of Millennium Agreements Forecasting System for Sub-Sahara African Water Project 38. Report commissioned by the Neth- Resources and Food Security.” Bulletin of the erlands Ministry of Foreign Affairs, Directorate American Meteorological Society 95 (6): 861–82. Sustainable Economic Development. Delft: EARS Stakhiv, E. Z. 2010. “Practical Approaches to Water Earth Environment Monitoring BV. Management under Climate Change Uncertainty.” Salas, J. D., B. Rajagopalan, L. Saito, and C. Brown. In Hydrocomplexity: New Tools for Solving Wicked 2012. “Special Section on Climate Change and Water Problems, edited by H. H. G. ­Savinje, Water Resources: Climate Nonstationarity and S. Demuth, and P. Hubert. IAHS Publication 338. Water Resources Management, Introduction.” Wallingford, U.K.: IAHS Press. Journal of Water Resources Planning and Manage- Subramanian, A., B. Brown, and A. Wolf. 2012. “Reach- ment 138 (5): 385–88. ing across the Waters: Facing the Risks of Coop- Serrat-Capdevila, A., and A. Mishra, eds. 2012. eration in International Waters.” Water Paper, “Rio+20 Water Security Recommendations.” World Bank, Washington, DC. Forum on Science, Technology and Innovation Verdin, J., C. Funk, G. Senay, and R. Choularton. 2005. for Sustainable Development, International “Climate Science and Famine Early Warning.” Council for Sciences, Rio de Janeiro, Brazil, Philosophical Transactions of the Royal Society B June 11–15. http://www.icsu.org/rio20/science- 360 (1463): 2155–68. and-technology-forum/programme/forum%20 Villarini, G., J. A. Smith, F. Serinaldi, J. Bales, P. D. recommendations/water-security. Bates, and W. F. Krajewski. 2009. “Flood Fre- Serrat-Capdevila, A., A. Browning-Aiken, K. Lansey, quency Analysis for Nonstationary Annual Peak T. Finan, and J. B. Valdés. 2009. “Increasing Socio- Records in an Urban Drainage Basin.” Advances in Ecological Resilience by Placing Science at the Water Resources 32 (8): 1255–66. Decision Table: The Role of the San Pedro Wang, J., Y. Hong, L. Li, J. J. Gourley, S. I. Khan, Basin Decision-Support System Model K. K. Yilmaz, R. F. Adler, F. S. Policelli, S. Habib, (Arizona).” Ecology and Society 14 (1): 37. http:// D. Irwin, A. S. Limaye, T. Korme, and L. Okello. www.ecologyandsociety.org/vol14/iss1/art37/. 2011. “The Coupled Routing and Excess Stor- Serrat-Capdevila, A., J. B. Valdes, F. Dominguez, S. age (CREST) Distributed Hydrological Model.” Rajgopal, and H. Gupta. 2013. “Characterizing Hydrological Sciences Journal 56 (1): 84–98. the Droughts of the New Century in the U.S. Wijnen, M., B. Augeard, B. Hiller, C. Ward, and Southwest: A Comprehensive Assessment from P. Huntjens. 2012. “Managing the Invisible: State-of-the-Art Climate Model Projections.” ­ Understanding and Improving G ­ roundwater International Journal of Water Resources Develop- Governance.” Water Paper, World Bank, ­ ment 29 (2): 152–71. ­Washington, DC. Serrat-Capdevila, A., J. B. Valdes, H. Gupta, and G. World Bank. 2009. “Implementation Completion and Schneier-Madanes. 2014. “Water Governance Results Report of the Environmental Protection Tools: The Role of Science, Decision Support Sys- and Sustainable Development of the Guarani tems, and Stakeholder Participation.” In Global- Aquifer System Project.” World Bank, Washing- ized Water: A Question of Governance, edited by ton, DC. G. Schneier-Madanes, ch. 15. Dordrecht: Springer Wu, B., L. Jiang, N. Yana, C. Perry, and H. Zeng. 2013. Netherlands. “Basin-Wide Evapotranspiration Management: Serrat-Capdevila, A., J. B. Valdes, and E. Z. Stakhiv. Concept and Practical Application in Hai Basin, 2013. “Water Management Applications for Satel- China.” Agricultural Water Management 145 lite Precipitation Products: Synthesis and Rec- (November): 145–53. ommendations.” Journal of the American Water Wu, B., J. Meng, Q. Li, N. Yan, X. Du, and M. Zhang. 2014. Resources Association 399 (1): 1–17. “Remote Sensing-Based Global Crop Monitoring: Shah, T. 2003. “Decentralized Water Harvesting and Experiences with China’s Cropwatch System.” Groundwater Recharge: Can These Save Saura- International Journal of Digital Earth 7 (2): 113–37. shtra and Kutch from Desiccation?” IWMI-Tata Yan, N., B. Wu, V. K. Boken, S. Chang, and L. Yang. Water Policy Program, Colombo, Sri Lanka. 2014. “A Drought Monitoring Operational System Shah, T., S. Darghouth, and A. Dinar. 2006. “Conjunc- for China Using Satellite Data: Design and Evalu- tive Use of Groundwater and Surface Water.” ation.” Geomatics, Natural Hazards, and Risk. doi: Agricultural and Rural Development Notes, Water 10.1080/19475705.2014.895964. for Food Team, World Bank, Washington, DC. Yepes, T. 2008. “Investment Needs for Infrastruc- Sheffield, J., E. Wood, N. Chaney, K. Guan, S. Sadri, ture in Developing Countries 2008–2015.” Latin X. Yuan, L. Olang, A. Amani, A. Ali, S. Demuth, America and Caribbean Region, World Bank, and L. Ogallo. 2014. “A Drought Monitoring and Washington, DC. C h apt e r 1 : K e y G l o b a l W at e r C h a l l e n g e s a n d t h e R o l e o f R e m ot e S e n s i n g   |  31 CHAPTER 2 The World Bank Group and Water INTRODUCTION WATER POLICY AND STRATEGIES Providing water services while sustainably Water has been one of the most important managing a scarce resource has been at the areas of World Bank lending. The 1992 World core of World Bank Group assistance in the Development Report (World Bank 1992) high- water sector. For the Bank, “water” comprises lights some of the difficulties encountered in both water resources management (WRM) this sector, exacerbated even then by rapid and services associated with water, such as population growth and urbanization in devel- water supply and sanitation, energy genera- oping countries. In response, the Bank tion (power plant cooling and other energy approved a policy paper presenting a frame- source needs besides hydropower), irrigation, work for improving the situation (World Bank drainage and flood management, as well as 1993) and drawing from the Dublin Statement environmental services. Water also plays a of the International Conference on Water and crucial role in other areas, from public health the Environment (ICWE 1992) as well as to urban and rural development. From the Agenda 21 (United Nations 1992). ­ discussion in chapter 1, it is clear that World The objectives of the Bank’s WRM policy are Bank activities can benefit considerably from to support countries’ efforts to reduce poverty the use of remote sensing (RS) technology and promote equitable, efficient, and sustain- and, in fact, have already done so. To assess the able development. This is done by ­sustaining the potential value for water-related activities, water environment while providing potable this chapter summarizes the Bank’s water water and sanitation facilities, providing drain- ­ policy, strategies, practice, and portfolio. age services and water for productive services,   33 and protecting people and property from floods. Therefore, the basic strategic challenge for It stresses a comprehensive framework for for- the Bank is finding ways to help clients to scale mulating country policies, taking into account up the impact of their own policies, institu- the interdependence of water resources. tions, and resources, so that the Bank’s A decade later, the 2003 Water Resources resources—whether finance or knowledge— Sector Strategy and the 2003 Water Supply are as effective as possible in helping clients to and Sanitation Sector Business Strategy improve their overall approaches. started guiding the Bank’s work in the water sector.1 Since 2003 practices have evolved to scale up the Bank’s assistance in water (see THE WATER GLOBAL PRACTICE table 2.1): from reengagement in high-risk, high-return infrastructure and stronger The World Bank Group consists of five special- emphasis on improving the delivery of water ized institutions: the International Bank for supply and sanitation services as well as the Reconstruction and Development (IBRD), the management of water resources to a growing International Development Association (IDA), focus on the role of climate change, urban the International Finance Corporation, the development, energy, agriculture, and ­disaster Multilateral Investment Guarantee Agency, risk management. and the International Centre for Settlement of The Bank’s water strategies are still rele- Investment Disputes. IBRD and IDA are com- vant frameworks for addressing today’s monly known as the World Bank, which, as of water-related challenges. However, in order July 1, 2014, has 14 Global Practices as well as to respond to clients’ increasing demand for 5  Cross-Cutting Solution Areas that aim to more and better-quality water by managing a bring best-in-class knowledge and solutions to complex series of trade-offs, the Water Global regional and country clients. Practice has adopted a more inclusive, inte- Through this new operating model, the grated, cross-sector approach to addressing World Bank Group aims to help countries to these challenges. Laid out in the Bank’s 2003 achieve the twin goals of (1) ending extreme Water Resources Sector Strategy, this poverty by 2030 and (2) promoting shared approach describes the main global water prosperity for the bottom 40 percent of the challenges and suggests steps that the Bank population in every developing country. could take to make water more inclusive, such The World Bank Group has been addressing as integrating water with energy, c ­limate, water issues globally through large-scale finan- agriculture, land use, and overall economic cial and technical assistance to countries. To development. This was reaffirmed in the 2010 meet the growing demand for investment midcycle review of the 2003 strategy (World financing driven by the best knowledge avail- Bank 2010). able, the World Bank Group launched a single, The Bank is engaged in a wide variety of integrated Water Global Practice in 2014. The activities dealing directly with water, including Water Global Practice brings together financ- support for water resources management, ing, implementation, and knowledge in one water supply and sanitation services, flood pro- platform that combines the Bank’s global tection, hydropower, irrigation and drainage, as knowledge and country investments. This well as a variety of activities partially or indi- model seeks to generate transformational rectly related to water, such as adaptation to solutions that help countries to grow sustain- ­ and mitigation of climate change, urban devel- ably into the twenty-first century. opment, agriculture, transport, energy develop- The World Bank Group’s strategy places the ment, and environmental protection. poor and most vulnerable people at the center 34  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k Table 2.1 Evolution of Key Principles over Time KEY PRINCIPLE 1993 WRM POLICY PAPER 2003 WATER RESOURCES 2003 WATER SUPPLY AND 2010 MIDCYCLE SECTOR STRATEGY SANITATION SECTOR IMPLEMENTATION REPORT BUSINESS STRATEGY Integration and Focuses on “modern” water Places water resources Establishes a link between Highlights climate change water resources resources management, that management at the sustainable water supply and adaptation and mitigation management is, considers independent center of sustainable sanitation services, better as well as need for management of water by growth, with emphasis management of water cross-sector links. various sectors inappropriate; on basin-wide efficiency resources, sanitation, and selects river basins as unit of in irrigation. wastewater, and analysis. environmental protection. Stakeholders and Emphasizes stakeholder Emphasizes “political Focuses on need to respond Emphasizes building client institutions participation in water economy of change.” to local demand and capacity for results-based resources management and complement local initiatives; decision making. need to respect the principle promotes private of “subsidiarity.” participation. Position on Promotes investments to Commits to reengage Recognizes that infrastructure Reaffirms emphasis on infrastructure improve water quality; with high-risk, high- is important but insufficient infrastructure, with efforts promotes investments to reward hydraulic for sustainability; switches to link quantity and quality increase supply only when infrastructure. focus to operator to infrastructure adequate demand performance and service investments. management is in place. quality. Economic and Emphasizes incentives and Emphasizes need for Emphasizes need for clear and Emphasizes need for more financial principles economic principles for water pricing, cost consistent financial policies, efficient water supply improving allocation and recovery, and utility affordability, and centrality of systems and support for enhancing quality. reform. cost recovery. low-cost, onsite sanitation. Source: World Bank 2013. Note: WRM = water resources management. of its work. Efforts aim to ensure that everyone to improve the livelihoods of millions of the has basic access to sustainable water and sani- poorest people. world’s ­ tation services and that management of water resources addresses water considerations in sectors such as agriculture, energy, disaster THE WATER PORTFOLIO risk management, and health. Finally, these efforts place water at the center of adaptation The World Bank portfolio is a valuable source of strategies to help countries to cope with the information on the profile of Bank activities and effects of climate change and build a more provides a snapshot of the budget allocated to resilient future for generations to come. strategic goals and priority s­ ectors. The activi- Robust solutions to complex water issues ties are classified by sector (using 10 different incorporate cutting-edge knowledge and codes), according to which part of the economy innovation. New knowledge products that received support,2 and by theme (66 in total), draw on the Bank’s global experiences and corresponding to the goals of Bank activities. partner expertise are filling the gaps in global Each sector is subdivided into subsectors or knowledge and transforming the design of sector codes, and the themes are grouped into water investment projects to deliver results. 11 categories.3 Each project in the portfolio indi- Multiyear, programmatic engagements in cates which sectors and themes it has been strategic areas are designed to make dramatic mapped to and the corresponding share of economic improvements in the long term and investment. In this publication, the sectors and C h apt e r 2 : T h e W or l d Ba n k G ro u p a n d W at e r   |  35 themes of a project that represent the largest Figure 2.1 Historical Water Lending, by Subsector, share are denominated “primary,” while those FY 2009–14 receiving smaller shares are called “secondary.” 5,000 Up to July 1, 2014, a Water Sector Board 4,000 was responsible for the quality of activities associated with water, sanitation, and flood US$ (millions) 3,000 protection, which included 10 sectoral codes 2,000 and 1  thematic code. Water-related projects, however, were mapped not only to the water 1,000 sector and theme, but also to other sectors 0 Fiscal Year Fiscal Year Fiscal Year Fiscal Year Fiscal Year Fiscal Year and themes such as agriculture, rural devel- 2009 2010 2011 2012 2013 2014 opment, and urban development. Water supply and sanitation Water resources management Hydropower Flood protection The following sections summarize the Irrigation Source: Water Portfolio Monitor, FY14 Quarter 2 Update. World Bank water portfolio to gain insight into the profile Business Warehouse database. of the Bank’s water-related operations. For Note: Data for the thematic code “water resources management” are not included in the total to avoid double counting, since water this publication, the water portfolio was resources management is a cross-cutting theme. divided into (a) lending and (b) analytical and advisory activities (AAAs). Figure 2.2 Historical Water Lending, by Region, FY 2009–14 Lending The active water portfolio as of April 30, 2014, 8,000 was worth US$33 billion.4 It included 272 active 7,000 6,000 projects in the following subsectors: water sup- 5,000 ply and sanitation, flood protection, irrigation, US$ (millions) 4,000 and hydropower. Water supply and sanitation 3,000 was the largest subsector, accounting for 63 2,000 percent of water lending. Irrigation was the 1,000 second largest (20 percent), followed by hydro- 0 Fiscal Year Fiscal Year Fiscal Year Fiscal Year Fiscal Year Fiscal Year power (9 percent) and flood protection 2009 2010 2011 2012 2013 2014a (8 ­percent). A little less than half (43 ­percent) of Africa East Asia and Pacific Latin America and the Caribbean Middle East and North Africa water financing goes to projects mapped to sec- Europe and Central Asia South Asia Source: Water Portfolio Monitor, FY14 Quarter 2 Update. World Bank tors other than water. The sectors and themes Business Warehouse database. with the largest share of water components are a. Pipeline includes projects with a “firm” and “likely” probability of agriculture and rural development (23 per- approval. Data as of January 10, 2014. cent), urban development (34  percent), and energy (10 percent). The water portfolio’s total complement financing operations, primarily value nearly doubled over the last five years. economic and sector work and technical assis- Lending in fiscal year 20145 (third quarter) rose tance. Economic and sector work comprises to US$6.9 b ­ illion,6 representing about 18 per- products published by the Bank that can cent of the Bank’s portfolio. inform and influence the planning and design Figures 2.1 and 2.2 show some of the subsec- of a country strategy, lending program, or pol- tor and geographic trends in water lending in icy and that can build the client’s analytical the past six years. capacity. Technical assistance activities strengthen local institutions, promote knowl- Analytical and Advisory Activities edge exchange, and prepare clients for reform The water portfolio also includes significant and program implementation. Technical assis- dollar amounts for AAAs that support and tance now constitutes more than two-thirds of 36  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k all AAAs (72 percent). In the period covered in projects and hydropower projects in the total figures 2.1 and 2.2, Africa and East Asia and the number of projects is close to 8 percent each. Pacific accounted for the largest share of AAAs. Table 2.3 shows the number of projects In fiscal year 2014 (as of the end of the third mapped to non-water sectors that were never- quarter), about 70 percent of AAA was dedi- theless considered water related because cated to water supply and sanitation.7 These water resources management was coded as a nonlending activities generally mirror the theme. Of these, the agriculture, fishing, and breakdown of investment operations. forestry sector is the non-water sector with the largest number of water-related projects. Sector and Theme Components in the Water Portfolio Table 2.3 Water-Related Projects with Non-Water During the period analyzed (between fiscal year Sector Codes 2002 and fiscal year 2012), the Bank used sector NON-WATER SUBSECTOR NUMBER OF and theme codes to classify projects.8 Table 2.2 WATER-RELATED shows the total number of water projects by PROJECTS water sector code. A total of 179 water supply Forestry 4 and sanitation projects and 176 general water, General agriculture, fishing, 73 sanitation, and flood protection projects were and forestry identified. Together, they represent nearly half Health 6 of all water-related projects. Irrigation and Public administration: 5 Agriculture, fishing, and drainage projects represent the third largest forestry sector, with 14.6 percent of the total number of Other non-water sectors with 11 projects. The share of both flood protection water resources management as a theme Total 99 Table 2.2 Total Water-Related Lending, by Water Source: Water Portfolio analysis. World Bank Business Warehouse Subsector database. WATER SUBSECTOR TOTAL NUMBER OF WATER SECTOR In the pool of 775 projects identified, other PROJECTS non-water sectors with water-related projects Flood protection 58 that did not have water resources manage- General water, sanitation, 176 ment coded as a theme (6 in total) were also and flood protection identified. The non-water sectors represented Hydropower 60 were strongly linked to climate change (1 proj- Irrigation and drainage 113 ect), environment and water resources man- Ports, water, and shipping a 12 agement (3), natural disaster management (1), and land administration and management (1). Public administration: Water, 28 sanitation, and flood protection Wastewater and sewerage 44 ANNEX 2A. DATA ANALYSIS Water supply and sanitation 179 METHODOLOGY OF THE SECTOR AND THEME COMPONENTS IN THE Total 670 WATER PORTFOLIO Source: Water Portfolio analysis. World Bank Business Warehouse database. a. Since 2011, ports, waterways, and shipping have been coded under Annex 2A is available online at https:// the transport sector. Before 2011, they belonged to water. Thus only openknowledge.worldbank.org/handle/10986 projects approved between fiscal year 2002 and fiscal year 2011 are considered as water related under this category. /22952. C h apt e r 2 : T h e W or l d Ba n k G ro u p a n d W at e r   |  37 NOTES 8. See annex 2A for a description of the methodology used in the data analysis. Annexes to this book are available online at 1. The Water Resources Strategy was rooted in the https://openknowledge.worldbank.org 1993 policy paper (World Bank 1993). Since 2003, /handle/10986/22952. water sector thinking has also been informed by a 2010 sector study on water and development (IEG 2010), an implementation progress report (World Bank 2010), and various other analytical work and portfolio reviews. REFERENCES 2. Agriculture, fishing, and forestry; public ­ administration, law, and justice; information ICWE (International Conference on Water and the and communications; education; finance; health Environment). 1992. “Dublin Statement on and other social services; energy and mining; Water and Sustainable Development Adopted ­ transportation; water, sanitation, and flood ­ nternational January 31, 1992, in Dublin, Ireland.” I ­ protection; and ­industry and trade. Conference on Water and the Environment, Dublin, January 26–31. 3. Economic management; public sector ­ governance; rule of law; financial and private sector ­ development; IEG (Independent Evaluation Group). 2010. “Water trade and integration; social p­ rotection and risk and Development: World Bank Support, 1997– management; social ­ development, gender, and 2007.” IEG Study Series, World Bank, Washing- inclusion; human development; urban develop- ton, DC. ment; rural development; and environment and United Nations. 1992. “Protection of the Quality and natural resources management. Supply of Freshwater Resources: Application of 4. The figures in this chapter are given only as Integrated Approaches to the Development, background for the process of evaluating the cur- ­ Management, and Use of Water Resources.” rent and future significance of remote sensing in In Agenda 21, vol. 2, ch. 18. New York: United the Bank’s water-related activities. They are also Nations Conference on Environment and used to illustrate orders of magnitude or geo- Development. graphic and time-based relative comparisons. They World Bank. 1992. World Development Report 1992: should not be considered official Bank figures. Development and the Environment. New York: 5. The Bank’s fiscal year runs from July 1 to June 30. Oxford University Press. 6. Data as of April 4, 2014. Fiscal year 2014 ———. 1993. “Water Resources Management.” Policy commitments include projected pipeline. Data ­ Paper, World Bank, Washington, DC. for the thematic code “water resources ———. 2010. “Sustaining Water for All in a Changing management” are not included in the total to Climate: World Bank Group Implementation avoid double counting, since water resources Progress Report of the Water Resources Sector management is a cross-cutting theme. Strategy.” World Bank, Washington, DC. 7. The strategic and analytical work conducted ———. 2013. “Water Vision.” Water Anchor by the Bank’s Water Sanitation Program is not Working Paper (unpublished), World Bank, included these figures. Washington, DC. 38  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k CHAPTER 3 The World Bank and Remote Sensing INTRODUCTION INTERNAL INITIATIVES This chapter presents an overview of how For a long time, remote sensing was a special- remote sensing (RS) has been used in the Bank ized niche outside the mainstream of the Bank’s to date. It begins with a summary of the inter- work. As partnerships for remote sensing have nal and external programs or “windows” become more commonplace and the range of through which RS data have been made avail- available products has widened, awareness has able for use in projects funded or managed by grown among practitioners at the World Bank. the World Bank, including those related to In recent years, some efforts have been made to water resources management. These pro- integrate initiatives and organize the use of this grams are often specific partnerships with tool within the Bank. This process has been a public or private entities such as the National learning experience, and not every application Aeronautics and Space Administration has been successful; nevertheless, capacity and (NASA), the European Space Agency (ESA), or experience are being built up. The following sec- corporations selling digital imagery or soft- tions summarize these activities within the Bank ware products. The nine windows identified and provide links to further resources. More are described briefly, and activities sponsored details may be found in the consolidated matrix by these programs are summarized; most of of those programs (see table 3A.1 in annex 3A, these programs are still active.1 The chapter available online at https://openknowledge then presents the results of a Bank portfolio .worldbank.org/handle/10986/22952. review conducted to identify those operations that have used RS products in some way, as Earth Observation for Development well as some considerations for future use Earth Observation for Development is intended derived from a limited survey of Bank staff. to be a single hub for all Earth observation   39 (EO) activities occurring within the Bank.2 was €1.3 million. In addition, ESA provided In combination with the GeoWB data portal,3 access to EO data from 15 satellite missions, for it is part of a process of building internal capac- a total value of €1 million. Those satellite mis- ity for managing geospatial data sets devel- sions included the European Remote Sensing oped through World Bank operations. Satellite (ERS), Envisat, RapidEye, the Satellite Earth Observation for Development seeks for Earth Observation (SPOT), Cosmo-Skymed, to capture and integrate knowledge products TerraSAR-X, Radarsat, GeoEye, and WorldView. based on RS data that had previously been Five dedicated, hands-on training workshops developed in isolation. It aims to create a uni- were organized in Brazil, Indonesia, Papua New fied source for existing RS products that can Guinea, and Zambia and at the headquarters of also be expanded to incorporate new products the Indian Ocean Commission. The ESA and as they are created. Its larger goal is to main- the World Bank published the results of stream RS data and products—making them EOWorld in 2013 (European Space Agency and available to the broader community of develop- World Bank 2013). ment practitioners, along with best practices, Following the success of the first round of lessons learned, and experiences pertaining to the program, ESA extended its financial and their use. This involves raising awareness of technical supervision support, launching a call the role that remote sensing can play in sus- for proposals for a new set of activities to pro- tainable development and of the range of exist- duce and deliver EO information services. ing RS products and services. It is aligned with These activities focused on four themes: the Bank’s recent “Open Data, Open Knowl- (a) urban development, (b) disaster risk man- edge, Open Solutions” policy reforms. agement, (c) forestry, and (d) oceans (World Currently, Earth Observation for Develop- Bank 2012). ment provides access to data and products from three sources: (a) the EOWorld partner- U.S. Government and World Bank ship with the European Space Agency, (b) the Agreement agreement between the World Bank and the A memorandum of understanding was signed U.S. government, involving several agencies, in March 2011 between the World Bank and and (c) the agreement between the World Bank the U.S. government. Its goal is “supporting and the Japan Aerospace Exploration Agency developing countries’ effort to create a water- (JAXA). Each of these is discussed briefly in secure world and to fight water scarcity and the following sections. poor water quality.” Under this agreement, where possible, U.S. EOWorld European Space Agency government agencies such as NASA, the While the partnership between the World National Oceanic and Atmospheric Adminis- Bank and the European Space Agency is tration (NOAA), U.S. Geological Survey, and anchored in the Bank’s Urban, Rural, and U.S. Department of Agriculture will provide RS Social Development Global Practice, it brings data and the means necessary to interpret and together expertise from all regions of the world employ them. and all the Global Practices of the World Bank. The following categories have been identi- The EOWorld partnership was established fied as priority areas for the use of RS data: in two stages. In 2008, the pilot program started climate variability and change, agricultural sys- ­ with 3 activities; in 2010, the partnership tems, and water systems planning and manage- expanded to involve 12 activities. Those 12 activ- ment. The data will support (a) sound ities were selected after a competitive “call for management of water resources, (b) reliable proposals.” The total value of the 12  activities and sustainable access to an acceptable quantity 40  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k and quality of water to meet human livelihood, objective is “to facilitate cooperative water ecosystem, and production needs, (c) efforts to resources management and development in the lower the risk of hydrologic events, and (d) reha- Nile River basin. This would be achieved bilitation of degraded watersheds. These RS through the provision of targeted technical tools, developed by U.S. government agencies, assistance to the initiative’s member countries hold potential for developing countries to and broader stakeholders, to facilitate coopera- improve productivity and reduce conflict, while tive activities, improve integrated water also increasing resilience to climate change.4 resources planning and management, and iden- Under this agreement, several knowledge- tify and prepare studies of potential investments exchange events have been organized to build of regional significance” (World Bank 2012). familiarity between NASA and World Bank The technical assistance to be provided staff and to begin developing tools and under NCORE will include geospatial analysis approaches to using remote sensing for to improve the analysis of existing RS data sets development. on wetlands, which will facilitate the prepara- tion of future investments. Additional efforts Japan Aerospace Exploration Agency include improving public access to the existing JAXA and the World Bank signed an agree- database, sharing knowledge among countries ment in 2008 for the use of data from the and institutions, and establishing a real-time Advanced Land Observation Satellite (ALOS). hydrometeorological portal. Developed and operated by JAXA, ALOS pro- vided high-resolution images of the regions of BOX 3.1 Latin America and the Caribbean where severe impacts of climate change were expected. The Nile Basin Initiative ALOS images and data were used in support of World Bank adaptation projects in Bolivia, The Nile Basin Initiative (NBI) is a cooperative, intergovernmental partner- Colombia, Ecuador, Mexico, the Andes region ship among the 10 countries whose territories occupy the basin of the of Peru, and the West Indies. These images Nile River in Africa: Burundi, the Democratic Republic of Congo, the Arab were used to detect changes in vulnerable eco- Republic of Egypt, Ethiopia, Kenya, Rwanda, South Sudan, Sudan, Tanzania, and Uganda; Eritrea has observer status. systems regionwide, which contributed to the The NBI provides a forum for engaging dialogue on the joint manage- development of adaptation programs in the ment of the river and its shared watershed, sharing information, and region. Images taken by ALOS of the tropical ­ building capacity. Notable among these efforts is the creation of a Nile glaciers in the Andes were used to assess gla- Decision Support System to integrate the relevant information to assist cier dynamics under an adaptation project in decision makers in formulating policy for the basin. The NBI also includes that area. As of April 15, 2008, total investment some funding for common activities, plus investments in water manage- ment at the subbasin level. In the 15 years since the NBI’s inception, more in adaptation in Latin America, including than a dozen projects have been completed in the Nile basin, often man- World Bank support, totaled US$90 million. In aged through the World Bank. 2011, JAXA officially terminated the operation RS and geospatial data and information products are an important of ALOS because of a failure of the satellite’s part of the information-sharing and capacity-building efforts of the NBI. power system. This effectively ended the Many of the investment projects undertaken have incorporated elements collaboration. of geospatial data gathering or improved data interpretation and have built institutional capacity to create and manipulate geospatial data. The NASA Nile Project is working with the Eastern Nile Technical Nile Cooperation for Results Project Regional Office (ENTRO), which is part of the NBI. Using Tropic Rainfall ­ The Nile Cooperation for Results Project Measuring Mission data, ENTRO provides flood forecasts for the Eastern (NCORE) is one of the latest investment proj- Nile basin. The NASA Nile Project has also produced analyses of the water ects carried out by the World Bank under the balance in the Nile basin using remote sensing data. Nile Basin Initiative (box 3.1). Its development c h apt e r 3 : T h e W or l d Ba n k a n d R e m ot e S e n s i n g   |  41 GeoWB, GeoCenter, and Spatial Help Desk The first phase of TIGER-NET focused on Facilities such as GeoWB and Spatial Help consultation, review, and analysis of user needs Desk provide visualization services for RS data and current technological capacity and demand and have created data repositories for projects for the application. During this phase it was across the World Bank, making RS data acces- concluded that the various institutions had sible to specific projects. GeoWB is an internal very similar system requirements but that their spatial data platform, managed by the World application requirements and information Bank’s GeoCenter, that enables data sharing needs varied according to the specific chal- and map visualizations. It was launched to lenges posed by different water basins. In its work on sustainable development in collabora- second phase, the program will aim to extend tion with Esri, the leading geographic informa- the number of water authorities involved as tion system (GIS) software provider. GeoCenter host institutions for the WOIS. not only supports the GeoWB data portal, but More details on World Bank projects with also provides GIS and mapping services. Spa- a  TIGER-NET component may be found in tial Help Desk elaborates maps and other spa- table 3A.2 in annex 3A (available online). tial products such as interactive files containing regular maps, three-dimensional maps, and NASA SERVIR spatial data analyses. SERVIR—the Regional Visualization and Moni- toring System—was launched in 2004 as a col- laborative effort of NASA, the U.S. Agency for EXTERNAL INITIATIVES International Development, the World Bank, and the Central American Commission for Envi- This section briefly discusses external initia- ronment and Development. It provides satellite- tives in which the Bank participates. based EO data and science applications to help developing countries in Central America, East TIGER-NET Africa, and the Himalayas to improve their envi- The Bank decided to participate in the TIGER ronmental decision making with regard to the Initiative, launched by ESA in March 2012, in nine societal benefit areas identified by the response to growing needs for information Group on Earth O ­ bservations—disasters, eco- related to integrated water resources manage- systems, energy, biodiversity, weather, water, cli- ment in Africa. TIGER-NET is a major compo- mate, health, and agriculture. Other partners nent of the TIGER Initiative5 and will run for within the U.S. government are NOAA, U.S. three years, with a total budget of €1.5 million. Environmental Protection Agency, U.S. Forestry TIGER-NET supports the assessment and Service, and U.S. Geological Survey. monitoring of water resources from the water- SERVIR facilitates decision making by gov- shed to the cross-border basin level aimed at ernment officials, managers, scientists, research- (a) developing an open-source Water Observa- ers, students, and the general public by providing tion and Information System (WOIS),6 for Earth observations and predictive models based monitoring, assessing, and taking stock of on data from orbiting satellites, ground-­ based water resources using EO data and (b) provid- observations, and forecast models. Since the ing capacity building and training to enable eventual goal of SERVIR is to become self-­ African water authorities to exploit the full sustaining (with host nation support), it works capacities offered by satellites such as Sentinel closely with governments and international and EO data. These EO products and services organizations.7 SERVIR has participated in are used to monitor, assess, and manage water training sessions, brown-bag lunch seminars, resources. and presentations at Bank-organized events. 42  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k Open Landscape Partnership Program insight into specific ways in which RS data The Open Landscape Partnership Program is a have been used. These include filling data gaps, global joint initiative of satellite data providers, serving as input for modeling to evaluate a distributors, processors, and end users.8 Its project’s impact, supporting basin planning in objective is to create a community of practice prefeasibility studies, or helping to boost proj- that will expand demand for open access, high- ect performance or operational quality.12 resolution satellite imagery (2-meter resolu- More details on the lending and AAAs tion and a 1-month frequency or better). The reviewed in this context may be found in data could be used to further public account- table  3A.2 in annex 3A (available online). ability, transparency, and sustainability of nat- Selected examples of EO applications in Uttar ural resources management for ecologically Pradesh, India, and in Malawi and Zambia are important areas. presented in appendix A of this publication. Subscribers to the platform’s pilot phase will get free Web access  to available World- General Trends View-2 satellite imagery for the stated area of A portfolio review identified 61 lending proj- interest, online mapping tools, and designated ects and 16 AAAs approved between July  1, server space. They will be able to use these 2001, and April 30, 2014, that used remote sens- assets to develop and  document their own ing (figure 3.1). The breakdown of total lending crowd-mapping projects of critical landscapes and AAA by subsector is similar to the in-depth and hotspots, in exchange for agreeing to con- breakdown of the portfolio (see chapter 2).13 tribute the documented results (including Web The analysis indicates that the use of RS maps) to the platform’s project library. The lat- applications in lending and AAA has increased ter would be available, through an online steadily over the years, especially since 2007. forum, for review, analysis, and discussion by Yet only a small share of all the water projects peer practitioners.9  identified in the period under review (about 10 percent) actually used, are using, or plan to use RS technologies or approaches in their RS APPLICATIONS IN WORLD BANK operations or AAA. WATER–RELATED PROJECTS AND ANALYTICAL AND ADVISORY Figure 3.1  Water-Related Lending and Analytical and Advisory Activities ACTIVITIES Using Remote Sensing, 1997–2013 16 Based on the results of the portfolio review dis- cussed in chapter 2,10 this section identifies the 14 Number of projects and AAAs areas where RS technologies have been used in 12 the Bank’s Water Global Practice and looks at 10 the areas where RS tools could be applied.11 8 The results presented in this section provide 6 an overview of the RS applications in Bank lending and analytical and advisory activity 4 (AAA) related to water, aimed at identifying 2 (a)  the water challenges that the RS applica- 0 tions address, (b) the current operational uses 1997 2001 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Year of RS tools, and (c) the relationship between Source: RS Portfolio analysis. World Bank data. RS tools and specific development objective(s) Note: Data for 2014 have not been taken into account because this analysis only covers the first four of the project or AAA. Thus the analysis gives months of 2014. AAAs = analytical and advisory activities. c h apt e r 3 : T h e W or l d Ba n k a n d R e m ot e S e n s i n g   |  43 About 78 percent of total lending and AAA management. Other than water resources combined are water-dedicated activities (see management, climate change and natural table 3A.2 in annex 3A (available online))— disaster management are the two cross- that is, activities whose investments in water or sectoral themes showing the largest number of the share of water-coded subsectors represent RS applications in water projects and AAA. 50 percent or more of the Bank’s total commit- Not only is this conclusion reflected in the ment for that particular project or activity. project, AAA sector, and theme, but the appli- As figure 3.2 shows, Africa has the highest cations themselves show that multivariate use of RS applications (39 percent), South Asia inputs from Earth observation have been con- has the second highest (18 percent), and the sidered alongside water-related variables. Middle East and North Africa ranks third Thus applications in both water-dedicated and (15  percent). In Africa, RS applications have non-water-dedicated projects14 and AAAs may been used at the basin and subregional level. As also encompass non-water components. discussed, much attention has been given to As shown in table 3.1, the share of water investing in RS technologies to address trans- resources management in water-related projects boundary watershed management challenges (lending) using remote sensing is 57 percent, often involving more than two countries or two whereas the combined share of climate change or more projects. This partly explains the rela- and natural disaster management themes is tively high number of countries and projects in 26  percent. Rural services and infrastructure, the Africa region using RS technologies. although relatively small, ranks third as primary Unlike the codes, which are sector specific, Sensing Table 3.1  Number of Projects Using Remote ­ themes can be attached to both water and in Water-Related Lending and Analytical and other subsectors (table 3.1). A significant share ­ Advisory Activities, by Primary Theme of water-related lending and AAA that have PRIMARY THEME LENDING AAA used (or planned to use) remote sensing have Other than water resources done so in cross-cutting areas nearly as often management as they have in the area of water resources Climate change 6 3 Figure 3.2  Lending and Analytical and Advisory Environmental policies and 0 1 institutions Activities Using Remote Sensing, by Bank Region Infrastructure services for 0 1 3% private sector development Land administration and 2 0 management 18% Natural disaster management 10 3 Pollution management and 1 0 39% environmental health Rural policies and institutions 1 0 15% Rural services and 5 0 infrastructure Urban services and housing 1 1 12% 8% for the poor 5% Water resources management 35 7 Africa Middle East and North Africa       East Asia and Pacific South Asia Europe and Central Asia World Total 61 16 Latin America and the Caribbean Source: RS Portfolio analysis. World Bank data. Source: RS Portfolio analysis. World Bank data. Note: AAA = analytical and advisory activity. 44  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k non-water theme. Among the AAAs, climate • Evaluation of project impact on agricultural change and natural disaster management com- water management. Integrated landscape bined share 38 percent of all water-related activ- management and agricultural intensifica- ities, while water resources management takes tion, climate-smart agriculture, and agri- the lead with 44 percent. cultural value chains Results by Subsector • Agricultural water-saving measures and Table 3A.2 in annex 3A (available online) lists support services. Irrigation planning and the entire sample of World Bank water- monitoring; reduction of nonbeneficial related projects and AAAs that have used (or evapotranspiration; farm-level resilience planned to use) remote sensing; it also to climate change, raising farm income by includes a brief d­ escription of the RS applica- increasing farm yields and output value; tions used in each project. Table 3.2 summa- planning and training tools at micro- rizes that table. Additional information about watershed levels; maps and climate infor- the relevant sectors is presented in annex 3A, mation for use by farmers in decision which highlights the attributes related to the making; agroclimatic advisory risk sys- use of remote sensing, their characteristics, tems; improved Web-based information on and any trends, by subsector. markets, postharvesting, and value addi- World Bank Potential Demand for RS tion; farm participatory field trials and Applications demonstrations for specific technologies; As discussed at the beginning of this chapter, and research management to strengthen several windows provide water-related RS the institutional arrangements for longer- assistance and products within the World Bank. term, needs-based research identification, At present, the nature of both actual and planned technology transfer, research quality assur- uses of RS applications varies widely within the ance, and coordination of rain-fed agricul- Water Global Practice: ture and watershed management research Table 3.2  Use of Remote Sensing in World Bank Lending and Analytical and Advisory Activities, by Water Subsector NUMBER OF PROJECTS NUMBER OF AAAs TOTAL (PROJECTS TOTAL CATEGORY  SECTOR PRIMARY SECONDARY TOTAL  PRIMARY SECONDARY TOTAL  + AAA) (%) 1 Flood protection 9 0 9 2 0 2 11 14 General water, sanitation, 19 3 22 10 3 13 35 45 2 and flood protection 3 Irrigation and drainage 16 3 19 0 0 0 19 25 Public administration: water, 3 0 3 0 0 0 3 4 4 sanitation, and flood protection Renewable energy and 3 0 3 0 0 0 3 4 5 hydropower 6 Wastewater and sewerage 1 0 1 0 0 0 1 1 7 Water supply and sanitation 2 1 3 1 0 1 4 5 General agriculture, fishing, 1 0 1 0 0 0 1 1 * and forestry   Total     61     16 77 100 Note: AAA = analytical and advisory activity. * Not a water subsector per se, but has water resources management as a theme. c h apt e r 3 : T h e W or l d Ba n k a n d R e m ot e S e n s i n g   |  45 • Use of modern, basin-wide water resources for forecasting and early warning systems information systems. Water information as well as for disaster preparedness, disas- system platforms ter management, and disaster response. Additionally, it could be used to improve • Feasibility studies. Irrigation projects, climate variability and change. resilience to ­ hydropower stations, and use of digital elevation models for reservoir inundation • Agricultural systems. The agriculture sec- models and site identification tor in developing countries is particularly • Basin planning, monitoring, and forecasting. vulnerable to climate change and can ben- Watershed planning and monitoring efit greatly from RS assistance, particularly toward (a) mapping evapotranspiration for • Transboundary options for flood risk miti- use in estimating water losses and monitor- gation. Pilot nonstructural flood prepared- ing irrigation water use and (b) monitoring ness and emergency response activities; the performance of cropping systems for regional flood forecasting, warning, and use in improving the management of both communication systems; regional data irrigated and rain-fed systems. sharing on flood operation mechanisms; urban mapping of buildings and infrastruc- • Water systems planning and management. ture; urban growth monitoring; regional Comprehensive planning of water systems assessment of water resources manage- requires the ability to estimate surface water ment on shared regional aquifers and groundwater fluxes in river basins. Existing RS systems can provide new tools to • Investment planning and basin decision sup- monitor or estimate various elements of the port systems. Systematic information base hydrologic cycle, including precipitation, and tools for water investments in systems evapotranspiration, flows, changes in avail- contexts; identification of different types able surface water and groundwater, water of infrastructure considered in the calcula- storage, aquifer recharge, and inundation. tion of water balance These data will also facilitate basin plan- • Institutional and community planning ning, inflow forecasting, systems operations, frameworks for addressing environmen- and water infrastructure management. tal and social issues. Basin-wide planning (capacity building and coordination of gov- ernment institutions in decision making ANNEX 3A. WORLD BANK REMOTE for the sustainable use and conservation of SENSING PROGRAMS water resources) and conservation of habi- tats and biodiversity. Annex 3A is available online at https:// openknowledge.worldbank.org/ handle/ However, a survey of task team leaders indi- 10986/22952. cated that remote sensing could also be useful in the following categories and operational areas:15 ANNEX 3B. METHODOLOGY AND • Climate variability and change. As the cli- RESULTS OF THE USE OF EARTH mate and its variability change, remote sens- OBSERVATION APPLICATIONS BY ing could be useful for countries seeking to WATER SUBSECTOR improve their capabilities in (a)  managing droughts and floods, (b)  reducing other forms of disaster risk, and (c)  mitigating Annex 3B is available online at https:// the impact of climate change. Against this openknowledge.worldbank.org/ handle/ background, remote sensing could be used 10986/22952. 46  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k NOTES 8. The Open Landscape Partnership Program is a joint initiative of Scanex Research and Develop- ment Center, Transparent World, and Digital- 1. However, a complete compilation of the data Globe, Inc.—the platform’s founding partners—in used in these activities, with a view to making all collaboration with NASA, OpenStreetMap, data available to World Bank staff and the p ­ ublic, ­ University Geoportals Consortium, the World lies in the domain of the Earth Observation for Bank, the World Resources Institute, the World Development and GeoWB initiatives, which are Wide Fund for Nature, Yandex, and various par- described later. It is outside the scope of this ticipants of the Critical Ecosystems Partnership publication. Fund, the Global Forest Watch 2.0, the Global 2. The term “Earth observation” is sometimes used TIGER Initiative, the Global Snow Leopard in a broad sense to include both in situ and RS ­ Initiative, and the Save Our Species program. See observations. However, this report uses Earth http://www.openlandscape.info. observation and remote sensing interchangeably, 9. For information on the platform, see http://www the former being more common in Europe and .openlandscape.info/index.php?option=com_cont the latter more common in North America. ent&view=article&id=18&Itemid=2. 3. GeoWB is an internal spatial data platform, 10. Updated to fiscal year 2014. managed by the Bank’s GeoCenter, that enables ­ 11. For more details on the methodological approach data sharing and map visualizations. used and the data analyzed, see annex 3B (avail- 4. For the media note prepared and presented able online at https://openknowledge.worldbank by the U.S. Department of State on this .org/handle/10986/22952. ­ collaboration, see http://www.state.gov/r/pa/prs/ 12. This “narrow” review fails to specify (a) the ps/2011/03/158835.htm. For the m ­ emorandum quality and quantity of data generated to fill of understanding, see http://www.state. information gaps, (b) whether the information gov/e/oes/158770.htm. For a fact sheet on the gathered has been validated or the models have agreement, see http://www.state.gov/r/pa/prs/ ­ been calibrated, (c) the resolutions used, and (d) ps/2011/03/158774.htm. the extent to which RS applications significantly 5. ESA launched the TIGER Initiative in 2002 in influenced a project’s performance or the deci- response to the urgent need for action high- sions about it. The limitation of this review pre- lighted by the Johannesburg World Summit on cludes an in-depth analysis of the effectiveness Sustainable Development and in the context of of each RS application. the Committee on Earth Observation Satellites. 13. See chapter 2 for the total number of projects The overall objective of the initiative is to help per subsector. However, unlike the results of the African countries to overcome the problems portfolio analysis, where the water supply and they face in the collection, analysis, and use of sanitation subsector represents a major share of water-related geo-information by exploiting the the whole portfolio, in this chapter this subsector advantages of EO technology. For more informa- represents one of the smallest portfolio shares tion on the TIGER Initiative, see http://www. using remote sensing. tiger.esa.int. 14. A water-dedicated project or activity is a project 6. A WOIS is a multipurpose system consisting of or activity whose share of lending commitment a storage container for geodata, EO data-pro- for water-related activities is greater than or equal cessing facilities for extracting and processing to 50 percent; a non-water-dedicated project or modeling tools for hydrologic modules, and data, ­ activity is a project or activity whose share of visualization and analysis tools. Open-source ­ lending commitment for water-related activities is software components like GRASS GIS, BEAM less than 50 percent. and NEST, Orfeo Toolbox, SWAT, PostGIS, and R scripts have been integrated so that all 15. These categories are derived from a small survey functionalities can be accessed as part of step- conducted among Bank task team leaders active by-step, fully ­ automated scripts. The system in the water sector to get a sense of their poten- architecture allows customization by users and tial demand for RS applications. Given the very system adaption and ­ scalability. More informa- limited nature of the survey, the results are only tion about WOIS can be found in European indicative. Space Agency (2014). 7. More information can be obtained from Open Government at NASA Open (http://www. nasa.gov/open) and from the NASA fact sheet, REFERENCES SERVIR: Connecting Space to Village” (http:// “­ www.nasa.gov/sites/default/files/638969main_ European Space Agency. 2014. Satellite Observa- SERVIR.pdf ). tions Supporting Integrated Water Resources c h apt e r 3 : T h e W or l d Ba n k a n d R e m ot e S e n s i n g   |  47 Management in Africa. Paris: European Space esamultimedia.esa.int/multimedia/publications/ Agency. http://www.tiger-net.org. ESA_WB_Partnership_Report_2013_complete/. European Space Agency and World Bank. 2013. World Bank. 2012. “European Space Agency (ESA): Earth Observation for Sustainable Development. World Bank Collaboration Earth Observation for Partnership Report. Paris: European Space Development.” World Bank, Washington, DC. Agency; Washington, DC: World Bank. http:// http://go.worldbank.org/IBJTNEU2U0. 48  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k CHAPTER 4 Key Data Needs for Good Water Management INTRODUCTION This chapter enumerates the variables included in this assessment and discusses how each one is rel- Given the challenges discussed in chapter 1 evant to specific types of water resources activi- and the World Bank water-related activities ties within World Bank sectors and themes. Next annotated in chapter 2, this chapter turns to it addresses the issue of data availability. Lastly, the data requirements and characteristics of a the changing character of operational hydrology range of water resources activities. It identi- is briefly reviewed, concluding with a look ahead. fies the hydrometeorological data that each specific kind of activity could use or benefit from if those data were available and links KEY DATA those activities to water-related sectors and subsectors of World Bank projects. Other Table 4A.1 in annex 4A (available online) pres- types of data such as land cover, land subsid- ents the hydrometeorological variables deemed ence, and topography are included in this crucial for a specific water resources activity. characterization, as they are relevant to hydro- The 17 variables considered in this analysis are logic applications as well. For simplicity’s sake, precipitation, temperature, evapotranspiration this chapter refers to all of these variables as (ET), normalized difference vegetation index hydrometeorological variables. (NDVI), streamflow, soil moisture, wind speed, Tables 4A.1 and 4A.2 in annex 4A (available groundwater recharge, groundwater level, online at https://openknowledge.worldbank.org/ surface water level, snow or ice cover, snow or ­ handle/10986/22952) give a detailed character- ice water equivalent, land cover change, pump- ization of the key hydrometeorological variables ing and groundwater change, land subsidence, that are necessary for various water resources elevation, and water quality. The annex gives a activities, ranging from policy and planning to brief description of these variables and explains design, operations, and disaster management. how to read table 4A.1. The variables that can   49 be estimated using Earth observation (EO)— hydropower potential of available water precipitation, evapotranspiration, soil mois- resources ture, vegetation cover, groundwater, surface • Design irrigation. Design water extrac- water, snow, and water quality—are described tion and efficient distribution through an in more detail in part II. irrigated area to satisfy crop water needs; topography, soil moisture, and evapotrans- Water Resources Activities, Sectors, piration are important variables in this and Themes design process “Water resources activities” are defined as the key efforts or tasks related to the planning, • Design wastewater. Design systems that design, operation, management, administration, will efficiently collect gray and black waters and governance of water resources. These activ- from the point of use and treat them to the ities can be relevant to the water sectors, sub- desired standard of quality before reusing sectors, or themes considered by the Bank in its or releasing them into the environment operations. In the absence of readily available • Design water supply systems. Design sys- information on the specific hydrometeorologi- tems that will efficiently supply water of cal variables used in these projects,1 the listed the desired quality for drinking and other activities constitute an implicit link between the uses in a specific area key data in table 4A.1 and the water-related Bank portfolio review presented in chapter 2. • Disaster management. Mitigate the risk The following activities are considered in and impact of disasters and manage disas- this publication. They are based on informa- ters when they occur tion derived from the portfolio review • Energy (other than hydropower). Develop described in chapter 2. a mix of energy sources, including eolic • Comprehensive spatial planning and (wind), solar (equipment production and land management. Plan how to use land electricity generation), biofuel (irrigation), resources to accommodate current and nuclear (refrigeration), and thermal energy future societal needs; design infrastruc- • Food security and crop monitoring. Monitor ture and allocate land to satisfy the need food production and availability, reduce for habitation, recreation, environment, human vulnerability, and operate in a industry, water, and energy transport framework that makes it possible to issue • Design environment. Develop designs that alerts and adopt mitigation measures in allow the preservation of natural ecosys- case of food insecurity or famine tems in human-intervened systems • Forest management. Manage for aes- • Design flood control. Develop designs that thetics, fish, recreation, urban values, contain floods up to a certain magnitude water, wilderness, wildlife, wood prod- and associated probability of occurrence; forest genetic resources, and other ucts, ­ use available data to calculate maximum purposes (timber extraction, planting ­ probable precipitation and maximum and replanting of various species, cutting probable flood; design hydrograph and roads and pathways through forests, and infrastructure accordingly—by safely rout- preventing fire) ing and dampening the event • Health issues. Plan, monitor, and manage • Design hydropower. Design hydropower vector-control issues, water quality, pollu- production facilities adapted to the tion sources, and others 50  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k • Marine and estuarine environments. Main- • Water resources planning. Develop future tain ecosystem services (storm protection, plans for managing water resources and wildlife and biodiversity preservation, necessary infrastructure, practices, and water treatment, fisheries, recreation, and regulations other purposes) • Water resources policy. Assess the dynamics • Operations environment. Manage to allow of water budget and resource availability in the preservation of natural ecosystems in order to issue policies to guide sustainable human-intervened systems management • Operations flood control. Operate reservoir • Water resources management. Satisfy the systems to regulate and dampen flood peaks demands for water and balance water and route them safely through the system demand and supply by implementing exist- ing policies • Operations hydropower. Maximize hydro- power production, while accounting for • Water resources strategy. Translate the other constraints principles of policy and politics into spe- cific actions, recognizing that strategy lies • Operations irrigation. Extract and distrib- between policy and planning ute water efficiently through irrigated areas to satisfy crop water needs • Watershed management. Manage soil and water conservation practices and manage • Operations wastewater. Treat gray and land use and land cover to ensure con- black water flows being produced in the tinuing ecosystem services and resource system, reuse treated water, and dispose of availability residues in an efficient way • Weather monitoring. Use current-state • Operations water supply systems. Operate variables and continuity and other equa- and ensure water supply to a specific level tions to predict future-state variables and of reliability future weather • Terrestrial and freshwater ecosystems. • Urban design and management. Design in a Maintain ecosystem services (flood pro- way similar to comprehensive planning but tection, wildlife and biodiversity preser- within a city. vation, water treatment, recreation, and other purposes) Hydrometeorological Variables The 17 variables selected are important for • Transboundary issues. Understand the understanding the hydrologic cycle in a spe- dynamics of use of transboundary water cific basin or region, for quantifying the avail- bodies, rivers, and aquifers, as well as the ability of water resources in space and through impacts of different uses on other regions seasons and years, and for understanding the and nations; use that knowledge and effects of human extraction and use. Water understanding in designing agreements allocation, water permits, and water use— and managing allocations to implement related to both surface water and groundwa- them coherently ter—should be based on an understanding of • Water resources administration. Assign regional water availability as well as the water rights and implement court rulings requirements of ecological flows and aquifer on disputes, pumping quotas, water fee levels. Sustainability is achieved when water collection, insurance plans, and the like uses do not jeopardize either the ability to C h apt e r 4 : K e y Data N e e d s f or G oo d W at e r M a n ag e m e n t   |  51 maintain the same level of use in the long term (given meteorological forcings and intercep- or the functionality of ecosystems to continue tion, that is, water intercepted by plant, leaf, offering the same level of ecosystem services. and branch surfaces), direct runoff, and Understanding the hydrologic cycle and the infiltration. Infiltrated water can recharge an ­ dynamics of water availability in space and aquifer or subsurface flow, remain in the soil time is essential to making such assessments. as moisture, or return to the atmosphere via The variables considered in this publication evapotranspiration. Vegetation cover (mea- were selected on the basis of their relevance sured with NDVI) is a very significant factor for a range of water resources activities in for transpiring soil moisture back into the World Bank water-related sectors, subsectors, atmosphere and influencing infiltration in and themes. The selected variables were com- times of rainfall. pared with other classifications of EO variables Direct runoff and subsurface flow may all deemed relevant to water management. For contribute to river streamflows, as well as example, the Group on Earth Observations baseflows, when the aquifer system is con- carried out an extensive review of user require- nected to the river. In these cases, groundwa- ments for critical water cycle observations ter levels are important because riparian areas (Friedl and Unninayar 2010; Friedl and Zell and wetlands may depend on groundwater. 2010; see also Lawford 2014).2 All users in that Groundwater recharge determines the rate at assessment ranked precipitation and soil mois- which aquifers are replenished; in natural ture observations as, respectively, the first and environments, it is determined by the sum of second most important variables across soci- contributions to river baseflow, evapotranspi- etal benefit areas. ration from connected riparian ecosystems, The selected variables include all of the 15 and groundwater flowing down the gradient variables of perceived priority at the global out of the region. Human pumping of ground- level, although slight differences exist between water captures flows that would otherwise the scope of the variables in Lawford (2014) become baseflow, riparian evapotranspiration, and in this publication—for example, Lawford or groundwater outflow or deplete the aquifer (2014) distinguishes between evapotranspira- storage (lowering groundwater levels and tion in lakes and wetlands as separate from inducing land subsidence). other kinds of evapotranspiration, between The extent of snow cover and snow water streamflow and river discharge to oceans, and equivalent are also important variables, as they between snow cover and glaciers. For practical reduce ET losses, increase soil infiltration, and purposes, and given its utilitarian nature, this constitute a source of water storage that regu- publication does not distinguish between dif- lates river flows through the spring and into ferent types of evapotranspiration or stream- the dry seasons on most continents. flow or between ice or snow cover and snow As the many variables that influence and water equivalent. Consequently, the variables govern the hydrologic cycle are spatially het- included in this publication coincide with erogeneous and vary in time, hydrologic mod- almost all of the “primary” essential water els are essential to understanding the variables as defined in Lawford (2014). hydrologic system and identifying the main All of these variables are highly relevant to drivers of its hydrologic behavior. While hydrologic assessments. As precipitation rep- ground observations provide point measure- resents the water input into a basin, variables ments, limited in time, which have to be gener- such as temperature and wind speed influence alized for large regions, remote sensing (RS) ET rates. Land cover controls the partitioning estimates provide spatial information at of precipitation between evapotranspiration varying time frequencies. Thus the addition of ­ 52  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k RS estimates benefits hydrologic modeling, patterns of human water use and their cou- both through historical analyses and real-time pled dynamics. Unfortunately, the availability simulations, indirectly supporting water of data for planning and management pur- resources management and planning. poses is usually less than optimal. Sometimes the quality and characteristics of the data are Relevance of Variables for Each Activity, such that they do not allow a good analysis Sector, and Theme and interpretation. The variables of relevance to each activity can This section starts by discussing the chal- be derived by specifying the type of measure- lenges of data collection, the decline of ground ments of the variable—for point (­ground-level) observation networks, and how ground mea- measurements and areal (RS) measurements, surements stack up against satellite observa- respectively. For example, precipitation can tions. It then provides an overview of be measured at ground level with “rain gauge operational hydrology to date. Finally, it networks” and remotely with ground-based ­ discusses the future of operational hydrology “radar” and different kinds of “satellite” and provides insights into how to bridge the ­sensors. gap between management practitioners, on the Longer time series are always desirable, but one hand, and RS products and applications, table 4A.1 in annex 4A (available online) pro- on the other hand. vides two or three values that can be understood The collection of ground-based observa- as “minimum length required,” “adequate for tional records is usually the responsibility of use,” and “optimal length,” with the caveat that national water resources government agen- these are approximate values. Available data cies and regional hydrometeorological ser- lengths are almost always shorter than desired. vices. Other agencies and institutions Finally, table 4A.2 in annex 4A (available (ministries of energy, agriculture, health, and online) provides an initial characterization of transportation) and sometimes even the existing data for the variables of interest. This private sector may also collect water-related ­ includes some insights into accessibility of data. Given the local nature of hydrometeoro- records, approximate time-series lengths of logical observations and the fact that the col- potentially available records, time intervals, lection and storage of data in central, national and an explanation of how specific variables archives are often not systematic, data are fre- are measured or estimated (based on quently kept in a fragmented way. In addition, RS-derived data). A detailed explanation of as data are seen as essential information on how values for precipitation, evapotranspira- resources and, thus, a source of power, public tion, soil moisture, vegetation cover, ground- or third-party access may be restricted or dif- water, surface water, snow, and water quality ficult to obtain. This applies especially to ­ stimated with Earth observation is given are e transboundary basins, whenever there is a in chapter 6 of this publication. conflict or tension regarding the allocation of water among the member states, as is the case of the Nile River. AVAILABILITY OF DATA As to data needs, a distinction can be made between the need for (a) long-term data To be able to manage something, one must records for strategic policy, planning, and know what it is that needs to be managed. design and (b) real-time data for monitoring Sustainable use of resources and resilient sys- and forecasting to serve operational manage- tems need data and a thorough understanding ment purposes as well as short- and medium- of natural hydrologic processes as well as term decision making. C h apt e r 4 : K e y Data N e e d s f or G oo d W at e r M a n ag e m e n t   |  53 Some of the factors hindering the use of agencies, the lack of training and capacity data from ground-based monitoring networks building for collecting data and managing data- for water resources operations and planning bases, and inaccessibility of measurement are the lack of real-time data and accessibility, locations due to logistical problems, safety ­ coupled with quality control. issues, and conflicts. Very few observation stations are equipped International organizations attempt to with telemetry systems that allow data to be over­ come these challenges by promoting transmitted in real time, resulting in a lack of ­ cooperation, data sharing, and capacity build- available real-time data. Most recorded data ing. Many initiatives have been developed for only become available after several days, that purpose, often aimed at a specific variable weeks, or months, with a lot of readings losing or type of data. The International Groundwa- most of their value for operational hydrology ter Resources Assessment Centre (IGRAC), for purposes. Many new networks in developed example, aims to assess global groundwater countries are being implemented with teleme- resources and share the information through a try capabilities, and crowd-sourcing efforts for centralized system.3 The Global Runoff Data hydrometeorological observations are under Centre (GRDC) is a repository of global stream- way. However, the situation in developing flow data records and can be accessed online.4 countries is direr, as the number of observation A more recent initiative attempts to stations is being reduced and existing ones are ­connect and link existing efforts and networks not being properly maintained. observing all types of hydrometeorological Data are not easily shared across agencies, variables. Established in 2001, the Global much less between nations, and they are not Terrestrial Network–Hydrology (GTN–H) even easily bought, as the data owners some- links existing networks and systems into a times use them to gain leverage or power. network of networks for integrated observa- While some national hydrometeorological tions of the global water cycle (figure 4.1).5 services in developing countries may consider ­ The GTN–H is a joint project of the Global data a source of revenue to relieve their diffi- Climate Observing System, the Climate and cult budgetary situation, in general, data are Water Department of the World Meteorologi- seen as both a public good and a strategic cal Organization, and the Global Terrestrial resource. Observing System. It is the largest association Many data records lack quality assurance of international hydrometeorological data and quality control, especially in developing centers and users worldwide. countries. Data records can have many flaws, In addition, the World Bank has a freely often due to the absence of a systematic data- available databank, which contains records on a retrieval methodology (data are collected broad range of topics and fields related to eco- only seasonally or when judged necessary, in nomic development, and a Climate Change the rainy season for instance), operator error Knowledge Portal.6 The portal contains histori- (missing data lead to data gaps or invented ­ limate cal data and model projections of future c data values), and deficient archiving (read- under different climate change scenarios. The ings are not properly referenced in time and historical data include temperature and precipi- space). In addition, data records from differ- tation records from observational stations par- ent locations may be very heterogeneous, due ticipating in the Global Historical Climatology to the lack of systematic procedures or Network and merged station-satellite historical ­uniform standards. data records from the U.S. National Centers for These problems in developing countries are Environmental Prediction. caused by several factors, among others, the The Food and Agriculture Organization also lack of proper funding for hydrometeorological has a freely available online database called 54  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k Figure 4.1 Components of the Global Terrestrial Network–Hydrology, 2013 Global Global Precipitation Runoff Climatological Data Centre Center GEMS/ Water River Precipitation discharge* Water Water quality International vapor* and BGC Groundwater FLUXNET fluxes Resources Assessment Evapo- Centre transpiration Atmosphere Groundwater* Food and Agriculture Water Hydrolare Organization use* Aquastat Lake levels, areas, and temperatures* Hydrosphere Snow cover, Centre National WGMS glaciers, and d’Études Spatiales/ ice caps* Legos Isotopic Soil composition moisture* National Snow and Ice GNIP Data Center ISMN GNIR Variable/ * GCOS Essential Climate Variable Global network/coverage defined and contact established Global network/coverage partly existing/identified and/or contact to be improved No global network/coverage identified Source: © World Meteorological Organization (http://gtn-h.unh.edu/). Used with permission. Permission required for further reuse. Note: GCOS = Global Climate Observing System; BGC = biogeochemical global climate (models); CNES = Centre National d’Études Spatiales; FAO = Food and Agriculture Organization; GEMS = Global Environmental Monitoring System; GNIP = Global Network of Isotopes in Precipitation; GNIR = Global Network of Isotopes in Rivers; GPCC = Global Precipitation Climatological Center; GRDC = Global Runoff Data Centre; IGRAC = International Groundwater Resources Assessment Centre; ISMN = International Soil Moisture Network; NSIDC = National Snow and Ice Data Center; WGMS = World Glacier Monitoring Service. AQUASTAT, developed by the Land and Water Decline of Ground-Based Observation Division.7 The main database provides five- Networks year averages for up to 70 variables, by country. The number of global, ground-based hydro­ Other databases have information on dams, meteorological observations has gradually institutions (by country), sediment yields in decreased since the 1980s (figure 4.2; rivers, water-related investments in Africa, and ­Shiklomanov, Lammers, and Vorosmarty 2002; irrigation investments around the world. Stokstad 1999). This is due to a combination of C h apt e r 4 : K e y Data N e e d s f or G oo d W at e r M a n ag e m e n t   |  55 Figure 4.2 Availability of Historical Monthly and Daily Discharge Data in the Global Runoff Data Centre Database, 2004 and 2014 a. Monthly data b. Daily data 2010 2000 1990 1980 1970 1960 1950 1940 1930 1920 1910 1900 8,000 7,000 6,000 5,000 4,000 3,000 2,000 1,000 0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 Number of stations Number of stations 2004 2014 2004 2014 Source: © GRDC (Global Runoff Data Centre). Used with permission. Further permission required for reuse. Note: While the historical size of the archive at the GRDC increased substantially between 2004 and 2014 in terms of both the number of stations and the volume of data available for the historical period, the number of available stations and data has declined since the 1980s. This decrease is due to several factors: (a) a decline in the number of monitoring stations; (b) long quality assurance process times; (c) lack of data sharing by country agencies; (d) increased operation of monitoring infrastructure by hydropower companies, which do not share the data due to its strategic value; and (e) decentralization of management and monitoring responsibilities, which multiplies the number of agencies that the GRDC has to interact with in order to obtain data updates (Ulrich Looser, GRDC head, personal communication). factors including budget constraints and the other locations to estimate variables in the ensuing lack of maintenance and operators as basin of interest. Remote sensing can also be well as the existence of political turmoil and used to monitor ungauged basins. The Predic- conflicts that sometimes destroy gauges, pre- tion in Ungauged Basins Initiative (Pomeroy, vent readings, or halt funding altogether. Even Whitfield, and Spence 2013; Seibert and Beven in stable, first-world countries, spending bud- 2009; Sivapalan 2003; Wagener and Montanari gets for in situ monitoring have shrunk, despite 2011) is a good example of efforts made to the call of the Intergovernmental Panel on Cli- overcome the problem of lack of data in mate Change for more in situ measurements ungauged or poorly gauged basins. Generally, (IPCC 1991). The discontinuation of readings prediction efforts for ungauged basins need in stations with long time series entails the loss data from gauged basins with similar charac- of “climate memory,” at a time when long-term teristics to create analogies based on hydro- records are becoming critical to documenting logic modeling, frequency analysis, statistical and understanding ­climate variability and change. correlations, parameter regionalization, and remote sensing. In general, such approaches Ungauged Basins allow for the characterization of hydrologic In ungauged basins, predictions are still possi- regimes, their variability, and tentative predic- ble through several approaches. Regionaliza- tions in ungauged basins, although the latter tion is a technique that attempts to fill the void are associated with significant levels of uncer- of missing data by using information from tainty due to their indirect nature. 