NASA-ISRO SAR (NISAR) Mission Science Users' Handbook

NISAR Science Users’ Handbook NISAR Science Users’ Handbook National Aeronautics and Space Administration

NASA-ISRO SAR Mission (NISAR) NASA-ISRO SAR (NISAR) Mission

National Aeronautics and Space Administration Science Users’ Handbook Jet Propulsion Laboratory California Institute of Technology Pasadena, California Current Project Planned Launch Date: Document Date: www.nasa.gov Phase: C January 2022 November 7, 2018, Version 1

CL# JPL 400-XXXX 12/18 NISAR Science Users’ Handbook i NISAR Science Users’ Handbook

CONTENTS

1 INTRODUCTION 1

2 SCIENCE FOCUS AREAS 5

2.1 SOLID EARTH PROCESSES: EARTHQUAKES, VOLCANOES, AND LANDSLIDES 6 2.2 ECOSYSTEMS: BIOMASS, DISTURBANCE, AGRICULTURE AND INUNDATION 8 2.3 DYNAMICS OF ICE: ICE SHEETS, GLACIERS, AND SEA ICE 9 2.4 APPLICATIONS 11 2.5 OCEAN STUDIES AND COASTAL PROCESSES 14 2.6 DISASTER MONITORING, ASSESSMENT AND RECOVERY 14

3 MISSION MEASUREMENTS AND REQUIREMENTS 17

3.1 MEASUREMENTS OF SURFACE DEFORMATION 17 AND CHANGE 3.2 LANDCOVER AND FOREST CHARACTERIZATION 18 WITH L-BAND SAR 3.3 REQUIREMENTS AND SCIENCE TRACEABILITY 19 3.4 SCIENCE TRACEABILITY TO MISSION REQUIREMENTS 23

4 MISSION CHARACTERISTICS 27

4.1 OBSERVING STRATEGY 27 4.2 REFERENCE SCIENCE ORBIT 27 4.3 MISSION PHASES AND TIMELINE 33 4.4 GROUND SEGMENT OVERVIEW 34 4.5 TELECOMMUNICATIONS 36 4.6 MISSION PLANNING AND OPERATIONS 38 4.7 INSTRUMENT DESIGN 39 4.8 FLIGHT SYSTEMS/SPACECRAFT 43 4.9 PROJECT STATUS 45 ii NISAR Science Users’ Handbook iii NISAR Science Users’ Handbook

5 MISSION DATA PRODUCTS 47 6.6 CRYOSPHERE PRODUCTS- ICE SHEETS 92 6.6.1 Theoretical basis of algorithm 92 5.1 L0 DATA PRODUCTS 49 6.6.2 Implementation approach for algorithm 94 5.2 L1 DATA PRODUCTS 50 6.6.3 Planned output products 100 5.3 L2 DATA PRODUCTS 53 5.4 DATA PRODUCT DELIVERY/HOW TO ACCESS NISAR DATA 53 6.7 CRYOSPHERE PRODUCTS- SEA ICE 100 6.7.1 Theoretical basis of algorithm 101 6 SCIENCE DATA PRODUCTS AND VALIDATION 55 6.7.2 Implementation approach for algorithm 105 APPROACHES 6.7.3 Planned output products 108 6.1 SOLID EARTH SCIENCE PRODUCTS: CO-SEISMIC, 55 TRANSIENT AND SECULAR DISPLACEMENTS 7 ERROR SOURCES 111 6.1.1 Theoretical basis of algorithm 55 7.1 POLARIMETRIC ERROR SOURCES 111 6.1.2 Implementation approach for algorithm 57 7.2 INTERFEROMETRIC ERROR SOURCES 112 6.1.3 Planned output products 62

6.2 ECOSYSTEMS PRODUCTS- BIOMASS 62 8 CALIBRATION AND VALIDATION 117 6.2.1 Theoretical basis of algorithm 63 8.1 BACKGROUND 117 6.2.2 Implementation approach for algorithm 68 8.2 CALIBRATION AND VALIDATION ACTIVITIES 118 6.2.3 Planned output products 73 8.3 CALIBRATION/VALIDATION ROLES AND RESPONSIBILITIES 124 6.3 ECOSYSTEMS PRODUCTS- DISTURBANCE 73 6.3.1 Theoretical basis of algorithm 75 9 CONCLUSIONS 127 6.3.2 Implementation approach for algorithm 77 6.3.3 Planned output products 82 10 ACKNOWLEDGEMENTS 129

6.4 ECOSYSTEMS PRODUCTS- INUNDATION 82 11 REFERENCES 131 6.4.1 Theoretical basis of algorithm 83 6.4.2 Planned output products 86 12 GLOSSARY 145 6.5 ECOSYSTEMS PRODUCTS- CROP MONITORING 87 6.5.1 Theoretical basis of algorithm 87 13 APPENDIX A: HISTORICAL BACKGROUND FOR NISAR 147 6.5.2 Implementation approach for algorithm 89 6.5.3 Planned output products 91 iv NISAR Science Users’ Handbook v NISAR Science Users’ Handbook

14 APPENDIX B: NISAR SCIENCE TEAM 149 17.2 ECOSYSTEMS 186 17.2.1 Biomass 187 14.1 NASA SCIENCE DEFINITION TEAM 149 17.2.2 Biomass disturbance and recovery 189 14.2 ISRO SCIENCE TEAM 150 17.2.3 Agricultural monitoring 190 14.3 PROJECT SCIENCE TEAM 151 17.2.4 Wetlands and inundation 192

15 APPENDIX C: KEY CONCEPTS 153 17.3 CRYOSPHERE 193 17.3.1 Ice sheets 195 15.1 BASIC RADAR CONCEPTS: RADAR IMAGING, POLARIMETRY, 153 AND INTERFEROMETRY 17.3.2 Glaciers and mountain snow 197 15.1.1 Synthetic Aperture Radar (SAR) 153 17.3.3 Sea ice 199 15.1.2 Polarimetry 154 17.3.4 Permafrost 202 15.1.3 Interferometric Synthetic Aperture Radar (InSAR) 155 17.4 APPLICATIONS 203 15.2 DEFORMATION-RELATED TERMINOLOGY 156 17.4.1 Ecosystems: Food security 206 15.2.1 Deformation and displacement 158 17.4.2 Ecosystems: Forest resource management 207 15.2.2 Strain, gradients, and rotation 159 17.4.3 Ecosystems: Wildland fires 208 15.2.3 Stress and rheology 159 17.4.4 Ecosystems: Forest disturbance 210 15.3 ECOSYSTEMS-RELATED TERMINOLOGY 159 17.4.5 Ecosystems: Coastal erosion and shoreline change 210 15.3.1 Biomass 159 17.4.6 Geologic Hazards: Earthquakes 211 15.3.2 Disturbance 160 17.4.7 Geologic Hazards: Volcanic unrest 212 15.3.3 Recovery 160 17.4.8 Geologic Hazards: Sinkholes and mine collapse 213 15.3.4 Classification 160 17.4.9 Geologic Hazards: Landslides and debris flows 213 17.4.10 Hazards: Anthropogenic technological disasters 213 16 APPENDIX D: BASELINE LEVEL 2 REQUIREMENTS 163 17.4.11 Critical Infrastructure: Levees, dams, aqueducts 214

16.1 LEVEL 2 SOLID EARTH 163 17.4.12 Critical Infrastructure: Transportation 214 16.2 LEVEL 2 CRYOSPHERE 163 17.4.13 Critical Infrastructure: Facility situational awareness 214 16.3 LEVEL 2 ECOSYSTEMS 165 17.4.14 Critical Infrastructure: Arctic domain awareness 215 16.4 LEVEL 2 URGENT RESPONSE 167 17.4.15 Maritime: Hurricanes and wind storms 216 17.4.16 Maritime: Sea ice monitoring 216 17 APPENDIX E: NISAR SCIENCE FOCUS AREAS 169 17.4.17 Maritime: Coastal ocean circulation features 217

17.1 SOLID EARTH 169 17.1.1 Earthquakes and seismic hazards 170 17.1.2 Volcano hazards 175 17.1.3 Landslide hazards 176 17.1.4 Induced seismicity 178 17.1.5 Aquifer systems 181 17.1.6 Glacial isostatic adjustment 186 vi NISAR Science Users’ Handbook vii NISAR Science Users’ Handbook

17.3 CRYOSPHERE 193 17.4.26 Underground Reservoirs: Oil and gas production 224 17.3.1 Ice sheets 195 17.4.27 Underground Reservoirs: Induced seismicity 224 17.3.2 Glaciers and mountain snow 197 17.4.28 Rapid damage assessment 226 17.3.3 Sea ice 199 17.3.4 Permafrost 202 18 APPENDIX F: IN SITU MEASUREMENTS FOR CAL/VAL 229

17.4 APPLICATIONS 203 19 APPENDIX G: RADAR INSTRUMENT MODES 235 17.4.1 Ecosystems: Food security 206 17.4.2 Ecosystems: Forest resource management 207 20 APPENDIX H: SCIENCE TARGET MAPS 255 17.4.3 Ecosystems: Wildland fires 208 17.4.4 Ecosystems: Forest disturbance 210 21 APPENDIX I: DATA PRODUCT LAYERS 257 17.4.5 Ecosystems: Coastal erosion and shoreline change 210 17.4.6 Geologic Hazards: Earthquakes 211 17.4.7 Geologic Hazards: Volcanic unrest 212 17.4.8 Geologic Hazards: Sinkholes and mine collapse 213 17.4.9 Geologic Hazards: Landslides and debris flows 213 17.4.10 Hazards: Anthropogenic technological disasters 213 A B 17.4.11 Critical Infrastructure: Levees, dams, aqueducts 214 17.4.12 Critical Infrastructure: Transportation 214 17.4.13 Critical Infrastructure: Facility situational awareness 214 17.4.14 Critical Infrastructure: Arctic domain awareness 215 17.4.15 Maritime: Hurricanes and wind storms 216 17.4.16 Maritime: Sea ice monitoring 216 17.4.17 Maritime: Coastal ocean circulation features 217 17.4.18 Maritime: Ocean surface wind-speed 217 17.4.19 Maritime: Iceberg and ship detection 218 17.4.20 Maritime: Oil spills 218 17.4.21 Maritime/Hydrology: Floods hazards 219 17.4.22 Hydrology: Flood forecasting 220 17.4.23 Hydrology: Coastal inundation 221 17.4.24 Hydrology: Soil moisture 222 17.4.25 Underground Reservoirs: Groundwater withdrawal 223 viii NISAR Science Users’ Handbook ix NISAR Science Users’ Handbook

Updated for Baseline Left-Looking Only Mode List of Authors

Falk Amelung Richard Forster Tapan Misra University of Miami University of Utah Space Applications Centre, NISAR, 2018. NASA-ISRO SAR (NISAR) Mission Science Users’ Handbook. NASA Indian Space Research Organization Gerald Bawden Jet Propulsion Laboratory. 263 pp. Margaret Glasscoe NASA Headquarters Jet Propulsion Laboratory, Frank Monaldo

California Institute of Technology National Oceanic and Adrian Borsa Atmospheric Administration Scripps Institution of Bradford Hager Oceanography, UC, San Diego Massachusetts Institute of Technology Sandip Oza Space Applications Centre, Sean Buckley Scott Hensley Indian Space Research Organization Jet Propulsion Laboratory, Jet Propulsion Laboratory, California Institute of Technology California Institute of Technology Matthew Pritchard Cornell University Susan Callery Benjamin Holt Jet Propulsion Laboratory, Jet Propulsion Laboratory, Eric Rignot California Institute of Technology California Institute of Technology University of California, Irvine

Manab Chakraborty Cathleen Jones Paul Rosen Space Applications Centre, Jet Propulsion Laboratory, Jet Propulsion Laboratory, Indian Space Research Organization California Institute of Technology California Institute of Technology

Bruce Chapman Ian Joughin Sassan Saatchi Jet Propulsion Laboratory, Applied Physics Laboratory, Jet Propulsion Laboratory, California Institute of Technology University of Washington California Institute of Technology

Anup Das Josef Kellndorfer Priyanka Sharma Space Applications Centre, Earth Big Data, LLC. Jet Propulsion Laboratory, Indian Space Research Organization California Institute of Technology Raj Kumar Andrea Donnellan Space Applications Centre, Marc Simard Jet Propulsion Laboratory, Indian Space Research Organization Jet Propulsion Laboratory, California Institute of Technology California Institute of Technology Rowena Lohman Ralph Dubayah Cornell University Mark Simons University of Maryland California Institute of Technology Zhong Lu Kurt Feigl Southern Methodist University Paul Siqueira University of Wisconsin University of Massachusetts, Franz Meyer Amherst Eric Fielding University of Alaska, Fairbanks Jet Propulsion Laboratory, Howard Zebker California Institute of Technology Stanford University

NISAR Science Users’ Handbook 1 NISAR Science Users’ Handbook

1 INTRODUCTION

The NASA-ISRO Synthetic Aperture Radar (SAR), or The Earth Science Division (ESD) within the Science NISAR (Figure 1-1), mission is a multi-disciplinary Mission Directorate (SMD) at NASA Headquarters has radar mission to make integrated measurements to directed the Jet Propulsion Laboratory (JPL) to manage understand the causes and consequences of land the United States component of the NISAR project. surface changes. NISAR will make global measure- ESD has assigned the Earth Science Mission Program ments of the causes and consequences of land Office (ESMPO), located at Goddard Space Flight NISAR PROVIDES surface changes for integration into Earth system Center (GSFC), the responsibility for overall program A MEANS OF models. NISAR provides a means of disentangling and management. clarifying spatially and temporally complex phenome- DISENTANGLING na, ranging from ecosystem disturbances, to ice sheet The NISAR mission is derived from the Deformation, AND CLARIFYING collapse and natural hazards including earthquakes, Ecosystem Structure and Dynamics of Ice (DESDynI) tsunamis, volcanoes and landslides. The purpose of radar mission concept, which was one of the four Tier 1 SPATIALLY AND this handbook is to prepare scientists and algorithm missions recommended in the 2007 Decadal Survey. To TEMPORALLY developers for NISAR by providing a basic description satisfy requirements of three distinct scientific commu- of the mission and its data characteristics that will nities with global perspectives, as well as address the COMPLEX allow them to take full advantage of this comprehen- potentials of the system for new applications, the NISAR PHENOMENA, sive data set when it becomes available. system comprises a dual frequency, fully polarimetric radar, with an imaging swath greater than 240 km. This RANGING FROM NISAR is a joint partnership between the National design permits complete global coverage in a 12-day Aeronautics and Space Administration (NASA) and the exact repeat to generate interferometric time-series ECOSYSTEM Indian Space Research Organization (ISRO). Since the and perform systematic global mapping of the changing DISTURBANCES, TO 2007 National Academy of Science “Decadal Survey” surface of the Earth. The recommended Lidar component report, “Earth Science and Applications from Space: of DESDynI will be accomplished with the GEDI mis- ICE SHEET COLLAPSE National Imperatives for the Next Decade and Beyond,” sion (Global Ecosystem Dynamics Investigation Lidar). AND NATURAL NASA has been studying concepts for a Synthetic NISAR’s launch is planned for January 2022. After a Aperture Radar mission to determine Earth change 90-day commissioning period, the mission will conduct HAZARDS INCLUDING in three disciplines: ecosystems (vegetation and the a minimum of three full years of science operations EARTHQUAKES, carbon cycle), deformation (solid Earth studies), and with the L-band SAR in a near-polar, dawn-dusk, frozen, cryospheric sciences (primarily as related to climatic sun-synchronous orbit to satisfy NASA’s requirements; TSUNAMIS, VOLCANOES [ Figure 1-1] drivers and effects on sea level). In the course of these ISRO requires five years of operations with the S-band studies, a partnership with ISRO developed, which SAR. If the system does not use its fuel reserved excess Artist’s concept of NASA–Indian Space Research AND LANDSLIDES. led to a joint spaceborne mission with both L-band capacity during the nominal mission, it is possible to Organization Synthetic Aperture Radar (NISAR) in orbit. The mission will produce L-band (at 25-cm wave- and S-band SAR systems onboard. The current 2018 extend mission operations further for either instrument. length) polarimetric radar images and interferometric Decadal Survey, “Thriving on Our Changing Planet: A data globally, and comparable S-band (at 12-cm Decadal Strategy for Earth Observation from Space”, NISAR’s science objectives are based on priorities wavelength) data over India and targeted areas around confirms the importance of NISAR and encourages the identified in the 2007 Decadal Survey and rearticulated the world. Credit: NASA/JPL-Caltech international partnership between NASA and ISRO. in the 2010 report on NASA’s Climate-Centric Architec- ture. NISAR will be the first NASA radar mission to systematically and globally study solid Earth, ice masses, 2 NISAR Science Users’ Handbook 3 NISAR Science Users’ Handbook

and ecosystems. NISAR will measure ice mass and the • Comprehensive biomass assessment in low TABLE 1-1. NISAR CHARACTERISTICS

land surface motions and changes, ecosystem distur- biomass areas where dynamics are greatest, Nisar Characteristic: Enables: bances, and biomass, elucidating underlying processes and global disturbance assessments, agricul- L-band (24 cm wavelength) Foliage penetration and and improving fundamental scientific understand- tural change, and wetlands dynamics, informing interferometric persistence ing. The measurements will improve forecasts and carbon flux models at the most critical spatial and S-band (12 cm wavelength) Sensitivity to light vegetation assessment of changing ecosystems, response of ice temporal scales; SweepSAR1 technique with Global data collection sheets, and natural hazards. NASA also supports use of • In combination with GEDI and other missions, imaging swath > 240 km the NISAR data for a broad range of applications that comprehensive global biomass to set the decadal benefit society, including response to disasters around boundary conditions for carbon flux models; Polarimetry (single/dual/quad) Surface characterization and the world. In addition to the original NASA objectives, biomass estimation NISAR WILL BE • Complete assessments of the velocity state of ISRO has identified a range of applications of partic- 12-day exact repeat Rapid sampling Greenland’s and Antarctica’s ice sheets, each ular relevance to India that the mission will address, THE FIRST NASA month over the mission life, as a key boundary 3 – 10 meters mode-dependent Small-scale observations including monitoring of agricultural biomass over India, condition for ice sheet models; SAR resolution RADAR MISSION TO monitoring and assessing disasters to which India responds, studying snow and glaciers in the Himalayas, • Regular monitoring of the world’s most dynamic 3 years science operations Time-series analysis SYSTEMATICALLY AND (5 years consumables) and studying Indian coastal and near-shore oceans. mountain glaciers; GLOBALLY STUDY • Comprehensive mapping of sea ice motion and Pointing control < 273 arcseconds Deformation interferometry All NISAR science data (L- and S-band) will be freely SOLID EARTH, ICE deformation, improving our understanding of Orbit control < 350 meters Deformation interferometry available and open to the public, consistent with the ocean-atmosphere interaction at the poles; MASSES, AND long-standing NASA Earth Science open data policy. > 30% observation duty cycle Complete land/ice coverage • A rich data set for exploring a broad range of With its global acquisition strategy, cloud-penetrating Left pointing capability Polar coverage, north and south ECOSYSTEMS. applications that benefit from fast, reliable, and capability, high spatial resolution, and 12-day repeat regular sampling of virtually any areas of interest 1 pattern, NISAR will provide a reliable, spatially dense SweepSAR is a technique to achieve wide swath at full resolution. See Section 4.7 for a more detailed description. on land or ice. These include infrastructure moni- time-series of radar data that will be a unique resource toring, agriculture and forestry, disaster response, for exploring Earth change (Table 1-1). aquifer utilization, and ship navigability.