56  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k It is difficult to predict whether the number (level of development for different water uses) of ground monitoring observations around the would best achieve the objective. Maas et al. world will continue to decline or start to rise in (1962) describe the program’s major accom- the near future, although it is likely that large plishments. Many of its methods for evaluating areas of the world will remain poorly gauged. and ranking design alternatives based on The decline of existing ground-based networks ­ economic efficiency, given a hydrologic con- in regions across the world has left satellite text, are still in use today. Using ground obser- observations to fill this void. However, it is a vational records, the planning and design of fallacy to think that the latter can just substi- infrastructure as well as management policies tute ground observations. Integrating and made use of statistical methods involving comparing measurements from the ground ­ stochastic hydrology, frequency analysis, and from space are very necessary but complex ­ probability distributions, and extreme values. and challenging tasks. Evolving from the narrow cost-benefit anal- ysis through the early inclusion of environ- mental considerations in management and OPERATIONAL HYDROLOGY planning—that is, the principles and standards TODAY of the Water Resources Council (1983)—the principles of international water resources Operational hydrology is the range of activities management are reflected in most regulatory attempting to measure and understand the frameworks of developed and some develop- water balance components for use in direct ing countries; the same ground observational practical applications of planning, design, and records are being used in the same way for management of water resources. This section hydrologic and hydraulic considerations. provides an overview of the current state-of- Data coverage and access are poor in many the-art of operational hydrology, focused spe- regions and tend to cluster around large infra- cifically on developing regions, and how it is structure projects that bring in the resources informing water management, planning, and for reliable monitoring networks, rather than water resources activities in general. around other, less costly initiatives in water Ground-based observation time series have supply and sanitation and in irrigation. With been and still are the rule for operational comprehensive planning and the integration hydrology, and RS hydrometeorological vari- of large infrastructure projects with other ables are rarely, if ever, used operationally— efforts, monitoring networks could be made and not merely experimentally or as relative better available to inform a range of water guidance—to support decision making. ­ resources activities. Present-day water resources planning, In developing regions of Africa, reservoirs design, and operations of hydraulic infrastruc- and other infrastructure are operated with ture and management are based on and have effective, traditional tools such as rule curves evolved from the Harvard Water Program developed from historical records. Opera- (1955–60). In that program, academicians and tions rarely incorporate real-time prediction senior federal and state agency employees data into the decision-making process. worked together on research and training for Regional centers focus on hydrometeorologi- designing and planning water resources sys- cal and agricultural research, such as tems. Tools and methods were developed that, AGRHYMET, with 9 member states in West given a certain planning objective, could Africa; the C ­ limate Services Center (CSC) of ­ determine what set of structural measures, the Southern Africa Development Commu- operating procedures, and water allocations nity (SADC), with 15 member states; the C h apt e r 4 : K e y Data N e e d s f or G oo d W at e r M a n ag e m e n t   |  57 Figure 4.3 Seasonal Forecasts Issued by Two Regional Centers act as a repository of data, although the data in 2013 are not easily shared, as they are still owned by the individual member states. One of their a. Seasonal rainfall forecast 30 main research efforts is to monitor hydrocli- matic conditions and drought and to prepare 25 seasonal forecasts. 30 Latitude (degrees) 20 40 30 45 35 20 It is also fairly common for these centers to 20 15 45 35 hold seasonal climate outlook forums with the 10 20 participation of collaborating institutions, in 35 40 40 45 20 LEGENDE S Supérieur which they integrate all of the climate informa- 5 N Normal tion available and produce a seasonal forecast I Inférieur 0 –20 –15 –10 –5 0 5 10 15 20 25 30 for the incoming rainy season for their region. Longitude (degrees) For instance, AGRHYMET integrates informa- b. Seasonal streamflow forecast BV Senegal tion from Columbia University’s International Prévision saisonnière hydrologique 2013 Research Institute for Climate and Society, the BV Gambia BV Niger BV Oubangui-Chari U.K. Met Office, Météo-France, the World BV Komadougou Lac Tchad BV Volta Meteorological Organization, the ­ African BV Comoé BV Bandama 35 BV Sassandra ­ Centre of Meteorological Applications for 35 40 BV Ouemé 40 BV Mono. 35 25 35 35 40 25 40 25 35 40 25 35 40 25 Development, agencies from regional member 40 35 25 25 40 25 35 35 40 35 states, and river basin organizations. 35 25 35 40 40 40 25 40 25 25 Forecasts are produced by assimilating dif- 25 ferent types of information based on sea sur- c. Seasonal rainfall forecast face temperatures and other climate data and 25 are usually issued two consecutive times, as 40 1 35 the rainy season approaches. Using a simple 40 3 2 35 11 format, the forecasts give the probabilities of 25 40 35 having an average, above-average, or below- 25 4 average rainfall in specific regions. Two fore- 7 10 casts, issued in 2013 by AGRHYMET-CILSS 8 40 5 35 9 25 (Permanent Inter-State Committee for 35 6 40 25 40 25 35 Drought Control in the Sahel) and SADC, respectively, are shown in figure 4.3. Being a probabilistic forecast, whatever the volume of Source: AGRHYMET Regional Center 2013. Panels a and b: © AGRHYMET Regional Center. Used with permission. Further permission required for reuse. Panel c: © actual rainfall is, the forecast is never wrong, Southern African Development Community (SADC)–Climate Services Centre. Used and its accuracy can only be assessed in the with permission. Further permission required for reuse. Note: Panels a and b show forecasts issued in 2013 by AGRHYMET-CILSS for long term. West Africa, Chad, and Cameroon. Panel c shows a forecast issued by SADC for A key question is how these forecasts are Southern Africa. The groups of three numbers on the panels represent probabilities in percentage of above-average (supérieur), average (normal), and below-average used, or put in practice, by the member states (inférieur) rainfall (a, c) or streamflow (b), respectively, for the region. The individual and the practitioner community. In two work- numbers in (c) represent areas with homogeneous rainfall. shops in Africa organized by the United Nations Educational, Scientific, and Cultural Intergovernmental Authority on Develop- Organization with representatives from ment (IGAD) in Eastern Africa’s Climate Pre- AGRHYMET, ICPAC, and SADC-CSC, the diction and Application Centre (ICPAC); and answer to this question remained elusive. the Regional Center for Mapping of Resources The representatives of the regional centers for Development. Many of these centers could did not know how the member states used the 58  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k forecast, and neither did the representatives In Asia, such regional climate outlook of the member states themselves. For several forums are much more recent, having started African basins, what may be lacking is an ­ in 2009 for South Asia and in 2013 for South- overall basin management plan with a clear east Asia. Their functioning and outputs are decision-­making process based on monitoring very similar to those described above. and observations. The Regional Integrated Multi-Hazard Similar hydrometeorological networks Early Warning System for Africa and Asia exist in Central America, South America, and (RIMES) is hosted by the Asian Institute of Asia.8 In Central America, more than 40 cli- Technology in Thailand and represents a matic forums have been held to date. These ­consortium of 31 member states, mainly around forums provide climate outlooks for the next the Indian Ocean and Central and Southeast- three months, and they are usually organized ern Asia, as well as regional international orga- three times a year—in the summer of the nizations and universities.9 The governing Northern Hemisphere at the beginning of the council is composed of “heads of National rainy season, at the end of the rainy season, Meteorological and Hydrological Services and and at the end of the year when cold fronts national scientific and technical agencies gen- arrive. The Central America Climate Forum is erating multihazard early warning informa- a working group directed by the Regional tion.” The key services are (a) earthquake and Committee of Water Resources of the Central tsunami watch provision; (b) weather, climate, American Integration System Secretariat, and hydrologic research and development; and with the participation of national hydromete- (c) capacity building in end-to-end early orological services, universities, private enti- warning. ties, and other Central American institutions (García 2014). The climate outlooks estimate the plausible precipitation and temperature, FUTURE OF OPERATIONAL obtained by statistical methods, compare the HYDROLOGY: TRANSLATING DATA estimates with analogous years, and analyze INTO INFORMATION results from global and regional models regarding sea surface temperatures and distri- The future of operational hydrology depends butions of wind, pressure, and precipitation. on the ability to extract relevant information These outlooks are intended to complement from the abundance of data from different the projections from the meteorological ser- sources with different degrees of accuracy and vices of individual member states. Once the precision and to use it for specific decision- forecasts have been published, working groups making purposes. Given the increasing of specialists from different sectors use them to amounts of RS data available and current tele- make recommendations for their sector, to communication capacities, it can be difficult prepare for the possibility of “above-normal,” for a manager to know what data sources to “normal,” or “below-normal” conditions. The use or trust and how to combine different types sectors concerned are agriculture, fisheries of information. For any particular planning, and aquaculture, health and nutrition, water design, or management decision, it will be and sanitation, risk management, and energy. essential to distill only the relevant informa- In addition, the reports suggest that each tion from all of the available data. member state reissue more specific recom- Chapter 7 of this publication provides a series mendations taking into account the particular of guidelines to help decision makers to decide contexts of each nation for each sector (SICA whether Earth observation may be useful and, if and OBSAN-R 2011). so, to choose the most suitable EO data sources. C h apt e r 4 : K e y Data N e e d s f or G oo d W at e r M a n ag e m e n t   |  59 It is also necessary to characterize the errors and they want to test it, but there is no model and uncertainty contained in hydrometeorolog- that will work uniformly well everywhere.” ical estimates, as well as in  data merged from This comment illustrates the need to com- different sources. Chapter  8 provides insight pare products and models. Given a specific into the accuracy and v ­alidation of the most need, what are the trade-offs between using common EO-estimated h ­ydrometeorological “simple” and complex models and between variables. The combined use of ground observa- using one set of input data or another? What tions and RS estimates in an integrated manner, models perform best for what purposes (flood specifying the uncertainty bounds on final prod- forecasting, low flow estimation, forecasting for ucts, guarantees that the best possible use will reservoir operations, irrigation, drought moni- be made of existing resources. Bayesian toring)?11 The characteristics of an RS applica- approaches assimilating different types of data tion for flood forecasting in a context where the with associated resolutions and uncertainties main considerations are short time steps, quick are appropriate for such purposes. response times, and accurate p ­ rediction of peak In addition, producing the best possible flows exceeding a certain magnitude will differ estimates by integrating different types of mea- strongly from those of an application to support surements can be tailored to specific manage- reservoir operations, which will be geared ment and decision-making purposes. What toward accurately p ­redicting water volumes will this information be used for, beyond scien- over longer time steps. The suitability or per- tific and research purposes (which are what formance of a specific application can only be most space missions are currently geared evaluated against a specific purpose. toward)? In other words, managers and deci- Whether an application can be evaluated sion makers need to have a detailed, specific depends on whether it can be calibrated and answer to the following questions: What type then validated. Even in ungauged basins, of information do you need to support your biases in rainfall12 and other variables can be decision-making process? How will you corrected—based on observations from neigh- change your decisions based on different fore- boring basins or regions—and these two data casts or information? While these questions sets can then be compared to ensure that the are two sides of the same coin, they engage dif- estimates to be used lie within an acceptable ferent thought processes. During the Pakistan range. The usability of RS products for deci- floods in 2010 (Mendoza et al. 2010), the infor- sion making and planning is determined by mation was there, but the mechanisms to act questions revolving around the degree of on it were not reliable. Other cases show that uncertainty (error estimates), accuracy (the tailoring information to management purposes extent to which errors are characterized), can also be a challenge in the developed world. precision (spatial and temporal resolution), Sometimes the specific tasks required to and timeliness of the data available (for use in attain the overall goals of water management near real time or as historical data). agencies—or the means by which they should If products are used as inputs for modeling be developed—are poorly defined.10 If the applications, it is important to know how errors management tasks and specific decisions are propagated through model calculations and required are well defined, tools can be tai- how uncertainties are compounded through lored to inform those decisions. It is logical model cascades. For example, due to the nonlin- that different models and applications will be earity of rainfall-to-runoff transformation and needed for different purposes and different the spatial variability of rainfall over a basin, questions. In speaking of hydrologic models, relative errors in satellite-derived precipitation an AGRHYMET official once acknowledged, estimates tend to be magnified in the value of “Everyone comes here with their own tool the flood peaks (Nikopoulos et al. 2010). Thus a 60  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k good understanding is needed of how a specific 5. For information on GTN–H, see http://gtn-h.unh application propagates input errors to the out- .edu/. 6. The databank is accessible at http:// put variables. Next, applications can be tested data.worldbank.org/. The Climate Change for reliability: How many times did the observa- Knowledge Portal is accessible at http:// tions fall within the uncertainty bounds of each sdwebx.worldbank.org/climateportal/index application’s predictions? .cfm?page=climate_data. Acknowledging the limitations of each 7. AQUASTAT is accessible at http://www.fao.org /nr/water/aquastat/main/index.stm. application and being transparent and up-front 8. A complete listing is available on the World about the uncertainty in its output variables Meteorological Organization’s website (http:// ­ form the basis of applicability. Part III of this www.wmo.int/pages/prog/wcp/wcasp/clips publication presents the results of a (limited) /outlooks/climate_forecasts.html). literature review regarding the validation and 9. For information on RIMES, see http://www .rimes.int/. accuracy of the most common EO-estimated 10. An ongoing study of the Water Global Practice hydrometeorological variables. funded by the Water Partnership Program and the Global Facility for Disaster Reduction and Recovery, in collaboration with the World Meteo- ANNEX 4A. KEY DATA FOR WATER rological Organization, tries to assess the current status of the national ­ meteorological and hydro- RESOURCES MANAGEMENT ­ ifferent regions of the world. logical services in d This study may identify needs for future support Annex 4A is available online at https://open in strengthening their c ­ apabilities to include knowledge.worldbank.org/ handle/10986 demand-driven activities in their operations. 11. The Bank’s Water Partnership Program held a /22952. “Flood Model Showcase” workshop in W ­ ashington, DC, on September 23–24, 2014, to present common flood problems and various ­ models and tools that NOTES could be used to inform decision making. The aim exposure and vulnerability to flood- was to reduce ­ 1. Which hydrometeorological variables, including related hazards and enhance understanding of those listed in table 4A.1, were considered in any each model’s “best” application, with an emphasis particular Bank operation is not specified in the characteristics of the information each tool on the ­ project portfolio databases. can provide. The report of the workshop is under 2. The initial report (Friedl and Unninayar 2010) preparation. sought to identify the priorities for Earth 12. Most RS estimates yield consistent observation from the user’s perspective in order to ­ underestimations or overestimations of a inform future EO strategy. It considers user classes variable with respect to its measured value on categorized by type and function. Major groups the ground. Bias correction techniques can help that use water information for decision making to remove these systematic biases. For example, were identified, and then a broad range of applica- satellite precipitation products consistently tend tions was identified within each of these groups. to overestimate rainfall in the tropics, as they Based on these categories, a list of EOs for the “observe” it well above the ground surface, while water social benefits area was generated for three some of the rainfall is likely to evaporate before spatial perspectives: global, regional, and local. reaching the ground. Bias correction helps to Of 45 observational types of variables i ­ dentified reconcile these estimates with direct ground as being useful for water-related decisions, measurements, as long as the biases are consistent variables with a perceived priority at the global 15 ­ over time. level were used to identify the most critical EO priorities across all social benefits areas. Lawford (2014, table 4) displays the list of variables based REFERENCES on extensively reviewed user needs for water data. The final report is by Friedl and Zell (2010). AGRHYMET Regional Center. 2013. Bulletin spécial 3. For information on the IGRAC, see http://www sur la mise à jour des prévisions des caractéristiques .un-igrac.org/. agro-hydro-climatiques de la campagne 4. For information on the GRDC, http://www.bafg d’hivernage 2013 en Afrique de l’Ouest, au Tchad et .de/GRDC/EN/Home/homepage_node.html au Cameroun 23 (3, July). C h apt e r 4 : K e y Data N e e d s f or G oo d W at e r M a n ag e m e n t   |  61 Friedl, L., and S. Unninayar. 2010. “GEO Task US-09- Pomeroy, J. W., P. H. Whitfield, and C. Spence, eds. 01a: Critical Earth Observations Priorities; Water 2013. “Putting Prediction in Ungauged Basins Societal Benefit Area.” Final SBA Report, Group into Practice.” Canadian Water Resources Asso- on Earth Observations, Geneva. http://sbageotask ciation, Ottawa. .larc.nasa.gov/Water_US0901a-FINAL Seibert, J., and K. J. Beven. 2009. “Gauging the .pdf. Ungauged Basin: How Many Discharge Friedl, L., and E. Zell. 2010. “GEO Task US-09-01a: ­ Measurements Are Needed?” Hydrology and Critical Earth Observation Priorities; Final Earth System Sciences 13 (June): 883–92. Report.” Group on Earth Observations, Geneva, Shiklomanov, A. I., R. B. Lammers, and C. J. October. http://sbageotask.larc.nasa.gov. Vorosmarty. 2002. “Widespread Decline in García, L. E. 2014. Personal communication. Hydrological Monitoring Threatens Panarctic IPCC (Intergovernmental Panel on Climate Research.” EOS Transactions 83 (2): 16–17. Change). 1991. The First Assessment Report of SICA (Sistema de Integración Centroamericana) and the Intergovernmental Panel on Climate Change. OBSAN-R (Observatorio Regional de Seguridad Cambridge, U.K.: Cambridge University Press. Alimentaria y Nutricional). 2011. XIII Foro de Lawford, R. 2014. The GEOSS Water Strategy: From Aplicación de los Pronósticos Climáticos a la Observations to Decisions. Geneva: Group on Seguridad Alimentaria y Nutricional: Perspectivas ­ Earth Observations. para el período mayo-julio 2011. Tegucigalpa, Maas, A., M. M. Hufschmidt, R. Dorfman, ­ Honduras: SICA and OBSAN-R. H. A. Thomas, S. A. Marglin, and G. Maskew Fair. Sivapalan, M. 2003. “Prediction in Ungauged Basins: 1962. Design of Water-Resource Systems, New A Grand Challenge for Theoretical Hydrology.” ­ Techniques for Relating Economic Objectives, Hydrological Processes 17 (15): 3163–70. Engineering Analysis, and Government Planning. Stokstad, E. 1999. “Scarcity of Rain, Stream Gages Cambridge, MA: Harvard University Press. Threatens Forecasts.” Science 285 (5431): Mendoza, G., J. Giovannettone, A. Willis, M. Wright, 1199–200. and E. Stakhiv. 2010. “Assessment of Pakistan’s Wagener, T., and A. Montanari. 2011. “Convergence August 2010 Flood: Interim Report.” Institute of of Approaches toward Reducing Uncertainty in Water Resources, U.S. Army Corps of Engineers, Predictions in Ungauged Basins.” Water Resources Alexandria, VA. Research 47 (6): W06301. Nikopoulos, E. I., E. N. Anagnostou, F. Hossain, M. Water Resources Council. 1983. “Economic and Gebremichael, and M. Borga. 2010. “Understand- Environmental Principles and Guidelines ing the Scale Relationships of Uncertainty for Water and Related Land Resources Propagation of Satellite Rainfall through a ­ Implementation Studies (Principles and Distributed Hydrologic Model.” Journal of Guidelines).” U.S. Water Resources Council, Hydrometeorology 11 (2): 520–32. Washington, DC. 62  |  P A R T I : W at e r a n d Eart h O b s e r v atio n s i n t h e W or l d Ba n k PART II Earth Observation for Water Resources Management Juan P. Guerschman, Randall J. Donohue, Tom G. Van Niel, Luigi J. Renzullo, Arnold G. Dekker, Tim J. Malthus, Tim R. McVicar, and Albert I. J. M. Van Dijk OVERVIEW Part II of this publication captures and expands the may arise in a specific context, such as that of the results reported in part I with the following aims: (a) World Bank. Chapter 6 discusses the state-of-the-art to connect World Bank needs in water resources in those areas and provides an overview of the perti- management (WRM) issue areas to the range of nent EO sensors with their respective specifications. products providing Earth observation (EO) informa- Chapter 7 then provides information on whether the tion regarding water resources; (b) to describe the use of Earth observation should be considered given current state-of-the-art of water resources–related the specific requirements for spatiotemporal data. It Earth observation and provide an overview of (cur- provides a simple decision framework for determining rent and future) EO sensors as well as measured how EO products might best be used to generate the water resources variables; and (c) to provide guid- required information and how to select the most suit- ance on how to decide whether Earth observation able EO data products for a specific WRM problem. may be useful for addressing a WRM issue and Moreover, it highlights guiding questions to ask, once approximate the likely accuracy of the variables EO options are deemed worth exploring for the WRM ­ estimated through Earth observation. issue at hand. To make it easier to navigate the material Part II may be read in two different ways, depend- presented in part II, chapter 7 includes a flowchart ing on the reader’s background. connecting all of the information. Persons new to monitoring and assessing WRM Those who are already familiar with the material areas using Earth observation may want to begin with covered in chapters 5 and 6 or who have sufficient chapter 5, which provides an idea of the issues that working knowledge of Earth observation may want to S T R U C T U R A L M E A S U R E S A G A I N S T T S U N A M I S   |  63 turn directly to the section of chapter 7 that is the WRM issues presented in c ­ hapter 5. If nec- relevant to a specific application (or, alterna- essary, the sensor-variable tables in chapter 5 tively, refer to the applications presented in and the section of chapter 6 covering the appli- appendix B). The guiding questions should help cation at hand may be of interest. Figure II.1 them to select the most appropriate solution to summarizes these two options. Figure II.1 Schematic Showing Two Possible Ways to Read Part II Science and World Bank Scope Potential solutions EO sensors Chapter 5 Chapter 6 Chapter 7 World Bank Verify most sensible Choose EO application application EO sensor(s) Note: Chapter 7 (green): deemed essential; chapter 6 (orange): deemed optional; chapter 5 (blue): sketches the World Bank context. EO = Earth observation. 64  |  E art h O bser v ation f or W ater  R esources M anagement CHAPTER 5 Earth Observations and Water Issues INTRODUCTION widely and range from local problems, such as the provision of drinking water or sewage This chapter provides an overview of the issue systems in a specific town, to larger-scale ­ areas related to water resources management challenges, such as the likely impacts of (WRM) in a given context, such as the World climate change on water availability in large ­ Bank, and discusses the data requirements for and often transboundary basins. addressing them, focusing on those variables that The sectors, subsectors, and themes that char- can be obtained or estimated through Earth acterize water-related operations in the Bank’s observation (EO). Besides the usual surface char- portfolio are described in part I (see also annex acteristics, such as topography, land subsidence, 2A available online at https://openknowledge and others mentioned in part I, Earth observa- .worldbank.org/handle/10986/22952). Each of tion can address eight key hydrometeorological these operations deals with particular water- variables1 relevant to WRM applications: precipi- related issues, which in some cases are common tation; evapotranspiration; soil moisture; vegeta- to more than one sector or subsector. Moreover, tion, land use, and land cover; groundwater; the sectoral classification identifies which part of surface water; snow and ice; and water quality. the economy is receiving support; it is used in part I as a convenient mechanism to identify the EO-RELATED WATER RESOURCES water-related activities in the Bank’s portfolio MANAGEMENT IN THE WORLD and to identify key hydrometeorological vari- BANK CONTEXT ables deemed necessary for each water resources activity. Part II focuses on issues, grouping sec- Part I assesses the activities funded by the tors and subsectors according to the nature of the World Bank to address the most challenging issue areas or topics they address (see box 5.1, water-related issues in the developing world. which was adapted from the Water Partnership The issues that those activities address differ ­Program’s classification).2   65 BOX 5.1 from measurements of thermal bands of the spectrum. Water-Related Topics and Subtopics Considered in the Soil moisture can be measured by sampling: World Bank Context weighing a sample of extracted soil, drying it, and weighing it again. The difference in weight • Water supply for rural or urban water users is the evaporated soil moisture. This approach •Sanitation and hygiene is the most direct method of measuring soil •Agricultural water management, in irrigation or in rain-fed agriculture moisture and has the least uncertainty, even •Water resources management and environmental ­ services, including though the sampling procedure and drying aquatic ecosystems, environmental flows, invasive aquatic plants, and water and climate change method can introduce errors. However, it is • Hydropower also very labor-intensive. Less direct methods of field measurement—for instance, using time-domain reflectometers—can increase the efficiency of soil moisture measurement but Similar to the classification of sector, subsec- require calibration and are more prone to tor, and theme, each of these topics deals with uncertainties (among other things, due to salt particular water-related issues, which, in some concentration and turbidity, as in this exam- cases, are relevant to more than one topic. For ple). Both field-based techniques only measure example, flood extent mapping and flood the conditions in a very small section of the ­ prediction are of interest to urban water supply, sample or around the sensor. environmental flows, and climate change. Except for the measurement of one- dimensional flows, such as river discharge, one of the main challenges of ground FIELD MEASUREMENT, EARTH ­observation networks is to capture the spatial OBSERVATION, AND MODELING variability of the variable being measured. A rain gauge measures the rainfall over a few Field-based or in situ measurements are gener- square centimeters. Usually, observations ally more direct than Earth observation—that from a few rain gauges are used, assuming is, they measure the biophysical variable of that they are representative of rainfall over interest using a measurement principle that the entire basin. These observations may be has fewer uncertainties and assumptions. The more or less accurate, depending on the measurements are, however, usually represen- extent of the storms, topography, and other tative of a smaller area than is observed by factors. Current satellite precipitation prod- ­ satellite sensors. For example, a ground rain ucts have resolutions usually ranging from gauge takes direct measurements of the rain 0.25° with an average value of rainfall for a that falls on it, while satellite estimates can be cell area of roughly 625 square kilometers to derived indirectly from durations of cloud top 0.04° with an average value of rainfall for a temperatures or more direct measurements of cell area of roughly 16 square kilometers. The rainfall between somewhere in the cloud and spatial footprints of the two types of observa- the ground surface. tions (ground versus satellite) are several Evapotranspiration can be measured directly orders of magnitude different, making direct using Eddy covariance methods (directly mea- comparisons difficult. suring relative humidity in ascending air flows) Earth observations by satellite-based sen- or evaporation pans (measuring water loss to sors, or satellite remote sensing (RS),3 can evapotranspiration), while remotely sensed overcome the problem of spatial representa- estimates of evapotranspiration are inferred tiveness and generally also provide continuous 66  |  P A R T I I : E art h O bser v ation f or W ater  R esources M anagement measurements in time. However, they often Active Passive (SMAP), launched on ­January 29, rely on indirect methods to derive the value of 2015, start providing new, more accurate esti- the biophysical variable of interest. For the mates, these still need to be validated, and the example of soil moisture, surface brightness reliability of new WRM applications needs to temperature measured by passive microwave be assessed.4 sensors is influenced by the soil moisture con- Field measurement and Earth observation ditions and can be used to estimate this impor- can complement each other to enhance and tant soil property. Yet using passive microwave overcome their respective weaknesses. How- sensors has some drawbacks: the spatial reso- ever, neither type of observation provides lution is coarse (depending on the sensor, direct information on the future or the past about 12–50 kilometers), only moisture in the (that is, before the observations were made). very top layer of soil (1–2 centimeters) affects Digital satellite remote sensing was first used brightness temperature, and vegetation and in the 1970s, but its use only became wide- surface water can confound the measurement. spread in the mid-1980s. Furthermore, neither The latter also facilitates its use for observing form of observation provides any direct infor- vegetation biomass and surface water, respec- mation on how specific interventions or tively (see chapter 6). scenarios might affect a variable of interest. ­ Thus observations from space need to be RS data or data products that blend remote analyzed, validated, and used in accordance sensing and ground observations are difficult with their limitations, as they can contain to read. Responding to the need for storing ­several types or errors. Sampling and measure- large amounts of gridded data over vast areas ment errors can occur due to the measurement and increasing periods of time, RS estimates of a variable in the wrong place (for example, are made available in files with binary, ASCII, rainfall at the cloud base instead of at the NetCDF, or other formats. These data files ground surface) and due to indirect estima- require programming skills (codes or software) tions and biases in measurement sensors, that are not necessarily available, much less resulting in errors in the magnitude of the rate widespread, in developing countries, in addi- being measured. These errors will be different, tion to hardware with a minimum computa- depending on the specific geographic and tional power. While visualizations and atmospheric setting. Satellite precipitation customized applications are often developed products have performed differently, depend- to make the reading of data more user friendly, ing on the type of rainfall mechanisms, topog- capacity building and perhaps additional strat- raphy, and geography involved. Soil moisture egies are needed to facilitate access to informa- estimates are influenced by the type of vegeta- tion contained in the data sets. tion and cloud cover and can contain large Efforts to produce data sets integrating errors. Thus case-by-case validation efforts are ground observational networks and RS obser- essential before applying them in real-world vations attempt to capitalize on the accuracy situations. and precision of point measurements in the Something similar applies to the observa- ground as well as the spatial representation tion of other variables. False alarms and missed provided by Earth observations. Products events are two other types of errors that are combining all available data in a region (that is, ­ difficult to correct without ground measure- rain gauge networks, radar, and satellite pre- ments or without complementary RS observa- cipitation estimates) into a gridded data set are tions. Even if new missions such as the Global the best possible representation at a specific Precipitation Measurement (GPM), launched spatiotemporal resolution of the true rainfall on February 27, 2014, or the Soil Moisture over the region, although they are not devoid C h apter 5 : E art h O bser v ations an d W ater  I ssues   |  67 of errors. Measuring and representing the used to estimate conditions in times when obser- “ground truth” accurately are still challenging. vations were not yet available or under varying Given the challenges of accurately captur- scenarios, although their outputs will only be as ing spatial variability, it is very difficult to pro- good as the physics and assumptions underpin- duce a spatially explicit “ground truth” ning them. Nevertheless, models represent our reference data set against which to compare best conceptual understanding of physical pro- satellite estimates. Ali, Lebel, and Amani cesses at any given time in history and provide (2005) demonstrate that errors of satellite insight into how components of the Earth sys- products in some settings are likely to be sig- tem interact. nificantly lower when the errors in gauge With the growing wealth of water informa- “ground truth” data and the covariance tion available from field networks, EO systems, between them are taken into account. An and computer models, much research in recent example of a data set integrating different years has been devoted to developing mathe- types of data is the Global Precipitation matical techniques and computing infrastruc- ­ Climatology Project’s One Degree Daily. An ture to bring the information together in ways exhaustive list of these types of data sets can be that enhance overall accuracy and utility found on the website of the International Pre- figure 5.1). Appendix B gives numerous exam- (­ cipitation Working Group.5 ples of experimental and operational systems These limitations can be overcome—to vary- that have exploited multiple data sets and infor- ing extents—with the aid of computer models. mation sources to improve the monitoring of These models can be predictive and can also be key water cycle variables, including merging Figure 5.1 Conceptual Depiction of Information-Integration Paradigm Referred to as Model-Data Fusion On-ground observations + relatively direct – sparse or infrequent – not predictive Biophysical models Satellite observations + predictive + full and frequent coverage + directly interpretable – relatively indirect + full and continuous coverage – not predictive – unhindered by reality Note: + = pros; – = cons 68  |  P A R T I I : E art h O bser v ation f or W ater  R esources M anagement field measurements and RS estimates of much like the resolution of a photograph. ­ precipitation in gauge-sparse landscapes and Related terms are (satellite) footprint and pixel constraining regional water balance through size,6 both expressed in units of distance at the multisensor calibration of a landscape hydrol- Earth’s surface (although the two are not nec- ogy model. A common thread is the increasing essarily equal). Temporal resolution refers to use of Earth observation in conjunction with the frequency with which repeat measure- models and field observation networks, where ments are available. A related term is revisit and when available, to fill the knowledge gap. time, which refers to the time period between In the absence of any field observations, subsequent satellite overpasses. This publica- certain analytical frameworks that exclusively tion considers the general categories of spatial use RS data may still provide fit-for-purpose and temporal resolution shown in box 5.2. information. This is especially beneficial for The key types of variables and their mini- countries with limited or no field observation mum spatial and temporal resolution require- networks. For example, drought monitoring ments can be evaluated for each WRM issue. and water quality systems can use a range of Table 5.1 identifies the main water issues that biophysical, “remotely sensed only” variables can be addressed with the aid of Earth obser- to provide useful synoptic information for vation and links them to the relevant Water decision makers and policy makers. Where Partnership Program topics (and subtopics, field observation networks have validated such where applicable). For each topic and subtopic, information, confidence in the use of Earth the pertinent water issues were derived from observation has increased. examples given on the program’s website and information provided in part I of this publica- tion.7 The results of this analysis are discussed RELEVANT VARIABLES PROVIDED below and summarized in tables 5.2 and 5.3 BY EO (a  rearrangement of table 5.2 that focuses on spatial and temporal resolution). Among all types of information potentially use- Some caveats are in order. First, the analysis ful for addressing WRM issues, many can be undertaken sometimes makes general obtained with the aid of EO techniques. Only in very few cases does the satellite imagery BOX 5.2 (almost) measure the actual variable of inter- est, such as surface albedo or surface ­turbidity. General Categories of Resolution and Examples of More typically, the observations are  used to Platforms Providing This Type of Data infer or estimate the variable—using some modeling technique, often referred to as the Spatial resolution: retrieval algorithm or observation model. •  S1: very fine, pixel size less than 10 meters (QuickBird, IKONOS) Table 4A.1 in annex 4A (available online) •  S2: fine, pixel size: 10–100 meters (Landsat, ASTER) compares World Bank water-related activities •  S3: medium, pixel size: 100–1,000 meters (MODIS, AVHRR) with the relevant variables that can be mea- •  S4: coarse, pixel size more than 1,000 meters (ASCAT, AMSR-E, GRACE) sured in situ or estimated with the aid of Earth Temporal resolution (revisit times): observation. In relation to Earth observation, it •  T1: near continuous, less than 3 hours (geostationary satellites) is important to consider the spatial and tempo- •  T2: high frequency, 3–24 hours (polar-orbiting broad-swath satellites ral resolution that the satellite imagery must such as MODIS, AVHRR) have for it to be useful for informing the issue •  T3: medium frequency, 1–30 days (Landsat) •  T4: occasional, once only or ad hoc (SRTM, tasked radar) at hand. Spatial resolution relates to the spatial detail that can be distinguished in the data, C h apter 5 : E art h O bser v ations an d W ater  I ssues   |  69 Table 5.1 Relationship between Water Issues and Water Topics and Subtopics in the World Bank Water Partnership Program 70  | WATER RESOURCES MANAGEMENT AND WATER SUPPLY ENVIRONMENTAL SERVICES AGRICULTURAL INVASIVE WATER RESOURCES SANITATION WATER AQUATIC ENVIRONMENTAL AQUATIC AND CLIMATE ISSUE RURAL URBAN AND HYGIENE MANAGEMENT ECOSYSTEMS FLOWS PLANTS CHANGE HYDROPOWER Identifying and monitoring water • • • • • reservoirs Monitoring and predicting water quality in • • • • • dams and reservoirs Mapping extent of flood • • • • • • • Predicting extent of flood • • • • • • • Monitoring extent of snow and glacial • • • • • • cover Mapping urban and rural infrastructure • • • • • Assessing water use efficiency in irrigated • • crops Monitoring rates of irrigation water use • • Monitoring rates of groundwater • • • extraction Mapping irrigated areas • • Monitoring crop production and food • • security Monitoring and forecasting drought • • • • • • Monitoring water quality of coastal • discharge Monitoring maritime pollution (for • example, oil spills) Identifying and monitoring groundwater- • • dependent ecosystems Monitoring river elevation • • • • • • Monitoring and controlling aquatic weeds • • • Conducting integrated assessment of water availability under climate change • scenarios Designing hydropower production • facilities Note: Bullet indicates where knowledge of a particular water issue is relevant to a specific water (sub)topic. Table 5.2 Overview of Water Issues and Relevant Variables Provided by Earth Observation VEGETATION SOIL AND LAND SURFACE SNOW WATER PRECIPITATION EVAPOTRANSPIRATION MOISTURE COVER GROUNDWATER WATER AND ICE QUALITY MODELING ISSUES S T S T S T S T S T S T S T S T OTHERS APPROACHES Identifying and monitoring                     S1, T3,           water reservoirs S2 T4 Monitoring and predicting                     S1, T2,     S1, T2,   Biogeochemical models water quality in dams and S2, T3 S2, T3 reservoirs S3 S3 Mapping extent of flood                     S2, T1,         Elevation   S3 T2, (DEM) T3 Predicting extent of flood S4 T1,       S4 T1,         S2, T2,         Elevation Hydrodynamic models T2 T2 S3 T3 (DEM) Monitoring extent of snow S4 T1,                       S2, T2,     Elevation   and glacial cover T2 S3, T3 (DEM) S4 Mapping urban and rural             S1, T4                     infrastructure S2 Assessing water use S4 T1,   S2, S3 T2, T3 S4 T1, S2, T3 S4 T2             River models efficiency T2 T2 S3 Monitoring rates of     S2, S3 T2, T3     S2, T3                   River models irrigation water use S3 Monitoring rates of     S2, S3 T2, T3         S4 T2               River models groundwater extraction Mapping irrigated areas S4 T1,   S2, S3 T2, T3 S4 T1, S2, T2,     S2, T2,         Elevation   T2 T2 S3 T3 S3 T3 (DEM) Monitoring crop S4 T1,   S2, S3 T2, T3 S2, T2,                   Crop or pasture growth production and food T2 S3 T3 models security Monitoring and forecasting S4 T1,   S2, S3 T2, T3 S4 T1, S2, T3 S4 T2 S2, T2, S2, T2,       Landscape water drought T2 T2 S3 S3 T3 S3 T3 balance models Monitoring water quality                             S1, T1,     of coastal discharge S2, T2, S3 T3 (Continued)   |  71 72  | Table 5.2 (Continued) VEGETATION SOIL AND LAND SURFACE SNOW WATER PRECIPITATION EVAPOTRANSPIRATION MOISTURE COVER GROUNDWATER WATER AND ICE QUALITY MODELING ISSUES S T S T S T S T S T S T S T S T OTHERS APPROACHES Monitoring maritime                             S1, T1,     pollution (for example, oil S2, T2, spills) S3 T3 Identifying and monitoring S4 T1, T2 S2, S3 T2, T3     S2, T3     S2, T2,             groundwater-dependent S3 S3 T3 ecosystems Monitoring river flow S4 T1 T2 S2, S3 T2, T3 S4 T1,     S4 T2     S2, T2,     Elevation Landscape water T2 S3 T3 (DEM) balance and river models Monitoring and controlling             S1, T2,             S1, T2,     aquatic weeds S2 T3 S2 T3 Conducting integrated S4 T1 T2 S2, S3 T2, T3 S4 T1, S2, T3 S4 T2 S2, T2, S2, T2,       Landscape water assessment of water T2 S3 S3 T3 S3 T3 balance and river availability under climate models change scenarios Designing hydropower             S1, T4                 Elevation   production facilities S2 (DEM) Note: S = spatial; T = temporal. S1, S2, S3, and S4 refer to the spatial resolution of the data, while T1, T2, T3, and T4 refer to the temporal resolution. They are defined as follows: S1, very fine (pixel size, less than 10 meters), S2, fine (pixel size, 10–100 meters), S3, medium (pixel size, 100–1,000 meters), and S4, low (pixel size, more than 1,000 meters), T1, near continuous (revisit time, less than 3 hours), T2, high frequency (revisit time, 3–24 hours), T3, medium frequency (revisit time, 1–30 days), T4, occasional (revisit time, once only or ad hoc). Blue indicates that the data are highly valuable, green indicates that they are valuable, and white indicates that they are not relevant. For more information on these issues, see the section in chapter 6 on the type of data obtained and the section in chapter 7 on determining the minimum required data requirements; both sections are divided into subsections on each variable. DEM = digital elevation model. Table 5.3  Overview of Water Issues and Relevant Variables Provided by Earth Observation   Rearranged to Focus on Spatial and Temporal Resolution SPATIAL TEMPORAL ISSUE S1 S2 S3 S4 T1 T2 T3 T4 MODEL Identifying and monitoring water reservoirs SW SW SW SW Monitoring and predicting water quality in SW SW SW SW SW Biogeochemical dams and reservoirs WQ models WQ WQ WQ WQ Mapping extent of flood SW SW SW SW SW DEM DEM DEM Predicting extent of flood P P P Hydrodynamic models SM SM SM SW SW SW SW DEM DEM DEM Monitoring extent of snow and glacial cover P P P S&I S&I S&I S&I S&I DEM Mapping urban and rural infrastructure V&LC V&LC V&LC Assessing water use efficiency P P P River models ET ET ET ET SM SM SM V&LC V&LC V&LC GW GW Monitoring rates of irrigation water use ET ET ET ET River models V&LC V&LC V&LC Monitoring rates of groundwater extraction ET ET ET ET River models GW GW Mapping irrigated areas P P P SM SM SM ET ET ET ET V&LC V&LC V&LC V&LC SW SW SW SW DEM DEM DEM Monitoring crop production and food P P P Crop or pasture security growth models ET ET ET ET V&LC V&LC V&LC V&LC Monitoring and forecasting drought P P P Landscape water balance models ET ET ET ET SM SM SM V&LC V&LC V&LC GW GW SW SW SW SW S&I S&I S&I S&I (Continued) C h apter 5 : E art h O bser v ations an d W ater  I ssues   |  73 Table 5.3 (Continued) SPATIAL TEMPORAL ISSUE S1 S2 S3 S4 T1 T2 T3 T4 MODEL Monitoring water quality of coastal WQ WQ WQ WQ WQ WQ discharge Monitoring maritime pollution (oil spills) WQ WQ WQ WQ WQ WQ Identifying and monitoring groundwater- P P P dependent ecosystems ET ET ET ET V&LC V&LC V&LC SW SW SW SW Monitoring river flow P P P Landscape water balance and river ET ET ET ET models SM SM SM GW GW S&I S&I S&I S&I DEM DEM DEM Monitoring and controlling aquatic weeds V&LC V&LC V&LC V&LC WQ WQ WQ WQ Conducting integrated assessment of water P P P Landscape water availability under climate change scenarios balance and river ET ET ET ET models SM SM SM V&LC V&LC V&LC GW GW SW SW SW SW S&I S&I S&I S&I Designing hydropower production facilities V&LC V&LC V&LC DEM DEM DEM Note: S1, S2, S3, and S4 refer to the spatial resolution of the data, while T1, T2, T3, and T4 refer to the temporal resolution. They are defined as follows: S1, very fine (pixel size, less than 10 meters), S2, fine (pixel size, 10–100 meters), S3, medium (pixel size, 100–1,000 meters), and S4, low (pixel size, more than 1,000 meters), T1, near continuous (revisit time, less than 3 hours), T2, high frequency (revisit time, 3–24 hours), T3, medium frequency (revisit time, 1–30 days), T4, occasional (revisit time, once only or ad hoc). Blue indicates that the data are highly valuable, green indicates that they are valuable, and white indicates that they are not relevant. DEM = digital elevation model; ET = evapotranspiration; GW = groundwater; P = precipitation; S&I = snow and ice; SM = soil moisture; SW = surface water; V&LC = vegetation and land cover; WQ = water quality. assumptions about natural water systems, on Australia’s large-scale, flood-harvesting water ­ supply and use, and the infrastructure ­private cotton farms, where water storage con- built to support the latter. Given the impor- tainers can measure kilometers across. tance of issues related to water for agriculture, Second, it is difficult to assess data require- this is particularly relevant where the charac- ments without considering the current state of teristics of the information that can be derived EO technology and methods of analysis; that is, in any specific application need to be compared even where EO applications have only been very carefully with the characteristics of the conceived in a theoretical sense, such ideas are farming systems involved. For example, farm usually constrained by the assumed limits to the dams are typically comparatively small struc- technology.8 This introduces a degree of circu- tures (often less than 100 meters across). How- larity in the analysis, particularly when consid- ever, notable exceptions do exist—for instance, ering the lowest spatial and temporal resolution 74  |  P A R T I I : E art h O bser v ation f or W ater  R esources M anagement that might still be useful. For example, if satel- water issue. Chapter 6 explains each variable, lite observation of soil moisture or water level detailing its relevance, the theoretical basis for were possible at a scale of meters and minutes, its estimation with Earth observation, and the it is likely that entirely new applications would current and future technologies available for be conceived and developed and that data its measurement. ­ requirements would be modified accordingly. The following list explains the information Tables 5.2 and 5.3 cross-reference the issues conveyed schematically in tables 5.2 and 5.3: addressed under the topic areas with relevant variables that can be measured or estimated • Identifying and monitoring water reservoirs. with the aid of Earth observation. The two Applications could include identifying tables present the same information, but water reservoirs for monitoring compli- arranged in different ways to facilitate inter- ance or observing water storage as part pretation. For each issue, the relevant variables of a drought warning system. They may that can be obtained from Earth observation also be used to observe water resources are listed. Each variable is classified according and climate change as well as hydropower. to its usefulness: green when considered “valu- Depending on the size of the reservoirs, able,” meaning that it is likely to be useful in high spatial resolution may be required (S1, addressing the issue at hand, blue when con- S2). Generally, slow water dynamics mean sidered “highly valuable,” meaning that using that a moderate f ­requency is likely to be EO may significantly improve the ability to required (T3, T4). address the issue, and white when deemed not • Monitoring and predicting water quality relevant. In addition, the most appropriate in dams and reservoirs. In addition to the spatial and temporal resolutions are listed. need to locate these water bodies (through As an example, consider the efficiency of remote sensing or other sources), Earth water use in crops, an important issue for agri- observation can be used at similar spatial cultural water management. Earth observation resolutions and temporal revisit times (S1, can provide information on evapotranspira- S2, S3, T2, T3) to quantify water quality. tion9 to estimate water use by crops and can This is particularly relevant for assessing also identify the location of the irrigated the health risks to human and animal popu- crops—both types of data may well be essential. lations who depend on these water bodies. These data can be obtained at fine and medium It may also be related to water resources, spatial resolution and with high and medium climate change, and hydropower. frequency (temporal resolution); the prefera- ble combination will depend on the nature of • Mapping flood extent. Floods are a haz- the application. Furthermore, Earth observa- ard to both rural and urban populations tion can be used to estimate rainfall, which is because they can affect the provision of particularly useful where the field rainfall mea- potable water. Flood extent can be moni- surement network is inadequate. Precipitation tored with Earth observation; normally, data from Earth observation are only available high to medium spatial resolution is at coarse spatial resolution, but with high or required, depending on the extent of the even near-continuous frequency. Finally, Earth flood and the physical characteristics of observation can provide potentially relevant the terrain. In large floodplains, medium information on the amount of moisture in the resolution (S3) may suffice, but high reso- top layer of soil and on the volume of ground- lution (S2) may be required in many other water, but again, only at coarse resolution. cases. Normally, high-frequency imagery Table 5.3 may be used as a guide for deter- (T2) is desired, but opportunistic acquisi- mining the data needs and availability for each tion of medium-frequency imagery (T3) C h apter 5 : E art h O bser v ations an d W ater  I ssues   |  75 can also be useful. Digital elevation models used by the crop. Efficient agricultural man- (DEMs) can help to identify flooded areas. agement ensures the sustainable use of water. The key variable to monitor is evapo- • Predicting flood extent. Besides mapping transpiration, which can be done at high flood extent when flooding occurs, pre- and medium spatial resolutions (S2, S3) and dicting flood extent is highly relevant to also at high- and medium-frequency revisit urban water supply and, of course, to disas- times (T2, T3), depending on the applica- ter management. DEMs are critical in this tion. Other useful variables are land cover, at context; in addition to weather forecasts, the same spatial resolution and at least once antecedent rainfall and soil moisture con- during the growing cycle (T3), and precipi- ditions can be very useful. Flood extent tation and soil moisture, typically at coarse can be predicted by considering previously spatial resolution (S4), but perhaps daily or flooded areas and estimating the associ- more frequently (T1, T2). ated recurrence times. It may be combined with hydrodynamic models to simulate • Monitoring rates of irrigation water use. water flows and flood extent during high- Similar to assessing water use efficiency, rainfall events upstream. monitoring irrigation water use requires • Monitoring extent of snow and glacial cover. estimating crop ET rates. Information on Many regions of the world obtain part the location and size (land cover mapping) of their water supply from melting snow of irrigated crops is useful. It is also impor- and ice. This is the case in high-latitude tant in relation to water resources and regions, in mountainous regions, and in ­climate change. valleys at the foothills of high mountains. • Monitoring rates of groundwater extraction. Measuring the area and water equivalent Water volumes extracted from groundwa- of snow and ice can help to estimate the ter cannot be estimated directly with Earth ­ volume of water runoff to be expected dur- observation. Gravimetric measurements ing spring and summer. Snow and ice can can provide coarse resolution (S4) esti- be measured with Earth observation at mates of groundwater, which can inform high, medium, and coarse spatial resolu- basin-wide changes in groundwater ­ levels. tion (S2,  S3, S4); high as well as medium In local studies, a combination of river frequency (T2, T3) are required and models and satellite ET estimates may possible. In addition to the direct mapping ­ help to constrain groundwater extraction of snow and ice areas, Earth observation of estimates, which are also of importance for precipitation can improve water equiva- urban water supply. lent estimates, and DEMs can also indicate where snow is likely to fall and persist. • Mapping irrigated areas. The location of crops can be determined by using land • Mapping urban and rural infrastructure. cover classification techniques at high or Applications include the identification of medium spatial resolution (S2, S3) and existing facilities and land cover mapping revisit times (T2, T3). Whether specific before construction. Generally, very high crops have been irrigated cannot be estab- or high spatial resolution imagery (S1, S2) lished directly with Earth observation, is needed, either from satellites or from unless water remains in the surface for long airborne imagery on occasion (T4). periods, as is the case of paddy rice. How- • Assessing water use efficiency in irrigated ever, it may be determined with ancillary crops. Water use efficiency is the ratio of information, such as the connectedness to agricultural produce to the amount of water surface water reservoirs or rivers, or with 76  |  P A R T I I : E art h O bser v ation f or W ater  R esources M anagement information regarding the estimated water monitored with Earth observation using balance deficit (that is, the difference methods similar to those described under between precipitation and evapotranspira- the previous item. tion) during the growing season, which can • Identifying and monitoring groundwater- be obtained with Earth observation. dependent ecosystems. Groundwater- • Monitoring crop production and food dependent ecosystems require access to ­security. Earth observation can be used to groundwater to meet some or all of their estimate crop production via vegetation water requirements. Their survival can be indexes, normally at high and medium spa- threatened by consumptive use of water for tial (S2, S3) and temporal (T2, T3) resolu- agriculture, mining, and other purposes. tions. Remotely sensed precipitation and These systems typically need to be identi- evapotranspiration can be useful too. Crop fied at high or medium spatial resolution growth models or pasture growth models (S2, S3) and can be supported by EO-based (in the case of livestock production) can land cover mapping and estimates of water also be useful and may be parameterized balance deficit. Mapping of open water can with EO data. help to distinguish ecosystems dependent on groundwater from those dependent on • Monitoring and forecasting drought. surface water inflows. Drought monitoring typically uses vegeta- tion index10 anomalies and precipitation • Monitoring river streamflow. River stream- data. However, remotely sensed total water flow can be monitored directly with Earth storage, surface soil moisture, and rainfall observation using radar altimetry, but are increasingly being incorporated into currently only for comparatively broad ­ drought monitoring systems. Forecasting ­ rivers and at a limited number of locations. drought requires landscape water balance Flows can also be modeled with landscape models that can be forced (up to the fore- and river models, which can be informed cast date) or calibrated with additional EO (forced or calibrated) with EO estimates information on rainfall, evapotranspira- of precipitation, soil moisture, evapo- tion, soil moisture, groundwater, and snow transpiration, snow and ice extent (where and ice, where relevant. relevant), and groundwater. ­ • Monitoring water quality of coastal dis- • Monitoring and controlling aquatic weeds. charge. Water quality in rivers can affect Aquatic weeds can be a challenging prob- marine ecosystems by discharging exces- lem affecting navigation, water supply, sive levels of sediments and nutrients. and habitats. They may be detected and These discharges can be monitored with mapped with land cover classification Earth observation either in rivers or estu- techniques and tend to be related to water aries themselves or in coastal waters. quality. This kind of monitoring is nor- The effects of the discharge, such as algal mally done, and necessarily so, at very high blooms, can also be detected. While inland or high spatial resolution (S1, S2). water bodies may require high or very high • Conducting integrated assessment of water spatial resolution (S1, S2), coastal environ- availability under climate change scenarios. ments typically require medium spatial Proper characterization and understand- (S3) and high temporal (T2) resolution. ing of the water balance and its drivers • Monitoring maritime pollution ( for exam- in the past and present are necessary to ple, oil spills). Coastal pollution can be predict the availability of water over large C h apter 5 : E art h O bser v ations an d W ater  I ssues   |  77 regions under climate change scenarios. 4. For the GPM, see http://www.nasa.gov/mission_ Almost all of the variables mentioned here pages/GPM/launch/. For the SMAP, see https:// smap.jpl.nasa.gov/. can be useful. Landscape and river mod- 5. For information on these types of data sets, see els are important for integrating the vari- http://www.isac.cnr.it/~ipwg/data/datasets1.html. ous observations and creating scenarios of 6. Spatial resolution is defined as the size of the future conditions. smallest individual component or dot (called a pixel) from which the image is constituted. • Designing hydropower production facilities. For instance, if a satellite’s resolution is stated Designing hydropower facilities requires as “5 meters,” each pixel in the imagery has a size of 5 meters by 5 meters. The footprint is the the availability of accurate DEMs, which area of the Earth covered by the microwave radia- can be obtained from airborne or satellite tion from a satellite dish (transponder). The size imagery. It may also benefit from mapping of the footprint depends on the location of the of land cover, including existing buildings satellite in its orbit, the shape and size of the beam produced by its transponder, and the distance and vegetation types, at very high or high from the Earth. resolution (S1, S2). 7. Therefore, this list is not exhaustive. For example, monitoring water reservoirs and water quality in dams and reservoirs may also relate to water resources and climate change and to hydropower, NOTES monitoring rates of irrigation water use may also relate to water resources and climate change, 1. As in part I, for simplicity’s sake, other types of monitoring groundwater extraction may also data relevant to hydrologic applications—such as relate to urban water supply, and so on. land cover, land subsidence, and topography—are 8. Henry Ford supposedly said about his cars, “If I also referred to as hydrometeorological variables. had asked people what they wanted, they would 2. The Water Partnership Program sometimes have said faster horses.” changes its classification slightly (see http://water 9. Evapotranspiration is the process by which water .worldbank.org/wpp). As of February 15, 2015, is transferred from the land to the atmosphere by the subtopics are water supply, sanitation, evaporation from the soil and other surfaces and irrigation and drainage, hydropower, and water ­ by transpiration from plants. resources management. The program’s “thematic 10. A vegetation index describes the greenness—the ­ highlights” are water resources management, relative density and health of vegetation—for each climate change, food security, energy s ­ ­ ecurity, pixel in a satellite image. water for environment, water supply and sanitation, integrated urban water management, ­ remote sensing, and disaster risk management. For practical purposes, part II uses the categories listed as topics, combined with the program’s REFERENCE action areas. 3. Although some define Earth observation as con- Ali, A., T. Lebel, and A. Amani. 2005. “Rainfall Estima- sisting of remote sensing and in situ measure- tion in the Sahel. Part I: Error Function.” Journal ments, this publication uses the terms Earth of Applied Meteorology 44 (11): 1691–706. observation and remote sensing interchangeably. 78  |  P A R T I I : E art h O bser v ation f or W ater  R esources M anagement CHAPTER 6 Earth Observations for Monitoring Water Resources INTRODUCTION and temporal domains. EMS sensors are defined by the extent, resolution, and density in each of This chapter provides an overview of the main those domains (table 6.1) plus the spatial domain variables that can be derived from satellite (Emelyanova et al. 2012). Earth observation (EO) and are relevant to the This data framework provides a means to water issues presented in chapter 5. Most EO assess the likely utility of different types of RS instruments obtain an image of radiation inten- data to estimate key biophysical variables in sity in specific portions of the electromagnetic particular applications. Because of the limits spectrum (EMS). The radiation is reflected on measurement and telecommunication tech- from the sun by the Earth’s surface, called opti- nology, there is typically a trade-off between cal remote sensing (RS); emitted by the Earth’s the performance that a sensor can achieve in surface itself, called passive remote sensing; or each of these dimensions. For example, imag- first emitted by the instrument and then ery obtained by the Moderate Resolution reflected from the surface, called active remote Imaging Spectroradiometer (MODIS) sensors sensing, such as radar. Exceptions to EMS has a temporal, spectral, and radiometric reso- imagers are satellite altimeters and the Gravity lution that is about an order of magnitude Recovery and Climate Experiment (GRACE) higher than that obtained by Landsat, but its gravimetry mission, whose primary measure- spatial resolution is an order of magnitude ments are distance to the Earth’s surface and lower (figure 6.1, panel a). between the two satellites, respectively. This framework only considers observa- tional characteristics, although important oper- ational considerations often also exist, such as CHARACTERISTICS OF SENSORS the following: Fundamental to the design of any EMS sensor • Data availability and the cost of purchase, are its characteristics in the spectral, radiometric, if any 79 Table 6.1 Data Framework Comprising Domain-Characteristic Elements DOMAIN EXTENT RESOLUTION DENSITY Spectral Portion(s) of the EMS being sampled Bandwidth(s)a Number of bands in a particular portion of the EMSb Radiometric Dynamic range of radiances Change in radiance due to change by Number of bits used across the (minimum and maximum radiance per one digital number dynamic range of radiances band) Temporal Recording period over which the Period of data acquisitiond Satellite repeat characteristicse data are availablec Source: Modified from McVicar and Jupp 2002. Note: EMS = electromagnetic spectrum. a. The narrower the bandwidth, the higher the spectral resolution. For example, hyperspectral sensors (Hyperion) have higher spectral density than broadband instruments (Landsat TM/ETM+), although they sample similar EMS ranges. b.  c. For some remotely sensed systems (AVHRR and Landsat TM), data have been recorded near-continuously for about 30 years. d. For remotely sensed images, this is a matter of seconds, which contrasts with meteorological data such as the daily rainfall totals.  or some applications using optical (that is, reflective and thermal) data, the availability of cloud-free images is an important consideration. Whereas the satellite’s e. F repeat characteristics do not change, cloud cover will change the effective temporal density of a site over time. Figure 6.1 Characteristics of MODIS and Landsat TM Data Domain a. Temporal density and spatial resolution MODIS data Landsat data Temporal (16 days) Temporal (daily) ial ) ial ) at 0 m at m Sp ×50 Sp ×30 0 Spatial (500×500 m) (50 Spatial (30 m×30 m) (30 b. Spectral extent, resolution, and density 1.0 Landsat TM 0.8 Modis Reflectance 0.6 Soil 0.4 Vegetation 0.2 Water 0 0.5 1.0 1.5 2.0 2.5 Wavelength (µm) c. Radiometric extent, resolution, and density, TM and MODIS infrared bands 600 Radiance (W m–2 sr–1 µm–1) MODIS: 0.17 W m–2 sr–1 µm–1 DN–1) 500 400 300 TM: 1.04 W m–2 sr–1 µm–1 DN–1) 200 100 0 256 512 1024 2048 4096 Digital number (DN) Source: Emelyanova et al. n.d. © Commonwealth Scientific and Industrial Research Organisation (CSIRO). Used with permission. Further permission required for reuse. Note: Panel a shows temporal density and spatial resolution. Panel b shows spectral extent, resolution, and density (darker colors represent the MODIS bands, while the lighter colors represent the Landsat TM bands). Panel c shows radiometric extent, resolution, and density for the TM and MODIS infrared bands. MODIS = Moderate Resolution Imaging Spectroradiometer; TM = thematic mapper. 80  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T • Data latency (the time that passes between surface in liquid form (rain), solid form (snow the actual observation and the moment the or hail), or a combined form (sleet). The ability data are made available) to quantify precipitation distributions in space and time is critical to establishing infrastruc- • Reliability (any guarantees with regard to ture to capture and store water resources for an future availability and latency, stability of ever-growing population. Due to its fine-scale the data characteristics, and the like) spatial and temporal variability, monitoring • Data format (size of the data files, requirements large-area precipitation challenges field-based for specialized skills, software, or hardware) measurement networks. A rain gauge can pro- • Degree of validation and acceptance vide an accurate estimate of precipitation at a (whether stakeholders will accept the data point in the landscape, but there is uncertainty being used or the quality of the data com- about whether this estimate is representative pared with data from alternative sources) of rainfall at some distance away from the gauge location. This problem is especially pro- • Interpretability and uncertainty (how nounced for particular rainfall regimes, such as unambiguous is the interpretation of the convective storms (figure 6.2). Space-based data in the context of a specific application) methods of estimating precipitation offer ways This list is not meant to be exhaustive. to fill the information “gap” either by merging with existing surface measurement networks— to constrain estimation between gauges—or by TYPES OF DATA OBTAINED FROM providing direct estimates when and where no EARTH OBSERVATION other information is available. This publication has adopted the definitions Estimating Space-Based Precipitation given in table 6.2. The remainder of this chapter Satellite-based estimation of precipitation explains how the main data products relevant to began in the 1970s with the advent of weather water resources monitoring are obtained from satellites. Multichannel radiometers aboard raw or processed data. Appendix B provides a geostationary satellite platforms provide visi- list of notable examples of information products. ble infrared (VIS) and thermal infrared (TIR) imagery of the Earth’s surface at medium (S3), Precipitation about 1-kilometer, spatial resolution and very Definition high (T1) temporal resolution. These satellite Precipitation is the process by which water data were used to generate the first set of pre- returns from the atmosphere to the Earth’s cipitation estimates for large areas of the globe, Table 6.2 Types of Data Obtained from Earth Observation DATA TYPE DESCRIPTION Raw data Sensor measurements as received directly from the satellite, formatted as “digital counts” Processed data Top of atmosphere (TOA) signal. Raw data processed to TOA data: conversion to real-world units, such as radiance (watt per steradian per square meter per nanometer [W·sr−1·m−2·nm−1]) or reflectance (%); signal calibration Surface signal. TOA data processed to surface-equivalent data: corrections applied to remove atmospheric and solar-sensor viewing-angle effects; scene stitching; geolocation and reprojection Data products Conversion of processed data into products that describe real-world (usually biophysical) variables such as chlorophyll concentration, leaf area, rainfall rate, surface temperature, and soil moisture mass Information products Conversion of data products into management-relevant information for decision support, for example, eutrophication state of Cobalt Lake, flood risk of the Emerald River delta, and sustainable irrigation rates in the Crimson basin Note: TOA = top of atmosphere. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  81 Figure 6.2 Space-Based Precipitation Measurements from TRMM Satellite a. Rainfall off the coast of Madagascar b. Rainfall over southeastern United States Source: NASA (http://pmm.nasa.gov/mission-updates/trmm-news/trmm-sees-severe-weather). Note: (a) Large convective rainfall storm off the northwest coast of Madagascar as detected by Tropical Rainfall Measuring Mission (TRMM) Satellite’s precipitation radar on April 3, 2014, at 01:43 UTC (Coordinated Universal Time); (b) a frontal rainfall system developing into a line of intense storms over southeast United States at 13:00 UTC on April 7, 2013. based on a relationship between cloud top these data are generally considered superior temperatures and precipitation rate (roughly, to those obtained from TIR observations. the lower the temperatures, the higher the rate Both microwave and thermal approaches to of precipitation). However, Arkin and Meisner estimating satellite-based precipitation have (1987) show that TIR-based estimates of pre- strengths and weaknesses. For example, TIR- cipitation are relatively poor, as the relation- based estimates from geostationary satellites ship between the cloud top temperature and have full-disk (global) coverage at near con- precipitation rate break down for resolutions tinuous (T1) temporal resolution of one-six to in time shorter than one day and resolutions in 1 hour and a coarse to medium spatial resolu- space lower than 2.5° in latitude and longitude. tion of less than 5 kilometers (S3 or S4), but The next advance in satellite-based estima- result in poor precipitation estimates at the tion of precipitation occurred in the 1980s high resolution. Conversely, passive micro- with the deployment of passive microwave wave estimates from polar-orbiting satellites sensors aboard polar-orbiting satellites. In are more accurate but cover less of the globe, contrast to the weak relationship underpin- have coarser spatial resolution of about 10–100 ning TIR-based precipitation, the scattering kilometers (S4), and have less frequent repeat and emission of passive microwave radiation coverage (T4) for any given sensor. by ice particles or rain droplets in clouds is The mid-1990s ushered in a new era of deriv- better understood and modeled (Kummerow ing multisatellite precipitation estimates, as algo- et al. 2001). Satellite-based, passive microwave rithms were developed that exploited the high brightness temperatures between 10–200 spatial and temporal coverage of the geostation- gigahertz have stronger relationships with ary TIR estimates with the more accurate passive precipitation, and the retrievals derived from microwave-based estimates, making the best of 82  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T both approaches (Huffman et al. 1997; Joyce et al. real time and in post–real time (known as 2004; Kubota et al. 2007; Sorooshian et al. 2000). research grade products). In 2012, the system Table 6.3 summarizes the characteristics of underwent a major transition from v6 (version some of the more commonly used satellite pre- 6) to v7 in which all products from the start of cipitation products (SPPs). A common feature production in December 1997 were reprocessed. of these products is that they all use both Studies have demonstrated the superiority of microwave and thermal EO data to generate TMPA v7 to its predecessor (Chen et al. 2013). precipitation estimates. On February 27, 2014, an H-IIA rocket The launch of the Tropical Rainfall Measur- from the Japan Aerospace Exploration ing Mission (TRMM) satellite in 1997 placed Agency (JAXA) launched into orbit the first the world’s first precipitation radar in orbit. satellite of the core observatory of the Global Precipitation radar provides detailed informa- Precipitation Measurement (GPM) mission, tion on the vertical structure (250-meter reso- building on and continuing the long history of lution) of rainfall and offers the most accurate space-based estimation of precipitation. GPM precipitation estimates from space (Kum- will provide a multisatellite view of global merow et  al. 2001). At a satellite orbit of precipitation at unparalleled spatial and tem- 350 kilometers in altitude, swath width of 215 poral coverage. Many of the techniques for kilometers, and orbital inclination limiting its estimating precipitation from space and for coverage to ±35° latitude, the data provided by blending results from multiple satellite sen- precipitation radar are far from global. How- sors have been honed over decades—from the ever, the quality of the precipitation estimates early, cloud top temperature methods of geo- makes precipitation radar a valuable source of stationary thermal Earth observation to the information for calibrating both passive micro- recent constellation of polar-orbiting micro- wave and TIR instrumentation across multiple wave imagers of the TMPA system. The GPM satellite platforms, thus extending the potential mission continues this legacy of space-based coverage of precipitation estimation. This is the monitoring of precipitation (Hou et al. 2013). basis for the TRMM Multisatellite Precipita- The core observatory of the GPM mission tion Analysis (TMPA) system, which generates has a design life of three years, with battery life quasi-global precipitation estimates going back of at least five years and extended mission life as far as January 1, 1998 (Huffman et al. 2007). until 2021 (Hou et al. 2013). Global precipitation The TMPA system sets the standard for the products will be generated at three-hourly operational production of global satellite- intervals, with a latency of three to four hours derived precipitation estimates. While peer sys- every day, by combining data from a “constella- tems (that is, rainfall analysis systems based tion” of current and planned microwave sensors primarily on satellite observations) may have through the Integrated Multisatellite Retrievals higher resolution in space and time than some of for GPM (IMERG) system (Huffman et al. 2013). the TMPA products (table 6.3), agencies throughout the world have used TMPA’s (quasi-) Satellite-Derived Precipitation Products operational status to feed into their rainfall anal- Table 6.3 lists the characteristics of some of the ysis systems (Mitra et  al. 2009; Rozante et  al. better-known global precipitation products that 2010) and to inform current and planned global are derived from multiple satellite sensors. The flood and drought monitoring systems (Pozzi International Precipitation Working Group et al. 2013; Wu et al. 2012). The suite of precipita- provides a more comprehensive list of SPPs, tion products from the TMPA system (including including single-source products, model reanal- precipitation radar only, microwave only, and yses (weather model–derived precipitation merged microwave-TIR products) is available in products), and gauge-only gridded estimates.1 chapter 6 : E arth O bservations for M onitoring W ater R esources   |  83 Table 6.3 Overview of Main Characteristics of Some Widely Used Global Satellite-Derived Precipitation Estimates MAIN REFERENCE SPATIAL TEMPORAL (NUMBER OF NAME PRIMARY SATELLITE(S) OR SENSOR(S) RESOLUTIONC FREQUENCY DATA LATENCY CITATIONS) ACCESS TMPAa – 3B40RT Combined TMI, SSM/I, AMSU, and AMSR-E data 0.25° 3 hours 6–7 hours Huffman et al. 2007 (967) Anonymous ftp site: ftp:// trmmopen.gsfc.nasa.gov/pub/ merged/3B40RT TMPA – 3B41RT Infrared brightness temperatures (geostationary) spatially 0.25° 3 hours 6–7 hours Huffman et al. 2007 Anonymous ftp site: ftp:// aggregated and calibrated to microwave rain rates trmmopen.gsfc.nasa.gov/pub/ merged/3B41RT TMPA – 3B42RT Thermal infrared (TIR)-rainfall calibrated with precipitation 0.25° 3 hours 6–7 hours Huffman et al. 2007 Anonymous ftp site: ftp:// radar and merged with TMI and whatever other passive trmmopen.gsfc.nasa.gov/pub/ microwave data are available merged/3B42RT TMPA – 3B42 TIR-rainfall calibrated with precipitation radar and merged 0.25° 3 hours 1–2 months Huffman et al. 2007 Links from http://disc.sci.gsfc. with TMI, SSM/I, AMSU, and AMSR-E data; monthly nasa.gov/ aggregates adjusted with rain gauge measurements TMPA – 3B42 daily Same as TMPA-3B42, but aggregated to daily rainfall totals 0.25° Daily 3–4 months Huffman et al. 2007 Links from http://disc.sci.gsfc. nasa.gov/ CMORPH – 30 minutes Passive microwave, including TMI, SSM/I, AMSU, AMSR-E 0.07° 30 minutes 18–24 hours Joyce et al. 2004 (473) ftp://ftp.cpc.ncep.noaa.gov (TIR used for motion vectors) CMORPH – 3-hourly 0.25° 3 hours 2–3 days Joyce et al. 2004 CMORPH – daily 0.25° Daily 2–3 days Joyce et al. 2004 QMORPH – 30 minutes 0.07° 30 minutes 3–4 hours Joyce et al. 2004 QMORPH – daily 0.25° 30 minutes 3–4 hours Joyce et al. 2004 PERSIANN TIR, passive microwave, including TMI, SSM/I, and AMSU 0.25° 1 hour About 2 days Sorooshian et al. 2000 http://chrs.web.uci.edu/ (252); Hsu et al. 1997 persiann/data.html 84  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T PERSIANN – CCS TIR data, geostationary 0.04° 1 hour About 1 hour Hong et al. 2004 http://hydis.eng.uci.edu/gawdi/ PERSIANN – CDR GridSat-B1 CDR TIR window (near 11 microns) 0.25° Daily About 3 months Ashouri et al. 2014 http://www.ncdc.noaa.gov/ cdr/operationalcdrs.html (requires user registration) PERSIANN – CONNECT TIR, passive microwave 0.25° 1 hour Sellars et al. 2013 http://chrs.web.uci.edu/ (precipitation objectsb) research/voxel/index.html GSMaP Passive microwave, including TMI, SSM/I, and AMSU 0.1° Kubota et al. 2007 (84) http://sharaku.eorc.jaxa.jp/ GSMaP/index.htm Note: Data latency refers to the minimum time period between satellite data acquisition and product available for download. — = not available; AMSR = Advanced Microwave Scanning Radiometer; AMSU = Advanced Microwave Sounding Unit; CCS = cloud classification system; CDR = climate data record; CMORPH = Climate Prediction Center MORPHing technique; GSMaP = Global Satellite Mapping of Precipitation; PERSIANN = Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks; QMORPH = variation of CMORPH; SSM = Special Sensor Microwave; SSM/I = Special Sensor Microwave Imager; TIR = thermal infrared; TMI = TRMM Microwave Imager; TMPA = TRMM Multisatellite Precipitation Analysis; TRMM = Tropical Rainfall Measuring Mission. a. Many TMPA-derived products provide monthly average rain rate and profile information at 0.5°–5° resolution. These products can be found at http://disc.sci.gsfc.nasa.gov/precipitation. Precipitation object is a four-dimensional data construct comprising geographic latitude and longitude, precipitation intensity, and time. This system applies a connectivity algorithm to precipitation objects through time, b.  which allows identification of individual precipitation events from the PERSIANN precipitation product archive. National Aeronautics and Space Administration missions that produce data parameters with a coarse spatial resolution typically report the resolution in geographic degrees or fractions of degrees. The size of a degree (or c.  fraction of a degree) depends on how close the measured area is to the equator and the poles. However, the verification statistics are reported makes them less desirable or credible to some at an aggregate scale (for example, national potential users than the direct measurements average), while performance of these products made by rain gauges. is spatially variable at the local scale. For exam- Precipitation estimates from microwave- ple, regions with orographic rainfall pose a based satellite observations are known to challenge to satellite retrieval of rainfall (due to underestimate light rainfall rates, typical of the light intensity of orographic rainfall), and precipitation resulting from orographic lift2 rainfall in those regions is typically underesti- and cold fronts, for example. This is due to the mated or missed in the satellite products. reduced contrast in brightness temperatures from the land surface and scattering layer for Known Issues low clouds. This can be a further impediment Satellite-derived precipitation estimates have to the adoption of SPPs, particularly in moun- the potential to improve spatially distributed tainous areas of the Earth’s surface. hydrologic model estimation and prediction Geostationary SPPs are based on cloud top (Gebremichael and Hossain 2010; Pan, Li, and temperatures. The underlying assumption Wood 2010). Unlike the isolated point mea- here is that a weak relationship exists between surements provided by rain gauges, satellite- the observed temperature of clouds and rain based precipitation estimates offer greater rate, the idea being that lower temperatures spatial coverage of rainfall estimation with indicate clouds extending higher up into the higher temporal frequency than many of the atmosphere than their surroundings. While current gauging networks. Radar rainfall this relationship may hold for strongly con- offers high-resolution (about 1 kilometer), vective systems, with cumulonimbus clouds high-frequency (about 10 minutes) precipita- extending into the stratosphere, the relation- tion estimates for areas within about a 150- to ship is less solid for rain-producing clouds 300-kilometer radius of the radar location. (for example, stratiform) in the lower to mid- However, the estimates are known to be dle parts of the troposphere. Furthermore, the affected by beam blockage and greater uncer- well-known misregistration between the tainty moving away from the radar. For these location of the cloud top and the rain front reasons, as Gourley et  al. (2010) and other further compounds the lack of reliability of studies have shown, they can give poorer esti- geostationary-based rainfall estimation. mates compared with some SPPs. Neverthe- Observation frequency is another issue with less, where radar data are available and well SPPs, especially for the detection of extreme calibrated, radar rainfall can be useful for rainfall events (for example, AghaKouchak small-scale hydrologic prediction. However, et al. 2011). Most modern SPPs are derived pri- much of the global land area is “unobserved” marily through passive microwave sensors by ground-based rainfall radar systems, aboard polar-orbiting satellites, each with a which limits their use in large-area (espe- repeat frequency of typically more than one cially ­continental or global scale) water day. Satellite constellations mitigate the issue resources ­assessment. somewhat by potentially providing many snap- The coarse spatial resolution of many of the shots of an area from multiple polar-orbiting SPPs currently available is considered one of the platforms. For example, the TMPA product impediments to their widespread adoption by 3B42RT (table 6.3) uses data from any available the hydrologic modeling community and water passive microwave sensor within a 90-minute resources managers. Moreover, the fact that the window on either side of the synoptic precipitation products are retrievals derived time  (which is at three-hour intervals over a from brightness temperature observations day). However, given the typically short chapter 6 : E arth O bservations for M onitoring W ater R esources   |  85 duration and very localized nature of extreme By exploiting the accuracy of station-level convective rainfall events, the event could pass rain gauge measurements and the spatial cov- undetected or be underrepresented in the erage of gridded rainfall products, the blending derived products. of these two sources of information mitigates Studies evaluating SPPs and precipitation the shortcomings of the respective data sets to forecast from numerical weather prediction produce improved precipitation estimates. models have shown that SPPs do compara- The statistical blending of satellite-derived tively well at detecting “summer” rainfall, precipitation products and rain gauge mea- characterized by convective weather sys- surements has only been explored relatively tems, whereas weather model forecasts are recently to generate high-resolution rainfall better for “winter” rain, which is largely estimates at continental scales (Chappell et al. stratiform (Ebert, Janowiak, and Kidd 2007; 2013; Mitra et  al. 2009; Renzullo et  al. 2011; Sapiano et al. 2010). The complementarity of Rozante et al. 2010; Vila et al. 2009; Xiong et al. the satellite- and model-derived precipita- 2008). When further combined with reanaly- tion has spurred some researchers to con- ses, the results are often a great deal improved sider combining the two sources of (Sheffield, Goteti, and Wood 2006). information, for example, as a simple ensem- Renzullo et  al. (2011) explore the role of ble mean of the data sets (Peña-Arancibia satellite precipitation to enhance gauge-based et al. 2013) or through more statistics-based analysis in Australia (figure 6.3). They exam- merging approaches (Sapiano, Smith, and ine several statistical methods for blending a Arkin 2008). For instance, the Asia-Pacific TMPA near-real-time product (3B42RT) with Water Monitor of the Commonwealth Scien- gauge measurements from approximately tific and Industrial Research Organisation 2,000 stations distributed across Australia (CSIRO) uses a blending method that empha- reporting daily (that is, 24-hour accumulated) sizes precipitation estimates from the rainfall in real time. The blending of satellite TRMM satellite product for areas closer to estimates with gauge data resulted in a clear the equator and weather model precipitation improvement over the use of satellite data estimates from the European Centre for alone, and the satellite data imparted more Medium-Range Weather Forecasts for areas realistic patterns of rainfall distribution in the toward the poles.3 blended product than in “smoother,” gauge- only analyses. However, the quantitative eval- Blended Satellite- and Gauge-Based uation of the blended satellite-gauge rainfall Precipitation Analyses product using the independent set of post- Rain gauge measurements are typically not real-time rain gauge observation revealed used to retrieve rainfall data from satellite- that the estimates were no better than the based platforms (either geostationary or polar gauge-only analyses. Subsequent investiga- orbiting) or to conduct numerical weather pre- tion (reported in the supplementary material diction (reanalysis), except by some products of Chappell et al. 2013) shows that the result to correct retrospective bias in the rainfall esti- was due largely to the fact that the evaluation mates. Blending multiple precipitation data occurred predominantly in well-gauged parts sets has been the practice for many years of the continent, where the gauge analysis has among researchers in rainfall radar (for exam- lower error. The satellite-derived product ple, gauge-corrected reflectivity; Krajewski was only likely to improve estimation for 1987) and satellite-derived precipitation (for parts of the country with fewer than four example, merged passive microwave and TIR gauges per 10,000 square kilometers (equiva- imagery; Huffman et al. 1997). lent to a 1° x 1° cell; figure 6.4). 86  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Figure 6.3 Daily Rainfall Estimates for March 1, 2010, in Australia a. Multisatellite b. Rain gauge c. Blended Rainfall (mm) 0 10 20 30 40 50+ Source: CSIRO 2011. © CSIRO. Used with permission. Further permission required for reuse. Note: Panel a is from a National Aeronautics and Space Administration multisatellite rainfall product. Panel b is from analysis of rain gauges. Panel c is from combining the gauge and satellite rainfall estimates. The rain front shown led to widespread flooding in southern Queensland and northern New South Wales. Figure 6.4 Distribution of Real-Time Rain Gauges and Areas Where Satellite-Derived Precipitation Is Likely to Improve Accuracy of Rainfall Estimation in Australia (a) (b) (c) Sources: Renzullo et al. 2011 (panels a and b); Global Precipitation Climatological Center (http://gpcc.dwd.de) (panel c). © Water Information Research and Development Alliance (WIRADA) (panels a and b). Used with permission. Further permission required for reuse. Note: Panel a shows the location of the approximately 2,000 rain gauges reporting 24-hour accumulated rainfall in real time (the distribution is typical for any given day of the year). Panel b depicts the number of days in a year where the density of rain gauges is less than four gauges per 1° x 1° (satellite-derived precipitation is likely to improve rainfall estimates in the white regions). Panel c shows the number of gauges per 1° x 1° grid cell from the Global Precipitation Climatological Center. The factors affecting when and where sat- with multiple sources of precipitation data (for ellite data are expected to produce better example, gauge, radar, and forecasts). Shef- rainfall estimation include type of topography field, Goteti, and Wood (2006) use global, and rainfall in addition to gauge density. How- 1°  x  1° daily precipitation data, derived from ever, in large parts of the globe, rain gauge the Global Precipitation Climatology Project’s networks are sparse (figure 6.4, panel c) and Special Sensor Microwave Imager (SSM/I) there is growing evidence to suggest that sat- and gauge observations, to correct modeled ellite-derived precipitation, together with daily rainfall reanalysis from the National Cen- weather model reanalysis estimates, can pro- ter for Environmental Prediction and the vide highly valuable rainfall estimates and National Center for Atmospheric Research.4 narrow the information gap. Furthermore, they use the three-hourly TMPA SPPs can enhance global precipitation esti- 3B42RT rain rates to disaggregate the daily mation when the data are used in conjunction data temporally into three-hourly rainfall chapter 6 : E arth O bservations for M onitoring W ater R esources   |  87 estimates globally. The result is a precipitation is energy limited. In the second stage, water is data set with improved accuracy compared limiting, and as the soil dries and plants close with model prediction and, in some instances, their stomata, actual ET declines. The water- satellite products alone. limited part of the actual ET process is com- plex, depending on both biology (where there Evapotranspiration is vegetation) and meteorology. However, when Definition water is not limiting, energy-limited actual ET Evaporation is the phase change from a liquid to is determined primarily by four principal mete- gas. Evapotranspiration (ET) may occur from orological drivers: net radiation, air tempera- the Earth’s surface (the soil, a water body, or ture, relative humidity, and wind (McVicar other type of surface), through plant leaves et  al., “Global Review and Synthesis,” 2012a; (termed transpiration), and from rainfall on the McVicar et al., “Less Bluster Ahead?” 2012b). surface of leaves (termed interception). While the term evapotranspiration covers these three com- Relevance ponents, interception is not explicitly used in the Actual evapotranspiration connects many of compound word. Evaporation, like precipitation, the Earth’s hydrologic and related environ- has the dimensions of depth per time, and com- mental processes at local, regional, and mon units are millimeters per hour, per day, or global scales. For example, actual ET links per year. When spatially integrated over an area the water balance to the energy balance, veg- such as a paddock, catchment, basin, or country, etation to hydrology, and hydrology to cli- the dimensions become volume per time, and mate. Actual evapotranspiration is both a common units are cubic meters per day. matter of environmental physics and biology, Actual evapotranspiration is difficult to mea- as it is governed by the conditions of the sure at a single location (Leuning et al. 2012), let physical environment at the surface and in alone to estimate accurately both spatially and the atmosphere, but also by photosynthesiz- temporally over large areas. This is different ing vegetation, which transpires water to than potential evapotranspiration, which can assimilate carbon (described later in this be readily calculated using commonly mea- section). Through vegetation, actual evapo- sured meteorological variables (Donohue, transpiration is the primary hydrologic McVicar, and Roderick 2010) or instrumental “lever” by which man can either inadver- equivalents of potential ET such as pan evapo- tently alter or actively manage the water ration, which can be readily measured (Roder- cycle. As such, it is relevant to water manage- ick, Hobbins, and Farquhar 2009a, 2009b). ment in agriculture, environmental services, The distinction between actual evapotrans- climate change, and hydropower (see tables piration, potential evapotranspiration, and pan 5.2 and 5.3). For models of actual ET to be evaporation is important (McMahon et  al. relevant to water management, they ideally 2013). Potential ET and pan evaporation are should have the following characteristics: estimates and measurements, respectively, of • Have coverage that is suited to the purpose atmospheric evaporative demand under envi- ronmental conditions with limitless access to • Be spatially and temporally dynamic at water, so they are not representative of actual moderate to high resolution evapotranspiration when and where the sur- • Accurately close the energy and mass bal- face is not saturated. Actual ET can be concep- ance (as a means of quality control). tualized as a two-stage process. In the first stage, following sufficient precipitation or irri- For the first two characteristics, optical EO gation, water is freely available, and actual ET data are often used for spatial modeling of 88  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T actual evapotranspiration. Closing the energy Empirical methods have also been developed and mass balance from Earth observation is to estimate actual ET from surface temperature generally problematic, however, because it derived from thermal EO data and a vegetation requires estimating 24-hour latent heat flux index (traditionally the normalized difference from as few as a single measurement made at vegetation index or NDVI). This method has a specific time of the day (Kalma, McVicar, been called the “triangle” or “trapezoid” and McCabe 2008; Van Niel et al. 2012). method, describing the general shape of the sur- face temperature versus NDVI data  space Theoretical Basis of Remote Sensing of Actual (Lambin and Ehrlich 1996). The extremes of the Evapotranspiration surface temperature axis of this data space form Empirical methods of estimating actual ET the “cool edge” and “warm edge,” representing take advantage of the numerous links that it has more or less actual ET, respectively. The NDVI to the energy balance, the water balance, and axis represents the amount of green vegetation vegetation, which allow for various functional cover. The end members of the data space, then, relationships to be established. For well- represent the maximum and minimum evapo- vegetated surfaces, the largest component of ration and the maximum and minimum transpi- actual ET is usually transpiration. This means ration. This method provides a linearization of that a good estimate of actual ET can some- the ratio of actual to potential ET (see Van Niel times be made using simple statistical relation- and McVicar 2004 for a detailed description). ships with remotely sensed vegetation indexes The triangle method is suited for estimating that reflect the dynamics of vegetation green- relative amounts of actual ET over local to ness. For example, Nagler et  al. (2007, 2009) regional areas, but is less adept at modeling and Yebra et al. (2013) find that the enhanced absolute amounts of actual ET over large basins vegetation index (EVI) scales actual ET well. or continents or comparing actual ET from one Although this simple relationship is likely to region to another. work well in many places, it is not useful The main advantage of empirical methods everywhere. For instance, actual ET over a is their simplicity; their main disadvantages water body might be very high, but this would are their reliance on ground measurements of be missed when relating actual evapotranspi- ration to EVI. As actual ET is also part of the Figure 6.5 Examples of Actual Evapotranspiration Estimates for Region in water balance, it should, at least for certain Western Queensland, Australia, during Flow Event in February 2004 times and places, have a strong relationship to a. CMRSET model b. PML model remotely sensed moisture indexes. In particu- lar, areas of the world that are water limited might be well modeled with a simple relation- ship to a metric of moisture availability. For example, to estimate actual ET over arid Aus- tralia, Guerschman et al. (2009b) use both the EVI and the global vegetation moisture index with monthly precipitation to define a coeffi- cient useful for scaling potential ET. They find that this method performs well compared to a Source: Reproduced with permission from King et al. 2011. © WIRADA. Used with permission. Further variety of other methods, most of which are permission required for reuse. much more complicated to implement (King Note: The CMRSET model (Guerschman et al. 2009b) uses remotely sensed vegetation and moisture et al. 2011). Figure 6.5 illustrates the effects of indexes. The PML model (Zhang et al. 2008) is based primarily on vegetation dynamics. The color scale is the same for the two images (blue-red = 0–5 millimeters per day). The location of the region basing ET estimates primarily on vegetation. shown is 139E-142E, 23S-26S. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  89 actual ET and, generally, the inability to be et al. 1998; Menenti and Choudhury 1993; Nor- improved via better process understanding. man et al. 2003; Su 2002). The energy balance That is, once more or better ground data approach, and in particular the two-layer become available, it will probably be possible model, has the inherent risk of being over- to improve these types of models by optimizing parameterized compared to the data likely to the fit, but they are mostly statistical in nature be available for modeling, especially over large so they have limited capacity to inform process basins or continents. Therefore, many attempts understanding. They also generally do not have been made to make the model parsimoni- allow for better estimation based on improved ous with the data available for modeling large understanding of the system. Nonetheless, areas (Roerink, Su, and Menenti 2000; Sobrino empirical approaches can be a practical way to et al. 2005). estimate actual ET. Energy balance methods often first estimate relative evaporation (that is, the evaporative Energy Balance Methods fraction) by defining “hot” and “cold” pixels The Earth’s surface is heated by solar radiation from the image (for example, SEBAL, the Sur- and loses this heat through long-wave radia- face Energy Balance Algorithm for Land; Bas- tion, sensible heat flux (that is, heating the air), tiaanssen et al. 1998) or by defining hypothetical and latent heat flux (that is, using energy to “dry” and “wet” conditions determining the evaporate water). Surface temperature can be theoretical limits to evaporation (for example, used to estimate sensible heat flux. If net radia- SEBS, the Surface Energy Balance System; Su tion (that is, incoming minus outgoing radia- 2002). For an algorithm like SEBAL, for exam- tion) and changes in heat storage can be ple, identifying appropriate “hot” or “cold” estimated as well (Zhu et al. 2014), latent heat pixels is paramount, making it sometimes flux can be calculated as the difference between suited to use in agricultural areas with adja- all of these terms because of the requirement cent dryland and irrigated types of land cover. for energy balance (that is, the law of conserva- In other environments with less hydrologic tion of energy). contrast or when the area of interest is too Most energy balance methods for estimating large (covering drastically different climate actual ET use satellite-derived radiometric or zones), however, representative “hot” or “cold” “skin” temperature data. Some of these models pixels may not be readily found, making SEBAL explicitly recognize that surface temperature less suitable. Furthermore, energy balance measured by satellite is sometimes insufficient methods usually do not explicitly consider to solve the energy balance accurately by mod- evaporation due to interception, even though eling two separate “layers,” one in the vegeta- interception can represent more than 20 per- tion canopy and one at ground surface, cent of precipitation for certain types of vege- generally called two-layer models. They tation (Miralles et al. 2010). approximate a set of simultaneous equations Most common applications of energy bal- that estimate an equal number of unknowns ance methods have used data from polar- (Friedl 1995; Jupp et  al. 1998), one of them orbiting platforms such as AVHRR (Advanced being the effective surface temperature that is Very High Resolution Radiometer), MODIS, required in the energy balance equation. and Landsat (table 6.4). Resultant values rep- Energy balance methods require modeling of resent an “instantaneous” flux, which available energy and aerodynamic resistance, requires scaling to actual ET integrated over a which are potential sources of uncertainty. longer time period (for example, a day or a Various implementations of the energy bal- month) to be relevant to hydrology. Subse- ance method have been developed (Anderson quent scaling of latent heat to daily or monthly et al. 2007, 2011; Bastiaanssen et al. 1998; Jupp actual ET is a source of considerable 90  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 6.4 Overview of Sensors Most Suitable for Estimating Actual Evapotranspiration from EO Data ORBIT AND RAW DATA COST DATE LAUNCHED SATELLITE SENSOR PIXEL SIZE SPECTRAL PER SQUARE (END) OR SYSTEM (METERS) BANDS REVISIT CYCLE KILOMETER (US$) PLANNED LAUNCH EMPIRICAL PM LAI REBM Polar orbiting MODIS 250–1,000 29 2 times a day Free 2000 ❶ ❶ ❶ VIIRS and JPSS 375–750 14 2 times a day Free 2012 ❶ ❶ ❶ AVHRR 1,000 4 Daily Free 1981 ❷ ❶ ❶ Landsat 5 TM 30–90 7 16 days Free 1985 (2012) ❶ ❶ ❶ Landsat 7 ETM+ 30–60 8 16 days Free 2000 ❶ ❷ ❷ Landsat 8 30–100 11 16 days Free 2013 ❶ ❶ ❶ Geostationary GOES (2nd and 3rd 1,000–4,000 4 15 minutes Free 1994 ❷ ❶ ❶ generation) Meteosat (2nd 1,000–3,000 7 15 minutes Free 2002 ❶ ❶ ❶ generation) Himawari-8 500–2,000 10 15 minutes Free 2014 ❶ ❶ ❶ Note: The suitability of each sensor to provide data useful for the three classes of models is shown with numbers and colors, as follows: ❶ highly suitable, ❷ suitable. AVHRR = Advanced Very High Resolution Radiometer; ETM+ = Enhanced Thematic Mapper Plus; GOES = geostationary operational environmental satellite; JPSS = Joint Polar Satellite System; MODIS = Moderate Resolution Imaging Spectrometer; PM LAI = Penman-Monteith leaf area index; REBM = resistance energy balance model; TM = Thematic Mapper; VIIRS = Visible/Infrared Imager Radiometer Suite. uncertainty (McVicar and Jupp 2002; Van EO data are not a pure solution to this problem Niel et al. 2011, 2012). However, because the due to issues regarding specific time-of-day energy balance approach makes use of acquisition, cloud cover, and differences “instantaneous” surface temperature, it is between the radiometric “skin” temperature directly suited to the use of geostationary received at the sensor and the effective surface data. For example, the algorithm disALEXI temperature that solves the energy balance at uses geostationary data over North America ground level. In days prior to operational EO- to observe the change in surface temperature based land surface temperature products, the during the morning (Anderson et  al. 2011), so-called Penman-Monteith “combination making it suitable for modeling flux, which is equation” was derived, eliminating the need to closer to the theoretical nature of the phe- estimate surface temperature (Monteith 1965, nomena being estimated. The advantage of 1981; Penman 1948). the energy balance method is that it counters The Penman-Monteith equation combines the main disadvantage of the empirical the aerodynamic formulation of actual evapo- approaches: energy balance methods are emi- transpiration and the energy balance with an nently suited to inform and be improved by approximation of the saturation vapor pres- better process understanding. sure calculated at surface temperature. The problem with using the Penman-Monteith Penman-Monteith Methods equation, however, is that, although the need One of the main obstacles to calculating actual to know surface temperature was eliminated, evapotranspiration from an energy balance it was replaced by a different unknown param- perspective is the need to derive the effective eter—the surface (or canopy) conductance surface temperature of an area. This is particu- (sometimes written in the form of resistance, larly problematic when the area of interest in which case it would be the surface or canopy (that is, a pixel) is heterogeneous (for example, resistance). While the number of unknown partly vegetated and partly bare soil). Thermal parameters remains unchanged, one chapter 6 : E arth O bservations for M onitoring W ater R esources   |  91 advantage is that this model allows for the Furthermore, the so-called FAO-56 method unknown parameter to be addressed in a dif- (Allen et al. 2007) can be seen as an intermedi- ferent way: through conductance. Conduc- ate method in that it calculates a hypothetical tance is a parameter associated with evapotranspiration for an idealized crop and transpiration and carbon assimilation in the then applies an empirical method to scale this process of photosynthesis, so it allows estima- hypothetical ET factor. The satellite-based tion of actual ET through vegetation method developed by Guerschman et  al. characteristics. (2009b) is akin to the FAO-56 approach. Since The Penman-Monteith equation is a process- it is suitable for use with higher-spatial- based model, so it is in a different category than resolution sensors like Landsat ETM+, it (and the empirical relationships primarily using sta- methods like it) can be used for estimating tistically fitted relationships between actual ET actual evapotranspiration at the level of an irri- and vegetation indexes. It is based, to  a large gation scheme, farm, or even a field (figure 6.6). degree, on the energy balance, but as surface Because there are various global EO-based conductance is commonly modeled through vegetation data sets, the Penman-Monteith remotely sensed vegetation products like a veg- equation is commonly used to estimate actual etation index or a leaf area index (LAI), it is also ET. However, just as is the case for purely considered to be in a different category than the empirical relationships, the Penman-Monteith energy balance methods discussed above. How- approach will not necessarily perform well ever, the three types of methods sometimes where actual ET is not driven primarily by tran- overlap, making their distinction less clear. spiration. To implement the approach over vast For example, surface conductance is an areas, it also relies on a model of available unknown, and only empirical methods are energy and aerodynamic conductance, which available to estimate it. Yebra et  al. (2013) may add considerable error or uncertainty compares fully empirical approaches with to  the estimation. The Penman-Monteith Penman-Monteith approaches, including approach models actual ET through the link approaches based on the MODIS leaf area between vegetation and energy balance; thus index and on the Guerschman et al. (2009b) the main impediment to its implementation is crop factor approach. They conclude that, the need to determine surface conductance. among these, the best approach to modeling However, due to the reliance of the Penman- actual ET is through the use of an empirical Monteith method on vegetation dynamics and relationship to estimate surface conductance stomatal conductance, the relationship between from vegetation indexes, where each of the this actual ET estimate and crop growth or three indexes tested (enhanced vegetation gross primary production modeling is closer. index [EVI], normalized difference vegetation Methods have been developed that use sat- index [NDVI], and the Guerschman Kc index) ellite LAI products to estimate surface conduc- has specific strengths and weaknesses. tance from an assumed (or optimized) leaf-level Alternatively, surface conductance can be stomatal conductance value. Some readily modeled using estimates derived from field available global vegetation LAI products from measurements, whether by upscaling leaf- Earth observation have promoted the develop- level measurements of stomatal conductance ment of LAI-scaled global ET estimates (Mu, (Kelliher et  al. 1995) or by using field-level Zhao, and Running 2011; Zhang et al. 2012). estimates of surface conductance derived from lysimeters or other water balance methods, or, Past, Present, and Future Sensor Availability more commonly in recent years, by using for Mapping Actual Evapotranspiration micrometeorological methods and flux tower Table 6.4 lists some existing and planned sen- Eddy covariance measurements, in particular. sors that can provide estimates of actual ET. 92  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Figure 6.6 Mapping of Actual Evapotranspiration Using High-Resolution Satellite Images for Part of Lower Gwydir Region in New South Wales, Australia a. January 31, 2005 b. April 3, 2005 ETa (mm/day) 0.0 3.5 7.0 Source: Emelyanova et al. 2012. © CSIRO. Used with permission. Further permission required for reuse. Note: The top row shows false color composites of the original Landsat TM (thematic mapper) imagery; the bottom row shows estimated actual evapotranspiration rates. The colors in the top row correspond to vigorous vegetation (green), open water (black-blue), and dry land (purple). Most optical sensors can be used for mapping models just described. Examples of these three actual ET for at least one of the three general categories, with relevant references, are pro- categories. Those sensors that include bands vided in table 6.5. As discussed above, each in the visible (VIS) and near-infrared (NIR) approach has strengths and weaknesses. The spectrum are generally suited to empirical simplicity of the empirical approach is offset methods using vegetation indexes and to the by its inability to inform and be informed by Penman-Monteith approach (which also process understanding. The ability of the Pen- requires meteorological data). Sensors having man-Monteith approach to estimate actual ET short-wave infrared (SWIR) bands in addi- better through vegetation dynamics is offset by tion to VIS-NIR bands allow for determining its inability to model water bodies or soil evap- empirical relationships with moisture indexes oration. The ability of the energy balance and for estimating actual ET from surfaces approach to inform process understanding and with no vegetation, including water bodies. use geostationary thermal data is offset by dif- Many of the optical sensors also record ther- ficulties in scaling instantaneous observations mal data, making them suitable for both to longer time periods and relative ET to abso- empirical and energy balance approaches lute ET as well as by model complexity. Fur- that use surface temperature. MODIS, Land- thermore, while the approaches have been sat, and VIIRS (Visible/Infrared Imager classified into three categories, specific Radiometer Suite) are examples of sensors ­ implementations may sometimes blur these that acquire the data useful for all three cate- ­distinctions. gories of models. Sensors like AVHRR do not While it is unlikely that any single approach have SWIR bands, but are useful in the vege- will be best suited to estimate actual ET for all tative index and surface temperature-based situations, a common relevant issue is having a approaches. system in place for robust and repeatable assess- ment of ET models. For instance, the ET inter- Existing RS-Based Data Products and comparison and evaluation framework within Services for Actual Evapotranspiration Australia was designed to assess eight conti- There are numerous EO-based implementa- nental models of actual ET to help to inform tions of the three categories of actual ET the Australian Water Resources Assessment chapter 6 : E arth O bservations for M onitoring W ater R esources   |  93 Table 6.5 Examples of Studies Using the Three General Classes of Actual ET Models TYPE OF MODEL AND EO ALGORITHM REFERENCES Empirical CMRSET Guerschman et al. 2009b Surface temperature versus NDVI Lambin and Ehrlich 1996 Actual ET versus EVI Nagler et al. 2007, 2009 Penman-Monteith Unnamed Yebra et al. 2013 Unnamed Cleugh et al. 2007 PML Leuning et al. 2012; Zhang et al. 2008 MODIS ET Mu, Zhao, and Running 2011 Energy balance SEBAL Bastiaanssen et al. 1998 SEBS Su 2002 ETWatch Wu et al. 2012 S-SEBI Roerink, Su, and Menenti 2000; Sobrino et al. 2005 dis(ALEXI) Anderson et al. 2007; Norman et al. 2003 NDTI Jupp et al. 1998; McVicar and Jupp 2002 Note: ALEXI = Atmosphere-Land Exchange Inverse; CMRSET = CSIRO MODIS Reflectance-based Scaling ET; EO = Earth observation; ET = evapotranspiration; EVI = enhanced vegetation index; MODIS ET = Moderate Resolution Imaging Spectrometer evapotranspiration; NDTI = normalized difference temperature index; NDVI = normalized difference vegetation index; PML = Penman-Monteith-Leuning; SEBAL = Surface Energy Balance Algorithm for Land; SEBS = Surface Energy Balance System; S-SEBI = Simplified Surface Energy Balance index. (AWRA; see appendix B). Key to this type of Soil Moisture assessment framework is some form of field Definition measurement, which can be used in the ET Soil moisture is defined as the amount of water model itself (for empirical methods) and for in the uppermost layers of the soil column, validation of all types of models. So-called Eddy where the definition of “uppermost” varies covariance flux data are often used for these depending on sensing technology or modeling purposes.5 application and can vary from the top 1 centi- For places where no ground measure- meter to the first 1 meter of soil or more. Soil ments exist, EO-based models of actual ET moisture is highly variable in space and time, can still be used, but in the absence of error and its importance to water resources is appar- assessment, their suitability for management ent via the link between key water balance purposes may not be known with certainty. terms and hydrologic processes in the soil col- Figure 6.7 illustrates the accuracy that might umn. As a measure of catchment antecedent be expected for a forested site, where actual moisture6 condition, soil moisture affects the ET is not overly challenging to estimate. Typ- amount of evaporation from soil, transpiration ically, at moderate to low resolution (about by vegetation, and partitioning of rainfall into 5  kilometers, monthly time step), actual ET infiltration and surface runoff. Soil moisture can be estimated to within 1 millimeter per has had a role in characterizing hydroclimate day or better. Of course, this will depend on and monitoring the effects of climate change site, algorithm, and data characteristics. (for example, drought monitoring; Bolten et al. Other important considerations are reliabil- 2010) for more than 20 years, but this role has ity, maturity, and complexity of the system been “formalized” only relatively recently by required to produce the estimate of actual ET. its listing among the World Meteorological 94  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T a. Estimated actual ET 5 4 ET flux (mm/day) 3 2 1 0 Figure 6.7 Comparison of Actual ET Estimates 2001 2002 2003 2004 2005 2006 Derived from the NDTI Model with Actual ET b. Measured actual ET Measurements from the Tumbarumba, NSW, Flux 2.0 Tower 1.5 Source: Reproduced from King et al. 2011. © WIRADA. Used with ET model – flux data 1.0 permission. Further permission required for reuse. 0.5 Note: In the top panel, the blue dots represent actual ET estimates 0 from the normalized difference temperature index model (Jupp et al. –0.5 1998; McVicar and Jupp 2002), which are compared to actual ET measurements from the Tumbarumba, NSW, flux tower (black –1.0 dashes). The red dots in the bottom panel show the differences –1.5 Mean: –0.2126 SD: 0.6692 between model estimates and the flux tower measurements (in units –2.0 of millimeters per day). ET = evapotranspiration; NDTI = normalized 2001 2002 2003 2004 2005 2006 difference temperature index. Organization’s Global Climate Observing Sys- through calibration and data assimilation (Zreda tem essential climate variables (Bojinski et al. et al. 2012). Both field-based and proximal sens- 2014). ing are valuable information sources in their Soil moisture monitoring has advanced con- own right, and they provide essential data for siderably over the last decade, with burgeoning the evaluation and calibration of satellite- innovative ground- and satellite-based technolo- derived and modeled soil moisture products, gies for large-area monitoring (for a comprehen- helping to build confidence in their accuracy. sive summary, see Ochsner et  al. 2013). These Global monitoring of soil moisture is only technologies include coordinated global net- achievable with satellite Earth observation in works of field-based sensors (Dorigo et al. 2013; conjunction with field-based soil moisture Smith et al. 2012) and a novel proximal sensing monitoring networks. Satellite soil moisture technique based on cosmic ray detectors (Desi- products have been derived from a contiguous lets, Zreda, and Ferré 2010; Zreda et al. 2012). series of space-borne sensors spanning 30 years While field-based detectors can measure (Liu et  al. 2012). However, the first dedicated moisture very accurately at various depths for a soil moisture monitoring mission from space point in the landscape, cosmic ray probes pro- was not launched until 2009 (Kerr et al. 2012). vide an integrated root-zone moisture measure- Time series of satellite soil moisture have been ment for an area of about 600 meters in diameter used in climate studies (Jung et  al. 2010; Liu (Desilets and Zreda 2013)—an area suitable for et  al. 2009; Seneviratne et  al. 2010) and have many agricultural and land management appli- refined our understanding of rainfall genera- cations. Cosmic ray probes have been applied tion processes (Taylor et al. 2012). beyond root-zone soil moisture sensing, includ- The assimilation of satellite soil moisture ing estimating aboveground biomass (Franz into land surface models has been shown to et al. 2013) and constraining land surface models improve soil water representation in the chapter 6 : E arth O bservations for M onitoring W ater R esources   |  95 models (Draper et  al. 2012; Renzullo et  al. thereafter with several field experiments involv- 2014) and led to improvements in estimated ing tower-mounted and airborne microwave evaporative fluxes, drainage, and discharge radiometers (Jackson and Schmugge 1989). (Brocca et al. 2012; Dharssi et al. 2011; Draper Until that point, the primary use of microwave et  al. 2011; Pipunic et  al. 2013; Reichle and instruments on satellites had been communica- Koster 2005). It is through integration with tions, monitoring of snow and sea ice extent, landscape hydrology models and field-based and atmospheric soundings of temperature and monitoring networks (via calibration and data moisture. It was only in the early 2000s that assimilation) that the satellite soil moisture these space-based microwave sensors started products offer greatest potential for monitor- being used to estimate soil moisture, with the ing large-area water resources, particularly as first global satellite soil moisture products avail- a constraint for parts of the Earth where tradi- able in 2002 (de Jeu and Owe 2003; Wagner tional ground observation networks have et al. 2003; figure 6.8). sparse, intermittent, or no coverage at all. Satellite soil moisture sensing technology is based on either radiometric measurements of Brief Summary of Soil Moisture emissions from the soil (the so-called passive Sensing from Space microwave approach) or radar technology The dielectric properties of soil are greatly transmitting pulse of electromagnetic radiation altered by the amount of liquid water present in to the Earth’s surface and measuring the back- the soil. The relationship between moisture in scattered signal (the so-called active approach). the soil and emitted radiation (about 1–20 giga- One of the defining characteristics separating hertz or a 1.5–30-centimeter region of the elec- active and passive sensors is their contrasting tromagnetic spectrum) has been conceptually spatial resolution: passive sensors require large understood since the 1970s and encapsulated in integrating areas for adequate signal-to-noise various physical models (Dobson et  al. 1985; (ratios)7 so the instantaneous field-of-view Wang and Schmugge 1980). The potential (pixel) has a resolution of 30–120 kilometers, for Earth observation to measure soil moisture whereas the resolution of active systems, for a on a small scale was demonstrated shortly given frequency, is a function of beam width, pulse duration, and satellite antenna length. Figure 6.8 Remote Sensing–Based Soil Moisture Monitoring This means that the resolution required to sus- tain a good signal-to-noise ranges from about 10 meters to 10 kilometers. Examples of active and passive satellite EO systems used in the production of global soil moisture products are listed and defined in table 6.6. They include SSM/I, TMI (TRMM Microwave Imager), AMSR-E (Advanced Microwave Scanning Radiometer for EOS), AMSR2 (Advanced Microwave Scanning Radiometer2), and SMOS AMSR-E soil moisture [m3 m–3] (Soil Moisture and Ocean Salinity Sensor) for 0 0.1 0.3 0.5 0.7 passive radiometry and ERS (European Remote Sensing Satellite), ASAR (Advanced Synthetic Source: CSIRO, using data from Owe, de Jeu, and Holmes 2008. © CSIRO. Used with permission. Further permission required for reuse. Aperture Radar), and ASACT (Advanced Scat- Note: Satellite soil moisture products for January 2, 2006, derived by applying the retrieval algorithm terometer) for active scatterometry. of Owe, de Jeu, and Holmes (2008) to the descending passes of the AMSR-E sensor aboard NASA’s Until recently, satellite soil moisture prod- Aqua satellite. AMSR-E = Advanced Microwave Scanning Radiometer for EOS; NASA = National Aeronautics and Space Administration. ucts were typically derived from X- and C-band 96  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 6.6 Overview of Key Characteristics of Soil Moisture Sensors Aboard Past, Current, and Near-Future Satellite Platforms SOIL MOISTURE RESAMPLED SATELLITE MISSION LIFE PRODUCT RESOLUTION SWATH GLOBAL KEY SENSOR PLATFORM SPAN PROVIDER TYPE UNITS (KILOMETERS) (KILOMETERS) COVERAGE REFERENCE DATA ACCESS 3 -3 Aquarius SAC-D June 2010– NSIDC Active and m m 100 390 Daily: 7-day Bindlish and ftp://n5eil01u.ecs.nsidc.org/ passive L compositea Jackson 2013 SAN/AQUARIUS/ band ASAR Envisat March ESA Active C band % 1 100–405 3–8 days Wagner et al. http://rs.geo.tuwien.ac.at/ 2002– April 2012 2008 products/ ASCAT MetOp October 2006– TUW Active C band % 12.5 550 1.5 days Wagner, Registered user: ftp.ipf.tuwien. (-A, -B) Lemoine, and ac.at Rott 1999 AMSR-E Aqua May NSIDC Passive X g cm-3 25 1,445 Daily Njoku et al. 2003 ftp://n4ftl01u.ecs.nasa.gov/ 2002– October band SAN/AMSA/ 2011 AMSR-E Aqua May VUA Passive C m3 m-3 25 1,445 Ascending and Owe, de Jeu, and Level-2 swath data: ftp://hydro1. 2002– October band descending Holmes 2008 sci.gsfc.nasa.gov/data/s4pa/ 2011 WAOB/LPRM_AMSRE_ SOILM2.00 2 Level-3 ascending 0.25° gridded data: ftp://hydro1.sci.gsfc.nasa.gov/ data/s4pa/WAOB/LPRM_ AMSRE_A_SOILM3.002/ Level-3 descending 0.25° gridded data: ftp://hydro1.sci.gsfc.nasa.gov/ data/s4pa/WAOB/LPRM_ AMSRE_D_SOILM3.002/ Level-4 assimilated root-zone soil moisture: ftp://hydro1.sci. gsfc.nasa.gov/data/s4pa/ WAOB/LPRM_AMSRE_D_ RZSM3.001/ AMSR-E Aqua May 2002– NSIDC Passive X m3 m-3 25 1,445 Ascending and Jones and ftp://sidads.colorado.edu/pub/ October 2011 band descending Kimball 2012 DATASETS/nsidc0451_AMSRE_ Land_Parms_v01/ AMSR2 GCOM-W1 May 2012– JAXA Passive C m3 m-3 50 1,450 Ascending and Koike 2013 https://gcom-w1.jaxa.jp/auth. band descending html AMSR2 GCOM-W1 May 2012– VUA Passive C m3 m-3 10 and 25 1,450 Ascending and Parinussa et al. http://globalchange.nasa.gov/ chapter 6 : E arth O bservations for M onitoring W ater R esources   |  97 band descending 2014 (Continued) Table 6.6 (Continued) SOIL MOISTURE RESAMPLED SATELLITE MISSION LIFE PRODUCT RESOLUTION SWATH GLOBAL KEY SENSOR PLATFORM SPAN PROVIDER TYPE UNITS (KILOMETERS) (KILOMETERS) COVERAGE REFERENCE DATA ACCESS AMI ERS (-1,-2) July 1991–June TUW Active C band % 25 500 3 days Wagner, http://rs.geo.tuwien.ac.at/ 2011 Lemoine, and products/ Rott 1999 MIRAS SMOS November ESA Passive L m3 m-3 50 1,000 3 days Kerr et al. 2012 Registered users: http://www 2009– band .esa.int/Our_Activities/ December 2014 (multiple- Observing_the_Earth/The_ angle) Living_Planet_Programme/ Earth_Explorers/SMOS/ Overview C-SAR Sentinel-1a April 2014– ESA Active C band % 0.004–0.08 80–400 5 days Unknown Unknown SMAP SMAP 2015– NSIDC Active and 11 1,000 1.5 days Entekhabi et al. NSIDC site passive L 2010 (launch 2015) band SSM/I DMSP October 1987– VUA Passive K 25 1,700 Daily Owe, de Jeu, and Registered users: http://www band Holmes 2008 .esa-soilmoisture-cci.org SMMR Nimbus November VUA Passive C, X 25 780 Ascending and De Jeu and Owe Registered users: http://www 1977–December band descending 2003 .esa-soilmoisture-cci.org 1987 TMI TRMM November VUA Passive X 25 878 Equatorial daily Owe, de Jeu, and Level-2 swath data: ftp://hydro1. 1997– band Holmes 2008 sci.gsfc.nasa.gov/data/s4pa/ WAOB/LPRM_TMI_SOILM2.001/ Level-3 0.25° gridded daytime data: ftp://hydro1.sci.gsfc.nasa. 98  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T gov/data/s4pa/WAOB/LPRM_ TMI_DY_SOILM3.001/ Level-3 0.25° gridded nighttime data: ftp://hydro1.sci.gsfc.nasa. gov/data/s4pa/WAOB/LPRM_ TMI_NT_SOILM3.001/ WindSat Coriolis January 2003– NOAA Passive C 25 1,025 1.5 days Li et al. 2010 http://www.ospo.noaa.gov/ band Products/land/smops/ Windsat Coriolis January 2003– VUA Passive C 25 1,025 1.5 days Parinussa Registered users: http://www band Holmes, and de .esa-soilmoisture-cci.org Jeu 2012 Note: Microwave bands L, C, X, and K correspond to frequency ranges 1–2, 4–8, 8–12, and 18–26 GHz, respectively. The end-of-life dates for some of the current missions are speculative. AMI = active microwave instrument; ASAR = Advanced Synthetic Aperture Radar; ASCAT = Advanced Scatterometer; AMSR-E = Advance Microwave Scanning Radiometer–Earth Observing System; C-SAR = Circular Synthetic Aperture Radar; ESA = European Space Agency; JAXA = Japan Aerospace Exploration Agency; MIRAS = Microwave Imaging Radiometer with Aperture Synthesis; NOAA = National Oceanic and Atmospheric Administration; NSDIC = National Snow and Ice Data Centre; SMAP = Soil Moisture Active Passive; SMM/I = Special Sensor Microwave Imager; SMMR = Scanning Multi-channel Microwave Radiometer; TMI =TRMM Microwave Imager; TUW = Vienna University of Technology; VUA = VU University Amsterdam; m3 m-3 = cubic meter of water per cubic meter of soil; g cm-3 = grams per cubic centimeter. a. Seven-day composite is based on images collected over 7 consecutive days. The benefit is that it eliminates most cloud cover found in daily images. microwave signals (8–12 and 4–8 gigahertz fre- moisture retrievals from active and passive sen- quency range, respectively), which means that sors due to their respective performance across their values correspond to emissions or back- different landscapes. For example, Dorigo et al. scatter from the top 1–2 centimeters of soil. (2010) derived the error structure of the AMSR- The launch of SMOS (Barré et al. 2008) ush- E (passive) and ASCAT (active) soil moisture ered in a new era of L-band (1–2 gigahertz or products over the globe using a statistical tech- 15–30 centimeters) sensing technology dedi- nique called triple collocation (Scipal et  al. cated to monitoring soil moisture in the top 5 2008). Examination of the error patterns centimeters of soil, which will continue with showed that AMSR-E errors were largest in the scheduled launch of Soil Moisture Active landscapes with a moderate to high level of tree Passive (SMAP) (Entekhabi et al. 2010). cover, due to the influence of vegetation on the The SMAP mission of the National Aero- emitted signal, while ASCAT errors were larg- nautics and Space Administration (NASA) est in dry arid areas, due to the scattering prop- launched in January 2015 is the first dedicated erties of dry soil and undulations (dunes) in soil moisture sensing mission to combine both those landscapes. Others have reported similar active and passive sources for high-resolution findings; for example, Draper et al. (2012) show (about 10 kilometers) mapping of soil moisture that assimilating ASCAT into NASA’s catch- for the globe on a daily basis. ment model led to significantly  less accurate estimates of root-zone moisture over highly Satellite Soil Moisture Products variable terrain compared to using AMSR-E. Satellite soil moisture products are generated Comparing the accuracy of satellite soil by different groups around the world, includ- moisture estimates with surface measure- ing government research agencies (NASA, ments is necessary to gain acceptance of the JAXA, European Space Agency [ESA], the products by the user community and often National Snow and Ice Data Center [NSIDC]) involves evaluations against field-based soil and universities (VU University of Amsterdam, moisture measurements, such as individual Vienna University of Technology). Different soil moisture products (Albergel et  al. 2011), groups have used different retrieval algorithms alternative soil moisture products from the to derive soil moisture from brightness tem- same sensor (Draper et al. 2009), or soil mois- perature observations by the same satellite ture products across sensors (Su et  al. 2011, sensors. For example, the University of Amster- 2013). However, care must be taken to use con- dam and NSIDC derive soil moisture from sistent definitions when comparing soil mois- C-band brightness temperature data from the ture values using model and field-based AMSR-E, but they employ different retrieval measurements. Differences may be observed schemes and radiative transfer model parame- due to incompatibility of soil moisture units, terization (detailed in Owe, de Jeu, and Holmes spatial resolution (that is, from point to pixel), 2008 and Njoku et al. 2003, respectively). Dif- sampling depth (emission depth), as well as ferent products may represent soil moisture differences in the product-processing meth- values quite differently; for instance, values ods. Given the range of potential sources of based on radiative transfer equations are typi- inconsistency, drawing conclusions from cally expressed in volumetric or gravimetric observed differences poses difficulties (Leroux units, while scatterometer-derived estimates et al. 2013; Wagner et al. 2003). are expressed in percentage wetness or degree Investigations often reveal that no one soil of saturation (0–100 percent). moisture product is “best” for all locations and Beyond differences in resolution, there is applications and that it is advisable to exploit demonstrated complementarity of satellite soil the complementarity between products chapter 6 : E arth O bservations for M onitoring W ater R esources   |  99 (active, passive, and modeled) and to generate Global maps of the freeze-thaw state have merged soil moisture estimates (Draper et  al. been derived from satellite microwave sensors 2012; Liu et  al. 2012; Renzullo et  al. 2014). spanning the last 30 years at 0.25° resolution Knowing how each source of soil moisture (Kim et  al. 2011). The freeze-thaw products data should be used to produce the most suit- are raster maps with three discrete classes: able merged estimates requires spatially frozen, thawed, and transitional, where the explicit quantification of the random errors of transitional class can be further divided into each product (see figure 6.9 for an example). transitional (a.m. frozen, p.m. thaw) or inverse- This is where the triple collocation technique transitional (a.m. thaw, p.m. frozen) using suc- has gained popularity in the community of sat- cessive day-night  passes. The SMAP mission ellite soil moisture data users (Dorigo et  al. aims to generate freeze-thaw coverage for the 2010; Miralles et  al. 2010; Scipal et  al. 2008; globe at 3-kilometer resolution with a two-day Yilmaz and Crow 2014; Zwieback et al. 2012). repeat cycle for latitudes above 50° north In addition to measuring moisture in the (McDonald, Kimball, and Kim 2010). uppermost layers of the soil column, satellite microwave sensors are used to map the land Considerations for Use in Water Management surface freeze-thaw state. Like soil moisture, Applications satellite mapping of global freeze-thaw state is Although satellite technologies only provide soil achieved by exploiting the large difference in moisture information for the very few top centi- dielectric properties between frozen and meters of soil, the data have been shown to be thawed surfaces. The freeze-thaw state of the useful in climate studies (Jung et  al. 2010; land surface is an important link between the Taylor et al. 2012), weather forecasting (Dharssi ­ hydrologic cycle and the carbon cycle via veg- et  al. 2011), and hydrologic prediction (Brocca etation dynamics, specifically plant phenology et al. 2010, 2012; Pauwels et al. 2002), especially (Kimball, McDonald, and Zhao 2006). in combination with land surface models by data assimilation (Draper et  al. 2012; Renzullo Figure 6.9 Comparing Error Estimates for Soil Moisture Products Derived et  al. 2014). Land surface model estimates of from Active and Passive Microwave Sensors Using Triple Collocation Technique root-zone moisture are demonstrably improved a. ASCAT (active) b. AMSR-E (passive) through the assimilation of satellite soil mois- ture. The value to water resources management (WRM) is further enhanced through the esti- mated constraint that the model imparts to other components of the water cycle (evapo- transpiration and runoff ). Root-zone soil mois- ture is useful for monitoring drought and modeling landscape ecology. However, the Error in satellite soil moisture coarse spatial resolution of the data is probably (relative wetness units 0–1) an impediment to widespread adoption, espe- cially in agricultural applications. Climate stud- 0.05 0.10 0.15 0.20 ies require long time series of harmonized soil Source: Adapted from Renzullo et al. 2014. moisture values. This has recently been achieved Note: Error estimates for soil moisture products from ASCAT (active) and AMSR-E (passive) microwave sensors for Australia derived using the popular triple collocation technique. The by Liu et al. (2012), using 30 years of soil mois- technique requires three independent estimates of soil moisture to infer the errors in the respective ture data derived from several satellites. data. The third product used (not displayed) was a top-layer soil moisture estimate from the Australian Water Resources Assessment (AWRA) landscape model. White spaces in the maps The complementarity of active and passive correspond to locations where the temporal dynamics of the three soil moisture data differed significantly and the triple collocation technique did not yield an estimate. AMSR-E = Advance soil moisture retrievals has been recognized. It Microwave Scanning Radiometer–Earth Observing System; ASCAT = Advanced Scatterometer. is envisaged that future application of satellite 100  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T soil moisture studies to climate, weather, and Earth observation can also be used to clas- water resources management will be based on sify vegetation into distinct types of vegeta- products that combine the typically higher tion cover that describe various combinations accuracy but coarser resolution of passive sen- of growth form (trees, shrubs, grasses), phe- sors with the higher spatial resolution but nology (deciduous, nondeciduous), and some- noisier signal of active sensors. Indeed, this is times climate types (temperate, arid, tropical). the motivation behind the SMAP mission, Data on type of vegetation cover provide a which will be the first dedicated soil moisture convenient, but static, means of summarizing monitoring mission carrying both active and the broad roles of different functional types of passive sensing systems. vegetation. Typical types may include forest, grassland, and cropland. When nonvegetation Vegetation and Vegetation Cover classes such as water, snow, urban, and bare Definition soil are included, the categorization might Vegetation is the collective term for the cover- better be referred to as land cover types. age of plants across land areas. Vegetation One of the most important roles of vegetation attributes and processes relate to the emergent in the water cycle is to modify evaporation rates. properties and functioning of those plants In order to photosynthesize, vegetation extracts when considered at the landscape scale. Vege- soil water and groundwater and evaporates it tation can be characterized quantitatively into the atmosphere as transpiration. As a result, using measures of height, canopy and stem positive relationships exist between transpira- density, and leaf area, among others. It can also tion rates and leaf photosynthetic capacity and be described qualitatively, as vegetation classes between transpiration rates and leaf area. Tran- or cover types, such as forest, croplands, spiration rates also tend to be positively related tundra, and the like. ­ to rooting depth: the deeper the rooting system, the greater the capacity of the soil to store water Relevance for plants, which means a greater proportion of The role of vegetation in the hydrologic cycle is precipitation has the potential to become tran- well established (Budyko and Miller 1974; spiration and a smaller proportion becomes Monteith 1972; Ol’dekop 1911; Rodríguez- runoff. Vegetation also affects the energy bal- Iturbe et al. 2001; Specht 1972), modifying the ance by changing surface albedo (that is, the direct role of the climate in the partitioning of reflectivity of the land surface). Albedo alters precipitation between evaporation and runoff. the amount of sunlight absorbed at the land sur- It has been estimated that between 80 and face, which can change ET rates. 90  percent of terrestrial evaporation is trans- In regions prone to soil water deficits, pired by vegetation (Jasechko et  al. 2013). remotely sensed information may be used to Besides impoundments, vegetation is the main assess the magnitude of the impact of such def- pathway by which humans modify the terres- icits on vegetation health and be interpreted in trial water balance. Far from being a passive terms of ecosystem health, range productivity, factor, vegetation has a significant and highly and crop production, for example. This dynamic impact on its surroundings, influenc- explains the direct use of vegetation EO data in ing just about all land surface processes. The some drought monitoring systems. ubiquitous, exposed, and temporally dynamic Vegetation can also affect pollutant mobiliza- nature of vegetation means that Earth observa- tion and hence water quality, particularly in tion is a powerful and cost-effective means of relation to nutrients and sediment loads. In gen- quantifying many of the key roles that vegeta- eral, for a given intensity of precipitation event, tion plays in the hydrologic cycle. the higher the amount of bare soil, the greater chapter 6 : E arth O bservations for M onitoring W ater R esources   |  101 the rate of soil erosion will be. So the larger the estimates of albedo, fractional foliage cover, total cover—including both foliage and litter fPAR, and LAI, if total leaf area is reasonably cover—the better water quality tends to be. low. The enhanced vegetation index, which is a Since vegetation does not directly deter- MODIS-derived product (Huete et al. 2002), is mine water supply or demand (as do precipita- increasingly used. It is also highly correlated tion, radiation, and others), but instead with LAI but is less sensitive to saturation (that modifies these things, information on vegeta- is, the diminished ability to estimate LAI accu- tion is typically used as one of numerous inputs rately in high leaf areas). into water balance models to represent the Table 6.7 outlines the most prominent indirect effect of vegetation on the water cycle. sources of these core vegetation-related vari- Some of these key vegetation attributes can- ables. Current sources are listed, as are not be observed directly with remote sensors. expected future sources. Historical sources are Instead, it is common to use structural attri- also shown, as these are important for looking butes of vegetation, aggregated into vegetation at the long-term dynamics in vegetation (for cover types, as surrogates. So, for example, tall drought monitoring, for example). The infor- vegetation with high foliage cover is typically mation in table 6.7 was obtained from the Com- associated with forests. Forests usually have mittee on Earth Observation Satellites’s Earth high leaf area and deep roots and so have Observation Handbook and the World Meteo- relatively high transpiration rates. Conversely, ­ rological Organization’s Observing Systems grasslands and pastures have low vegetation, Capability Analysis and Review Tool, both of with more variable leaf area, shallower roots, which are excellent resources.8 and relatively low transpiration rates. A­ ssigning Numerous studies evaluate and compare general attributes to broad types of vegetation some of these products within and across differ- is the basis for using data on vegetation cover ent satellite sensors, globally and regionally class in most water cycle analyses. (Beck et al. 2011; Hill et al. 2006; Morisette et al. 2006; Tucker et al. 2005). Often, a product devel- Remote Sensing of Vegetation oped regionally will outperform, for that region, The most important and widely used satellite- a product derived globally. For this reason, cau- derived, vegetation-related variables are albedo, tion is sometimes warranted when using glob- LAI, and fPAR (the fraction of sunlight absorbed ally derived products in regional applications. by foliage). Of these, only albedo is more or less Other vegetation characteristics can also be directly observed by the satellite. The remaining of use to WRM applications. Leaf photosyn- attributes can only be inferred using related thetic capacity is an important predictor of metrics or EO-driven modeling. Generally crop growth. It is not generally, and certainly speaking, any sensor that measures at least not routinely, estimated using Earth observa- infrared and NIR reflectance can be used to pro- tion, as it requires hyperspectral sensors. It is, duce these products. Leaf area is almost univer- however, an emerging EO product (Houborg sally represented by the LAI (the area of leaves et al. 2013; Wu et al. 2009). per unit of ground area). It is also nearly linearly Rooting depth cannot be observed with related to fPAR and fractional foliage cover (the remote sensing. Yet there are methods for esti- fraction of ground covered by green foliage), mating rooting depth indirectly from water when leaf areas are low (that is, LAI less than 6). balance models (Ichii et al. 2006), which can be LAI and fPAR are typically derived from driven by Earth observation, often via remotely NDVI, which is calculated from infrared and sensed evapotranspiration. These methods NIR reflectance values. Methods have been typically use rooting depth as a tuning param- developed to use NDVI to produce approximate eter in either water balance or gross primary 102  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 6.7 Overview of Sensors Most Suitable for Estimating Vegetation and Land Cover DATA CURRENCY SPATIAL AND FUNCTIONAL MISSION MISSION NAME RESOLUTION REVISIT PERIOD TYPE OF SENSOR INSTRUMENTS (SHORT) (METERS) (DAYS) ACCESSIBILITY LAUNCH DATE END DATE NDVI ALBEDO fPAR LAI Archival Optical TM Landsat 5 30 16 Open July 1982 June 2013 MSS Landsat 1-3 80 18 Open July 1972 September ❶ ❶ ❶ ❶ 1983 ❷ ❷ ❷ ❷ AVHRR/2 NOAA 7-14 1,100 1 Open June 1981 Current ❶ ❷ ❶ ❹ Active microwave X-band SAR TanDEM-X 16 11 Open June 2010 X-band SAR TerraSAR-X 16 11 Open June 2007 ❷ ❹ ❷ ❹ S-band SAR HJ-1C 20 31 Open November 2012 ❷ ❹ ❷ ❹ SAR RADARSAT-2 25 24 Constrained December 2007 ❷ ❹ ❷ ❹ (RADARSAT-2) ❷ ❹ ❷ ❹ Optical MSI RapidEye 6.5 1 Open August 2008 ASTER Terra 15 16 Open December 1999 ❷ ❷ ❷ ❷ OLI Landsat-8 30 16 Open February 2013 ❶ ❶ ❶ ❶ ETM+ Landsat 7 30 16 Open April 1999 ❶ ❶ ❶ ❷ Hyperion NMP EO-1 30 16 Open November 2000 October ❶ ❶ ❶ ❷ 2014 ❶ ❶ ❶ ❶ AWiFS RESOURCESAT-2 55 26 Open April 2011 LISS-III RESOURCESAT-2 55 26 Open April 2011 ❶ ❶ ❶ ❶ (Resourcesat) ❶ ❶ ❷ ❷ MISR Terra 250 16 Open December 1999 MODIS Aqua 250 16 Open May 2002 ❷ ❷ ❷ ❷ MODIS Terra 250 16 Open December 1999 ❶ ❶ ❶ ❶ AVHRR/3 NOAA-18 1,100 1 Open May 2005 ❶ ❶ ❶ ❶ AVHRR/3 NOAA-19 1,100 1 Open February 2009 ❶ ❷ ❶ ❹ VEGETATION SPOT-5 1,150 26 Unknown May 2002 December ❶ ❷ ❶ ❹ 2014 ❶ ❷ ❶ ❷ VIIRS Suomi NPP 1,600 16 Open October 2011 (Continued) ❶ ❶ ❶ ❶ chapter 6 : E arth O bservations for M onitoring W ater R esources   |  103 Table 6.7 (Continued) DATA CURRENCY SPATIAL AND FUNCTIONAL MISSION MISSION NAME RESOLUTION REVISIT PERIOD TYPE OF SENSOR INSTRUMENTS (SHORT) (METERS) (DAYS) ACCESSIBILITY LAUNCH DATE END DATE NDVI ALBEDO fPAR LAI Future Active microwave SAR (RCM) RADARSAT C-1 50 12 Constrained 2018 SAR (RCM) RADARSAT C-2 50 12 Constrained 2018 ❷ ❹ ❷ ❹ SAR (RCM) RADARSAT C-3 50 12 Constrained 2018 ❷ ❹ ❷ ❹ Optical LISS-III RESOURCESAT- 5.8 26 Open 2015 ❷ ❹ ❷ ❹ (Resourcesat) 2A ❶ ❶ ❷ ❷ MSI Sentinel-2 A 10 10 Open 2015 (Sentinel-2) ❶ ❷ ❶ ❷ MSI Sentinel-2 B 10 10 Open 2016 (Sentinel-2) ❶ ❷ ❶ ❷ MSI Sentinel-2 C 10 10 Open 2020 (Sentinel-2) ❶ ❷ ❶ ❷ LISS-IV RESOURCESAT- 23.5 26 Open 2015 2A ❶ ❶ ❷ ❷ HISUI ALOS-3 30 60 Unknown 2016 AWiFS RESOURCESAT- 55 26 Open 2015 ❶ ❶ ❶ ❶ 2A ❶ ❶ ❶ ❶ OLCI Sentinel-3 A 300 27 Open 2015/16 OLCI Sentinel-3 B 300 27 Open 2017 ❶ ❷ ❷ ❷ 104  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T OLCI Sentinel-3 C 300 27 Open 2020 ❶ ❷ ❷ ❷ VIIRS JPSS-1 1,600 16 Open 2017 ❶ ❷ ❷ ❷ Sources: Committee on Earth Observation Satellites (CEOS) Earth Observation Handbook (http://www.eohandbook.com/) and the WMO Observing Systems Capability Analysis and Review Tool (http://www.wmo-sat.info/ ❶ ❶ ❶ ❶ oscar/). Note: NDVI = normalized difference vegetation index; fPAR = fraction of absorbed photosynthetically active radiation; LAI = leaf area index. The suitability of each sensor to provide useful data is shown with numbers and colors, as follows: ❶ highly suitable, ❷ suitable, ❹ not suitable. productivity models, when all other parame- of the main satellite programs used for vegeta- ters are known with reasonable certainty. tion sensing provide these vegetation or land Vegetation litter (that is, dead vegetation) cover maps as precalculated information prod- cover can now be detected fairly routinely ucts available for off-the-shelf use (table 6.8). from most of the main land-observing satellite Among these, only the MODIS Land Cover Type platforms that have at least some SWIR capacity. product provides dynamic annual mapping. It is, however, best detected from hyperspectral Vegetation height can be estimated using sensors (Guerschman et  al. 2009a). Observa- either optical photogrammetry, satellite-borne tions of litter cover also enable estimates to be LiDAR (Simard et al. 2011), or synthetic aper- made of the fraction of bare soil (that is, the ture radar (SAR; Kellndorfer et al. 2004; Weg- ground area not covered in vegetable matter). muller and Werner 1997). Vegetation biomass is the mass of live plant Vegetation Cover Classes tissue. Aboveground biomass is usually esti- Vegetation cover classes are identified using mated from optical Earth observation, radar, combinations of remotely sensed variables, often LiDAR, or a combination of these, using empir- in conjunction with ancillary data such as cli- ical conversion functions (Goetz et  al. 2009; mate and land use maps and field observations. Lucas et al. 2010). More recently, aboveground Remotely sensed estimates of vegetation height biomass has also been estimated using the veg- from radar or Laser Imaging, Detection, and etation optical depth index, which is a relative Ranging (LiDAR) and biomass, likewise from measure of aboveground vegetation water con- radar or LiDAR, are useful for distinguishing tent derived from passive microwave remote between structurally distinct types of vegetation. sensing (Andela et al. 2013). The main EO method for deriving classes of vegetation cover, however, is by classifying the Example Applications temporal (seasonal) dynamics in leaf area. For Earth observation of vegetation has an impor- example, when LAI has high seasonal variability, tant role to play in providing information for it is reasonable to assume that this is caused by meteorological and agricultural drought short-lived vegetation that dies back (including monitoring systems, by focusing directly on crops) or by deciduous woody vegetation. Many the impact of drought on vegetation using Table 6.8 Examples of Global Vegetation Cover Maps SPATIAL NAME SENSOR AGENCY COVER CLASSES RESOLUTION CURRENCY SOURCE UMD Land Cover AVHRR University of 14 1°, 8 kilometers, 1998 Hansen et al. 2000 Classification Maryland 1 kilometer MODIS Land MODIS Terra and U.S. Geological 17 500 meters Yearly, 2001–12 Friedl et al. 2010 Cover Type Aqua Survey (MCD12Q1) ESA-GlobCover MERIS European Space 22 300 meters 2009 Agency and Catholic University of Louvain FROM-GLC Landsat TM and Tsinghua 11 30 meters Gong et al. 2012 ETM University Source: CEOS 2015, OSCAR database. Note: AVHRR = Advanced Very High Resolution Radiometer; ESA = European Space Agency; ETM = Enhanced Thematic Mapper; FROM-GLC = Finer Resolution Observation and Monitoring of Global Land Cover; MERIS = Medium Resolution Imaging Spectrometer; MODIS = Moderate Resolution Imaging Spectrometer; TM = Thematic Mapper; UMD = University of Maryland. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  105 time-series analysis of information on vegeta- normalized difference water index, derived tion “greenness,” where the current anomaly from MODIS (figure 6.10)10 in greenness is compared with the long-term • The Australian Bureau of Meteorology’s mean value (or an alternative reference value, Climate Maps, which provide AVHRR such as the same time last year). Typically, NDVI maps11 vegetation drought monitoring approaches are combined with other approaches (for an • Princeton University’s African Drought example, see Mu et  al. 2012). Such analyses Monitor, which uses MODIS NDVI12 useful for monitoring and predicting food are ­ • The University of Montana’s global ter- shortages and for targeting investments in restrial drought severity index, which uses agricultural infrastructure. The following are the MODIS evapotranspiration and NDVI some examples of current drought monitor- products.13 ing systems that include vegetation Earth ­observation: Actual evaporation rates can be estimated using the Penman-Monteith model (Monteith • The U.S. government’s Global Drought 1981). As discussed in the section on evapo- Information System, which uses AVHRR transpiration, this requires estimating a “sur- NDVI, among several other satellite- face conductance” parameter, which is related derived vegetation indexes9 primarily to vegetation characteristics. Spa- • The European Commission’s European tially explicit, temporally varying estimates of Drought Observatory, which uses fPAR, evapotranspiration can be made across large derived from MERIS (Medium Resolu- areas when the surface conductance parame- tion Imaging Spectrometer), and the ter is driven by using RS information to assign “typical” values to land cover classes or by Figure 6.10 Combined Drought Indicator for Europe, Mid-March 2014 using LAI or NDVI directly (Leuning et  al. 2008; Zhang et  al. 2010). Yebra et  al. (2013) review the suitability of alternative MODIS vegetation EO data for estimating evapotrans- piration and find that NDVI and EVI produce the best predictions of canopy conductance. The majority of spatially explicit hydrologic models need to incorporate the role of vegeta- tion in the water cycle and do so by including some vegetation-specific parameters. As with the estimation of evapotranspiration, the infor- mation used may be fully dynamic and quanti- tative (continuously varying fields of LAI, rooting depths, and so forth). More typically, though, vegetation is described as types of veg- etation cover (with static boundaries), assign- © European Drought Observatory (EDO) 2013 ing vegetation-specific characteristics to each Source: © European Drought Observatory. Used with permission. Further permission required for type of cover. This does not require satellite reuse. observations for the period of analysis, which Note: The combined drought indicator is based on anomalies in the standard precipitation index, modeled soil moisture, and remotely sensed fPAR. Yellow is “watch,” where the index is anomalously has obvious advantages for predicting presatel- low; orange is “warning,” where low rainfall translates into a soil moisture anomaly; red is “alert,” lite or future conditions. These characteristics when these two conditions are accompanied by an fPAR anomaly. fPAR = fraction of absorbed photosynthetically active radiation. can be scalars (that is, static) or variables (that 106  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T is, dynamic, for example, a prescribed seasonal combined with crop phenological signatures pattern), with remotely sensed data being a (that is, timing and rates of green-up and senes- prime source of information for the latter. cence) within the greenness signal to identify The AWRA system (van Dijk 2010; van Dijk the area of land that has been irrigated within a and Renzullo 2011) is used by the Australian given region (Conrad et al. 2011; Ozdogan et al. Bureau of Meteorology for water resources 2010; Pervez and Brown 2010 for examples and assessment and accounting and is one example more information). ­ Figure 6.12 shows a map of of a spatial water balance estimation model an irrigated area in the Indian Krishna basin, that uses remotely sensed vegetation informa- derived from MODIS imagery. tion. The landscape is divided into “tree” and “herbaceous” cover types (derived from Groundwater AVHRR imagery), and each is assigned a fixed Definition rooting depth but spatially and seasonally Groundwater is the water contained in the varying LAI (derived from MODIS imagery). saturated zone—the subsurface volume Figure 6.11 shows an example AWRA output. below the water table—where water fills the Techniques exist to use remotely sensed cracks and pores of rock, sediment, and soil. vegetation information for assessing the area of Groundwater can be recharged by rainfall, irrigated agriculture. Such mapping is impor- snowmelt, irrigation, and rivers. It dis- tant, as there can be a significant difference charges when water resurfaces through between the “irrigable area” (that is, the area springs and wells, flows into lakes, streams, that is equipped with infrastructure for irriga- and the ocean, or is extracted by vegetation. tion) and the area actually irrigated at any given Groundwater moves at varying speeds, time. Techniques are based on the concept depending on the storage pressure and the that, in semiarid and arid environments at porosity of the storage medium, among other least, regional time-series analysis of greenness things. Aquifers are subsurface layers where (LAI, fPAR, or NDVI) can identify areas that are unusually green and contrast with the sur- Figure 6.12 Map of Irrigated Land Cover Types in the Krishna Basin, India rounding landscape. This information may be 75°0'0"E 78°0'0"E 81°0'0"E Figure 6.11 Map of AWRA-Derived Total Annual Landscape Water Yields in 2011–12 for Tasmania, 18 °0'0"E 18 °0'0"E Australia Annual landscape water yield (millimeters) 3200 2400 1800 1200 900 Legend 600 Class1: Water bodies 15°0'0"E 15°0'0"E 400 Class2: Shrublands mix with rangelands Class3: Rangelands mix with rain-fed 300 Class4: Rain-fed agriculture 200 Class5: Rain-fed + groundwater 100 Class6: Minor irrigated (light/tank) Class7: Irrigated + surface water + HOBART 50 groundwater – continuous crops 0 Class8: Irrigated – surface water – High data Scale double crop 0 25 50 100 km Class9: Forests uncertainty 75°0'0"E 78°0'0"E 81°0'0"E Source: Bureau of Meterology (BoM) 2012. © BoM. Used with Source: Gumma, Thenkabail, and Nelson 2011. permission. Further permission required for reuse. Note: Irrigated land cover types are shown in green and red. This map is derived from Moderate Note: AWRA = Australian Water Resources Assessment. Resolution Imaging Spectrometer (MODIS) imagery 2000–01. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  107 groundwater is confined, sometimes under When groundwater reaches the surface in an pressure, by adjacent rock and clay layers of otherwise dry landscape, the additional water low permeability. supply can be detected through its enhance- ment of surface evaporation rates and vegeta- Relevance tion productivity. Groundwater-dependent Groundwater is a critical source of water for ecosystems are tied to this process. Areas where human consumption and agriculture, especially evaporation or vegetation productivity or cover where surface water is scarce or polluted. It also are higher than what would be expected for the moderates streamflow, producing the longer- given precipitation can be detected using term baseflow component of total flows, which remotely sensed evaporation or vegetation decouples flows somewhat from the variability cover. Details on how these two attributes can inherent in the climatic drivers of streamflow. be remotely sensed may be found in the section The two greatest risks to groundwater supplies on evapotranspiration and ground cover, and an are overextraction and pollution. example of such an analysis is provided in Approximately 43 percent of all irrigated chapter 7. An excellent resource on the use of ­ agriculture depends on groundwater (Siebert Earth observation in groundwater applications et  al. 2010), and this proportion is rising Meijerink et al. (2007). is ­ rapidly—often resulting in unsustainable ­ Satellite gravimetry—satellites that measure extraction rates (Gleeson et  al. 2012; Wada gravity fields—is able to detect changes in these et al. 2010; Wada, van Beek, and Bierkens 2012). fields between subsequent overpasses. With Groundwater is also a major source of drinking suitable postprocessing to remove the effects water, with, for example, 51 percent of the U.S. of phenomena such as tides and tectonic move- population relying on groundwater.14 Ground- ments, postglacial rebound, atmospheric com- water is the primary source of water in the position, and changes in the mass or other Middle East and Northern Africa (where fossil features of surface water, gravimetric observa- groundwater reserves are used) and in coun- tions can provide information on changes in tries such as Denmark, Jamaica, Portugal, and subsurface water mass. Currently there is one Slovakia with high-yielding sources, often in set of gravity measurement satellites—the limestone or “karst” aquifers (FAO 2013). GRACE mission by NASA and the German Aerospace Centre, which was launched in Remote Sensing of Groundwater 2002 (Tapley et  al. 2004). This satellite mis- As groundwater lies below the land surface, sion is currently operating seven years beyond there are currently no EO techniques for direct its intended five-year lifetime, and the quality observation of groundwater level. The main of its observations is slowly degrading. A indirect techniques are through satellite grav- follow-on GRACE mission is planned for 2017. ity field mapping (gravimetry) and radar inter- GRACE’s coarse spatial resolution (about ferometry (table 6.9). The former measures 400 kilometers) means that it can be used only changes in the regional gravity field, while the for large, basin-scale applications. Nonethe- latter measures changes in land surface eleva- less, its unique ability to monitor integrated tion. Both techniques assume that there is a changes in water storage everywhere makes it relationship between changes in gravity fields a valuable sensor. Several reviews have been or surface elevations, respectively, and changes conducted of the application of GRACE to in groundwater storage. assess water storage (Güntner 2008; Ramillien, In water-limited landscapes, changes in shal- Famiglietti, and Wahr 2008; Syed et al. 2008) low groundwater may also be inferred by exam- and groundwater depletion (Leblanc et  al. ining the effects on other surface processes. 2009; Rodell, Velicogna, and Famiglietti 2009). 108  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 6.9 Overview of Sensors Most Suitable for Estimating Groundwater DATA CURRENCY AND SPATIAL REVISIT FUNCTIONAL TYPE OF RESOLUTION PERIOD LAUNCH SURFACE SENSOR SENSOR PLATFORM (METERS) (DAYS) ACCESSIBILITY DATE GRAVITY FIELD HEIGHT Current Active microwave X-band SAR TanDEM-X 1–16 11 Open June 2010 X-band SAR TerraSAR-X 1–16 11 Open June 2007 ❹ ❷ SAR 2000 COSMO-SkyMed 1 1–100 16 Constrained June 2007 ❹ ❷ December ❹ ❷ SAR 2000 COSMO-SkyMed 2 1–100 16 Constrained 2007 ❹ ❷ October SAR 2000 COSMO-SkyMed 3 1–100 16 Constrained 2008 ❹ ❷ November SAR 2000 COSMO-SkyMed 4 1–100 16 Constrained 2010 ❹ ❷ December SAR (RADARSAT-2) RADARSAT-2 8–172 24 Constrained 2007 ❹ ❷ PALSAR-2 (ALOS-2) ALOS-2 10 14 Constrained March 2014 Gravimetric GRACE GRACE 400,000 1 Open March 2002 ❹ ❷ Future ❶ ❹ Active microwave C-band SAR Sentinel-1 B 10–50 12 Open 2015 C-band SAR Sentinel-1 C 11–50 12 Open 2019 ❹ ❷ SAR (RCM) RADARSAT C-1 8–100 Constrained 2018 ❹ ❷ SAR (RCM) RADARSAT C-2 8–100 24 Constrained 2018 ❹ ❷ SAR (RCM) RADARSAT C-3 8–100 Constrained 2018 ❹ ❷ Gravimetric GRACE II GRACE FO 400,000 1 Open 2017 ❹ ❷ ❶ ❹ Note: The suitability of each sensor to provide useful data is shown with numbers and colors, as follows: ❶ highly suitable, ❷ suitable, ❹ not suitable. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  109 Recently, GRACE observations were used Second, the use of RS observations to provide along with ocean, lake, and river water level management-relevant groundwater information altimetry to constrain fully spatial estimates of is still in the development stage—it has not yet the water balance for the globe at 100-kilometer reached a stage of maturity, where the data prod- resolution (van Dijk et al. 2013). ucts are generated routinely or operationally or Satellite radar interferometry allows mea- where the product has been widely tested and surements of very small changes in soil surface accepted by the scientific and practitioner elevation that can help to detect changes in communities. groundwater storage. The technique relies on Lastly, while the GRACE-derived gravimet- the change in the distance between the satel- ric data provide new and potentially valuable lite and a given location on the Earth’s surface insights into groundwater-related processes, the between successive satellite overpasses. These data are of exceptionally coarse resolution and changes can be measured very accurately (to generally restricted to large, basin-scale applica- less than 1 centimeter) using SAR instruments. tions of no finer than 300-kilometer resolution. Becker (2006) and Galloway and Hoffmann (2007) provide reviews of different applica- Surface Water tions of interferometry to groundwater charac- Definition terization and monitoring and demonstrate Because of its particular relevance for water interferometry’s utility for supporting ground- assessment and water cycle studies, surface water management directly or by improving water is treated separately from other types of the ability to model groundwater. SAR can be land cover. Water bodies can vary greatly in used to monitor seasonal and long-term size and duration. This section focuses on the changes in groundwater storage, provided the following: relationship between vertical surface move- • Natural or man-made reservoirs, which ment and groundwater storage, known as can range from water bodies such as small Terzhagi’s Principle, can be quantified. Changes ponds of a few square meters to large lakes in vertical surface movement must be inter- of several thousands of square kilometers. preted carefully and regionally, as there may Generally, these water bodies change in be many context-specific considerations— area and volume relatively slowly in time such as whether the local geological materials (that is, in a matter of weeks to months). deform at all in response to changes in groundwater mass and whether there is much • Surface water due to flooding, which can interference from local vegetation. range from small overbank floods near water streams to very large floods covering Limitations hundreds of square kilometers. In general By its nature, groundwater cannot be observed terms, floods are more dynamic in time directly using Earth observation, so the use of than reservoirs and can change in area and remote sensing here is inferential and has limi- volume in a matter of hours or days. tations. First, RS observations can provide information on groundwater levels or recharge Relevance and discharge rates only when combined with Monitoring surface water areas is relevant to other sources of observations—both RS and applications linked to agriculture, urban field data. This entails the use of models, where water use, and flood mitigation. Surface the actual data product (levels or recharge and water may be used for irrigation as well as discharge rates) is a modeled variable derived for human or animal consumption in both from an array of inputs. rural and urban areas. Many ecosystems, 110  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T such as wetlands, depend on regular flood- discrimination of water from soil and  vegeta- ing, and their health can be compromised if tion more problematic (although these same too much surface water is diverted to other properties can be exploited for deriving water uses. quality parameters, as discussed in the section Earth observation can be used for estimat- on optical water quality). In the SWIR region ing the area of such reservoirs and floods. The (about 900–2,500 nanometers), water quality reservoir size that satellite sensors can mea- does not interfere, and any water body will sure depends on the spatial resolution and the reflect very low amounts of radiation. area-to-perimeter ratio of the reservoir. Esti- One major disadvantage of using optical mating the total volume of water available in imagery is that the images are subject to cloud reservoirs requires making assumptions about contamination. This is particularly problem- the depth of such areas, based on local bathym- atic in tropical regions during the monsoon etry measurements, although in some cases season, when cloud-free imagery may be rare. radar or LiDAR altimetry can also be used to In addition, optical sensors are poorly suited to track water levels. detecting water under dense canopies, such as Monitoring flooding events with remote in the Amazon or Congo basins, where much sensing involves the same physical principles of the floodplains may be located in inundated as monitoring water in reservoirs. However, forests (Mayaux et al. 2002; Mertes et al. 1995). the main difference is that, generally, flood Many algorithms exist for mapping surface events are more dynamic in time and therefore water areas. These include the use of simple require the use of imagery acquired with high threshold values in a spectral band (Overton temporal repetition and, even more critical, 2005; Powell, Letcher, and Croke 2008), com- not prone to cloud obscuring. binations of two bands such as the normalized Measuring surface water elevation can pro- difference water or vegetation indexes (Brak- vide estimates of changes in the total volume of enridge and Anderson 2006; Sakamoto et  al. water in reservoirs and wetlands and river dis- 2007), and, in some cases, ancillary variables to charge, although this is currently only possible in improve the detection of water in the presence wide rivers (that is, several hundreds of meters). of topographic shading effects (Guerschman et al. 2011; Ordoyne and Friedl 2008). Theoretical Basis for Remote Sensing Radar and passive microwave imagery are of Surface Water: Estimating Area and (by very good approximation) not affected by Water Level clouds or water vapor and therefore can pro- There are two principal ways to estimate area: vide useful information on surface water under optical imaging and radar and passive micro- clouds. In addition, radar is better suited than wave imaging. optical sensors for detecting water under The main optical characteristic of water is dense canopies (Rosenqvist et al. 2002). that it absorbs most of the incoming solar radia- The backscatter coefficient15 of smooth open tion in the visible and infrared regions and water bodies is low, which allows discrimina- therefore reflects less radiation than other tion of water from land using radar. However, landscapes. This characteristic has been SAR is susceptible to wind-induced waves, exploited since the mid-1970s (Rango and which increase scattering back to the sensor, Anderson 1974; Rango and Salomonson 1974). creating difficulties for detecting surface water In the VIS-NIR wavelengths (about 400–900 (Smith and Alsdorf 1998). Complications also nanometers), sediments, chlorophyll, and other arise when there is vegetation above the water elements affecting water quality can modify the surface. This dramatically increases backscatter spectral signal and, in some cases, make the and can create uncertainties in automated chapter 6 : E arth O bservations for M onitoring W ater R esources   |  111 ­ apping. Interferometric coherence from mul- m and driven by a series of constraints. For exam- titemporal observations is another alternative ple, for measuring water in reservoirs, spatial to delineate surface water accurately, for exam- resolution is usually the most important factor, ple, from the ERS-1, ERS-2 tandem mission. so high- or very high-resolution sensors— Water bodies have significantly lower bright- either optical or radar—are more suitable than ness temperatures than their surroundings, and medium-resolution sensors. As mentioned, the emissivity polarization difference is gener- optical sensors are subject to clouds, but cur- ally large, which makes it feasible to detect rent high-resolution radar needs to be tasked. water bodies with microwave measurements For water in floods (generally larger in area (De Jeu 2003). Possibly the greatest disadvan- than reservoirs and more rapidly changing tage of passive microwave sensors is their over time), timely acquisition is more valuable, coarse resolution, which has hampered their so medium-spatial-resolution optical and radar adoption for environmental monitoring. Single sensors may be considered more suitable. “pixel” calibration against field discharge mea- Most optical sensors can be used to map surements has been used for monitoring dis- surface water. Sensors that include bands in charge during (relatively large) flood events, the SWIR are the best suited to the task, as however. water bodies unambiguously absorb most of Several types of active sensors, including the radiation on those wavelengths. Examples laser, profiling radar, interferometric SAR, and of such sensors include MODIS, Landsat, and swath radar, are able to characterize water VIIRS. Optical sensors with bands only in the levels. Laser systems emit a pulse of light (nor- ­ VIS-NIR can also be used to map surface water, mally VIS or NIR) and measure the time that although they are less suitable for differentiat- the echo takes to return to the sensor. Radar ing water from other types of land cover such altimeters work on a similar principle. Inter- as wet or dry soil, particularly when the water ferometric SAR uses multiple images to esti- contains many suspended sediments or chlo- mate changes in elevation (and in terrain). rophyll. Examples of such sensors include These techniques have been used to measure AVHRR, QuickBird, and IKONOS. ocean levels since the early 1990s. Over land, All of these sensors generally have a trade- the accuracy of the level measurements off between spatial and temporal resolution. depends on the size of the water bodies being Landsat and MODIS, for example, have a simi- measured; over rivers, surfaces are about lar spectral ability to identify surface water, but 10 centimeters at best and more typically about while Landsat is able to do so at high spatial 50 centimeters. With increased averaging over resolution (30-meter pixels), MODIS does so at large lakes (more than 100 square kilometers), medium resolution (250- or 500-meter pixels, accuracy improves to 3–4 centimeters (Alsdorf, depending on the bands used; see ­ figure 6.13). Rodríguez, and Lettenmaier 2007). This makes At the same time, MODIS can capture about satellite altimetry suitable for monitoring large two images of the same area per day (depend- water bodies, particularly in remote areas ing on the latitude) from the Terra and Aqua where field-based gauging is not available. satellites, whereas Landsat revisits each site every 16 days. These differences need to be Sensors for Surface Water Remote Sensing considered when assessing the ability of each Table 6.10 provides a list of existing and sensor to monitor surface water area. planned sensors that can produce data for esti- The spatial resolution and cloud- mating the area and height of surface water. penetrating ability of radar makes it particu- The classification of the suitability of each sen- larly useful for mapping the extent of surface sor for mapping water area and water height in water during flood events. However, such reservoirs and floods is somewhat subjective acquisitions need to be tasked, making 112  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 6.10 Overview of Sensors Most Suitable for Mapping Surface Water Extent and Height DATA CURRENCY SPATIAL REVISIT SURFACE SURFACE AND FUNCTIONAL MISSION MISSION NAME RESOLUTION PERIOD LAUNCH WATER IN WATER IN WATER TYPE OF SENSOR INSTRUMENTS (SHORT) (METERS) (DAYS) ACCESSIBILITY DATE END DATE RESERVOIRS FLOODS LEVEL Archival Optical TM Landsat 5 30 16 Open July 1982 June 2013 MSS Landsat 1-3 80 18 Open July 1972 September ❶ ❷ ❹ 1983 ❷ ❷ ❹ AVHRR/2 NOAA 7-14 1,100 1 Open June 1981 Active microwave SAR (RADARSAT-1) RADARSAT-1 25 24 Constrained November March 2013 ❸ ❸ ❹ 1995 ❷ ❷ ❹ PALSAR ALOS 7–44 14 Constrained January 2006 September 2012 ❶ ❷ ❹ ASAR (stripmap) Envisat 30 Constrained March 2002 April 2012 ASAR (wide swath) Envisat 150 Open March 2002 April 2012 ❶ ❷ ❹ AMI-SAR ERS-1 30 35 Constrained July 1991 March 2000 ❷ ❶ ❹ AMI-SAR ERS-2 30 35 Constrained April 1995 July 2011 ❶ ❷ ❹ Radar altimetry NRA  TOPEX/Poseidon 10 Open January 1992 January ❶ ❷ ❹ 2005 ❹ ❹ ❶ GFO-RA GFO 17 Open February 1998 October 2008 ❹ ❹ ❶ Poseidon-2 JASON-1 10 Open December July 2013 2001 ❹ ❹ ❶ Current Optical OSA IKONOS 3.3 Constrained September 1999 ❷ ❸ ❹ GIS GeoEye 1.6 Constrained September 2008 ❷ ❸ ❹ MSI RapidEye 6.5 1 Open August 2008 ASTER Terra 15 16 Open December ❷ ❸ ❹ 1999 ❶ ❷ ❹ OLI Landsat-8 30 16 Open February 2013 ETM+ Landsat 7 30 16 Open April 1999 ❶ ❷ ❹ MODIS Aqua 250 16 Open May 2002 ❶ ❷ ❹ MODIS Terra 250 16 Open December ❷ ❶ ❹ chapter 6 : E arth O bservations for M onitoring W ater R esources   |  113 1999 ❷ ❶ ❹ (Continued) Table 6.10 (Continued) DATA CURRENCY SPATIAL REVISIT SURFACE SURFACE AND FUNCTIONAL MISSION MISSION NAME RESOLUTION PERIOD LAUNCH WATER IN WATER IN WATER TYPE OF SENSOR INSTRUMENTS (SHORT) (METERS) (DAYS) ACCESSIBILITY DATE END DATE RESERVOIRS FLOODS LEVEL Optical, cont. AVHRR/3 NOAA-18 1,100 1 Open May 2005 AVHRR/3 NOAA-19 1,100 1 Open February 2009 ❸ ❷ ❹ VEGETATION SPOT-5 1,150 26 Unknown May 2002 ❸ ❷ ❹ VIIRS Suomi NPP 1,600 16 Open October 2011 ❶ ❷ ❹ Active microwave SAR (RADARSAT-2) RADARSAT-2 25 24 Constrained December ❷ ❷ ❹ 2007 ❶ ❷ PALSAR-2 (stripmap) ALOS-2 10 14 Constrained June 2014 PALSAR-2 (ScanSAR) ALOS-2 100 14 Open June 2014 ❶ ❷ ❹ C-SAR (stripmap) Sentinel-1A 4x5 Constrained April 2014 ❷ ❷ ❹ C-SAR (IW) Sentinel-1A 5x20 Constrained April 2014 ❶ ❷ ❹ Radar altimetry Poseidon-3 JASON-2 Open June 2008 ❶ ❶ ❹ Future ❹ ❹ ❶ Optical MSI (Sentinel-2) Sentinel-2 A 10 10 Open 2015 MSI (Sentinel-2) Sentinel-2 B 10 10 Open 2016 ❶ ❷ ❹ MSI (Sentinel-2) Sentinel-2 C 10 10 Open 2020 ❶ ❷ ❹ Radar altimetry KaRIN SWOT Open 2020 ❶ ❷ ❹ Poseidon-3B JASON-3 Open 2015 ❹ ❹ ❶ 114  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T ❹ ❹ ❶ and pixel size indicate the size of the water bodies potentially detectable and the temporal repetition. Note: The ability of each sensor to detect surface water in reservoirs or floods and water height is shown with colors and numbers as ❶ highly suitable, ❷ suitable, ❸ potentially suitable, and ❹ not suitable. The revisit cycle Figure 6.13 Example of Satellite Imagery Captured can only be measured when and where it is during Flood Event in Northern New South Wales, sufficiently covered by the tracks. Australia A most promising future development is the a. Landsat imagery b. MODIS imagery Surface Water and Ocean Topography Mis- sion, which is scheduled for launch in 2020. It will include a radar altimeter, an interferome- ter (Ka-Band Radar Interferometer [KaRIN]), and a microwave radiometer (Rodríguez and Estéban-Fernández 2010). Applications This section highlights some notable examples of research applications. Hess (2003) uses Japa- nese Earth Resources Satellite 1 SAR data to Source: Adapted from Guerschman et al. 2011; WIRADA 2012. © CSIRO. map inundation in the Amazon basin during the Used with permission. Further permission required for reuse. © high- and low-water seasons at 100-meter reso- WIRADA. Used with permission. Further permission required for reuse. Note: On the left, Landsat imagery and, on the right, Moderate lution. Papa et  al. (2010) combine passive Resolution Imaging Spectrometer (MODIS) imagery. Top-row figures (SSM/I) and active (ERS) microwave with opti- show the surface reflectance in false color; bottom-row figures show an object-oriented classification of surface water in red (bottom left) cal (AVHRR) imagery to describe the global pat- and the open water likelihood index as an estimate of the fraction of terns of surface water extent from 1993 to 2004. the pixel covered with water (bottom right). Papa et al. (2010) develop a technique to com- existing radar sensors impractical for routine bine these disparate data sources, which over- global monitoring at high resolution (less lap only partially in time, to intercalibrate the than 100 meters). So far, the highest resolu- surface water estimates. They report a slight tion available for radar in routine mode for decrease in the global inundated area for the part of the globe is about 1 kilometer (ASAR period analyzed, mainly in the tropics. global monitoring [GM]). The C-band Syn- In an example application, CSIRO com- thetic Aperture Radar (C-SAR) instrument on bined Landsat- and MODIS-based mapping board the Recently launched Sentinel-1 mis- with flow measurements to produce the sion is intended to improve that to global cov- Murray-Darling basin floodplain inundation erage at 5-by-20-meter resolution in the model (Overton et  al. 2011). This model pro- ScanSAR-Interferometric wide-swath mode. vides a regional-scale model of the spatial Lake and reservoir altimetry is normally extent of floodplain inundation under ecologi- obtained using the radar altimetry instru- cally significant flood return periods. The ments on the Jason (1 and 2), TOPEX (Ocean model was developed using the flow scenarios Topography Experiment)/Poseidon, and modeled under the Murray-Darling Basin Sus- ERS-2 satellites. The Ice, Clouds, and Eleva- tainable Yields Project (see appendix B), allow- tion Satellite (ICESat) provided useful LiDAR ing mapping under different climate and measurements with accuracies of 3 centime- development scenarios. ters over footprints of 70 meters, but an instru- Several data services provide flooding or ment failed after launch, and the satellite is lake and reservoir levels in near real time, inactive. An ICESat-2 is planned for launch in including the following: 2016. A general disadvantage of altimetry is • The Dartmouth Flood Observatory16 that it does not provide full coverage but instead measures along the orbit track, which • The Near-Real-Time Global MODIS Flood means that the height of surface water level Mapping 17 chapter 6 : E arth O bservations for M onitoring W ater R esources   |  115 • Crop Explorer18 also a source of floodwater. Apart from the natu- ral, seasonal cycle of flooding associated with • LEGOS HydroWeb19 spring snowmelt, snow can also be a source of • The European Space Agency’s River & severe flooding when unusually warm condi- Lake.20 tions occur prior to or during the normal spring melt. Both the temperature and liquid water Some of these examples are discussed fur- content of snow (snow wetness) are good indi- ther in appendix B. cators of how close the snowpack is to melting and therefore are important for forecasting Snow floods, along with measures of snow extent and Definition depth or snow water equivalent. Snow cover exists where the accumulation of snow is sufficient for the land surface to have a Remote Sensing of Snow reasonably continuous layer of snow. The accu- In the visible wavelengths, snow is generally mulation and melting of snow cover provide an highly reflective (that is, has a high albedo), important supply of freshwater across many which makes it relatively easy to detect, as it mountainous and high-latitude (mainly north- contrasts with the surrounding landscape. In ern hemisphere) regions. Outside of the areas thermal infrared wavelengths, it also possesses permanently covered with snow, snow cover easily recognizable features, often being colder provides a supply of water only in the spring and than its surroundings and therefore emitting summer. Important attributes of snow cover are less radiation in these wavelengths. Snow its areal extent, thickness, and water content. cover also affects the microwave radiation The electromagnetic and structural properties emitted from the Earth. It alters the attenua- of snow are of particular importance in this con- tion of microwaves, and analysis of the attenu- text as their unique characteristics allow for ation patterns can reveal important details remote sensing of their extent, depth, and mass. about the depth, composition (that is, the solid Since the same mass of water can take up dif- and liquid fractions), and structure of the snow ferent volumes when frozen (depending on its pack. This attenuation can be measured using structure as snow or ice), it is useful to describe passive microwave sensors, which detect snow in terms of snow water equivalent—the microwaves emitted from the Earth’s surface mass or depth of water obtained when a certain after passing through the overlying snow pack, volume of snow is melted. While glaciers and or using active microwave sensors (typically other terrestrial ice bodies provide important SAR), which detect the backscatter of sensor- sources of freshwater, the remote sensing of ice emitted microwave radiation. bodies is not covered in this review. For an The reliability of remotely sensed, snow- excellent overview of the remote sensing of related information decreases as cloud cover, both snow cover and glaciers, see Rees (2006). tree or forest cover, and terrain complexity increase. Low sun illumination angles, typical of Relevance the higher northern latitudes, reduce the qual- As snow contains freshwater, meltwater from ity of remotely sensed information. Frei et  al. snow cover provides an important source of (2012) provide a useful overview of remotely water for consumption, irrigation, and power sensed snow products. Table 6.11 outlines the generation in many parts of the globe. As a most prominent remote sensors relevant to source of water, it is highly seasonal, so surface measuring snow (cover) extent, snow moisture, impoundments are often constructed to capture and snow water equivalent. Both current and and store meltwater across seasons. Snow is expected future sources are listed. The 116  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 6.11 Overview of Sensors Most Suitable for Mapping Snow Extent, Snow Moisture, and Snow Water Equivalent DATA CURRENCY SPATIAL SNOW AND FUNCTIONAL RESOLUTION REVISIT PERIOD ACCESSIBILITY WATER SNOW TYPE OF SENSOR SENSOR PLATFORM (METERS) (DAYS) AND COST LAUNCH DATE END DATE SNOW EXTENT EQUIVALENT MOISTURE Current Active microwave SAR 2000 COSMO-SkyMed 1–100 16 Constrained June 2007 June 2014 1 ❷ ❶ ❷ SAR 2000 COSMO-SkyMed 1–100 16 Constrained December December 2 2007 2014 ❷ ❶ ❷ SAR 2000 COSMO-SkyMed 1–100 16 Constrained October 2008 October 3 2015 ❷ ❶ ❷ SAR 2000 COSMO-SkyMed 1–100 16 Constrained November 4 2010 2017 November ❷ ❶ ❷ X-band SAR TanDEM-X 1–16 11 Open June 2010 December 2015 ❶ ❶ ❶ X-band SAR TerraSAR-X 1–16 11 Open June 2007 December 2015 ❶ ❶ ❶ SAR RADARSAT-2 8–172 24 Constrained December April 2015 (RADARSAT-2) 2007 ❷ ❷ ❷ Hyperspectral Hyperion NMP EO-1 30 16 Open November October 2000 2014 ❶ ❹ ❹ Optical LISS-IV RESOURCESAT-2 5.8 26 Constrained April 2011 April 2016 MSI RapidEye 6.5 1 Open August 2008 August ❷ ❹ ❹ 2019 ❷ ❹ ❹ ASTER Terra 15 16 Open December October 1999 2015 ❶ ❹ ❹ LISS-III RESOURCESAT-2 23.5 26 Open April 2011 April 2016 LISS-III RESOURCESAT- 23.5 26 Open October 2015 October ❷ ❹ ❹ 2A 2020 ❷ ❹ ❹ ALI NMP EO-1 30 16 Open November October 2000 2014 ❷ ❹ ❹ ETM+ Landsat 7 30 16 Open April 2099 January 2017 ❷ ❹ ❹ OLI Landsat 8 30 16 Open February 2013 May 2023 AWiFS RESOURCESAT-2 55 26 Open April 2011 April 2016 ❷ ❹ ❹ AWiFS RESOURCESAT- 55 26 Open October 2015 October ❷ ❹ ❹ 2A 2020 ❷ ❹ ❹ chapter 6 : E arth O bservations for M onitoring W ater R esources   |  117 (Continued) Table 6.11 (Continued) DATA CURRENCY SPATIAL SNOW AND FUNCTIONAL RESOLUTION REVISIT PERIOD ACCESSIBILITY WATER SNOW TYPE OF SENSOR SENSOR PLATFORM (METERS) (DAYS) AND COST LAUNCH DATE END DATE SNOW EXTENT EQUIVALENT MOISTURE Optical, cont. OLS DMSP F-16 560 0.5 Constrained October 2003 October 2014 ❷ ❹ ❹ OLS DMSP F-17 560 0.5 Constrained November June 2014 2006 ❷ ❹ ❹ OLS DMSP F-18 560 0.5 Constrained October 2009 April 2014 OLS DMSP F-15 560 0.5 Constrained December May 2014 ❷ ❹ ❹ 1999 ❷ ❹ ❹ AVHRR/3 NOAA-18 1,100 1 Open May 2005 December 2015 ❷ ❹ ❹ AVHRR/3 NOAA-19 1,100 1 Open February 2009 March 2016 ❷ ❹ ❹ AIRS Aqua 13,000 16 Open May 2002 October 2015 ❷ ❹ ❹ MSU-GS Elektro-L N1 1,000–4,000 Open January 2011 December 2018 ❷ ❹ ❹ MODIS Aqua 250–1,000 16 Open May 2002 October 2015 ❶ ❹ ❹ MODIS Terra 250–1,000 16 Open December October 1999 2015 ❶ ❹ ❹ VIIRS Suomi NPP 400–1,600 16 Open October 2011 March 2017 118  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T ❶ ❹ ❹ Passive microwave AMSU-A Aqua 48,000 16 Open May 2002 October 2015 ❶ ❷ ❹ AMSU-A NOAA-18 48,000 1 Open May 2005 December 2015 ❶ ❷ ❹ SSM/I DMSP F-15 15,700– 0.5 Constrained December May 2014 68,900 1999 ❷ ❷ ❹ SSM/IS DMSP F-16 25,000– 0.5 Constrained October 2003 October 42,000 2014 ❷ ❷ ❹ SSM/IS DMSP F-17 25,000– 0.5 Constrained November June 2014 42,000 2006 ❷ ❷ ❹ SSM/IS DMSP F-18 25,000– 0.5 Constrained October 2009 April 2014 42,000 ❷ ❷ ❹ AMSR-E Aqua 5,000–50,000 16 Open May 2002 October 2015 ❶ ❶ ❶ Thermal TIRS Landsat 8 100 16 Open February 2013 May 2023 (Continued) ❷ ❹ ❹ Table 6.11 (Continued) DATA CURRENCY SPATIAL SNOW AND FUNCTIONAL RESOLUTION REVISIT PERIOD ACCESSIBILITY WATER SNOW TYPE OF SENSOR SENSOR PLATFORM (METERS) (DAYS) AND COST LAUNCH DATE END DATE SNOW EXTENT EQUIVALENT MOISTURE Future Active microwave PALSAR-2 ALOS-2 10 14 Unknown March 2014 March (ALOS-2) 2019 ❹ ❷ ❹ C-band SAR Sentinel-1 B 10–50 12 Open December May 2023 2015 ❷ ❶ ❷ C-band SAR Sentinel-1 C 11–50 12 Open March 2019 June 2026 SAR (RCM) RADARSAT C-1 8–100 Constrained July 2018 ❷ ❶ ❷ 2025 November ❷ ❷ ❷ SAR (RCM) RADARSAT C-2 8–100 24 Constrained July 2018 2025 November ❷ ❷ ❷ SAR (RCM) RADARSAT C-3 8–100 Constrained July 2018 2025 November ❷ ❷ ❷ C-band SAR Sentinel-1 A 9–50 12 Open March 2014 January 2021 ❷ ❶ ❷ Optical PRISM-2 ALOS-3 1 60 Unknown December December (ALOS-3) 2015 2020 ❷ ❹ ❹ LISS-IV RESOURCESAT- 5.8 26 Constrained October 2015 October 2A 2020 ❷ ❹ ❹ HYSI CARTOSAT-3 12 Constrained July 2017 July 2022 (Cartosat- ❶ ❹ ❹ 3/3A) OLS DMSP F-19 560 0.5 Constrained March 2014 March 2019 ❷ ❹ ❹ MSU-GS Arctica 1,000–4,000 1 Open December December 2015 2018 ❷ ❹ ❹ MSU-GS Elektro-L N2 1,000–4,000 — Open June 2014 June 2019 MSU-GS Elektro-L N3 1,000–4,000 — Open December December ❷ ❹ ❹ 2015 2022 ❷ ❹ ❹ VIIRS JPSS-1 400–1,600 16 Open January 2017 March 2024 ❶ ❹ ❹ Passive microwave SSM/IS DMSP F-19 25,000– 0.5 Constrained March 2014 March 42,000 2019 ❷ ❷ ❹ Sources: Committee on Earth Observation Satellites (CEOS) Earth Observation Handbook (http://www.eohandbook.com/) and the World Monitoring Organization Observing Systems Capability Analysis and Review Tool (http:// www.wmo-sat.info/oscar/). chapter 6 : E arth O bservations for M onitoring W ater R esources   |  119 Note: The suitability of each sensor to provide useful data is shown with numbers and colors, as follows: ❶ highly suitable, ❷ suitable, and ❹ not suitable. SWE = snow water equivalent. — = not available. University of Utah and the University of Califor- provided on a daily, weekly, and monthly nia, Santa Barbara, have generated a MODIS- basis for the Northern Hemisphere. This based snow cover product specifically for use in information product first became available areas with complex terrain (Painter et al. 2009). in 1979 and still exists today (Pulliainen 2006; Takala et  al. 2011). The NSIDC pro- SNOW EXTENT duces an AMSR-E-based product for the The areal extent of snow cover can be detected world at 25-kilometer resolution that starts using optical, near infrared, and microwave in mid-2002 (Kelly et  al. 2003). A snow- sensors or a combination of these. Operational depth product for China has been generated snow cover maps are currently produced with by the Environmental and Ecological Sci- MODIS and AVHRR imagery using visible and ence Data Center for West China. This is a infrared sensors. The strength of visible and 25-kilometer resolution product spanning infrared sensors is their relative abundance 1978 to 2006 that is derived from passive and ease of access. They are, however, sensitive microwave data (SMMR and SMM/I; see to cloud cover, which can be common at high Che et al. 2008). More current snow depth, altitudes or in environments with significant extent, and snow water equivalent data snow cover. products have recently been developed for Various snow extent products are available. northern China (Dai et al. 2012). The two most widely used infrared- and NIR- based products are the MODIS product suites Optical Water Quality and Macrophytes and the ice mapping system (IMS). The Definition MOD10 suite of products provides daily, eight- For practical purposes, “inland waters” are day, and monthly estimates of global snow defined as inland surface waters, including cover at 500-meter and 0.05° resolutions (see rivers, lakes, artificial reservoirs, and estuar- Hall et al. 2002, 2010; Salomonson and Appel ies and their associated wetlands. “Water 2004). The National Snow and Ice Data Center quality” refers to the physical, chemical, and produces the interactive multisensor snow and biological content of water and may vary ice mapping data product. This is a daily prod- geographically and seasonally, irrespective uct for the Northern Hemisphere at 4-kilome- of the presence of specific pollution sources. ter and 24-kilometer resolutions (Helfrich Many factors affect water quality. No single et al. 2007; Ramsay 1998). measure exists for good water quality. There- SNOW WATER EQUIVALENT fore, the term “water quality” does not Active and passive microwave sensors are describe an absolute condition but rather a the primary means of detecting snow depth, condition relative to the use or purpose of snow water equivalent, and snow wetness. the water (for example, for drinking, irriga- The great advantage of microwave sensors is tion, industrial, recreational, or environmen- that they are not sensitive to cloud cover and tal purposes). Water that is suitable for can detect more than snow extent alone. irrigation, for instance, may not meet drink- Their disadvantages are that they are sensi- ing water standards. Thus “water quality” tive to the presence of trees and, in the case refers to the natural state of water bodies and of passive sensors, have relatively low spatial to their response to a combination of stress- resolutions. Dietz et  al. (2011) provide an ors such as changes in land use; nutrient excellent review of microwave-based meth- inputs; contamination from farming prac- ods for detecting snow. tices, industrial activity, and urbanization; The ESA produces the GlobSnow snow and changes in hydrology, flow regimes, and water equivalent data product, which is climate. 120  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Remote Sensing of Water Quality amount of light through the water column Earth observation can only be used directly to to above the surface. assess a subset of water quality variables, often Earth observation cannot directly assess referred to as optical water quality variables, water quality parameters that do not have a including concentrations of the following: direct expression in the optical response of • Chlorophyll (milligrams per cubic meter), the water body. These parameters include which is an indicator of phytoplankton many chemical compounds such as nutri- biomass, trophic, and nutrient status and ents. However, in some cases, nonoptical the most widely used index of water qual- products may be estimated through infer- ity and nutrient status globally ence, proxy relationships, or data assimila- tion with remotely sensed optical properties • Cyanophycocyanin (milligrams per cubic of products such as nitrogen, phosphate, meter) and cyanophycoerythrin (milli- organic and inorganic micropollutants, and grams per cubic meter), which are indica- dissolved oxygen. However, these relation- tors of cyanobacterial biomass common in ships are stochastic, may not be causal, and harmful and toxic algal blooms may have a limited range of validity. By mak- • Colored dissolved organic matter (per ing use of the combined information in meter absorption at 440 nanometers), directly measurable optical properties, it is which is the optically measurable com- possible to derive information about eutro- ponent of dissolved organic matter in the phication, environmental flows, and carbon water column, sometimes used as an indi- and primary productivity. cator of organic matter and aquatic carbon • Total suspended matter (milligrams per Relevance cubic meter) and nonalgal particulate Access to clean, safe drinking water is a key matter, which are important for assess- determinant of quality of life and is linked ing the quality of drinking water and con- directly to human health. Depending on the trolling the light characteristic of aquatic use to which the water is put, polluted or con- environments. taminated water may not be regarded as a usable resource. Similarly, as contaminant con- Additionally, the following conditions can centration is often related to water volume and be estimated: flow, water quality is ultimately linked to water • Vertical light attenuation (per meter) and quantity. Water supply and sanitation are thus turbidity, which measure the underwater essential components of any integrated light field and are important for assessing approach to malnutrition and poverty reduc- the degree of light limitation, rates of pri- tion, and water quality is a key related chal- mary production, species composition, and lenge in sustainable development. other ecosystem responses The quality of water is affected by stressors including urbanization, population growth, • Emergent and submerged macrophytes land use change, deforestation, farming, down to depth visibility, which are impor- overexploitation, and contamination from tant indicators of wetland and aquatic eco- extractive industries in the mining and energy system health and function sectors. As such, the relevance of water qual- • Bathymetry (meters), which can estimate ity issues will change in different settings, and water depth when the bottom or bottom their impact will ultimately depend on the cover of a water body reflects a measurable water’s intended use. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  121 Water quality monitoring is a key source of trends in water quality for several decades and information for ensuring that both human and to develop suitable reports to address specific ecosystem health are not compromised and questions raised by decision and policy makers. for determining the water’s suitability for other purposes (irrigation, industry). Nation- Theoretical Basis for Remote Sensing of states require information on water quality to Inland Water Quality inform key policy and legislative requirements Earth observation of the water quality param- that may include assessments against water eters identified above is achieved through opti- quality guidelines and targets, national cal means principally in the VIS-NIR spectrum water  quality management strategies, water (about 400–900 nanometers). The light reach- resources assessments, state of the environ- ing the surface of a water body consists of ment reporting, and strategies formulating direct sunlight and diffuse skylight after scat- adaptive responses to climate change. How- tering and absorption have interacted in the ever, even developed countries (such as Aus- atmosphere (figure 6.14). At the surface, this tralia and the United States) may not have any light is either reflected by the surface or nationally coordinated water quality monitor- refracted as it passes across the air-water inter- ing programs, and the authorities may instead face. Within the water column, the water itself rely on individual states to provide such infor- and different particulate and dissolved water mation; moreover, frameworks for disseminat- column constituents transform the light by ing such information are often lacking transmitting, absorbing, or scattering the altogether or poorly developed (Dekker and down-welling light. Of the light that is scat- Hestir 2012). tered, a proportion may be backscattered in an Despite international efforts to monitor upward direction and pass across the water-air global inland water quality, existing data are interface at the right angle to be observed by scarce and declining, have poor geographic and airborne or satellite sensors once it has again temporal coverage, may lack quality assurance passed through the atmosphere. and control, and may be of questionable accu- In the visible region (about 400–900 nano- racy (Srebotnjak et al. 2012). The international meters), the influence of sediments, chloro- coordinating group, the Group on Earth Obser- phyll, and colored dissolved organic matter vations (GEO), recognizes the value of Earth interacts to modify the shape and amount of observation for improving understanding of the spectrally reflected signal (Kirk 2011); RS global water quality, its hotspots, and trends; water quality algorithms largely take advan- for ensuring food and energy security; for facil- tage of these variations in the “shape” of spec- itating poverty reduction; for protecting the tral reflectance. In wavelengths longer than health of humans and ecosystems; and for 900 nanometers, water itself is such a strong maintaining biodiversity. GEO has formed the absorber that very little radiation is reflected Inland and Near-Coastal Water Quality from water bodies (figure 6.15). For this reason, Remote Sensing Working Group to promote the water quality variables listed above are the development of improved optical water often referred to as “optical water quality quality products (GEO 2011). variables.” Through the provision of synoptic, consis- The algorithms for translating the mea- tent, and comparable data, Earth observation sured spectral reflectance from a water body has the opportunity to overcome some of the to water quality variables include empirical gaps and deficiencies in current, field-based approaches (Tyler et  al. 2006; Wang et  al. water quality monitoring efforts. Sufficient 2009); semi-empirical approaches (Gons archives of EO data now exist to monitor global 1999; Härmä et  al. 2001); and physics-based, 122  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Figure 6.14 Schematic of the Light Interactions That Drive Optical EO Involving the Air, Water, and Substrate Sensor A Direct sunlight T M O S P H E Remote sensing signal R I Diffuse C sunlight A B S O R P T I O N Reflection at water surface Refraction at water surface A A Q B U S Aquatic A O plants T R I P C T I Coral O N Floor of water body Source: CSIRO Land and Water (A. G. Dekker and H. Buettikofer). Note: EO = Earth observation. Figure 6.15 Typical Reflectance Spectrum from semi-analytical spectral inversion methods Eutrophic Inland Water Body and Regions in Which (Brando et  al. 2012; Lee et  al. 1998). These Different Water Quality Parameters Influence the three methods are outlined below and subse- Shape of That Spectrum quently compared with regard to their need 12 for field measurements as well as their reli- Suspended 10 ability, accuracy, maturity, and complexity solids ­ hapter 7). (see c 8 Reflectance (%) Suspended Empirical approaches statistically relate 6 solids field samples of the optical water quality vari- ables to radiance or reflectance values mea- Phycocyanin 4 Colored dissolved Chlorophyll sured by a satellite or airborne sensor. There is 2 organic matter no need to understand the underlying physical relationships in such algorithms (such as 0 atmospheric and underwater light processes). 400 500 600 700 800 Wavelength (nanometer) However, they do require coincident field mea- surements to calibrate the relationships for chapter 6 : E arth O bservations for M onitoring W ater R esources   |  123 specific water bodies and, as such, struggle show improved accuracy for estimating water when water column constituents lie outside column composition (Dekker, Vos, and Peters the range on which the pertinent statistical 2001), are capable of assessing the error in the relationship is based (in both space and time) estimation of water quality constituents, are and are not easily adapted to new satellite sen- repeatable over time and space, are transfer- sors. Empirical methods are also less reliable able to new water bodies and other sensors, when undertaking retrospective monitoring, and can be applied retrospectively to image especially when the characteristics of lake archives (Dekker et  al. 2006; Odermatt et  al. water quality may change and end up outside 2012). This means that retrospective monitor- the range of those on which the empirical rela- ing of changes in optical water quality is possi- tionship is based. ble to assess the impacts and mitigation of Semi-empirical algorithms improve over various stressors to the system. pure empirical approaches by choosing the most A recommended pathway for longer-term appropriate single or spectral band combination operational use is to develop a robust, semi- to estimate the water column constituent. They analytical inversion method for application can also partly annul some of the atmospheric globally. Semi-empirical methods can be used and water surface effects. Semi-empirical algo- in the interim, as they often are reasonably rithms, however, also suffer from extrapolation robust for a category of water types and for a errors beyond the range of constituents single EO sensor system. Empirical approaches observed, the need to establish new, semi- are only useful as proof of concept. In general, empirical algorithms when switching s ­ ensors or they are not recommended if all optically water bodies, and the lack of reliability in retro- active substances (chlorophyll, colored dis- spective monitoring when characteristics of solved organic matter, total suspended solids, lake water quality change. They are therefore cyanophycocyanin, cyanophycoerythrin and less accurate than fully empirical methods. the resulting physical properties of turbidity, The water quality variables retrieved using Secchi disk depth, and vertical light attenua- empirical and semi-empirical algorithms tion) need to be determined. include total suspended matter, suspended inorganic matter, colored dissolved organic Mapping Inland Aquatic Macrophytes matter, turbidity, transparency, chlorophyll, Table 6.12 highlights the abilities of current and cyanophycocyanin pigments (Matthews and future optical sensors to differentiate 2011). With few exceptions (such as Minnesota among the growth habits of different macro- lakes in the United States; Olmanson, Brezonik, phytes. In addition to providing valuable habi- and Bauer 2011), neither approach offers sig- tat to multiple freshwater ecosystem species, nificant confidence for application in a national emergent wetland vegetation has extremely monitoring system (Dekker and Hestir 2012). high rates of net primary production and The Minnesota lakes method worked because evapotranspiration, drives a large portion of it was supported by a vast, citizens’ science- wetland carbon formation and storage, and based field measurement effort. plays an important role in wetland sediment Semi-analytical inversion algorithms are stability and accretion (Byrd et al. 2014; Zhou built around knowledge of the underlying and Zhou 2009). Floating and submersed physics of light transfer in waters and use the plants provide important structuring for fresh- inversion of predictions of light reflecting from water ecosystems, influencing the physical and a water body, generated by forward radiative chemical environment and food web (Liu et al. transfer models, to estimate key water quality 2013; Meerhoff et  al. 2003; Santos, Anderson, constituents simultaneously. Such approaches and Ustin 2011; Vanderstukken et al. 2014). 124  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 6.12 Existing and Near-Future Satellite Sensor Systems of Relevance for Inland and Near-Coastal Water Quality SPECTRAL BANDS WATER QUALITY VARIABLES MACROPHYTES DATA SENSOR (WATER- CURRENCY FUNCTIONAL RELEVANT REVISIT RAW DATA AND TYPE SPECTRAL FREQUENCY COST PER FUNCTIONAL (OPTICAL SPATIAL RANGE, CYCLE (ONCE SQUARE TYPE OF AND NEARBY RESOLUTION 400–1,000 EVERY X KILOMETER LAUNCH END TURB FLOATING- SENSOR INFRARED) (PIXEL SIZE) NANOMETERS) DAYS) (US$) DATE DATE CHL CYP TSM CDOM Kd SD EMERGENT LEAVED SUBMERSED Archival Ocean-coastal MERIS 1.2 and 0.3 15 2 days Free March May kilometers 2002 2012 ❶ ❶ ❶ ❶ ❶ ❶ ❸ ❹ ❹ Mid-spatial LANDSAT 1 to 7 30 meters 4 16 days Free July 1972 Surface resolution bloom ❸ ❷ ❷ ❷ ❷ ❷ ❸ ❸ SAR ERS 1 30 meters C band, VV 35 days Free 1991 (ERS-1) 2000 polarization (ERS-1) ❹ ❹ ❹ ❹ ❹ ❹ ❸ ❹ ❹ ERS 2 30 meters C band, VV 35 days Free 1995 2011 polarization (ERS-2) (ERS-2) ❹ ❹ ❹ ❹ ❹ ❹ ❸ ❹ ❹ ALOS PALSAR 10–100 meters L band, HH, VV, 46 days Commercial 2006 2011 HV, VH or R&D ❹ ❹ ❹ ❹ ❹ ❹ ❶ ❹ ❹ Current Coarse spatial MODIS-A&T 1 kilometer 9 Daily Free December resolution 1999 ❶ ❷ ❶ ❶ ❶ ❶ ❹ ❹ ❹ MODIS-A&T 500 meters 2 Daily Free December 1999 ❹ ❹ ❷ ❷ ❷ ❷ ❹ ❸ ❸ MODIS-A&T 250 meters 2 Daily Free December 1999 ❷ ❹ ❷ ❹ ❷ ❷ ❸ ❸ ❸ OCM-2 300 meters 15 2–3 days Free September 2009 ❶ ❶ ❶ ❶ ❶ ❶ ❸ ❹ ❹ Suomi-VIIRS 740 meters 7 Twice daily Free October 2011 ❶ ❷ ❶ ❶ ❶ ❶ ❹ ❹ ❹ Geostationary SEVIRI on MSG 1 kilometer 2 96 per 24 hours Free 2002 ❸ ❹ ❷ ❹ ❷ ❷ ❹ ❹ ❹ GOCI 500 meters 8 Half hourly Free 2010 ❶ ❸ ❶ ❷ ❶ ❶ ❹ ❹ ❹ Himawari-8 500 meters–2 4 10 minutes Free 2014 kilometers possible ❸ ❹ ❶ ❸ ❷ ❶ ❹ ❹ ❹ Mid-spatial LANDSAT 8 30 meters 5 16 days Free September Surface resolution 2013 bloom ❸ ❷ ❷ ❷ ❷ ❷ ❷ ❷ (Continued) |  125 126  | Table 6.12 (Continued) SPECTRAL BANDS WATER QUALITY VARIABLES MACROPHYTES DATA SENSOR (WATER- CURRENCY FUNCTIONAL RELEVANT REVISIT RAW DATA AND TYPE SPECTRAL FREQUENCY COST PER FUNCTIONAL (OPTICAL SPATIAL RANGE, CYCLE (ONCE SQUARE TYPE OF AND NEARBY RESOLUTION 400–1,000 EVERY X KILOMETER LAUNCH END TURB FLOATING- SENSOR INFRARED) (PIXEL SIZE) NANOMETERS) DAYS) (US$) DATE DATE CHL CYP TSM CDOM Kd SD EMERGENT LEAVED SUBMERSED High spatial IKONOS, 2–4 meters 3 to 4 Programmable: 5 to 15 October Surface resolution QuickBird, 60 days to 2–3 1999– bloom ❸ ❷ ❷ ❷ ❷ ❷ ❷ ❷ SPOT-5 , 6 days GeoEye RapidEye 6.5 meters 5 Daily 1.5 August Surface 2008 bloom ❸ ❷ ❷ ❷ ❷ ❷ ❷ ❷ WORLDVIEW-2 2 meters 8 Programmable: 30 October spectral– 0.5 60 days to 1 day 2009 ❷ ❷ ❶ ❷ ❶ ❶ ❶ ❶ ❶ meter in black and white SAR Radarsat 2 30 meters C band, fully 24 days Commercial 2007 polarimetric ❹ ❹ ❹ ❹ ❹ ❹ ❷ ❹ ❹ ALOS 2 3–100 meters L band, HH, VV, 14 days Commercial May 2014 HV, VH ❹ ❹ ❹ ❹ ❹ ❹ ❶ ❹ ❹ Sentinel 1 5–20 meters C band, HH, VV, 12 days Free March HV, VH 2014 ❹ ❹ ❹ ❹ ❹ ❹ ❷ ❹ ❹ Future High spatial Sentinel 2 10- to 60-meter 10 10 days per Free 2014 Surface resolution bands sensor; 5 days bloom ❷ ❶ ❶ ❶ ❶ ❷ ❷ ❷ with 2 Sentinel-2 sensors WORLDVIEW-3 1.24 meters 8 Programmable: 30 2014 spectral–0.50 60 days to 1 day ❷ ❷ ❶ ❷ ❶ ❶ ❶ ❶ ❶ meter in black and white Ocean-coastal Sentinel-3 300 meters 21 Daily (with 2 Free 2015 satellites) ❶ ❶ ❶ ❶ ❶ ❶ ❸ ❹ ❹ (Continued) Table 6.12 (Continued) SPECTRAL BANDS WATER QUALITY VARIABLES MACROPHYTES DATA SENSOR (WATER- CURRENCY FUNCTIONAL RELEVANT REVISIT RAW DATA AND TYPE SPECTRAL FREQUENCY COST PER FUNCTIONAL (OPTICAL SPATIAL RANGE, CYCLE (ONCE SQUARE TYPE OF AND NEARBY RESOLUTION 400–1,000 EVERY X KILOMETER LAUNCH END TURB FLOATING- SENSOR INFRARED) (PIXEL SIZE) NANOMETERS) DAYS) (US$) DATE DATE CHL CYP TSM CDOM Kd SD EMERGENT LEAVED SUBMERSED Hyperspectral EnMap 30 meters 90 Programmable Free 2017 (once per 4 ❶ ❶ ❶ ❶ ❶ ❶ ❷ ❷ ❷ days) PRISMA 20 meters 60 25 days Free 2017 spectral–2.5 ❶ ❶ ❶ ❶ ❶ ❶ ❷ ❷ ❷ meters in black and white SAR Cosmo-Skymed 5–100 meters X band, HH, VV, 16 days Commercial HV, VH or R&D ❹ ❹ ❹ ❹ ❹ ❹ ❸ ❹ ❹ TerraSAR-X/ 0.25–40 meters X band 3 days Commercial 2007 (2010) Tandem-X or R&D ❹ ❹ ❹ ❹ ❹ ❹ ❸ ❹ ❹ phytoplankton functional types; integrated products could be eutrophication index; water quality index, algal bloom index; carbon contents and flux; contaminant estimation. CHL = chlorophyll; CYP = cyanobacterial pigments such as Note: The suitability of each sensor to provide useful data is shown with numbers and colors, as follows: ❶ highly suitable, ❷ suitable, ❸ potentially suitable, ❹ not suitable. Products in development are coarse particle-size distributions and cyanophycocyanin and cyanophycoerythrin; TSM = total suspended matter; CDOM = colored dissolved organic matter; Kd = vertical attenuation of light coefficient; Turb = turbidity; SD = Secchi disk transparency. R&D = research and development. |  127 Routine mapping of the biophysical Common SAR wavelength bands include X parameters of macrophytes—derived from (3-centimeter wavelength), C (5.6-centimeter), high-resolution optical satellite or airborne S (10-centimeter), L (23-centimeter), and imagery in lakes and shallow, lentic environ- P  (75-centimeter) bands, and common SAR ments—has value for assessing cover and the detectors may be set up to receive defined effectiveness of management practices in polarizations in the same (horizontal transmit controlling excessive aquatic plant growth. and horizontal receive [HH] or vertical trans- Macrophytes may be separated into three mit and vertical receive [VV]) or cross-polar- groups, based on their principal growth hab- ization modes (horizontal transmit and vertical its—submersed, floating-leaved, and emer- receive [HV] or vertical transmit and horizon- gent—and the mapping of species by growth tal receive [VH]). Longer microwave wave- habit using both airborne and satellite data lengths (L band) penetrate further into can be reasonably accurate (Hunter et  al. canopies, and differences in polarization 2010; Malthus and George 1997; Tian et  al. behavior may also help to detect differences in 2010). The mapping is done largely on the specific vegetation canopy (Martínez and Le basis of reflectance values in NIR wavebands, Toan 2007). which are much stronger from emergent and The intensity of the radar backscatter is floating-leaved species. Hyperspectral data related directly to the roughness and, com- can differentiate several aquatic plant asso- bined with volumetric scattering, wavelength, ciations (Tian et  al. 2010) and be used to and polarization, provides specific vegetation detect submersed aquatic species, even in responses and hence information on canopy highly turbid environments (Hestir et  al. characteristics (Evans et  al. 2010; Kasischke 2008; Santos et  al. 2012), as can the use of and Bruhwiler 2003; Klemas 2013). Common LiDAR and textural analysis of image data satellite-borne SAR systems, as well as their (Proctor, He, and Robinson 2013; Verrelst characteristics and abilities to differentiate et al. 2009). Differentiation of species, how- emergent vegetation, are highlighted in ever, currently poses a greater challenge. table  6.12. However, because wetlands are Because of the high spatial and phenological highly spatially heterogeneous, the large foot- variability of aquatic macrophytes, high- print provided by most SAR systems also limits spectral-resolution data are needed to dis- their ability to discriminate wetland plant spe- criminate communities adequately (Klemas cies successfully from space. 2013) and measure the biogeochemical fea- tures needed for species discrimination and Applications physiological function (Santos et  al. 2012; Systematic examples of truly operational mon- Ustin et al. 2004). itoring of inland water quality beyond that SAR data also have value in offering applied to single water bodies are lacking, weather-independent monitoring of aquatic reflecting the challenges in applying more sim- macrophytes and wetlands as well as flooding ple empirical and semi-empirical algorithms. extent (Silva et al. 2008). Dielectric signal dif- Using empirical methods, Olmanson, Bauer, ferences arise from the presence of water as and Brezonik (2008) compiled a comprehen- surface water and within vegetation, thus mak- sive water clarity database assembled from ing it possible to detect dry and flooded vegeta- Landsat imagery over 1985–2005 for more tion and, hence, to map the extent of flooding than 10,500 Minnesota lakes larger than and of emergent vegetation (Costa 2004; Evans 8  hectares in surface area. This study high- et al. 2010). The lack of penetration of micro- lighted the geographic patterns in clarity waves into water prevents detection of sub- linked to land use at the level of both individual mersed macrophyte species. lake and eco-region.21 128  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Algorithm development to allow applica- quality. In improving the design of such assess- tion beyond a single inland water body is only ments, the following are key considerations: now being addressed in some research proj- ects targeting larger lakes with ocean color • Temporal sampling to represent the dynam- sensors (Global Lakes Sentinel Services ics of water quality and the range of condi- [GLaSS] and GloboLakes). The monitoring of tions that can occur over diurnal, seasonal, water quality conditions across the Great Bar- and annual cycles (droughts and flood- rier Reef World Heritage Park offers the best ing) as well as to develop a time series for example of the potential to deliver water trend analysis. Retrospective process- quality products derived from semi-analytical ing of satellite images, with archives dat- inversion algorithms. MODIS data are used to ing back to the mid-1980s, may also reveal derive concentrations of key water quality temporal changes, trends, and anomalies constituents for the reef on a daily basis, and across inland water and near-coastal water the data are delivered via the Australian systems. Bureau of Meteorology’s Marine Water Qual- • Spatial sampling to represent water bodies ity Dashboard.22 under consideration and provide under- The following data services and research standing of system processes such as projects provide water quality products heterogeneity, environmental flows, inter- derived from Earth observation: relationships between water bodies, and • Downstream services of the European catchment runoff effects. Union and ESA Copernicus Programme23 End user requirements should determine • Marine water quality and forecasting by the optimal spatial sampling scheme, but the European Union and ESA’s Environ- logistical, operational, and financial con- ment Monitoring Services24 straints usually prevent the optimal sampling scheme from being realized. Extensive dis- • Monitoring of harmful algal bloom in Lake tances, for instance, may make capturing the Erie by the National Oceanic and Atmo- spatial distribution of measurements using spheric Administration (NOAA)25 field-based methods unfeasible. EO-derived • Development of a harmful algal bloom water quality information, albeit for a more advisory and forecasting capability by the limited set of parameters, may be used to over- European Union’s ASIMUTH (Applied come the challenges in water quality sampling Simulations and Integrated Modelling for schemes based solely on field-based the Understanding of Toxic and Harmful approaches to provide complementary over- Algal Blooms) project26 sight of water quality conditions and trends. In future, capacity building should focus on inte- • Global Earth observation for inte- grating EO data and field-based observations grated water resource assessment by and on developing early warning tools for algal EartH2Observe.27 blooms. These examples are discussed further in Table 6.12 provides an overview of existing appendix B. and upcoming satellite sensor systems of rele- vance for monitoring inland water quality Past, Present, and Future Sensor Availability internationally and their suitability for mea- for Inland Water Mapping suring optical water quality variables. While In many countries, field-based water quality policy, legislative, environmental, and climate monitoring efforts are insufficient to provide change drivers should steer the development national-scale assessments of inland water of an operational system for inland water chapter 6 : E arth O bservations for M onitoring W ater R esources   |  129 quality monitoring, the ideal satellite sensor for smaller or narrow water bodies, reducing system for inland water quality does not exist; the need for high-resolution imagery and thus there are trade-offs between spatial, temporal, also reducing cost. spectral, and radiometric characteristics. Thus Spectral resolution (the number, width, and having satellite sensors available for detecting placing of spectral bands) ultimately deter- and monitoring retrospective, current, and mines the amount and accuracy of water future inland water quality is necessary for quality variables that are discernable from a developing regional, national, and transbound- water body (table 6.12). Sensors with few ary inland water quality monitoring systems bands may only be used to detect total sus- using Earth observation. pended matter, vertical light attenuation, Sec- Different satellite systems show different chi disk transparency, turbidity, and colored trade-offs between temporal frequency (once a dissolved organic matter if a blue spectral day to once a year), spatial resolution (2-meter band is available. Algal pigments such as chlo- to 1.2-kilometer pixels), spectral resolution rophyll may also be detected. However, at low (and the related issue of more water quality concentrations, accuracy will be low, as broad variables at higher confidence levels), radio- spectral bands cannot discriminate the more metric resolution (how accurate and how narrow features of pigment spectral absorp- many levels of reflectance are measurable), and tion from other absorbing and backscattering the cost of acquiring unprocessed satellite data materials in the water column. As the number (ranging from US$0 to about US$30 per square of narrower and more suitably positioned kilometer). This also influences their useful- spectral bands increases (MODIS, MERIS, ness for inland water quality assessment. and Ocean Colour Monitor [OCM]-2), chloro- Tools are needed for reporting information phyll becomes an accurately measurable vari- on water quality at a variety of scales (conti- able, and types of phytoplankton pigment nental, transboundary, regional, and national). such as cyanobacterial pigments may become However, satisfying this need is challenging detectable. given the multiple-size scales of inland water Radiometric resolution determines the low- bodies with respect to the different spatial and est level of radiance or reflectance that the spectral resolutions offered by the different sensor can reliably detect per spectral band. satellite sensors. As the spectral and spatial resolution Spatial resolution has consequences for increases, the useful signal relative to noise in imaging small water bodies such as small- or the data decreases, but this trade-off in spec- medium-width river systems. In such situa- tral, spatial, and radiometric resolution is tions, high-spatial-resolution imagery (with countered by improvements in detector tech- pixel sizes of 2 to 10 meters) may be the only nology where, in general, more modern sen- option, possibly leading to significant data sors have a higher radiometric sensitivity acquisition costs. A multiple-resolution overall than older sensors. approach is most cost-effective where coarse (but frequent) satellite imagery is used for larger lakes, reservoirs, and river sections and NOTES high-resolution imagery is acquired and pro- 1. For more information on the data sets of the cessed only when necessary. Rather than imag- International Precipitation Working Group, see ing all water bodies, a “virtual station concept” http://www.isac.cnr.it/~ipwg/data/datasets.html. approach could systematically image a selec- 2. Orographic lift occurs when an air mass is forced from a low elevation to a higher elevation as tion of water bodies that represent the associ- it moves over rising terrain. As the air mass ated aquatic ecosystem (be it natural or gains altitude, it quickly cools down adiabati- artificial). This would be especially effective cally, which can raise the relative humidity to 130  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T 100 percent and create clouds and, under the right 21. For information on the project, see http://water conditions, precipitation. .umn.edu/lwc/index.html. 3. For more information on the Asia-Pacific Water 22. For the Marine Water Quality Dashboard, see Monitor, see http://eos.csiro.au/apwm/. http://www.bom.gov.au/marinewaterquality/. 4. For information on the Global Precipitation 23. For an overview of the EU-ESA Copernicus ­ Climatology Project, see http://www.gewex.org/ Programme Downstream services, see http:// gpcp.html. gmesdata.esa.int/web/gsc/core_services/ 5. For information on Eddy covariance flux data, downstream_services. see http://www.fluxnet.ornl.gov/. 24. For more information on the Copernicus Pro- 6. Antecedent moisture is a term from the fields of gramme’s Environment Monitoring Services, see hydrology and sewage collection and disposal http://www.myocean.eu/web/26-catalogue that describes the relative wetness or dryness -of-services.php. of a watershed or sanitary sewershed. Anteced- 25. For information from NOAA’s Great Lakes Envi- ent moisture conditions change continuously ronmental Laboratory, see http://www.glerl.noaa and can have a very significant effect on the flow .gov/res/Centers/HABS/lake_erie_hab/lake_erie responses in these systems during wet weather. _hab.html. 7. Signal-to-noise ratio is the ratio of the level of the 26. For the ASIMUTH project, see http://www signal carrying real information to the level of the .asimuth.eu/. signal carrying spurious information (as a result 27. For EartH2Observe, see http://www of defects in the system). .earth2observe.eu/. 8. For the Earth Observation Handbook database, see http://www.eohandbook.com/. For the Observing Systems Capability Analysis and Review Tool, see http://www.wmo-sat.info/oscar/. REFERENCES 9. For the Global Drought Information System, see www.drought.gov. AghaKouchak, A., A. Behrangi, S. Sorooshian, K. Hsu, 10. For the European Drought Observatory, see and E. Amitai. 2011. “Evaluation of Satellite- http://edo.jrc.ec.europa.eu/. Retrieved Extreme Precipitation Rates across the 11. For the Climate Maps, see http://www.bom.gov Central United States.” Journal of Geophysical .au/jsp/awap/ndvi/. Research 116 (January): D02115. 12. For the African Drought Monitor, see http:// Albergel, C., E. Zakharova, J.-C. Calvet, M. Zribi, drought.icpac.net/. M. Pardé, J.-P. Wigneron, N. Novello, Y. Kerr, 13. For the global terrestrial drought severity index, A. Mialon, and N. Fritz. 2011. “A First Assessment see http://www.ntsg.umt.edu/project/dsi. of the SMOS Data in Southwestern France Using in Situ and Airborne Soil Moisture Estimates: The 14. For the Groundwater Foundation, see http:// CAROLS Airborne Campaign.” Remote Sensing of www.groundwater.org/get-informed/basics/ Environment 115 (10): 2718–28. groundwater.html. Allen, R. G., M. Tasumi, A. Morse, R. Trezza, 15. Backscatter is the portion of the outgoing radar J. L. Wright, W. Bastiaanssen, W. Kramber, signal that the target redirects directly back I. Lorite, and C. W. Robison. 2007. “Satellite-Based toward the radar antenna, backscattering is the Energy Balance for Mapping Evapotranspiration process by which backscatter is formed, and back- with Internalized Calibration (METRIC) Applica- scatter coefficient is a measure of the reflective tions.” Journal of Irrigation and Drainage Engi- strength of a radar target. neering 133 (4): 395–406. 16. For the Dartmouth Flood Observatory, see http:// Alsdorf, D. E., E. Rodríguez, and D. P. Letten- floodobservatory.colorado.edu/. maier. 2007. “Measuring Surface Water from 17. For Near-Real-Time Global MODIS Flood Map- Space.” Reviews of Geophysics 45 (2): RG2002. ping, see http://oas.gsfc.nasa.gov/floodmap/. Andela, N., Y. Y. Liu, A. I. J. M. van Dijk, R. A. M. de 18. For Crop Explorer, see http://www.pecad.fas Jeu, and T. R. McVicar. 2013. “Global Changes in .usda.gov/cropexplorer/global_reservoir/. Dryland Vegetation Dynamics (1988 and 2008) 19. For LEGOS (Laboratoire d’Etudes en Géophy- Assessed by Satellite Remote Sensing: Compar- sique et Océanographie Spatiales) HydroWeb, see ing a New Passive Microwave Vegetation Density http://www.legos.obs-mip.fr/en/soa/hydrologie Record with Reflective Greenness Data.” Biogeo- /hydroweb/. sciences 10 (10): 6657–76. 20. For the European Space Agency’s East Rivers & Anderson, M. C., W. P. Kustas, J. M. Norman, Lakes, see http://tethys.eaprs.cse.dmu C. R. Hain, J. R. Mecikalski, L. Schultz, .ac.uk/RiverLake/. M. P. González-Dugo, C. Cammalleri, G. d’Urso, chapter 6 : E arth O bservations for M onitoring W ater R esources   |  131 A. Pimstein, and F. Gao. 2011. “Mapping Daily Remotely Sensed Soil Moisture Retrievals for Evapotranspiration at Field to Continental Scales Operational Agricultural Drought Monitoring.” Using Geostationary and Polar Orbiting Satellite IEEE Journal on Selected Topics in Applied Earth Imagery.” Hydrology and Earth System Sciences Observations and Remote Sensing 3 (1): 57–66. 15 (1): 223–39. Brakenridge, R., and E. Anderson. 2006. “MODIS- Anderson, M. C., J. M. Norman, J. R. Mecikalski, Based Flood Detection, Mapping, and Measure- J. A. Otkin, and W. P. Kustas. 2007. “A Climatologi- ment: The Potential for Operational Hydrological cal Study of Evapotranspiration and Moisture Applications.” Transboundary Floods Reducing Stress across the Continental United States Based Risks through Flood Management 72 (1): 1–12. on Thermal Remote Sensing: 1. Model Formula- Brando, V. E., A. G. Dekker, Y-Je Park, and T. Schroeder. tion.” Journal of Geophysical Research 112 (D10): 2012. “Adaptive Semi-Analytical Inversion of D11112. Ocean Color Radiometry in Optically Complex Arkin, P. A., and B. N. Meisner. 1987. “The Relationship Waters.” Applied Optics 51 (15): 2808. between Large-Scale Convective Rainfall and Brocca, L., F. Melone, T. Moramarco, W. Wagner, Cold Cloud Cover over the Western Hemisphere V. Naeimi, Z. Bartalis, and S. Hasenauer. 2010. during 1982–1984.” Monthly Weather Review “Improving Runoff Prediction through the 115 (1): 51–74. Assimilation of the ASCAT Soil Moisture Prod- Ashouri, H., H.-L. Hsu, S. Sorooshian, and uct.” Hydrology and Earth System Sciences 14 (10): D. K. Braithwaite. 2014. “PERSIANN-CDR: 1881–93. Daily Precipitation Climate Data Record from Brocca, L., T. Moramarco, F. Melone, W. Wagner, Multi-Satellite Observations for Hydrological S. Member, S. Hasenauer, and S. Hahn. 2012. and Climate Studies.” Bulletin of the Ameri- “Assimilation of Surface- and Root-Zone ASCAT can Meteorological Society 96 (1). doi:10.1175/ Soil Moisture Products into Rainfall–Runoff BAMS-D-13-00068.1. Modeling.” IEEE Transaction on Geoscience and Barré, H. M. J. P., B. Duesmann, Y. H. Kerr, and Remote Sensing 50 (7): 2542–55. S. Member. 2008. “SMOS: The Mission and the BoM (Bureau of Meteorology). 2012. “Australian System.” IEEE Transactions on Geoscience and Water Resources Assessment.” Australian Bureau Remote Sensing 46 (3): 587–93. of Meteorology, Melbourne. Bastiaanssen, W. G. M., M. Menenti, R. A. Feddes, and A. A. M. Holtslag. 1998. “A Remote Sensing Sur- Budyko, M. I., and D. Miller. 1974. “Climate and face Energy Balance Algorithm for Land (SEBAL). Life.” International Geophysics Series 18. Bondi, 1. Formulation.” Journal of Hydrology 212-213 ­Australia: Geniza. (December): 198–212. Byrd, K. B., J. L. O’Connell, S. Di Tommaso, and Beck, H. E., T. R. McVicar, A. I. J. M. van Dijk, M. Kelly. 2014. “Evaluation of Sensor Types and J. Schellekens, R. A. M. de Jeu, and L. A. Bruijn- Environmental Controls on Mapping Biomass zeel. 2011. “Global Evaluation of Four AVHRR- of Coastal Marsh Emergent Vegetation.” Remote NDVI Data Sets: Intercomparison and Sensing of Environment 149 (June): 166–80. Assessment Against Landsat Imagery.” Remote doi:10.1016/j.rse.2014.04.003. Sensing of Environment 115 (10): 2547–63. Chappell, A., L. J. Renzullo, T. H. Raupach, and Becker, M. W. 2006. “Potential for Satellite Remote M. Haylock. 2013. “Evaluating Geostatistical Sensing of Ground Water.” Ground Water 44 (2): Methods of Blending Satellite and Gauge Data to 306–18. Estimate Near Real-Time Daily Rainfall for Aus- tralia.” Journal of Hydrology 493 (June): 105–14. Bindlish, R., and T. J. Jackson. 2013. “Soil Moisture Products Using Aquarius/SAC-D Observations.” CEOS (Committee on Earth Observation Satellites). Algorithm Theoretical Basis Document Version 2015. The Earth Observation Handbook. http:// 2.0. U.S. Department of Agriculture Hydrology www.eohandbook.com/. and Remote Sensing Lab, Beltsville, MD. http:// Che, T., X. Li, R. Jin, R. Armstrong, and T. J. Zhang. nsidc.org/data/docs/daac/aquarius/pdfs/Aquar- 2008. “Snow Depth Derived from Passive Micro- ius_VSM_ATBD_UsersGuide.pdf. wave Remote-Sensing Data in China.” Annals of Bojinski, S., M. Verstraete, T. C. Peterson, C. Richter, Glaciology 49 (1): 145–54. A. Simmons, and M. Zemp. 2014. “The Concept of Chen, S., Y. Hong, J. J. Gourley, G. J. Huffman, Y. Tian, Essential Climate Variables in Support of Climate Q. Cao, B. Yong, P.-E. Kirstetter, J. Hu, J. Hardy, Research, Applications, and Policy.” Bulletin of the Z. Li, S. I. Khan, and X. Xue. 2013. “Evaluation American Meteorological Society 95 (9): 1431–43. of the Successive V6 and V7 TRMM Multisatel- doi:10.1175/BAMS-D-13-00047.1. lite Precipitation Analysis over the Continental Bolten, J. D., W. T. Crow, X. Zhan, T. J. Jackson, and United States.” Water Resources Research C. A. Reynolds. 2010. “Evaluating the Utility of 49 (12): 1–13. 132  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Cleugh, H. A., R. Leuning, Q. Mu, and S. W. Running. Dharssi, I., K. Bovis, B. Macpherson, and C. Jones. 2007. “Regional Evaporation Estimates from Flux 2011. “Operational Assimilation of ASCAT Surface Tower and MODIS Satellite Data.” Remote Sens- Soil Wetness at the Met Office.” Hydrology and ing of Environment 106 (3): 285–304. Earth System Sciences Discussions 8 (2): 4313–54. Conrad, C., R. R. Colditz, S. Dech, D. Klein, and Dietz, A. J., C. Kuenzer, U. Gessner, and S. Dech. 2011. P. L. G. Vlek. 2011. “Temporal Segmentation of “Remote Sensing of Snow: A Review of Available MODIS Time Series for Improving Crop Classifica- Methods.” International Journal of Remote Sensing tion in Central Asian Irrigation Systems.” Interna- 33 (13): 4094–134. tional Journal of Remote Sensing 32 (23): 8763–78. Dobson, M. C., F. T. Ulaby, M. T. Hallikainen, and Costa, M. P. F. 2004. “Use of SAR Satellites for Map- M. A. El-Rayes. 