Anticipated scientific results over the course of the mission include: • Comprehensive assessment of motion along plate boundaries that are on land, identifying areas of increasing strain, and capturing signatures of several hundred earthquakes that will contribute to our understanding of fault systems; [ Figure 1-2] • Comprehensive inventories of global volcanoes, NISAR will image Earth’s dynamic surface over time. NISAR will provide information on changes in ice sheets and their state of activity and associated risks; glaciers, the evolution of natural and managed ecosystems, earthquake and volcano deformation, subsidence from groundwater and oil pumping, and the human impact of these and other phenomena (all images were open source). NISAR Science Users’ Handbook 5 NISAR Science Users’ Handbook

2 SCIENCE FOCUS AREAS

Earth’s land and ice surface is constantly changing and 3. Apply NISAR’s unique data set to hazard identifi- interacting with its interior, oceans and atmosphere. In cation and mitigation;

response to interior forces, plate tectonics deform the 4. Providing information to support disaster response surface, causing earthquakes, volcanoes, mountain and recovery. building and erosion. These events shaping the Earth’s 5. Provide observations of relative sea level rise from surface can be violent and damaging. Human and nat- melting land ice and land subsidence. ural forces are rapidly modifying the global distribution and structure of terrestrial ecosystems on which life NISAR will provide systematic global measurements to depends, causing steep reductions in species diversity, characterize processes, frequent measurements to un- endangering sustainability, altering the global carbon derstand temporal changes, and a minimum three-year cycle and affecting climate. Changes in ice sheets, sea duration to estimate long-term trends and determine ice, and glaciers are key indicators of these climate ef- subtle rates and rate changes. NISAR will serve the IN RESPONSE TO fects and are undergoing dramatic changes. Increasing objectives of a number of major science disciplines melt rates of landfast ice contribute to sea level rise. INTERIOR FORCES, and will meet the needs of a broad science community with numerous applications, including earthquakes, PLATE TECTONICS NISAR addresses the needs of Solid Earth, Ecosystems, volcanoes, landslides, ice sheets, sea ice, snow and and Cryospheric science disciplines, and provides data DEFORM THE glaciers, coastal processes, ocean and land parameter for many applications. NISAR is an all-weather, global retrieval, and ecosystems. In addition, NISAR will play a SURFACE, CAUSING geodetic and polarimetric radar-imaging mission with role in monitoring and assessment of natural disasters the following key scientific objectives: EARTHQUAKES, such as floods, forest fires and earthquakes. 1. Determine the likelihood of earthquakes, volcanic VOLCANOES, eruptions, landslides and land subsidence; NISAR will meet the needs of a broad end-user com- MOUNTAIN BUILDING, 2. Understand the dynamics of carbon storage and munity with numerous applications, including assess- uptake in wooded, agricultural, wetland, and ing geologic and anthropogenic hazards, monitoring AND EROSION. permafrost systems; critical infrastructure for risk management, supporting

3. Understand the response of ice sheets to climate agriculture and forestry agencies, identifying pollu- change, the interaction of sea ice and climate, and tion in coastal waters, and evaluating ground surface impacts on sea level rise worldwide. changes associated with fluid extraction, e.g., groundwater withdrawal during droughts or elevation changes

Applications key objectives are to: associated with oil or gas production. In addition, NISAR will play a role in response to and recovery from natural 1. Understand dynamics of water, hydrocarbon, disasters such as floods, forest fires and earthquakes. and sequestered CO2 reservoirs, which impact societies;

2. Provide agricultural monitoring capability to support sufficient food security objectives; Jesse Kelpszas/Shutterstock 6 NISAR Science Users’ Handbook 7 NISAR Science Users’ Handbook

NISAR observations will address several science and rapid ice loss and significantly accelerated rates provide information about the underlying processes at at a given point on the fault, but the characteristics applications area that the 2018 Decadal Survey recom- of sea-level rise, and in the case of land, changes work. NISAR will uniquely address several questions are also best constrained by combining coseismic mends progress in: in strain rates that impact and provide critical posed in NASA’s Challenges and Opportunities for Earth displacement information, such as from NISAR, with

• Determining the extent to which the shrinking of insights into earthquakes, volcanic eruptions, Surface and Interior (Davis et al, 2016) Report: seismic data (e.g., Pritchard et al., 2006; 2007; glaciers and ice sheets, and their contributions landslides and tectonic plate deformation. 1. What is the nature of deformation associated with Duputel et al., 2015). These estimates of fault slip to sea-level rise is accelerating, decelerating or plate boundaries, and what are the implications parameters then provide key input into mechanical remaining unchanged; 2.1 SOLID EARTH PROCESSES: for earthquakes, tsunamis and other related models of faults and the surrounding crust and upper EARTHQUAKES, VOLCANOES, mantle, estimates of stress change on neighboring • Quantifying trends in water stored on land (e.g., natural hazards? AND LANDSLIDES faults, and inform our basic understanding of regional in aquifers) and the implications for issues such 2. How do tectonic processes and climate variabil- seismic hazards. as water availability for human consumption and Society’s exposure to natural hazards is increasing. ity interact to shape Earth’s surface and create irrigation; Earthquakes threaten densely populated regions on the natural hazards? Measurements of secular velocities within tectonic U.S. western coast, home to about 50 million citizens • Understanding alterations to surface characteris- 3. How do magmatic systems evolve, under what plate boundary regions place constraints on models and costly infrastructure. Volcanic eruptions endanger tics and landscapes (e.g., snow cover, snow melt, conditions do volcanoes erupt, and how do erup- of fault physics, contributing to estimates of long- many areas of the Earth and can disrupt air travel. landslides, earthquakes, eruptions, urbanization, tions and volcano hazards develop? term seismic hazard. NISAR will enable imaging Many natural hazards subtly change and deform the land-cover and land use) and the implications 4. What are the dynamics of Earth’s deep interior, Earth’s plate boundary zones at depth, sampling the land surface resulting in catastrophic events such for applications, such as risk management and and how does Earth’s surface respond? range of different tectonic styles, capturing plate as landslides. Properly preparing for, mitigating, and resource management; boundaries at different stages of the earthquake responding to nature’s disasters requires detecting, • Assessing the evolving characteristics and health NISAR will also address to varying degrees several cycle, and informing regional assessments of seismic measuring, and understanding these slow-moving of terrestrial vegetation and aquatic ecosystems, questions posed in the 2018 Decadal Survey by the hazard. processes before they trigger a disaster. which is important for understanding key conse- Earth Surface and Interior Panel: quences such as crop yields, carbon uptake and Detecting and quantifying transient deformation NISAR provides an opportunity to monitor, mitigate, and 1. How can large-scale geological hazards be accu- biodiversity; plays an essential role in improving our under- respond to earthquakes, volcanoes and landslides that rately forecast in a socially relevant timeframe? standing of fundamental processes associated with • Examining movement of land and ice surfaces result in land surface deformation (Figure 2-1). The 2. How do geological disasters directly impact the tectonics, subsurface movement of magma and to determine, in the case of ice, the likelihood of magnitude and dynamics of these surface expressions Earth system and society following an event? volcanic eruptions, landslides, response to changing 3. How will local sea level change along coastlines surface loads and a wide variety of anthropogenic around the world in the next decade to century? phenomena. Aseismic and post-seismic fault slip transients, volcanic and landslides deformation, and 4. What processes and interactions determine the local subsidence and uplift due to migration of crustal “Head” rates of landscape change? fluids occur globally over temporal and spatial scales 5. How much water is traveling deep underground, ranging from sub-daily to multi-year, and tens of and how does it affect geological processes and “Neck” meter to hundreds of kilometers. Many eruptions are “Toe” water supplies? U.S./Mexico border preceded by surface deformation induced by moving magma in the subsurface. However, we find that Measuring displacements associated with earthquakes 2010 M 7.2 EL MAYOR–CUCAPAH MOUNT ETNA, ITALY SLUMGULLIAN LANDSLIDE, w periods of magma movement do not always result in EARTHQUAKE COLORADO is essential for describing which parts of a fault have an eruption. Systematic measurement of deformation ruptured and which have not, but may have been [Figure 2-1] over volcanoes should help clarify the reason. Simi- brought closer to failure, and for constraining estimates NISAR will measure surface deformation to determine the likelihood of earthquakes, volcanic eruptions and larly, many landslides move intermittently, and may of the distribution of fault slip in the subsurface. Seis- landslides. (Left) 2010 Mw 7.2 El Mayor – Cucapah earthquake shown in L-band UAVSAR at north end of have periods of increased rates of slow sliding before mic data provides estimates of other rupture character- rupture. Surface fracturing and right-lateral displacement is apparent (after Donnellan et al, 2018 submitted; catastrophic run out. NISAR will enable detection and istics, such as the speed at which the rupture propa- image NASA/JPL-Caltech/GeoGateway). (Middle) Mount Etna deformation from the C-band ERS-1 satellite inventory of slow moving landslides, enabling better (after Lundgren et al; 2004; image ESA/NASA/JPL-Caltech). (Right) Slumgullian landslide inversion of L-band gates along the fault and the rate at which slip occurs understanding of variations in movement and how UAVSAR from four images in April 2012 (Delbridge et al; 2016). mass movement is triggered. 8 NISAR Science Users’ Handbook 9 NISAR Science Users’ Handbook

2.2 ECOSYSTEMS: BIOMASS, NISAR will address the following questions: quantitative understanding of the role of terrestrial The upcoming NISAR mission will provide dependable DISTURBANCE, AGRICULTURE • How do changing climate and land use in forests, ecosystems in atmospheric CO2 absorption is limited by observations throughout the year and at repeat periods AND INUNDATION wetlands and agricultural regions affect the car- large uncertainties in two areas: estimates of current that are on par with the cycles that biomass, distur- bon cycle and species habitats? carbon storage in above ground forest biomass, and bance, agriculture and inundation undergo. Hence, The world’s growing population is experiencing unprec- large uncertainties in changes in biomass. From 1990 the NISAR mission will serve as a new foundation for edented changes to our climate through intensifying • What are the effects of disturbance on ecosystem to 2000, the global area of temperate forest increased observing these important ecological environments. events such as floods, droughts, wildfires, hurricanes, functions and services? by almost 3 million hectares per year, while deforesta- tornadoes and insect infestations and their related The NISAR radar is able to image the landscape using tion in the tropics occurred at an average rate exceed- 2.3 DYNAMICS OF ICE: ICE SHEETS, health effects. These impacts are putting pressure on the unique capability of its radio waves that penetrate ing 12 million hectares per year. Uncertainty in biomass GLACIERS, AND SEA ICE our landscapes and ecosystems that generate food, into the forest canopy and scattering from large woody change is greatest in tropics and more uncertain fiber, energy and living spaces for a growing global components (stems and branches) that constitutes the NISAR will address how ice masses interrelate with than changes in forested area (Millennium Ecosystem population. It is imperative to understand the connec- bulk of biomass and carbon pool in forested eco- global climate and sea level (Figure 2-3). Ice sheets Assessment Synthesis Report, 2005). tions between natural resource management and ecosystems (Figure 2-2). The sensitivity of backscatter and glaciers are the largest contributors to sea level system responses to create a sustainable future. The measurements at different wavelengths and polariza- rise with a potential to raise sea level by several tens To feed a growing population of more than 8 billion, 2018 Decadal Survey asks, “What are the structure, tions to the size and orientation of woody components, of centimeters, or more than one meter in the coming food production and supply occur on a global basis. In function, and biodiversity of Earth’s ecosystems, and and their density, makes radar sensors suitable for century. Summer sea ice cover is decreasing drastically order to better guide policy and decision making, na- how and why are they changing in time and space?” It measurements of live, above-ground, woody biomass and may vanish entirely within the next decades. Over tional and international organizations work to transpar- also specifically calls out the need to quantify biomass (carbon stock) and structural attributes, such as volume the satellite period of observations, Arctic sea ice has ently monitor trends and conditions of agriculture in a and characterize ecosystem structure to assess carbon and basal area. NISAR will resolve vegetation biomass thinned, shifted from a predominately perennial to timely basis. Because of the variable nature of planting uptake from the atmosphere and changes in land cover, over a variety of biomes, including low-biomass and seasonal ice cover, and reduced in extent at the end and harvesting practices, efforts such as this are man- and to support resource management. regenerating forests globally, will monitor and identify of summer by nearly 30 percent. Collectively, these power intensive and time-consuming tasks. changes of forest structure and biomass from dis- effects mean that despite their remote location, chang- NISAR radar data will address the distribution of turbances such as fire, logging, or deforestation, and es in ice have global economic and health implications During natural disasters, first responders often look vegetation and biomass to understand changes and characterize post disturbance recovery. as climate changes. The 2018 Decadal Survey prioritiz- to NASA to provide timely and valuable information trends in terrestrial ecosystems and their functioning as es observations in “Understanding glacier and ice sheet to assist their work to mitigate damage and assess carbon sources and sinks and characterize and quan- Changes and degradation of terrestrial ecosystems are contributions to rates of sea-level rise and how likely destruction by these common tragic events. Many tify changes resulting from disturbance and recovery. leading to steep reductions in biodiversity. Additionally, they are to impact sea-level rise in the future.” It asks, federal agencies and university researchers that study “How much will sea level rise, globally and regionally, wetlands have difficulty evaluating the health of our over the next decade and beyond, and what will be the waterways and wetlands due to lack of information re- role of ice sheets and ocean heat storage?” NISAR will Azimuth (km) N garding the ebb and flow of flood waters during normal 0.0 2.0 4.0 6.0 8.0 address the following questions: 0.0 and extreme seasonal inundation. Incidence Angle (deg) • Will there be catastrophic collapse of the major 30 0.5 ice sheets, including Greenland and West Antarctic 35 Among the organizations that respond to flooding di- 1.0 and, if so, how rapidly will this change occur? 40 sasters are state and local agencies, as well as federal Boundary 1.5 45 Orchard agencies, such as Federal Emergency Management • What will be the resulting time patterns of

Slant Range (km) Rice Water sea-level rise at the global and regional level? 2.0 50 Agency (FEMA), the National Oceanic and Atmospheric Dryland Crop Urban Administration (NOAA), and the United States Geolog- • How are mountain glaciers and ice caps world- 0 5 10 20 km [ Figure 2-2] ical Survey (USGS). International aid in the event of wide changing in relation to climate, and what is natural disasters caused by flooding often includes data NISAR will determine the contribution of Earth’s biomass to the global carbon budget and characterize their impact on sea level now and in the future? sharing arrangements to help our allies respond to the ecosystem disturbance and impacts on biodiversity. (Left) Delineated forest stands using L-band ALOS-2 • How rapidly will the Arctic sea ice cover continue humanitarian crises that flooding can cause. radar using PolInSAR in northern Sweden. Forest stands are delineated in white (Neumann et al; 2012). to thin and to decrease in summer extent? (Right) Classified crops in southeast China using L-band ALOS data (Zhang et al; 2009). 10 NISAR Science Users’ Handbook 11 NISAR Science Users’ Handbook

tions between the ice, ocean, and atmosphere, and structure of the mantle, with different structural models 0 m yr–1 11,000 m yr–1 their future behavior. NISAR observations of ice motion predicting different patterns of surface deformation. over both the Arctic and Antarctic sea ice covers will enable a unique, comprehensive examination of the 2.4 APPLICATIONS significantly different responses to climate forcing that NISAR will add a tremendous data set to create new are occurring between the two polar regions. and greatly improve upon many applications that use earth observation data (see Appendix E, section Mountain glaciers and ice caps are among the most 17.4). With frequent, repeated observations over important indicators of climate change, and further- hazard-prone areas, NISAR will provide substantial data more provide fresh water. The Himalayas is the largest to guide development of applications and associated and highest mountain range in the world and plays a scientific studies. All NISAR data products will be freely significant role in the regional hydrological cycle and available through a web portal. This way, the nation’s climate in the central and south Asia. The Himalayan investment in land surveys remotely acquired from region, comprising the highest number of mountain space can be widely used by a variety of agencies and glaciers in the world, has a unique mass-energy individuals. exchange regime that may have serious impact on climate change. Systematic observations of snow-ice NISAR will support applications across five main areas, extent, density and thickness, glacial inventory and namely critical infrastructure monitoring (Figure 2-4), glacial movements and mass-balance of glaciers will hazard assessment (e.g. earthquakes, volcanoes, improve our understanding of underlying processes landslides, floods, sinkholes, etc.), maritime and coastal acting on them and their future behavior under global waters situational awareness, ecosystem services, climate change scenarios. NISAR radar, with its greater and underground reservoir management (Figure 2-5), [ Figure 2-3] NISAR will measure changes in glacier and ice sheet motion, sea ice, and mountain glaciers to determine penetration depth, large swath, and frequently repeated the latter of which encompasses water, oil, and gas how global climate and ice masses interrelate and how melting of land ice raises sea level. (Top left) observations, will enable the study of snow and the reservoirs. In addition to urgent response, NISAR can be Canadian RADARSAT mission show the rapid speed up of Jakobshavn Isbrae in Greenland between global distribution of glaciers at much improved spa- used to collect pre-event information for mitigation and February 1992 and October 2000 (Joughin et al, 2004a). (Top right) Ice flow of the Antarctic ice sheet tio-temporal scales. from ALOS PALSAR, Envisat ASAR, RADARSAT-2, and ERS-1/2 satellite radar interferometry (Rignot et al; monitoring, as well as to collect data during and after 2011a). (Bottom left) UAVSAR L-band sea ice image, which includes old ice, first year ice, and an open disasters or other impactful events support response Earth is continuously readjusting to redistribution of lead. (Bottom right) Surface velocity map for the Wrangell-St. Elias Mountains, the Chugach Mountains/ and recovery. In many cases, operational products can water and ice masses associated with the retreat of the Kenai Peninsula, the Alaska Range, and the Tordrillo Range using L-band radar (Burgess et al, 2013). be derived from NISAR data, particularly for ecosystems Pleistocene ice sheets and ongoing melting of remain- services where hourly-to-daily updates are not needed. ing glaciers, ice caps and ice sheets. The readjustment, In addition, NISAR’s Applications discipline area encom- also known as glacial-isostatic adjustment (GIA), passes science topics not addressed directly be NISAR Flow rates of outlet glaciers around many parts of will measure velocities of the Greenland and Antarctic includes both an instantaneous elastic response to re- science requirements, e.g., soil moisture measurement, Greenland and Antarctica have increased significantly, ice sheets through time, will determine the time- cent changes in load as well as a delayed visco-elastic snow inventory, and atmospheric sciences. More de- more than doubling in some cases. These accelerations varying position of the grounding line around Antarctica response associated with changes in ice loading that tails about individual applications are given in Appendix and increased melt rates have caused glacier and ice and will monitor the extent and stability of buttressing occurred thousands of years ago. The resulting surface E, section 17.4. sheet margins to thin by up to tens of meters per year ice shelves. deformation from glacial-isostatic adjustment has as ice is lost to the sea. Much of this ice (e.g., floating important implications for our ability to predict relative NISAR will be a reliable source over the life of the mis- ice shelves) acts as a buttress holding back interior ice. Sea ice is another component of the Earth cryosphere sea level rise, which is the difference between the sion for proactive planning for disasters and monitoring Loss of this buttressing introduces instability of these system that is changing rapidly and in ways that can water and land surfaces, a quantity that captures not the development of conditions that could lead to failure. ice sheets, which will likely lead to a more rapid rise in affect climate worldwide. Comprehensive observations just sea level rise but also land elevation change from The stresses induced during a disaster can lead to sea level. NISAR will provide temporally and geograph- of sea-ice extent, motion, concentration and thickness, processes like GIA. Accurate sea level rise predictions failure of compromised structures, and a significant ically comprehensive observations to characterize and derived from multiple satellite observations, including are also tied to our understanding of the rheological part of risk management involves prevention of failure understand ice sheet and glacier dynamics. NISAR data NISAR, will improve our understanding of the interac- 12 NISAR Science Users’ Handbook 13 NISAR Science Users’ Handbook

during disasters, e.g., avoiding overtopping during high understanding will be enabled by measurement of [ Figure 2-5]

water or failure of levees during an earthquake. The surface deformation that NISAR will provide. NISAR will measure changes impact of disasters, sea level change, land subsidence, in i reservoirs. Subsidence and ground movement like landslides or other slope NISAR will be a game-changer for many applications due to ground water with- failures is increasing rapidly with growing population by providing time series of changing conditions to dis- drawal measured with C-band and development in high-risk areas. Monitoring regions entangle long-term impact from seasonal changes and ERS-1 northeast of Los prone to these disasters before they occur can improve episodic, event-induced impact. An inventory of chang- Angeles (Image created by INSAR DEFORMATION MAP risk management by identifying the tell-tale signs of es can be made to assess the rates at which these Gilles Peltzer, JPL/UCLA). processes that lead to these disasters, such as regions systems are changing and associated risks through accumulating elastic strain that will lead to a major deformation measurements at fine spatial and temporal earthquake or the signatures of magma migrating in sampling over the three-year nominal mission lifetime. the subsurface near a volcanic edifice. This improved NISAR will assist in monitoring slow onset disasters like 0

SHADED VIEW OF DEFORMATION

15 cm VERTICAL EXAGGERATION: 20,000 [ Figure 2-4]

NISAR will measure changes in infrastructure. Rate of ground movement determined from L-band UAVSAR along 38°2’10”N one of the levees that 3 km prevents flooding of an island DISTORTED STREET MAP in the Sacramento-San Joaquin Delta (Deverel 2016).

38°2’0”N droughts and large-scale crop failures based on track- geologic, anthropogenic, or climate-related causing changes over a long time period using the frequent, es require adequate sampling to resolve temporal repeated observations. Water resource management variation of displacements with seasonal or finer InSAR Line-of-Sight Velocity mm/yr and critical infrastructure monitoring will be revolution- resolution. Annual cycles result from water withdrawal <–40 ized by access to these data to inform short and long and recharge in aquifer systems, or from climate-in- –35 - –40 term planning (Fig. 2-4). NISAR will be used to monitor duced patterns, such as the freezing and thawing of –30 - –35 levees, dams and aquifers that are under stress from the active layer overlying permafrost in the arctic and –25 - –30 38°1’50”N –20 - –25 groundwater over-utilization; areas where fluid injection sub-arctic regions. Human-induced deformations, such –15 - –20 into and withdrawal from the subsurface is causing as those caused by oil and gas mining or degradation –10 - –15 changes to the surface and potentially affecting water of transportation infrastructure, can occur over many –5 - –10 quality; and differentiation of anthropogenic causes of different time scales, and their identification and 0 - –5 N >0 subsidence from that related to the underlying geology differentiation require resolving processes at much of a region. better than the annual time scale. One example of this 0 0.05 0.1 0.2 Miles need is measurement of surface displacement above 121°45’50”W 121°45’40”W 121°45’30”W Small surface deformation signals, e.g., subsidence, hydrologic aquifers, for which it is necessary to sepa- require long time series to accurately determine slow rate the inelastic subsidence that permanently reduces movement. Measurements sensitive to change from the storage capacity of an aquifer from the annual 14 NISAR Science Users’ Handbook 15 NISAR Science Users’ Handbook

subsidence/inflation due to water use patterns. Since air travel, and impact regional agriculture. Today, the 2.6 OCEAN STUDIES AND COASTAL areas due to its sensitivity to landform structures, mois- proper management of the aquifer system depends on economic and human impacts are growing as popu- PROCESSES ture content and high land-water contrast. NISAR will maintaining the long-term storage of the system, NISAR lation pressure drives development in high-risk areas provide a unique opportunity to study coastal features NISAR will acquire both L-SAR and S-SAR data over must be able to distinguish among these components. and as climate change increases the intensity and and map shoreline changes through high repeat cycle, areas of interest to India, primarily in and around India A similar statement applies to permafrost-induced deg- frequency of severe weather events. synoptic coverage of coastal areas. and in the Arctic and Antarctic. One of the focus areas radation of roads or other structures, where long-term for NISAR data will be to study coastal processes over subsidence trends relating to permafrost decay need to NISAR has a requirement to deliver data for urgent NISAR operating in L-band and S-band will be sensitive India to address the questions: be separated from quasi-seasonal deformation signals response on a best effort basis. Following a disaster to the ocean roughness with wide dynamic range, • How are Indian coastlines changing? DISASTERS WILL BE caused by the freeze-thaw cycle of the overlying active or in anticipation of a forecasted event, NISAR will be enabling study of oceanic internal waves, current fronts layer (Liu et al; 2010; Liu et al; 2012). In both exam- programmed for high priority data acquisition, downlink • What is the shallow bathymetry around India? and upwelling zones. NISAR at L-band will image most DIRECTLY MONITORED ples, the measurements are similar to those needed to and processing to provide low latency information to water at the land-sea coastal interface globally be- • What is the variation of winds in India’s coastal determine secular velocities along tectonic boundaries, support urgent response. There is the unavoidable cause the radar will be turned on prior to reaching land. AND ASSESSED FOR waters? except that the horizontal component of displacement delay between when a disaster occurs and the next im- In many areas, this will enable mapping of surface wind SUPPORT OF EMER- is small and therefore the emphasis needs be on aging opportunity, so NISAR will add to the set of Earth speed, coastal bathymetry, and near-coast surface fea- A large percentage of the world’s population resides accurate determination of the vertical component with observing instruments in space that can respond to tures related to currents and eddies. In coastal regions, GENCY RESPONSE AS A near the coasts and derives their livelihood from the sufficiently high resolution to pinpoint critical areas disasters, shortening overall the time to data delivery. the repeated and regular measurement of surface wind coastal regions, and this is particularly true in India and most heavily impacted in order to allocate resources for Appendix E, section 17.4, provides information on the speed can map wind speed climatology, important for MISSION GOAL OF southeast Asia. Coastal regions, being at the confluence targeted remediation. many types of disasters to which NISAR can contribute the siting of offshore wind power turbines. In addition, NISAR, MOVING BEYOND of land, sea and atmosphere, are subjected to various significant response information. Nearly the full range the high target-to-background contrast at L-band will natural forces and processes resulting in erosion of and The NISAR program has developed a Utilization Plan of disasters can be addressed, from floods to fires to help in identification of oil slicks and ships in the open THE SCIENCE VALUE deposition at the coasts. To understand the nature and outlining how the mission will engage with the end user earthquakes, volcanos, landslides, and even oil spills as well as coastal ocean. magnitude of coastal processes; replace; with, periodic PROVIDED BY BET- community to advance the use of its data for practical and dam collapse. mapping and monitoring of coastal erosional and applications (https://nisar.jpl.nasa.gov/files/nisar/ TER UNDERSTANDING depositional landform features, shoreline changes and NISAR_Utilization_Plan.pdf). There are also a series of Disasters like floods, forest fires and coastal and coastal habitats are required. SAR has been proven to OF THE PROCESSES white papers highlighting individual applications (www. oceanic oil spills can be monitored using radar images be a useful tool for mapping and monitoring of coastal nisar.jpl.nasa.gov/applications). Those topics, among that are provided as mission goal of NISAR. Some other INVOLVED, WHICH CAN others are included in Appendix E, section 17.4. disasters result directly in ground movement, such as ALSO LEAD TO BETTER ground rupture during an earthquake. In these cases, 2.5 DISASTER RESPONSE deformation measurements of the disaster area can FORECASTING AND RISK dramatically improve determination of the scope of Disasters will be directly monitored and assessed for ASSESSMENT. the event, leading to better assessment for targeting support of emergency response as a mission goal of response assets and more efficient recovery. Further- NISAR, moving beyond the science value provided by more, the same data used to monitor ground deforma- better understanding of the processes involved, which tion in disaster-prone regions can be used to detect can also lead to better forecasting and risk assessment. large-scale surface disruption, which can be used to Natural disasters, like floods and earthquakes, cause develop synoptic high-resolution damage proxy maps. thousands of fatalities and cost billions annually. Nearly Such damage proxy maps can aid emergency natural ten percent of the world’s population lives in low lying disaster response throughout the globe regardless of coastal areas subject to flooding. Large earthquakes the level of local infrastructure, so that the response can cause damage hundreds of kilometers from their coordinators can determine from afar where to send epicenter, impacting a wide area. Volcanic eruptions responders within the disaster zone. destroy cities and towns, eject ash clouds that disrupt NISAR Science Users’ Handbook 17 NISAR Science Users’ Handbook

3 MISSION MEASUREMENTS AND REQUIREMENTS

NISAR will utilize the techniques of synthetic aper- 3.1 MEASUREMENTS OF SURFACE ture radar interferometry and polarimetry to measure DEFORMATION AND CHANGE surface deformation and change of the Solid Earth, The technique of Interferometric Synthetic Aperture Cryosphere and Ecosystems. For a brief introduction to Radar (InSAR) uses coherent processing of radar basic radar concepts, including radar imaging (SAR), signals collected over the same scene at two different polarimetry and interferometry, refer to Appendix C. times to derive surface deformation from the change in There are also a wide variety of resources available the relative phase of the two returns (Figure 3-1; Rosen to learn more about the technology and techniques of et al., 2000; Hanssen, 2001). The radar instruments on NISAR. Here are some examples: NISAR will operate as repeat-pass InSAR to measure • https://arset.gsfc.nasa.gov/disasters/webinars/ surface deformation of land and ice-covered surfac- intro-SAR es. An InSAR satellite passing over a location before • https://saredu.dlr.de/ and after an event, such as an earthquake, tectonic deformation, volcanic inflation or ice sheet motion, at • https://www.unavco.org/education/profession- exactly the same point in inertial space (zero baseline), al-development/short-courses/ measures how the ground shifts between passes, • https://www.asf.alaska.edu/asf-tutorials/

PASS #1 PASS #2

Field of view Field of view NISAR

= 24 cm Phase Initial difference surface

Post-earthquake surface

[ Figure 3-1] InSAR measures surface deformation by measuring the difference in the phase of the radar wave between the two passes if a point on the ground moves and the spacecraft is in the same position for both passes (zero baseline). InSAR deformation geometry is demonstrated in these figures at the left and right. On Pass #1, a surface of interest is imaged and the radar satellite measures the phase φ1 (x,y) between the satellite and the ground along the line-of-sight (LOS) direction. Later at Pass #2, the satellite makes another measurement φ2 (x,y) between the satellite and the ground. If the ground moves between passes, the phase difference Δφ (x,y) is proportional to the ground deformation between passes along the LOS direction. zlikovec/Shutterstock 18 NISAR Science Users’ Handbook 19 NISAR Science Users’ Handbook

via a radar interferogram. This is the product of the provide meaningful observations for a broad diversity of science requirements, a secondary S-band SAR, built the implementation of the mission: The NISAR Mission first image with the complex conjugate of the second ecosystems with a timely revisit period. With a resource by ISRO, will provide opportunities in collecting must fulfill these requirements to be successful1 . ISRO (Donnellan et al., 2008). The interferogram measures such as NISAR, and distributed under NASA’s open data dual-frequency observations over key sights in India places additional requirements on the L-band system to the difference in phase of the radar wave between two policy, the NISAR mission will support improved man- and others that are distributed globally. The mission acquire data over science areas of interest to India that passes, which is sensitive to ground motion directed agement of resources and understanding of ecosystem itself includes a large diameter (12 m) deployable re- are above and beyond the NASA requirements, includ- along the radar line of sight. An InSAR image of the processes. flector and a dual frequency antenna feed to implement ing coastal bathymetry and ocean winds, geology over point-by-point phase difference of the wave on the the SweepSAR wide-swath mapping system, which is India, and coastal shoreline studies (Table 3-1). Unlike surface is used to create a map of the movement of NISAR polarization configurations will enable accurate the enabling technology to allow for global access, fast the NASA requirements, the quantitative values asso- the surface over time. In this way, ground deformation estimation of vegetation above ground biomass up to revisit, frequent temporal sampling, and full resolution. ciated with these measurements are characterized as along the line-of-sight (LOS) direction on the scale of 100t/ha. In polarimetric backscatter measurements, The polarimetric capability of the NISAR system pro- goals; it is the collection of the data toward these goals a fraction of the radar wavelength can be resolved as forest components (stems, branches, and leaves) are vides dual-polarized (dual-pol) global observations for that drives the ISRO requirements. There are no explicit long as the phase coherence between the signals is scatterers within the footprint of the radar beam that every cycle and the potential for quad-pol observations requirements on science measurements at S-band, just maintained (Zebker and Goldstein, 1986; Gabriel et interact with the incoming waves. The size (volume) in India and the U.S. The dual-pol system is based on a statement identifying the impact such measurements al., 1989). The radar instrument can take observations and the dielectric constant (moisture or wood density) transmitting a horizontally or vertically polarized wave- can make, leaving open a range of options for exploring through cloud cover, without sunlight, and can measure and orientation and morphology of the scatterers deforms and receiving signals in both polarizations. Over its potentials. sub-centimeter changes. termine the magnitude and polarization of the reflected land surfaces, the transmit polarization will principally waves. As a result, the backscatter radar energy at be horizontally polarized, and receive will be over both Table 3-1 shows an overview of the L1 baseline 3.2 LANDCOVER AND FOREST linear polarizations is a function of the forest volume vertical and horizontal polarizations, resulting in polar- requirements for the mission. Baseline requirements CHARACTERIZATION WITH and biomass. The shape of this function depends on ization combinations known as HH and HV to describe represent the full complement of science requested by L-BAND SAR the wavelength, polarization, forest type, and moisture the configuration. NASA of the NISAR Mission. The NISAR project teams

NISAR will serve to estimate above ground biomass, conditions. The relationship varies with vegetation type at JPL and ISRO use the L1 requirements to develop a identify croplands and inundated extent, and detect and environmental conditions (e.g., soil moisture and For a limited set of targets, the NISAR mission will detailed set of L2 project requirements, which govern forest disturbances. The overall ecosystem science roughness), but with multiple polarizations and repeat- make fully polarimetric measurements (i.e., fully pola- the implementation in such a way that by meeting the community will greatly benefit from the mission, ed measurements, the biomass can be determined with rimetric, or quad-pol) by alternating between transmit- L2/L1 requirements, the requirements will be met. The which is characterized by high frequency revisit time high accuracy. ting H-, and V-polarized waveforms and receive both L2 science requirements are described in Appendix D. (12 days) and L-band capabilities. By their fundamental H and V (giving HH, HV, VH, VV imagery). Polarization NASA and ISRO have jointly coordinated all require- nature, ecosystems are driven by the hydrologic and Changes in forest structure observed by the system, combinations, such as dual- and quad-pol, allow for ments at Level 1 and Level 2. Lower level requirements seasonal cycles, and hence undergo dynamic changes whether due to natural cycles, or human or natural a fuller characterization of ground-target’s responses are generated by the NASA and ISRO project teams throughout the year. When combined with the need disturbances, will provide key measurements to assess to the SAR. Variations in the polarimetric responses of independently. The teams coordinate hardware and for monitoring changes in these systems, through fire, the role and feed-back of forests in the global carbon targets to different combinations of polarization can activities through interface documents. drought, encroachment, deforestation or otherwise, it cycle. Every 12 days, the NISAR mission will resolve be related to the physical characteristics of the target is important to detect and demarcate these regions in severity and time of disturbance with a 1 hectare and reflecting energy back to the radar, and hence can be The requirement on 2-D Solid Earth and ice sheet order to provide quantitative measures of inventory and 12-day spatial temporal resolution. NISAR’s rapid used for classifying target type and performing quanti- displacement covers a range of lower level require- change that affect the many services that Ecosystems revisit time will provide timely identification of cropland tative estimates of the target state. ments on the ability to measure deformation of land. offer to populations worldwide. NISAR’s dynamic obser- status, estimation of soil moisture, and the monitoring The science described above for deformation relies on 3.3 REQUIREMENTS AND SCIENCE vations and compilation of a new historical record will of flooding and inundation extent. time-series of data acquired regularly and with fast TRACEABILITY provide an important resource throughout the mission’s sampling. This range of science can be specified as In addition to the basic resource of measuring radar individual requirements on velocities or strain rates, but lifetime and beyond. NASA and ISRO have developed a joint set of require- reflectivity, for ecosystems, the NISAR mission has a that would lead to a large number of L1 requirements. ments for NISAR. These agency level requirements number of other capabilities that will be useful for the This requirement is written with the foreknowledge Among the important features of the mission charac- are known as “Level 1” (L1) requirements and control teristics are its wide swath, high resolution, 12-day discipline. Among these features is the capability of of flow-down to a repeat-pass interferometry and repeat orbit cycle and dual-frequency (L- and S-band) performing repeat-pass interferometry and in the collection of polarimetric data. While the core capability of 1There are also a set of threshold requirements, which define the minimum complement of science considered to be worth the investment. capability. These features will allow the mission to Baseline requirements can be relaxed toward thresholds when implementation issues lead to loss of performance. At this point in Phase C, the payload is the L-band SAR used to meet all of NASA NISAR continues to work toward the baseline requirements. 20 NISAR Science Users’ Handbook 21 NISAR Science Users’ Handbook

TABLE 3-1. LEVEL 1 BASELINE REQUIREMENTS they move. This is a proven technique, but there is no cations are binary (e.g. agriculture/non-agriculture) and