1985. “Microwave Dielectric ping Zonation of Vegetation Communities in the Behavior of Wet Soil: Part II; Dielectric Mixing Amazon Floodplain.” International Journal of Models.” IEEE Transactions on Geoscience and Remote Sensing 25 (10): 1817–35. Remote Sensing GE-23 (1): 35–46. CSIRO (Commonwealth Scientific and Industrial Donohue, R. J., T. R. McVicar, and M. L. Roderick. Research Organization). 2011. Water: Science and 2010. “Assessing the Ability of Potential Evapora- Solutions for Australia. Canberra: CSIRO Publish- tion Formulations to Capture the Dynamics in ing. http://www.publish.csiro.au/pid/6557.htm. Evaporative Demand within a Changing Climate.” Journal of Hydrology 386 (1-4): 186–97. Dai, L., T. Che, J. Wang, and P. Zhang. 2012. “Snow Depth and Snow Water Equivalent Estimation Dorigo, W. A., K. Scipal, R. M. Parinussa, Y. Y. Liu, from AMSR-E Data Based on a Priori Snow Char- W. Wagner, R. A. M. de Jeu, and V. Naeimi. 2010. acteristics in Xinjiang, China.” Remote Sensing of “Error Characterisation of Global Active and Pas- Environment 127 (December): 14–29. sive Microwave Soil Moisture Datasets.” Hydrol- ogy and Earth System Sciences 14 (12): 2605–16. de Jeu, R. A. M. 2003. “Retrieval of Land Surface Parameters Using Passive Microwave Remote Dorigo, W. A., A. Xaver, M. Vreugdenhil, A. Gruber, Sensing.” Vrije Universiteit, Amsterdam. A. Hegyiová, A. D. Sanchis-Dufau, and M. Drusch. 2013. “Global Automated Quality Control of In de Jeu, R. A. M., and M. Owe. 2003. “Further Valida- Situ Soil Moisture Data from the International tion of a New Methodology for Surface Moisture Soil Moisture Network.” Vadose Zone Journal and Vegetation Optical Depth Retrieval.” Interna- 12 (3). doi: 10.2136/vzj2012.0097. tional Journal of Remote Sensing 24 (22): 4559–78. Draper, C., J.-F. Mahfouf, J.-C. Calvet, E. Martin, and Dekker, A. G., V. E. Brando, J. M. Anstee, S. Fyfe, W. Wagner. 2011. “Assimilation of ASCAT Near- T. J. M. Malthus, and E. V. Karpouzli. 2006. Surface Soil Moisture into the SIM Hydrological “Remote Sensing of Seagrass Ecosystems: Use of Model over France.” Hydrology and Earth System Spaceborne and Airborne Sensors.” In Seagrass Sciences 15 (12): 3829–41. Biology, Ecology, and Conservation, edited by Draper, C. S., R. H. Reichle, G. J. M. De Lannoy, A. Larkum, R. Orth, and C. Duarte. Heidelberg: and Q. Liu. 2012. “Assimilation of Passive and Springer Verlag. Active Microwave Soil Moisture Retriev- Dekker, A. G., and E. L. Hestir. 2012. Evaluating the als.” Geophysical Research Letters 39 (4). Feasibility of Systematic Inland Water Quality doi:10.1029/2011GL050655. Monitoring with Satellite Remote Sensing. Water Draper, C. S., J. P. Walker, P. J. Steinle, R. A. M. de Jeu, for a Healthy Country National Research Flagship. and T. R. H. Holmes. 2009. “An Evaluation of Canberra: CSIRO. https://publications.csiro.au/ AMSR–E Derived Soil Moisture over Australia.” rpr/download?pid=csiro:EP117441&dsid=DS10. Remote Sensing of Environment 113 (4): 703–10. Dekker, A. G., R. J. Vos, and S. W. M. Peters. 2001. Ebert, E. E., J. E. Janowiak, and C. Kidd. 2007. “Com- “Comparison of Remote Sensing Data, Model parison of Near-Real-Time Precipitation Esti- Results and In Situ Data for Total Suspended Mat- mates from Satellite Observations and Numerical ter (TSM) in the Southern Frisian Lakes.” Science Models.” Bulletin of the American Meteorological of the Total Environment 268 (1-3): 197–214. Society 88 (1): 47–64. Desilets, D., and M. Zreda. 2013. “Footprint Diameter Emelyanova, I. V., T. R. McVicar, T. G. Van Niel, for a Cosmic-Ray Soil Moisture Probe: Theory L. T. Li, and A. I. J. M. van Dijk. 2012. “On Blend- and Monte Carlo Simulations.” Water Resources ing Landsat-MODIS Surface Reflectances in Two Research 49 (6): 3566–75. Landscapes with Contrasting Spectral, Spatial, Desilets, D., M. Zreda, and T. P. A. Ferré. 2010. and Temporal Dynamics.” WIRADA Project 3.4: “Nature’s Neutron Probe: Land Surface Technical Report, CSIRO Water for a Healthy Hydrology at an Elusive Scale with Cosmic Country Flagship, Australia. https://publications Rays.” Water Resources Research 46 (11): 1–7. .csiro.au/rpr/download?pid=csiro:EP128838&ds doi:10.1029/2009WR008726. id=DS1. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  133 Emelyanova, I. V., T. G. Van Niel, and T. R. McVicar. Gleeson, T., Y. Wada, M. F. P. Bierkens, and L. P. H. van n.d. “Clarifying Terminology in Multi-Sensor Beek. 2012. “Water Balance of Global Aquifers Data Fusion Literature: A Taxonomic Approach.” Revealed by Groundwater Footprint.” Nature 488 CSIRO, Canberra. (7410): 197–200. Entekhabi, D., E. G. Njoku, P. O’Neill, K. H. Kel- Goetz, S. J., A. Baccini, N. T. Laporte, T. Johns, logg, W. T. Crow, W. N. Edelstein, J. K. Entin, W. Walker, J. Kellndorfer, R. A. Houghton, and S. D. Goodman, T. J. Jackson, J. Johnson, M. Sun. 2009. “Mapping and Monitoring Carbon J. Kimball, J. R. Piepmeier, R. D. Koster, Stocks with Satellite Observations: A Comparison N. Martin, K. C. McDonald, M. Moghaddam, of Methods.” Carbon Balance and Management S. Moran, R. Reichle, J.-C. Shi, M. W. Spencer, 4 (1): 1–7. S. W. Thurman, T. Leung, and J. Van Zyl. 2010. Gong, P., J. Wang, L. Yu, Y. Zhao, Y. Zhao, L. Liang, “The Soil Moisture Active Passive (SMAP) Z. Niu, X. Huang, H. Fu, S. Liu, C. Li, X. Li, W. Fu, Mission.” Proceedings of the IEEE 98 (5): C. Liu, Y. Xu, X. Wang, Q. Cheng, L. Hu, W. Yao, 704–16. H. Zhang, P. Zhu, Z. Zhao, H. Zhang, Y. Zheng, L. Ji, Evans, T. L., M. P. F. Costa, K. H. Telmer, and T. S. F Y. Zhang, H. Chen, A. Yan, J. Guo, L. Yu, L. Wang, Silva. 2010. “Using ALOS/PALSAR and RADAR- X. Liu, T. Shi, M. Zhu, Y. Chen, G. Yang, P. Tang, SAT-2 to Map Land Cover and Seasonal Inunda- B. Xu, C. Giri, N. Clinton, Z. Zhu, J. Chen, and tion in the Brazilian Pantanal.” IEEE Journal of J. Chen. 2012. “Finer Resolution Observation and Selected Topics in Applied Earth Observations and Monitoring of Global Land Cover: First Mapping Remote Sensing 3 (4): 560–75. Results with Landsat TM and ETM+ Data.” Interna- FAO (Food and Agriculture Organization). 2013. tional Journal of Remote Sensing 34 (7-7): 2607–54. AQUASTAT Database. Rome: Food and Agricul- Gons, H. J. 1999. “Optical Teledetection of Chlorophyll ture Organization of the United Nations (FAO). A in Turbid Inland Waters.” Environmental Sci- Franz, T. E., M. Zreda, R. Rosolem, B. K. Hornbuckle, ence and Technology 33 (7): 1127–32. S. L. Irvin, H. Adams, T. E. Kolb, C. Zweck, and Gourley, J. J., Y. Hong, Z. L. Flamig, L. Li, and J. Wang. W. J. Shuttleworth. 2013. “Ecosystem-Scale 2010. “Intercomparison of Rainfall Estimates from Measurements of Biomass Water Using Cosmic ­ Radar, Satellite, Gauge, and Combinations for a Ray Neutrons.” Geophysical Research Letters Season of Record Rainfall.” Journal of Applied 40 (15): 3929–33. Meteorology and Climatology 49 (3): 437–52. Frei, A., M. Tedesco, S. Lee, J. Foster, D. K. Hall, Guerschman, J. P., M. J. Hill, L. J. Renzullo, D. J. Bar- R. Kelly, and D. A. Robinson. 2012. “A Review rett, A. S. Marks, and E. J. Botha. 2009a. “Estimat- of Global Satellite-Derived Snow Products.” ing Fractional Cover of Photosynthetic Vegetation, Advances in Space Research 50 (8): 1007–29. Non-Photosynthetic Vegetation, and Bare Soil in Friedl, M. A. 1995. “Modeling Land Surface Fluxes the Australian Tropical Savanna Region Upscaling Using a Sparse Canopy Model and Radiometric the EO-1 Hyperion and MODIS Sensors.” Remote Surface Temperature Measurements.” Journal of Sensing of Environment 113 (5): 928–45. Geophysical Research 100 (D12): 25435–46. Guerschman, J. P., A. I. J. M. Van Dijk, G. Mattersdorf, Friedl, M. A., D. Sulla-Menashe, B. Tan, A. Schneider, J. Beringer, L. B. Hutley, R. Leuning, R. C. Pipunic, N. Ramankutty, A. Sibley, and X. Huang. 2010. and B. S. Sherman. 2009b. “Scaling of Potential “MODIS Collection 5 Global Land Cover: Algo- Evapotranspiration with MODIS Data Repro- rithm Refinements and Characterization of New duces Flux Observations and Catchment Water Datasets.” Remote Sensing of Environment 114 (1): Balance Observations across Australia.” Journal of 168–82. Hydrology 369 (1-2): 107–19. Galloway, D., and J. Hoffmann. 2007. “The Applica- Guerschman, J. P., G. Warren, G. Byrne, L. Lymburner, tion of Satellite Differential SAR Interferometry- N. Mueller, and A. I. J. M. Van Dijk. 2011. MODIS- Derived Ground Displacements in Hydrogeology.” Based Standing Water Detection for Flood and Hydrogeology Journal 15 (1): 133–54. Large Reservoir Mapping: Algorithm Development Gebremichael, M., and F. Hossain, eds. 2010. Satellite and Applications for the Australian Continent. Rainfall Applications for Surface Hydrology. Dor- Canberra: CSIRO. drecht: Springer Science+Business Media. Gumma, M. K., P. S. Thenkabail, and A. Nelson. 2011. GEO (Group on Earth Observations). 2011. “Progress “Mapping Irrigated Areas Using MODIS 250 Report on GEO Inland and Near-Coastal Water Meter Time-Series Data: A Study on Krishna Quality Remote Sensing Working Group.” https:// River Basin (India).” Water 3 (1): 113–31. www.earthobservations.org/documents/sbas/ Güntner, A. 2008. “Improvement of Global Hydrologi- wa/201101_coastal_and_inland_wq_remote_sens- cal Models Using GRACE Data.” Surveys in Geo- ing_wg_progress_report.pdf physics 29 (4-5): 375–97. 134  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Hall, D. K., G. A. Riggs, J. L. Foster, and S.V. Kumar. Houborg, R., A. Cescatti, M. Migliavacca, and 2010. “Development and Evaluation of a Cloud- W. P. Kustas. 2013. “Satellite Retrievals of Leaf Gap-Filled MODIS Daily Snow-Cover Prod- Chlorophyll and Photosynthetic Capacity for uct.” Remote Sensing of Environment 114 (3): Improved Modeling of GPP.” Agricultural and For- 496–503. est Meteorology 177 (1): 10–23. Hall, D. K., G. A. Riggs, V. V. Salomonson, N. E. DiGi- Hsu, K. L., X. G. Gao, S. Sorooshian, and H. V. Gupta. rolamo, and K. J. Bayr. 2002. “MODIS Snow- 1997. “Precipitation Estimation from Remotely Cover Products.” Remote Sensing of Environment Sensed Information Using Artificial Neural Net- 83 (1-2): 181–94. works.” Journal of Applied Meteorology 36 (9): Hansen, M. C., R. S. DeFries, J. R. G. Townshend, and 1176–90. R. Sohlberg. 2000. “Global Land Cover Classifica- Huete, A., K. Didan, T. Miura, E. P. Rodriguez, X. Gao, tion at 1 Km Spatial Resolution Using a Classifi- and L. G. Ferreira. 2002. “Overview of the Radio- cation Tree Approach.” International Journal of metric and Biophysical Performance of the Remote Sensing 21 (6-7): 1331–64. MODIS Vegetation Indices.” Remote Sensing of Härmä, P., J. Vepsalainen, T. Hannonen, T. Pyhalahti, Environment 83 (1-2): 195–213. J. Kamari, K. Kallio, K. Eloheimo, and S. Kopo- Huffman, G. J., R. F. Adler, P. Arkin, A. Chang, R. Fer- nen. 2001. “Detection of Water Quality Using raro, A. Gruber, and U. Schneider. 1997. “The Simulated Satellite Data and Semi-Empirical Global Precipitation Climatology Project (GPCP) Algorithms in Finland.” Science of the Total Envi- Combined Precipitation Dataset.” Bulletin of the ronment 268 (1-3): 107–21. American Meteorological Society 78 (1): 5–20. Helfrich, S. R., D. McNamara, B. H. Ramsay, Huffman, G. J., R. F. Adler, D. T. Bolvin, G. J. Gu, Baldwin, and T. Kasheta. 2007. “Enhancements T. ­ E. J. Nelkin, K. P. Bowman, Y. Hong, E. F. Stocker, to, and Forthcoming Developments in, the Inter- and D. B. Wolff. 2007. “The TRMM Multisatellite active Multisensor Snow and Ice Mapping System Precipitation Analysis (TMPA): Quasi-Global, (IMS).” Hydrological Processes 21 (12): 1576–86. Multiyear, Combined-Sensor Precipitation Esti- Hess, L. 2003. “Dual-Season Mapping of Wetland mates at Fine Scales.” Journal of Hydrometeorol- Inundation and Vegetation for the Central Ama- ogy 8 (1): 38–55. zon Basin.” Remote Sensing of Environment 87 (4): Huffman, G. J., D. T. Bolvin, D. Braithwaite, K. Hsu, 404–28. R. Joyce, and P. Xie. 2013. NASA Global Precipi- Hestir, E. L., S. Khanna, M. E. Andrew, M. J. Santos, tation Measurement (GPM) Integrated Multi- J. H. Viers, J. A. Greenberg, S. S. Rajapakse, and SatellitE Retrievals for GPM (IMERG). Algorithm S. L. Ustin. 2008. “Identification of Invasive Veg- Theoretical Basis Document, Version 4.1. etation Using Hyperspectral Remote Sensing in Hunter, P. D., D. J. Gilvear, A. N. Tyler, N. J. Willby, and the California Delta Ecosystem.” Remote Sensing A. Kelly. 2010. “Mapping Macrophytic Vegetation in of Environment 112 (11): 4034–47. doi:10.1016/j Shallow Lakes Using the Compact Airborne Spec- .rse.2008.01.022. trographic Imager (CASI).” Aquatic Conservation: Hill, M. J., U. Senarath, A. Lee, M. Zeppel, J. M. Night- Marine and Freshwater Ecosystems 20 (7): 717–27. ingale, R. J. Williams, and T. R. McVicar. 2006. Ichii, K., H. Hashimoto, M. A. White, C. Potter, “Assessment of the MODIS LAI Product for Aus- L. R. Hutyra, A. R. Huete, R. B. Myneni, and tralian Ecosystems.” Remote Sensing of Environ- R. R. Nemani. 2006. “Constraining Rooting ment 101 (4): 495–518. Depths in Tropical Rainforests Using Satellite Hong, Y., R. F. Adler, F. Hossain, S. Curtis, and Data and Ecosystem Modeling for Accurate Simu- G. J. Huffman. 2007. “A First Approach to Global lation of Gross Primary Production Seasonality.” Runoff Simulation Using Satellite Rainfall Estima- Global Change Biology 13 (1): 67–77. tion.” Water Resources Research 43 (8): W08502. Jackson, T. J., and T. J. Schmugge. 1989. “Passive doi:10.1029/2006WR005739. Microwave Remote Sensing System for Soil Mois- Hong, Y., K.-L. Hsu, S. Sorooshian, and X. Gao. 2004. ture: Some Supporting Research.” IEEE Transac- “Precipitation Estimation from Remotely Sensed tions on Geoscience and Remote Sensing 27 (2): Imagery Using an Artificial Neural Network 225–35. Cloud Classification System.” Journal of Applied Jasechko, S., Z. D. Sharp, J. J. Gibson, S, J. Birks, Y. Yi, Meteorology 43 (12): 1834–52. and P. J. Fawcett. 2013. “Terrestrial Water Fluxes Hou, A.Y., R. K. Kakar, S. Neeck, A. A. Azarbarzin, Dominated by Transpiration.” Nature 496 (7445): C. D. Kummerow, M. Kojima, R. Oki, K. Naka- 347–50. mura, and T. Iguchi. 2013. “The Global Precipita- Jones, L. A., and J. S. Kimball. 2012. “Daily Global tion Measurement (GPM) Mission.” Bulletin of the Land Surface Parameters Derived from AMSR-E.” American Meteorological Society 95 (5): 701–22. National Aeronautics and Space Administration, chapter 6 : E arth O bservations for M onitoring W ater R esources   |  135 Distributed Active Archive Center, National Snow Passive Microwave Remote Sensing.” Geoscience and Ice Data Center, Boulder, CO. and Remote Sensing 49 (3): 949–60. doi:10.1109/ Joyce, R. J., J. E. Janowiak, P. A. Arkin, and P. P. Xie. TGRS.2010.2070515. 2004. “CMORPH: A Method That Produces Kimball, J. S., K. C. McDonald, and M. Zhao. 2006. Global Precipitation Estimates from Passive “Spring Thaw and Its Effect on Terrestrial Vegeta- Microwave and Infrared Data at High Spatial and tion Productivity in the Western Arctic Observed Temporal Resolution.” Journal of Hydrometeorol- from Satellite Microwave and Optical Remote ogy 5 (3): 487–503. Sensing.” Earth Interactions 10 (21): 1–22. Jung, M., M. Reichstein, P. Ciais, S. I. Seneviratne, King, E. A., T. G. Van Niel, A. I. J. M. Van Dijk, Z. Wang, J. Sheffield, M. L. Goulden, G. Bonan, A. Cescatti, M. J. Paget, T. Raupach, J. P. Guerschman, J. Chen, R. de Jeu, A. J. Dolman, W. Eugster, V. Haverd, T. R. McVicar, I. Miltenberg, M. R. Rau- D. Gerten, D. Gianelle, N. Gobron, J. Heinke, pach, L. J. Renzullo, and Y. Zhang. 2011. Actual J. Kimball, B. E. Law, L. Montagnani, Q. Mu, Evapotranspiration Estimates for Australia: Inter- B. Mueller, K. Oleson, D. Papale, A. D. Richardson, comparison and Evaluation. Water for a Healthy O. Roupsard, S. Running, E. Tomelleri, N. Viovy, Country National Research Flagship. Canberra: U. Weber, C. Williams, E. Wood, S. Zaehle, and Commonwealth Scientific and Industrial Research K. Zhang. 2010. “Recent Decline in the Global Organisation. Land Evapotranspiration Trend Due to Limited Kirk, J. T. O. 2011. Light and Photosynthesis in Aquatic Moisture Supply.” Nature 467 (7318): 951–54. Ecosystems. 3d ed. Cambridge, U.K.: Cambridge Jupp, D. L. B., G. L. Tian, T. R. McVicar, Y. Qin, and University Press. F. Li. 1998. Soil Moisture and Drought Monitoring Klemas, V. 2013. “Remote Sensing of Coastal Wet- Using Remote Sensing I: Theoretical Background land Biomass: An Overview.” Journal of Coastal and Methods. Technical Report 95. Canberra: Research 29 (5): 1016–28. CSIRO Earth Observation Centre. Koike, T. 2013. “Level 2 (Soil Moisture).” In Descrip- Kalma, J. D., T. R. McVicar, and M. F. McCabe. 2008. “Estimating Land Surface Evaporation: A Review tions of GCOM-W1 AMSR2 Level 1R and Level 2 of Methods Using Remotely Sensed Surface Algorithms, ch. 8. Japan Aerospace Exploration Temperature Data.” Surveys in Geophysics 29 (4): Agency Earth Observation Research Center. 421–69. http://suzaku.eorc.jaxa.jp/GCOM_W/data/doc/ NDX-120015A.pdf. Kasischke, E. S., and L. P. Bruhwiler. 2003. “Emis- sions of Carbon Dioxide, Carbon Monoxide, Krajewski, W. F. 1987. “Cokriging Radar-Rainfall and and Methane from Boreal Forest Fires in 1998.” Rain Gauge Data.” Journal of Geophysical Research Journal of Geophysical Research 107 (D1): FFR 92 (D8): 9571–80. 2-1–FFR 2-14. Kubota, T., S. Shige, H. Hashizume, K. Aonashi, Kelliher, F. M., R. Leuning, M. R. Raupach, and N. Takahashi, S. Seto, H. Masafumi, N. Yukari, E. D. Schulze. 1995. “Maximum Conductances for T. U. Takayabu, N. Katsuhiro, I. Koyura, K. Misako, Evaporation from Global Vegetation Types.” Agri- and O. Ken’ichi. 2007. “Global Precipitation Map cultural and Forest Meteorology 73 (1-2): 1–16. Using Satellite-Borne Microwave Radiometers by Kellndorfer, J., W. Walker, L. Pierce, C. Dobson, the GSMaP Project: Production and Validation.” J. A. Fites, C. Hunsaker, J. Vona, and M. Clutter. IEEE Transactions on Geoscience and Remote 2004. “Vegetation Height Estimation from Shuttle Sensing 45 (7): 2259–75. Radar Topography Mission and National Eleva- Kummerow, C., Y. Hong, W. S. Olson, S. Yang, R. F. Adler, tion Datasets.” Remote Sensing of Environment J. McCollum, R. Ferraro, G. Petty, D. B. Shin, and 93 (3): 339–58. T. T. Wilheit. 2001. “The Evolution of the Goddard Kelly, R. E., A. T. Chang, L. Tsang, and J. L. Foster. Profiling Algorithm (GPROF) for Rainfall Estima- 2003. “A Prototype AMSR-E Global Snow Area tion from Passive Microwave Sensors.” Journal of and Snow Depth Algorithm.” IEEE Transactions Applied Meteorology 40 (11): 1801–20. on Geoscience and Remote Sensing 41 (2): 230–42. Lambin, E. F., and D. Ehrlich. 1996. “The Surface Tem- Kerr, Y. H., S. Member, P. Waldteufel, P. Richaume, perature-Vegetation Index Space for Land Cover J. P. Wigneron, S. Member, and S. Delwart. 2012. and Land-Cover Change Analysis.” International “The SMOS Soil Moisture Retrieval Algorithm.” Journal of Remote Sensing 17 (3): 463–87. IEEE Transactions on Geoscience and Remote Leblanc, M. J., P. Tregoning, G. Ramillien, S. O. Tweed, Sensing 50 (5): 1384–403. and A. Fakes. 2009. “Basin-Scale, Integrated Kim, Y., J. S. Kimball, K. C. McDonald, and J. Glassy. Observations of the Early 21st Century Multiyear 2011. “Developing a Global Data Record of Daily Drought in Southeast Australia.” Water Resources Landscape Freeze/Thaw Status Using Satellite Research 45 (4): W04408. 136  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Lee, K., L. Carder, C. D. Mobley, R. G. Steward, and Martínez, J. M., and T. Le Toan. 2007. “Mapping of J. S. Patch. 1998. “Hyperspectral Remote Sensing Flood Dynamics and Spatial Distribution of Vege- for Shallow Waters: I. A Semi-Analytical Model.” tation in the Amazon Floodplain Using Multitem- Applied Optics 37 (27): 6329–38. poral SAR Data.” Remote Sensing of Environment Leroux, D. J., Y. H. Kerr, P. Richaume, and R. Fieuzal. 108 (3): 209–23. 2013. “Spatial Distribution and Possible Sources Matthews, M. W. 2011. “A Current Review of Empiri- of SMOS Errors at the Global Scale.” Remote cal Procedures of Remote Sensing in Inland and Sensing of Environment 133 (June 15): 240–50. Near-Coastal Transitional Waters.” International doi:10.1016/j.rse.2013.02.017. Journal of Remote Sensing 32 (21): 6855–99. Leuning, R., E. van Gorsel, W. J. Massman, and P. R. Mayaux, P., G. F. Rauste, Y. M. Simard, and S. Saatchi. Isaac. 2012. “Reflections on the Surface Energy 2002. “Large-Scale Vegetation Maps Derived Imbalance Problem.” Agricultural and Forest from the Combined L-Band GRFM and C-Band Meteorology 56 (April 15): 65–74. CAMP Wide Area Radar Mosaics of Central Leuning, R., Y. Q. Zhang, A. Rajaud, H. Cleugh, and Africa.” International Journal of Remote Sens- K. Tu. 2008. “A Simple Surface Conductance Model ing 23 (7): 1261–82. to Estimate Regional Evaporation Using MODIS McDonald, K. C., J. S. Kimball, and Y. Kim. 2010. “The Leaf Area Index and the Penman-Monteith Equa- Soil Moisture Active/Passive (SMAP) Freeze/ tion.” Water Resources Research 44 (10): W10419. Thaw Product: Providing a Crucial Linkage Li, L., P. W. Gaiser, B.-C. Gao, R. M. Bevilacqua, between Earth’s Water and Carbon Cycles.” In T. J. Jackson, E. G. Njoku, C. Rudiger, J.-C. Calvet, AGU Fall Meeting Abstracts, vol. 1, 1219. Washing- and R. Bindlish. 2010. “WindSat Global Soil Mois- ton, DC: American Geophysical Union. ture Retrieval and Validation.” IEEE Transactions McMahon, T. A., M. C. Peel, L. Lowe, R. Srikanthan, on Geoscience and Remote Sensing 48 (5): 2224–41. and T. R. McVicar. 2013. “Estimating Actual, doi: 10.1109/TGRS.2009.2037749. Potential, Reference Crop, and Pan Evaporation Liu, X., Y. Zhang, Y. Yin, M. Wang, and B. Qin. Using Standard Meteorological Data: A Pragmatic 2013. “Wind and Submerged Aquatic Vegeta- Synthesis.” Hydrology and Earth System Sciences tion Influence Bio-Optical Properties in Large 17 (April 10): 1331–63. Shallow Lake Taihu, China.” Journal of Geo- McVicar, T. R., and D. L. B. Jupp. 2002. “Using Covari- physical Research: Biogeosciences 118 (2): 713–27. ates to Spatially Interpolate Moisture Availability doi:10.1002/jgrg.20054. in the Murray-Darling Basin: A Novel Use of Liu, Y. Y., W. A. Dorigo, R. M. Parinussa, R. A. M. de Remotely Sensed Data.” Remote Sensing of Envi- Jeu, W. Wagner, M. F. McCabe, and A. I. J. M. van ronment 79 (2-3): 199–212. Dijk. 2012. “Trend-Preserving Blending of Passive McVicar, T. R., M. L. Roderick, R. J. Donohue, L. T. Li, and Active Microwave Soil Moisture Retrievals.” T. G. Van Niel, A. Thomas, J. Grieser, D. Jhajharia, Remote Sensing of Environment 123 (August): Y. Himri, N. M. Mahowald, A. V. Mescherskaya, 280–97. A. C. Kruger, S. Rehman, and Y. Dinpashoh. Liu, Y. Y., A. I. J. M. van Dijk, R. A. M. de Jeu, and 2012a. “Global Review and Synthesis of Trends in T. R. H. Holmes. 2009. “An Analysis of Spatiotem- Observed Terrestrial Near-Surface Wind Speeds: poral Variations of Soil and Vegetation Moisture Implications for Evaporation.” Journal of Hydrol- from a 29-Year Satellite-Derived Dataset over ogy 416-417 (January 24): 182–205. Mainland Australia.” Water Resources Research McVicar, T. R., M. L. Roderick, R. J. Donohue, and 45 (7): 1–12. T. G. Van Niel. 2012b. “Less Bluster Ahead? Eco- Lucas, R., J. Armston, R. Fairfax, R. Fensham, A. Accad, hydrological Implications of Global Trends of J. Carreiras, J. Kelley, P. Bunting, D. Clewley, Terrestrial Near-Surface Wind Speeds.” Ecohy- S. Bray, D. Metcalfe, J. Dwyer, M. Bowen, T. Eyre, drology 5 (4): 381–88. M. Laidlaw, and M. Shimada. 2010. “An Evalua- Meerhoff, M., N. Mazzeo, B. Moss, and L. Rodríguez- tion of the ALOS PALSAR L-Band Backscatter— Gallego. 2003. “The Structuring Role of Free- Above Ground Biomass Relationship Queensland, Floating Versus Submerged Plants in a Subtropical Australia: Impacts of Surface Moisture Condi- Shallow Lake.” Aquatic Ecology 37 (4): 377–91. tion and Vegetation Structure.” IEEE Journal of doi:10.1023/B:AECO.0000007041.57843.0b. Selected Topics in Applied Earth Observations and Meijerink, A. M. J., D. Bannert, O. Batelaan, Remote Sensing 3 (4): 576–93. M. W. Lubczyński, and T. Pointet. 2007. Remote Malthus, T. J., and D. G. George. 1997. “Airborne Sensing Applications to Groundwater. IHP-VI Remote Sensing of Macrophytes in Cefni Res- Series on Groundwater 16. Paris: UNESCO. ervoir, Anglesey, UK.” Aquatic Botany 58 (3-4): Menenti, M., and B. J. Choudhury. 1993. “Parameter- 317–32. ization of Land Surface Evaporation by Means chapter 6 : E arth O bservations for M onitoring W ater R esources   |  137 of Location Dependent Potential Evapora- Nagler, P. L., E. P. Glenn, H. Kim, W. Emmerich, transpiration and Surface Temperature Range.” R. L. Scott, T. E. Huxman, and A. R. Huete. 2007. In Exchange Processes at the Land Surface for “Relationship between Evapotranspiration and a Range of Space and Time Scales, edited by Precipitation Pulses in a Semiarid Rangeland H.-J. Bolle, R. A. Feddes, and J. D. Kalma, 561–68. Estimated by Moisture Flux Towers and MODIS Yokohama: International Association of Hydro- Vegetation Indices.” Journal of Arid Environments logical Sciences. 70 (3): 443–62. Mertes, L. A. K., D. L. Daniel, J. M. Melack, B. Nelson, Nagler, P., K. Morino, R. S. Murray, J. Osterberg, and L. A. Martinelli, and B. R. Forsberg. 1995. “Spatial E. Glenn. 2009. “An Empirical Algorithm for Esti- Patterns of Hydrology, Geomorphology, and Veg- mating Agricultural and Riparian Evapotranspira- etation on the Floodplain of the Amazon River in tion Using MODIS Enhanced Vegetation Index Brazil from a Remote Sensing Perspective.” Geo- and Ground Measurements of ET. I. Description morphology 13 (1-4): 215–32. of Method.” Remote Sensing 1 (4): 1273–97. Miralles, D. G., J. H. Gash, T. R. H. Holmes, R. A. M. Njoku, E. G., T. L. Jackson, V. Lakshmi, T. Chan, and de Jeu, and A. J. Dolman. 2010. “Global Canopy S. V. Nghiem. 2003. “Soil Moisture Retrieval from Interception from Satellite Observations.” Jour- AMSR-E.” IEEE Transactions on Geoscience and nal of Geophysical Research: Atmospheres 115 Remote Sensing 41 (2): 215–29. (D16122): 8. Norman, J. M., M. C. Anderson, W. P. Kustas, Mitra, A. K., A. K. Bohra, M. N. Rajeevan, and A. N. French, J. Mecikalski, R. Torn, G. R. Diak, T. N. Krishnamurti. 2009. “Daily Indian Precipita- T. J. Schmugge, and B. C. W. Tanner. 2003. tion Analysis Formed from a Merge of Rain-Gauge “Remote Sensing of Surface Energy Fluxes Data with the TRMM TMPA Satellite-Derived at 101-M Pixel Resolutions.” Water Resources Rainfall Estimates.” Journal of the Meteorological Research 39 (8): 1221. Society of Japan 87A: 265–79. Ochsner, T. E., M. H. Cosh, R. H. Cuenca, W. A. Dorigo, Monteith, J. L. 1965. “Evaporation and Environment.” C. S. Draper, Y. Hagimoto, Y. H. Kerr, K. M. Lar- Symposia of the Society for Experimental Biology son, Eni G. Njoku, E. E. Small, and M. Zreda. 19 (2): 205–34. 2013. “State of the Art in Large-Scale Soil ———. 1972. “Solar Radiation and Productivity in Tropi- Moisture Monitoring.” Soil Science Society of cal Ecosystems.” Journal of Applied Ecology 9 (3): America Journal 77 (6): 1888–919. doi:10.2136/ 747–66. sssaj2013.03.0093. ———. 1981. “Evaporation and Surface Temperature.” Odermatt, D., A. Gitelson, V. E. Brando, and Quarterly Journal of the Royal Meteorological Soci- M. Schaepman. 2012. “Review of Constituent ety 107 (451): 1–27. Retrieval in Optically Deep and Complex Waters from Satellite Imagery.” Remote Sensing of Envi- Morisette, J. T., F. Baret, J. L. Privette, R. B. Myneni, ronment 118 (3): 116–26. J. E. Nickeson, S. Garrigues, N. V. Shabanov, M. Weiss, R. A. Fernandes, S. G. Leblanc, Ol’dekop, E. M. 1911. “On Evaporation from the Surface M. Kalacska, G. A. Sanchez-Azofeifa, M. Chubey, of River Basins.” Transactions on Meteorological B. Rivard, P. Stenberg, M. Rautiainen, P. Voipio, Observations 4. T. Manninen, A. N. Pilant, T. E. Lewis, Olmanson, L. G., M. E. Bauer, and P. L. Brezonik. J. S. Iiames, R. Colombo, M. Meroni, L. Busetto, 2008. “A 20-Year Landsat Water Clarity Census W. B. Cohen, D. P. Turner, E. D. Warner, of Minnesota’s 10,000 Lakes.” Remote Sensing of G. W. Petersen, G. Seufert, and R. Cook. 2006. Environment 112 (11): 4086–97. doi:10.1016/ “Validation of Global Moderate-Resolution LAI j.rse.2007.12.013. Products: A Framework Proposed within the Olmanson, L. G., P. L. Brezonik, and M. E. Bauer. 2011. CEOS Land Product Validation Subgroup.” IEEE “Evaluation of Medium- to Low-Resolution Sat- Transactions on Geoscience and Remote Sensing ellite Imagery for Regional Lake Water Quality 44 (7): 1804–17. Assessments.” Water Resources Research 47 (9). Mu, Q., M. Zhao, J. S. Kimball, N. G. McDowell, and doi: 10.1029/2011WR011005. S. W. Running. 2012. “A Remotely Sensed Global Ordoyne, C., and M. A. Friedl. 2008. “Using MODIS Terrestrial Drought Severity Index.” Bulletin Data to Characterize Seasonal Inundation Pat- of the American Meteorological Society 94 (1): terns in the Florida Everglades.” Remote Sensing 83–98. of Environment 112 (11): 4107–19. Mu, Q., M. Zhao, and S. W. Running. 2011. “Improve- OSCAR (Observing Systems Capability Analysis and ments to a MODIS Global Terrestrial Evapo- Review Tool) (database). World Meteorological transpiration Algorithm.” Remote Sensing of Organization, Geneva, Switzerland. http://www Environment 115 (8): 1781–800. .wmo-sat.info/oscar/. 138  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Overton, I. C. 2005. “Modelling Floodplain Inundation Penman, H. L. 1948. “Natural Evaporation from Open on a Regulated River: Integrating GIS, Remote Water, Bare Soil, and Grass.” Proceedings of the Sensing and Hydrological Models.” River Research Royal Society of London. Series A, Mathematical and Applications 21 (9): 991–1001. and Physical Sciences 193 (1032): 120–45. Overton, I. C., T. M. Doody, D. Pollock, J. P. Guer- Pervez, M. S., and J. F. Brown. 2010. “Mapping Irri- schman, G. Warren, W. Jin, Y. Chen, and gated Lands at 250-m Scale by Merging MODIS B. Wurcker. 2011. The Murray–Darling Basin Data and National Agricultural Statistics.” Remote Floodplain Inundation Model (MDB-FIM). Tech- Sensing 2 (10): 2388–412. nical Report. Canberra: Commonwealth Scientific Pipunic, R. C., J. P. Walker, A. W. Western, and and Industrial Research Organisation. C. M. Trudinger. 2013. “Assimilation of Multiple Owe, M., R. A. M. de Jeu, and T. Holmes. 2008. “Multi- Data Types for Improved Heat Flux Prediction: A sensor Historical Climatology of Satellite-Derived One-Dimensional Field Study.” Remote Sensing of Global Land Surface Moisture.” Journal of Geo- Environment 136 (September): 315–29. physical Research 113 (F1): F01002. Powell, S. J., R. A. Letcher, and B. F. W. Croke. 2008. Ozdogan, M., Y. Yang, G. Allez, and C. Cervantes. 2010. “Modelling Floodplain Inundation for Environ- “Remote Sensing of Irrigated Agriculture: Oppor- mental Flows: Gwydir Wetlands, Australia.” Eco- tunities and Challenges.” Remote Sensing 2 (9): logical Modelling 211 (3–4): 350–62. 2274–304. Pozzi, W., J. Sheffield, R. Stefanski, D. Cripe, R. Pul- Painter, T. H., K. Rittger, C. McKenzie, P. Slaughter, warty, J. V. Vogt, and R. Lawford. 2013. “Toward R. E. Davis, and J. Dozier. 2009. “Retrieval of Sub- Global Drought Early Warning Capability: pixel Snow Covered Area, Grain Size, and Albedo Expanding International Cooperation for the from MODIS.” Remote Sensing of Environment 113 Development of a Framework for Monitoring and (4): 868–79. Forecasting.” Bulletin of the American Meteoro- Pan, M., H. Li, and E. Wood. 2010.  “Assessing the logical Society 94 (6): 776–85. Skill of Satellite-Based Precipitation Estimates Proctor, C., J. H. He, and V. Robinson. 2013. “Texture in Hydrological Applications.” Water Resources Augmented Detection of Macrophyte Species Research 46 (9): W09535. Using Decision Trees.” ISPRS Journal of Photo- Papa, F., C. Prigent, F. Aires, C. Jimenez, W. B. Rossow, grammetry and Remote Sensing 80 (June): 10–20. and E. Matthews. 2010. “Interannual Variabil- Pulliainen, J. 2006. “Mapping of Snow Water Equiva- ity of Surface Water Extent at the Global Scale, lent and Snow Depth in Boreal and Sub-Arctic 1993–2004.” Journal of Geophysical Research: Zones by Assimilating Space-Borne Microwave Atmospheres 115 (D12): L11401. Radiometer Data and Ground-Based Observations.” Parinussa, R. M., T. R. H. Holmes, and R. A. M. de Remote Sensing of Environment 101 (2): 257–69. Jeu. 2012. “Soil Moisture Retrievals from the Ramillien, G., J. S. Famiglietti, and J. Wahr. 2008. WindSat Spaceborne Polarimetric Microwave “Detection of Continental Hydrology and Glaciol- Radiometer.” IEEE Transactions on Geoscience ogy Signals from GRACE: A Review.” Surveys in and Remote Sensing 50 (7): 2683–94. doi:10.1109/ Geophysics 29 (4): 361–74. TGRS.2011.2174643. Ramsay, B. H. 1998. “The Interactive Multisensor Parinussa, R. M., G. Wang, T. R. H. Holmes, Y. Y. Liu, Snow and Ice Mapping System.” Hydrological Pro- A. J. Dolman, R. A. M. de Jeu, T. Jiang, P. Zhang, cesses 12 (10-11): 1537–46. and J. Shi. 2014. “Global Surface Soil Moisture Rango, A., and A. T. Anderson. 1974. “Flood Hazard from the Microwave Radiation Imager On Board Studies in the Mississippi River Basin Using Remote the Fengyun-3B Satellite.” International Journal of Sensing.” Water Resources Bulletin 10 (5): 1060–81. Remote Sensing 35 (19): 7007–29. doi:10.1080/0143 1161.2014.960622. Rango, A., and V. V. Salomonson. 1974. “Regional Flood Mapping from Space.” Water Resources Pauwels, V. R. N., R. Hoeben, N. E. C. Verhoest, F. P. De Research 10 (3): 473–84. Troch, and P. A. Troch. 2002. “Improvement of TOPLATS-Based Discharge Predictions through Rees, W. G. 2006. Remote Sensing of Snow and Ice. Assimilation of ERS-Based Remotely Sensed Soil Boca Raton, FL: Taylor and Francis. Moisture Values.” Hydrological Processes 16 (5): Reichle, R., and R. Koster. 2005. “Global Assimilation 995–1013. of Satellite Surface Soil Moisture Retrievals into Peña-Arancibia, J. L., A. I. J. M. Van Dijk, L. J. Ren- the NASA Catchment Land Model.” Geophysical zullo, and M. Mulligan. 2013. “Evaluation of Research Letters 32 (2): L02404. Precipitation Estimation Accuracy in Reanalyses, Renzullo, L., A. Chappell, T. Raupach, P. Dyce, M. Li, Satellite Products, and an Ensemble Method for and Q. Shao. 2011. An Assessment of Blended Sat- Regions in Australia and South and East Asia.” ellite-Gauge Precipitation Products for Australia. Journal of Hydrometeorology 14 (4): 1323–33. Water for a Healthy Country National Research chapter 6 : E arth O bservations for M onitoring W ater R esources   |  139 Flagship. Canberra: Commonwealth Scientific Santos, M., L. Anderson, and S. Ustin. 2011. “Effects and Industrial Research Organisation. of Invasive Species on Plant Communities: An Renzullo, L. J., A. I. J. M. Van Dijk, J.-M. Perraud, Example Using Submersed Aquatic Plants at the D. Collins, B. Henderson, H. Jin, A. B. Smith, and Regional Scale.” Biological Invasions 13 (2): 443– D. L. McJannet. 2014. “Continental Satellite Soil 57. doi:10.1007/s10530-010-9840-6. Moisture Data Assimilation Improves Root-Zone Santos, M. J., E. L. Hestir, S. Khanna, and S. L. Ustin. Moisture Analysis for Water Resources Assess- 2012. “Image Spectroscopy and Stable Isotopes ment.” Journal of Hydrology 519 (D): 2747–62 Elucidate Functional Dissimilarity between Rodell, M., I. Velicogna, and J. S. Famiglietti. 2009. Native and Nonnative Plant Species in the Aquatic “Satellite-Based Estimates of Groundwater Deple- Environment.” The New Phytologist 193 (3): 683– tion in India.” Nature 460 (August 20): 999–1002. 95. doi:10.1111/j.1469-8137.2011.03955.x. Roderick, M. L., M. T. Hobbins, and G. D. Farquhar. Sapiano, M. R. P., J. E. Hanoiak, W. Shi, R. W. Hig- 2009a. “Pan Evaporation Trends and the Terres- gins, and V. B. S. Silva. 2010. “Regional Evaluation trial Water Balance. II. Energy Balance and Inter- through Independent Precipitation Measure- pretation.” Geography Compass 3 (2): 761–80. ments: USA.” In Satellite Rainfall Applications for —. 2009b. “Pan Evaporation Trends and the Terrestrial Surface Hydrology, edited by M. Gebremichael Water Balance. I. Principles and Observations.” and F. Hossain, 169–204. Dordrecht: Springer Geography Compass 3 (2): 746–60. Science+Business Media. Rodríguez, E., and D. Estéban-Fernández. 2010. “The Sapiano, M. R. P., T. M. Smith, and P. A. Arkin. Surface Water and Ocean Topography Mission 2008. “A New Merged Analysis of Precipitation (SWOT): The Ka-Band Radar Interferometer ­ Utilizing Satellite and Reanalysis Data.” Journal of (KaRIn) for Water Level Measurements at All ­Geophysical Research 113 (D22): 1–17. Scales.” Proceedings of SPIE—The International Scipal, K., T. Holmes, R. A. M. de Jeu, V. Naeimi, Society for Optical Engineering 7826: 782614–18. and W. Wagner. 2008. “A Possible Solu- Rodríguez-Iturbe, I., A. Porporato, F. Laio, and tion for the Problem of Estimating the Error L. Ridolfi. 2001. “Plants in Water-Controlled Eco- Structure of Global Soil Moisture Data systems: Active Role in Hydrologic Processes and Sets.” Geophysical Research Letters 35 (24). Response to Water Stress; I. Scope and General DOI: 10.1029/2008GL035599. Outline.” Advances in Water Resources 24 (7): Sellars, S., P. Nguyen, W. Chu, X. Gao, K. Hsu, and 695–705. S. Sorooshian. 2013. “Computational Earth Sci- Roerink, G. J., Z. Su, and M. Menenti. 2000. “S-SEBI: ence: Big Data Transformed into Insight.” EOS A Simple Remote Sensing Algorithm to Estimate Transactions American Geophysical Union 94 (32): the Surface Energy Balance.” Physics and Chem- 277–78. istry of the Earth, Part B: Hydrology, Oceans, and Seneviratne, S. I., T. Corti, E. L. Davin, M. Hirschi, Atmosphere 25 (2): 147–57. E. B. Jaeger, I. Lehner, B. Orlowski, and Rosenqvist, Å., M. Shimada, B. Chapman, L. Dutra, A. J. Teuling. 2010. “Investigating Soil Moisture- S. Saatchi, and O. Tanaka. 2002. “Introduction Climate Interactions in a Changing Climate: A from the Guest Editors.” International Journal of Review.” Earth-Science Reviews 99 (3-4): 125–61. Remote Sensing 23 (7): 1215. Sheffield, J., G. Goteti, and E. F. Wood. 2006. “Develop- Rozante, J. R., D. S. Moreira, L. G. G. de Goncalves, and ment of a 50-Year High-Resolution Global Dataset D. A. Vila. 2010.  “Combining TRMM and Surface of Meteorological Forcings for Land Surface Mod- Observations of Precipitation: Techniques and eling.” Journal of Climate 19 (13): 3088–111. Validation over South America.” Weather and Siebert, S., J. Burke, J. M. Faures, K. Frenken, Forecasting 25 (3): 885–94. J. Hoogeveen, P. Döll, and F. T. Portmann. 2010. Sakamoto, T., N. Van Nguyen, A. Kotera, H. Ohno, “Groundwater Use for Irrigation—A Global Inven- N. Ishitsuka, and M. Yokozawa. 2007. “Detect- tory.” Hydrology and Earth System Sciences, Dis- ing Temporal Changes in the Extent of Annual cussions 7 (3): 3977–4021. Flooding within the Cambodia and the Vietnam- Silva, T. F., M. F. Costa, J. Melack, and E. L. M. Novo. ese Mekong Delta from MODIS Time-Series 2008. “Remote Sensing of Aquatic Vegetation: Imagery.” Remote Sensing of Environment 109 (3): Theory and Applications.” Environmental Moni- 295–313. toring and Assessment 140 (1-3): 131–45. Salomonson, V. V., and I. Appel. 2004. “Estimating Simard, M., N. Pinto, J. B. Fisher, and A. Bac- Fractional Snow Cover from MODIS Using the cini. 2011. “Mapping Forest Canopy Height Normalized Difference Snow Index.” Remote Globally with Spaceborne Lidar.” Journal of Geo- Sensing of Environment 89 (3): 351–60. physical Research 116 (G4): 4021. 140  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Smith, L. C., and D. E. Alsdorf. 1998. “Control on Sedi- Tapley, B. D., S. Bettadpur, J. C. Ries, P. F. Thompson, ment and Organic Carbon Delivery to the Arctic and M. M. Watkins. 2004. “GRACE Measure- Ocean Revealed with Space-Borne Synthetic ments of Mass Variability in the Earth System.” Aperture Radar: Ob’ River, Siberia.” Geology Science 305 (5683): 503–05. 26 (5): 395–98. Taylor, C. M., R. A. A. de Jeu, F. Guichard, P. P. Har- Smith, A. B., J. P. Walker, A. W. Western, R. I. Young, ris, and W. A. Dorigo. 2012. “Afternoon Rain K. M. Ellett, R. C. Pipunic, R. B. Grayson, L. Siri- More Likely over Drier Soils.” Nature 489 (7416): wardena, F. H. S. Chiew, and H. Richter. 2012. 423–26. “The Murrumbidgee Soil Moisture Monitoring Tian, Y. Q., Q. Yu, M. J. Zimmerman, S. Flint, and Network Data Set.” Water Resources Research M. C. Waldron. 2010. “Differentiating Aquatic 48 (7): 1–6. doi:10.1029/2012WR011976. Plant Communities in a Eutrophic River Using Sobrino, J. A., M. Gomez, J. C. Jimenez-Munoz, Hyperspectral and Multispectral Remote Sens- A. ­ Olioso, and G. Chehbouni. 2005. “A Simple ing.” Freshwater Biology 55 (8): 1658–73. Algorithm to Estimate Evapotranspiration from Tucker, C. J., J. E. Pinzon, M. E. Brown, D. A. Slayback, DAIS Data: Application to the DAISEX Cam- E. W. Pak, R. Mahoney, E. F. Vermote, and N. El paigns.” Journal of Hydrology 315 (1-4): 117–25. Saleous. 2005. “An Extended AVHRR 8-Km NDVI Sorooshian, S., K. L. Hsu, X. Gao, H. V. Gupta, B. Imam, Dataset Compatible with MODIS and SPOT and D. Braithwaite. 2000.  “Evaluation of PERSI- Vegetation NDVI Data.” International Journal of ANN System Satellite-Based Estimates of Tropical Remote Sensing 26 (20): 4485–98. Rainfall.” Bulletin of the American Meteorological Tyler, A. N., E. Svab, T. Preston, M. Présing, and Society 81 (9): 2035–46. A. Kovács. 2006. “Remote Sensing of the Water Specht, R. L. 1972. “Water Use by Perennial Evergreen Quality of Shallow Lakes: A Mixture Modelling Plant Communities in Australia and Papua New Approach to Quantifying Phytoplankton in Water Guinea.” Australian Journal of Botany 20 (3): Characterized by High-Suspended Sediment.” 273–99. International Journal of Remote Sensing 27 (8): 152. Srebotnjak, T., G. Carr, A. de Sherbinin, and C. Rick- Ustin, S. L., D. A. Roberts, J. A. Gamon, G. P. Asner, and wood. 2012. “A Global Water Quality Index and R. O. Green. 2004. “Using Imaging Spectroscopy Hot-Deck Imputation of Missing Data.” Ecological to Study Ecosystem Processes and Properties.” Indicators 17 (June): 108–19. BioScience 54 (6): 523. doi:10.1641/0006-3568 Su, Z. 2002. “The Surface Energy Balance System (2004)054[0523:UISTSE]2.0.CO;2. (SEBS) for Estimation of Turbulent Heat Fluxes.” Vanderstukken, M., S. A. J. Declerck, E. Decaestecker, Hydrology and Earth System Sciences 6 (1): 85–100. and K. Muylaert. 2014. “Long-Term Allelopathic Su, Z., P. de Rosnay, J. Wen, L. Wang, and Y. Zeng. 2013. Control of Phytoplankton by the Submerged “Evaluation of ECMWF’s Soil Moisture Analyses Macrophyte Elodea Nuttallii.” Freshwater Biology Using Observations on the Tibetan Plateau.” Jour- 59 (5): 930–41. doi:10.1111/fwb.12316. nal of Geophysical Research: Atmospheres 118 (11): Van Dijk, A. I. J. M. 2010. The Australian Water 5304–18. Resources Assessment System. Technical Report Su, Z., J. Wen, L. Dente, R. van der Velde, L. Wang, 3. Landscape Model (version 5.0) Techni- Y. Ma, K. Yang, and Z. Hu. 2011. “The Tibetan Pla- cal Description. Water for a Healthy Country teau Observatory of Plateau Scale Soil Moisture National Research Flagship. Canberra: CSIRO. and Soil Temperature (Tibet-Obs) for Quantifying Van Dijk, A. I. J. M., and L. J. Renzullo. 2011. “Water Uncertainties in Coarse Resolution Satellite and Resource Monitoring Systems and the Role of Sat- Model Products.” Hydrology and Earth System Sci- ellite Observations.” Hydrology and Earth System ences 15 (7): 2303–316. Sciences 15: 39–55. Syed, T. H., J. S. Famiglietti, M. Rodell, J. Chen, and Van Dijk, A. I. J. M., L. J. Renzullo, Y. Wada, and C. R. Wilson. 2008. “Analysis of Terrestrial Water P. Tregoning. 2013. “A Global Water Cycle Reanal- Storage Changes from GRACE and GLDAS.” ysis (2003–2012) Reconciling Satellite Gravimetry Water Resources Research 44 (2): W02433. and Altimetry Observations with a Hydrological Takala, M., K. Luojus, J. Pulliainen, C. Derksen, Model Ensemble.” Hydrology and Earth System J. Lemmetyinen, J. P. Karna, J. Koskinen, and Sciences Discussions 10: 15475–523. B. Bojkov. 2011. “Estimating Northern Hemi- Van Niel, T. G., and T. R. McVicar. 2004. “Current and sphere Snow Water Equivalent for Climate Potential Uses of Optical Remote Sensing in Rice- Research through Assimilation of Space-Borne Based Irrigation Systems: A Review.” Australian Radiometer Data and Ground-Based Measure- Journal of Agricultural Research 55 (2): 155–85. ments.” Remote Sensing of Environment 115 (12): Van Niel, T. G., T. R. McVicar, M. L. Roderick, 3517–29. A. I. J. M. van Dijk, J. Beringer, L. B. Hutley, and chapter 6 : E arth O bservations for M onitoring W ater R esources   |  141 E. van Gorsel. 2012. “Upscaling Latent Heat Flux Wang, J. R., and T. J. Schmugge. 1980. “An Empirical for Thermal Remote Sensing Studies: Compari- Model for the Complex Dielectric Permittivity son of Alternative Approaches and Correction of of Soils as a Function of Water Content.” IEEE Bias.” Journal of Hydrology 468-469 (October 25): Transactions on Geoscience and Remote Sens- 35–46. ing GE-18 (4): 288–95. Van Niel, T. G., T. R. McVicar, M. L. Roderick, Wegmuller, U., and C. Werner. 1997. “Retrieval of Veg- A. I. J. M. van Dijk, L. J. Renzullo, and E. van etation Parameters with SAR Interferometry.” Gorsel. 2011. “Correcting for Systematic Error IEEE Transactions on Geoscience and Remote in Satellite-Derived Latent Heat Flux due to Sensing 35 (1): 18–24. Assumptions in Temporal Scaling: Assessment WIRADA (Water Information Research and Develop- from Flux Tower Observations.” Journal of ment Alliance). 2012. Water Information Research Hydrology 409 (1-2): 140–48. and Development Alliance: Science Symposium Verrelst, J., G. W. Geerling, K. V. Sykora, and Proceedings, Melbourne, Australia, August 1–5, J. G. P. W. Clevers. 2009. “Mapping of Aggregated 2011. Water for a Healthy Country National Floodplain Plant Communities Using Image Research Flagship. Canberra: CSIRO. Fusion of CASI and Lidar Data.” International Wu, C., Z. Niu, Q. Tang, W. Huang, B. Rivard, and Journal of Applied Earth Observation and Geoin- J. Feng. 2009. “Remote Estimation of Gross Pri- formation 11 (1): 83–94. mary Production in Wheat Using Chlorophyll- Vila, D. A., L. G. G. de Gonçalves, D. L. Toll, and Related Vegetation Indices.” Agricultural and J. R. Rozante. 2009. “Statistical Evaluation of Forest Meteorology 149 (6-7): 1015–21. Combined Daily Gauge Observations and Rainfall Wu, H., R. F. Adler, Y. Hong, Y. Tian, and F. Policelli. Satellite Estimates over Continental South Amer- 2012. “Evaluation of Global Flood Detection ica.” Journal of Hydrometeorology 10 (2): 533–43. Using Satellite-Based Rainfall and a Hydrologic Wada, Y., L. P. H. van Beek, and M. F. P. Bierkens. 2012. Model.” Journal of Hydrometeorology 13 (4): “Nonsustainable Groundwater Sustaining Irri- 1268–84. gation: A Global Assessment.” Water Resources Xiong, A.-Y., P.-P. Xie, J.-Y. Liang, Y. Shen, M. Chen, Research 48 (6): 2055. R. J. Joyce, J. E. Janowiak, and P. A. Arkin. 2008. Wada, Y., L. P. H. van Beek, C. M. van Kem- “Merging Gauge Observations and Satellite Esti- pen, J. W. T. M. Reckman, S. Vasak, and mates of Daily Precipitation over China.” In Pro- M. F. P. Bierkens. 2010. “Global Depletion of ceedings of the 4th Workshop of the International Groundwater Resources.” Geophysical Research Precipitation Working Group (IPWG), Beijing, Letters 37 (20): L20402. China, October 13–17. Bologna: Institute of Atmo- Wagner, W., G. Lemoine, and H. Rott. 1999. “A Method spheric Sciences and Technology. http://www for Estimating Soil Moisture from ERS Scatterom- .isac.cnr.it/~ipwg/meetings/beijing- 2008/. eter and Soil Data.” Remote Sensing of Environ- Yebra, M., A. I. J. M. Van Dijk, R. Leuning, ment 70 (2): 191–207. A. R. Huete, and J. P. Guerschman. 2013. “Evalu- Wagner, W., C. Pathe, M. Doubkova, D. Sabel, ation of Optical Remote Sensing to Estimate A. Bartsch, S. Hasenauer, G. Blöschl, K. Sci- Actual Evapotranspiration and Canopy Conduc- pal, J. Martínez-Fernández, and A. Löw. 2008. tance.” Remote Sensing of Environment 129 “Temporal Stability of Soil Moisture and Radar (February 15): 250–61. Backscatter Observed by the Advanced Synthetic Yilmaz, M. T., and W. T. Crow. 2014. “Evaluation of Aperture Radar (ASAR).” Sensors 8 (February): Assumptions in Soil Moisture Triple Collocation 1174–97. Analysis.” Journal of Hydrometeorology 15 (3): Wagner, W., K. Scipal, C. Pathe, D. Gerten, W. Lucht, 1293–302. doi:10.1175/JHM-D-13-0158.1. and B. Rudolf. 2003. “Evaluation of the Agree- Zhang, K., J. S. Kimball, R. R. Nemani, and ment between the First Global Remotely Sensed S. W. Running. 2010. “A Continuous Satellite- Soil Moisture Data with Model and Precipitation Derived Global Record of Land Surface Evapo- Data.” Journal of Geophysical Research 108 (D19): transpiration from 1983 to 2006.” Water Resources 4611. Research 46 (9): W09522. Wang, J.-J., X. X. Lu, S. C. Liew, and Y. Zhou. 2009. Zhang, Y. Q., F. H. S. Chiew, L. Zhang, R. Leuning, “Retrieval of Suspended Sediment Concentra- and H. A. Cleugh. 2008. “Estimating Catch- tions in Large Turbid Rivers Using Landsat ment Evaporation and Runoff Using MODIS ETM+: An Example from the Yangtze River, Leaf Area Index and the Penman-Monteith China.” Earth Surface Processes and Landforms 34 Equation.” Water Resources Research 44 (10). (8): 1082–92. DOI: 10.1029/2007WR006563 142  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Zhang, Y., R. Leuning, F. H. S. Chiew, E. Wang, Organic Matter in Complex Freshwater Envi- L. Zhang, C. Liu, F. Sun, M. C. Peel, Y. Shen, and ronments.” Remote Sensing of Environment 140 M. Jung. 2012. “Decadal Trends in Evaporation January): 766–78. doi:10.1016/j.rse.2013.10.015. (­ from Global Energy and Water Balances.” Journal Zreda, M., W. J. Shuttleworth, X. Zeng, C. Zweck, of Hydrometeorology 13 (February): 379–91. D. Desilets, T. Franz, and R. Rosolem. 2012. “COS- Zhou, L., and G. Zhou. 2009. “Measurement and MOS: The COsmic-Ray Soil Moisture Observing Modelling of Evapotranspiration over a Reed System.” Hydrology and Earth System Sciences (Phragmites australis) Marsh in Northeast 16 (11): 4079–99. China.” Journal of Hydrology 372 (1-4): 41–47. Zwieback, S., K. Scipal, W. Dorigo, and W. Wagner. doi:10.1016/j.jhydrol.2009.03.033. 2012. “Structural and Statistical Properties of the Zhu, W., Q. Yu, Y. Q. Tian, B. L. Becker, T. Zheng, and Collocation Technique for Error Characteriza- H. J. Carrick. 2014. “An Assessment of Remote tion.” Nonlinear Processes in Geophysics 19 (1): Sensing Algorithms for Colored Dissolved 69–80. chapter 6 : E arth O bservations for M onitoring W ater R esources   |  143 CHAPTER 7 Assessing the Characteristics of Required and Available Earth Observation Data INTRODUCTION of EO technologies and techniques is either not suitable or less apparent and needs to be aug- For many potential applications, Earth obser- mented with hydrologic modeling or field- vation (EO) data products will be immediately based metering to be useful. and obviously useful for improving water This chapter provides guidelines to help resources management (WRM) and water project leaders to decide whether Earth monitoring. Examples include using evapo- observations may be useful and, if so, what transpiration (ET) to assess crop water use in the most suitable data sources to consider irrigated regions, chlorophyll estimation to would be. For each water resources applica- monitor water quality in water bodies that tion area, issues related to accuracy, avail- provide domestic water supply, and satellite ability, maturity, complexity, and reliability rainfall to estimate the amount and duration of are briefly discussed. The chapter aims to rainfall in ungauged regions. provide a simple framework to help decision In these examples, as in many others, care- makers to determine, for a given WRM issue, ful consideration of the spatial resolution, how EO products might best be employed to temporal frequency, data latency, and longev- generate the required information and how ity of the satellite systems is needed to select to select the EO data products with the most the most appropriate EO product from the appropriate characteristics or specifications. often wide range of products available, taking The focus lies on what questions to ask once into account the specific WRM problem to be it has been concluded that exploring EO addressed. Furthermore, there are as many, if options for the WRM problem at hand is not more, WRM issues where the application worthwhile.   145 EO-based solutions are not always applica- potentially be useful. If it is decided that they ble. For this reason, the chapter begins by pro- could be useful, certain questions must be asked viding some precursor questions meant to regarding the data characteristics. The chapter clarify whether EO data products could begins by summarizing these questions and Figure 7.1    Decision tree Questions to ask Chapter and tables Guidelines for What WRM questions need to be answered? See chapter 5 and tables 5.1, Define the nature of the Determining WRM problem What are the policy or regulatory drivers? 5.2, and 5.3 (WRM issues and Whether to Use   Who are the stakeholders and beneficiaries? EO variables). EO Products Note: EO = Earth Is metering available? observation; What is the condition of the data networks? WRM = water Define the status of What are the impediments to sharing, resources management. existing observation collating, and archiving the data? networks What has been done in the past? Any monitoring? Modeling? Justification Will complement ground-based monitoring networks or serve as the sole information Do you need source? No to use Earth observation Will it be used in conjunction with data? modeling? Are the EO data stream(s) suitable for Do not long-term WRM decision support? Yes use EO Can EO See tables 5.1, 5.2, and 5.3. Suitability potentially provide the See chapter 6 and tables 6.3 What variables can EO provide? No required data? through 6.12. Are data products readily available? Examples in appendix B. Yes Spatial resolution What is the appropriate pixel size? Temporal frequency Sensor dependent. See tables How frequent do these observations need 6.3 through 6.12. to be? Record length How far back in time does your data record need to go? Determine minimum In situ data requirement required data How much in situ data are used in data characteristics product? Reliability What is the certainty associated with the supply of that product across space and through time? Product specific. See tables 7.2 Accuracy through 7.14. What is the uncertainty associated with the data estimates? Maturity How established is the data product? Complexity What level of complication is involved in the process of converting the EO data into the data product? Can the EO Use EO product meet the data No Yes product(s) requirements? 146  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T then putting them in the context of their suit- BOX 7.1 ability for WRM decision support. It then sum- marizes the suitability of the EO data products Guiding Questions to Aid in the Decision Whether described in chapter 6, guided by the questions to Use Earth Observation for Water Resources outlined in this chapter. The information con- Management tained in this chapter and the key questions to 1. Nature of the problem ask are summarized in the flowchart presented • What WRM questions need to be answered? in figure 7.1. • What are the policy or regulatory drivers of these questions? • Who are the stakeholders and beneficiaries of a solution to the WRM problem? ESTABLISHING THE ROLE OF EARTH OBSERVATION TO SUPPORT WRM 2. Existing data and observation networks DECISION MAKING • What metering is currently available? • What is the condition of the data networks? What Questions Need to Be Answered? • Are there any impediments to sharing, collating, and archiving the Earth observation cannot provide an appropri- data (such as transboundary issues)? ate solution for all WRM problems. For this • What, if anything, has been done in the past to address the issues reason, box 7.1 poses some questions that can be at hand? asked to determine the nature of the WRM • Any monitoring? Modeling? problem under investigation and the sources of • Can Earth observation fill an information gap? information essential to the overall decision- • Will it complement field-monitoring networks or serve as the making process. Points 1 and 2 of box 7.1 are dis- sole source of information? cussed in greater detail in the following • Will it be used in conjunction with modeling? paragraphs. The questions listed under point 3 • Are the EO data stream(s) suitable for long-term WRM decision are related not to Earth observation but to insti- support? tutional goals and capability; though important, 3. Sustaining and maintaining WRM decision support and monitoring they are not addressed in this publication. programs • Is there capability to adopt a solution in the short and longer Nature of the WRM Problem term? It is important to begin by asking, what WRM • What key national organizations and international experts could questions need to be answered? Outlining the be potential partners regarding Earth observation? specific set of WRM questions is critical to • What is the local capability to adopt new techniques and tech- establishing the scope of the investigation, nologies? including the geographic extent, amount of • What computing infrastructure, if any, is needed? Is it available, investment (What can be feasibly achieved?), and who owns it? and expectations for monitoring and reporting • To what degree will local expertise require training in new techniques and technologies? programs beyond a project of fixed duration. Government policy and regulatory drivers • What level of national versus international resourcing will be ­ required? related to the broad WRM issues to be addressed probably exist, perhaps in response to an environmental crisis. The higher-level WRM statements and policy will need to be place another constraint on the solutions that translated to very specific information and are feasible. For example, if the goal is to reporting requirements. The characteristics of respond rapidly to an impending natural disas- key stakeholders and their information ter, this will eliminate some of the EO products requirements and communication options may with greater latency. chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  147 Existing Data and Observation Networks production system. This includes aspects such It is important to assess any existing (past or as the mission lifetime of the satellite, the options current) field observation networks and data. in case of mission failure, any redundancy in data Sometimes a better solution may be to install a streams through other sensors, service-level network of field-based sensors, upgrade exist- agreements, and ongoing provision of a stable ing networks, or rehabilitate former gauging data product with known and unchanging (or networks. Of course, such a solution has impor- perhaps improved) characteristics. Of course, tant implications that need to be considered, these aspects only need to be considered if “live” for example, regarding spatial coverage and information systems are to be developed, for cost-efficiency (for example, capital costs, instance, for long-term d­ ecision-making support ongoing maintenance, and data management for water resources management and consistent and processing). Alternatively, the current net- monitoring over time. work might be appropriate in principle, but sharing measurements across organizational or jurisdictional boundaries—within or between DESCRIBING THE nations—may be challenging. Conducting a CHARACTERISTICS OF EO DATA thorough analysis of the status and access to PRODUCTS existing observation networks and data is a valuable first step toward identifying any To identify which data products might be requirements for Earth observation to fill data suitable for a given application, numerous gaps (see chapter 4 for additional information). data characteristics can be assessed that relate Although there are exceptions, EO data to spatial and temporal attributes, accuracy, products are usually not directly suitable for and reliability. Eight core data characteristics addressing WRM problems unless augmented are useful for establishing the suitability of with additional data. Field measurements, data for a predetermined task: spatial resolu- where and when they are available, are critical tion, frequency and timing, record length, in the development of any EO data product and field data requirements, data product reliabil- in the assessment of accuracy and uncertainty ity, data product accuracy, data product matu- (often called “validation,” the subject of the rity, and data product complexity (table 7.1). chapters in part III). They are also highly valu- Three of these characteristics are sensor spe- able for correcting (or “calibrating”) the prod- cific in that they depend on the satellite sen- uct in regional applications or for enhancing sor from which they are derived, whereas the the resolution (for example, through statistical downscaling). Table 7.1  Major Characteristics of Data Products and Even more valuable, but also more compli- Their Type of Dependence cated, is to use EO data along with field data DATA PRODUCT and hydrologic computer models and with CHARACTERISTIC DEPENDENCE knowledge of the errors in each of these to Spatial resolution Sensor specific constrain the hydrologic estimation or predic- Temporal frequency Sensor specific tion (see chapters 4 and 11 for additional infor- Record length Sensor specific mation). Such a model-data fusion approach Field data requirements Product specific makes the EO data more directly relevant and Reliability Product specific valuable to the WRM variables of interest. Accuracy Product specific Finally, the prospects for ongoing data collec- Maturity Product specific tion and access need to be considered if they are Complexity Product specific to form a component of an operational data 148  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T five other characteristics are product specific dates. Often, however, satellite sensors acquire in that they depend more on how the data data long past their mission end dates (for product was generated than on their sensor example, Landsat-5). Generally speaking, the source. These eight characteristics are defined longer the record length, the older the satellite in this section. and its associated technology. Thus there is usually a trade-off between record length and Spatial Resolution other data attributes, such as accuracy, spatial Sometimes called the spatial frequency or resolution, and number of spectral bands. image resolution, the spatial resolution of EO data refers to the pixel size of the image. Spa- Field Data Requirements tial resolution of data determines the precision Some EO data products are generated solely with which the spatial variation of the observed from satellite observations, and some are gen- phenomenon can be captured. A relatively erated from a combination of satellite and large pixel size will capture less spatial varia- field-based observations. The latter have much tion than a relatively small pixel size. higher input data requirements and so are When a pixel straddles two (or more) dis- more dependent on the availability of suitable tinct ground features (such as a water body and field data. They are also generally more com- adjacent vegetation), the pixel captures a mix- plex to generate and can become limited to the ture of the signals from both features and is specific locations and times that the field data referred to as a “mixel.” Mixels make image represent: interpretation more difficult, and the larger the • Low field-based (data) requirement prod- pixel size the more mixels the data are likely to ucts do not use field data or use it only for contain. validation purposes. Temporal Resolution • Medium field-based (data) requirement The temporal resolution of data (or the tempo- products need field data to calibrate the EO ral frequency) refers to how often a sensor data or use a moderate amount of field data makes observations of a given location. In the to derive the final EO product itself (such case of polar-orbiting satellites, frequency is as when river gauge data are combined related to overpass frequency and is typically with satellite-derived flood extent data to measured in days. The frequency of geostation- estimate flood volumes). ary satellites is much higher, being measured in • High field-based (data) requirement prod- minutes to hours. Relatively high-frequency ucts incorporate multiple sources of field observations are able to capture the dynamics data (for example, most ET and soil mois- in fast-changing processes better than rela- ture data products), sometimes in complex tively low-frequency observations. data assimilation systems. For some applications, the time of observa- tion can be important to ensure that the obser- vations occur at the same time each day or at Reliability specific times of the day, such as at noon. Data product reliability refers to the certainty of supply of that product across space and Record Length through time. The greater the spatial cover- The record length refers to how long the record age, the more frequently the product is of data is. This is typically a function of the updated; the greater the number of options for period of operation of the satellite and so is sourcing the product, the higher the reliability determined by the mission launch and end of the product. chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  149 • Low reliability describes a product that well-established science and can be gauged by is tailor-made for a specific time, region, ­ doption: its level of validation, acceptance, and a or application or is generated by only one • Low maturity indicates that the product is organization. still in an experimental stage. • Medium reliability describes a product that • Medium maturity indicates that the prod- typically has wide (global) coverage and is uct is developmental in that the underlying frequently updated but comes from only science is mature but the product’s conver- one source organization. sion to being operational is still in progress. • High reliability describes a product with • High maturity refers to a proven—widely global coverage that is frequently updated tested and adopted—operational product. and can be sourced from multiple indepen- dent organizations. Complexity Accuracy Data product complexity describes the level of The accuracy of data products is an estimation complication involved in converting the EO- of the uncertainty associated with the data esti- processed data into the data product. Com- mates. Accuracy can be described in absolute plexity is a function of, for example, the terms (that is, in physical units such as number of methodological steps involved, ­ millimeters per year) or in relative terms (usu- the number and type of input data sources, the ally as a percentage). For example, if an estimate level of mathematics involved, the volume of of evaporation of 200 millimeters per year has data to be processed, and the technical exper- an error of 10 percent (therefore, an accuracy of tise required: 90 percent), the real value could be as low as • Low complexity indicates no function or 180 or as high as 220 millimeters per year. a very simple function for converting pro- Rarely will the accuracy of an EO data prod- cessed satellite data into the data product, uct be as high as that of an equivalent field mea- requiring basic technical expertise. surement. Despite a generally lower accuracy, EO products can still be an important data • Medium complexity indicates a moderately source, as EO imagery can provide information complex method. with greater spatial extent, spatial density, or • High complexity indicates a highly com- temporal frequency than most field-based plex method for generating the data (point-based) observation networks. For this product, requiring advanced technical or reason, the combination of EO and field data computational expertise. generally provides the best information out- comes. Part III provides additional information about validation of EO estimates of precipita- tion, evapotranspiration, soil moisture, snow DETERMINING THE cover and snow water equivalent, surface water CHARACTERISTICS OF MINIMUM levels and streamflows, and streamflow out- REQUIRED EO DATA puts from models using EO inputs. By analyzing the information required to Maturity address a specific issue at hand, it should be pos- Product maturity relates to how established sible to translate these requirements into the data product is. A well-established, or minimum required data characteristics or spec- ­ mature, product is generally founded on ifications that can be used to assess the 150  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 7.2  Guiding Questions for Determining the BOX 7.2 Minimum Requirements of EO Data Products CHARACTERISTIC GUIDING QUESTIONS Screening for Adequacy of Field Observations Justification Do you need to use EO data? • Are the data well described—that is, is it clear what was measured, Suitability Can EO provide the required how, where, and when? data products? • Are the right variables measured? Spatial resolution What is the appropriate pixel size? • Are the data of sufficient spatial density across your area of interest? Temporal frequency How frequent do these • Do the data cover the period of interest? observations need to be? • Are the measurements frequent enough? Record length How far back in time does your • Are the measurements available throughout the time period, with- data record need to go? out important gaps? Reliability Do you need guaranteed • Are the data of known and suitable accuracy? continuation of data supply into the future? • Are the data guaranteed to be free from bias and manipulation? Accuracy What degree of accuracy is • Are the data available in digital form and in an interpretable needed in the data products? format? ­ Maturity Do you want to use only data • Are the data publicly available or is it clear they will be made avail- products that are commonly able by their custodian? used? • Do the data have to continue being collected into the future? Note: EO = Earth observation. suitability of EO products. This section des­ Chapter 5, especially tables 5.1–5.3, dis- cribes those minimum requirements. Table  7.2 cusses the water-related issues for which Earth presents a list of questions for determining the ­ observation may be useful, and chapter 6 pro- requirements. vides information about each of the EO appli- Sometimes the answer to a question might cation areas relevant to water resources not be obvious. For instance, there may be management and about existing and future EO open questions about the overall approach to systems (tables 6.3–6.12). the WRM issue at hand or there may be a potential degree of circularity between what Suitability: Can EO Provide the Required EO data are available and what the required Information? characteristics are. Under such circumstances, Not all variables and processes can be mea- consultation with an EO area expert is likely to sured with Earth observation, whether directly be beneficial. or through inference using a model. For this reason, a key question to ask is whether the Justification: Do You Need to Use EO Data? required data or information products can be When field observations are accessible and generated from remote sensing at all. Table 6.2 sufficiently informative, it may well be possible in chapter 6 provides an overview of the most to answer WRM questions directly, without commonly derived products suitable for WRM using EO data. Relevant questions to consider applications. are listed in box 7.2. If the required data product is not readily If the answer to any of these questions is available, it may still be possible to derive such negative, it may be worth exploring the poten- a product from existing processed data, but tial usefulness of EO data products, either by this is likely to require engaging EO expertise. themselves or, more typically, in conjunction In that case, it would need to be determined if with field observations and computer models. and how a desired data product might be chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  151 derived from processed data. Determining Over what time scales does the phenomenon suitability requires a deeper understanding of of interest vary, or how long does it take for the the characteristics of individual satellite sen- phenomenon to vary significantly when con- sors and the relationship between the observa- sidering the intended purpose? As a bare mini- tion and the variable of interest. Consultation mum, satellite observations should be available with an EO expert should quickly settle at least at the same frequency as the variation whether there is any such prospect. in the phenomenon of interest. For short-lived events, the exact timing of observation is also Spatial Resolution: What Is the   likely to be important. If the dynamics of a pro- Appropriate Pixel Size? cess or event are important, a frequency sub- Spatial resolution is an important and almost stantially less than the duration of the event universal characteristic of EO data. In each will be necessary. application it will be necessary to consider the For instance, seasonal flooding may last for a minimal distances over which the phenome- month. However, if you are interested in peak non of interest (precipitation, soil moisture, or flood extent and that peak only lasts a day, a sin- water quality) varies or distance over which gle daily satellite measurement will suffice to any variations in the phenomenon would capture it, but the timing of the measurement become significant for the purpose at hand. As will be critical and may be difficult to achieve. a rule of thumb, the “pixel size” (the character- Alternatively, if the advance and recession of the istic length of one image pixel as measured on flood are of interest, regular (weekly or even the Earth’s surface) should be no more than a daily) measurements will be required. Another quarter of the length over which the phenome- example is an algal bloom in a lake that lasts a non varies and preferably finer. For example, if few days and can only be detected by a satellite the phenomenon of interest is total crop evapo- sensor that has high frequency or that can be ration from fields that are typically about 600 x pointed at an area of interest and therefore tar- 600 meters in size, the pixel size of the ET get a specific area. The latter, of course, requires product should be no larger than 150 meters. that there be sufficient time between knowledge Alternatively, if the variations in evaporation of the event and acquisition of the imagery, and within that field are of interest, a resolution on it usually comes at a cost. Conversely, if the sea- the order of 10 meters might be required. sonal pattern of algal levels across a year is of Tables 5.2 and 5.3 in chapter 5 provide interest, as few as four images may be sufficient. details on the spatial resolution of the main EO The satellite data product tables in chap- data products suited to water resources man- ter  6 (tables 6.3 through 6.12) provide details agement. The pixel size listed generally reflects on the revisit times of the main EO data prod- the smallest pixel size of the sensor from which ucts suited to water resources management. the data are derived. Some practitioners may be inclined to use Record Length: How Far Back Does the the highest resolution, but the cost of doing so Data Record Need to Go? can sometimes be very high and the value Analysis of changes (trends or shifts) in the added may not be worth the cost. This caveat behavior of a system will require a record that is also applies to temporal frequency. sufficiently long to establish such changes with confidence. Similarly, accurate estimation of the Temporal Resolution: How Frequent Do mean and variance of a particular variable will the Observations Need to Be? require a sufficiently long record for calculation. The temporal resolution needs to suit the The question as to what length of record is suffi- nature of the question asked—for example, cient for these purposes cannot be answered, but 152  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T some record length is required. As a general rule, BOX 7.3 a minimum of 15  years of observations is often required before trends in natural phenomena can Examples of Accuracy Parameters be analyzed properly (see chapter 4 and table • Absolute units (20 gigaliters per day streamflow) 4A.1 in annex 4A, available online at https:// • Relative units (10 percent of estimated water use) openknowledge.worldbank.org/ handle • Temporal correlation (for detecting trends or anomalies, the cor- /10986/22952, for additional information); the relation with ground-based measurements or independent data) World Meteorological Organization defines “cli- • Spatial classification (kappa statistics or confusion matrix for mate” as pertaining to a period of at least 30 years. mapping irrigated area) ­ Alternatively, the application may provide • Detection of extremes (detection scores for peak rainfall or flood near-real-time information and therefore per- extent) haps only the most recent period (for example, a day or month) is of interest, although often such information will need to be considered in a historical context. will remain available into the future with little Finally, the particular application may only or no interruptions? Or are the data required need data for a very specific period of interest, for a one-off, project-based study, with no such as the 2013 growing season, one drought- ­ follow-on monitoring being anticipated? This flood cycle, or a water year. This reduces the will determine whether it is possible to use demands on record length. only data derived from stable, likely long-term The satellite data product tables in ­ chapter 6 satellite missions with a track record of reli- (tables 6.3 through 6.12) describe the record ability or whether it is possible to include data lengths (launch and end dates, if applicable) of from short-term or experimental missions (the the main EO data products suited to water majority of EO satellites even today). It may be resources management. important to ascertain whether there is a long- term plan to ensure that satellite sensors and Accuracy: How Good Do the   the data stream will be available into the future. Data Have to Be? This is particularly important when deciding What is the acceptable tolerance of error in the whether to invest in the infrastructure data product for the purpose at hand? There quired for operational satellite imagery and re­ are many possible ways in which to express ­ geographic information system processing, accuracy. These will depend on the character- ­ perhaps including information validation pro- istics and intended use of the data product. grams and Web-based data services or other Examples are given in box 7.3. forms of information products. Conventional ways of validating EO data However, officially available information typically focus on the first two aspects—­ about mission continuity should only be used precisely the standard type of information that as general guidance: a current continuation generally is provided. However, it can be very policy may be changed in future, whereas a challenging to obtain accurate information on mission that currently has no official prospect other aspects, sometimes even from experts on of continuation may be replaced by a compa- the particular data source. rable sensor with identical, similar, or even better characteristics in future. Arguably the Reliability: Is Continued Supply of Data most reliable test of the risk to investment is into the Future Essential? redundancy: if several missions make the same This is an important question to answer. Is or quite similar observations, the associated assurance required that the EO data source risks are usually correspondingly lower. chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  153 The satellite data product tables in chapter 6 unchangeable decision-support system. Know- (tables 6.3 through 6.12) provide some details ing which government department or agency on mission reliability for the main EO data is ultimately likely to be responsible for main- products used in water resources management. taining any ongoing monitoring program and reporting the information helps to assess the Maturity: Can Data Products Be Limited to available capacity and preparedness to adopt Well-Established Products? EO-based solutions. Maturity refers to the degree to which an EO An EO solution may also require input from data product has been evaluated by the other national or international agencies research or management community. With (through the provision of observations or data maturity comes a better understanding of the products). Identifying and securing such key accuracy and