Attribute 2-D Solid Earth 2-D Ice Sheet & Sea Ice Biomass Disturbance Cropland, reliable source of data to perform it with. are distinct from the biomass disturbance and recovery Displacement Glacier Displ. Velocity Inundation Area classifications in the previous requirement. The requirement on biomass and disturbance states Duration 3 years 3 years 3 years 3 years 3 years 3 years that the mission measure global biomass and its The urgent response requirement for NISAR is written Resolution 100 m 100 m 5 km 100 m 100 m 100 m disturbance and recovery, but only specifies an accu- to ensure that the mission has some capability for grid (1 ha) (1 ha) (1 ha) racy for the low-density woody biomass. The global disaster response built into it, but one that does not Accuracy 3.5 (1+L1/2) mm 100 mm or 100 m/day 20 Mg/ha for 80% for areas 80% requirement on biomass and disturbance/recovery drive the costs for development or operations. NISAR is or better, 0.1 km better over 70% or better areas of losing > 50% classification allows a specification of the details of disturbance primarily a science mission, but radar imaging systems < L <50 km, over of fundamental over 70% biomass canopy cover accuracy 70% of areas sampling of areas < 100 Mg/ha and recovery at L2 but requires global observations at are among the most useful space remote sensing interest intervals L1. Thus, in regions of high-density woody biomass, assets for understanding disasters because they can

Sampling 12 days or bet- 12 days or 3 days or Annual Annual 12 days where there are no explicit accuracy requirements, deliver reliable imagery day or night, rain or shine, that ter, over 80% better better or better measurements must be made to ensure the capture of are not obscured by smoke or fire. of all intervals, disturbance and recovery. < 60-day gap over mission ISRO has identified a number of science goals that do The requirement on cropland and inundation area is an not fall in the joint Baseline requirements as summa- Coverage Land areas Global ice Arctic and Global areas Global areas Global areas predicted to sheets and Antarctic of woody of woody of crops overall classification requirement of ecosystems of par- rized in Table 3-2 and articulated above. The mea- move faster glaciers sea ice biomass cover biomass and wetlands ticular interest to the science community. The classifi- surement metrics in the table are specified as goals than 1 mm/yr, cover volcanoes, reservoirs, TABLE 3-2. ISRO L-BAND BASELINE GOALS glacial rebound, landslides Attribute Coastal Wind Bathymetry Coastal Shoreline Geological Sea Ice Velocity Position Features Characteristics Urgent 24-hour tasking 24/5 24/5 24/5 24/5 24/5 Response 5-hour data Best-effort Best-effort Best-effort Best-effort Best-effort Duration 3 years 3 years 3 years 3 years 3 years Latency delivery basis (BEB) basis (BEB) basis (BEB) basis (BEB) basis (BEB) Best-effort basis Resolution 1 km grid 100 m grid 10 m 10 m 10 m

Accuracy 2 m/s over at 20 cm over at 5 m over at least N/A N/A least 80% of least 80% of 80% of areas of specifies the sampling and accuracy understood to be L1 requirement is specified with different resolution areas of interest areas of interest interest achievable. The accuracy is controlled largely by the and accuracy requirements. As with land deformation, Sampling 6 days or better Every 6 months 12 days or 90 days or better, 12 days noise introduced by the atmosphere, which the project the intent of this requirement is to design a system that better with at least two or better cannot control. The intent of this requirement is to delivers reliably regularly sampled interferometrically viewing geometries design a system that reliably delivers regularly sampled viable data on ascending and descending orbit passes Coverage Oceans within India’s coast India’s coastal Selected regions Seas surrounding interferometrically viable data on ascending and de- as needed to achieve the science at a particular target. 200 km of to an offshore shoreline including paleo- India’s Arctic and scending orbit passes as needed to achieve the science As with solid Earth requirements, the L1 capability as India’s coast distance where channels in Rajas- Antarctic polar the depth of the than, lineaments stations at a particular target. As such, the L2 requirements defined allows for a flow-down to a set of L2 require- ocean is 20 m or and may improve one aspect of the L1 requirements at the ments that meet the ice-sheet science objectives. less structural studies in Himalayas and in expense of another (e.g., resolution vs. accuracy). Deccan plateau The requirement on sea ice velocity is also a deforma- The requirement on 2-D ice sheet and glacier dis- tion requirement but is called out separately because it placement covers a range of lower level requirements relies on different kinds of measurements with different on the ability to measure deformation of ice. It is a sampling and accuracy requirements. In this case, similar geodetic measurement as for the solid Earth the intent of the requirement is to observe the poles requirement above, but the environment has a different regularly and track features in the radar imagery as influence on the ice-covered regions than land, so the 22 NISAR Science Users’ Handbook 23 NISAR Science Users’ Handbook

TABLE 3-3. SCIENCE TRACEABILITY MATRIX because it is difficult to quantify how well they can All disciplines also require global reach so that entire

Science Measurement be met. NISAR will collect L-band SAR data needed to systems can be characterized – e.g.; all of Amazonia, Science Requirements Instrument Projected Mission support these goals. all of Greenland and Antarctica, or all of the “ring of Requirements Performance Requirements Objectives Physical Parameters Observables (Spatial and (Top Level) fire.” For global access and fast revisit, a wide-swath Temporal) 3.4 SCIENCE TRACEABILITY TO or steerable mapping system is required.

Determine the Annual biomass Radar reflec- Frequency L-band 1215-1300 MHz Seasonal global coverage MISSION REQUIREMENTS contribution of at 100 m reso- tivity radio- per science target mask All disciplines also require many samples in time (i.e., Earth’s biomas lution and 20% metrically The Science Traceability Matrix (STM) connects the to the global accuracy for bio- accurate to Polarization Dual-pol Quad-pol 6 samples per season every cycle) to reduce noise sources associated with carbon budget mass less than 0.5 dB science requirements to instrument and mission 100 Mg/ha Resolution 5 m range 3-m range environmental variability – e.g., soil moisture changes Ascending/descending requirements in a succinct table (Table 3-3). Due to 10 m azimuth 8-m azimuth – so a solely steerable mapping system generally will the breadth of the science goals for NISAR, and the not suffice. A wide-swath mapping system such as Geolocation 1 m 0.5 m Maximum incidence angle interplay between instrument and mission operations accuracy diversity ScanSAR (Moore et al., 1981) or SweepSAR (Freeman scenarios to meet the science goals as just described, et al., 2009) is required for global access, fast revisit, Relative 0.1dB 0.05 dB 3-year mission it is difficult to capture traceability in a way that the radiometric and frequent temporal sampling. accuracy sensitivities of science requirements to mission capa-

Annual distur- Absolute 0.5 dB 0.2 dB bilities and vice versa, is transparent. bance/recovery radiometric All disciplines require looks, spatial averaging of intrin- at 100 m reso- accuracy sic resolution SAR or InSAR data, to reduce speckle and lution All disciplines – Solid Earth, Cryosphere, and Ecosys- other local noise effects. To meet the demanding accu- Range -15 dB -18 dB tems – require long wavelengths. For Ecosystems, long ambiguities racy requirements described below, the system must wavelengths are needed to maximize the sensitivity to have fine resolution in both image dimensions to create Azimuth -15 dB -20 dB dual-pol biomass variability. For solid Earth and cryospheric de- ambiguities sufficient looks to average. A ScanSAR system that has formation, long wavelengths are preferred to minimize ISLR -15 dB -20 dB dual-pol reduced resolution in the along-track dimension will not the effects of temporal change of the surface; it takes suffice, particularly given the limited allowable range NESO -25 dB -25 dB a larger change of the surface to create significant resolution at L-band. Therefore, a SweepSAR wide- Access Global Global decorrelation when the wavelength is long. swath mapping system, which allows wide swath while Determine the Surface displace- Radar group Frequency L-band 1215-1300 MHz Every cycle sampling maintaining resolution in the along-track dimension, is causes and ments to 20 mm and phase All disciplines benefit from polarimetry – while eco- consequences over 12 days delay differ- Polarization Single-pol Quad-pol Ascending/descending required for global access, fast revisit, frequent tempo- of changes of systems demand polarimetry to meet their objectives, ences on 12 ral sampling and full resolution. SweepSAR is explained Earth’s surface day centers Resolution 4-m range 3 m range Global coverage per science deformation science can take advantage of polarimetry and interior 10-m azimuth 8 m azimuth in greater detail in section 4.7. target mask to characterize environmental effects, like soil moisture Repeat 12 days or less 12 days Non-tidal cycle repeat variations, and potentially optimize correlation in vege- interval (d) At this highest level, the system described in the next tated regions. Swath width 2880/d km 252 km Reconfigurable section is necessary to meet the objectives of all disciplines. Determine how Incidence 33-46 degrees 32-47 degrees Left/right viewing for Anta- All disciplines are interested in mapping dynamic climate and ice angle range for d=12 ractic/Arctic coverage masses inter- processes – ones that can change from week to week, The general observational characteristics – wide swath, relate and raise Pointing 273 arcsec 273 arcsec (TBC) Orbit repeatability to <500 m or instantaneously, such as when a storm front hits, a sea level control fast repeat, fine resolution and multiple polarizations glacier surges or an earthquake strikes. In that sense, – represent the most basic flow down from science Surface displace- Radar group Repeat 12 days or less 12 days Every opportunity sampling all disciplines are interested in regular sampling with ments to 100 m delay dif- interval (d) requirements to mission and instrument requirements. over 3 days ferences on the fastest revisit time achievable given the constraints Table 3-4 summarizes the key and driving require- 3-day centers Swath width 2880/d km 252 km Complete sea-ice coverage of the project. (240 km at d=12) ments on the mission system to satisfy the science

Respond to Hazard- Radar imagery Any of above Re-target hazard area to requirements. harzards dependent previously acquired mode imaging within 24 hours

Deliver data after acquistion within 5 hours 24 NISAR Science Users’ Handbook

TABLE 3-4. OVERVIEW OF KEY AND DRIVING REQUIREMENTS

Key & Driving Requirement Why Is It Challenging? Why Is It Needed?

Interferometry capability Interferometry requires that the space- Interferometry is needed to obtain geodetic between any two repeated craft be (a) controlled in its orbit to measurements at the required spatial sam- acquisitions better than 350 m positioning through- pling out the mission and (b) controlled in its pointing to a small fraction of a degree

Fast sampling (6 days) and Implies that the accessible field-of- Fast sampling is required to observe Earth’s interferometric revisit regard of the instrument covers the most dynamic and poorly understood pro- (12 days) over all Earth’s > 240-km ground track spacing cesses without aliasing. The repeat period land surfaces chosen for NISAR is a balance between covering interesting and practical regions (from an observation planning point of view)

Frequent sampling over Given the multiplicity of disciplines, In addition to fast sampling, many samples most of Earth’s surface the only way to acquire sufficient are needed throughout the mission to defeat data to meet coverage and accuracy the noise sources that limit accuracy. To first requirements is for the radar to have a order, more data is needed to average errors field of view equal to its field of regard down to an acceptable level (> 240 km). This requires specialized hardware to create an extra-wide swath at full resolution (Sweep SAR)

Polarimetry Polarimetry requires additional hard- Classification of surfaces and estimation ware, mass and power resources, add- of biomass cannot be done at the required ing to complexity and cost accuracies without a polarimetric capability

Signal-to-noise ratio The mission must be designed with Many of Earth’s surface types are poor sufficient power and antenna gain to reflectors. When reflection is low, the noise observe dim targets adequately dominates the measurement and leads to less accurate results

Radiometric predictability over time Knowledge of the signal level enables Knowing the signal level is important to the quantitative associations to be made absolute radar cross-section measurements between radar signals and geophys- used to derive biomass and classifications ical parameters. This drives design of structure stiffness and electrical tolerances in the radar NISAR Science Users’ Handbook 27 NISAR Science Users’ Handbook

4 MISSION CHARACTERISTICS

This section describes NISAR mission attributes that number and frequency of acquisitions needed in any are important to the use and interpretation of the given time interval, radar modes to be used, the season data. These attributes include the observing strategy (if relevant), and whether to observe on the ascending – including areas of acquisition, mode of operation, or descending pass or both. frequency of coverage, the orbit, and the radar observational capabilities. The mission operations design and All NASA requirements can be met exclusively with THE NISAR constraints, which can influence science acquisition the NASA-provided L-band radar system. In addition to OBSERVATORY planning and execution, are also described. This design the NASA science requirements, ISRO scientists have includes NASA and ISRO’s plans to respond to urgent specified targets of interest in India and its surrounding WILL LAUNCH events. coastal waters. These areas have similar attributes as ON AN ISRO those defined by the NASA SDT for global targets. The 4.1 OBSERVING STRATEGY ISRO requirements combine L-band and S-band obser- GEOSYNCHRONOUS vations. Operating the L- and S-band radars simul- The NISAR mission aims to achieve global coverage of taneously will provide unique data and also minimize SATELLITE LAUNCH land where biomass exists (which is nearly everywhere mode conflicts over India. However, the programmatic on land, except the polar areas), full coverage of land VEHICLE MARK II guideline is to not require simultaneous operation, but and sea ice at both poles and in mountains, frequent to make it an implementation goal. FROM SATISH coverage of land areas that are deforming rapidly, and regular infrequent global coverage elsewhere to be pre- DHAWAN SPACE 4.2 REFERENCE SCIENCE ORBIT pared to respond to events that are unusual, such as CENTER IN mid-plate earthquakes. Global coverage requirements The NISAR observatory will launch on an ISRO Geosyn- for NISAR science are specified in the L1 requirements chronous Satellite Launch Vehicle (GSLV) Mark II from SRIHARIKOTA, INDIA. (Figure 4-1). Each of the L2 science requirements (list- Satish Dhawan Space Center (SDSC) in Sriharikota, ed in Appendix D) specifies a measurement objective, India. Launch services will be provided by ISRO, which an accuracy of that measurement, and the area of will manage launch vehicle development and provide interest or target area over which the measurement all necessary technical documentation. The current tar- must be made. get launch date is December 2021. The baseline orbit was selected to satisfy scientific and programmatic Science targets are proposed by each of the three requirements, and has the following characteristics: NISAR scientific discipline teams (Solid Earth, Eco- 747 km altitude, 98.4 degrees inclination, sun-synchro- systems and Cryosphere), in the form of geographical nous, dawn-dusk (6 PM ascending node), and a total polygons and nominal radar modes (see Appendix G for repeat cycle of 173 orbits in 12 days. NISAR’s 747-km NISAR target maps by each discipline). With these tar- altitude orbit, consisting of 173 orbits/cycle, will allow gets and the L2 measurement accuracy requirements for global coverage every 12 days, as shown in Figure stated in Appendix D in mind, an observing strategy 4-2 and Table 4-1. can be devised, which takes into account the desired 28 NISAR Science Users’ Handbook 29 NISAR Science Users’ Handbook

[ Figure 4-1] [ Figure 4-2]

Global coverage for ascend- A snapshot of the Reference ing (top) and descending Science Orbit orbital elements (bottom) passes showing at the first ascending equator instrument modes and crossing are given in the target areas. The coverage following table and are specified addresses the three science in an Earth-Centered True disciplines of solid Earth Equator and Equinox of Epoch deformation, ecosystems, coordinate frame purple mesh. and cryosphere. Many appli- During every 12-day repeat cations objectives will also cycle, NISAR will execute be satisfied through these 173 orbits, which will provide observations. global coverage of the Earth.

TABLE 4-1. ORBITAL ELEMENTS AT THE FIRST ASCENDING EQUATOR CROSSING FOR THE NISAR REFERENCE SCIENCE ORBIT. NISAR WILL ORBIT THE EARTH IN A NEAR-POLAR (98.4 DEGREES INCLINATION), SUN-SYNCHRONOUS ORBIT

ORBITAL ELEMENT VALUE (osculating ) VALUE (mean) INSTRUMENT MODES Semi-Major Axis (km) 7134.54476 7125.48662 Background Land (L-SAR only) Antarctic Region (Joint mode)

Eccentricity North America SNWG Target (L-SAR only) ISRO Sea Ice (Joint mode) 0.0012677 0.0011650

ISRO Agriculture + Instrument Calibration Sites Land Ice (L-SAR only) Inclination (deg) 98.40508 98.40508 (Joint mode)

Sea Ice (L-SAR only) Systematic Coverage (Joint mode) Longitude of Node (deg) -19.61438 -19.61601 Systematic Coverage and Deformation Low Data Rate Study Mode (L-SAR only) (Joint mode) Argument of Periapsis (deg) 68.40031 89.99764 Urban Areas (L-SAR only) Agriculture/Sea Ice (S-SAR only) True Anomaly (deg) -68.40237 -89.99818 High Res Deformation-Landslide/Urban (Joint mode) 30 NISAR Science Users’ Handbook 31 NISAR Science Users’ Handbook

During science operations, NISAR will fly within a plans on executing maneuvers over the long ocean [ Figure 4-4] Actual Trajectory Through Diamond

diamond-shaped orbital corridor defined for each of the passes (Atlantic and Pacific) as much as possible not to Actual versus repeat cycle’s 173 orbits and tied to the rotating Earth impact science data collection. reference trajectory (Figure 4-3). This corridor is defined to enable accurate for NISAR as correlation of science observations from pass-to-pass The NISAR spacecraft will accommodate two fully maintained within and cycle-to-cycle, supporting assessment of changes capable synthetic aperture radar instruments (24 cm the diamond. in the science targets. The dimensions of the diamond wavelength L-SAR and 10 cm wavelength S-SAR), each were calculated as an upper bound on acceptable error designed as array-fed reflectors to work as SweepSAR produced by a non-zero baseline between passes/cy- scan-on-receive wide swath mapping systems. The cles between three primary factors (Rosen et al., 2000) spacecraft will launch on an ISRO GSLV-II launch

of phase unwrapping error, geometric decorrelation and vehicle into a polar sun-synchronous dawn dusk orbit. RGT topographic leakage, but ultimately dominated by the The mapping scenario calls for frequent sampling over former (phase unwrapping error, i.e., high fringe rate in broad areas to create time series and allow for noise regions of large topographic relief). reduction through stacking methods. Thus, a high-rate Reference Trajectory [ Figure 4-3] instrument and data downlink system are required. The The center of the Diamond is defined by the 173- average capacity of the envisioned data downlink is of During science operations, orbit reference trajectory (referred to as the reference the order of 26 Tbits per day, supporting the instru- NISAR will fly within a science orbit), which is fixed to the Earth’s surface and ments which can produce at L-band from 72 Mbps in TABLE 4-2. OVERVIEW OF NISAR MISSION CHARACTERISTICS diamond-shaped orbital corridor defined for each of the is exactly repeated every 12 days. The diamond can its lowest bandwidth mode to over 1500 Mbps in the ELEMENT DESCRIPTION repeat cycle’s 173 orbits and be thought of as a fixed altitude, longitude and latitude most demanding high-bandwidth, multi-polarization Proposed Launch Date Late 2021 tied to the rotating Earth. profile that spans the entire repeat cycle; a conceptual mode. Tables 4-2 and 4-3 summarize the overall mis- representation of this corridor is shown in Figure 4-4. sion characteristics. Orbit 12-day exact repeat, sun-synchronous, dawn-dusk, polar, 747 km altitude To maintain the diamond, the JPL Navigation team Mission Duration 3 years nominal, with extended mission fuel reserve

ACTUAL ORBIT Science Data Downlink • 30-45 minutes of data downlink per orbit at 3.5 Gbps data Perturbed by all forces; Drag make up Approach rate through polar ground stations (DMU) maneuvers and inclination adjust • 1 Gbps direct downlink to India over Indian ground stations maneuvers (IAM) are used to control actual orbit relative to reference orbit. ±325m Observation • L-band multi-mode global radar imaging Vertical Approach • S-band multi-mode targeted radar imaging • Dual-frequency capable west east • ~240 km swath for all modes Actual SMA > • Full pol, multiple bandwidths up to 80 MHz Reference SMA • Near-zero Doppler pointing, fixed boresight REFERENCE ORBIT • Left-looking for the entire mission, relying on the Includes full gravity field but no drag, international SAR constellation to fill in coverage DMU ±250m luni-solar, or solar radiation pressure around the Arctic pole Horizontal effects. Defines latitude, longitude, and altitude profile that exactly repeats for Mapping Approach Under study – current approach defines a reference mission every 12-day/173-orbit repeat cycle. with fixed modes over broad target areas. Actual SMA < 80m Reference SMA (margin)

Path represents Ascending Equator crossings (one per orbit).

Contolling to the above diamond within the dashed lines allows sufficient margin to accommodate variations in position due to REFERENCE off-nominal inclination and eccentricity. This is not related to GROUND density variations or maneuver execution errors. TRACK 32 NISAR Science Users’ Handbook 33 NISAR Science Users’ Handbook

TABLE 4-3. MAJOR MISSION AND INSTRUMENT CHARACTERISTICS FOR NISAR 4.3 MISSION PHASES AND TIMELINE craft checkout and instrument checkout. Philosophically,

PARAMETERS S-BAND L-BAND the sub-phases are designed as a step-by-step buildup Figure 4-4 provides a high-level overview of the NISAR in capability to full observatory operations, beginning mission timeline, and Table 4-5 provided more details Orbit 747 km with 98° inclination with the physical deployment of all deployable parts on the different phases of the mission. Repeat Cycle 12-days (notably the boom and radar antenna, but not includ-

ing the solar arrays which are deployed during launch Time of Nodal Crossing 6 AM/6 PM LAUNCH PHASE phase), checking out the engineering systems, turning Frequency 3.2 GHz ± 37.5 MHz 1.257 GHz ± 40 MHz The NISAR Observatory will be launched from ISRO’s on the radars and testing them independently and then

conducting joint tests with both radars operating. Available Polarmetric Modes Single Pol (SP): HH or VV SP: HH or VV Satish Dhawan Space Centre (SDSC), also referred to Dual Pol (DP): HH/HV or VV/VH DP: HH/HV or VV/VH as Sriharikota High Altitude Range (SHAR), located in Compact Pol (CP): RH/RV CP: RH/RV Sriharikota on the southeast coast of the Indian pen- SCIENCE OPERATIONS PHASE Quasi-Quad Pol (QQP): HH/HV Quad Pol (QP): HH/HV/VH/VV insula, on the GSLV Mark-II expendable launch vehicle and VH/VV The science operations phase begins at the end of contributed by ISRO. The target launch readiness date Available Range Bandwidths 10 MHZ, 25 MHz, 37.5 MHz, 5 MHZ, 20 MHz, 40 MHz, 80 MHz commissioning and extends for three years and contains is December 2021. The launch sequence encompasses 75 MHz (Additional 5 MHz iono band for all data collection required to achieve the L1 science 20 & 40 MHz modes at other end the time interval that takes the observatory from the objectives. During this phase, the science orbit will be of pass-band) ground, encapsulated in the launch vehicle fairing, maintained via regular maneuvers, scheduled to avoid or Swath Width to after separation, and ends with the completion of >240 Km (except for QQP Mode) >240 Km (except for 80MHz BW) minimize conflicts with science observations. Extensive solar array deployment and the observatory in an Spatial Resolution 7m (Az); 3m–24m (Slant-Ra) 7m (Az); 3m–48m (Slant-Ra) calibration and validation (CalVal) activities will take Earth-pointed attitude and in two-way communication place throughout the first 5 months, with yearly updates Incidence Angle Range 37–47 deg 37–47 deg with the ground. The launch sequence is a critical of 1-month duration. event. Noise Equivalent s° -25 dB (baseline) -25 dB (for required full-swath 20 dB (threshold) modes) - The observation plan for both L- and S-band instru- COMMISSIONING PHASE Ambiguities < -20 dB for all modes < -23 dB swath average in SP ments, along with engineering activities (e.g., ma- except QQP or DP modes The first 90 days after launch will be dedicated to neuvers, parameter updates, etc.), will be generated < -17 dB swath average in QP modes commissioning, or in-orbit checkout (IOC), the objec- pre-launch via frequent coordination between JPL and tive of which is to prepare the observatory for science ISRO. This plan is called the reference mission; the Data and Product Access Free and Open operations. Commissioning is divided into sub-phases science observations alone within that reference mission of initial checkout (ISRO engineering systems and JPL are called the reference observation plan (ROP). The NASA contributions include the L-band SAR instrument, NASA and ISRO will share science and engineering engineering payload checkout), deployments, space- schedule of science observations will be driven by a including the 12-m diameter deployable mesh reflector data captured at their respective downlink stations, and and 9-m deployable boom and the entire octagonal each organization will maintain their own ground pro- instrument structure. In addition, NASA is providing cessing and product distribution system. The science [ Figure 4-5] MONTHS YEARS a high capacity solid-state recorder (approximately teams and algorithm development teams at NASA and 0 1 2 3 4 5 6 1 2 3 9 Tbits at end of life), GPS, 3.5 Gbps Ka-band telecom ISRO will work jointly to create a common set of prod- Mission timeline and phases for NISAR. The mission timeline system, and an engineering payload to coordinate uct types and software. The project will deliver NISAR LAUNCH for NISAR will be divided into command and data handling with the ISRO spacecraft data to NASA and ISRO for archive and distribution. DEPLOYMENTS launch, a 90-day commissioning DE-COMMISSIONING control systems. ISRO is providing the spacecraft and NASA and ISRO have agreed to a free and open data or in-orbit checkout period, COMMISSIONING launch vehicle, as well as the S-band SAR electronics policy for these data. followed by 3 years of nominal SCIENCE OPERATIONS (3 YEARS) to be mounted on the instrument structure. The coor- science operations, and 90 days 5 MONTHS dination of technical interfaces among subsystems is a of decommissioning CAL/VAL major focus area in the partnership. 34 NISAR Science Users’ Handbook 35 NISAR Science Users’ Handbook

variety of inputs, including L- and S-band target maps, Routine operations of NISAR are dominated by Orbit SVALBARD (Ka) ALASKA (Ka) radar mode tables, and spacecraft and ground-sta- maintenance maneuvers, science observations and TROMSØ NOC tion constraints and capabilities. This schedule will data-downlink. Additional activities will include contin- be determined by JPL’s mission planning team, and uous pointing of the solar array to maximize power and the project will endeavor to fly the reference mission, continuous zero-doppler steering of the spacecraft. SCI PROC & DAAC* which includes these science observations exactly as JPL GSFC planned pre-launch (accommodating for small timing changes based on the actual orbit). Periodic updates are possible post-launch which will lead to a new reference mission. BIAK (S-LEOP)

TABLE 4-4. NISAR MISSION PHASES Ground Station

MISSION PHASE START DATE DURATION BOUNDARY END STATE Control Center LUCKNOW (S) MAURITIUS (S) Launch (L) December 2021* 1 day + ~40 minutes Spacecraft in target orbit, power positive, NASA Provided (L - 24 hours) in two-way communication ISRO Provided SAC

Commissioning L + ~40 minutes 90 days All systems ready to begin science data * Cloud facility location, SHADNAGAR (Ka) collection and co-location of DAAC, PUNTA ARENAS (Ka) NRSC AWS US West 2. BANGALORE (S) SDSC (launch) Science Operations L + 90 days Mission objectives are complete ANTARCTICA (S, Ka) 3 years SRIHARIKOTA ISTRAC (S-LEAP) Decommissioning L + 3.25 years 90 days Spacecraft in disposal orbit and passivated

DECOMMISSIONING PHASE GROUND DATA SYSTEM [ Figure 4-6] facility includes computer hardware dedicated to oper- products may be migrated to the production system to

Decommissioning phase begins after the three-years The GDS includes the tracking stations, data cap- NISAR ground stations ational data production. In addition, the SDS facility is generate larger areas of Level 3 products. of the primary science phase and after any extended ture services, the communications network and end (including the NASA Near- planned as a cloud-based hybrid SDS, with all elements To facilitate the software development process, the SDS operations phases (e.g., NASA Senior Review) have party services (Figure 4-6). The stations, services and Earth Network stations cloud-enabled. This allows for some processing to be will establish a mechanism for developmental instanc- concluded. This phase extends for 90 days. NASA communications are NASA multi-mission capabilities in Alaska, Svalbard and done at JPL and some to be distributed to the external es of the SDS to be made available to the algorithm deorbit and debris requirements are not applicable for managed by the Goddard Space Flight Center (GSFC). Punta Arenas; ISRO cloud. The science and algorithm development teams development and science teams. These developmental NISAR, however the project must comply with ISRO’s The GDS will send the raw science data to the Science stations in Antarctica, will have access to cloud instances separate from the instances will be logically separate from the production guidelines to safely end the mission. ISRO adheres to Data System (SDS), which converts the downlinked Shadnagar, Bangalore, production instances to enhance algorithmic accuracy system but will allow development and testing of the the IADC Space Debris Mitigation Guidelines, IADC-02- raw data into Level 0a and Level 0b data that are the Lucknow, Mauritius, Biak), and performance. software that will be used to automatically generate the control center and launch 01, Revision 1, September 2007. starting point for the science data processing. science data products once NISAR is in orbit. location (Sriharikota The SDS is controlled through a cloud-based produc- (SDSC), India). 4.4 GROUND SEGMENT OVERVIEW SCIENCE DATA SYSTEM tion management system at the Jet Propulsion Labo- NASA NEAR-EARTH NETWORK (NEN) ratory (JPL) in Pasadena, California. JPL is responsible The NISAR ground segment consists of the Ground The SDS converts the Level 0b data into L1/L2 science NISAR will downlink both to ISRO ground stations (see for implementation of software to generate Level 1 Data System (GDS), Science Data System (SDS), and data products that the NISAR mission provides to the below) and to NASA Near Earth Network (NEN) stations. radar instrument data products and Level 2 products. Mission Planning & Operations System. The GDS and science community for research and applications. The For the NASA stations, Ka-band antennas will be used The science team is responsible for generating Level 3 SDS manage the end-to-end flow of data from raw SDS facility is designed to process data efficiently at one or more complexes. The specific antenna com- geophysical data products for calibration and valida- data to fully processed science data products. and distribute data products in a timely manner to the plexes currently identified are Alaska, United States; tion purposes. As funds permit, software for Level 3 community as required to meet mission objectives. The Svalbard, Norway; Punta Arenas, Chile; and Troll, Antarctica.

2. For data product levels see https://science.nasa.gov/earth-science/earth-science-data/data-processing-levels-for-eosdis-data-products 36 NISAR Science Users’ Handbook 37 NISAR Science Users’ Handbook

JPL MISSION OPERATIONS CENTER (MOC) Antarctica. The station in Shadnagar is also referred to

as the Shadnagar Acquisition Network, or SAN. ~8 Tbits/day* Ka-band: 2.88 Gbps** JPL will perform mission operations from multiple ~8 passes/day Science Ka-band: 4 Gbps** buildings at JPL in Pasadena, California, all of which Science & Engineering SATISH DHAWAN SPACE CENTRE (SDSC), are considered to make up the Mission Operations ~26 Tbits/day* SHRIHARIKOTA RANGE (SHAR) ~20 passes/day Center (MOC). The existing multi-mission Earth Orbiting NISAR Missions Operation Center (EOMOC) will provide opera- SDSC SHAR, with two launch pads, is the main launch ISRO Ka-band S-band: Uplink 4 Kbps Station (NRSC) Downlink 4 & 32 Kbps tions teams with consoles, workstations, and voice and center of ISRO, located at 100 km north of Chennai.

video displays. Navigation and GPS operations will be SDSC SHAR has the necessary infrastructure for ISRO Element NASA Ka-band JPL/NASA Element conducted from other JPL locations. launching satellites into low Earth orbit, polar orbit and ISRO S-band Stations Stations (ISTRAC) * Information Rate (Goddard NEN) geostationary transfer orbit. The launch complexes ** Symbol Rate JPL SCIENCE DATA PROCESSING FACILITY provide complete support for vehicle assembly, fueling, checkout and launch operations. JPL science data processing will be done using the JPL

SDS. SDS software and storage will be hosted by cloud WIDE AREA NETWORKS (WANS) [ Figure 4-7] and information rate of 3.45 Gbps via a JPL provided KA-BAND COMMUNICATIONS services, likely Amazon Web Services (AWS) in Oregon. NISAR telecommunications transmitter. ISRO supplies the Ka-band electronics Wide Area Networks (WANs) will be used for long-dis- ISRO’s NRSC facility operates an Earth science downlinks include Ka-band downlink and a 0.7m High Gain Antenna (HGA) mounted on the NASA DISTRIBUTED ACTIVE ARCHIVE tance exchanges among NISAR facilities. All WANs will link and processing center in Shadnagar, India, near to NASA and ISRO stations spacecraft’s nadir surface to be used by both ISRO and CENTERS (DAACS) consist of circuits carrying TCP/IP-based traffic. Hyderabad. This facility is the primary center for ISRO at 4 Gbps and 2.88 Gbps, JPL Ka-band transmitters, through a JPL provided and Ka-band communications from the observatory during NASA’s Earth Observing System (EOS) operates Distrib- respectively, and S-band uplink controlled switch. The antenna gimbal and control of 4.5 TELECOMMUNICATIONS nominal science operations. NRSC plans to place a uted Active Archive Centers (DAACs) around the United and downlink from and to ISRO the gimbal will be provided by ISRO. There will be 15 ground stations. Ka-band reception antenna here within the existing States and has been interoperating with foreign sites. The NISAR observatory’s telecommunications system to 20 downlink sessions per day, with average session Integrated Multi-Mission Ground segment for Earth Ob- For NISAR, the Alaska Satellite Facility (ASF) DAAC has provides for one uplink path and three downlink paths. duration of less than 10 minutes. Note that there servation Satellites (IMGEOS) facility at SAN. ISRO also been selected. The DAAC will utilize AWS cloud services The uplink path is from ISRO’s command center at are separate Ka-band telecom transmitters, but they plans to use another Ka-band ground station (Bharati for processing, storage and distribution. ISTRAC through the observatory’s S-band antenna share the same Ka-band antenna. This system is fully in Antarctica) for science data downlinks. Primary mounted on the ISRO spacecraft bus. The three down- redundant and cross-strapped except for the antenna playback of science data, however, will utilize NASA ISRO TELEMETRY, TRACKING AND link paths are as follows: tracking and engineering and Ka-band gimbal. stations of the NEN at the ASF and Svalbard (Norway). COMMAND NETWORK (ISTRAC) telemetry, from the same S-band antenna back down These stations are shown in Table 4-5 and Figure 4-8. to ISRO’s spacecraft operations center at ISTRAC; The ISRO ISTRAC facility in Bangalore will be used for instrument data from both L- and S-band systems, spacecraft operations and to schedule and operate a TABLE 4-5. NISAR KA-BAND GROUND STATIONS through the shared spacecraft Ka-band antenna set of S-band Telemetry, Tracking and Commanding NASA/ISRO STATION ID USAGE PLAN LATITUDE (°) LONGITUDE (°) ALTITUDE (M) (provided by ISRO) to ISRO’s NRSC facilities near Hyder- (TTC) stations. abad via ISRO’s Ka-band ground stations at Shadnagar NASA Alaska AS3 Primary Site: Maximum Utilization 64.859 °N 147.854 °W 431 and Antarctica; and the same instrument data and NATIONAL REMOTE SENSING CENTRE NASA Svalbard SG2 Primary Site: Maximum Utilization 78.230 °N 15.398 °E 499 (NRSC) engineering telemetry through the shared spacecraft NASA Punta Arenas PA Backup/Secondary Site – As Needed 52.938 °S 70.857 °W 17 Ka-band antenna to NASA NEN stations, (Figure 4-7).

The ISRO National Remote Sensing Centre (NRSC) ISRO Shadnagar SAN Primary Site 17.028 °N 78.188 °E 625 operates an Earth science acquisition, processing and ISRO’s 2.88 Gbps Ka-band system provides for science ISRO Antarctica ANT Primary Site 69.394 °S 76.173 °E 0 dissemination center in Hyderabad, India. For NISAR, data downlink to Indian ground stations with an effec- this center operates two Ka-band stations as part of tive information rate of 2.0 Gbps. Ka-band downlink their Integrated Multi-Mission Ground segment for to NASA ground stations will be at 4.0 Gbps with Earth Observation Satellites (IMGEOS), one near NRSC in Shadnagar, India, and another remote station in 38 NISAR Science Users’ Handbook 39 NISAR Science Users’ Handbook

[ Figure 4-8] MECHANISMS • Workshops Locations of NISAR • Web-form AGENCIES Ka-band ground • Webex stations (NASA NISAR PROJECT COMMUNITY JOINT SCIENCE TEAM & stations in Alaska, SCIENTISTS INPUTS/ TEAM (NASA & PROGRAM Svalbard and Punta REQUESTS ISRO) OFFICE Arenas, and ISRO APPLICATIONS stations in Shadnagar DEVELOPERS Assess impacts on and Antarctica reference observation are shown). plan, mission ops and PROVIDE FEEDBACK ON RECONCILIATION team resources

[ Figure 4-9] Mission operations will be a joint JPL-ISRO effort. polarizations transmitted, and cross-polarizations. A