suitability of the product for spe- partnerships up-front may be critical to suc- cific purposes and some pedigree and accep- cess. To evaluate the resources available and tance where its use has been successful. required, the following aspects of information Restricting the type of data products used to technology may be worth considering: those that are well established and in common • Infrastructure for acquiring the data (via use reduces the risk of nondelivery and disap- the Internet) pointment. Operational products—those that are readily available and have been widely • Storage of the data and backup facilities adopted across the WRM community—are • Implementation and maintenance of the generally restricted to mature products, or WRM system conversely, maturity comes with increased adoption across the community. Obviously, these infrastructure aspects also However, interested parties may be willing have implications for the human resources to use emerging (experimental or developmen- required to maintain and use them. In addition, tal) products because they provide information area expertise will be required on an ongoing that is otherwise not available and be willing to basis to interpret and report the information. accept some degree of uncertainty related to Training may be needed, as well as ongoing user product accuracy, suitability, and future avail- support in the transition from the research (or ability. In such cases, undertaking a pilot or development) environment to the operational case study may be worth considering before implementation of the solution. These aspects attempting to implement an operational data all depend on the complexity of the solution. service. If this is done in communication with the research and management community, such projects in themselves can rapidly achieve DETERMINING THE GENERALIZED greater maturity and acceptance. CHARACTERISTICS OF EO DATA PRODUCTS Complexity: What Data Management and Analysis Capacity Is Available? Now that the core characteristics have been Prior to pursuing an EO-based solution, it is suggested for describing the suitability of EO probably beneficial to establish who will be data products for water resources manage- responsible for running the WRM decision ment and guidelines have been provided for support or monitoring program and to evalu- determining what data characteristics are ate their mandate, resources, and capabilities. required for the application at hand, this sec- Ongoing “live” monitoring systems will be tion outlines the characteristics of available EO more demanding to maintain than an data products. Following the same format as 154  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T table 5.2 in chapter 5 and the discussion in range of options may be available for any given chapter 6, where EO-derived data products are type of data product, which may represent deci- summarized in eight broad types of infor­ sions about, for example, the trade-off between mation, this section contains eight tables resolution and accuracy (averaging over larger presenting the characteristics, respectively, of ­ areas or longer periods helps to increase the precipitation, evapotranspiration, soil mois- signal-to-noise ratio) or the ability to process, ture, vegetation and vegetation cover types, manage, and download the product (Internet groundwater, surface water, snow, and water speeds quickly become a bottleneck in using EO quality data products. data). Therefore, some broad generalizations Some of these core characteristics are are made in the following tables. For more pre- related directly to the satellite sensor from cise assessments, the documentation of individ- which the data are derived and are discussed in ual products will need to be referred to, and chapter 5. These sensor-specific characteristics product experts may need to be consulted. are spatial resolution, frequency, and record length. Assessments of the remaining five char- Precipitation acteristics are summarized in the tables below, Table 6.3 in chapter 6 provides an overview of while accuracy is discussed in part III. the range and characteristics of precipitation Beyond the constraints of the sensor obser- products derived from EO data and their rela- vations used, the characteristics of derived data tion to alternative sources of precipitation products also depend on the choices made in the data. The accuracy of the different products process of generating the data product. A wide varies with the season. Table 7.3 provides Table 7.3  Field Data Requirements and Characteristics of EO-Based Precipitation Products FIELD DATA COMMENTS ON PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS Rain gauge High Low Bias between ±0.2 millimeter High Medium Global gauge analyzes analysis day-1 (that is, as high as 60% for coarse spatial resolution; some regions); considered to be daily to monthly estimates; benchmark; accuracy decreases local or continental away from gauge location analyses typically about 1–10 kilometers; accuracy decreases with distance from gauge location Radar rainfall High Low Bias between ±0.5 millimeter High Medium Beam blockage day-1 (30–40% accuracy); (topography effects) subhourly rain rates; coverage hampers quality of limited, and estimates uncertain estimate; higher resolution at distance from radar (in space and time) compared to satellite products, but patchy coverage for large-area applications TIR, Low High Bias between ±2 millimeters High High Based on weak relationship geostationary day-1 (often greater than 100% between cloud top error), best at estimating small temperatures and rain rate; convective rainfall systems generally considered poorer-quality estimate than PMW; low latency (that is, real-time products possible) (Continued) chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  155 Table 7.3 (Continued) FIELD DATA COMMENTS ON PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS PMW, Low High Bias between ±1.5 millimeters High High High-quality retrievals, but polar orbiting day-1 (about 100% error); better much coarser resolution than models at estimating than TIR; difficulty in convective rainfall systems over capturing orographic or warmer months; patchy light rainfall; requires coverage multiple PMW platforms for more complete coverage and needs to be calibrated with in-orbit precipitation radar Merged Medium Medium Bias between ±1 millimeter High High Coarse resolution but TIR-PMW day-1; better at estimating greater coverage than convective rainfall systems over PMW alone; subdaily and warmer months; global near-real-time estimation coverage possible Merge High High Bias reduced to between ±0.75 High High Coarse resolution; could TIR-PMW millimeter day-1 on average be downscaled further gauge (often less than 100% error); with additional gauge data performance as with merged or analyses; greater data TIR-PMW; long latency latency Model Low High Bias between ±1 millimeter High High Coarse spatial resolution reanalysis day-1; better than PMW at estimating stratiform rainfall systems typical of cooler months Merged model, High Medium Bias between ±0.5 millimeter High High Coarse spatial resolution; satellite and day-1 (often less than 100% requires access to multiple gauge analysis error); coarse resolution data from multiple agencies Note: EO = Earth observation; TIR = thermal infrared; PMW = passive microwave. summary estimates of the bias (the average dif- more detailed, quantitative analyses, including ference between product and gauge observa- the categorical statistics (probability of detec- tions), taken from International Precipitation tion, false alarm ratio). Chapter 9 discusses the Working Group (IPWG) validation pages.1 validation of precipitation estimates derived The values reported on the IPWG valida- from remote sensing (RS). tion pages and in table 7.3 are regional aver- ages. At the aggregate level, the errors can Evapotranspiration sometimes be of the same magnitude as the For an overview of the generation of actual ET- rainfall value itself or much higher (even related data products from EO data, satellite greater than 100 percent relative error). The sensors suitable for generating such products, quality of satellite (and indeed of numerical and sensor-specific data characteristics, see weather prediction and gauge-based rainfall) chapter 6, specifically table 6.4. In chapter 6, estimates varies with geographic location, and three broad classes of actual ET estimation it is recommended that persons interested in approaches are defined that make use of satellite precipitation products consult the remote sensing: empirical, PM (Penman-­ IPWG websites or the literature (Ebert, Janow- Monteith) leaf area index (LAI), and resistance iak, and Kidd 2007; Sapiano et al. 2010) for energy balance model. Empirical methods seek 156  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T to define statistical relationships between For some specifics about reliability, accuracy, commonly observed EO data or products, usu- maturity, and complexity, see the references in ally either vegetation indexes or surface tem- the section on evapotranspiration in chapter 6. perature. The PM LAI approach uses the Chapter 9 discusses the validation of RS- Penman-Monteith “combination equation” derived ET estimates. and EO-based vegetation characteristics (usu- ally LAI) to model surface conductance. Soil Moisture Energy balance approaches mostly use EO- For an overview of the range of soil moisture based land surface temperature to estimate products generated from EO data, the data char- sensible heat flux, which can then be used acteristics, and their use, see chapter 6 and along with an estimate of the available energy table 6.6. Metrics of interest are given in table 7.5. to approximate latent heat flux. The absolute accuracy of satellite soil moisture Obviously, the metrics described in table 7.4 products is very rarely (if ever) of interest. For will depend on the specifics of the algorithms example, preprocessing of the data eliminates sys- and characteristics of the data sets used. An tematic differences between model estimates and attempt is made here to summarize actual ET observations prior to assimilation. The soil mois- products that are operationally available at ture estimates derived from Earth observation either global or continental scale on a monthly only represent the first centimeters of the surface. or shorter time step, which usually relates to Moreover, most applications (for example, in about 5-kilometer or finer spatial resolution. drought monitoring) require knowledge of soil Table 7.4  Field Data Requirements and Characteristics of EO-Based Evapotranspiration Products FIELD DATA COMMENTS ON PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS Empirical High Medium Usually better than Medium Low Limited ability to be 1 millimeter per day; most improved via better reliable when and where the process understanding; actual ET is dominated by the usually requires field EO metric from which the calibration, which may statistical relationship was only be regionally defined; for example, an applicable empirical relationship between actual ET and a moisture index would work best under water-limited conditions PM LAI Medium High Usually better than High Medium Limitations when ET is 1 millimeter per day; generally not dominated by reliable for places and times transpiration (that is, when transpiration is the main open water or soil source of actual ET; has been evaporation); accurate implemented operationally estimation of spatially over the globe and temporally varying conductance is difficult Resistance Medium High Usually better than High High Usually requires scaling energy 1 millimeter per day; generally instantaneous estimates balance reliable for estimating to daily or longer time model instantaneous flux and thus steps; may suffer from eminently suited for use with over-parameterization geostationary data Note: EO = Earth observation; ET = evapotranspiration; PM LAI = Penman-Monteith leaf area index. chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  157 Table 7.5  Field Data Requirements and Characteristics of EO-Based Soil Moisture Products FIELD DATA COMMENTS ON PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS Active Low Medium Higher spatial resolution, High Medium Higher spatial resolution hampered by noise; accuracy than PMW but can be affected in areas of highly significantly noisier; variable terrain terrain effects PMW Low High Generally considered more High High Low spatial resolution; accurate than data from active affected by dense systems; poor performance vegetation and biased in over areas of dense vegetation the vicinity of coast or open-water bodies Combined Low Medium Merged data, either through Medium High Requires multiple active–PMW joint assimilation or statistical sensors; SMAP is only combination, better than mission (planned) where individual products alone; a satellite will have both well-known complementarity active and PMW sensors of the two sources on one platform Assimilated Low Medium Assimilating surface model Low High Only way to get into land products into land surface root-zone moisture; surface models has been shown to however, it is still models improve root-zone moisture experimental (SMAP estimation by 30–80% provides root-zone moisture product by assimilating satellite surface models into land surface models) Note: EO = Earth observation; PMW = passive microwave; SMAP = Soil Moisture Active Passive. moisture relative to some threshold or historical Vegetation and Vegetation Cover frequency. Therefore, correlation is a better stan- For an overview of the generation of vegetation dard metric for evaluating soil moisture products. and vegetation cover–related data products For the products listed in table 6.6, values range from EO data, satellite sensors suitable for gen- from 0.6 to 0.9, depending on the product and erating such products, and sensor-specific data where it is evaluated. ­Chapter 9 discusses the vali- characteristics, see chapter 6 and table 6.7. dation of RS-derived soil moisture estimates. There are four vegetation-related products, as Table 7.6  Field Data Requirements and Characteristics of EO-Based Vegetation and Vegetation Cover Products FIELD DATA COMMENTS ON PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS Albedo Low (validation only) High 5–10% Medium Low NDVI Low (validation only) High Usually 5–10% Medium Low Relationship to real-world values is sensor specific LAI Low (validation only) High Decreases as LAI Medium Medium Performs best over increases; cannot low-density canopies (that detect change above is, LAI lower than 3–4) values around 10 fPAR Low (validation only) High 5–10% Medium Low Note: EO = Earth observation; NDVI = normalized difference vegetation index; LAI = leaf area index; fPAR = fraction of absorbed photosynthetically active radiation. 158  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T shown in table 7.6: albedo, the normalized dif- about reliability, accuracy, maturity, and com- ference vegetation index (NDVI), the leaf area plexity, see the references cited in the section index, and the fraction of photosynthetically on groundwater in chapter 6. active radiation absorbed by green leaves (fPAR). Surface Water A fifth product is vegetation cover, which is a For an overview of the generation of surface qualitative classification of vegetation based on water–related data products from EO data, the broad structural, climatic, or functional charac- satellite sensors suitable for generating such teristics. Because vegetation cover data are so products, and the sensor-specific data charac- varied in what they represent and how they are teristics, see chapter 6 and table 6.10. In derived, they are not included in table 7.6. table 7.8, both reservoir area and flood extent refer to the delimitation of the area covered Groundwater with standing water, although they differ in For an overview of the generation of ground- size and temporal dynamics: reservoir area can water-related data products from EO data and range from a few square meters (as in the case satellite sensors suitable for generating such of small ponds) to large lakes of several thou- products, see chapter 6 and table 6.9. The main sands of square kilometers and generally EO approaches for estimating groundwater are change in area and volume relatively slowly in through satellite gravity field mapping (gra- time. In general terms, floods are more dynamic vimetry) and radar interferometry. The former in time than reservoirs and can change in area measures changes in the regional gravity field, and volume in a matter of hours or days. Chap- while the latter measures changes in land sur- ter  9 discusses the validation of RS-derived face elevation (table 7.7). For some specifics surface water estimates. Table 7.7  Field Data Requirements and Characteristics of EO-Based Groundwater Products FIELD DATA COMMENTS ON PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS Gravity Low Medium-low (GRACE Suitable for very large Medium High Spatial resolution for field is operating seven areas; nominal precision obtaining a reliable years beyond its is 1 gravity value signal (about intended lifetime; a 400 kilometers) is follow-on mission is limited to very large planned for 2017) basins; may not be suitable in areas of tectonic rebound Surface High Low (no product per Varying accuracy Medium High Requires the height se, requires depending on relationship between interpretation for interpreter’s skill and vertical surface each instance) understanding of movement and regional geology, groundwater storage groundwater systems, to be quantified; and surface conditions changes in vertical surface movement are limited to regional interpretation, requiring a specialist Note: EO = Earth observation; GRACE = Gravity Recovery and Climate Experiment. chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  159 Table 7.8  Field Data Requirements and Characteristics of EO-Based Surface Water Products FIELD DATA COMMENTS ON INDICATOR REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS Reservoir area Low (validation only) Medium High (kappa greater than 90%) High Medium No global, continuously in classifications updated product Flood extent Low (mapping Medium High (kappa greater than 90%) High Medium One global product; flooded areas) to in most situations; kappa flooded area limited by medium (estimating 50–90% possible when water cloud cover; optical river discharge) is obscured by vegetation methods give poor (flooded forests) results in flooded forests Water level Low (validation only) High Altimetry accuracy dependent High Low Limited to large on sensor and size of water reservoirs only body; can be from about 10 centimeters to 50 centimetersa Note: EO = Earth observation. a. A good pragmatic source of information about the accuracy of altimetry data can be found at http://www.pecad.fas.usda.gov/cropexplorer/global_reservoir/validation.htm. Snow chapter  6) are listed again in box 7.4 for easy For an overview of the generation of snow- reference. related data products from EO data, satellite The additional criteria that can be used to sensors suitable for generating such products, determine whether EO is appropriate for a and sensor-specific data characteristics, refer particular water quality application are sum- to chapter 6 and table 6.11. Metrics related to marized in this section. The information is pre- snow cover, snow water equivalent, and snow sented in tables 7.10–7.12 covering empirical, moisture are given in table 7.9. Chapter 9 dis- semi-empirical, and physics-based inversion cusses the validation of RS-derived estimates methods, respectively. of snow cover and snow water equivalents. Table 7.10 presents empirical methods (where a statistical relationship is established Water Quality between the spectral bands used and the field- The six water quality variables that can be based measurement of the variable, without determined directly from EO data (see necessarily being a causal relationship). This Table 7.9  Field Data Requirements and Characteristics of EO-Based Snow Products FIELD DATA COMMENTS ON PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY LIMITATIONS Snow extent Low Medium 10–20% error Relatively Medium Affected by cloud or fraction of mature cover snow cover Snow water Medium Medium 20–30% error in flat areas; Mature for Medium for Terrain is a major equivalent very large in mountainous flat areas; flat areas; determinant of product areas low for complex for quality; also affected by mountainous mountainous prior knowledge of areas areas snow properties such as density, particle size, and shape Snow High Low Low Very low Complex moisture Note: EO = Earth observation. 160  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T Table 7.10  Field Data Requirements and Characteristics of EO-Based Water Quality Products: Empirical Methods FIELD DATA PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY COMMENTS ON LIMITATIONS CHL High Medium 50–70% High Low Empirical methods are only valid for field-based ranges; are not transportable to other water bodies; may provide spurious results; simultaneous acquisition of the in situ measurement during overpass of satellite is an absolute requirement to establish the empirical relationship CYP High Medium 40–60% Medium Medium Same as previous CDOM High Low 40–60% Low High Same as previous TSM High High 80% High Low Same as previous Kd High High 80% Medium Medium Same as previous Turb/SD High High 70–80% Medium Low Same as previous Source: Matthews 2011. Note: EO = Earth observation; CHL = chlorophyll; CYP = cyanobacterial pigments; CDOM = colored dissolved organic matter; TSM = total suspended matter; Kd = vertical attenuation of light coefficient; Turb/SD = turbidity/Secchi disk transparency. ­ BOX 7.4 medium suitability for automation across large areas. Water Quality Variables Directly   Table 7.12 refers to physics-based inversion Determined by Earth Observation methods (also known as semi-analytical inver- sion methods): all variables are assessed simul- Directly assessed: taneously in one spectral inversion. This • Chlorophyll method provides physics-based consistency of • Cyanobacterial pigments results and is most suitable for automation • Colored dissolved organic matter across large areas. • Total suspended matter Chapter 6 lists the water quality variables Indirectly assessed: that EO can provide, and table 6.12 provides an • Vertical attenuation of light coefficient inventory of EO satellites with their capabili- • Turbidity/Secchi disk transparency ties and suitability. Whether it is possible to process the EO data to retrieve quantitative water quality information will depend on the availability and method is the least suitable for automation quality of field data with which to calibrate the across large areas unless accompanied by a sig- relationships. The quality will depend on nificant, ongoing field measurement activity whether the field data cover all variables of across most water bodies present. concern (see chapter 6) and whether these Table 7.11 refers to semi-empirical methods coincide closely with the times of satellite (where a causal relationship is established overpasses. between the spectral bands used and the vari- Without any field data, empirical methods able assessed). This method is less prone to (where an empirical relationship is estab- providing spurious results, although results lished between field data and EO image pixel may have significantly higher errors outside values) will not work for quantitative assess- the field-based range. This method has ments. However, it may be possible to apply chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  161 Table 7.11  Field Data Requirements and Characteristics of EO-Based Water Quality Products: Semi-Empirical Methods FIELD DATA PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY COMMENTS ON LIMITATIONS CHL Medium High 60–80% High Low Semi-empirical methods may be extrapolated beyond field-based ranges, although nonlinear effects do occur; may be transportable to other similar water bodies; reduced requirement for field measurement simultaneous with satellite overpass; requires good atmospheric and water surface glint correction for time series assessments CYP Medium Medium 50–70% Medium Medium Same as previous CDOM Medium Medium 50–70% Medium High In waters with high organic particulate matter and high algal contents, the CDOM absorption signal is masked TSM Medium High 80% High Low Same as CHL and CYP Kd Medium High 80% Medium Medium Same as previous Turb/SD Medium High 70–80% Medium Medium Same as previous Sources: Matthews 2011; Odermatt et al. 2012. Note: EO = Earth observation; CHL = chlorophyll; CYP = cyanobacterial pigments; CDOM = colored dissolved organic matter; TSM = total suspended matter; Kd = vertical attenuation of light coefficient; Turb/SD = Turbidity/Secchi disk transparency. ­ Table 7.12  Field Data Requirements and Characteristics of EO-Based Water Quality Products: Physics-Based Inversion Methods FIELD DATA PRODUCT REQUIREMENTS RELIABILITY ACCURACY MATURITY COMPLEXITY COMMENTS ON LIMITATIONS CHL Low High 60–80% Medium High High complexity: requires good atmospheric and water surface glint correction CYP Low Medium 55–75% Low High Same as previous CDOM Low Medium 70% Medium High Same as previous TSM Low High 80% Medium High Same as previous Kd Low High 85% Medium High Same as previous Turb/SD Low High 80% Medium High Same as previous Source: Odermatt et al. 2012. Note: EO = Earth observation; CHL = chlorophyll; CYP = cyanobacterial pigments; CDOM = colored dissolved organic matter; TSM = total suspended matter; Kd = vertical attenuation of light coefficient; Turb/SD = turbidity/Secchi disk transparency. ­ spectral band indexes derived from the litera- the parameterizations available globally. Some ture to indicate relative measures of water generic approaches to assessing water quality quality. An alternative is to apply physics- are becoming available from the National based inversion techniques, which will Aeronautics and Space Agency (NASA) and require a high level of expertise but can be the European Space Agency (ESA), for the automated more easily in the long run. The coarse spatial resolution MODIS (Moderate parameterization of the physics-based inver- Resolution Imaging Spectrometer) and sion model will need to be based on expert MERIS (Medium Resolution Imaging Spec- assessment of the types of water present and trometer) sensors.2 162  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T If appropriate field data are available in improve the water quality of a large lake so that some or all of the water bodies of interest, it it may once again be in the condition to (a) pro- may be possible to develop and apply empirical vide water suitable for drinking (untreated for methods and extrapolate these relationships to livestock but treated for human consumption), other nearby water bodies for which field data (b) support a local, small-scale freshwater fish- do not exist. It is essential that the range of ing industry, (c) maintain biodiversity, and field measurements (concentrations of sus- (d)  allow recreational use. The lake has pended matter, chlorophyll, and others) be suf- recently degraded from a mesotrophic system ficiently large to be representative; it is not to a hypertrophic system prone to algal blooms. useful to apply empirical algorithms of clear The lake displays potentially harmful algal glacier-fed lakes to turbid downstream rivers blooms in spring, summer, and fall, which are or lakes prone to algal blooms, for example. probably caused by eutrophication as a result With physics-based inversion methods, it of direct and diffuse sources of agricultural and may be possible to refine the parameterization horticultural use of land around the lake as of the waters that do have field data and to well as diffuse and point sources of untreated, extrapolate these values downstream or to primary sewage water. Field measurements of nearby water bodies. However, although electrical conductivity and alkalinity or acidity physics-­based methods are less prone to error are the only recently available water quality than empirical relationships, they do need a data. Anecdotal evidence suggests that the lake suitable parameterization of initial values. has transitioned from a clear productive lake With access to a sufficient quantity of rele- 10 years ago to a turbid, algal bloom–­dominated vant field data for the water bodies of concern, lake today. the question of whether to use Earth observa- At this stage, the first questions to ask are as tion becomes more relevant: Are the field data follows: Do you want to do EO-based retro- sufficient (regarding time range, frequency, spective assessment of your water system to spatial representativeness, timeliness) to understand its evolution through time or do inform water quality management decision you need EO measurements now to inform you making on their own? The following questions of the water quality situation today and into should be asked at this stage: How far back do the future? Table 7.13 lists a further set of the field data archives go with respect to the appropriate questions for informing the deci- EO archives (see table 6.12)? Are all of the vari- sion about whether to proceed on the basis of ables required available in the field data or not? EO data. Can Earth observation provide the extra vari- This example illustrates how the informa- ables that are required? Are real-time data tion contained in the tables in chapters 5 and 6 products required? may be used together with the guiding ques- tions presented in this chapter to determine the most suitable EO sensor and method for WORKED EXAMPLES assessing the development of water quality over the last 10 years for a lake undergoing To illustrate how the decision process might environmental change. proceed in practice, this section describes two Two real-world examples in the research fictional case studies. literature provide some in-depth information on how MODIS and MERIS were used to Improving Lake Water Quality assess similar conditions in lakes in China You are project manager with responsibility (Hu et al. 2010) and South Africa (Matthews, for developing a program that will help to Stewart, and Lisl 2012), respectively. chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  163 Table 7.13  Guiding Questions for Determining the Characteristics of Required EO Data Products: Water Quality Example GUIDING QUESTIONS CHARACTERISTIC QUALIFIED ANSWER Do you need to use EO Justification Yes, if there is no other source of information on water quality going back 10 years. data? Information from EO data is needed as it is the only archival information of a (semi) quantitative nature available. Can EO provide the required Suitability Yes, retrospective information on chlorophyll, cyanophycocyanin, suspended matter, data products? turbidity, Secchi disk transparency, and vertical attenuation coefficient of light through time and space is key to understanding what aquatic ecosystem processes occurred in the last 10 years. However, each satellite sensor will differ in terms of ability to differentiate water quality variables based on its spectral resolution (see table 6.12). What is the appropriate Spatial resolution Given that the lake is 40 x 8 kilometers, lies in the subtropics (with a wet cloudy season pixel size? often obscuring the lake and a dry season with clear sky conditions) and a minimum period of interest of the last 10 years, a study of tables 5.1, 5.2, and 6.12 shows that the MODIS, MERIS, and Landsat sensor image data are the appropriate ones to use. How frequent do these Temporal Coarser-scale MODIS and MERIS data offer a higher temporal frequency of coverage. observations need to be? frequency However, under cloud-free conditions, Landsat may offer sufficient frequency. How far back in time does Record length The length of the archive available and the period of interest will determine the suitability your data record need to go? of each satellite sensor. Do you need guaranteed Reliability Use tables 5.1, 5.2, and 6.12 to identify the sensor systems with continuing future data continuation of data supply supply. into the future? What degree of accuracy is Accuracy The capability to measure all water quality variables in table 6.12 increases from Landsat to needed in the data MODIS to MERIS on the basis of their spectral characteristics. The accuracy will generally products? be highest for MERIS. Do you want to use only Maturity EO algorithms for water quality products are summarized in tables 7.10–7.12. Progressing data products that are from empirical methods (requiring a sufficient number of simultaneous field commonly used? measurements synchronous with a satellite overpass) to semi-empirical measurements, to semi-analytical methods, reliability and accuracy increase, but complexity also increases, while maturity decreases. The relevant question for this specific case study is: Do you need qualitative assessment of change or do you need the most reliable concentration estimates? In this case, a qualitative assessment of the satellite archive where the transition in time and space can be mapped from a mesotrophic, clear lake to a hypertrophic, algal bloom– dominated lake over a span of about 10 years does not require accurate water quality retrievals but does require frequent images with the capability to see when cyanobacteria start dominating the system. This leads to the conclusion that you should focus on the MERIS archive, using off-the-shelf products available through the BEAM software package. Note: EO = Earth observation. Achieving More Sustainable Basin Water and the environment by ensuring an adequate Management flow of water. The second hypothetical example is a river The main question that needs to be answered basin where irrigated agriculture is practiced is how much water is taken from surface water regularly. Water is diverted from the river and and groundwater. Additional questions include also used from groundwater to irrigate crops. how much water is used by crops, how much is Natural vegetation, particularly wetlands, suf- lost in transportation to crops (through ineffi- fer from reduced water supply, especially in ciencies in canals), what is the variation of water drier than average years. A new project intends use between years (particularly wet and dry to improve the long-term sustainability of both years), how does water use vary across the the rural communities that depend on crops region, how much water is used by each type of 164  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T crop, and who uses the water. Ideally, this infor- points is very poor. In addition, it is suspected mation should be available for the previous that a large amount of water is taken from the 15–20 years in order to characterize the spatial rivers and pumped from the ground illegally. and temporal variability in water use. Precipitation and other meteorological data are What is the status of the information net- also scarce. To inform whether to proceed on works? In this hypothetical example, the infor- the basis of EO data, table 7.14 lists additional mation available from river gauges and diversion pertinent questions. Table 7.14  Guiding Questions for Determining the Required EO Data Product Characteristics:   Efficiency of Agricultural Water Use Example GUIDING QUESTIONS CHARACTERISTIC QUALIFIED ANSWER Do you need to use EO data? Justification EO is an appropriate tool for estimating crop extent and water use, which complements the existing, field-based information (or replaces it when such information is not available). Can EO provide the required data? The Suitability Yes, retrospective information on evapotranspiration, rainfall, and required products? soil moisture can be obtained or generated. What is the appropriate pixel size? Spatial resolution The required cell size depends on the size of the actual crop paddocks. It is assumed that images of tens of meters resolution are sufficient, such as Landsat TM/ETM or similar. Data of coarser resolution, such as MODIS, can be of use as well, as it can provide a higher temporal repeatability. Data from these two satellites can be used to estimate evapotranspiration. RS estimates of precipitation and soil moisture can complement the information available for the project, albeit at coarser resolution. How frequent do these observations need Temporal frequency The ET estimates should ideally include as many observations to be? during the crop-growing cycle as possible. In practice, this will be limited by the data available. If using Landsat or similar sensor (to provide the highest spatial resolution possible), data are available every 16 days, but cloud cover (depending on location and season) will determine how often a useful observation is available. Coarse spatial resolution data such as MODIS can provide more frequent information (as it passes daily). A data-blending technique could be used. How far back in time does your data Record length It is worth doing a retrospective analysis of water use in the last record need to go? years to understand the interannual variability and trends in water use. How far back in time depends on the particular circumstances of the region under study and on the availability of data (both field based and from satellite) in the past. Data from the Landsat TM sensor are available from 1986–87 onward, so potentially almost 30 years of continuous observations at 30-meter resolution are available. Do you need guaranteed continuation of Reliability Yes, if the project intends to maintain a system that can provide data supply into the future? information on water use in the area into the future. If it is a one-off study looking at the present and past, continuation of data supply is not needed. What degree of accuracy is needed in the Accuracy The accuracy of ET estimates from EO is equal to or better than data products? 1 millimeter per day. This accuracy is generally adequate for assessing water balance in actively growing crops. Do you want to use only data products Maturity There is currently no operational ET product at high spatial that are commonly used? resolution (tens of meters). Some research agency may need to develop it for the study area. Note: EO = Earth observation; ET = evapotranspiration; RS = remote sensing. chapter 7 : A ssessing the C haracteristics of R e q uire d an d A vai l ab l e E arth O bservation Data   |  165 NOTES Matthews, M. W. 2011. “A Current Review of Empirical Procedures of Remote Sensing in Inland and Near-Coastal Transitional Waters.” International 1. For the validation pages, see http://www.isac.cnr Journal of Remote Sensing 32 (21): 6855–99. .it/~ipwg/. Matthews, M. W., B. Stewart, and R. Lisl. 2012. 2. For NASA, see oceancolor.gsfc.nasa.gov; for ESA’s “An Algorithm for Detecting Trophic Status BEAM software, see www.brockmann-consult. (Chlorophyll-A), Cyanobacterial-Dominance, de/cms/web/beam/. Surface Scums, and Floating Vegetation in Inland and Coastal Waters.” Remote Sensing of Environment 124 (September): 637–52. REFERENCES Odermatt, D., A. Gitelson, V. E. Brando, and M. Schaepman. 2012. “Review of Constituent Ebert, E. E., J. E. Janowiak, and C. Kidd. 2007. Retrieval in Optically Deep and Complex Waters “Comparison of Near-Real-Time Precipitation from Satellite Imagery.” Remote Sensing of Estimates from Satellite Observations and Environment 118 (3): 116–26. Numerical Models.” Bulletin of the American Sapiano, M. R. P., J. E. Hanoiak, W. Shi, Meteorological Society 88 (1): 47–64. R. W. Higgins, and V. B. S. Silva. 2010. Hu, C., Z. Lee, R. Ma, K. Yu, D. Li, and S. Shang. “Regional Evaluation through Independent 2010. “Moderate Resolution Imaging Precipitation Measurements: USA.” In Satellite Spectroradiometer (MODIS) Observations of Rainfall Applications for Surface Hydrology, Cyanobacteria Blooms in Taihu Lake, China.” edited by M. Gebremichael and F. Hossain, Journal of Geophysical Research 115(C4): C04002. 169–204. Dordrecht: Springer Science+Business doi:10.1029/2009JC005511. Media. 166  |  P A R T I I : E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E S M A N A G E M E N T PART III Validation of Remote Sensing– Estimated Hydrometeorological Variables Eleonora M. C. Demaria and Aleix Serrat-Capdevila OVERVIEW Satellite-estimated hydrometeorological variables evapotranspiration, water levels (in large rivers, are increasingly available at spatial and temporal lakes, estuaries, and oceans), changes in aquifer mass scales suitable for different research and operational (levels), topography (subsidence), temperature, snow applications in the fields of agriculture, hydrology, cover, snow water equivalent, and many water qual- meteorology, and water quality and supply, among ity parameters such as chlorophyll, cyanobacterial others. In parallel, an increase in computational indicators, colored dissolved organic matter, and sus- power has allowed the development of a broad range pended matter. In addition, land surface and hydro- of scientific and operational applications that help logic models are used to assimilate satellite estimates with understanding the climate on Earth, forecast- to simulate river flows, crops, landslides, and vector- ing weather and hydrologic events, and improving borne diseases, to name just a few (Fernández-Prieto natural resources management. The continuous et al. 2012; Guilloteau et al. 2014; Hong and Adler growth and improvement in the quality of the avail- 2008; Serrat-Capdevila, Valdes, and Stakhiv 2013). able remote sensing (RS) measurements provide sci- Current and planned satellite missions are of great entists with an unprecedented capability to observe interest to natural resources managers in data-poor and evaluate different components of the water cycle countries who can use RS estimates for their short- at spatial scales ranging from local to global. and long-term planning when a lack of ground net- Satellite missions routinely measure or estimate— works undermines the feasibility and quality of more or less accurately—precipitation, soil moisture, natural resources evaluations and forecasting.   167 S T R U C T U R A L M E A S U R E S A G A I N S T T S U N A M I S   |  However, validity or “ground truth” of RS ground, and errors in the algorithms used, for products is one of the main characteristics to instance, to convert brightness temperatures be taken into account when considering their and radar signatures into amounts of precipita- potential use. Satellite estimations are prone to tion (Demaria et al. 2014). Additionally, coarse several sources of uncertainty, which can sig- ground networks make the calibration of satel- nificantly affect the quality of the variables lite estimates difficult or even impossible in to  be forecasted. The three main sources of many regions around the world. uncertainty in satellite estimations are retrieval Part III is organized as follows. Chapter 8 errors, sampling errors, and inadequate ground discusses the main challenges of using ground observations (figure III.1). Uncertainties may observations to validate RS estimates of hydro- also arise from the need for model calibration, logic variables, and describes the methodologi- different spatial scales, and bias correction of cal approach used to evaluate the reliability of the estimated values prior to being used for RS products at different spatiotemporal scales. water resources applications. Chapter 9 evaluates the performance of RS Sampling errors result from discontinuities products for measuring meteorological vari- in space and time between two consecutive sat- ables, and chapter 10 focuses on the use of ellite passages. In the case of precipitation esti- remotely sensed variables in combination with mates, a satellite takes snapshots of the cloud hydrologic or hydrodynamic models for esti- fields (reflectivity) at specified times through- mating streamflow. Chapter 11 provides a syn- out the day. Numerical algorithms are subse- thesis of the main take-home messages. quently used to extrapolate those measurements in space and time to obtain daily totals. Retrieval errors stem from sources such as noise in the REFERENCES instrument measurements, improper calibra- tion of the sensor, the sensor’s inability to delin- Demaria, E. M. C., B. Nijssen, J. B. Valdés, Rodriguez, and F. Su. 2014. “Satellite D. A. ­ eate rainy and dry areas, errors in the transfer Precipitation in Southeastern South America: of information between the satellite and the How Do Sampling Errors Impact High Flow Simulations?” I ­ nternational Journal of River Basin Figure III.1 Main Sources of Uncertainty in Satellite- Management 12 (1): 1–13. Estimated Hydrologic Variables Fernández-Prieto, D., P. van Oevelen, Z. Su, and W. Wagner. 2012. “Advances in Earth Observation Model calibration for Water Cycle Science.” Hydrology and Earth Bias correction Spatial scales System Sciences 16 (2): 543–49. Guilloteau, C., M. Gosset, C. Vignolles, M. Alcoba, Y. M. Touree, and J. Lacaux. 2014. “Impacts of Satellite-Based Rainfall Products on Predicting Spatial Patterns of Rift Valley Fever Vectors.” Ground Retrieval Journal of Hydrometeorology 15 (4): 1624–35. observations error Uncertainty Hong, Y., and R. F. Adler. 2008. “Predicting Global remote sensing Landslide Spatiotemporal Distribution: Integrat- estimates ing Landslide Susceptibility Zoning Techniques and Real-Time Satellite Rainfall Estimates.” International Journal of Sediment Research 23 (3): 249–57. Serrat-Capdevila, A., J. B. Valdes, and E. Z. Stakhiv. Sampling 2013. “Water Management Applications for error Satellite Precipitation Products: Synthesis and Recommendations.” Journal of the American Water Resources Association 50 (2): 509–25. 168  |  E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E M A N A G E M E N T CHAPTER 8 Challenges of Remote Sensing Validation INTRODUCTION of both types of data: (a) direct point observa- tions of rainfall that makes it to the ground sur- An important challenge is how to reconcile face and (b) radar and satellite estimates that remote sensing (RS) estimates with ground provide the spatial distribution of the rainfall. measurements, as they can be observations of a Thus validation efforts rely on reference very different nature. For example, rain gauge data sets consisting of ground observations, measurements represent the rainfall in a few radar or other satellite data, or a combination square centimeters—with a time interval or of observation types to measure and charac- aggregation that varies from seconds to daily terize the errors in satellite estimates. In and a spatial characterization limited by the regions where the density of ground observa- number of rain gauges—and are often not avail- tions is high (mostly in the United States and able in real time (especially in developing- some European countries), gridded precipita- country settings). Satellite rainfall estimates are tion products from interpolated ground obser- indirect measures (from infrared, passive vation data are available, frequently at daily microwave, or radar sensors), often having a and monthly aggregations. However, in the spatial resolution1 ranging from 0.04° to 0.25°— absence of ground measurements of certain with a precipitation value representative of an variables, RS data can also be evaluated against area 16 and 625 square kilometers, respec- model-generated data, bearing in mind that tively—and time steps ranging from half an model-generated data may contain signifi- hour to three hours or a day. The different spa- cantly more errors than ground observations. tial footprints of the reference data sets pose This is the case of values for evapotranspira- the most difficulties in evaluating and validat- tion or soil moisture content that have been ing RS estimates. For this reason, the best rep- estimated with hydrologic land surface mod- resentation of “ground truth” is an assimilation els. In these cases, the hydrologic model has   169 been calibrated and validated for the region from the initial selection, based on two criteria: using available or proxy observations, but the (a) articles that used only ground observations errors contained in the output data may be sig- for the validation process and (b) articles that nificant and need to be acknowledged. The used similar metrics to measure the errors in use of model data (as a substitute for observa- satellite estimates. This culling yielded the fol- tions) for evaluation and validation purposes lowing selection: 24 articles about precipita- was not considered in this exercise. tion, 19 articles about evapotranspiration (ET), In a validation effort, different types of 17 articles about soil moisture, 19 articles about errors may be considered. For example, RS snow water equivalent and snow depth, and 15 estimates of precipitation may contain three articles about surface water levels and types of errors: missed events (no detection of streamflows. events), false alarms (detection of rainfall not The uncertainty of the satellite estimates recorded on the ground), and errors in the rain was grouped, when possible, by (a) temporal rate magnitude of correctly detected rainfall scale, ranging from daily to annual, (b) spatial events. However, the errors from ground mea- scale, from point or cell to basin, and (c) vari- surement networks should also be considered, ability on a global scale. A tabular summary of and reference data sets need to be bench- reliability indicators—the root mean squared marked (Anagnostou et al. 2010). The true error (RMSE), the bias, and the correlation errors in satellite estimates are significantly coefficient (CC)—is presented for each key lower when the errors in ground networks and hydrologic variable. As pointed out, this review the covariance between errors in the two types covered only scientific studies that used of observations are acknowledged (Ali, Lebel, ground observations for the validation pro- and Amani 2005). For each type of hydromete- cess. Additionally, it included as many world- orological variable considered, similar caveats wide validation sites as possible to obtain the apply to errors and limitations of both remote geographic variability of the uncertainty. and in situ observations. All of these errors and To unify the results and make meaningful potential limitations need to be accounted for recommendations, the bibliographic analysis in the evaluation and use of RS estimates. focused on the magnitude-of-error indicators: RMSE, bias, and CC. These are usually defined as follows:2 METHODOLOGICAL APPROACH n 1 The review of validation efforts of remote RMSE 5 n  (S 2 O ) (8.1) t51 t t 2 sensing (RS) estimates of key hydrometeoro- logical variables proved challenging. The pub- lications included and reviewed in this part use Bias 5 St 2 Ot,(8.2) a wide range of metrics and approaches in their validation efforts for each case study’s CC 5 location. Special attention was given to articles  S O 2 S  O n t t t t published in peer-reviewed scientific journals, 2 2 n   O 2  O t 2 n   S 2  S t t 2 t which are indexed based on the impact of the cited research on the field of study. (8.3) For the present review, 205 articles were initially selected from the scientific literature where St denotes satellite estimations and Ot published in the last 11 years (2003–14). To denotes ground observations. Bias can also be facilitate comparisons, 94 articles were culled expressed as a percentage of the observed value. 170  |  P A R T I I I : V ali d ation of R emote S ensing – E stimate d H y d rometeorological V ariables Chapter 10 gives (a) an overarching view NOTES of validation efforts to date for each key vari- able, (b) some indications regarding the con- 1. National Aeronautics and Space Administration text in which these were carried out, and (NASA) missions that produce data parameters with a coarse spatial resolution typically report (c) a sense of the range of findings. Only the the resolution in geographic degrees or fractions variables most relevant to hydrologic appli- of degrees. The size of a degree (or fraction of a cations and water resources management degree) depends on how close the measured area is to the equator and the poles. The spatial area of a were included: precipitation, evapotranspi- 1° x 1° square (that is, 1° of latitude x 1° of longitude) ration, soil moisture, snow cover and snow gets smaller the closer you get to the poles. water equivalent, water surface levels, and 2. RMSE measures the differences between (sample streamflow simulations using satellite-esti- or population) values predicted by a model or esti- mated precipitation. mator and the values actually observed. The bias (of an estimator) is the difference between the The review of streamflow simulation estimator’s expected value and the actual value of applications in chapter 11 had to deal with the the parameter being estimated. CC represents the fact that most publications had used different degree of linear dependence of two variables and always lies between −1 and +1. hydrologic modeling approaches—lumped versus distributed models, different forcing variables and approaches for model calibra- REFERENCES tion, raw versus bias-corrected satellite esti- mates, and different model specificities—as Ali, A., T. Lebel, and A. Amani. 2005. “Rainfall Estima- well as varied geographic locations, including tion in the Sahel. Part I: Error Function.” Journal of Applied Meteorology 44 (11): 1691–706. flat or mountainous terrain and diverse Anagnostou, E. N., V. Maggioni, E. I. Nikolopoulos, weather regimes. Thiemig et al. (2014) pro- T. Meskele, F. Hossain, and A. Papadopoulos. vide a good framework for the evaluation of 2010. “Benchmarking High-Resolution Global RS applications for modeling streamflow. Satellite Rainfall Products to Radar and Rain- Gauge Rainfall Estimates.” IEEE Transactions on Groundwater estimations were not included Geoscience and Remote Sensing 48 (4): 1667–83. in the review because of their limitations Famiglietti, J. S., M. Lo, S. L. Ho, J. Bethune, (described in the section on groundwater in K. J. Anderson, T. H. Syed, S. C. Swenson, C. R. de chapter 6). The Gravity Recovery and Climate Linage, and M. Rodell. 2011. “Satellites Measure Experiment (GRACE) mission developed by Recent Rates of Groundwater Depletion in California’s Central Valley.” Geophysical Research the National Aeronautics and Space Adminis- Letters 38 (3). doi: 10.1029/2010GL046442. tration and the German Aerospace Center can Feng, W., M. Zhong, J. M. Lemoine, R. Biancale, measure changes in the Earth’s mass at a H. T. Hsu, and J. Xia. 2013. “Evaluation of monthly time step and at spatial scales of about Groundwater Depletion in North China Using the Gravity Recovery and Climate Experiment 250–300 kilometers, which yield spatial and (GRACE) Data and Ground-Based Measure- temporal resolutions too coarse for planning ments.” Water Resources Research 49 (4): 2110–18. and management purposes. However, aquifer- Frappart, F., L. Seoane, and G. Ramillien. 2013. “Valida- level measurements are sparse, and validation tion of GRACE-Derived Terrestrial Water Storage from a Regional Approach over South America.” efforts require the use of hydrologic models as Remote Sensing of Environment 137 (October): 69–83. ground truth (Wahr, Swenson, and Velicogna Thiemig, V., R. Rojas, M. Zambrano-Bigiarini, 2006). Despite their limitations, satellite esti- V. Levizzani, and A. DeRoo. 2014. “Validation mations have been used successfully to analyze of Satellite-Based Precipitation Products over long-term trends in changes in groundwater Sparsely Gauged African River Basins.” Journal of Hydrometeorology 13 (6): 1760–83. levels on a regional or continental scale Wahr, J., S. Swenson, and I. Velicogna. 2006. “Accuracy (Famiglietti et al. 2011; Feng et al. 2013; of GRACE Mass Estimates.” Geophysical Research Frappart, Seoane, and Ramillien 2013). Letters 33 (6). doi: 10.1029/2005GL025305. C hapter 8 : C hallenges of R emote S ensing V ali d ation   |  171 CHAPTER 9 Validation of Remote Sensing Data PRECIPITATION retrieval (Hsu et al. 1997; Huffman et al. 2001) or based on blended methods that use multi- Precipitation is the input variable most com- satellites and multi-sensors (Huffman et al. monly found in hydrologic applications and 1997, 2007). This section evaluates the accu- processes related to the water cycle. Histori- racy of satellite-based precipitation estimates cally, the main source of precipitation data has based on the following products: been observations from ground gauge networks • Two products from the National Aeronau- and in some places, if available, from precipita- tics and Space Administration (NASA): tion radars. However, well-­ functioning ground Tropical Rainfall Measuring Mission networks are limited to the industrial coun- (TRMM) and Multisatellite Precipita- tries and are extremely sparse in the underde- tion Analysis (TMPA) real-time product veloped parts of the globe. Moreover, network (TMPA 3B42RT) and TMPA 3B42 densities are relatively low over thinly popu- lated, high-latitude areas. Satellite-based pre- • One product from the Climate Prediction cipitation estimates are naturally of most Center: the Climate Prediction Center interest in those parts of the world where MORPHing technique (CMORPH) ground observation networks are sparse. This • One product from the University of is even more true if the population density in California, Irvine: Precipitation Estimation those areas is relatively high (and thus highly from Remotely Sensed Information Using vulnerable to hydrologic extremes), entailing Artificial Neural Networks (PERSIANN) relatively high water needs and use. Various precipitation products already • One product from the Japan Science and exist that are either based exclusively on (visi- Technology Agency (JAXA): Global Satel- ble, infrared, or passive microwave) satellite lite Mapping of Precipitation (GSMaP).   173 Table 9A.1 in annex 9A (available online at Extreme Precipitation Events https://openknowledge.worldbank.org/handle In data-poor regions where intense storms are /10986/22952) presents key aspects of the sat- responsible for substantial economic losses, ellite products used in the validation process of large numbers of displaced people, and a flood- the scientific literature reviewed. For each related death toll, satellite estimations of pre- journal article reviewed, the table gives the cipitation would be very valuable for the study geographic location of the validation site, the and monitoring of such destructive meteoro- temporal scale (that is, daily, monthly, annual, logical phenomena. However, based on three- or seasonal), and the magnitude of the root hour satellite estimations, Mei et al. (2014) find mean squared error (RMSE), bias, and correla- that warm-season extreme precipitation val- tion coefficient (CC). ues (that is, those above the 90th percentile) are poorly correlated with ground observa- Orographic Effects on Estimated   tions—correlation values ranging from 0 for Precipitation CMORPH to 0.51 for TMPA 3B42RT v7. Dur- In mountainous areas, satellite sensors have ing the cold season, the correlation is even trouble capturing orographic precipitation weaker, with an average value of 0.16. The and the effects of rain shadow. In addition, as RMSE values range from 0.38 to 0.98 ­millimeter ground networks are sparse or nonexistent during the warm season and from 0.54 to in such areas, sensor calibration and valida- 0.86 millimeter during the cold season. tion are difficult. In the western Black Sea TMPA 3B42 systematically underestimates region of Turkey, where the complex topog- the magnitude of tropical cyclones in Australia raphy is a major factor in the genesis of pre- by −15 percent for rainfall intensities in the range cipitation, studies have found that the volume of 50–75 millimeters per day and by −40 percent of monthly precipitation is 50 percent less for intensities higher than 200 millimeters per (within a 50-kilometer range) on the leeward day (Chen et al. 2013). Similarly, for tropical (drier) side of the mountain range than on cyclones in the southeastern United States, the windward (wetter) side, due to the rain TMPA products 3B42 and 3B42RT show biases shadow effect (Derin and Yilmaz 2014). On in the ±25 and ±50  percent range of observa- average, satellite products tend to underesti- tions, respectively (Habib, Henschke, and Adler mate observations on the windward side of 2009).2 In southeastern South America, where mountains by −18 percent (negative bias) the most intense precipitation on Earth has during the warm and dry s ­ eason and by as been documented (Zipser et al. 2006), satellite much as −53 percent during the cold and products fail to capture the magnitude of aver- humid season. Satellites tend to overestimate age precipitation of meso-scale convective sys- precipitation observations, with the excep- tems—thunderstorm systems with a spatial tion of CMORPH estimates, on the leeward range of 100 kilometers or more. In the case of side, on average, by +2  percent during the “pure” satellite products, the CMORPH esti- warm season and by +25 percent during the mate biases range from −70 to +60 millimeters cold season.1 Since warm orographic pro- per day, while the PERSIANN estimate biases cesses cannot always be detected by passive range from −55 to +25 millimeters per day. Even microwave or infrared sensors (Dinku et al. the TMPA 3B42 rainfall estimates show biases 2007), biases in daily products range from in the range of −60 to +50 millimeters per day. −9.5 (warm season) to −51.8  percent (cold This is surprising since this satellite product is season) on the windward side and from routinely bias-corrected using ground observa- +7.25 percent (warm season) to +38.3 percent tions. This suggests that, at least in certain (cold season) on the leeward side. regions of the world, this post-processing 174  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S correction does not necessarily offer much of an storms), with values ranging from 0.19 to 0.72 improvement over the uncorrected data sets and a median value of 0.32 (figure 9.1, blue col- (Demaria et al. 2011). umn). Winter storms, which are characterized by warm top clouds with insufficient ice for sat- Seasonal Precipitation ellite sensors to detect precipitation, are respon- In Australia, infrared–passive microwave sat- sible for satellite misses and an increase in the ellite products such as TMPA 3B42RT, number of false alarms. For more intense pre- ­ CMOPRH, and PERSIANN have performed cipitation (higher than 20 ­ millimeters per day), better in the tropics during the summer satellite products show biases during the sum- months (December–January), when rain is mer, especially in the mid-latitude regions, mostly of a convective nature, than in midlati- because they cannot observe the rapid temporal tudes, where the accuracy of satellite sensors evolution of most convective storms (Ebert, deteriorates slightly. Since several validation Janowiak, and Kidd 2007). studies have been performed at the seasonal level, figure 9.1 shows the correlation coeffi- Summary cient between infrared–passive microwave The findings for the validation of precipitation satellite products and ground observations products are as follows: for the summer and winter, respectively, • In regions of complex topography, such as based on a subset of the summary data pro- mountainous regions, satellite products vided in table 9A.1 in annex 9.A (available tend to underestimate precipitation on online). Three-hourly, daily, monthly, and the windward side of the mountain (−18 annual validation correlation coefficients are and −53 percent during the warm-dry and not included in the plot. cold-wet season, respectively) and to over- During summertime, correlation coefficients estimate precipitation on the leeward side range from 0.35 to 0.85, with a median value of (+2 and +25 percent during the warm-dry 0.65 (figure 9.1, green column).3 In the winter and cold-wet season, respectively). months, they deteriorate slightly (most likely because of the nonconvective nature of winter • Satellite products have trouble estimating extreme precipitation events such as tropi- cal and subtropical storms, with biases of Figure 9.1  Correlation Coefficients between ±25 and ±50 percent for the TMPA prod- Observed and Satellite-Estimated Precipitation ucts 3B42 and 3B42RT, respectively. 1.0 • Satellite sensors are better at capturing 0.9 convective, summer precipitation in the 0.8 Correlation coefficient (CC) tropics and midlatitudes (CC ranges from 0.7 0.35 to 0.85, with a median of 0.65) than 0.6 winter precipitation (CC ranges from 0.19 0.5 to 0.72, with a median of 0.32), which is 0.4 usually of a nonconvective nature. 0.3 0.2 0.1 EVAPOTRANSPIRATION 0.0 Summer (n = 15) Winter (n = 13) Evapotranspiration (ET)—through evapora- Note: The satellite products used include CMORPH, PERSIANN, tion from the soil, rainfall intercepted by plants, GPROF 6.0, RFE 2.0, and TMPA 3B42 (both v6 and v7). n = sample size. The black horizontal line in both columns represents the median value. and plant transpiration—is a key component of chapter 9 : V alidation of R emote S ensing  D ata   |  175 the coupling between the atmosphere and the different spatial scales (Jiménez, Prigent, and Earth surface. In most ET estimation methods, Aires 2009; Jiménez et al. 2011; Mueller et al. the driving parameter is net radiation, and the 2011, 2013). vapor pressure deficit is used to calculate water vapor transfer. In recent years, several ET data Validation of Remotely Sensed   sets have been developed based on in situ ET Estimates ground data or satellite retrievals. Satellite Table 9A.2 in annex 9A (available online) pres- imagery, increasingly available at fine spatial ents a comparative summary of different ET and temporal resolutions, has generated infor- estimates. In general, satellite products derived mation that has allowed the development of on a monthly time scale have stronger agree- ET estimation schemes. While satellite remote ment with ground observations than those sensing provides reasonable estimates of dif- derived on a daily basis. Moreover, estimates ferent land surface fluxes, it does not measure show better agreements in humid (subtropical) evapotranspiration directly. Instead, the scien- than in arid and semiarid regions, as shown for tific community relies on retrieval algorithms the African continent, where ET products sys- to integrate those fluxes and simulate evapo- tematically overestimate reference values transpiration’s variability. (Trambauer et al. 2014). However, the uncer- Several methods, of different degrees of tainty band ranges from −30 percent underesti- complexity, have been developed using mation to +20 percent overestimation, most schemes that balance empirical and physically likely as a result of model ­ deficiencies—more based components. The simplest method specifically, the failure to account fully for (direct method) uses thermal infrared to infer changes in soil moisture resulting from plant temperature in the atmosphere, which is then transpiration and forest rainfall  interception used along with ground temperature measure- (Miralles et al. 2011). The  RMSE  ranges from ments to estimate ET rates. These methods are 0.26 millimeter to 3 millimeters per day, with an sensitive to cloud conditions and to errors in average value of 0.94 millimeter per day and a the ground- and satellite-measured tempera- standard deviation of 0.73  millimeter per day. ture values. Deterministic methods use Comparisons with ground observations (flux soil-vegetation-atmosphere transfer (SVAT) models, which can potentially be linked to cli- mate and hydrologic models, but require accu- Figure 9.2  Correlation Coefficients rate RS estimates of evapotranspiration and between Observed and Satellite-Estimated the estimation of several model parameters. Evapotranspiration 1.0 SVAT models and RS data can be combined into more complex data assimilation processes 0.9 (Courault, Seguin, and Olioso 2005). 0.8 Correlation coefficient (CC) As is the case of most hydrometeorological 0.7 variables estimated with satellites, the lack of 0.6 ground reference data is one of the main cul- 0.5 prits of the estimates’ uncertainty (Wang and 0.4 Dickinson 2012). To mitigate the impact of the 0.3 lack of observations, an international initiative 0.2 was launched in the previous decade to evalu- 0.1 ate and compare existing land ET products.4 0.0 n = 22 The project aims to create a global, multiyear Note: n = sample size. The black horizontal line in the column benchmark data set of evapotranspiration at represents the median value. ­ 176  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S towers) worldwide suggest a robust linear cor- required to validate satellite products and relation (median value of CC = 0.83) between introduce significant sampling uncertainty satellite-estimated ET and ground observations (Crow et al. 2012). figure 9.2), despite the large biases.5 (­ The theoretical basis for using remote sen- sors to measure soil moisture content is based Summary on the contrast between the dielectric proper- The findings for the validation of evapotrans- ties of the dry soil material and the water. Water piration products are as follows: has a large dielectric constant (of about 80); when this is added to the dry soil matrix (dielec- • Satellite observations can estimate the tric constant of about 4), the soil’s dielectric main drivers of evapotranspiration (tem- constant rises significantly and the emission perature, latent heat, sensible heat) on a and scattering properties of the soil change (de global scale and thus can be very valuable in Jeu et al. 2008). The validation of RS soil mois- meeting the need for global ET estimates. ture is challenging due to the disparity between • However, large discrepancies in the esti- the spatial scales of the satellite and those of in mates indicate that land evapotranspira- situ observations. Conventional soil moisture tion is, and will remain, one of the most observations provide point measurements, uncertain components of the water bal- while satellite observations provide estimates ance, with biases ranging from −30 to +10 covering a much larger spatial area (Su et al., percent and average RMSE values of 0.94 2011 and 2013). Moreover, soil moisture has a millimeter per day (±0.73). relatively large spatial and temporal variability, related to the presence or absence of vegetation coverage. Soil emissions tend to be attenuated SOIL MOISTURE by the vegetation canopy, resulting in decreased sensitivity of sensors to variations in soil mois- Soil moisture—water stored in the soil—­controls ture. Extremely dry soils, such as are found in the partitioning of available energy into sensi- desert regions, can also introduce uncertainty in ble and latent heat fluxes and influences the the sensors’ measurements because of higher evolution of weather and hydrologic processes backscatter (de Jeu et al. 2008). Examples of in a basin. In recent decades, soil moisture has soil moisture sensors aboard different satellites routinely been estimated with several satellite are provided in table 6.6 in chapter 6. sensors (Dorigo et al. 2010). However, the lack of soil moisture observations that can be used Validation of Remotely Sensed Soil for validation remains a fundamental problem. Moisture Estimates To address this issue, an international network Table 9A.3 of annex 9A (available online) has been created to support efforts aimed at shows the results of validation efforts using establishing and maintaining a global in situ satellite-estimated soil moisture and ground soil moisture database, which is essential for observations from intensive field campaigns the scientific community to be able to validate and existing networks in Australia, France, and improve global satellite observations.6 Italy, Spain, and the United States as well as in Typically, existing and planned ground-based Asia and West Africa. The RMSE values7 range soil moisture networks cover areas ranging from 0.01 to 0.36 cubic meter of water per from 100 square kilometers to 10 million cubic meter of soil (m3/m3) and have a mean square kilometers. However, because of the value of 0.11 (±0.09) m3/m3. However, most spatial variability of observed soil moisture, RMSE values are small (figure 9.3, panel a), as coarser networks often lack the resolution 69 percent of the studies evaluated have RMSE chapter 9 : V alidation of R emote S ensing  D ata   |  177 Figure 9.3  Errors in RS Estimations of Soil Moisture correlation coefficients larger than 0.7 and a. RMSE b. Bias 40 percent of the sites have correlation coeffi- 0.4 0.4 cients larger than 0.6. Figure 9.4 shows the dis- tribution of the correlation coefficients 0.3 0.3 grouped by type of sensor (passive, active, and combined). While the correlation coefficients RMSE (m3/m3) Bias (m3/m3) 0.2 0.2 of the active sensors show slightly less disper- 0.1 sion than those of the passive sensors, both 0.1 types of sensors have similar mean CC values, 0.0 indicating that, for the studies included in this 0.0 –0.1 report, both methods are comparably effective 0 5 10 15 20 0 5 10 15 20 at retrieving soil moisture data from space. Su Validation experiments Validation experiments Note: The x-axis shows individual satellite products. For the validation exercise, only 17 scientific et al. (2011, 2013) provide additional examples studies were considered. Since all of the satellite products from each study were included in the of RS soil moisture validations, focusing on the analysis, n = 22; RMSE = root mean squared error. Tibetan Plateau. values less than 0.15 m3/m3. Satellite estimates Summary tend to overestimate the value of observations, The findings for the validation of soil moisture as shown by the mean positive bias of 0.04 are as follows: (±0.05) m3/m3. As is the case with the RMSE, • Satellite estimates are only representative biases range from −0.09 to 0.13 m3/m3, with of the top 5 centimeters of the soil layer, 82 percent of the validation sites showing posi- which can limit their applicability. tive values (figure 9.3, panel b). The correlation coefficients between obser- • Lack of ground observations limits vations and RS estimates range from 0.11 to ­ satellite-derived estimate validation to a 0.96. Despite a mean CC value of 0.58 (±0.19), few locations and to special field campaigns. around 15 percent of the validated sites have • The soil moisture satellite products ana- lyzed yielded a mean RMSE of 0.11 (±0.09) m3/m3 and a mean positive bias of 0.04 Figure 9.4  Correlation Coefficients between (±0.05) m3/m3. Observed and Satellite-Estimated Soil Moisture, by Type of Sensor • RS estimations of soil moisture are promis- 1.0 ing, considering the mean correlation coef- 0.9 ficient of 0.58 (±0.19). 0.8 Correlation coefficient (CC) 0.7 0.6 SNOW COVER AND SNOW   0.5 WATER EQUIVALENT 0.4 0.3 Accurate information on snow in the winter- 0.2 time is an important component of spring and 0.1 summer soil moisture predictions. The actual 0.0 values of these parameters, in turn, have an Passive + active Active Passive impact on precipitation patterns, hydrologic (n = 47) (n = 22) (n = 25) extremes (floods and droughts), wildlife dynam- Note: n = sample size. The black horizontal line represents the median value of the corresponding sample. ics, and water supply. Natural ecosystems rely 178  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S heavily on spring streamflows for important and Terra satellites. An example of a fractional transitional stages in their life cycle. Despite product is the MYD10A1 Fractional (Rittger, their importance for natural resources manage- Painter, and Dozier 2013), which has the advan- ment, in situ snow measurements are sparse, tage that its estimates of snow depth can be and given the high spatial variability in snow directly compared with ground observations. distribution, remote sensing constitutes an In contrast, binary estimates are validated using invaluable source of global spatially distributed a so-called contingency matrix (also called a snow estimates. confusion matrix), which counts the number of Uncertainty in RS-derived snow estimates, coincidences between satellite and ground in addition to sampling and retrieval errors, measurements of “snow” and “no snow.” The results from snow reflectance, forest transmis- National Oceanic and Atmospheric Adminis- sivity, forest reflectance (of an opaque canopy), tration’s Advanced Very High Resolution Radi- and snow-free ground reflectance for different ometer (AVHRR) and the U.S. Air Force classes of land cover (Dong, Walker, and Defense Special Sensor Microwave Imager Houser 2005). In addition, cloud cover has a (SSM/I) also offer fractional satellite products. large impact on the overall accuracy of satellite- derived snow cover estimates. For example, Validation of Remotely Sensed Snow Maurer et al. (2003) report that the accuracy of Cover and Snow Water Equivalent the MODIS daily snow cover mapping algo- Estimates ­ rithm under clear sky conditions is more than Table 9A.4 of annex 9A (available online) sum- 80 percent. To reduce the impact of cloud marizes the errors recorded in satellite-esti- cover on snow, cloud masks are routinely mated snow cover and snow water equivalent developed to identify the areas where land for different satellite products. In densely for- products should be retrieved based on the ested areas of Canada, the uncertainty of satel- amount of cloud obstruction (Hall et al. 2002). lite products ranges from −25 to +10 percent Validation studies indicate that satellite (Foster et al. 2005). In these areas, a dense for- sensors have higher accuracy in plains areas, est canopy diminishes the ability of the satel- with little or no forest cover, than in forested lite to determine the amount of snow areas in the northern latitudes. The forest underneath it. cover masks the emission of microwaves by Despite the lack of agreement on the magni- snow. As is the case of precipitation estimates, tude and sign of the errors, the linear correla- complex topography significantly affects the tion between ground observations and satellite quality of snow data retrieval. estimations for the 15 studies included in this Unlike the satellite-derived estimates dis- analysis has a median value of 0.53 (±0.22) cussed above, satellite sensors can estimate figure 9.5, panel a). The RMSE ranges from 13 (­ snow cover and snow water equivalent in two to 75 millimeters, with a mean value of 32.3 mil- ways: (a) as binary estimates (that is, snow or limeters (±20.2). A mean negative bias of no snow), where the sensor only detects the −4.4 millimeters indicates that satellites under- presence or absence of snow on the ground estimate observations. However, the bias (but cannot estimate depth of snow or snow shows a high variability and has a standard water equivalent), and (b) as fractional esti- deviation of 26.7 millimeters.8 mates of snow-covered area, based on mixing By contrast, snow cover products compare different satellite spectral bands. favorably with ground observations, thanks to Satellite estimates available in a “binary” for- improvements in spatial and temporal resolu- mat are MOD10A1 and MYD10A1 Binary, both tion and in cloud mapping. Figure 9.5, panel b, derived from the MODIS on board the Aqua shows that the median agreement value, or chapter 9 : V alidation of R emote S ensing  D ata   |  179 Figure 9.5  Correlation Coefficients and Snow • Since the largest snow accumulations Mapping Agreement between Observed and occur at higher elevations, sparsely distrib- Satellite-Estimated Snow Water Equivalent and uted snow stations contribute significantly Snow Cover to the uncertainty. a. Snow water equivalent b. Snow cover 1.0 100 • The high accuracy of satellite sensors in 0.9 90 detecting the presence of snow on the 0.8 ground is partially due to improvements in Correlation coefficient (CC) 0.7 80 cloud mapping techniques. Ongoing efforts Agreement (%) 0.6 70 to combine satellite estimates with ground 0.5 observations have the potential to reduce 60 the uncertainty in future products. 