Flowchart showing steps to Day-to-day observatory operations will be conducted cross polarization mode, for example, receives the be followed for long-term at the ISRO Telemetry Tracking and Command Network horizontally polarized component of the return signal 4.6 MISSION PLANNING AND impacts to L1/L2 science requirements. If resource vio- re-planning of Reference (ISTRAC) center in Bangalore. ISTRAC monitors and when vertically polarized pulses were transmitted, and OPERATIONS lations or performance impacts are identified, iteration Observation Plan. This process controls the spacecraft, downlinking spacecraft telem- vice versa. From the NISAR science orbit, the instru- will be required. Since nearly all objectives are best satisfied with will be followed periodically etry to a local archive from where JPL can pull data ment’s pointing accuracy is such that the L-SAR data regular repeated observations of any given science (roughly every 6 months) as needed. All science data is downlinked via the JPL can be used to produce repeat-pass interferograms JPL will develop the coordinated observation plan that target, the NASA-ISRO joint science team will create an for updating the plan Ka-band telecom, initially processed, and archived first sensitive to large-scale land deformation rates as small takes into account spacecraft power, maneuvers, data during operations. overall science observation strategy that establishes a in the JPL Science Data System, and then in the ASF as 4 mm/year. throughput sizing and availability of downlink channels. nominal repetitive observing baseline prior to launch. DAAC, from where ISRO can pull the data as-needed. In That plan will be sent to ISRO for uplink to, and execu- It is anticipated that the Joint Science Team will alter addition, a subset of L-band and S-band data (speci- To meet the requirements of all science disciplines, the tion on, the observatory. JPL manages all L-band SAR the nominal observation plan during the course of the fied by SAC) will be downlinked directly to India (NRSC L-SAR radar instrument is designed to deliver fast sam- instrument operations, with the ISRO uplink station mission. Applications and other government users ground station) via the spacecraft Ka-band telecom. pling, global access and coverage, at full resolution and serving as a pass-through for L-band instrument com- may also request plan changes. The project team will with polarimetric diversity. The technological innovation mands. ISRO manages all S-band SAR operations. All strive towards accommodating these within the project 4.7 INSTRUMENT DESIGN that allows this performance is the scan-on-receive instrument operations are guided by the coordinated constraints. These post-launch updates to the Refer- “SweepSAR” design, conceived and refined jointly with observation plan, with specific commands/sequences The L-band Synthetic Aperture Radar (L-SAR) instru- ence Observation Plan will be applied on a quarterly engineering colleagues at the German Space Agency to implement the plan developed by the respective ment is the focus of the NASA-chartered science goals or semi-annual frequency basis, with accommodation (DLR) under the DESDynI study phase. organizations. Navigation is led by JPL, with maneuver for NISAR. To meet these goals, it will be heavily utilized of urgent response requests in response to natural design provided from JPL to ISRO to implement the during the mission. Current mission scenarios have the hazards and other emergencies (Figure 4-9). SweepSAR (Figure 4-10) requires the ability to receive maneuvers. Maneuver implementation is fed back to instrument on and collecting data for 45-50% per orbit The Joint Science Team will rely on Mission Oper- the echoed signal on each element independently, JPL as input for the next maneuver planning process. on average, with peaks as high as 70%. ations and the Project Science Team to understand such that localized echoes from the ground can be In the same vein, JPL provides the telecom sequence the implications of any changes to the observation tracked as they propagate at the speed of light across for the NASA-provided Ka-band telecom subsystem The L-SAR is a side-looking, fully polarimetric, plan. Changes will be specified through target/mode/ the swath. As an echo moves from receive element to used for all science data downlink, while ISRO feeds interferometric synthetic aperture radar operating at a attributes as is currently done. Mission Operations receive element, the signals from neighboring elements back to JPL the ISRO-provided Ka-band telecom wavelength of 24 cm (Rosen et al., 2015). The L-SAR is Team will then rerun the mission scenario simulation to must be combined to form a continuous record of the subsystem downlink contacts. JPL is responsible for capable of 242-km swaths, 7-m resolution along track, examine resource (power, thermal, data downlink, cost) echo. Given the width of the swath (~244 km), returns producing the required science data specified by NASA 2-8 m resolution cross-track (depending on mode) and constraint violations. The Project Science Team will from two or more echoes must be processed simulta- and delivering them to NASA DAAC(s). The ISRO NRSC can operate in various modes including quad-polari- apply the updated Candidate Observation Plan through neously. This operation is best performed using digital will process and distribute the required science data metric modes, i.e., transmitting in both vertical and the science performance models to see if there are any combining techniques, so the received echo is digitized specified by ISRO. horizontal polarizations, and receiving in both the same 40 NISAR Science Users’ Handbook 41 NISAR Science Users’ Handbook

[ Figure 4-10] Receive be mechanically and electrically separated, to keep the vary the pulse rate in order to move the gaps around coupling manageable. over time. The data can then be processed to gapless Sweep-SAR technique illustration Rx Signals of enabling “SweepSAR” concept, Processed imagery by interpolating across the gaps. Individually which allows full-resolution, Filtering, decimation, calibration estimation and Transmit multi-polarimetric observations combining are done in a set of FPGAs or ASICs on each Over most of the world, the instruments will be operat- across an extended swath radar. This complication exists for both L-band and ed independently. The requirements for range resolu- (> 240 km). By transmitting S-band and leads to a multiplicity of parallel processing tion, polarization and radar modes supported by the in- energy across the full feed efforts in the spaceborne electronics. The SweepSAR strument are science target dependent. The instrument aperture, a wide swath is technique was demonstrated in an airborne configura- supports a fixed set of polarizations and bandwidth All Tx On illuminated on the ground. tion to show its efficacy (Hensley et al., 2010). combinations of those listed in Table 4-6. The physical Each patch element on the feed layout of the payload is depicted in Figure 4-11. can receive independently, With SweepSAR, the entire incidence angle range allowing localization in time, hence is imaged at once as a single strip-map swath, at During data collection, the observatory performs near space, of the return echo scattered Radar Flight Path full resolution depending on the mode, and with full zero Doppler steering to compensate for the Earth’s from the ground. Note: Transmit and scanning receive events polarization capability if required for a given area of the rotation during observations. The mission will be overlap in time and space. Along- interest. Azimuth resolution is determined by the 12-m conducted in a left-only mode of operation to better reflector diameter and is of order 8 m. optimize science return, with the expectation that other track offset shown is for clarity Radar Swath of presentation only. sensors can achieve science in the high Arctic regions. [ Figure 4-11] Because the radar cannot receive echoes during trans- NISAR instrument mit events, there are one or more gaps in the swath if The radar is designed to operate in a variety of modes physical layout. the radar’s pulse rate is fixed. NISAR has the ability to to satisfy the various science objectives; these may

immediately upon reception, filtered, decimated, and instrument up to the radiated energy from the feed then sent to a signal combiner. aperture. Thereafter, both will share the same reflector, with a nearly identical optical prescription (F/D=0.75).

On transmit, the entire radar feed aperture is illumi- Because a distributed feed on a reflector-feed antenna Boom nated, which creates a narrow strip of radiated energy has a single focus, much of the radiated and received Attach Radar Antenna Point on the 12-m reflector that illuminates the full 242 km energy is not at the focus. Since S-band wavelength is Structure swath on the ground. On receive, the echo illuminates 2.5 times shorter than L-band, yet the feed is the same Star Tracker L-SAR RF Feed the entire reflector, and that energy is focused down length to achieve identical swath coverage, the S-band L-SAR Instrument Aperture S-SAR RF to a particular location on the radar feed aperture system has greater deviations from the focus. Thus, Controller Feed Aperture Radar Antenna (Internal) depending on the timing of the return. The narrowness the design has been iterated to derive the best offset, Reflector Radar Antenna Boom of the receive beam on the ground (due to the wide tilt and phasing of each radar to balance the perfor- reflector illumination) minimizes ambiguity noise so mance across the two systems. This analysis has been I3K Spacecraft Bus that individual pulse can be tracked separately as they done independently by the JPL and ISRO teams, then L-SAR Digital Electronics sweep across the feed. cross-compared to validate. Solar Arrays L-SAR Transmit/ The SweepSAR L-band and S-band radars are being For the radars to operate together as a dual-frequency Receive S-SAR RF Radar Modules and designed to work independently or together. The system, it is necessary to share oscillator and timing Instrument Digital Structure and L-band hardware will be built at JPL, and the S-band information to lock their pulse repetition frequency Electronics Electronics electronics portion at ISRO. The feed apertures at L- together, which will be done with simple interfaces. and S-band are also built by JPL and ISRO, respective- Another concern is the coupling between the feed ly, phase-matched to their respective electronics and apertures. In the current design, the two apertures will cabling. In this sense, each radar is a self-contained 42 NISAR Science Users’ Handbook 43 NISAR Science Users’ Handbook

TABLE 4-6. SUPPORTED POLARIZATIONS AND BANDWIDTH COMBINATIONS • Use of L-band and S-band jointly will allow a shows the fully integrated and deployed observatory

ELEMENT DESCRIPTION good estimate of the ionosphere using dual-band system. The 12-meter Radar Antenna Reflector (RAR) mitigation techniques (Rosen et al., 2010). is at top, supported by the Radar Antenna Boom (RAB). Operational SweepSAR scan-on-receive Implementation • Use of L-band and S-band jointly to extend the The boom is attached to the radar instrument structure (RIS), which is itself attached to the ISRO I3K spacecraft Configuration • 12-m diameter mesh reflector used for both L- and S-band range of sensitivity for biomass estimation and bus. Extending on either side of the bus are two solar • S-band 2 x 24 / L-band 2 x 12 patch array, one TR module per surface deformation, and aid in estimating soil patch-pair subarray per polarization moisture. arrays each with three panels that together supply • Independent S- and L-band electronics with timing synchroniza- approximately 4,000 W of power when illuminated (i.e., tion for possible simultaneous operations • Use of L-band and S-band jointly to study differen- • Digitization at each receive array element followed by real-time at all times when not in eclipse or off Sun-pointing). tial surface roughness and volume scattering ef- combining The radar payload integration (L-band and S-band in- fects, improving classification of natural surfaces. Radar Center Frequency S-band 3200 MHz; L-band 1260 MHz, simultaneous operations tegration) will occur at JPL, and the overall observatory possible • Use of L-band and S-band jointly or separately integration will occur at ISRO Satellite Center (ISAC) in to study decorrelation rates of natural surfaces, Bangalore, India. The main elements of the system are Realizable Bandwidths • 5 MHz (L) • 10 MHz (S) improving the utility of interferometry for change illustrated in Figures 4-11 and 4-12. • 25 MHz (S); 20+5 MHz split spectrum (L) detection and change classification. • 37.5 MHz (S); 40 MHz (L) • > 5 MHz (S); 80 MHz (L) ISRO provides the spacecraft bus, which includes These capabilities will provide researchers with a all systems required for central command and data Realizable Single-pol through quad-pol, including compact-pol and split-band fundamentally new global (at L-band) and globally handling, uplink and downlink, propulsion, attitude Polarizations dual-pol distributed (at S-band) data set for research. It is im- control, solar arrays, the S-band radar electronics, and Incidence Angle Range ~34-48 degrees portant to note that the system downlink is at present a Ka-band telecom system and gimbaled High Gain fully tasked, so opportunities for dual-band collection Performance • < -20 dB NES0 depending on mode Antenna (HGA) dish. ISRO also provides the launch • < -15 to -20 dB ambiguities variable across swath must be balanced against alterations to the nominal vehicle. NASA/JPL provides the L-band radar electron- • 3-10 m range resolution, sub-pixel geolocation; ~ 7 m azimuth observation plan. ics, the deployed boom and radar reflector, a high- resolution capacity/high-speed Solid State Recorder (SSR), the 4.8 FLIGHT SYSTEMS/SPACECRAFT GPS, high-rate Ka-band Payload Communication include single polarization (horizontal or vertical only) While the global measurements largely will be at Subsystem (PCS), the pyro firing system for boom and The NISAR flight system design, development, integra- modes, dual polarization (e.g., transmit in horizontal L-band, there will also be regular acquisitions at antenna deployments, and a Payload Data System tion, testing and operations are a joint venture, with polarization and receive in both horizontal and vertical S-band over India. As the mission evolves, insights into (PDS) that monitors and controls the JPL systems equivalent-scale contributions from both JPL and ISRO. polarization) modes, quad polarization (transmit in the most beneficial uses of S-band in place of L-band and handles communications between all of the JPL The suite of flight systems consists of the launch vehi- both and receive in both) modes, special “quasi-quad” or as a dual-frequency system will be gained, with the systems and the ISRO spacecraft bus. cle and free-flying observatory. The NISAR observatory modes simulating quad polarization modes with a observation plan modified accordingly. is designed around the core payloads of L- and S-band lower data rate, circular polarization modes, and com- The NISAR science requirements levy special functional SAR instruments, designed to collect near-global radar binations of any of the above (one for L-band, and a dif- The Shuttle Imaging Radar-C was the first orbiting requirements on the heritage ISRO spacecraft and data over land and ice to satisfy the L1 science goals. ferent for S-band). Table 18-1 in Appendix F shows the multi-frequency, multi-polarization SAR around Earth its associated mission operations. Both L-band and In addition to the two radar instruments, the NISAR pay- available modes for the L-SAR and S-SAR instruments. and demonstrated the value of having multiple wave- S-band radar payloads require substantial average load includes a global positioning system (GPS) receiver For each of the observation targets, there is a single lengths. Possible benefits include: power for operation on-orbit, which leads to a space- for precision orbit determination and onboard timing mode (polarization, bandwidth, radar band combination) • Use of S-band in polar regions can reduce the references, a solid-state recorder, and a high-rate data that is used over that area. For overlapping targets, impact of the ionosphere, since the S-band signal downlink subsystem to enable transmission of the such as background land and U.S. agriculture, the more will be five times less sensitive than L-band to high-volume science data to the ground. Figure 4-11 capable mode is a superset of capability of the other ionospheric perturbations. mode. Transition between these modes is seamless, which nearly eliminates data loss. 44 NISAR Science Users’ Handbook 45 NISAR Science Users’ Handbook

[ Figure 4-12] Radar 4.9 PROJECT STATUS and some flight model hardware testing are presented Antenna Solar Spacecraft in stowed Boom and plans for building the remaining flight hardware are Array The NISAR project conducted a successful Preliminary configuration. Panels described. At the time of writing, all subsystem reviews Radar Design Review (PDR) on June 21-23, 2016. Key Deci- leading up to CDR have been successfully completed. Structure sion Point-C (KDP-C), the review to confirm the mission Remaining reviews include the full radar instrument for detailed design and development, was held on CDR (April 2018), the project CDR (October 2018), and August 23, 2016. Shortly thereafter the project entered the Mission Systems CDR (October 2019). The planned the Design and Build Phase (Phase C), and the project launch date for the mission is late 2021. team has been working toward the Critical Design Re- view (CDR), where the results of the engineering model ISRO I3K Spacecraft Radar Antenna L-SAR Electronics Bus Launch Vehicle Reflector Adapter

craft design with large deployable solar arrays. The To date, the spacecraft design has been optimized to baseline science observation plan calls for up to 26 accommodate all key and driving requirements, and Tb (Terabits) per day of radar data collection, downlink refined technical analyses show that predicted and processing. This plan drives the spacecraft design performance meets science needs. The solar arrays to include a Ka-band telecom system to accommodate have the required capability, plus an extra string of the high bandwidth requirements. The spacecraft AOCS cells for contingency. The Ka-telecom system is sized (Attitude and Orbit Control Subsystem) is designed to to handle the throughput baseline up to 26 Tb per address several critical science-enabling functions: day, though the margins are tight for many of the 1) it must fly along the same orbit to within narrow elements of the data system, many of which are part tolerances (500 m) over the life of the mission; 2) it of the ground system. For pointing control, rigid-body must be able to control the attitude of the observatory analysis shows that the system is controllable to the as a whole to point at a fixed angular location relative required accuracy. Further flexible body analysis will be to an ideal orbit track and nadir at any given point on performed in the coming years to further characterize orbit; 3) it must be able to slew and hold attitude to requirements compliance. Though the nominal plan observe Earth from both sides of the orbit plane. For calls for left-looking observations only, the spacecraft orbit control, there is sufficient fuel to accommodate is being designed to be capable of operating either at least 5 years of operations at the chosen altitude. looking to the left or right. This would support contin- The propulsion system is agile enough to perform the gency operations, or imaging of the higher latitudes of necessary small orbit control maneuvers every few the Arctic possibly in an extended mission phase. days that are required to maintain the strict orbital tube requirements. JPL augments the ISRO spacecraft capability with GPS receivers, providing GPS time message and a 1pps (pulse per second) signal to the spacecraft and radar instruments. NISAR Science Users’ Handbook 47 NISAR Science Users’ Handbook

5 MISSION DATA PRODUCTS

NISAR data products will be organized by product level, and crop area delineation, as well as additional ecosys- with Level 0 being a raw form of data and Level 3 tem and land-cover applications that may be developed being a geocoded derived science product in physical during the NISAR mission. Ancillary data needed to cre- units. The NISAR L0A product is the received raw data ate these products, such as orbits and calibration files with metadata added to support storage at the DAAC. are included in the metadata layers of these products. The NISAR L0B product is a refined version of the For interferometry, the dense field of range and azimuth radar signal data with transmission artifacts removed. offsets, suitable for local resampling to account for NISAR L1 products will include all products in radar substantial motion between scenes, are also included (range-doppler) coordinates, including the Single Look as metadata at L1 and L2. The metadata layer descrip- Complex (SLC), Multi-Look Detected (MLD), unwrapped tions are in Tables 20-1 – 20-8 in Appendix H. THE NISAR DATA (UNW) and wrapped nearest-time interferograms (IFG), and polarimetric images. The NISAR L2 products will be The NISAR data product levels have been defined in PRODUCT LEVELS geocoded versions of all the L1 products (except MLD accordance with the NASA EOSDIS criteria for sci- HAVE BEEN DEFINED and the wrapped interferogram). The NISAR Science ence product processing level classification (https:// Data System (SDS) team will generate the L-band L0- science.nasa.gov/earth-science/earth-science-data/ IN ACCORDANCE L2 products (Table 5-1), and the NISAR Project Science data-processing-levels-for-eosdis-data-products). The WITH THE NASA and Science Teams will generate the L3 products at L0-L2 products will consist of two major components: selected calibration/validation sites distributed globally. self-describing binary data and quick-look metadata. EOSDIS CRITERIA FOR The binary data component is based on the HDF-EOS5 SCIENCE PRODUCT Level 1 products, including the SLC, IFG, Nearest-Time specification (Klein and Taaheri, 2016), an HDF5-based (UNW), and the L2 geocoded versions of these products format that has the following advantages: PROCESSING LEVEL (produced for all except IFG) will be relevant for studies • Open, self-describing format of solid Earth deformation and cryospheric sciences. CLASSIFICATION. • Supports hierarchical tree data arrangement The NISAR ecosystem products include the L2 Geocod- ed Single Look Complex (GSLC), the L1 polarimetric • Supported by GIS and database software covariance matrix in range-doppler coordinates (COV), • Provides flexibility to support any binary data and the L2 polarimetric covariance matrix in geocod- format making it scalable to support all levels of ed map coordinates (GCOV). The L2 geocoded SLC NISAR products product, generated from the L1 SLC product, enables • Widely used for a range of NASA EOS missions users to perform amplitude as well as interferometric (e.g., MODIS, AIRS, TRMM, CERES, MISR, GSSTF, analysis directly on a geocoded grid. Depending on the and Aquarius) polarimetric acquisition mode (single, dual, or quad), the GSLC product can have 1, 2, or 4 complex-valued layers. Based on the polarimetric acquisition mode In general, each L0-L2 product will be distributed as (single, dual, or quad) and processing option (symme- a single HDF-EOS5 granule. The NISAR quick-look trized or non-symmetrized cross-polarimetric channel), metadata accompanying the binary data will be in an the polarimetric covariance matrix can have from 1, 3, XML-based format. or 6 complex-valued layers. These products primarily support the NISAR ecosystem requirements of biomass Level 3-4 processing will be conducted by the NISAR estimation, disturbance detection, inundation mapping, science team. Measurements will include biomass, trekandshoot/Shutterstock 48 NISAR Science Users’ Handbook 49 NISAR Science Users’ Handbook

TABLE 5-1. NISAR L-BAND L0-L2 PRODUCTS disturbance/recovery maps, ice and land displacements 5.1 L0 DATA PRODUCTS

PRODUCT PRODUCT SCOPE DESCRIPTION and velocity fields, all in geocoded coordinates. These The NISAR SDS will produce two types of Level 0 data. LEVEL NAME products will also be delivered to NASA DAAC; however, The L0A product is the received raw data with metada- Incoming Data (Raw) Global Raw downlinked data delivered [ ] they will be generated only over selected regions of the Figure 5-1 ta added to support storage at the DAAC. Although the to SDS with metadata added world for calibration and validation purposes. for archiving Data Product Levels. Prod- L0A dataset will be publicly available, this downlinked ucts through Level 2 will be raw data will not be directly useable by the scientific L0 Global Radar Signal Data Corrected, aligned, and produced for the entire mission Figure 5-1 shows the overall data products that will (RSD) time-ordered radar pulse community. The L0B product is a refined version of the data set. Products at higher be generated by the project and delivered to the NASA data derived from RAW radar signal data with transmission artifacts removed. products and used for further levels will be produced by the DAAC. The detailed product description, including all the The project will process all L- and S-band data ac- processing science team for calibration data layers that will be specified in the L0-L2 products, quired over the NASA downlink network to L0B, which and validation purposes. is given in tables in Appendix H. Range-Doppler Single Global Standard L1 product that will is a reformatted, organized and regularized version Look Complex (SLC) be used to generate all higher level products

Multi-Look Detected Global Multi-looked amplitude prod- (MLD) uct in ground range coordi- LEVEL 0 LEVEL 1 LEVEL 2 LEVEL 3-4 (VALIDATION ONLY) nates.

Nearest-Time Interfer- Antarctica and Multi-looked flattened (WGS84 L1 Ancillary Data (Orbits ogram (IFG) Greenland. Nearest ellipsoid) interferogram Level 0, Level 1, Level 2, Level 3-4 Altitude) In Auxilliary Data pair in time and co- with topographic fringes in Validation, Browse Products, and (Atmosphere Correction, Catalog Delivered To DACC pol channels only. range-Doppler coordinates. DEM, etc.) Woody Biomass For VP Regions <100 t/ha Nearest-Time Un- Global except Ant- Multi-looked, unwrapped wrapped Interferogram arctica and Green- differential Interferogram in Calibrated Single-Look Vegetation Disturbance VP Complex (SIC) and (UNW) land. Nearest pair range-Doppler coordinates. Radar Product in time and co-pol Reformatted multilocked images.* channels only. Raw Data Unwrapped/wrapped Interferograms* polari- Geocoded Wetlands VP metric images In radar Polarimetric Images Inundation Area Polarimetric Covari- Global and all chan- Polarimetric covariance Time Varying Parameter coordinates ance Matrix (COV) nels. Single/dual/ matrix (1, 3, or 6 layers) in Files VP quad pol. range-Doppler coordinates. (Multiple Modes) Crop Area

Geocoded SLC (GSLC) Global and all Geocoded L1 SLC product channels. using the MOE state vectors Vegetation and a DEM. Structure*

Geocoded Near- Global except Ant- Geocoded multi-looked un- Ground Deformation/ VP est-Time Unwrapped arctica and Green- wrapped differential Inter- Rates L2 Geocoded Interferogram land. Nearest pair ferogram. Same as UNW but PRODUCTS Single-Lock Com- (GUNW) in time and co-pol resampled onto a UTM grid. VP plex (SLC) Images Change Proxy channels only. Instrument and Geocoded

Unwrapped/Wrapped Geocoded Polarimet- Global and all chan- Geocoded polarimetric cova- Ecosystems Interferograms Ice Sheet/Glacier VP ric Covariance Matrix nels. Single/dual/ riance matrix (1, 3, or 6 layers) Velocity And Velocity Solid Earth Def. (GCOV) quad pol. using the MOE state vectors Change and a DEM. Dynamics of Ice VP Sea Ice Velocity VP = Validation Product

NASA-SELECTED DAAC: Alaska Satellite Facility * Experimental 50 NISAR Science Users’ Handbook 51 NISAR Science Users’ Handbook

of the instrument science data coming down in the to assist in further processing into L1/L2 products: width of the mode and the size of the antenna. Other L1 The L1 SLC is used to derive other L1 and L2 products.

science telemetry. L0B data is a basic input to a SAR • Nominal pulse/chirp characteristics and actual data products, including interferograms, interferometric This product will contain individual binary raster layers image formation processor and is typically the starting replicas of transmitted chirps correlation maps, and polarimetric backscatter, will be representing complex signal return for each polariza- point for many SAR scientists. derived from the SLC product. The interferograms and tion layer. The SLC data corresponding to the auxiliary • Doppler centroid estimate correlation maps will be formed from nearest-in-time 5 MHz sub-band is stored in a similar format but in INCOMING RAW DATA (L0A) • Orbit and attitude data pairs of data sets. Given N interferometrically viable a separate data group within the HDF-EOS5 product • Geographic coordinate boundaries data sets, one can produce N(N-1)/2 unique interfero- granule. The SLC product is also packed with input, The L0A data product represents a collection of time- grams, an impractically large number of interferograms instrument and processing facility information; pro- tagged raw data packets and telemetry information • I/Q bias estimates to produce and store given that not all are typically cessing, calibration and noise parameters; geolocation downlinked to the GD, and typically do not have any • Calibration antenna patterns and related used in scientific analysis. Forming nearest-in-time grid; and data quality flags. The SLC product complex overlap. The data are ordered in time but not all com- information pairs yields only N-1 interferograms, each of which is floating point backscatter is beta nought with second- munication artifacts, missing data and synchronization • Calibration noise records typically used in further analysis. Another L1 product, ary layer LUTs provided to convert to sigma naught and errors are necessarily corrected. The raw signal data the Multi-Looked Detected imagery, will be derived gamma naught. from the primary imaging band and the auxiliary 5 MHz • Channel delay calibration estimates from the SLC by taking its modulus on a pixel-by-pixel sub-band are interleaved in this product and are not • Polarimetric compensation matrix basis and averaging. The SLC and MLD products con- NEAREST-TIME INTERFEROGRAM (IFG) yet decomposed into corresponding I/Q channels yet. tain look up tables for radiometric ellipsoid correction. A multi-polarization L0A product will contain layers The L0A-to-L0B processor aligns or rearranges the raw The L1 IFG product represents the ellipsoid height The project plans to use the Medium-fidelity Orbit corresponding to each polarization (for both the primary radar signal data to ease further processing and does corrected, wrapped interferogram generated from two Ephemeris (MOE) product, available within one day of and auxiliary band). Data are compressed by the radar not modify the actual signal data (i.e., operations like L1 range-Doppler SLCs in the range-Doppler geometry acquisition for L1 and L2 processing, as it is nearly using a block floating point quantization algorithm, and RFI removal is not applied at this stage). Raw signal of the earlier acquisition. The data is arranged on a as accurate as the final orbit product and reduces the L0A product will maintain this compressed state. data and metadata corresponding to the main imaging uniformly spaced, increasing zero-Doppler azimuth processing latency. band and the auxiliary 5 MHz sub-band are stored in time and increasing slant range grid. The IFG product Each L0A raw data product will be reduced to a L0B is primarily meant for detecting grounding lines and separate data groups within the HDF-EOS5 product RANGE-DOPPLER SINGLE LOOK product for use in further processing by the SDS. Each is only generated for acquisitions over Antarctica and granule. A multi-polarization L0B product will contain COMPLEX (SLC) radar pulse will be tagged with its own metadata (e.g., layers corresponding to each polarization (for both the Greenland. WGS84 ellipsoid is used as the reference receive time and PRF). The data are arranged on an in- primary and auxiliary band). Each radar pulse will be This product refers to the standard range-Doppler surface for flat earth correction and the products are creasing azimuth time and increasing slant range grid. tagged with its own metadata (e.g., receive time and geometry SLC imagery that are operationally delivered multi-looked to a posting of 25 meters on the ground. The downlinked data are packaged into L0A product on PRF). The data are arranged on an increasing azimuth by SAR sensors around the world (often referred to as The auxiliary 5 MHz sub-band is used to apply an reception. L0A data are primarily for archive purposes; time and increasing slant range grid. The L0B product Level 1.1 by ESA and JAXA). The L1 SLC product will be ionospheric phase screen during processing. it is anticipated that users interested in image process- is the primary input for L1 product generation. distributed in the zero-Doppler radar geometry conven- ing will start with L0B. tion. The L0B-to-L1 processor will handle PRF changes The L1 IFG product will contain individual binary raster 5.2 L1 DATA PRODUCTS within a data granule and the output imagery will be layers representing complex numbers with the ampli- RADAR SIGNAL DATA (L0B) on a grid characterized by a fixed set of starting slant tude representing coherence and the phase represent- There are multiple L1 products to support the NISAR range, starting azimuth time, azimuth time interval, ing interferometric phase for each co-pol channel. In The L0B product consists of aligned and filled raw science disciplines. The L1 SLC data product is the out- and slant range spacing values, to allow for easy addition to the metadata of the original L1 SLC gran- radar signal data that are used to derive higher level put of a SAR image formation processor. It is calibrated interpolation of the auxiliary 5 MHz sub-band layers to ules, lookup tables for the perpendicular and parallel science products. The block floating point quantized for time and phase delays in the radar and propagation match the primary image layers. All the primary image baseline components, range and azimuth offsets are samples from L0A raw data product are decoded and path (using a standard atmosphere/ionosphere model), layers for a multi-polarization or multi-frequency prod- also included. The interferogram will be flattened to the packed into complete range lines in the L0B product. and for antenna pattern effects and measured pointing uct will be generated on a common azimuth time-slant ellipsoid. A DEM will be used for fine registration. Sampling Window Start Time (SWST) shifts for the radar offsets. Each science target may require a different res- range grid. pulses are aligned and each pulse is annotated with olution and set of polarizations, hence the product will mode and PRF changes as well as missing data infor- accommodate multiple modes. This product is created mation. The following metadata are added at this stage at the fullest resolution possible, given the range band- 52 NISAR Science Users’ Handbook 53 NISAR Science Users’ Handbook

NEAREST-TIME UNWRAPPED INTERFERO- nel combinations for one L1 range-Doppler single-, inputs. In addition to the metadata fields provided in GEOCODED NEAREST-TIME UNWRAPPED GRAM (UNW) dual- or quad-pol SLC product in the range-Doppler the original L1 SLC, additional lookup tables that allow INTERFEROGRAM (GUNW) geometry. The polarimetric channels are multiplied in users to transform ground range coordinates to slant The L1 UNW product represents the unwrapped, multi- The L2 GUNW product is derived from the L2 UNW a lexicographic polarimetric basis. The COV product range coordinates are also included in the HDF-EOS5 looked differential interferogram generated from two L1 product by projecting it onto a DEM in the UTM system also contains the multilooked backscatter of single pol product granule. range-Doppler SLCs in the range-Doppler geometry of (like Landsat) at 25 meter posting. The data is arranged data. The COV products are radiometrically calibrated the earlier acquisition. The data is arranged on a union a uniformly spaced, north-south and west-east for spread loss and antenna pattern in all polarimetric 5.3 L2 DATA PRODUCTS formly spaced, increasing zero-Doppler azimuth time aligned UTM/WGS84 grid. All the lookup tables includ- channels. Radiometric ellipsoid correct is also applied. and increasing slant range grid. For every ingested L1 Level 2 products are geocoded versions of all the L1 ing phase corrections are transformed from image The physical quantity distributed with the COV products SLC product, an archived L1 SLC product correspond- products (except MLD and the wrapped interferogram coordinates to geographic coordinates. is the radar brightness beta nought. The products are ing to the same imaging geometry and nearest in time IFG) derived from the L1 images. The GSLC prod- multi-looked to a posting of approximately 25 meters is identified and an UNW processing job is launched. uct contains look up tables for radiometric ellipsoid GEOCODED POLARIMETRIC COVARIANCE on the ground. The data is arranged on a uniformly The UNW product is generated between co-pol chan- correction. The GCOV product has radiometric terrain MATRIX (GCOV) spaced, increasing zero-Doppler azimuth time and nels and for all regions other than Greenland and Ant- correction and includes a full resolution projection increasing slant range grid. The L2 GCOV product is derived from the L1 COV prod- arctica. Digital Elevation Models (DEMs) will be used for angle layer. uct by projecting it onto a DEM in the UTM system (like producing these data products which are multi-looked MULTI-LOOK DETECTED IMAGE (MLD) Landsat) at 25 meter posting. The product also contains to a posting of 25 meters on the ground. GEOCODED SINGLE LOOK COMPLEX the multilooked backscatter for single pol data. Product This product refers to the standard ground range- (GSLC) locations are terrain corrected and the radar cross The L1 UNW product will contain individual binary Doppler geometry multi-looked imagery that is oper- The L2 GSLC product is derived from the L1 section is corrected for terrain-dependent incident raster layers representing single precision floating point ationally delivered by SAR sensors around the world Range-Doppler SLC product and projected onto a DEM angles. The data is arranged on a uniformly spaced, unwrapped phase for each co-pol channel. In addi- (often referred to as Level 1.5 by ESA and JAXA). All in the UTM system (like Landsat). The data is arranged north-south and west-east aligned UTM/WGS84 grid. tion, byte layers with quantized coherence, geometry L1 MLD products will be distributed in the zero- on a uniformly spaced, north-south and west-east The GCOV product is distributed in a lexicographic po- masks and connected component information, and Doppler geometry convention. MLD products are aligned UTM/WGS84 grid. The spacing of the GSLC larimetric basis. All the lookup tables are transformed floating-point layers corresponding to the amplitudes derived from the L1 SLC products by incoherent aver- product in East and North directions will be comparable from image coordinates to geographic coordinates. The of master and slave acquisitions are included in the aging of intensity in the azimuth time-slant range grid to the full resolution original L1 SLC product. The GSLC physical quantity distributed with the GCOV products is HDF-EOS5 granule. In addition to the metadata of the to provide 25 meter ground resolution data and then product can be directly overlaid on a map or combined gamma naught. original L1 SLC granules, lookup tables for parallel projecting the data to an azimuth time-ground range with other similar GSLC products to derive interfero- and perpendicular baseline components, range and grid, under a constant ellipsoid height assumption. The grams and change maps, for example. 5.4 DATA PRODUCT DELIVERY/HOW TO azimuth offsets are also included. Additional metadata data are arranged on a uniformly spaced, increasing ACCESS NISAR DATA will include lookup tables for various phase corrections zero-Doppler azimuth time and increasing ground The L2 GSLC product will contain individual binary (e.g., solid Earth tides, ECMWF tropostatic dry delay range grid. The MLD product backscatter amplitude One Earth Science Data Center (ESDC) has been raster layers representing complex signal return for and ECMWF tropostatic wet delay, ionospheric phase is beta nought with secondary layer LUTs provided to designated by NASA’s Earth Science Division to archive each polarization layer. The GSLC product granule will screen). These phase corrections are not applied to convert to sigma nought and gamma nought. and distribute NISAR science data: the Alaska Satellite contain all the same metadata as the source L1 SLC the data but are available to users for application in Facility (ASF). ASF will receive validated NISAR science product but with the lookup tables referenced to geo- post-processing workflows. The L1 MLD product will contain individual binary ras- data products from the SDS, along with algorithm graphic coordinates instead of image coordinates. The ter layers representing multi-looked signal amplitude source code and ancillary data used in deriving the GSLC product includes a byte layer indicating quality, POLARIMETRIC COVARIANCE for each polarization layer. The L1 MLD product will products and provide long-term archiving and distribu- water bodies and shadow-layover. GSLC product will be MATRIX (COV) not include any layers corresponding to the auxiliary tion services for the general produced globally. 5 MHz sub-band in the HDF-EOS5 product granule. The The L1 COV product represents the multi-looked lookup tables corresponding to the original L1 SLC are cross-product between all possible polarization chan- modified to use ground range image coordinates as NISAR Science Users’ Handbook 55 NISAR Science Users’ Handbook

6 SCIENCE DATA PRODUCTS AND VALIDATION APPROACHES

Section 5 describes the Level 0 to Level 2 data prod- contiguous/overlapping SAR interferograms over all ucts that will be produced operationally by the NISAR time periods of interest. project and made available globally to the science community. This section describes the Level 3 or 4 In the first approach, Global Positioning System (GPS) products that the science team will produce in selected InSAR-derived surface displacements will be compared areas to show that the science requirements of the with point observations of surface motion from collo- mission will be met in each discipline area: Solid Earth, cated continuous Global Navigation Satellite System Ecosystems, Cryosphere. Science requirements have (GNSS) stations (GPS and continuous GPS, or cGPS, a different “level” scheme than products: The science are components of GNSS. Since all requirements are THE SCIENCE TEAM team uses Level 3/4 science data products to validate written in terms of relative displacements (sampling USES LEVEL 3/4 Level 2 science requirements, which are enumerated in the deformation field at individual points), comparisons Appendix D. This section describes the theoretical basis are done on the differences of observed surface motion SCIENCE DATA of the algorithms to be used to create science prod- (from both InSAR and GNSS) between GNSS station lo- PRODUCTS TO ucts for each of the Level 2 science requirements by cations within the scene. For a GNSS station network of discipline, and the anticipated methods to validate each N stations, this will yield N(N-1)/2 distinct observations VALIDATE of these requirements. for comparison, distributed across a range of length LEVEL 2 SCIENCE scales. As discussed below, the methodology differs 6.1 SOLID EARTH SCIENCE PRODUCTS slightly depending on if the comparison is performed REQUIREMENTS. directly on interferograms (Requirement 663) versus Solid Earth science products will be produced for basis functions derived from sets of interferograms co-seismic, transient, and secular displacements. The (Requirements 658/660), but the underlying premise is three primary NISAR Solid Earth L2 requirements on the same: that GNSS provides a sufficiently high-quality secular, co-seismic and transient deformation rates, time series to validate InSAR observations. This ap- that drive the L3 products needed for calibration and proach is appropriate where measurable displacement validation, are listed in Appendix D. is occurring across the cal/val region and the GPS/ GNSS network is sufficiently dense to capture most of 6.1.1 Theoretical Basis of Algorithm the expected spatial variability of the signal. APPROACH TO VALIDATING SOLID EARTH In the second approach, which is appropriate for L2 REQUIREMENTS negligibly deforming regions, the autocorrelation of noise in NISAR interferograms will be examined without Two separate approaches will be used by the NISAR comparison to GPS/GNSS, under the assumption that Science Team for validating the Solid Earth L2 re- surface deformation is essentially zero at all relevant quirements, both of which require the generation of a spatial scales. This method involves differencing InSAR standard set of NISAR L3 data products consisting of displacement observations between a large set of surface displacement time series for selected areas randomly chosen pixel pairs and confirming that the that sample a range of vegetation types, topographic estimates are statistically consistent with there being relief, and strain rates. Generation of these products, as no deformation within the scene. discussed in Section 6.1.2, requires a set of temporally trekandshoot/Shutterstock 56 NISAR Science Users’ Handbook 57 NISAR Science Users’ Handbook