0.4 0.3 50 0.2 40 SURFACE WATER LEVELS AND 0.1 STREAMFLOWS 0.0 30 n = 15 n = 24 Note: n = sample size. The horizontal black lines in both columns Since stream gauges are distributed sparsely ­ represent mean values. around the globe, using remote sensing to characterize river flow is extremely useful in river basins with extensive flood plains, in wet- accuracy, of estimated snow cover is 70 per- lands, and in braided rivers, where multiple cent (sample size = 24). The standard deviation river channels make it prohibitive to install of estimated snow cover for the products several gauging stations. Satellite estimates of reviewed is 23.3 percent, indicating that binary surface water levels are also useful for flood satellite products are quite effective at deter- forecasting, especially in transboundary river mining whether there is snow on the ground. basins, where the hydrologic information gen- erated from upstream areas in the basin is not Summary shared with the downstream partners (Bianca- The findings for the validation of snow cover maria, Hossain, and Lettenmaier 2011). In and snow water equivalent are as follows: regions where ground-based data are difficult to obtain due to funding shortages or political • The uncertainty in satellite estimates of unrest, satellite data may be available in near snow water equivalent and snow depth is real time for implementation in flood forecast- still high for the available products. The ing models (Coe and Birkett 2004). RMSE between observed and satellite-­ Satellite sensors can estimate surface water estimated snow water equivalent has a levels in rivers and wetlands by using the high mean value of 32.3 millimeters (±20.2). reflectivity of water. Therefore, unless the Satellite estimates tend to underestimate microwave pulses emitted by water bodies are observations by −4.4  millimeters, on aver- intercepted by vegetation, small changes in age. Moreover, the bias has a relatively water-level depths can be measured with cen- high variability and a standard deviation of timeter-scale accuracy (Alsdorf and Lettenma- 26.7 millimeters. ier 2003). The Ocean Topography Experiment • Satellites can successfully estimate ground (TOPEX)/Poseidon mission and the Japanese snow cover, as reflected by the 70 percent Earth Resources Satellite 1 (JERS-1) synthetic (±23) agreement of satellite products with aperture radar (SAR) mission carry onboard ground observations. radar altimeters designed to operate over 180  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S water and ice surfaces. An altimeter radar con- to 1.5 meters. The results suggest that the accu- tinuously emits microwave pulses toward the racy of the estimates is larger during periods surface of the Earth, and the time that passes with high water levels: 2 percent RMSE9 for between the pulse emission and the echo Lake Chad in Africa and 15-centimeter RMSE reception is used to estimate the height of the for wetland flooding in the Amazon River topographic surface (Birkett 1998; Hess et  al. basin. During the rest of the year, the errors are 2003). as high as 10 percent in Lake Chad and −2.2 Streamflows cannot be measured from meters in the Amazonian wetlands (Birkett space. Instead, satellite sensors can measure 2000; Hess et al. 2003). water levels, channel width, channel slope, and Recent work by Hossain et al. (2014) has flow velocity, among others, and models or sta- demonstrated the feasibility of implementing a tistical relationships between these variables five-day lead time water-level forecast system can then be used to estimate channel flows in the Brahmaputra River basin using Jason-2 (Bjerklie et al. 2003). In addition, infrequent estimates. Altimetry measurements to forecast satellite overpasses, limited sampling fre- stages used to force a hydrodynamic model quency (determined by the distance between inside Bangladesh yield forecast results with measurements along the satellite orbit), and RMSE values ranging from 0.2 (± 0.2) meters incomplete spatial coverage make estimations for the monsoon season and 0.7 (± 0.4) meters uncertain. One of the main constraints is the for the dry season, when compared with a pos- need to use hydraulic models (or statistical terior “nowcasting” using observed stages. correlations) (a) to route water levels along the These results strongly suggest that in large riv- river channel to compare gauge-height obser- ers, water-level altimetry measurements can vations with satellite measurements in virtual be used as inputs for hydrodynamic flow prop- stations (points where the satellite altimeter agation models and thus can be especially use- measurements are taken), as these rarely coin- ful in transboundary settings. cide with the location of a gauging station, and Despite the successful implementation of (b) to propagate water levels downstream to the system in Bangladesh, table 9A.5 shows forecast flows for the areas of interest. that satellite estimates of river water levels have RMSEs ranging from 0.27 meter to Validation of Remotely Sensed Surface 1.1 meters, which can be considered poor when Water Level Estimates compared with ground-based gauge measure- The most widely used satellites for estimating ments. However, these values do allow com- streamflow are the TOPEX/Poseidon mission parison of the interannual and seasonal developed by NASA and the French Space variability of water heights across the basin Agency, the JERS-1 SAR developed by JAXA, (Birkett et al. 2002). GRACE developed by NASA and the German Aerospace Center, and Envisat (no longer Summary operational) developed by the European Space The findings for validation of surface water Agency. and streamflow are as follows: Changes in the water level of lakes and wet- lands can be estimated by remote sensing with • The uncertainty in satellite estimates of relatively high reliability (see table 9A.5 of water surface is still high for the available annex 9A, available online), as evidenced by a products. The RMSE in lake water levels is median correlation coefficient of 0.96 (±0.24). 0.22 (±0.45) meter, on average, with greater The error in lake water levels is 0.28 (±0.45) accuracy during periods with high water meter on average and can range from 0.04 meter levels. chapter 9 : V alidation of R emote S ensing  D ata   |  181 • Infrequent satellite overpasses and incom- REFERENCES plete spatial coverage mean that hydrody- namic models have to be used to propagate Alsdorf, D. E., and D. P. Lettenmaier. 2003. “Track- estimated river water levels from the satel- ing Fresh Water from Space.” Science 301 (5639): 1491–94. lite virtual stations to locations of interest Biancamaria, S., F. Hossain, and D. P. Lettenmaier. downstream. 2011. “Forecasting Transboundary River Water Elevations from Space.” Geophysical Research ­Letters 38 (11): art. 4. Birkett, C. M. 1998. “Contribution of the TOPEX ANNEX 9A. SUMMARY OF NASA Radar Altimeter to the Global Monitoring SCIENTIFIC LITERATURE ON of Large Rivers and Wetlands.” Water Resources SATELLITE PRODUCTS INCLUDED Research 34 (5): 1223–39. IN VALIDATION ———. 2000. “Synergistic Remote Sensing of Lake Chad: Variability of Basin Inundation.” Remote Sensing of Environment 72 (2): 218–36. Annex 9A is available online at https://open Birkett, C. M., L. A. K. Mertes, T. Dunne, M. H. Costa, knowledge.worldbank.org/ handle/10986 and M. J. Jasinski. 2002. “Surface Water Dynam- /22952. ics in the Amazon Basin: Application of Satellite Radar Altimetry.” Journal of Geophysical Research: Atmospheres 107 (D20): 8059. Bjerklie, D. M., S. L. Dingman, C. J. Vorosmarty, NOTES C. H. Bolster, and R. G. Congalton. 2003. “Evaluat- ing the Potential for Measuring River Discharge 1. These average values are based on the paper by from Space.” Journal of Hydrology 278 (1-4): Derin and Yilmaz (2014) and are not shown in 17–38. table 9A.1 in annex 9A (available online). Chen, Y., E. E. Ebert, K. J. E. Walsh, and N. E. 2. These real-time products are described in table Davidson. 2013. “Evaluation of TMPA 3B42 Daily 6.3 in chapter 6. Precipitation Estimates of Tropical Cyclone 3. The values shown were taken from table 9A.2 in Rainfall over Australia.” Journal of Geophysical annex 9A (available online). The winter and sum- Research: Atmospheres 118 (21): 11966–78. mer values were analyzed (and are shown) sepa- Coe, M. T., and C. M. Birkett. 2004. “Calculation of River rately to highlight the seasonal differences in the Discharge and Prediction of Lake Height from Sat- correlation coefficient. ellite Radar Altimetry: Example for the Lake Chad 4. For information on this initiative, see http://www Basin.” Water Resources Research 40 (1) : 27–33. .iac.ethz.ch/groups/seneviratne/research/Land Courault, D., B. Seguin, and A. Olioso. 2005. “Review Flux-EVAL. on Estimation of Evapotranspiration from Remote 5. The RMSE values expressed in millimeters per Sensing Data: From Empirical to Numerical day do not say much about the performance Modeling Approaches.” Irrigation and Drainage of the RS estimates. For that reason, the figures Systems 19 (3): 223–49. in this section only highlight the correlation Crow, W. T., A. A. Berg, M. H. Cosh, A. Loew, coefficient. B. P. Mohanty, R. Panciera, P. de Rosnay, D. Ryu, 6. For information on the network and efforts to and J. P. Walker. 2012. “Upscaling Sparse Ground- ­ create a soil moisture database, see http://ismn Based Soil Moisture Observations for the Valida- .geo.tuwien.ac.at/ismn/. tion of Coarse-Resolution Satellite Soil Moisture 7. In this context, only RMSE values with m3/m3 Products.” Reviews of Geophysics 50 (2): RG2002. units are considered. de Jeu, R. A. M., W. Wagner, T. R. H. Holmes, 8. Many of the reviewed journal articles pro- A. J. Dolman, N. C. van de Giesen, and J. Friesen. vide a range of values, and the means of those 2008. “Global Soil Moisture Patterns Observed ranges were used to arrive at these figures. The by Space Borne Microwave Radiometers and Scat- underlying values are not shown in the tables of terometers.” Surveys in Geophysics 29 (4): 399–420. annex 9A. Demaria, E. M. C., D. A. Rodríguez, E. E. Ebert, P. Salio, 9. While the RMSE is usually expressed in the unit F. Su, and J. B. Valdes. 2011. “Evaluation of Meso- of the dependent variable, it can also be expressed scale Convective Systems in South America Using as a percentage. Multiple Satellite Products and an Object-Based 182  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S Approach.” Journal of Geophysical Research: Sensed Information Using Artificial Neural Net- Atmospheres 116 (D8). doi: 10.1029/2010JD015157. works.” Journal of Applied Meteorology 36 (9): Derin, Y., and K. K. Yilmaz. 2014. “Evaluation of Mul- 1176–90. tiple Satellite-Based Precipitation Products over Huffman, G. J., R. F. Adler, P. Arkin, A. Chang, Complex Topography.” Journal of Hydrometeorol- R. Ferraro, A. Gruber, J. Janowiak, A. McNab, ogy 15 (4): 1498–516. B. Rudolf, and U. Schneider. 1997. “The Global Dinku, T., P. Ceccato, E. Grover-Kopec, M. Lemma, Precipitation Climatology Project (GPCP) S. J. Connor, and C. F. Ropelewski. 2007. “Vali- Combined Precipitation Dataset.” Bulletin of the dation of Satellite Rainfall Products over East American Meteorological Society 78 (1): 5–20. ­ Africa’s Complex Topography.” International Huffman, G. J., R. F. Adler, D. T. Bolvin, G. Gu, Journal of Remote Sensing 28 (7): 1503–26. E. J. Nelkin, K. P. Bowman, Y. Hong, E. F. Dong, J. R., J. P. Walker, and P. R. Houser. 2005. Stocker, and D. B. Wolff. 2007. “The TRMM “­Factors Affecting Remotely Sensed Snow Water Multisatellite Precipitation Analysis (TMPA): Equivalent Uncertainty.” Remote Sensing of Quasi-Global, Multiyear, Combined-Sensor ­Environment 97 (1): 68–82. ­ Precipitation E­ stimates at Fine Scales.” Journal of ­Hydrometeorology 8 (1): 38–55. Dorigo, W. A., K. Scipal, R. M. Parinussa, Y. Y. Liu, Huffman, G. J., R. F. Adler, M. M. Morrissey, W. Wagner, R. A. M. de Jeu, and V. Naeimi. 2010. D. T. Bolvin, S. Curtis, R. Joyce, B. McGavock, “Error Characterisation of Global Active and Pas- and J. Susskind. 2001. “Global Precipitation at sive Microwave Soil Moisture Datasets.” Hydrol- One-Degree Daily Resolution from Multisatellite ogy and Earth System Sciences 14 (12): 2605–16. Observations.” Journal of Hydrometeorology 2 (1): Ebert, E. E., J. E. Janowiak, and C. Kidd. 2007. “Com- 36–50. parison of Near-Real-Time Precipitation Esti- Jiménez, C., C. Prigent, and F. Aires. 2009. “Toward an mates from Satellite Observations and Numerical Estimation of Global Land Surface Heat Fluxes Models.” Bulletin of the American Meteorological from Multisatellite Observations.” Journal of Society 88 (1): 47–64. Geophysical Research: Atmospheres 114 (D6). Foster, J. L., C. J. Sun, J. P. Walker, R. Kelly, A. Chang, doi: 10.1029/2008JD011392. J. R. Dong, and H. Powell. 2005. “Quantifying the Seneviratne, Jiménez, C., C. Prigent, B. Mueller, S. I. ­ Uncertainty in Passive Microwave Snow Water M. F. McCabe, E. F. Wood, W. B. ­ Rossow, Equivalent Observations.” Remote Sensing of Envi- Balsamo, A. K. Betts, P. A. Dirmeyer, G. ­ ronment 94 (2): 187–203. J. B. Fisher, M. Jung, M. Kanamitsu, R. H. Reichle, Habib, E., A. Henschke, and R. F. Adler. 2009. M. Reichstein, M. Rodell, J. Sheffield, K. Tu, “Evaluation of TMPA Satellite-Based Research and K. Wan. 2011. “Global Intercomparison of and Real-Time Rainfall Estimates during Six 12 Land Surface Heat Flux Estimates.” Journal Tropical-Related Heavy Rainfall Events over of Geophysical Research: Atmospheres 116 (D2). Louisiana, USA.” Atmospheric Research 94 (3): doi: 10.1029/2010JD014545. 373–88. Maurer, E. P., J. D. Rhoads, R. O. Dubayah, and Hall, D. K., G. A. Riggs, V. V. Salomonson, N. E. D. P. Lettenmaier. 2003. “Evaluation of the DiGirolamo, and K. J. Bayr. 2002. “MODIS Snow-Covered Area Data Product from MODIS.” Snow-Cover Products.” Remote Sensing of Hydrological Processes 17 (1): 59–71. Environment 83 (1-2): 181–94. Mei, Y., E. N. Anagnostou, E. I. Nikolopoulos, and Hess, L. L., J. M. Melack, E. Novo, C. C. F. Barbosa, and M. Borga. 2014. “Error Analysis of Satellite M. Gastil. 2003. “Dual-Season Mapping of Wet- Precipitation Products in Mountainous Basins.” land Inundation and Vegetation for the Central Journal of Hydrometeorology 15 (5): 1778–93. Amazon Basin.” Remote Sensing of Environment Miralles, D. G., R. A. M. de Jeu, J. H. Gash, 87 (4): 404–28. T. R. H. Holmes, and A. J. Dolman. 2011. Hossain, F., A. H. Siddique-E-Akbor, L. C. Mazumder, “Magnitude and Variability of Land Evaporation S. M. ShahNewaz, S. Biancamaria, H. Lee, and and Its Components at the Global Scale.” C. K. Shum. 2014. “Proof of Concept of an Hydrology and Earth System Sciences 15 (March): Altimeter-Based River Forecasting System for 967–81. Transboundary Flow Inside Bangladesh.” Mueller, B., M. Hirschi, C. Jimenez, P. Ciais, IEEE Journal of Selected Topics in Applied P. A. Dirmeyer, A. J. Dolman, J. B. Fisher, Earth Observations and Remote Sensing 7 (2): M. Jung, F. Ludwig, F. Maignan, D. G. Miralles, 587–601. M. F. McCabe, M. Reichstein, J. Sheffield, Hsu, K. L., X. G. Gao, S. Sorooshian, and H. V. Gupta. K. Wang, E. F. Wood, Y. Zhang, and 1997. “Precipitation Estimation from Remotely S. I. Seneviratn. 2013. “Benchmark Products chapter 9 : V alidation of R emote S ensing  D ata   |  183 for Land Evapotranspiration: Landflux-EVAL Su, Z., J. Wen, L. Dente, R. van der Velde, L. Wang, Multi-Data Set Synthesis.” Hydrology and Earth Y. Ma, K. Yang, and Z. Hu. 2011. “The Tibetan Pla- System ­Sciences 17 (10): 3707–20. teau Observatory of Plateau Scale Soil Moisture Mueller, B., S. I. Seneviratne, C. Jimenez, T. Corti, and Soil Temperature (Tibet-Obs) for Quantifying M. Hirschi, G. Balsamo, P. Ciais, P. Dirmeyer, Uncertainties in Coarse Resolution Satellite and J. B. Fisher, Z. Guo, M. Jung, F. Maignan, Model Products.” Hydrology and Earth System M. F. McCabe, R. Reichle, M. Reichstein, ­Sciences 15: 2303–16. M. Rodell, J. Sheffield, A. J. Teuling, Trambauer, P., E. Dutra, S. Maskey, M. Werner, K. Wang, and E. F. Wood. 2011. “Evaluation of F. Pappenberger, L. P. H. van Beek, and S. Uhlen- Global Observations-Based Evapotranspira- brook. 2014. “Comparison of Different Evaporation tion Datasets and IPCC AR4 Simulations.” Estimates over the African Continent.” Hydrology Geophysical Research Letters 38 (6). doi: and Earth System Sciences 18 (1): 193–212. 10.1029/2010GL046230. Wang, K., and R. E. Dickinson. 2012. “A Review of Rittger, K., T. H. Painter, and J. Dozier. 2013. “Assess- Global Terrestrial Evapotranspiration: ment of Methods for Mapping Snow Cover from Observation, Modeling, Climatology, and MODIS.” Advances in Water Resources 51 Climatic Variability.” Reviews of Geophysics 50 (2). (January): 367–80. doi: 10.1029/2011RG000373. Su, Z., P. de Rosnay, J. Wen, L. Wang, and Y. Zeng. 2013. Zipser, E. J., D. J. Cecil, C. Liu, S. W. Nesbitt, and “Evaluation of ECMWF’s Soil Moisture Analyses D. P. Yorty. 2006. “Where Are the Most Using Observations on the Tibetan Plateau.” Intense Thunderstorms on Earth?” Journal of Geophysical Research: Bulletin of the American Meteorological Atmospheres 118 (11): 5304–18. Society 87 (8): 1057–71. 184  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S CHAPTER 10 Validation of Streamflow Outputs from Models Using RS Inputs INTRODUCTION performance of satellite-driven simulated streamflows include (a) bias correction (remov- In recent decades, remote sensing (RS)–derived ing biases) of satellite estimates prior to run- precipitation data have become increasingly ning a hydrologic model and (b) recalibration available at temporal and spatial scales that are of hydrologic models using satellite rainfall useful for hydrologic purposes such as stream- inputs. Calibration of hydrologic models flow monitoring and forecasting, drought fore- accounts for many factors beyond parameter casting, and water resources management values, such as structural model inadequacies, (WRM). While chapter 10 reviews error evalu- availability of spatial and temporal input, and ation and validation efforts of specific satellite errors in input data. Thus in order to obtain products, this chapter focuses on the evaluation optimal simulation results, it is necessary to of errors in streamflow estimates obtained calibrate hydrologic models with the bias- using remotely sensed variables, such as pre- corrected RS estimates that will be used to run cipitation and water surface elevation, in com- such models for monitoring or predictive pur- bination with hydrologic or hydrodynamic poses (Serrat-Capdevila et al. 2013). models. This chapter reviews the errors in stream- Due to the highly nonlinear responses in flow estimates obtained from two approaches: the hydrologic cycle, errors in RS-derived pre- (a) rainfall-runoff modeling using satellite cipitation estimates can be amplified in some rainfall estimates and other meteorological fluxes (evapotranspiration, streamflows) and variables and (b) water-level altimeter dampened in others. Methods to improve the ­ measurements and hydrodynamic models.   185 A synthesis of the results reported in the litera- basins in Australia, Brazil, and the Republic of ture is presented in table 9A.6 (in annex  9A, Korea, ranging in size from 32 to 6,500 square available online at https://openknowledge kilometers. They compare statistics at monthly, .worldbank.org/handle/10986/22952), When 10-day, and daily intervals. They report that evaluating the ­performance of streamflow sim- performance statistics are worse for daily ulations, three main errors have to be analyzed: simulations using satellite products than for ­ the bias (errors in the mean), differences in simulations using rain gauges, except in a few variability (errors in representation of the cases, where both types of simulations some- observed variability), and correlation errors times perform equally well. They also find that (errors in the timing of simulated responses or the magnitude of errors increases as basins get events). These errors can be mathematically smaller. expressed in several metrics (which vary Gourley et al. (2011) evaluate streamflow across the ­ literature). In order to gain an simulations using real-time rain gauges, a denser understanding of the sources and nature of Micronet gauge network, radar (unadjusted and errors in hydrologic simulations, Gupta et al. stage IV, that is, gauge adjusted), Precipitation (2009) decompose the mean squared error Estimation from Remotely Sensed Information into three terms: the error in mean, the error in Using Artificial Neural Networks-Cloud Classi- variance, and the error in correlation: fication System (PERSIANN-CCS), and Tropi- cal Rainfall Measuring Mission-real time MSE 2so1  (so)2  (so)2, (TRMM-RT) (figures 10.1–10.3). They argue (10.1) that recalibration based on potentially biased where s is the mean of the satellite estimates; satellite data would “yield better simulations for o is the mean of the ground observations; s is the wrong reasons” and that it is better to cali- the standard deviation of the satellite esti- brate the model using ground observations of mates; 0 is the standard deviation of the rainfall data. This argument can be reversed ground observations; and  is the correlation easily to defend the use of satellite data rather coefficient (CC) between satellite data and the than ground data for the purpose of calibration, reference observed data. as a model in principle always tends to produce In a study of three satellite precipitation better results when run with the same forcing products (SPPs) over the African continent, data as were used in its calibration. In addition, Serrat-Capdevila et al. (2016) show how bias how many rain gauges are needed to represent correction of satellite precipitation can correct the truth accurately over the entire basin? As errors in the mean and variance terms of pre- calibration addresses a range of issues and cipitation, but not in the correlation term. attempts to compensate for errors in the input data (present in either satellite or ground obser- vations), for optimal performance, hydrologic STREAMFLOW SIMULATIONS models are ideally calibrated with the type of USING RAINFALL-RUNOFF forcing data they will be using in simulations MODELING (Serrat-Capdevila et al. 2014). The framework proposed by Thiemig et al. (2013) demonstrates Tobin and Bennett (2014) use the Tropical this argument and is the one to use. Rainfall Measuring Mission (TRMM) Multisat- PERSIANN-CCS yields streamflows with a ellite Precipitation Analysis (TMPA) 3B42 very small fractional bias (figure 10.1, panel b), (research product, nonreal time) and the but a root mean squared error (RMSE) compa- ­ Climate Prediction Center MORPHing tech- rable to that of TRMM-RT (figure 10.2, panels c nique (CMORPH) to force simulations in 10 and d). The relatively small fractional bias could 186  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S Figure 10.1 Fractional Bias of Streamflow Simulations Forced by Rainfall Algorithms a. Rain gauge adjustments to RS algorithms b. Incorporation of downscaled microwave data in PERSIANN-CCS 150 150 100 100 Fractional bias (%) Fractional bias (%) 50 50 0 0 –50 –50 –100 –100 –150 –150 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 (95) (75) (50) (25) (10) (7.5) (5) (4) (3) (95) (75) (50) (25) (10) (7.5) (5) (4) (3) Discharge threshold (cubic meters per second) Discharge threshold (cubic meters per second) (percentage) (percentage) RT-RESAMPLE Radar PERSIANN-CCS-RT PERSIANN-CCS-MW V6-RESAMPLE Stage IV c. Available real-time, operational rainfall algorithms d. Rescaling reference rainfall to 0.258 resolution of TRMM 150 150 100 100 Fractional bias (%) Fractional bias (%) 50 50 0 0 –50 –50 –100 –100 –150 –150 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 (95) (75) (50) (25) (10) (7.5) (5) (4) (3) (95) (75) (50) (25) (10) (7.5) (5) (4) (3) Discharge threshold (cubic meters per second) Discharge threshold (cubic meters per second) (percentage) (percentage) Gauge Stage IV ARS-RESAMPLE ARS Micronet RT-RESAMPLE PERSIANN-CCS-RT RT-RESAMPLE TRMM-3B42RT V6-RESAMPLE TRMM-3B42V6 Source: Adapted from Gourley et al. 2011. © American Meteorological Society (AMS). Used with permission. Further permission required for reuse. Note: The rainfall algorithms used are indicated in the legend for each panel. Scores are plotted as a function of flow exceedance threshold. be the result of many errors (including false and three hours), but still using rain gauge alarms and missed events) that compensate values as calibration forcing. As a result, the each other when aggregated. Thus it can be MRE improved to −10 from an MRE of −30 accompanied by a high RMSE, which still indi- and −40, depending on the magnitude of dis- cates poor performance. charge (no improvement was seen in the The improved performance, in terms of RMSE or in the bias). The MRE of PERSIANN- Micronet-relative efficiency (MRE)—that is CCS ranged from −17 to −25, ­ performing bet- improvements over using Micronet—for ter than the original runs with 3B42RT TMPA 3B42RT was achieved by recalibrating (before model recalibration at 0.25° and the model at the resolution of TMPA (0.25° three-hour resolution). C hapter 1 0 : V alidation of S treamflo w O u tp u ts from M odels Using R S   I np u ts   |  187 Figure 10.2 RMSE of Streamflow Simulations Forced by Rainfall Algorithms a. Rain gauge adjustments to RS algorithms b. Incorporation of downscaled microwave data in PERSIANN-CCS 120 120 100 100 (cubic meters per second) (cubic meters per second) Root mean squared error Root mean squared error 80 80 60 60 40 40 20 20 0 0 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 (95) (75) (50) (25) (10) (7.5) (5) (4) (3) (95) (75) (50) (25) (10) (7.5) (5) (4) (3) Discharge threshold (cubic meters per second) Discharge threshold (cubic meters per second) (percentage) (percentage) RT-RESAMPLE Radar PERSIANN-CCS-RT PERSIANN-CCS-MW V6-RESAMPLE Stage IV c. Available real-time, operational rainfall algorithms d. Rescaling reference rainfall to 0.258 resolution of TRMM 120 120 100 100 (cubic meters per second) (cubic meters per second) Root mean squared error Root mean squared error 80 80 60 60 40 40 20 20 0 0 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 (95) (75) (50) (25) (10) (7.5) (5) (4) (3) (95) (75) (50) (25) (10) (7.5) (5) (4) (3) Discharge threshold (cubic meters per second) Discharge threshold (cubic meters per second) (percentage) (percentage) Gauge Stage IV TRMM-3B42RT V6-RESAMPLE RT-RESAMPLE PERSIANN-CCS-RT RT-RESAMPLE ARS Micronet TRMM-3B42V6 ARS-RESAMPLE Source: Adapted from Gourley et al. 2011. © American Meteorological Society (AMS). Used with permission. Further permission required for reuse. Note: The rainfall algorithms used are indicated in the legend for each panel. Scores are plotted as a function of flow exceedance threshold. RMSE = root mean squared error. Hossain and Anagnostou (2004) examine prediction error by a factor of three. Extending the impact of passive microwave rainfall these results to short-duration, extreme flood- retrieval frequency and sampling errors on producing storms is one goal of the Gourley flood prediction uncertainty in a medium-size et al. (2011) study. Sangati and Borga (2009) find basin in northern Italy using a semi-distributed that spatial rainfall aggregation has a significant hydrologic model. Regarding temporal sam- effect on simulations of peak discharge for pling frequencies, they find that three-hour extreme flooding events. rainfall retrievals yield similar flood prediction Gourley et al. (2011) also show that seasonal uncertainties as do hourly inputs, but the performance and statistics are not representa- ­ six-hour rainfall retrievals amplify the runoff tive of extreme events: all satellite products 188  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S Figure 10.3 Relative Efficiency of Streamflow Simulations Forced by Rainfall Algorithms a. Rain gauge adjustments to RS algorithms b. Incorporation of downscaled microwave data in PERSIANN-CCS 0 0 Micronet-relative efficiency Micronet-relative efficiency –10 –10 –20 –20 –30 –30 –40 –40 –50 –50 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 (95) (75) (50) (25) (10) (7.5) (5) (4) (3) (95) (75) (50) (25) (10) (7.5) (5) (4) (3) Discharge threshold (cubic meters per second) Discharge threshold (cubic meters per second) (percentage) (percentage) RT-RESAMPLE Radar PERSIANN-CCS-RT PERSIANN-CCS-MW V6-RESAMPLE Stage IV c. Available real-time, operational rainfall algorithms d. Rescaling reference rainfall to 0.258 resolution of TRMM 0 0 Micronet-relative efficiency Micronet-relative efficiency –10 –10 –20 –20 –30 –30 –40 –40 –50 –50 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 >0.5 >0.8 >1.4 >7.0 >10.0 >12.7 >15.7 >17.6 >21.1 (95) (75) (50) (25) (10) (7.5) (5) (4) (3) (95) (75) (50) (25) (10) (7.5) (5) (4) (3) Discharge threshold (cubic meters per second) Discharge threshold (cubic meters per second) (percentage) (percentage) Gauge Stage IV TRMM-3B42RT V6-RESAMPLE RT-RESAMPLE PERSIANN-CCS-RT RT-RESAMPLE ARS-RESAMPLE TRMM-3B42V6 Source: Adapted from Gourley et al. 2011. © American Meteorological Society (AMS). Used with permission. Further permission required for reuse. Note: The rainfall algorithms used are indicated in the legend for each panel. Scores are plotted as a function of flow exceedance threshold. perform very poorly for a 500-year extreme Thiemig et al. (2013) evaluate simulations in event. The ranking of simulation performance two African basins (Volta and Baro-Akobo) and from the seasonal analysis actually reverses for four sub-basins with various SPPs using the the extreme event. They do not address why Kling-Gupta efficiency (KGE), an error metric recalibrating the model—by aggregating the that combines three components—bias, vari- reference data set to the resolution of the satel- ability, and correlation (Kling, Fuchs, and lite product—improves the simulations. This Paulin 2012)—as follows: ­ is likely due to the averaging of errors and parameters in the coarser resolution of the sat- KGE  1  (r  1)r  (  1)2  (  1)2 ellite product.  (10.2) C hapter 1 0 : V alidation of S treamflo w O u tp u ts from M odels Using R S   I np u ts   |  189 s 5. Running simulations with bias-corrected  o (10.3) SPP-specific calibration (to determine the CVs s/ s combined benefits of bias-correction and   CV , (10.4) o o/ o SPP-specific recalibration). where s is the mean of the satellite estimates; In the lowlands, performance is good or o is the mean of the ground observations, s is intermediate for African Rainfall Estimation the standard deviation of the satellite estimates; Algorithm Version 2 (RFE2) and TRMM, but o is the standard deviation of the ground poor for CMORPH and PERSIANN. In the observations; r is the linear correlation (Pear- mountainous basin, CMORPH performs bet- son product-moment) coefficient between sat- ter. Most of the poor and very poor perfor- ellite data and the reference observed data;  mance can be attributed to bias and variability the bias ratio; and  is the variability ratio (errors in mass balance and shape of between the coefficients of variation, CV. distributions). They use a structured approach to bench- CMORPH and PERSIANN clearly benefit mark improvements in the simulation’s perfor- strongly from both bias correction and model mance in those two basins: recalibration, and these processes correct mainly the bias term. In other words, CMORPH 1. Calibrating the hydrologic model with and PERSIANN contain significant biases that interpolated rain gauge data can be corrected. Bias correction is more effec- 2. Running SPP simulations with reference tive than recalibration (which is unable to cor- (gauge) calibration for each satellite prod- rect mass balance) at correcting products with uct (to determine the intrinsic value of raw large biases and yielding improved simula- SPPs) tions. For products without large biases, model recalibration yields more significant improve- 3. Running simulations with SPP-specific ments than bias correction, an intuitive result. calibration (to determine the value of reca- Finally, as recommended by Serrat-Capdevila libration and raw SPPs) et al. (2013), the combined use of bias correc- 4. Running simulations (with reference cal­ tion and recalibration of hydrologic models ibration parameters) for bias-corrected with bias-corrected SPP data yields the best SPPs possible performance. In a flow-forecasting system for the Yellow River—a Sino-Dutch cooperation project— Rosema et al. (2008) use RS data (from hourly BOX 10.1 visual and thermal infrared bands) for river Validation of Streamflow Simulations Using   basin management, including energy and Rainfall-Runoff Modeling water balance, drought monitoring, and flow and flood forecasting. This large project has • The performance of hydrologic applications using RS data can be custom-made satellite precipitation retrievals highly variable, depending on basin size, geography, topography, and modeling as well as a forecasting system. and storm systems. The means employed for this project probably • Hydrologic simulations will generally yield better results if RS in- go beyond the resources of most similar stud- puts are bias corrected (if they contain biases) and if the hydro- ies. The water resources forecasting system logic models are recalibrated with the same type of input data that (flow forecasting) yields correlations of 0.8 to will be used in these models for predictive purposes. 0.94 for the sub-basins, with Nash-Sutcliffe efficiency (NSE) coefficients of 0.77 to 0.84.1 190  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S Their high-water forecasting system (flood BOX 10.2 forecasting) yields correlations of 0.75 to 0.80 (slightly lower than for flow forecasting) and Validation of Streamflow Simulations Using Remotely NSEs of 0.71 to 0.79 (see table 9A.6, available Sensed Water Levels Upstream online). Box 10.1 summarizes the findings for vali- Estimating water levels in large river basins using altimeters is a more direct dating streamflow simulations using rainfall- way to monitor stage height and streamflow (and make predictions down- stream) than using only basin-wide rainfall-runoff models, which can be runoff modeling. particularly complex in the case of very large basins. An operational forecast system based on satellite surface water altim- etry to drive flow propagation models has extended forecast lead times in Bangladesh from 3 days to 8 or 10 days, with an RMSE of 0.7 meter at the STREAMFLOW SIMULATIONS India-Bangladesh border. BASED ON REMOTELY SENSED WATER LEVELS UPSTREAM Flood-prone developing countries usually lack the in situ hydrologic data necessary to imple- RMSE values ranging from 0.2 to 0.7 meter at ment flood forecasting systems. In the case of selected river stations (Hossain, Siddique-E- transboundary basins, downstream countries Akbor, Mazumder et al. 2014). Currently, a fore- are usually “blind” to what is happening in the cast system with lead times of 8 to 10 days has upper part of the basin, because of the lack of shown an RMSE of 0.7 meter at the India-Ban- international cooperation and scarcity of gladesh border (Hossain, Siddique-E-Akbor, ground observation networks. Yigzaw et al. 2014). These results indicate that For instance, presently there is no mecha- countries with large transboundary rivers could nism for the B­ angladeshi government to receive implement operational forecast systems with timely information on upstream conditions of currently available and planned altimeter mis- the Ganges-Brahmaputra basin. Stream mea- sions to manage water risks in flood-prone surements at the borders where the rivers enter regions. The authors argue that satellite radar the country only allow the Bangladeshi govern- altimetry is probably more valuable in large riv- ment to forecast water levels downstream with ers than rainfall-runoff simulations using satel- a lead time of two to three days at most. Recent lite precipitation estimates to anticipate the work by Hossain, Siddique-E-Akbor, Mazum- occurrence of high-water conditions in the der et al. (2014) and by Hossain, Siddique-E- basin (see box 10.2). Akbor, Yigzaw et al. (2014) has demonstrated the feasibility of implementing an 8- to 10-day ahead water-level forecast system in the Brah- NOTE maputra River basin using Jason-2 estimates. Measurements of river surface levels upstream 1. The Nash-Sutcliffe coefficient of efficiency is used to assess the performance of hydrologic models in in India and a hydrodynamic model (the Hydro- replicating observed streamflows and is defined as logic Engineering Centers River Analysis Sys- follows: tem, HEC-Ras) are used to predict how the T observed water levels upstream will propagate t  1(Qo t t 2   Qm)  , E1  T to areas downstream. In operational forecasts t1 t (Qo  Qo)2 during the high-flow season of August 2012, a where Qo is the mean of observed discharge, Qm is five-day water-level forecast system had aver- modeled discharge, and Qto is observed discharge age errors ranging from −0.4 to 0.4 meter, with at time t. C hapter 1 0 : V alidation of S treamflo w O u tp u ts from M odels Using R S   I np u ts   |  191 REFERENCES Rosema, A., M. De Weirdt, S. Foppes, R. Venneker, S. Maskey, Y. Gu, W. Zhao, C. Wang, X. Liu, S. Rao, D. Dai, Y. Zhang, L. Wen, D. Chen, Y. Di, S. Qiu, Gourley, J. J., Y. Hong, Z. L. Flamig, J. Wang, Q. Wang, L. Zhang, J. Liu, L. Liu, L. Xie, R. Zhang, H. ­Vergara, and E. N. Anagnostou. 2011. J. Yang, Y. Zhang, M. Luo, B. Hou, L. Zhao, “­Hydrologic Evaluation of Rainfall Estimates L. Zhu, X. Chen, T. Yang, H. Shang, S. Ren, from Radar, Satellite, Gauge, and Combinations on F. Sun, Y. Sun, F. Zheng, Y. Xue, Z. Yuan, H. Pang, Ft. Cobb Basin, Oklahoma.” Journal of Hydrome- C. Lu, G. Liu, X. Guo, D. Du, X. He, X. Tu, W. Sun, teorology 12 (5): 973–88. B. Bink, and X. Wu. 2008. Satellite Monitoring Gupta, H. V., H. Kling, K. K. Yilmaz, and G. F. Marti- and Flow Forecasting System for the Yellow River nez. 2009. “Decomposition of the Mean Squared Basin. Sino-Dutch Cooperation Project ORET Error and NSE Performance Criteria: Implica- 02/09-CN00069 Scientific Final Report. Delft, tions for Improving Hydrological Modeling.” Netherlands: EARS Earth Environment the ­ Journal of Hydrology 377 (1-2): 80–91. Monitoring BV, December. ­ Hossain, F., and E. N. Anagnostou. 2004. “Assessment of Sangati, M., and M. Borga. 2009. “Influence of Rainfall Current Passive-Microwave- and Infrared-Based Spatial Resolution on Flash Flood Modelling.” Satellite Rainfall Remote Sensing for Flood Predic- Natural Hazards and Earth System Sciences 9 (2): tion.” Journal of Geophysical Research: Atmospheres 575–84. doi:10.5194/nhess-9-575-2009. 109 (D7). Republished with an errata in 2005. Serrat-Capdevila, A., J. B. Valdes, and E. Z. Stakhiv. Hossain, F., A. H. Siddique-E-Akbor, L. C. Mazumder, 2014. “Water Management Applications for Satel- S. M. ShahNewaz, S. Biancamaria, H. Lee, and lite Precipitation Products: Synthesis and Rec- C. K. Shum. 2014. “Proof of Concept of an ommendations.” Journal of the American Water ­ Altimeter-Based River Forecasting System Resources Association 50 (2): 509–25. doi: 10.1111/ for Transboundary Flow Inside Bangladesh.” IEEE Journal of Selected Topics in Applied Earth jawr.12140. Observations and Remote Sensing 7 (2): 587–601. ­ Serrat-Capdevila, A., M. Merino, J. B. Valdes, and M. Hossain, F., A. H. M. Siddique-E-Akbor, W. Yigzaw, Durcik. 2016. “Evaluation of the Performance S. Shah-Newaz, M. Hossain, L. C. Mazumder, T. of Three Satellite Precipitation Products over Ahmed, C. K. Shum, H. Lee, S. Biancamaria, F. J. Africa.” University of Arizona, submitted to Turk, and A. Limaye. 2014. “Crossing the Valley Atmospheric Research. of Death: Lessons Learned from Implementing Thiemig, V., R. Rojas, M. Zambrano-Bigiarini, and an Operational Satellite-Based Flood Forecasting A. De Roo. 2013. “Hydrological Evaluation of System.” Bulletin of the American Meteorological Satellite-Based Rainfall Estimates over the Volta Society 95 (8): 1201–07. and Baro-Akobo Basin.” Journal of Hydrology 499 Kling, H., M. Fuchs, and M. Paulin. 2012. “Runoff (August): 324–38. Conditions in the Upper Danube Basin under an Tobin, K. J., and M. E. Bennett. 2014. “Satellite Pre- Ensemble of Climate Change Scenarios.” Journal cipitation Products and Hydrologic Applications.” of Hydrology 424-425 (March 6): 264–77. Water International 39 (3): 360–80. 192  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S CHAPTER 11 The Bottom Line On the basis of the literature review of the hydrologic models have been recalibrated state-of-the-art of remote sensing (RS) for with the same type of input data that will be hydrologic simulations, several tentative con- used in these models for predictive purposes. clusions may be drawn regarding the use of • Accuracy and performance vary depend- Earth observation (EO) to support water ing on climate, topography, the variable management applications: estimated, time aggregation, and basin • Satellite estimations are prone to several size. The tables presented in annex 9A sources of uncertainty, which can signifi- available online at https://openknowledge (­ cantly affect the quality of the variables to .worldbank.org/handle/10986/22952) give be forecast. a good idea of the most suitable RS prod- ucts for hydrologic applications, the con- • To inform natural resources managers texts in which they can be most useful, and about the usefulness of RS products for when more caution is warranted in the face their decision-making processes, it is of greater uncertainties. imperative to evaluate the reliability of those products at different spatiotemporal Finally, satellite estimations are overall scales. well correlated with ground observations (­figure 11.1)—showing median correlation • RS data in case studies and applications coefficient (CC) values of 0.55 (±0.25) for should always be used with ground data precipitation, 0.83 (±0.17) for evapotranspira- ­ when available and with some level of tion, 0.58 (±0.19) for soil moisture, and 0.53 validation. (±0.21) for snow water equivalent.1 Precipita- • Hydrologic simulations generally yield tion shows a broad range of CC values, which better results if RS inputs have been bias can be attributed to differences in the valida- corrected (if they contain biases) and the tion sites (which include mountain and plain   193 ­ l ocations distributed across the globe). observations. Despite the strong correlation Evapotranspiration and soil moisture valida- found for ET estimates, biases in the range of tion efforts are limited to fewer validation −30 to +10 percent (relative to observed sites and to field experiments, which can ­ values) have been reported in the literature partly explain the better linear relationship (see the section on evapotranspiration in between satellite-derived products and chapter 10). Similarly, for soil moisture, mean root mean squared estimate values of 0.11 (±0.09) cubic meter by cubic meter and a Figure 11.1 Correlation Coefficients between Ground Observations and Satellite Estimates mean positive bias of 0.04 (±0.05) cubic meter by cubic meter have been calculated (see the 1.0 section on soil moisture in chapter 9), indicat- 0.8 ing that, despite strong correlations, the uncertainty of satellite estimates is still large. Correlation coefficient (CC) 0.6 Snow water equivalent estimates also show a 0.4 large variability, with bias values ranging from −20 to +20 percent (see the section on snow 0.2 water equivalent in chapter 9). 0.0 –0.2 NOTE –0.4 Precipitation Evapotranspiration Soil moisture Snow water 1. Error margins given in parentheses refer to stan- (n = 30) (n = 22) (n = 47) equivalent dard deviation values. These cannot be extracted (n = 15) from the corresponding tables but were computed Note: n = sample size. The black horizontal line represents the median value. separately. 194  |  P A R T I I I : V A L I D A T I O N O F R E M O T E S E N S I N G – E S T I M A T E D H Y D R O M E T E O R O L O G I C A L V A R I A B L E S PART IV Concluding Remarks WATER AND DEVELOPMENT WRM. One key reason for this appears to be the lack of familiarity among the WRM community with Good water resources management (WRM) and available EO products and the ways in which they planning are essential to sustain economic and can be used to address WRM issues. This publication human development as well as to maintain the reviews the state of the art in the use of remote sens- health of the socioecological systems of which ing (RS) for water resources applications, guided by humans are a part. Especially in developing nations, the general scope and requirements of the World water supply and sanitation and a healthy environ- Bank’s Water Global Practice. ment form the basis of successful poverty reduction Important topics like water supply for rural or strategies. With that ultimate goal in mind and to urban water users, sanitation and hygiene, agricul- face other global water resources challenges, con- tural water management, WRM and environmental tributions are needed to bridge the gap between services, and hydropower can be informed by eight existing technologies and operational applications variables that may contribute to and modify water in support of the planning, design, operation, and resources management. These variables are precipi- management of water resources. tation, evapotranspiration, soil moisture, vegetation and vegetation cover, groundwater, surface water, snow and ice, and water quality. An understanding POTENTIAL OF REMOTE SENSING of these eight biophysical parameters as well as of There is great potential for space-based Earth the theoretical basis for their estimation through observation (EO) to enhance the capability to moni- Earth observation is important. Equally important is tor the Earth’s vital water resources, especially in to have a list of current and near-future sensors data-sparse regions of the globe. Despite this poten- that  can provide such information, indicates their tial, EO data products are currently underused in suitability for water resources management, and,   195 where  appropriate, describes existing data can work toward characterizing in detail the products that are produced on a regular basis. climate- and water-sensitive decisions in their planning and management. CHALLENGES A WORD OF CAUTION Given the dynamic nature of Earth observa- tion, it is no less important to have a way to Despite its limitations, the literature review keep this list up to date. The number of EO presented in this publication reflects the state applications is growing as rapidly as the num- of the art of remote sensing for hydrological ber of new, space-based technology, satellite purposes. On that basis, several statements can missions, and data products. In addition, EO be made regarding the use of Earth observation sensors are becoming more sophisticated, to support water management applications: more sensitive, and more agile (as illustrated • It is imperative to evaluate the validity of by on-demand programming for image acqui- RS data products at different spatiotempo- sition from commercial, high-resolution ral scales if they are to be of use for decision ­ sensors). The algorithms that translate top-of- making. atmosphere EO data to ground-level informa- tion are evolving rapidly. • Validity and performance vary depending on Field measurement systems are also becom- climate, topography, the variable being esti- ing more sophisticated as new information mated, time aggregation, and basin size. It is technology, telemetry, and sensing solutions good to know the contexts in which RS data are developed. Moreover, methods to integrate can be most useful and when to be particu- observations and models through model-data larly alert to greater uncertainties than usual. fusion are being developed rapidly. While this • Satellite estimations are generally well cor- bodes well for the usefulness of Earth observa- related with ground observations. Despite tion for water resources management, it also these strong correlations, however, the means that some of the information in this uncertainty of some satellite estimates may publication will become outdated over the still be large (but that may also be the case next few years. The reader may therefore still of ground measurements). need to seek advice from area experts on the most recent developments and solutions. MAKING DECISIONS Numerous reports and publications on hydrologic applications of remote sensing The decision whether to use Earth observation focus on the tools (products and models), but to address a spatiotemporal information few publications focus on the needs of the requirement should be based on criteria regard- practitioners and the characteristics of the ing the accuracy, availability, maturity, com­ decisions that such tools could be informing. plexity, and reliability as well as the validity of Hence there is a great gap in the adoption of required data. The suitability of using Earth such tools by practitioners. To some extent, observation for addressing a WRM need will this is normal, as it is difficult to incorporate also depend on whether it is the only source of new, uncertain information into a decision data (in which case, the suitability of Earth process, especially if neither the uncertainty observation is clear); whether EO information nor the reliability of the source is well quanti- augments existing, but sparse, in situ informa- fied. While scientists and providers can work tion (in which case Earth observation will still toward including uncertainty and reliability be a critical source of information, providing the estimates, practitioners in developing regions spatiotemporal framework for maximizing the 196  |  E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E M A N A G E M E N T value of existing information); whether other From the satellite-sensor point of view, relevant data exist; and whether Earth observa- much coordination takes place via the Commit- tion is needed mainly for its spatiotemporal tee on Earth Observing Systems, whereas the aspects (in which case Earth observation will Group on Earth Observations plays a global only add value if its relevance, coverage, and coordinating role for the end users of this infor- accuracy significantly improve the information mation. Other agencies with a need for EO- derived from in situ data). derived information at multiple resolutions having global coverage are, for example, the A WORD OF HOPE United Nations Environmental Programme, the World Health Organization, and the Food and Some organizations like the World Bank have Agriculture Organization; coordination with funded or supported projects using EO infor- these organizations could be highly effective. mation. Through their water resources proj- ects, these organizations could potentially be among the world’s largest adopters of remote DOWNSCALING TO THE LOCAL sensing in water resources management. They CONTEXT may want to consider whether a coordinated Given these considerations and the information approach to remote sensing for WRM applica- gathered for this publication, the following are tions could increase the effectiveness and effi- suggested for helping developing-country prac- ciencies in executing their projects. For titioners to bridge the gap between scientific- example, if multiple projects involve similar academic and real-world uses of RS technology: applications, EO data sources, and EO tech- niques, it may be possible to use available 1. Technical support for mainstreaming the resources more efficiently by developing a sin- knowledge on how to make the best pos- gle data infrastructure (for example, for an sible use of remote sensing as a tool for entire region or transboundary basin). Simi- the water sector in particular. larly, if the same data are required repeatedly 2. Technical orientation and definition of in projects or if monitoring applications are clear procedures and criteria to assess considered, it may be worthwhile to develop a the usability of RS products for decision centralized data infrastructure that keeps such making and planning conditioned by data sets up to date. Some of the options could uncertainty (error estimates), accuracy be to establish in-house EO capabilities or to (characterization of errors), precision partner with institutions or consortia that have (spatial and temporal resolution), timeli- regional or global outreach. ness, and validity of the data. This could A phased approach—where several specific include the quality and quantity of data EO applications are chosen in an area of fre- generated to fill the information gaps, quent WRM activity and are developed in a whether the information gathered has generic manner that may subsequently be rep- been validated or calibrated, the resolu- licated elsewhere—is a possible pathway to tion used, or the level to which remote widespread uptake and implementation. Such sensing has significantly influenced judicious planning of demonstration projects, project performance. involving areas with a clear need for EO- derived information across relevant WRM 3. Knowledge about errors and uncer- application areas, possibly together with other tainty. If products are used as inputs for relevant agencies, could create synergy while modeling applications, it is important rapidly strengthening this area of activity. to know how errors are propagated or P A R T I V : C onc l u d ing R emar k s   |  197 compounded through model calcula- • What improvements in the hydrome- tions and what uncertainty is contained teorological information being used— in the output variables. or what new information—would make the greatest positive difference? 4. Technical orientation on reliability assessments of applications, including • What changes in decision making EO estimates. Given a specific appli- (decision thresholds and decision cation designed to model or predict a process) would result if such new variable, reliability evaluations should information were available? be performed to see in how many simu- • What changes in the institutional lated historical events the observations framework would be needed to obtain fell within the uncertainty bounds of the and be able to use this new informa- application’s predictions. Such a reliabil- tion and to make these changes in ity assessment would allow for improved decision making? characterization of the application’s limitations. 6. Financial support windows for special- 5. A good characterization of the planning ized technical assistance to individual and decision processes to be informed by projects or groups of projects. This sup- RS data and applications. If new develop- port could include financial support for ments bring about new decision-making the development of specific applications processes, these processes should like- that could benefit many projects in the wise be well characterized. Starting from a same or similar regions. comprehensive, basin-wide development 7. Financial and institutional support for plan, identify the specific management data repositories and RS data libraries and planning decisions to be made and of different products with potentially then characterize the climate- or water- built-in applications, for easy use by sensitive decisions that RS products could project teams. inform and what the benefits of such infor- mation would be. The following questions 8. Practical guidelines—such as those may be relevant for that purpose: offered in this publication—from the user’s point of view. These guidelines • What are the climate- or water-­ would include when in situ measure- sensitive management decisions that ments and RS applications would be ­ client country ministries, depart- more operationally advantageous and ments, and agencies are confronted when value would be added by using with in their water management and one as a complement to the other (RS as planning cycles? complementary to in situ measurements • What degree of accuracy and preci- or vice versa), taking into account their sion is required in each of these deci- relevance, availability, level of detail, and sions, and how much uncertainty can the accuracy required as well as develop- be tolerated? ing countries’ capabilities. • What are the consequences of a mis- 9. Financing mechanisms for in-country take caused by faulty data and what capacity building to improve decision kind and degree of failure can be making and better characterize decisions. tolerated? This would help to identify the value of 198  |  E A R T H O B S E R V A T I O N F O R W A T E R R E S O U R C E M A N A G E M E N T potential data and their relevance for a the co-production of knowledge by scholars specific decision-making activity. and practitioners. This publication provides a guide to WRM professionals considering the use of Earth OUTLOOK observation. Essential questions are provided A good understanding of the answers to these that must be answered to help to navigate and questions can inform the design of special tools evaluate the abundance of EO-based options for specific purposes. As new information and data products, including the likely validity becomes available, it may give new insights of water resource variables estimated though into how to apply this information in practice Earth observation. The focus is on appropriate within a specific management and planning questions to ask once it has been concluded setting. Thus communication between scien- that exploring EO options for the WRM prob- tists, researchers, and practitioners should be a lem at hand is worthwhile. A flowchart pre- two-way street. sented in chapter 7 offers a “road map” for this The World Bank and other development purpose (figure 7.1). banks, United Nations agencies such as the It is hoped that the information collected World Meteorological Organization, and other in this publication will contribute to a greater international entities could play a role in clos- and more judicious use of EO data in global ing the gap between science application efforts WRM issues, thereby helping to alleviate pov- and operational decision-making needs. In erty, promote sustainable growth, and addition, they could promote and facilitate increase the efficient use of the world’s water data sharing, capacity-building strategies, and resources. P A R T I V : C onc l u d ing R emar k s   |  199 APPENDIX A Examples of Earth Observation Applications in World Bank Projects P050647 UTTAR PRADESH WATER SECTOR RESTRUCTURING P122770 UTTAR PRADESH WATER SECTOR RESTRUCTURING PHASE 2 (POTENTIAL USE OF REMOTE SENSING) PROJECT DETAILS Team task leaders Winston Yu, Anju Gaur Contact Winston Yu, Anju Gaur Status P050647 (2001–11; closed) P126703 (2012–20; active) Description/objectives P050647: To set up an enabling institutional and policy framework for water sector reform in Uttar Pradesh State for integrated water resources management and to initiate irrigation and drainage subsector reforms to increase and sustain water and agricultural productivity in the state. P122770: To strengthen the institutional and policy framework for integrated water resources management for the entire state and to increase agricultural and water productivity by supporting farmers in targeted irrigation areas. Project component related to In past projects, evaluating project performance has been weak, and some level of monitoring and remote sensing evaluation has been required, leading to the need to adjust the project design during implementation. Funds were provided to recruit third-party expertise for the monitoring and evaluation of each component of the project (P050647). As a result, benchmarking, remote sensing, geographic information system (GIS), and participatory monitoring and evaluation were carried out for different components of the project. Baseline data collected during preparation and implementation of each specific intervention were used to assess project impact through the collection and analysis of similar information at specific points in time during the project period and, eventually, in other project areas. Use of remote sensing These projects use remote sensing (RS) tools to establish a strong monitoring and evaluation system in order to assess project progress and impact. The use of remote sensing in P050647 was intended to produce a geospatial evaluation tool to assess the progress and impact of agricultural and water projects supported by the World Bank in India. In this case, a selected pilot study was carried out in the Jaunpur Branch System to serve as a benchmark for other project areas. Window/initiative Not applicable   201 REMOTE SENSING INFORMATION Input (data type, source, 1. Landsat data. Images for the study area were identified and acquired through the Landsat Program resolution, etc.) website to complement overall project objectives of distributing a multitemporal, multispectral, and multiresolution range of imagery appropriate for irrigation impact analysis. Due to cloud cover in the region, additional sensor data sets—Landsat Thematic Mapper (TM), Landsat Enhanced Thematic Mapper Plus (ETM+), and Global Land Surveys (GLS)—were used to fill the scanline gaps. 2. MODIS (Moderate Resolution Imaging Spectroradiometer) data. Daily global imagery provided spatial resolutions of 250-meter (red and NIR1) and 500-meter (blue, green, NIR2, SWIR1, and SWIR2). The MDOO9A1 data sets in 2000 to 2010 were acquired from the U.S. Geological Survey’s Earth Resources Observation Systems Data Center website. The following three indexes were calculated for each MODIS eight-day composite: (1) normalized differential vegetation index (NDVI), (2) enhanced vegetation index (EVI), and (3) land surface water index (LSWI) using surface reflectance values from the blue, green, red, NIR1, and SWIR bands.  Model (source, variables, MODIS time-series analysis. The vegetation phenological analyses were calculated using the seasonal selection criteria) dynamics of the three indexes—EVI, NDVI, and LSWI—from 2000 to 2010. The analysis included cropping intensity (number of crops per unit area in a year), length of growing season, and beginning and ending of the growing season. For identifying multiple cropping cycles in an image pixel, the temporal profile of the indexes was analyzed by applying a computational algorithm to all of the individual pixels for delineating the number of cropping cycles in a year. Mapping multiple cropping areas. Multiple cropping areas were assessed for a regular calendar year (January–December). Given the nature of the cropping season in India and monsoon patterns, the cropping calendar was remapped from July to June for 11 years starting in 2000–01 at 500-meter spatial resolution. Annual vegetation anomalies. Annual vegetation anomalies were calculated by subtracting the annual mean NDVI from the long-term mean (2000–10). The main objective was to see the dynamics of cropland vegetation at annual intervals compared to the long-term average. Land use/land cover change analysis. Advanced Wide Field Sensor (AwiFS) based on land use maps was used to quantify the change in land use that occurred between 2004–05 and 2008–09. The change in area was further analyzed at head reach, middle reach, and tail ends to see changes at each distribution. Crop intensity. Crop intensity was estimated before and after project implementation as follows: cropping intensity = (gross cropped area / net sown area) x 100. Higher cropping intensity means that a higher portion of the net area is being cropped more than once during one agricultural calendar year. This also implies higher productivity per unit of arable land during one agricultural calendar year. Dynamics of the crop phenology. Satellite images were used from the MODIS sensor. For each eight-day composite image, the EVI and LSWI were calculated using surface reflectance values from the blue, red, near infrared (NIR, 841–875 nanometers), and shortwave infrared (SWIR, 1,628–1,652 nanometers) bands. The MODO9A1 files include quality control flags to account for various image artifacts (for example: clouds, cloud shadow). In addition, blue band reflectance was used to eliminate further contaminated observations (such as clouds, aerosols). Annual maximum values of EVI were selected for pixels from all of the remaining good observations in a year, and the dates for annual maximum EVI and LSWI were recorded. The study used seasonal maximum values of EVI and LSWI (magnitude) and date of seasonal maximum vegetation index (timing) as a measure for crop phenology. Output (results: maps, • Mapping of multiple cropping areas (single, double, and triple) for 2004–05 and 2008–09 indexes, etc.) • Mapping of annual vegetation anomalies from the long-term mean (2000–10) and spatial distribution of vegetation dynamics • Analysis of land use and land cover change • Identification of differences in crop intensity between 2004–05 and 2008–09. Collaboration A partnership was formed between the Uttar Pradesh Irrigation Department and the Remote Sensing Agency in Uttar Pradesh. 202  |  appen d i x A : E x amp l es of E arth O bservation A pp l ications in W or l d B an k P rojects Outcome relevant to the Potential for using remote sensing from the activity described above: objective of the component • RS-based analysis indicates that project intervention improved the vegetation health and distribution across the basin, which is a good indicator of increased productivity. • The methodology was overseen by a Bank team and proven useful to the client. • A partnership was formed between the Uttar Pradesh Irrigation Department and the Remote Sensing Agency in Uttar Pradesh. • This work led to an actual component in a new operation that will use this methodology. • The approach developed in this pilot study, though data intensive, is efficient with respect to the amount of fieldwork that would be required to do similar analysis. • Such a methodology can be replicated easily in other operations. • As in this case, RS methodologies can be an effective approach (especially when using free RS data) to monitoring agricultural performance in large geographic areas and potentially be mainstreamed into monitoring and evaluation approaches for irrigation projects. P114949 ZAMBIA WATER RESOURCES DEVELOPMENT P117617 SHIRE RIVER BASIN DEVELOPMENT PROJECT P104446 MALAWI DISASTER RISK REDUCTION AND RECOVERY PROJECT P102459 ZAMBIA IRRIGATION DEVELOPMENT AND SUPPORT PROJECT PROJECT DETAILS Team task leaders Marcus Wishart, Pieter Waalewijn, Kremena M. Ionkova, Indira Ekanayake Contact Marcus Wishart, Pieter Waalewijn, Kremena M. Ionkova, Indira Ekanayake, Nagaraja Harshadeep Status P114949 (2013–18; active) P117617 (2012–18; active) P104446 (2007–10; closed) P102459 (2010–18; active) Description/objectives P114949: To support the implementation of an integrated framework for development and management of water resources in Zambia. P117617: To generate sustainable social, economic, and environmental benefits by effectively and collaboratively planning, developing, and managing the Shire River basin’s natural resources. P104446: To increase yields per hectare and volume of products marketed by smallholders benefiting from investments in irrigation in selected sites served by the project. P102459: To increase yields per hectare and volume of products marketed by smallholders benefiting from investments in irrigation in selected sites served by the project. Project component related to These projects use Earth observation (EO) tools to map small water bodies in Zambia, assess water quality in remote sensing Lake Malawi, and assess erosion patterns in some areas in the Shire River basin of Malawi. Satellite Earth observation has added value to the task of making inventories of small water bodies, which are often sources of irrigation water for rural communities in Zambia. The network of ground measurements and inventories of these water bodies are often incomplete, sparse, or difficult to maintain. Conversely, the use of EO tools allowed the mapping of small reservoirs, which made more efficient use of existing ground measurements. This component focused on rural communities that will benefit from improved small-scale water resources infrastructure and basin planning. In Lake Malawi and nearby lakes Malombe and Chilwa, the existing ground data for assessing water quality are limited and inadequate. For example, it is critical to assess the sediment loads in these lakes and rivers accurately since high sediment loads have caused problems for hydroelectric power stations in the past. However, based on existing ground measurements, it is difficult to assess the hydrologic status of the basin. On the contrary, information derived from Earth observation can supplement ground measurements to improve watershed management in some catchments of the lake. appen d i x A : E x amp l es of E arth O bservation A pp l ications in W or l d B an k P rojects   |  203 Soil erosion has been a major concern in Malawi due to population growth, deforestation, and development of new settlements. EO information provides accurate and up-to-date information on land use and changes in land use in order to optimize planning for water resource investments, flood mitigation, and watershed management in selected catchments of the Shire basin. The experience of these projects shows that EO information and modern satellite products can be used to find innovative approaches to prioritize investments. Use of remote sensing • Identification, mapping, and cataloguing of small-scale water bodies, reservoirs, and lake extensions based on SAR (synthetic aperture radar) data, including their evolution over time • Production of Lake Malawi water quality products, including lake surface temperature measurements as well as historical water-level records • Estimation of soil loss and erosion using very high-resolution optical data (SPOT5) from 2005 to 2010. Window/initiative EOWorld, TigerNET REMOTE SENSING INFORMATION Input (data type, source, Small reservoir mapping (Zambia). Landsat, Advanced Synthetic Aperture Radar (ASAR) imagery. resolution, etc.) Monitoring of Lake Malawi. Envisat- Medium Resolution Imaging Spectrometer (MERIS) data were used to evaluate key water quality parameters, including chlorophyll-a as a proxy for biomass, total suspended matter concentrations and Kd (attenuation coefficient) as a proxy for turbidity and transparency, and colored dissolved organic matter as a proxy for the presence of humic substances. Shire River basin. The estimation of soil loss and erosion within Malawi’s Shire River basin was based on SPOT5 acquisitions from 2005 to 2010 and covered 10,798 square kilometers in 17 land use classes. Model (source, variables, Lake Malawi water quality: BEAM (Basin Economic Allocation Model); WISP (Water Information System selection criteria) Platform) Output (results: maps, indexes) • Identification and mapping of small reservoirs and assessment of relevant storage evolution over time • Land cover and land use maps and deforestation rates • Erosion maps • Water quality maps. Collaboration European Space Agency, Netherlands Geomatics and Earth Observation B.V. (Netherlands), Technical University of Delft (Netherlands), Water Insight (Netherlands) Outcome relevant to the EO information was used to assist the prioritization of investments, the monitoring of lakes and basins, basin objective of the component planning for water resource investments, flood mitigation and risk reduction, and watershed management in selected catchments. 204  |  appen d i x A : E x amp l es of E arth O bservation A pp l ications in W or l d B an k P rojects APPENDIX B Examples of Water Information Product Generation Systems INTRODUCTION uses Moderate Resolution Imaging Spectrome- ter (MODIS) 250-meter data to map surface The main text highlights the data needed to water areas and compare them with historical address the most pressing water issues in the imagery to detect flood occurrence. The Dart- developing world and explains where and mouth Flood Observatory also uses time series how Earth observation (EO) can help to pro- of passive microwave daily observations at vide this information. This appendix lists and selected locations to estimate river discharge briefly describes some water information sys- (River Watch). In each of these locations, an tems that are notable examples of the integra- empirical linear model has been fitted between tion of ground observations, EO data, and observed discharge and the passive microwave models. This is a subjective selection, not a signal, providing an estimate of discharge in comprehensive list of all existing systems. near real time when floods occur. The service allows users to access historical events. Using the same algorithms as those devel- FLOOD WARNING AND oped by the Dartmouth Flood Observatory, the MONITORING SYSTEMS National Aeronautics and Space Administra- tion (NASA) has implemented a near-real-time Dartmouth Flood Observatory service called Global MODIS Flood Mapping.2 The Dartmouth Flood Observatory provides This service allows users to download rasters historical and near-real-time monitoring of of surface water and flood water in several for- large flood events worldwide.1 The service mats (figure B.1).   205 Figure B.1  Map a. Surface water extent Showing Surface Water Extent in a Flood Event in Bolivia in March 2014 and Discharge Estimate from Passive Microwave Source: Brakenridge et al. 2014. http://floodobservatory .colorado.edu/Rapid Response/2014Bolivia4117/ 2014Bolivia.html. License: Creative Commons Attribution CC BY 3.0. Flooded during this event (MODIS) Flooded during this event (Landsat 8) Previously flooded (MODIS) Permanent surface water 15-day centered Satellite daily Low-flow 1.33-year 5-year 10-year 30-year average discharge threshold flood flood flood flood b. Discharge estimate 2,500 Discharge, cubic meters/sec 2,000 1,500 1,000 500 0 1/1/13 3/1/13 5/1/13 7/1/13 9/1/13 11/1/13 1/1/14 3/1/14 5/1/14 7/1/14 9/1/14 11/1/14 1/1/15 3/1/15 5/1/15 Global Flood and Landslide   Measuring Mission (TRMM) Multisatellite Monitoring Precipitation Analysis (TMPA) sensor to moni- The Global Flood and Landslide Monitoring tor rainfall accumulation across the globe, web service uses the Tropical Rainfall excluding high-latitude regions (figure B.2).3 It 206  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s Figure B.2  Example Outputs from the Global Flood and Landslide Monitoring System GODDARD SPACE FLIGHT CENTER + NASA Homepage + ABOUT TRMM + NEWS + PUBLICATIONS + SEARCH + CONTACTS + DATA + IMAGE POLICY Current Heavy Rain, Flood and Landslide Estimates (Rain information from Real-Time TRMM Multi-Satellite Precipitation Analysis [TMPA/3B42]) 9 APR 2014 0300 UTC (Observation time of last date processed) See HEAVY RAIN AREA maps See POTENTIAL LANDSLIDE maps Click on the maps below for regional displays with more information Source: Goddard Space Flight Center, National Aeronautics and Space Administration (NASA). http://trmm.gsfc.nasa.gov/publications_dir/ potential_flood_hydro.html. uses the satellite observations for estimating (NOAA), the U.S. Department of Agriculture flood risk (Hong et al. 2007; Wang et al. 2011) (USDA), the National Drought Mitigation Cen- and potential landslide sites (Hong, Adler, and ter (NDMC), and the University of Nebraska, Huffman 2006, 2007). Lincoln.4 The system uses climatic, hydrologic, and soil condition observations from more than 350 contributors around the United States and SOIL MOISTURE AND DROUGHT expert opinions from 11 climatologists to pro- MONITORING SYSTEMS duce the weekly drought condition map. This evaluation product is qualitative (and to some U.S. Drought Monitor extent subjective), not quantitative. Several The U.S. Drought Monitor provides a weekly external data sources that use Earth observation map of drought conditions across the United are also tapped to determine drought intensity: States and is produced jointly by the National • Vegetation drought response index, pro- Oceanic and Atmospheric Administration duced by the NDMC and the U.S. Geological A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s   |  207 Survey, combines data on the average per- • Soil moisture anomaly, which is obtained centage of seasonal greenness and start of from modeling season anomaly from the Advanced Very • The fraction of photosynthetically active High Resolution Radiometer (AVHRR) radiation absorbed by vegetation, obtained normalized difference vegetation index from the Medium Resolution Imaging (NDVI) with other biophysical and climate Spectrometer (MERIS) sensor. data (Brown et al. 2008; Gu et al. 2008) • Evaporative stress index, produced by the Australian Water Availability Project U.S. Department of Agriculture, is retrieved The Australian Water Availability Project via the energy balance using remotely sensed (AWAP) monitors the state and trend of the land surface temperature time-change sig- terrestrial water balance of the Australian nals and data from the geostationary opera- ­continent. The system uses the Waterdyn25M tional environmental satellites (GOES) model (­Raupach et al. 2009), which includes remotely sensed fraction of absorbed photo- • Vegetation health index, produced glob- syntetically active radiation to estimate vege- ally by NOAA, is calculated by combining a tation cover and surface temperature, aimed scaled NDVI (vegetation condition index) at improving the estimation of evapotranspi- with a scaled brightness temperature index ration fluxes. The AWAP system provides (temperature condition index), both derived weekly and monthly estimates of all the water from AVHRR balance components, including soil moisture • NDVI greenness maps, produced for the in two soil layers, transpiration, runoff, and Wildland Fire Assessment System, are deep drainage. These estimates are operation- derived from AVHRR ally available from 2007 onward; historical model runs have been produced for 1900– • Precipitation analysis, by the National 2011 to generate continental estimates of the Weather Service, is produced by merging water balance components. The Australian rainfall radar and gauge data Bureau of Agricultural and Resource Eco- • Groundwater and soil-moisture data from nomics and Sciences uses the AWAP system the Gravity Recovery and Climate Experi- to report weekly on soil moisture conditions ment (GRACE), produced by NASA, are across the country (figure B.3). assimilated into a land surface model. The dominance of AVHRR observations may IRRIGATION WATER USE AND CROP be evident and can be explained by the long GROWTH MONITORING SYSTEMS time series required to distinguish drought con- ditions of different severity. FieldLook Fieldlook, a system run by the eLEAF Com- European Drought Observatory pany in the Netherlands, provides satellite- The European Drought Observatory uses derived information to farmers.6 Weekly meteorological data and vegetation indexes estimates of biomass production, carbon diox- obtained from remote sensing to provide con- ide intake, leaf area index, and vegetation index tinuous drought assessments over Europe.5 are provided to subscribers, mostly in the Three indexes are combined: Netherlands, but also in some Eastern Euro- • The standardized precipitation index, pean countries. Key to the system is the ET which measures the rainfall anomaly from Tool, which is an adaptation of model evapo- observations transpiration across large areas based on the 208  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s BoM Interpolated Long-Term Observations (to 2008-01-07) MET: BoM Daily Observations (from 2008-01-08) AWAP | CSIRO | BoM | BRS | SEACI Figure B.3  Example AWAP 2007-present Weekly Series Legal Notice and Disclaimer Solar Temp Temp MODEL: Waterdyn25M Output of Selected Select All None Some Rainfall Soil moisture Soil moisture Total Transpiration Soil Potential Local Surface Deep Drainage Sensible Heat Latent Variables Generated Irradiance Max Min (Upper) (Lower) Evaporation Evaporation Evaporation Discharge Runoff Heat Physical Units Percent Rank Dates: Start 2007-01-01 Stop 2014-02-17 Order: Earliest First Latest First Image Size 100 % Redisplay by the AWAP System Rainfall Soil moisture (Upper) Soil moisture (Lower) Total Evaporation Transpiration Soil Evaporation Potential Evaporation Local Discharge Surface Runoff Deep Drainage Sensible Heat Latent Heat Source: Commonwealth 0 5 10 0 0.5 10 0.5 10 3 60 2 40 0.75 1.5 0 5 10 0 5 10 0 5 10 0 0.5 1.0 0 175 350 0 125 250 Scientific and Industrial Date Week Research Organisation ending 20140223 (CSIRO). http://www.eoc .csiro.au/awap/. © CSIRO. Week Used with permission. ending 20140216 Further permission required for reuse. Week ending 20140209 Week ending 20140202 Week ending 20140126 Week ending 20140119 Week Surface Energy Balance Algorithm for Land Crop Explorer and GeoGLAM:   (SEBAL) model. Crop Monitor The Crop Explorer system is run by the U.S. Irrigateway Department of Agriculture and provides a irriGATEWAY, a system run by the Common- global assessment and seasonal forecasts of wealth Scientific and Industrial Research crop growth and production.8 The website Organisation (CSIRO), aims to improve deci- allows zooming into continents and regions sion making for agricultural water resources and displays several observed or modeled management.7 Among the tools run by the ­ variables—including rainfall and soil mois- system, most relevant in this context are the ture—obtained from the World Meteorological crop coefficient (Kc) maps for irrigation dis- Organization and the U.S. Air Force Weather tricts (see figure B.4), which are generated Agency. It also provides vegetation indexes and using NDVI calculated from Landsat imag- anomalies from the vegetation (Satellite for ery. Actual evapotranspiration estimates are Earth Observation, SPOT) and MODIS sensors. generated for selected irrigation areas, and Crop Explorer also provides interactive access the ratio of actual to potential evapotranspi- to graphs and maps of reservoirs and lake levels ration is calculated to provide Kc. The data from the Jason-2 and Envisat s ­ ensors.9 for individual paddocks are extracted and The system provides an interactive map that sent automatically via text messages to farm- allows users to select a lake or reservoir and dis- ers who have subscribed to the service. plays the time-series data from either sensor These farmers, in turn, use the information showing height variation. Users also have the to refine the irrigation volumes and bench- option of downloading the data in ASCII format. mark their water use against that of other The GeoGLAM Crop Monitor is a joint ini- irrigators who subscribe to the service. tiative involving NASA and the Goddard Space Water providers can also use the system as Flight Center, the U.S. Department of Agricul- an auditing tool. ture and the Foreign Agricultural Service, A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s   |  209 Figure B.4  Example Output of a Crop- Coefficient (Kc) Map Produced by irriGATEWAY Source: CSIRO. http://www Home | Tools | Publications | Projects | Weatherstations | About | Contact .irrigateway.net/tools/ Kc Maps Home | Methodologies and FAQ kcmap/Location.aspx?loc =mia. © CSIRO. Used with Murrumbidgee Irrigation Area (prototype) permission. Further permission required for Choose a map by date: Legend reuse. 2010 09 06 Load 1.1 0.3 Map Satellite You are viewing the Kc map for 06/09/2010. Science Systems and Applications, Inc., and the food shortages. CropWatch estimates crop area University of Maryland.10 It uses MODIS NDVI and yields and also assesses drought and crop data to monitor croplands globally and pro- conditions. vides detailed maps at 250-meter resolution of vegetation index anomalies. Users can select a SNOW EXTENT point or polygon and obtain time-series data of NDVI from 2000 until the present. NOAA’s National Snow Analysis Several U.S agencies produce a variety of CropWatch satellite-derived snow products that range CropWatch, China’s global crop monitoring from regional to global in scale and from daily system, uses EO data combined with selected to monthly in frequency: field data to determine key crop production • The NOAA National Environmental Satel- indicators, including crop acreage, yield and lite Data and Information Service North- production, crop condition, cropping intensity, ern Hemisphere snow extent maps crop-planting proportion, total food availability, and status and severity of droughts. Results are • The National Snow and Ice Data Centre combined to analyze the balance between sup- (NSIDC) Northern Hemisphere EASE-Grid ply and demand for various food crops and, if Weekly Snow Cover and Sea Ice Extent needed, provide early warning against possible product 210  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s Figure B.5  Snow Snow Depth Depth for the 2014-04-08 06 UTC Continental United States on April 8, 2014 Source: National Opera- tional Hydrologic Remote Sensing Center. http://www .nohrsc.noaa.gov/nsa/. inches 1000s of ft 0 0.39 2 3.9 9.8 20 39 59 98 197 295 394 787 0 1.6 3.3 4.9 6.6 8.2 9.8 11 13 15 0 1 5 10 25 50 100 150 250 500 750 1000 2000 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 cm Elevation (km) • The NSIDC Near-Real-Time SSM/I EASE- instruments include the optical imager and Grid Daily Global Ice Concentration and thermal sounder. Snow Extent product MODIS Snow Data Product • The U.S. National Operational Hydro- NASA produces the MODIS suite of global logic Remote Sensing Center’s (NOHRSC) snow products, composed of products covering National Snow Analysis.11 a range of spatial resolutions (from 500 meters These products are made specifically for to 0.25°) and temporal resolutions (daily, eight- application to hydrologic analyses (figure B.5). day, and monthly). Snow cover is described as The analyses provide daily, operational moni- fractional cover; snow albedo is also available. toring of snow across the United States, includ- Together these are designated as the MOD10 ing snow depth, snow water equivalent, and data product. MOD10 data are derived from snow melt.12 the visible and infrared channels on MODIS Products provide information on the physi- and use the normalized difference snow index cal properties of snow by combining satellite-, (NDSI), which compares the differences in airborne-, and field-based observations with reflectance between green and mid-infrared snow models. Satellite imagery comes from wavelengths (Hall et al. 2002). both geostationary (GOES) and polar-orbiting operational environmental satellites (POES) GlobSnow operated by both NOAA and the European The European Space Agency (ESA) funded Organisation for the Exploitation of Meteoro- GlobSnow-1, which produced hemispheric, logical Satellites (EUMETSAT). The POES long-term, daily, weekly, and monthly records instruments include the optical AVHRR, ther- of snow cover and snow water equivalent. This mal High Resolution Infrared Radiation task is now being continued through the Sounder (HIRS/3), and microwave sensors ­ GlobSnow-2 Project. The snow cover data are from the Advanced Microwave Sounding Unit based on optical data from Envisat’s Advanced and Mitsubishi Heavy Industries. The GOES Along Track Scanning Thermal Radiometer A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s   |  211 (AASTR) and European Remote Sensing that the information they provide has found ­Satellite (ERS-2) sensors, while the snow water wide uptake in water management.” Van Dijk equivalent record is based on the time series of and Renzullo (2011) define “spatial water measurements by two different space-borne resources monitoring systems” as software passive microwave sensors (the Scanning Mul- that integrates observations into models to tichannel Microwave Radiometer [SMMR] produce spatial estimates of current (and past) and the Special Sensor Microwave Imager water resources distribution. A few examples [SSM/I]). The snow water equivalent product of such systems are given below. combines satellite-based passive microwave measurements with weather station data North American and GLOBAL Land Data through a data assimilation scheme. Assimilation Systems The North American Land Data Assimilation China Meteorological Administration Systems (NLDAS) combines data from multi- The China Meteorological Administration oper- ple sources within models to produce gridded ationally monitors the snow cover of China and maps of land surface states and fluxes.15 Among the Northern Hemisphere.13 Monitoring is done the states and fluxes reported, and of interest using a combination of optical and microwave for regional hydrology, are soil moisture, data from the geostationary and polar-orbiting streamflow, runoff, and evapotranspiration. satellites Fengyun-2D (FY-2D), Fengyun-2E The modeling framework behind the NLDAS (FY-2E), and Fengyun-3B (FY-3B). These prod- system includes the Mosaic, Noah, Sacra- ucts have been favorably compared against the mento, and VIC models (Xia, Mitchell, Ek, MODIS snow products (Yang et al. 2014). Cosgrove et al. 2012; Xia, Mitchell, Ek, Shef- field et al. 2012). The NLDAS uses remotely Central Asia Snow Melt Forecasting sensed information of downward shortwave Snowmelt is a critical water resource for many radiation from the GOES-8 satellite and the of the arid countries in Central Asia. Having Climate Prediction Center’s MORPHing suitable capacity to predict snowmelt is there- technique (CMORPH) for estimating precipi- fore important for water and food security in tation. The NLDAS Drought Monitor provides this region. The Regional Centre for Hydrol- estimates of soil moisture, snow water equiva- ogy in Central Asia, a Swiss-backed initiative, is lent, total runoff, streamflow, evapotranspira- tasked with (among other things) forecasting tion, and precipitation for the continental snowmelt across the five member countries United States. It also provides forecasts of (Kazakhstan, the Kyrgyz Republic, Tajikistan, these variables of up to six months. Using an Turkmenistan, and Uzbekistan).14 Forecasts approach similar to NLDAS, a global version are produced by combining satellite imagery (GLDAS) has been developed.16 (AVHRR), expert opinion, ground observa- Satellite observations are used in GLDAS tions, and modeling. directly and indirectly. In particular, the mete- orological forcing data are derived from “a  combination of NOAA/GDAS atmospheric WATER RESOURCES MONITORING analysis fields, spatially and temporally disag- SYSTEMS gregated NOAA Climate Prediction Center Merged Analysis of Precipitation (CMAP) According to Van Dijk and Renzullo (2011), fields, and observation-based radiation fields “Few satellite data are used in only a handful of derived using the method of the Air Force operational surface water resources monitor- Weather Agency’s AGRicultural METeorologi- ing systems. There appears to be little evidence cal modeling system.” 212  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s Australian Water Resources Assessment model. It shares a common heritage with the System Australian Water Resources Assessment Land- The Australian Water Resources Assessment scape (AWRA-L) model but is applicable to a System (AWRA) uses a series of coupled land- wider range of conditions. Processes such as scape, groundwater, and river models to provide evapotranspiration, soil and groundwater consistent water information for Australia.17 movement, and streamflow are represented for The Australian Bureau of Meteorology uses the two vegetation classes in each 1° grid cell (forest AWRA system, along with other data sources, to and nonforest cover). The climate data that are produce the Australian Water Resources Assess- fed into the model are a combination of several ment and annual National Water Account. sources, which are blended to obtain the best The AWRA provides consistent water infor- estimates of past and current conditions. The mation on climatic conditions and landscape model is forced by “ERA-Interim” weather characteristics, patterns and variability in forecast model reanalysis data from the Euro- water availability over time, surface water and pean Centre for Medium-Range Weather Fore- groundwater status, floods, streamflow salinity casts. For low latitudes, these are combined and inflows to wetlands, and urban and agri- with near-real-time TRMM multisensor pre- cultural water use. Previous AWRA reports cipitation analysis data (TMPA 3B42 RT) (Huff- such as the 2012 assessment (BoM 2013) used a man et al. 2007) to improve estimates of grid-based landscape model, AWRA-L (Van convective rainfall (Peña-Arancibia et al. 2013). Dijk 2010; Van Dijk and Warren 2010) to pro- duce information on the landscape water bal- ance (figure B.6). More recently, this has been WATER RESOURCES ASSESSMENT coupled with a continental groundwater model AND SCENARIO STUDIES (AWRA-G) and a river water accounting model (AWRA-R), which will be used in future Murray-Darling Basin Sustainable   reports. Yields Project In 2007 and 2008, CSIRO led a consortium to Asia-Pacific Water Monitor assess the likely impacts of climate change on The Asia-Pacific Water Monitor, an experi- the surface water and groundwater resources mental water balance monitoring system of the Murray-Darling basin. This region cov- developed by CSIRO and Australian National ers 1 million square kilometers and supplies at University, provides near-real-time water bal- least 40 percent of Australia’s agricultural pro- ance estimates for the Asia-Pacific region and duction. The Murray-Darling Basin Sustain- interprets these in a historical context.18 Maps able Yields Project delivered the most show precipitation, streamflow, catchment comprehensive and complex whole-of-basin water storage, and actual and potential evapo- water assessment ever undertaken in Australia transpiration. Information is presented as and was probably the world’s first regarding actual values, deciles, anomalies, and percent- the scale of assessment. age of average and is available for daily totals The project, funded by the Australian and 30-day averages. National Water Commission, reported on water The Asia-Pacific Water Monitor is based on a availability and water use under historical and water balance model that is updated daily using likely future climates, together with a consider- weather data derived from a mix of field and ation of possible changes in farm dams and for- satellite measurements and weather forecasts. estry. It brought together nearly 200 people The model used in the monitor is the World- from more than 15 organizations and assembled Wide Water Resources Assessment (W3RA) a complex, computer-based model of the basin’s A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s   |  213 Figure B.6 Example Landscape water flows Summary Output Region average Difference from 1911–2012 Decile ranking with respect to the from AWRA for 2012 long-term annual mean 1911–2012 record in the Murray-Darling Rainfall 651 mm +40% 10th—very much above average Basin Evapo- transpiration 559 mm +29% 10th—very much above average Source: Reproduced from Landscape BoM 2013. © Bureau of water yield 65 mm +110% 10th—very much above average Meteorology. Used with Streamflow (at selected gauges) permission. Further permission required for Annual total Predominantly above average flow throughout the region and numerous stream reuse. flow: gauges in the upper Darling River with very much above average flow Salinity: Annual median electrical conductivity predominantly below 1,000 μS/cm throughout the region Flooding: Major floods in the upper Darling River and in the Lachlan and Murrumbidgee rivers Surface water storage (comprising about 88% of the region’s total capacity of all major storages) Total 30 June 2012 30 June 2011 Change accessible accessible % of total accessible % of total accessible % of total capacity volume capacity volume capacity volume capacity 30,192 GL 25,230 GL 84% 22,006 GL 73% +3,224 GL +9% Wetlands inflow patterns (for selected wetlands) Currawinya Lakes and Very much above average flows in the normally wettest month of Paroo River wetlands: February and above average flows in December and March Gwydir wetlands: Very much above average flows in November, December, and February Macquarie Marshes: Very much above average flows in March and April Barmah–Millewa Forest: Very much above average flows in March, above average flows in July and August Groundwater (in selected aquifers) Levels: Predominantly rising groundwater levels in the northern aquifers, variable to stable groundwater level trends in the southern aquifers Salinity: Nonsaline groundwater (< 3,000 mg/L) in most aquifers in the uphill areas, mostly saline (≥ 3,000 mg/L) in the downhill basin aquifers Urban water use (Canberra) Total sourced in 2011–12 Total sourced in Change Restrictions 2010–11 44 GL 41 GL +3 GL (+7%) Permanent Water Conservation Measures Annual mean soil moisture (model estimates) Spatial patterns: Predominantly very much above average throughout the region, with some areas of above average in the south and east of the region Temporal patterns in Consistently very much above average during the year regional average: water resources. This was achieved by linking Murray-Darling basin: it has changed the flood- 40 existing and new models of surface and ing regimes that support nationally and interna- groundwater supplies and extractions across tionally important floodplain wetland systems, the basin’s 18 individual regions. reduced the total water flow at the Murray The project found that water resources mouth by 61 percent, and caused the river to development has profoundly affected the cease flowing through the Murray mouth 214  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s 40 percent of the time, compared with 1 percent It is a tool to assist the consideration of ecosys- of the time before water resources had started tem groundwater requirements in natural being developed. It also found that the impacts resources management, including water plan- of climate change by 2030 are uncertain. How- ning and environmental impact assessment. ever, surface water availability across the entire The atlas was funded by the Australian govern- basin is more likely to decline than increase. ment and developed by a consortium of private The project intensively used rainfall-runoff and public organizations; it is hosted by the models, together with past climate observa- Bureau of Meteorology.19 tions and future climate scenarios. It also used Development of the atlas used a wide range EO information to draw up a set of river water of data, field surveys, observations, and aca- balance accounts (Kirby et al. 2008). These demic and management expertise and required accounts were used to evaluate the uncertainty an extensive geographic information system in preexisting river hydrology models that framework to integrate these different sources were used in the scenario studies (Van Dijk of information. MODIS and Landsat observa- et al. 2008). The EO information used included tions played a critical role in the project, par- the following: ticularly for the many regions where detailed field observations were not available. • Irrigated cropping areas, derived by com- Specifically, the accuracy of 250-meter res- bining NDVI patterns with agricultural olution MODIS-derived estimates of evapo- statistics (BRS 2006) transpiration (Guerschman et al. 2009) was • Dynamic data on the extent of permanent enhanced using GRACE observations, and the and semi-permanent surface water areas seasonal patterns of evapotranspiration were combined with rainfall information to identify • Estimates of evapotranspiration from open areas likely to be reliant on external water water, irrigated land, wetlands, and dry- inputs other than rainfall. The resulting infor- land (Guerschman et al. 2008, 2009). mation was combined with inundation map- Satellite observations were also involved in ping (Guerschman et al. 2011) to identify determining forest cover and changes in forest surface water–fed ecosystems (Barron et al. cover (Furby 2002), which served as input for 2014). Furthermore, spatial classification of the scenario modeling. seasonal Landsat NDVI and wetness patterns were used to enhance mapping spatially in a National Atlas of Groundwater-Dependent subset of regions (figure B.7). Ecosystems One of the issues of concern in groundwater Water Quality, Potential Harmful Algal management is how to avoid damage to Blooms, and Aquaculture groundwater-dependent ecosystems. The Several programs combine field data, models, National Atlas of Groundwater-Dependent and EO data in a data-data fusion—that is, nei- Ecosystems presents the current knowledge of ther a model-data fusion nor a model-data groundwater-dependent ecosystems across assimilation—to gain more insight into the Australia and was developed to improve under- phenomena observed or to predict potentially standing of these ecosystems and facilitate harmful algal blooms. The European Union how they are considered in water resources and ESA Copernicus Programme’s website management. The atlas displays ecological and provides a substantial overview of what is pos- hydrogeological information on ecosystems sible in the near future for EO-based informa- that are known to depend on groundwater and tion services and provides scoping information ecosystems that potentially use groundwater. for current and near-future applications.20 A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s   |  215 Figure B.7  Example HOME ABOUT CONTACTS Search View from the Atlas Australian Government Bureau of Meteorology NSW VIC QLD WA SA TAS ACT NT AUSTRALIA GLOBAL ANTARCTICA of Groundwater- Dependent Bureau Home > Water Information > GDE Atlas Home > GDE Atlas Map Ecosystems, Hosted Water Information Regulations Standards News and events About by the Bureau of Meteorology Source: © Bureau of Meteorology. Used with permission. Further permission required for About the Atlas | text version | maximise map reuse. Quick Search Note: The likely presence of Layers groundwater-dependent Groundwater dependent ecosystems... ecosystems is shown in dark Reliant on surface... colors around Mont Reliant on subsurface... All ecosystem features Gambier, a karst region on No ecosystems analysed the border between Victoria Subterranean (Caves and aquifers) and South Australia. All ecosystem features T No ecosystems analysed A B Inflow dependent ecosystems (IDEs) L E IDE (rivers, springs, wetlands),... O IDE (vegetation),... F Gridded ID layer C O Gridded Remote Sensing Layer N T Base map E N Places T S State and Territory borders Roads Legend Advanced Search 0 20 40km Location Current scale 1,000,537 Marine Water Quality and Forecasting Southwest Shelf–Ocean, Mediterranean The Copernicus Programme of the European Sea, and Black Sea Union and the ESA has been funding the • Parameters and variables: ocean tempera- MyOcean Programme since 2009. MyOcean ture, ocean salinity, ocean currents, sea ice, (2009–12) and now MyOcean2 (2012–14) are sea level, winds, ocean optics, ocean chem- committed to developing and running a istry, ocean biology, and ocean chlorophyll European service based on a worldwide capacity for ocean monitoring and forecast- • Product type: forecast, near-real-time, ing, using observations data, modeling, and multiyear, time-invariant products (either assimilation systems.21 MyOcean offers reli- from observations or modeling). able and easy access to valuable core infor- mation about the ocean. The service is The Australian eReefs Marine Water intended to serve any user requesting generic Quality Dashboard information on the ocean, but especially Using the latest technologies to collate field- downstream service providers, who use the based and EO-derived information and new information as input for their value added and integrated modeling, eReefs has started services to end users. The interactive cata- producing powerful visualization, communi- logue allows users to select products accord- cation, and reporting tools.22 The Marine ing to the following: Water Quality Dashboard provides access to • Seven geographic areas: Global–Ocean, Arc- archival and real-time data on ocean color and tic Ocean, Baltic Sea, Atlantic-European sea surface temperature for the entire Great Northwest Shelf–Ocean, Atlantic-European Barrier Reef.23 It provides reef information 216  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s akin to that provided by the Bureau of Meteo- A major objective of the Water Framework rology for weather. This information could Directive is to establish an integrated, spatially benefit government agencies, reef managers, explicit monitoring and management system policy makers, researchers, industry, and local for all waters. Information such as that pre- communities. sented in this atlas could support monitoring The eReefs Project delivers the following: and management of Lake IJssel. In addition, this set of measurements and model results for • Expanded and improved monitoring data 2003, an unusually sunny, hot, and dry year, is • Measurement technologies and data deliv- ideal for investigating the relation between cli- ery tools (for example, mobile and Internet mate change (meteorological conditions, input tools) by the IJssel River) and water quality in Lake • A suite of new and integrated models IJssel. across paddock, catchment, estuary, reef lagoon, and ocean Harmful Algal Blooms The experimental Lake Erie Harmful Algal • A framework to explore the impact of mul- Bloom Bulletin was developed to provide a tiple factors such as temperature, nutrients, weekly forecast for microcystis blooms in turbidity, and acidity, and to communi- western Lake Erie.24 Many different species of cate this information to those who will be single-celled organisms live in the Great Lakes, affected by it including algae. When certain conditions are • An interactive visual picture of the reef and present, such as high levels of nutrients or its component parts, accessible to all light, these organisms can reproduce rapidly to produce a dense population of algae, called a • Citizen science initiatives to engage the bloom. Some of these blooms are harmless, but broader community on the health of the reef when the blooming organisms contain toxins, • Targeted communication products to other noxious chemicals, or pathogens, they allow the public to interact with the reef— become harmful. Harmful algal blooms can contributing monitoring information and cause the death of nearby fish, foul up nearby learning about the reef. coastlines, and produce harmful conditions for aquatic life as well as humans. Inland Water Quality If a harmful bloom is detected, scientists In 2005, the local management authority of the will issue a forecast bulletin. The bulletin largest freshwater lake in the Netherlands, Lake depicts the current location and future move- IJssel, asked the Institute for Environmental ment of harmful algal blooms and categorizes Studies (Vrije Universiteit) to demonstrate the the intensity on a weekly basis. This research status of operational spatial monitoring and project aims to determine the factors control- modeling. The results are summarized in an ling microcystin production and develop atlas of Lake IJssel (IJsselmeer). The atlas con- methods for determining cyanobacteria tains water quality products from SeaWiFS for blooms from satellite imagery. Imagery is cur- the year 2003. For the summer, it provides fort- rently available, but it is not yet able to discrim- nightly median maps for chlorophyll-a; for the inate toxic microcystis blooms from other algal winter period, it contains monthly median blooms within the images. The combined field maps. These data are compared to field data data and satellite image data produced from and model simulation results. The capacities of the initial efforts are critical first steps in the MERIS on monitoring chlorophyll-a are also characterization of bloom dynamics and the illustrated in the special maps section. development of future bloom forecasting tools. A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s   |  217 The Applied Simulations and Integrated scanner. These data also serve as input for Modelling for the Understanding of Toxic and growth models. Currently, Smartshell provides Harmful Algal Blooms (ASIMUTH) aims to three basic services: (a) site selection, (b) real- develop forecasting capabilities to warn of time monitoring, and (c) production monitor- impending hazardous blooms in five European ing and projection. countries.25 Through the ASIMUTH project, scientists and industry from five countries along Europe’s Atlantic Margin have formed a NOTES network to produce the first realistic advisory 1. For information on the Dartmouth Flood and forecasting capability as a downstream Observatory, see http://floodobservatory.colorado service to the European aquaculture industry. .edu/. The early warning of severe blooms will allow 2. For information on Global MODIS Flood fish and shellfish farmers to adapt their culture Mapping, see http://oas.gsfc.nasa.gov/floodmap/. and harvesting practices in time, so as to reduce 3. For information on Global Flood and Landslide Monitoring, see http://trmm.gsfc.nasa.gov/ potential losses. publications_dir/potential_flood_hydro.html. ASIMUTH is the first step toward develop- 4. For information on the Drought Monitor, see ing short-term hazardous algal bloom alert http://droughtmonitor.unl.edu/. systems for Atlantic Europe. This will be 5. For information on the European Drought achieved using information on the most cur- Observatory, see http://edo.jrc.ec.europa.eu/. rent marine conditions (weather, water char- 6. For information on Fieldlook, see http://www .mijnakker.nl/. acteristics, toxicity, harmful algal presence), 7. For information on irriGATEWAY, see http:// combined with local numerical predictions. www.irrigateway.net/. ASIMUTH will use geospatial products from 8. For information on Crop Explorer, see http:// the MyOcean project to initiate the models www.pecad.fas.usda.gov/cropexplorer/. developed during the project. Experts from 9. For information on the interactive graphs and each country will evaluate data from the moni- maps, see http://www.pecad.fas.usda.gov/ cropexplorer/global_reservoir/. toring programs, satellite images,  and model 10. For information on GeoGLAM Crop Monitor, see output to produce bulletins to inform the pub- http://www.geoglam-crop-monitor.org/. lic and the aquaculture sector. The bulletins 11. For information on the NOHRSC National Snow produced will present the current state of haz- Analysis, see http://www.nohrsc.noaa.gov/nsa/. ardous algal blooms in each area and the likeli- 12. For information on the National Operational Hydrologic Remote Sensing Center, see http:// hood of a toxic or harmful event of target www.nohrsc.noaa.gov/nsa/. species in the following week. 13. For information on the China Meteorological Administration, see http://cmdp.ncc.cma.gov.cn/ Aquaculture Monitoring/en_snow_ice.php. Smartshell is a real-time, online tool that pro- 14. For information on the Regional Centre for Hydrology in Central Asia, see http://www.rch- vides information on the water quality of aralsea.ch/index.html. coastal areas, aimed at the aquaculture sec- 15. For information on the NLDAS, see http://ldas tor.26 It uses maps of chlorophyll and sediment .gsfc.nasa.gov/nldas. concentrations as well as transparency derived 16. For information on the GLDAS, see http://ldas from satellite data and ancillary data such as .gsfc.nasa.gov/gldas/. wind force and direction data, water depth, 17. For information on the AWRA, see http://www .bom.gov.au/water/awra. and temperature. If required, frequent and 18. For information on the Asia-Pacific Water flexible field measurements can be done with Monitor, see http://eos.csiro.au/apwm. the Water Insight Spectrometer with three 19. For information on the atlas, see http://www.bom radiometers (WISP-3) handheld water quality .gov.au/water/groundwater/gde/. 218  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s 20. For an overview of the Copernicus Programme, Guerschman, J., A. Van Dijk, T. McVicar, T. Van see http://gmesdata.esa.int/web/gsc/ Niel, L. Li, Y. Liu, and J. Peña-Arancibia. core_services/downstream_services. 2008. Water Balance Estimates from Satellite 21. For information on MyOcean, see http://www Observations over the Murray-Darling Basin. .myocean.eu/web/26-catalogue-of-services.php. Report to the Australian Government from the 22. For information on eReefs, see http://www.bom CSIRO Murray-Darling Basin Sustainable Yields .gov.au/environment/eReefs_Infosheet.pdf. Project. Canberra: Commonwealth Scientific and Industrial Research Organisation. 23. For the Marine Water Quality Dashboard, see Guerschman, J. P., A. I. J. M. Van Dijk, G. Mattersdorf, http://www.bom.gov.au/marinewaterquality/. J. Beringer, L. B. Hutley, R. Leuning, R. C. Pipunic, 24. For information on the Lake Erie Harmful Algal and B. S. Sherman. 2009. “Scaling of Potential Bloom Bulletin, see http://www.glerl.noaa.gov/res/ Evapotranspiration with MODIS Data Centers/HABS/lake_erie_hab/lake_erie_hab.html. Reproduces Flux Observations and Catchment 25. For information on ASIMUTH, see http://www Water Balance Observations across Australia.” .asimuth.eu/en-ie/Pages/default.aspx. Journal of Hydrology 369 (1-2): 107–19. 26. For information on Smartshell, see www Guerschman, J. P., G. Warren, G. Byrne, L. Lymburner, .smartshellservices.com. N. Mueller, and A. I. J. M. Van Dijk. 2011. MODIS- Based Standing Water Detection for Flood and Large Reservoir Mapping: Algorithm Development and Applications for the Australian Continent. REFERENCES Canberra: Commonwealth Scientific and Industrial Research Organisation. Barron, O. V, I. Emelyanova, T. G. Van Niel, D. Pollock, Hall, D. K., G. A. Riggs, V. V. Salomonson, and G. Hodgson. 2014. “Mapping Groundwater- N. E. DiGirolamo, and K. J. Bayr. 2002. “MODIS Dependent Ecosystems Using Remote Sensing Snow-Cover Products.” Remote Sensing of Measures of Vegetation and Moisture Dynamics.” Environment 83 (1-2): 181–94. Hydrological Processes 28 (2): 372–85. doi:10.1002/ Hong, Y., R. F. Adler, F. Hossain, S. Curtis, and hyp.9609. G. J. Huffman. 2007. “A First Approach to Global BoM (Bureau of Meteorology). 2013. “Australian Water Runoff Simulation Using Satellite Rainfall Resources Assessment 2012.” Australian Bureau of Estimation.” Water Resources Research 43 (8): Meteorology, Melbourne. W08502. Brakenridge, G. R., D. Slayback, A. J. Kettner, Hong, Y., R. Adler, and G. J. Huffman. 2006. F. Policelli, T. De Groeve, and S. Cohen. 2014. “Evaluation of the Potential of NASA Multi- “Rapid Response Mapping and Measurement of Satellite Precipitation Analysis in Global Landslide the 2014 Flooding in Bolivia.” Dartmouth Flood Hazard Assessment.” Geophysical Research Letters Observatory, University of Colorado, Boulder, CO, 33 (22): L22402. doi:10.1029/2006GL028010. USA (http://floodobservatory.colorado.edu/Rapid ———. 2007. “Use of Satellite Remote Sensing Data in Response/2014Bolivia4117/2014Bolivia.html). the Mapping of Global Landslide Susceptibility.” Brown, J. F., B. D. Wardlow, T. Tadesse, M. J. Hayes, Natural Hazards 43 (2): 245–56. doi:10.1007/ and B. C. Reed. 2008. “The Vegetation Drought s11069-006-9104-z. Response Index (VegDRI): A New Integrated Huffman, G. J., R. F. Adler, D. T. Bolvin, G. Approach for Monitoring Drought Stress in Gu, E. J. Nelkin, K. P. Bowman, Y. Hong, Vegetation.” GIScience & Remote Sensing 45 (1): E. F. Stocker, and D. B. Wolff. 2007. “The TRMM 16–46. doi:10.2747/1548-1603.45.1.16. Multisatellite Precipitation Analysis (TMPA): BRS (Bureau of Rural Sciences). 2006. 1992/93, Quasi-Global, Multiyear, Combined-Sensor 1993/94, 1996/97, 1998/99, 2000/01 and 2001/02 Precipitation Estimates at Fine Scales.” Journal of Land Use of Australia, Version 3. Bureau of Rural Hydrometeorology 8 (1): 38–55. Sciences, Canberra. Kirby, J. M., A. I. J. M. Van-Dijk, J. Mainuddin, Furby, S. 2002. “Land Cover Change: Specification for J. L. Peña-Arancibia, Y. Liu, S. Marvanek, and Remote Sensing Analysis.” Australian Greenhouse L. T. Li. 2008. River Water Balance Accounts Office, Canberra. across the Murray-Darling Basin, 1990–2006. Gu, Y., E. Hunt, B. Wardlow, J. B. Basara, J. F. Brown, A report to the Australian Government from the and J. P. Verdin. 2008. “Evaluation of MODIS CSIRO Murray-Darling Basin Sustainable Yields NDVI and NDWI for Vegetation Drought Project. Canberra: Commonwealth Scientific and Monitoring Using Oklahoma Mesonet Soil Industrial Research Organisation. Moisture Data.” Geophysical Research Letters Peña-Arancibia, J. L., A. I. J. M. Van Dijk, L. J. Renzullo, 35 (22): L22401. doi:10.1029/2008GL035772. and M. Mulligan. 2013. “Evaluation of A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s   |  219 Precipitation Estimation Accuracy in Reanalyses, System. Technical Report 4. Landscape Model Satellite Products, and an Ensemble Method for (version 0.5) Evaluation against Observations. Regions in Australia and South and East Asia.” Canberra: Commonwealth Scientific and Journal of Hydrometeorology 14 (4): 1323–33. Industrial Research Organisation. Raupach, M. R., P. R. Briggs, V. Haverd, E. A. King, Wang, J., Y. Hong, L. Li, J. J. Gourley, S. I. Khan, M. Paget, and C. M. Trudinger. 2009. Australian K. K. Yilmaz, R. F. Adler, F. S. Policelli, S. Habib, Water Availability Project (AWAP): CSIRO D. Irwin, A. S. Limaye, T. Korme, and L. Okello. Marine and Atmospheric Research Component: 2011. “The Coupled Routing and Excess Storage Final Report for Phase 3. Canberra: Bureau of (CREST) Distributed Hydrological Model.” Meteorology and Commonwealth Scientific and Hydrological Sciences Journal 56 (1): 84–98. Industrial Research Organisation. Xia, Y., K. Mitchell, M. Ek, B. Cosgrove, J. Sheffield, Van Dijk, A. I. J. M. 2010. The Australian Water L. Luo, C. Alonge, H. Wei, J. Meng, B. Livneh, Resources Assessment System. Technical Q. Duan, and D. Lohmann. 2012. “Continental-Scale Report 3. Landscape Model (version 5.0) Water and Energy Flux Analysis and Validation for Technical Description. Water for a Healthy North American Land Data Assimilation System Country National Research Flagship. Canberra: Project Phase 2 (NLDAS-2): 2. Validation of Model- Commonwealth Scientific and Industrial Simulated Streamflow.” Journal of Geophysical Research Organisation. Research: Atmospheres 117 (D3): 3110. Van Dijk, A. I. J. M., J. M. Kirby, Z. Paydar, G. Xia, Y., K. Mitchell, M. Ek, J. Sheffield, B. Cosgrove, Podger, M. Mainuddin, S. Marvanek, and J. E. Wood, L. Luo, C. Alonge, H. Wei, J. Meng, Pena-Arancibia. 2008. Uncertainty in River B. Livneh, D. Lettenmaier, V. Koren, Q. Duan, Modelling across the Murray-Darling Basin. K. Mo, Y. Fan, and D. Mocko. 2012. “Continental- Report to the Australian Government from the Scale Water and Energy Flux Analysis and CSIRO Murray-Darling Basin Sustainable Yields Validation for the North American Land Data Project. Canberra: Commonwealth Scientific and Assimilation System Project Phase 2 (NLDAS-2): Industrial Research Organisation. 1. Intercomparison and Application of Model Van Dijk, A. I. J. M., and L. J. Renzullo. 2011. “Water Products.” Journal of Geophysical Research: Resource Monitoring Systems and the Role of Atmospheres 117 (D3): 3109.  Satellite Observations.” Hydrology and Earth Yang, J., L. Jiang, J. Shi, S. Wu, R. Sun, and H. Yang. System Sciences 15 (January): 39–55. 2014. “Monitoring Snow Cover Using Chinese Van Dijk, A. I. J. M., and G. Warren. 2010. The Meteorological Satellite Data over China.” Remote Australian Water Resources Assessment Sensing of Environment 143 (March 5): 192–203. 220  |  A p p e n d i x B : E x a m p l e s o f W a t e r I n f o r m a t i o n P r o d u c t G e n e r a t i o n S y s t e m s Index Note: b, f, n, and t refer to box, figure, note, and African Drought Monitor table, respectively. Princeton University, 107 Surface Hydrology Group, A Princeton University, 18 AAAs (analytical and advisory activities), water African Rainfall Estimation Algorithm portfolio, 36–37, 43–46 Version 2 (RFE2), 190 aboveground biomass, 106 AGRHYMET, 57–59 accuracy. See also errors AGRHYMET-CILSS (Permanent Inter-State of EO data products, 6, 150, 153 Committee for Drought Control of ET estimation, 95, 156–157 in the Sahel), 58 of precipitation estimates with RS data, agriculture. See also crops 155–156, 173–174 RS applications for, 46, 74, 76–77 of rain-gauge measurements, 87–89 water efficiency example, 165t7.14 of snow cover with RS, 179–180 WRM, 210–211 of soil moisture estimation, 157–158 Ahmed, T., 191 with spectral resolution, 131 Aires, F., 116 of surface water and streamflows with RS, albedo, 102, 103 180–181 algal blooms, 130, 216, 218–219 active microwave sensors, 117, 121 Ali, A., 68 active remote sensing, 80 ALOS (Advanced Land Observation Satellite), 41 active scatterometry EO systems, 97, 98–99t6.6 altimetry instruments, 116, 180–181 active sensors, surface water, 113, 114–115t6.10, Amani, A., 68 116–117 American Water Resources Association, 17–18 active soil moisture products, 100 AMSR-E, NSIDC, 100, 121 actual evapotranspiration, 3, 89–95 Anagnostou, E. N., 174, 186, 188 adaptive management, 26–28 analytical and advisory activities (AAAs), Advanced Land Observation water portfolio, 36–37, 43–46 Satellite (ALOS), 41 antecedent moisture, 132n6 Advanced Very High Resolution Radiometer applications, EO, 5–6 (AVHRR), 91, 209 Applied Simulations and Integrated Modeling Africa for the Understanding of Toxic and Harmful operational hydrology in, 57–59 Algal Blooms (ASIMUTH) Project, European RS applications in, 44 Union, 130, 219   221 aquaculture, 216, 219 bias correction, RS, 61n12 AQUASTAT, Food and Agriculture Bink, B., 190 Organization, 54–55 biomass, aboveground, 106 aquatic weeds, monitoring and controlling, 77 blended satellite- and gauge-based precipitation areal RS measurements, variables, 53 analyses, 87–89 Arkin, P. A., 83 Borga, M., 174, 188 ASCAT soil moisture products, 100 Brezonik, P. L., 129 Asian Institute of Technology, 59 Brown, B., 28 Asia-Pacific Water Monitor of the Commonwealth Brown, C., 16 Scientific and Industrial Research buffers, 29n3 Organisation (CSIRO), 87, 210 Bureau of Meteorology, Australia, 107, 108, 130, ASIMUTH (Applied Simulations and Integrated 214, 216–218 Modeling for the Understanding of Toxic and Harmful Algal Blooms) Project, European C Union, 130, 219 Cardwell, H., 18 Australia. See also Murray-Darling basin, Australia C-band Synthetic Aperture Radar (C-SAR), 116 actual ET, 90, 94 f 6.6, 96 f 6.7 Central America Climate Forum, 59 basin water assessment, 214–216 Central America Flash Flood Guidance, 20 Bureau of Meteorology, 107, 108, 130, Central Asia Snow Melt Forecasting, 213 214, 216–218 centralized data structure, 197 Daily Rainfall Estimates for March 1, 2010, Chang, S., 19 88 f 6.3 Chappell, A., 87 Distribution of Real-Time Rain Gauges and Characteristics of MODIS and Landsat TM Data Areas Where Satellite-Derived Precipitation Domain, 81 f 6.1 is Likely to Improve Accuracy of Rainfall Chen, D., 190 Estimation, 88 f 6.4 Chen, X., 190 drought monitoring system, 107 China, 23, 121, 211 eReefs project, 130, 217–218 China Meteorological Administration, 213 map of AWRA-derived total annual landscape chlorophyll, 122, 131 water yields for, 108 f 6.11 climate adaptation strategies, 27 rainfall estimates, 87, 88 f 6.3 Climate and Water Department of the satellite imagery during flood event, 116 f 6.13 World Meteorological Organization, 54 Tumbarumba, NSW, 96 f 6.7 climate change, 15–17, 77–78 Western Queensland, 90 Climate Change Knowledge Portal, Australian National Water Commission, 214 World Bank, 54 Australian Water Availability Project (AWAP), 209 climate forums, 59 Australian Water Resources Assessment (AWRA) Climate Maps, Bureau of Meteorology, system, 108 f 6.11, 214, 215 f  B.6 Australia, 107 Availability of Historical Monthly and Daily Climate Prediction Center MORPHing technique Discharge Data in the Global Runoff Data (CMORPH), 174–175, 186, 190 Center Database, 56 f 4.2 Climate Research Unit, time-series gauge AVHRR (Advanced Very High Resolution precipitation data set, 17 Radiometer), 91, 209 Climate Services Center (CSC), Southern Africa AWAP (Australian Water Availability Project), 209 Development Community, 57–58 AWRA (Australian Water Resources Assessment) climate variability and change, system, 108, 214, 215 f  B.6 RS applications in, 46 CMORPH (Climate Prediction Center MORPHing B technique), 174–175, 186, 190 backscatter terminology, 132n15 coarse spatial resolution, 171n1 Bales, J., 16 coastal discharge, monitoring water quality of, 77 Bangladesh, Institute of Water Modeling, 20 coastal pollution, monitoring water quality of, 77 Bates, P. D., 16 Combined Drought Indicator for Europe, 107 f 6.10 bathymetry, 122 Committee on Earth Observing Systems, 197 Bauer, M. E., 129 common-pool resources, management of, 27 Becker, M. W., 111 Commonwealth Scientific and Industrial Research Bennett, M. E., 186 Organisation (CSIRO), 87, 210 Beringer, J., 90 Comparing Error Estimates for Soil Moisture Biancamaria, S., 191 Products Derived from Active and Passive 222  |  I N D E X Microwave Sensors Using Triple Collocation flooding in near real time, 116–117 Technique, 101 f 6.9 groundwater, 108–111 Comparison of Actual ET Estimates Derived hydrometeorological variables, 49–57 from the NDTI Model with Actual ET hyperspectral, 129 Measurements from Tumbarumba, NSW, lake levels in near real time, 116–117 Flux Tower, 96 f 6.7 optical water quality, 121–125 complexity, data product, 150, 154 precipitation, 82–89, 173–175 Components of the Global Terrestrial Network- regionalization technique, 56 Hydrology, 55 f 4.1 reservoir levels in near real time, 116–117 comprehensive spatial planning and land RS and ground observation, 67–69, 148 management, 50 snow cover, 117–121, 178–180 Conceptual Depiction of Information-Integration soil moisture, 95–102, 177–178 Paradigm Referred to as Model-Data Fusion, surface water, 111–117, 180–182 68 f 5.1 translating into information, 59–61 conjunctive use of surface water and groundwater, types from EO, 82t6.2 20–22, 26–28 vegetation and vegetation cover, 102–108 Copernicus Programme, European Union and water quality, 130 ESA, 216, 217 Data Framework Comprising Domain- Correlation Coefficients and Snow Mapping Characteristic Elements, 81t6.1 Agreement between Observed and Satellite- Dech, S., 121 Estimated Snow Water Equivalent and Snow dekad cold cloud duration, 24 Cover, 180 f 9.5 dekad relative evapotranspiration, 24 Correlation Coefficients between Ground Dekker, A. G., 15 Observations and Satellite Estimates, 194 f  11.1 De Lannoy, G. J. M., 100 Correlation Coefficients between Observed and Dente, L., 178 Satellite-Estimated Evapotranspiration, DeRoo, A., 186, 189 176 f 9.2 De Weirdt, M., 190 Correlation Coefficients between Observed and Di, Y., 190 Satellite-Estimated Precipitation, 175 f 9.1 Dietz, A. J., 121 cosmic ray probes, as moisture measurement, 96 disALEXI, 92 Crop Explorer system, 210–211 disaster management, 50 crops. See also agriculture dissolved organic matter, 122 assessing water use efficiency in irrigated, 76 Distributed Model Intercomparison Project, 20 growth monitoring, 209–211 Distribution of Real-Time Rain Gauges and Areas monitoring and food security, 50 Where Satellite-Derived Precipitation is monitoring production and food security, 77 Likely to Improve Accuracy of Rainfall CropWatch System (China), 23, 211 Estimation in Australia, 88 f 6.4 C-SAR (C-band Synthetic Aperture Radar), 116 Dominguez, F., 190 CSC (Climate Services Center), Southern Africa Dorfman, R., 57 Development Community, 57–58 Dorigo, W. A., 100, 101 CSIRO (Asia-Pacific Water Monitor of the Draper, C. S., 100 Commonwealth Scientific and Industrial droughts Research Organisation), 87, 210 AGRHYMET-CILSS, 58 Curley, E., 17 monitoring systems, 18–20, 77, 106–107, 208–209 cyanophycocyanin, 122 Du, D., 190 cyanophycoerythrin, 122 Durcik, M., 186 Dyce, P., 87 D Dai, D., 190 E Daily Rainfall Estimates for March 1, 2010, Earth observation (EO). See also remote sensing; in Australia, 88 f 6.3 satellite remote sensing dams and reservoirs, water quality in, 75 accuracy, 6, 150, 153 Dartmouth Flood Observatory, 206 aid in determining usage of, 6b0.3, 7 f  ES.1, data. See also validation of RS data 145–148, 196–197 centralized structure, 197 challenges of, 196 droughts, 208–209 data product characteristics, 148–150 Eddy covariance flux, 93, 95 determining characteristics of minimum ET, 89–95 required data, 150–154 I N D E X   |  223 field-based measurements combined with, GlobSnow, 121, 212–213 66–69, 148 TIGER initiative, 19, 42, 47n5 outlook for, 199 European Union, 130, 216, 217 potential of, 1, 5–6, 195–196 evaporation, 3 terminology use of, 7n1, 46n2, 78n3 evaporative stress index, 209 types of data from, 82t6.2 evapotranspiration (ET) water management applications supported by, accuracy, 95, 156–157 193–194, 196 definition of, 3, 78n9 World Bank projects using, 1–3, 65–66, 201–204 EO data on, 75, 89–95 Earth Observation for Development, 39–40 estimation of, 22 Eastern Nile Technical Regional Office (ENTRO), field data requirements and characteristics 41b3.1 of EO-based products, 156–157 economy, political, 27 RS of, 24 Eddy covariance flux data, 93, 95 three general classes of models, 95t6.5 Edwards, P., 16 validation of RS data on, 175–177 eLeaf Company, 209 EVI (enhanced vegetation index), 90, 93 electromagnetic spectrum (EMS) sensors, 80 Evolution of Key Principles over Time, 35t2.1 Emmerich, W., 90 examples empirical methods Actual Evapotranspiration Estimates for Region ET, 90–95 in Western Queensland, Australia, during field data requirements and characteristics of Flow Event in February 2004, 90 f 6.5 EO-based water quality products, 123–125, General Categories of Resolution and Examples 129, 160–163 of Platforms Providing This Type of Data, EMS (electromagnetic spectrum) sensors, 80 69b5.2 energy balance methods, ET, 91–92, 94–95 Global Vegetation Cover Maps, 106t6.8 energy development, 50 Output of Selected Variables Generated by the enhanced vegetation index (EVI), 90, 93 AWAP System, 210 f  B.3 ENTRO (Eastern Nile Technical Regional Office), Outputs from the Global Flood and Landslide 41b3.1 Monitoring System, 208 f  B.2 Environmental and Ecological Science Data Satellite Imagery Captured during Flood Event Center for West China, 121 in Northern South Wales, Australia, 116 f 6.13 environmentally sustainable growth, 24–25 Studies Using the Three General Classes of environment design, 50 Actual ET Models, 95t6.5 environment operations, 51 Summary Output from AWRA for 2012 in the Envisat, 181 Murray-Darling Basin, 215 f  B.6 EO. See Earth observation View from the Atlas of Groundwater-Dependent EOWorld partnership, 40 Ecosystems, 217 f  B.7 eReefs Marine Water Quality Dashboard, Existing and Near-Future Satellite Sensor Systems 130, 217–218 of Relevance for Inland and Near-Coastal errors. See also accuracy Water Quality, 126–128t6.12 in EO data, 153 extreme events in hydrometeorological variables, 167–168 droughts, 17–20, 58, 77 in RS data, 68 floods, 17–20 in snow cover and snow water equivalent precipitation, 175 estimations, 179–180 in soil moisture estimations, 100–101, 178 f  9.3 F in streamflow estimates, 185–186 Famine Early Warning System Network, U.S. in validating RS, 170 Agency for International Development ESA. See European Space Agency (USAID), 19, 23 estuarine and marine environments, 51 FAO (Food and Agriculture Organization), 54–55 ET. See evapotranspiration FAO-56 method, 93 ET Tool, 209–210 farming. See agriculture; crops European Drought Observatory, European field-based measurements Commission, 107 EO combined with, 66–69, 148 European Drought Survey, 209 EO data characteristics, 149 European Space Agency (ESA) ET products, 156–157 Copernicus Programme, 216, 217 groundwater products, 159 EOWorld partnership, 40 precipitation products, 155–156 224  |  I N D E X snow products, 160 GeoCenter, 42 soil moisture products, 157–158 GeoGlam Crop Monitor, 210–211 surface water products, 160t7.8 geostationary orbiting operational environmental vegetation and vegetation cover products, satellites (GOES), 212 158–159 GeoWB, 42, 47n3 water quality products, 160–165 German Aerospace Center, 109, 181 FieldLook, 209–210 Gessner, U., 121 Flamig, Z. L., 86, 186, 188 glacial cover. See snow cover flash flood threat index, 19, 20 GLDAS (GLOBAL Land Data Assimilation “Flood Model Showcase” workshop, Water Systems), 213 Partnership Program, 61n11 Glenn, E. P., 90 floods. See also surface water global change, impacts of, 15–17 alert systems, 19–20 global climate models (GCMs), 15 control design, 50 Global Climate Observing System, 54 control operations, 51 Global Drought Information System, data services for, 116–117 U.S. Government, 107 forecasting applications, 19–20 Global Facility for Disaster Reduction and mapping extent of, 75–76 Recovery, 61n10 prediction, 76, 188 Global Flood Alert System, International Flood satellite imagery of event, 116 f 6.13 Network, 20 snowmelt used in forecasting, 117 Global Flood and Landslide Monitoring, 207–208 upstream water level simulations, 191 Global Flood and Landslide Monitoring, NASA warning and monitoring systems, 206–208 and Goddard Space Flight Center, 20 Food and Agriculture Organization (FAO), Global Historical Climatology Network, 54 54–55, 93 Global Initiative on Remote Sensing for Food Early Solutions for Africa Micros-Insurance Water Resources Management, 1 Project of the Netherlands Ministry of GLOBAL Land Data Assimilation Systems Development Cooperation, 23–24 (GLDAS), 213 food security and crop monitoring, 50 Global MODIS Flood Mapping, 206, 207 f  B.1 food-water-energy nexus, 22–24 Global Precipitation Measurement (GPM) footprint, satellite, 69, 78n6 mission, 84 Foppes, S., 190 Global Runoff Data Center (GRDC), 54, 56 f 4.2 Ford, Henry, 78n8 global satellite-derived precipitation estimates, forecasting applications 85t6.3 floods, 19–20 global terrestrial drought severity index, prediction errors, 19 University of Montana, 107 producing, 58–59 Global Terrestrial Network-Hydrology (GTN-H), Foreign Agricultural Service, 210–211 54, 55 f 4.1 forest management, 50 Global Terrestrial Observing System, 54 Foster, J., 117 global vegetation cover maps, 106t6.8 fPAR, 103 GlobSnow snow water equivalent data product, Fractional Bias of Streamflow Simulations Forced ESA, 121, 212–213 by Rainfall Algorithms, 187 f  10.1 Goddard Space Flight Center, 20, 210–211 freeze-thaw coverage, 101 GOES (geostationary orbiting operational Frei, A., 117 environmental satellites), 212 French Space Agency, 181 Goteti, G., 88 freshwater and terrestrial ecosystems, 51 Gourley, J. J., 86, 186, 188 funding sources, water infrastructure, 25–26 GPM (Global Precipitation Measurement) mission, 84 G GRACE (Gravity Recovery and Climate Galloway, D., 111 Experiment) mission, NASA, German gauge-based precipitation analyses, blended Aerospace Center, and Envisat, 80, 109, 111, satellite and, 87–89, 186 171, 181, 209 GCMs (global climate models), 15 gravimetry, satellite, 110t6.9, 111 General Categories of Resolution and Examples of Gravity Recovery and Climate Experiment Platforms Providing This Type of Data, 69b5.2 (GRACE) mission, NASA, German Aerospace general circulation models, 15 Center, and Envisat, 80, 109, 111, 171, 181, 209 GEO (Group on Earth Observations), 52, 123, 197 GRDC (Global Runoff Data Center), 54, 56 f 4.2 I N D E X   |  225 Great Barrier Reef, 217–218 Hydro-Estimator of Central America Flash Flood Great Barrier Reef World Heritage Park, 130 Guidance, 19–20 green growth, 24–25 Hydrological Research Centre, 20 greenness maps, NDVI, 209 hydrologic modeling, RS and, 52–53 ground-based observation networks, 53–57 hydrologic regimes, changes in, 16 ground-level measurements, variables, 53 hydrology, operational, 57–61 ground observation networks and EO, 66–69 Hydrology and Water Resources Department, “ground truth” measurements, 6, 7n3, 68, University of Arizona, 18 168, 169–171 hydrometeorological networks, 57–59 groundwater hydrometeorological variables conjunctive use of surface water and, availability of data, 53–57 20–22, 26–28 EO and, 65, 78n1 data from EO observation, 108–111 key data, 49–53 definition of, 4 validation efforts, 167–171 ecosystems dependent on, 77, 216, 217 f  B.7 hydropower field data requirements and characteristics design, 50 of EO-based products, 159 operations, 51 monitoring extraction with EO, 76 production facilities design, 78 recharge, 52 hyperspectral data, 129 Group on Earth Observations (GEO), 52, 123, 197 GTN-H (Global Terrestrial Network-Hydrology), I 54, 55 f 4.1 ice mapping system (IMS), 121 Gu, Y., 190 IGRAC (International Groundwater Resources Guerschman, J. P., 90, 93, 107 Assessment Center), 54 Guerschman Kc index, 93 IMS (ice mapping system), 121 guidelines India, 20–22, 108 f 6.12 for determining EO needs, 7 f  ES.1, 145–148 information-integration paradigm, 68 f 5.1 for determining minimum requirements of EO infrastructure data products, 150–154 funding gap, 25–26 for determining to use EO products, 5–6 mapping with EO, 76 efficiency of agricultural water use example for water, 24–25 determining EO products, 165t7.14 Inland and Near-Coastal Water Quality Remote water quality example for determining EO Sensing Working Group, GEO, 123 products, 163, 164t7.13 inland aquatic macrophytes, 122, 125, 129 Guo, X., 190 inland water definition of, 121 H mapping, 130–131 Hall, D. K., 117 quality, 123–125, 126–128t6.12, 218 Hanjra, M. A., 23 in-situ measurements, EO combined with, 66–69 Hartfield, K., 17 Institute of Water Modeling (Bangladesh), 20 Harvard Water Program, 57 Integrated Flood Analysis System, International He, X., 190 Centre for Water Hazard and Risk health issues, 50 Management, 20 Hess, L., 116 interception, 89 Hestir, E. L., 15 internal World Bank programs, RS data using, Historical Water Lending, by Subsector 39–43 and Region, Fiscal Year 2009-14, International Center for Integrated Water 36 f 2.1, 36 f 2.2 Resources Management of the United Hoffman, J., 111 Nations Educational, Scientific, and Cultural Hong, Y., 86, 186, 188 Organization (UNESCO), 18 Hossain, F., 188, 191 International Centre for Integrated Mountain Hossain, M., 191 Development (Nepal), 20 Hou, B., 190 International Flood Network, 20 Hu, Z., 178 International Groundwater Resources Assessment Huete, A. R., 90, 93, 107 Center (IGRAC), 54 Hufschmidt, M., 57 International Precipitation Working Group, 68 Hutley, L. B., 90, 93 irrigated land cover types in Krishna Basin, India, Huxman, T. E., 90 108 f 6.12 226  |  I N D E X irriGATEWAY, CSIRO, 210, 211 f  B.4 Liu, L., 190 irrigation Liu, Q., 100 design, 50 Liu, X., 190 mapping with EO, 76–77 Liu, Y. Y., 100, 101 monitoring systems, 209–211 long-term data, hydrometeorological variables, operations, 51 53–54 water use efficiency assessments with EO, 76 Lower Gwydir Region, NSW, Australia, 94 f 6.6 Lu, C., 190 J Luo, M., 190 Japan Aerospace Exploration Agency (JAXA), 41, 84, 181 M Japanese Earth Resources Satellite 1 (JERS-1), Ma, Y., 178 JAXA, 180–181 Maas, A., 57 de Jeu, R. A. M., 100, 101 macrophytes, 122, 125, 129 Jimenez, C., 116 Map of AWRA-Derived Total Annual Landscape justification, of using EO data, 151 Water Yields in 2011-12 for Tasmania, Australia, 108 f 6.11 K Map of Irrigated Land Cover Types in the Krishna Kelly, R., 117 Basin, India, 108 f 6.12 KGE (Kling-Gupta efficiency), 189–190 maps and mapping Kim, H., 90 actual evapotranspiration, 93–94, 94 f 6.6 Kling-Gupta efficiency (KGE), 189–190 inland aquatic macrophytes, 122, 125, 129 Krajewksi, W. F., 16 Surface Water Extent in a Flood Event in Bolivia Krishna Basin, India, 108 f 6.12 in March 2014 and Discharge Estimate from Kuenzer, C., 121 Passive Microwave, 207 f  B.1 Marglin, S. A., 57 L marine and estuarine environments, 51 LAI (leaf area index), 93, 103 Marine Water Quality Dashboard, Bureau Lake Erie Harmful Algal Bloom Bulletin, 218 of Meteorology, Australia, 130, 217–218 Lake IJssel, Netherlands, 218 maritime pollution, monitoring, 77 lake levels, data services for, 116–117 Marsh, S., 17 lake water quality improvement example, Maskew Fair, G., 57 163, 164t7.13 Maskey, S., 190 land cover changes, 16 Mattersdorf, G., 90, 93 land management and spatial planning, 50 Matthews, E., 116 Landsat, 80, 81 f 6.1, 91, 113 maturity, product, 150, 154 landscape water yields, AWRA-derived total Mazumder, L. C., 191 annual, 108 f 6.11 McCabe, M. F., 101 landslide monitoring, 206–208 McMahon, A., 26 Langsdale, S., 18 Medium Resolution Imaging Spectrometer Lawford, R., 52, 61n2 (MERIS), 162, 163, 209 leaf area index (LAI), 93, 103 Mei, Y., 174 Lebel, T., 68 Meisner, B. N., 83 Lee, H., 191 meltwater, 117 Lee, S., 117 Merino, M., 186 Leichenko, R. M., 23 MERIS (Medium Resolution Imaging lending, water portfolio, 36–37 Spectrometer), 162, 163, 209 Lending and Analytical and Advisory Activities microwave sensors, 117, 121 Using Remote Sensing, by Bank Region, 44 f 3.2 Minnesota lakes method, 125, 129 Leuning, R., 90, 93, 107 model-data fusion, 68 f 5.1 Levizzani, V., 186, 189 modeled soil moisture products, 96 Li, L., 86 modeling applications, 60–61 Li, M., 87 Moderate Resolution Imaging Spectometer light interactions that drive optical EO involving (MODIS) air, water, and substrate, 124 f 6.14 characteristics of, 81 f 6.1 Limaye, A., 191 description of, 80 Liu, G., 190 drought monitoring systems that use, 107 Liu, J., 190 energy balance methods using data from, 91 I N D E X   |  227 flood mapping, 206, 207 f  B.1 non-water-dedicated project or activity, 44, 47n14 image derived from, 108 f 6.12 normalized difference vegetation index (NDVI), Land Cover Type, 106 90, 93, 209 snow cover and snow water equivalent, North American Land Data Assimilation Systems 121, 179, 212 (NLDAS), 213 for surface water, 113 NSE (Nash-Sutcliffe efficiency), 190, 191n1 for water quality, 130, 162 NSIDC, 100, 121 Murray-Darling basin, Australia Number of Projects Using Remote Sensing in floodplain inundation model, 116 Water-Related Lending and Analytical and Sustainable Yields Project, 214–216 Advisory Activities, by Primary Theme, 44t3.1 MYD10A1 Fractional product, 179 MyOcean Programme, 217 O O’Brien, K. L., 23 N Ocean Topography Experiment (TOPEX)/ Naeimi, V., 100 Poseidon mission, NASA and French Space Nagler, P. L., 90 Agency, 180–181 NASA. See National Aeronautics and Space Olmanson, L. G., 129 Administration Open Landscape Partnership Program, 43, 47n8 Nash-Sutcliffe efficiency (NSE), 190, 191n1 operational hydrology, 57–61 National Aeronautics and Space Administration optical imaging, 112 (NASA) optical remote sensing, 80 basin monitoring, 18 optical water quality, 5, 121–125 GeoGlam Crop Monitor, 210–211 orographic lift, 131–132n2 GRACE mission, 80, 109, 111, 171, 181, 209 orographic precipitation, 174 NASA-USAID-SERVIR Program, 18, 20 overextraction, groundwater, 109 Nile Project, 41b3.1 Overview of Key Characteristics of Soil Moisture SERVIR Program, 18, 20, 42 Sensors aboard Past, Current, and Near- SMAP mission, 100 Future Satellite Platforms, 98–99t6.6 TMPA, 84, 85t6.3, 86–89, 174 Overview of Main Characteristics of Some Widely TOPEX/Poseidon mission, 181 Used Global Satellite-Derived Precipitation National Atlas of Groundwater-Dependent Estimates, 85t6.3 Ecosystems, 216, 217 f  B.7 Overview of Sensors Most Suitable for Estimating National Center for Atmospheric Research, 88 Actual Evapotranspiration from EO Data, National Center for Environmental Prediction, 88 92t6.4 National Drought Mitigation Center (NDMC), 208 Overview of Sensors Most Suitable for Estimating National Oceanic and Atmospheric Administration Vegetation and Land Cover, 104–105t6.7 (NOAA), 20, 130, 208, 211–212 Overview of Sensors Most Suitable for Mapping National Snow Analysis, NOAA, 211–212 Snow Extent, Snow Moisture, and Snow National Snow and Ice Data Center, 121 Water Equivalent, 118–120t6.11 natural capital, 24 Overview of Sensors Most Suitable for Mapping NBI (Nile Basin Initiative), 41b3.1 Surface Water Extent and Height, NCORE (Nile Cooperation for Results Project), 41 114–115t6.10 NDMC (National Drought Mitigation Center), 208 Overview of Sensors Suitable for Estimating NDTI model, actual ET estimates compared to Groundwater, 110t6.9 actual ET measurements from, 95, 96 f 6.7 Overview of Water Issues and Relevant Variables NDVI (normalized difference vegetation index), Provided by Earth Observation, 71–72t5.2 90, 93, 209 Overview of Water Issues and Relevant Variables Nepal, International Centre for Integrated Provided by EO Rearranged to Focus on Mountain Development, 20 Spatial and Temporal Resolution, 73–74t5.3 Netherlands, 23–24, 218 Nikolopoulos, E. I., 174 P Nile Basin Initiative (NBI), 41b3.1 pan evaporation, 89 Nile Cooperation for Results Project (NCORE), 41 Pang, H., 190 NLDAS (North American Land Data Assimilation Papa, F., 116 Systems), 213 Parinussa, M., 100, 101 NOAA (National Oceanic and Atmospheric Part II, ways to read, 63–64 Administration), 20, 130, 208, 211–212 passive microwave imagery, radar and, 112–113 non-sector codes, water-related projects with, 37 passive microwave sensors, 117, 121, 207 f  B.1 228  |  I N D E X passive radiometry EO systems, 97, 98–99t6.6 efficiency of agricultural water use example for passive remote sensing, 80 determining EO products, 165t7.14 passive soil moisture products, 100 water quality example for determining EO Penman-Monteith methods, 92–93, 94–95, 107 products, 163, 164t7.13 Permanent Inter-State Committee for Qureshi, M. E., 23 Drought Control in the Sahel (AGRHYMET-CILSS), 58 R PERSIANN (Precipitation Estimation radar from Remotely Sensed Information altimetry, 116, 180–181 Using Artificial Neural Networks), interferometry, 109, 110t6.9, 111 UC Irvine, 174–175, 190 and passive microwave imaging, 112–113 PERSIANN-CCS (Precipitation Estimation from rainfall estimates, 86 Remotely Sensed Information Using Artificial radiometric resolution, 131 Neural Networks-Cloud Classification rainfall System), UC Irvine, 186–187 estimates in Australia, 87, 88 f 6.3 phased approach, 197 runoff modeling using streamflow physics-based inversion methods, field data simulations, 185–191 requirements and characteristics of EO-based rain-gauges distribution, in real-time, 88 f 6.4 water quality products, 123–125, 161–163 rain shadow, 174 Pipunic, R. C., 90, 93 Rajagopal, S., 190 pixel size, satellite, 69, 78n6 Rajagopalan, B., 16 POES (polar-orbiting operational environmental Rao, S., 190 satellites), 212 Raupach, T., 87 point measurements, variables, 53 real-time data, hydrometeorological polar-orbiting operational environmental variables, 53–54 satellites(POES), 212 real-world uses, RS, 197–199 political economy, 27 record length, data, 149, 152–153 pollution, maritime, monitoring water reflectance spectrum from Eutrophic inland water quality of, 77 body and regions in which different water river basin water management sustainability quality parameters influence the shape of that example, 164–165 spectrum, 124 f 6.15 river water balance accounts, 216 reform cycle, for funding gap, 26 potential evapotranspiration, 3, 89 Regional Centre for Hydrology in Central Asia, 213 precipitation Regional Committee of Water Resources of analysis, 209 the Central American Integration System data from EO observation, 82–89 Secretariat, 59 definition of, 3 Regional Integrated Multi-Hazard Early Warning field data requirements and characteristics of System (RIMES), 59 EO-based products, 155–156 regionalization technique, 56 seasonal, 175 Regional Visualization and Monitoring System, 18, validation of RS data on, 173–175 20, 42 Precipitation Estimation from Remotely Sensed Reichle, R. H., 100 Information Using Artificial Neural Relationship between Water Issues and Water Networks (PERSIANN), UC Irvine, 174–175, Topics and Subtopics in the World Bank 186–187, 190 Water Partnership Program, 70t5.1 Precipitation Estimation from Remotely Sensed Relative Efficiency of Streamflow Simulations Information Using Artificial Neural Forced by Rainfall Algorithms, 189 f 10.3 Networks-Cloud Classification System reliability, data product, 149–150, 153–154 (PERSIANN-CSS), UC Irvine, 186–187 remote sensing (RS). See also Earth observation; Prigent, C., 116 satellite remote sensing; validation of RS data Princeton University, 107 active, 80 actual evapotranspiration, 94–95 Q background on, 1 Qiu, S., 190 bias correction, 61n12 questions false alarms, 19 for determining EO needs, 6b0.3, 145–148 field-based measurements combined with, for determining minimum requirements of EO 66–69 data products, 150–154 funding gap, 26 I N D E X   |  229 global change effects, 16–17 satellite-derived precipitation products (SPPs), institutional frameworks, 27–28 186, 189–190 optical, 80 satellite imagery during flood event, NSW, passive, 80 Australia, 116 f 6.13 potential of, 195–196 satellite remote sensing. See also Earth real-world uses of, 197–199 observation; remote sensing; sensors terminology use of, 7n1, 46n2, 78n3 ET, 175–177 usability of, 60 gravimetry, 109, 110t6.9, 111 water infrastructure, 24–25 of hydrometeorological variables, 167–168 for water scarcity, 12–13 of inland and near-coastal water quality, 122, in World Bank water-related projects, 2–3, 33, 126–128t6.12, 129, 130–131 39–46, 45t3.2 of precipitation, 82–84, 85t6.3, 86–89, 173–175 Remote Sensing-Based Soil Moisture of snow cover, 117–121, 178–180 Monitoring, 97 f 6.8 of soil moisture, 4, 66–67, 95–102, 177–178 Ren, S., 190 streamflow data validation, 111–117, 185–191 Renzullo, L., 87 surface water levels data validation, 111–117 reservoirs. See surface water of vegetation and vegetation cover, 102–108 revisit time, 69 Satellite Water Monitoring and Forecasting RFE2 (African Rainfall Estimation Algorithm System for the Yellow River, 13 Version 2), 190 saturated zone, 4 RIMES (Regional Integrated Multi-Hazard Early Schematic of the Light Interactions That Drive Warning System), 59 Optical EO Involving the Air, Water, and river basin water management sustainability Substrate, 124 f 6.14 example, 164–165 Schneier-Madanes, G., 17 river streamflow, monitoring, 77 Science Systems and Applications, Inc., 211 river water balance accounts, 216 Scipal, K., 100 RMSE measures Scott, R. L., 90 definition of, 171n2 Screening for Adequacy of Field Observations, ET, 176 151b7.2 lake water levels, 181 Seasonal Forecasts Issued by Two Regional precipitation, 174–175 Centers, 58 f 4.3 snow cover and snow water seasonal precipitation, 175 equivalent, 179–180 SEBAL (Surface Energy Balance Algorithm for soil moisture, 177–178 Land), 91, 209–210 streamflow, 186–191 sector codes, water-related lending by, 37 Streamflow Simulations Forced by Rainfall semi-empirical methods, field data requirements Algorithms, 188 f 10.2 and characteristics of EO-based water quality Robinson, D. A., 117 products, 123–125, 161–163 Rodriguez, D. J., 26 sensors. See also Earth observation; MODIS; Rojas, R., 186, 189 remote sensing; satellite remote sensing rooting depth estimate methods, 103, 106 for actual evapotranspiration, 91, 92t6.4, 93–94 Rosema, A., 190 for groundwater, 110t6.9 de Rosnay, P., 178 of inland and near-coastal water quality, 122, Rossow, W. B., 116 126–128t6.12, 129 RS. See remote sensing for inland water mapping, 130–131 rural infrastructures, mapping with EO, 76 for optical water quality, 123 for snow extent, snow moisture, and snow water S equivalent, 117, 118–120t6.11, 121 SADC (Southern Africa Development for soil moisture, 4, 66–67, 95–102 Community), 57–58 for surface water, 113, 114–115t6.10, 116–117 Saito, L., 16 for vegetation and land cover, 102–108 Salas, J. D., 16 Serinaldi, F., 16 sampling. See also examples Serrat-Capdevila, A., 186, 190 errors with hydrometeorological SERVIR Program, 18, 20, 42 variables, 167–168 ShahNewaz, S. M., 191 measurements, 66 Shang, H., 190 temporal, 130 Shao, Q., 87 Sangati, M., 188 Sheffield, J., 88 230  |  I N D E X Sherman, B. S., 90, 93 Surface Energy Balance Algorithm for Land Shum, C. K., 191 (SEBAL), 91, 209–210 Siddique-E-Akbor, A. H., 191 Surface Hydrology Group, Six Optical Water Quality Variables That Can Be Princeton University, 18 Derived from EO Data, 5bES.2 surface water SMAP (Soil Moisture Active Passive) mission, conjunctive use of groundwater and, 20–22, 26 100, 101, 102 data from EO observation, 111–117 Smartshell, 219 definition of, 4, 7n2 Smith, J. A., 16 field data requirements and characteristics of snow cover EO-based products, 159, 160t7.8 data from EO observation, 117–121 map of flood event, 207 f  B.1 definition of, 4–5 transboundary collaboration, 28–29 extent products, 211–213 validation of RS data, 180–182 field data requirements and characteristics Surface Water and Ocean Topography Mission, 116 of EO-based products, 160 sustainable management as hydrometeorological variable, 52 adaptive, 26–28 monitoring, 76 basin water example, 164–165 sensors, 117, 118–120t6.11, 121 conjunctive use of surface water and validation of RS data on, 178–180 groundwater, 21–22 soil moisture investing in, 24–25 data from EO observation, 95–102 water-related challenges, 12 definition of, 4 of water supply, 25 estimates, 20, 22 SVAT (soil-vegetation-atmosphere transfer), 176 field-based measurements, 66–67 field data requirements and characteristics T of EO-based products, 157–158 Tasmania, Australia, 108 f 6.11 monitoring systems, 208–209 Tedesco, M., 117 validation of RS data on, 177–178 temporal resolution, 69, 73–74t5.3, 149, 152 Soil Moisture Active Passive (SMAP) mission, temporal sampling, water quality 100, 101, 102 assessments, 130 soil-vegetation-atmosphere transfer (SVAT), 176 Tennessee Technological University, 20 Southern Africa Development Community terrestrial and freshwater ecosystems, 51 (SADC), 57–58 Terzhagi’s Principle, 111 space-based precipitation, 82–84, 85t6.3 Thiemig, V., 186, 189 space-based soil moisture sensing technology, Thomas, H. A., 57 4, 66–67, 96–101 TIGER initiative, European Space Agency, Spatial Help Desk, 42 19, 42, 47n5 spatial planning and land management, 50 TIGER-NET, 42 spatial resolution, 69, 73–74t5.3, 78n6, 131, time-series gauge precipitation data set, Climate 149, 152 Research Unit, 17 spatial sampling, water quality assessments, 130 TMPA (TRMM Multisatellite Precipitation spatial variability measurements, 66–69 Analysis) system, NASA, 84, 85t6.3, 86–89, spatial water resources monitoring systems, 174, 186–187, 207–208 213–214 Tobin, K. J., 186 spectral resolution, 131 TOPEX (Ocean Topography Experiment)/ SPPs (satellite-derived precipitation products), Poseidon mission, NASA and French Space 84, 86, 88, 186, 189–190 Agency, 180–181 standardized precipitation index, 209 total suspended matter, 122 Stephenson, K., 18 Total Water-Lending, by Water Subsector, 37t2.2 streamflows, validation of RS data, 180–182, transboundary collaboration, 28–29 185–191 transboundary issues, 51 Su, Z., 178 transpiration, 3, 89 Subramanian, A., 28 trapezoid method, 90 suitability, of using EO data, 151–152 triangle method, 90 Sun, F., 190 triple collocation technique, comparing error Sun, W., 190 estimates for soil moisture products derived Sun, Y., 190 from active and passive microwave sensors surface altimeter measurements, 20 using, 101 f 6.9 I N D E X   |  231 TRMM (Tropical Rainfall Measuring Mission) van den Berg, C., 26 Satellite-based precipitation measurements, Van Dijk, A. I. J. M., 90, 93, 101, 107, 213 83 f 6.2, 84, 186, 190 vegetation and vegetation cover Tu, X., 190 classes, 106 Tumbarumba, NSW, Australia, 96 f 6.7 data from EO observation, 102–108 turbidity, 122 definition of, 4 Turk, F. J., 191 drought monitoring, 106–107 two-layer model, energy balance method, 91 field data requirements and characteristics Types of Data Obtained from Earth Observation, of EO-based products, 158–159 82t6.2 indexes, 78n10, 90, 93 Typical Reflectance Spectrum from Eutrophic vegetation drought response index, 208–209 Inland Water Body and Regions in Which vegetation health index, 19, 209 Different Water Quality Parameters Influence van der Velde, R., 178 the Shape of That Spectrum, 124 f 6.15 Venneker, R., 190 Vergara, H., 186, 188 U vertical light attenuation, 122 UNESCO (United Nations Educational, Scientific, Villarini, G., 16 and Cultural Organization), 18, 20 ungauged basins, 56–57 W United Nations Wagner, W., 100, 101 Educational, Scientific, and Cultural Wang, C., 190 Organization (UNESCO), 18, 20 Wang, J., 86 FAO (Food and Agriculture Organization), Wang, L., 178 54–55, 93 Wang, Q., 190 University of Amsterdam, 100 wastewater, 50, 51 University of Arizona, 18 water. See also floods; groundwater; inland water; University of California, Irvine, 174–175 irrigation; surface water University of Maryland, 211 food-water-energy nexus, 22–24 University of Montana, 107 freshwater and terrestrial ecosystems, 51 University of Nebraska, Lincoln, 208 infrastructure, 24–25 upstream water levels, streamflow lake water quality improvement example, simulations for, 191 163, 164t7.13 urban design and management, 51 lending, water portfolio, 36–37 urban infrastructures, mapping with EO, 76 meltwater, 117 U.S. Agency for International Development optical water quality, 5, 121–125 (USAID), 19, 20 relevant variables provided by EO, 69, 71–72t5.2, U.S. Department of Agriculture (USDA), 208, 210–211 73–74t5.3, 75–79 U.S. Drought Monitor, 208–209 river basin water management sustainability Use of Remote Sensing in World Bank Lending and example, 164–165 Analytical and Advisory Activities, by Water river streamflow, monitoring, 77 Subsector, 45t3.2 river water balance accounts, 216 U.S. government upstream water levels, streamflow Global Drought Information System, 107 simulations for, 191 World Bank agreement with, 40–41 wastewater, 50, 51 U.S. snow analysis, 211–212 World Bank water observations, 1–3 World Bank Water Partnership Program topics V ans subtopics, 70t5.1 Valdes, B., 186 water clarity database, 129 Valdés, J. B., 190 water-dedicated project or validation of RS data activity, 44, 47n14 ET, 175–177 Waterdyn25M Model, 209 field-based measurements, 148 Water Global Practice, World Bank Group, hydrometeorological variables, 168, 169–171 34–35, 61n10 precipitation, 173–175 water holes, 19 snow cover and snow water equivalent, 178–180 water infrastructure, 24–26 soil moisture, 177–178 water lending, 36–37, 43–45 streamflows, 180–182, 185–191 Water Observation and Information System surface water levels, 180–182 (WOIS), 42, 47n6 232  |  I N D E X Water Partnership Program, World Bank, 61n11, Wolf, A., 28 69, 70t5.1, 74–78 Wood, E. F., 88 water portfolio, 35–37, 43–46 World Bank/World Bank Group water quality agreement with U.S. Government, 40–41 algal blooms, 130, 216, 218–219 Climate Change Knowledge Portal, 54 of coastal discharge, 77 EO applications in projects by, 1–3, 65–66, definition of, 5, 121 201–204 field data requirements and characteristics EO-related WRM in context of, 65–66 of EO-based products, 160–165 institutions of, 34 inland, 218 Relationship between Water Issues and Water marine, 217–218 Topics and Subtopics in the World Bank monitoring, 75 Water Partnership Program, 70t5.1 optical data obtained from EO, 121–125 RS in water-related projects, 2–3, 33, 39–46, 45t3.2 RS of, 13–15 Water Global Practice, 34–35, 61n10 Water-Related Lending and Analytical and Water Partnership Program of the Water Global Advisory Activities Using Practice, 1, 61n11, 69, 70t5.1, 74–78 Remote Sensing, 43 f 3.1 water portfolio, 35–37 water-related projects, 2–3, 33, 37t2.3, Water-Related Topics and Subtopics Considered 43–46, 66b5.1 in the World Bank Context, 66b5.1 water requirement satisfaction WRM (See water resources management) index (WRSI), 18–19 World Development Report (World Bank, 1992), 33 water reservoirs, identifying and monitoring, 75 World Meteorological Organization, 20 Water Resources Council, 57 WRM. See water resources management water resources management (WRM) WRSI (water requirement activities, 51 satisfaction index), 18–19 aid in determining EO needs, 6b0.3, 7 f  ES.1, Wu, B., 19 145–148 Wu, X., 190 and development, 195 EO data on, 148 X EO used to support, 193–194, 196 Xie, L., 190 satellite soil moisture sensing technology Xue, Y., 190 usage in, 101–102 Water Global Practice, 34–35 Y World Bank context of EO-related, 65–66 Yan, N., 19 World Bank objectives for, 33–34 Yang, J., 190 water resources monitoring systems, 213–214 Yang, K., 178 Water Resources Sector Strategy (World Bank), 34 Yang, L., 19 water scarcity, 12–13 Yang, T., 190 Water Sector Board, World Bank, 36 Yebra, M., 90, 93, 107 watershed management, 51 Yigzaw, W., 191 Water Supply and Sanitation Sector Business Yuan, Z., 190 Strategy (World Bank), 34 water supply systems, 50, 51 Z water systems planning and management, RS Zambrano-Bigiarini, M., 186, 189 applications in, 46 Zeng, Y., 178 water use efficiency assessments, 76 Zhang, L., 190 weather monitoring, 51 Zhang, R., 190 Wen, J., 178 Zhang, Y., 190 Wen, L., 190 Zhao, L., 190 Western Queensland, Australia, 90 Zhao, W., 190 WOIS (Water Observation and Information Zheng, F., 190 System), 42, 47n6 Zhu, L., 190 I N D E X   |  233 ECO-AUDIT Environmental Benefits Statement The World Bank Group is committed to reducing Council (FSC)–certified paper, with nearly all its environmental footprint. In support of this containing 50–100 percent recycled content. commitment, the Publishing and Knowledge The recycled fiber in our book paper is either Division leverages electronic publishing options unbleached or bleached using totally chlorine-free and print-on-demand technology, which is (TCF), processed chlorine-free (PCF), or enhanced located in regional hubs worldwide. Together, elemental chlorine-free (EECF) processes. these initiatives enable print runs to be lowered More information about the Bank’s and shipping distances decreased, resulting in environmental philosophy can be found at http:// reduced paper consumption, chemical use, www.worldbank.org/corporateresponsibility. greenhouse gas emissions, and waste. The Publishing and Knowledge Division follows the recommended standards for paper use set by the Green Press Initiative. The majority of our books are printed on Forest Stewardship