L2 REQUIREMENT 658 – SECULAR L2 REQUIREMENT 660 – COSEISMIC All the Solid Earth requirements call for a minimum vegetation covers (forest, shrub, bare surface, etc.), DEFORMATION RATE DISPLACEMENTS spatial coverage component. Validation of this com- seasonality (leaf on/off, snow, etc.), and terrain slopes. ponent will rely on a combination of assessing the The NISAR Science Team will select a set of cal/val re- To validate relative secular deformation rates (or To validate NISAR’s ability to recover relative coseismic coverage of basic InSAR-quality data and ensuring that gions to be used for this requirement and will list those velocities) from NISAR, Line-of-Sight (LOS) velocity displacements of 100 mm and larger within a scene, the required measurement accuracy is achieved in a sites in the NISAR cal/val plan. data will be used for each pixel in a target region. step functions in surface displacements are estimated suite of selected but comprehensive regions. Many Separate LOS velocities will be generated for ascending at the time of the earthquake from the InSAR and GNSS of these regions will be automatically evaluated as GENERALIZED TIME SERIES ANALYSIS and descending passes to meet the requirement for time series. The simplest version of the InSAR estimate part of the targeted sites for the transient deformation two components of motion over each target location. is a coseismic interferogram spanning the earthquake, The InSAR and GNSS comparisons described above requirement. Although the requirement specifies that the validation assuming negligible post-seismic deformation. Greater will be performed in the framework of generalized span 3 years of data, the NISAR Science Team can accuracy can be obtained by modeling the time series time series analysis, whereby information in each time L2 REQUIREMENT 663 – TRANSIENT series is characterized by one or more underlying basis perform the validation for periods shorter than 3 years using appropriate basis functions (e.g., a secular DISPLACEMENTS provided annual effects are mitigated by using data displacement rate, a Heaviside time function at the functions. The problem is cast as an overdetermined that span multiples of 1 year, or by explicitly modeling time of the earthquake, and an exponential postseismic To validate the L2 requirements on transients, 12-day least squares (LSQ) estimation problem, from which and removing the seasonal displacements. The relative response) and using the offset thus obtained. A similar interferograms will be produced from both descending parameters can be inferred for the simultaneous fit of vector velocity between any two points in the scene will analysis can be done for the GPS time series. and ascending tracks over diverse target sites where various components to the time series, on a sta- be taken as the difference in the LOS velocity at those GNSS observations are available. The two components tion-by-station or pixel-by-pixel basis. Implementation points. In validation approach #1, relative displacements of vector displacement, ascending and descending, will of this approach is described in Section 6.1.2. between each pair of GNSS stations within the SAR be validated separately. These components––which include secular velocities, In validation approach #1, LOS velocity product will be footprint and less than 50 km apart will be calculated. seasonal sinusoids, temporal offsets, and post-seismic used to calculate the relative InSAR velocity between To do the comparison, GNSS coseismic displacements For approach #1, unwrapped interferograms will be exponential decay––represent much of the non- each pair of GNSS stations within the SAR footprint that will be estimated by estimating the amplitude of a used at 100-m resolution to produce point-to-point stochastic variance in the time series and are well-suit- are less than 50 km apart. For subsequent comparison, Heaviside basis function at the time of the earthquake relative LOS measurements (and their associated ed to the specific validation targets. For instance, for the accompanying GNSS velocity differences will be for the 3-component GNSS positions, and the InSAR uncertainties) between GNSS sites. Position obser- Requirement 658 (secular deformation) the veloc- generated by taking the 3-component GNSS position displacements in the same way. The GNSS 3-compo- vations from the same set of GNSS sites and at the ity component of these fits will be used, while for time series, projecting them into the InSAR LOS direc- nent displacements are then projected into the InSAR InSAR acquisition times will be projected into the LOS Requirement 660 (coseismic deformation) the velocity, tion, estimating the GNSS LOS velocities, and differenc- line of sight and differenced to obtain the relative GNSS direction and differenced pairwise. These will be com- Heaviside (instantaneous step), and exponential/ ing the GNSS LOS velocities between all stations pairs. displacements between all station pairs. To test pared to the point-to-point InSAR LOS measurements logarithmic components will be used. To perform the To test NISAR’s fulfillment of the 2 mm/y specification, NISAR’s fulfillment of the 4(1+L½) mm specification, using a methodology similar to that used for validating validations, estimates of the fit parameters for these InSAR and GNSS relative velocity estimates for each the InSAR and GNSS relative displacement estimates co-seismic displacements (as described above), except functions (rather than the raw time series themselves) pair will be differenced, mean and standard deviation of are differenced for each pair of GNSS station locations, that the accuracy specification is 3(1+ ½L ) mm over will be used for the statistical comparisons of InSAR all residuals will be calculated, and a t-test will be per- distance L between stations is calculated, mean and 0.1 km < L < 50 km. To validate the noise in individual and GNSS. formed to check whether the mean error is statistically standard deviation of all residuals is calculated, and a interferograms in Approach #2, interferograms over a consistent with a value ≤ 2 mm/y. t-test is performed to check whether the mean error is set of non-deforming sites will be utilized. In practice, 6.1.2 Implementation Approach for statistically less than 4(1+L½) mm over length scales characterization of transient deformation will usually be Algorithm improved by examining longer time series of interfer- Validation approach #2 is identical to approach #1 0.1 km < L < 50 km (e.g., ≤ 5 mm at 0.1 km and GENERATION OF TIME SERIES FROM SETS ograms – the approach described here validates the except that the relative velocities are determined for ≤ 32 mm at 50 km). OF INTERFEROGRAMS random pairs of InSAR pixels within a scene, and requirement that short timescale or temporally complex the statistics are calculated directly from the InSAR Validation approach #2 is similar to approach #1 except transients can be characterized with a single interfer- The time series analysis will be performed using the estimates. The cal/val regions to be used for both that the relative displacements are determined for ogram. Generic InSAR Analysis Toolbox (GIAnT) (Hetland et al., approaches will be defined by the NISAR Science Team random pairs of InSAR pixels within a scene that does 2012, Agram et al., 2013), which is openly download- and listed in the NISAR cal/val plan. not include a significant earthquake, and the statistics Comprehensive validation requires transient sites able from http://earthdef.caltech.edu. This toolbox are calculated directly from the InSAR estimates. possessing different deformation characteristics (e.g., has been used in many studies including interseismic volcanoes, landslides, aquifers, hydrocarbons, etc.), deformation along the San Andreas Fault (Jolivet et al., 58 NISAR Science Users’ Handbook 59 NISAR Science Users’ Handbook

2015) and will continue to be updated (with separate • A processing configuration file that includes pro- [ ] Figure 6-1 Unwrapped Common Coherence Metadata documentation) and openly released on a regular basis. cessing parameters such as coherence thresholds, NISAR L3 product Interferograms Mask GIAnT is distributed with implementations of SBAS (Be- flags for applying phase corrections, etc., to allow generation

rardino et al., 2002, Doin et al., 2011) as well as Time- for region-specific customization. workflow. Stack Preparation Fun and MInTS (Hetland et al., 2012) techniques. The • Optional: Atmospheric delay metadata layers approach that will be used for the generation of NISAR In the current concept, L2 data will be provided as L3 products is akin to the TimeFun technique (Hetland Timeseries coregistered stacks of unwrapped interferograms. et al., 2012) implemented in GIAnT and allows for an TimeseriesEstimation Hence, no separate coregistration is planned during Estimation explicit inclusion of key basis functions (e.g., Heaviside stack preparation. The output of the stack preparation functions, secular rate, etc.) into the InSAR inversion. step is a self-contained HDF5 product that is handed There may be a small number of pixels where the Weather Optional Tidal off for further processing. Models Corrections Models classification is indeterminate. For example, at the given incidence angle, it is not possible to conclusively TIMESERIES ESTIMATION AND PARAME- classify the data. For those cases, the classification TERIZATION Model Timeseries would be Cx,y=4 or Ck=4. 1 describes the workflow that Dictionary Decomposition will be followed for L3 product generation. The timeseries (i.e., the unfiltered displacement of As shown in Figure 6-1, the L3 product generation each pixel vs. time) is estimated from the processed workflow includes the following consecutive steps: stack using an SBAS or similar approach, and then Visualization parameterized. In practice, GIAnT combines the two STACK PREPARATION steps of SBAS and model-based parameterization. As we expect high-quality orbital control for NISAR, we within GIAnT for implementing weather model-based or regional models such as NARR provide estimates In this initial processing step, all the necessary L2 anticipate that the set of interferograms will typically interferometric phase delay corrections. PyAPS is well of the air temperature, the atmospheric pressure and unwrapped interferogram products are gathered, orga- include all nearest-neighbor (i.e., ~12-day pairs) and documented, maintained and can be freely downloaded the humidity as a function of elevation on a coarse nized and reduced to a common grid for analysis with skip-1 interferograms, so the SBAS step will often be (http://pyaps.googlecode.com; PyAPS is included in resolution latitude/longitude grid. In PyAPS, this 3D dis- GIAnT. For operational NISAR processing, the following somewhat trivial. GIAnT distribution). PyAPS currently includes sup- tribution of atmospheric variables is used to determine information from the L2 products are used in the stack port for ECMWF’s ERA-Interim, NOAA’s NARR (North the atmospheric phase delay on each pixel of each preparation step: OPTIONAL CORRECTIONS American Regional Reanalysis, National Oceanographic interferogram. • Unwrapped interferograms (either in radar or and Atmospheric Administration) and NASA’s MERRA Phase distortions related to solid Earth and ocean tidal ground coordinates) prepared using the software (Modern-Era Retrospective Analysis, Goddard Space For a given GAM dataset, grid points overlapping with effects as well as those due to temporal variations in (Rosen et al., 2012). Flight Center, NASA) weather models. the spatial coverage of the SAR scene are selected. the vertical stratification of the atmosphere can be • Corresponding coherence layers (also generated Atmospheric variables are provided at precise pres- mitigated using the approaches described below. At using ISCE). Following Doin et al. (2009) and Jolivet et al. (2011), sure levels. These values are vertically interpolated to this point, it is expected that these corrections will not tropospheric delay maps are produced from atmo- a regular grid between the surface and a reference • Perpendicular baseline associated with the inter- be needed to validate the mission requirements, but spheric data provided by Global Atmospheric Models. altitude, z_ref, above which the delay is assumed to ferograms. they may be used to produce the highest quality data This method aims to correct differential atmospheric be nearly unchanged with time (~30,000 m). Then, the • A radar simulation file containing the pixels’ products. Typically, these are applied to the estimated delay correlated with the topography in interferometric delay function on each of the selected grid points of elevation. time series product rather than to the individual inter- phase measurements. Global Atmospheric Models the GAM is computed as a function of height. The LOS ferograms since they are a function of the time of each • A file containing radar incidence angles. (hereafter GAMs), such as ERA-Interim (European single path delay (δL)s s (z) at an elevation z is given radar acquisition. LOS • Shadow, layover and land/water mask layers Center for Medium-Range Weather Forecast), MERRA by (Doin et al., 2009, Jolivet et al., 2011): corresponding to the interferograms. Optional atmospheric correction utilizes the PyAPS z 10−6 ⎧k R ref ⎛ ⎛ R ⎞ e e ⎞ ⎫ (Jolivet et al., 2011, Jolivet and Agram, 2012) module S ⎪ 1 d d ⎪ δLLOS (z) = ⎨ P(z) − P zref + k2 − k1 + k3 2 dz⎬ cos(θ) g ( ( )) ∫ ⎜ ⎜ R ⎟ T T ⎟ ⎩⎪ m z ⎝ ⎝ v ⎠ ⎠ ⎭⎪ (6.1-1) 60 NISAR Science Users’ Handbook 61 NISAR Science Users’ Handbook

which includes a constant offset (a), velocity (v), and where θ is the local incidence angle, Rd = 287.05 using the appropriate LOS vector. Equation 6-3 can be approaches are needed to understand the limits of −1 −1 −1 −1 amplitudes (c) and phases ( ) of annual ( ) and J kg K and Rd = 287.05 J kg K are the dry air and j ϕj ω1 solved as a conventional weighted LSQ problem for the performance as completely as possible given existing semiannual ( ) sinusoidal terms. Where needed we water vapor specific gas constants, gm is a weighted ω2 maximum likelihood model, where the L2 norm of the limitations on resources and the distribution of GNSS can include additional complexity, such as coseismic average of the gravity acceleration between z and zref, P weighted misfit is minimized (e.g., Aster et al., 2018): networks. is the dry air partial pressure in Pa, e is the water vapor and postseismic processes parameterized by heaviside T −1 (step) functions H and postseismic functions F (the m d – Gm C d – Gm partial pressure in Pa, and T is the temperature in K. minϕ ( ) = ( ) d ( ) A: GNSS-InSAR direct comparison: Here parameterized −1 −1 latter typically exponential and/or logarithmic). B (t), R, The constants are k1 = 0.776 K Pa , k2 = 0.716 K Pa , ⊥ (6.1-4) time series from InSAR and GNSS are compared, across 3 2 −1 , and Δz are, respectively, the perpendicular com- and k3 = 3.75 ∙ 10 K Pa . θ the length scales described in the L2 requirements. ponent of the interferometric baseline relative to the Here, the data covariance matrix, Cd, is constructed Gradients of the relevant time series parameters (i.e., The absolute atmospheric delay is computed at each first date, slant range distance, incidence angle and using the empirical estimate of correlation from each velocity, v) are calculated between all possible pairs of topography error correction (e.g., Fattahi and Amelung, SAR acquisition date. For a pixel ai at an elevation z at contributing interferogram over the appropriate subset cGPS locations within a validation region, resulting in 2013) for the given pixel. acquisition date i, the four surrounding grid points are of pixels (i.e., masking out water bodies and regions the vectors Δmest,GNSS and Δmest,InSAR. For all these pairs, selected and the delays for their respective elevations that are decorrelated, such as agricultural fields) and unpaired two-sample t-tests (Snedecor and Cochran, This parameterization of ground deformation has a long are computed. The resulting delay at the pixel ai is then superscript T denotes matrix transpose. Only pixels that 1989) are performed to test the null hypothesis that the the bilinear interpolation between the delays at the heritage in geodesy, particularly in analysis of GNSS are coherent in most interferograms are used as input two estimates with their respective errors are from the time series as well as more recently with InSAR data four grid points. Finally, the absolute delay maps of the to the construction of Cd. The solution for this overde- same population. These tests are performed at the 95% InSAR partner images are combined to produce the dif- (e.g., Blewitt, 2007, Hetland et al., 2012, Agram et al., termined minimization problem can be written as confidence level. ferential delay maps used to correct the interferograms. 2013). For validation purposes, we will perform the − g Details and validation of the PyAPS approach are avail- same parameterization on any lowpass-filtered GNSS m est = G d (6.1-5) B: InSAR residual analysis: Using only InSAR data, the able in Doin et al. (2009) and Jolivet et al. (2012). time series used in the analysis, after projecting the residuals w are analyzed, calculated by subtracting GNSS into the InSAR line of sight. where the estimated displacement model mest,InSAR from the ⎡ T –1 –1⎤ T –1 Optional corrections for solid Earth and ocean-tide − −g =⎡ ⎤ observations d, G = ⎣ G⎣ Cd G⎦ ⎦ G Cd G loadings will be done using the SPOTL model (Agnew, Thus, given either an ensemble of interferograms or (6.1-6) −g −gT 2012). To facilitate an accurate representation of ocean the output of SBAS (displacement vs. time), the LSQ C w m = GGm C est,InSAR d G – d (6.1-8) problem can be written as tides, SPOTL provides access to a collection of global The full covariance on the estimated parameters, Cm, and regional ocean models and allows for an easy can be estimated from Empirical structure functions, S , is calculated from the Gm = d w combination of these models. It also includes methods (6.1-3) –g−g –gT−gT residuals w for a subsequent analysis of signal noise CCmm= G CCddGG to convert computed loads into harmonic constants, where G is the design matrix (constructed out of the (6.1-7) as a function of spatial scale. We define the semivario- and to compute the tide in the time domain from these different functional terms in Equation 6-2 evaluated gram S as the variance of the difference between two constants. either at the SAR image dates for SBAS output, or With this formulation, we can obtain GPS and InSAR points separated by distance r, between the dates spanned by each pair for interfer- velocity estimates and their formal uncertainties (in- 2 DECOMPOSITION OF INSAR TIME SERIES ograms), m is the vector of model parameters (the cluding in areas where the expected answer is zero). S ( r ) = E ⎡ f x − f x − r ⎤ ⎣⎢( ( ) ( )) ⎦⎥ INTO BASIS FUNCTIONS coefficients in Equation 6-2) andd is the vector of (6.1-9) observations. For GNSS time series, G, d, and m, are VALIDATION PROCEDURE FOR NISAR SOLID Given a time series of InSAR LOS displacements, the constructed using values evaluated at single epochs EARTH L2 REQUIREMENTS such that the covariance between two points corre- observations for a given pixel, U(t), can be parameter- corresponding to the GNSS solution times, as for SBAS sponds to: ized as: Once displacement parameters are derived from GNSS InSAR input. For comparison with InSAR observations, 2 S (r) U(t) = a + vt (mest,GNSS) and InSAR (mest,InSAR) via equations (6.1- the 3D GNSS time series are projected to the radar LOS Cn (r) = σ − +c cos ω t − φ + c cos ω t − φ 2) through (6.1-7), two complementary approaches 2 1 ( 1 )1 2 ( 2 2 ) (6.1-10) Neq (here referred to as A and B) can be used to validate B (t) 2 + h + f F t − t H t − t + ⊥ Δz + residual where σ is the variance of the noise within the data ∑( j j j ( j )) ( j ) the L2 requirements discussed in this document. Both j=1 Rsinθ set (Williams et al., 1998).

(6.1-2) 62 NISAR Science Users’ Handbook 63 NISAR Science Users’ Handbook

To calculate Cn(r) for a residual (w), w is first detrended Tables and/or figures of comparisons showing LSQ The NISAR mission is designed to measure above commonly described as a homogeneous medium hav- at the scale of the full imaging swath (~240 km) to solutions and error estimates of velocities and offsets ground woody vegetation biomass at a spatial resolu- ing a complex dielectric constant (e) that is a function meet the stationarity assumption inherent to covariance as a function of baseline length from both InSAR and tion of 100 m (1 ha), annually, over the lifetime of the of the volumetric soil moisture, mv, as well as the soil theory. To detrend, a linear plane is fitted and removed GNSS observations. mission. This will provide fine-grain products of carbon texture, temperature, and bulk density; several empiri-

from the data. Subsequently, the structure function Sw is stocks and changes required for understanding and cal models exist for this relationship (Dobson and Ulaby calculated according to (Lohman & Simons, 2005) 6.2 ECOSYSTEMS PRODUCTS– quantifying global carbon cycle. An upper threshold of 1986; Hallikainen et al., 1985; Mironov et al., 2004; BIOMASS 100 Mg/ha is set to reflect the sensitivity of L-band Peplinski et al., 1995). Studies of soil surface scatter- 1 nx ny 2

Sw,dx,dy = ∑ ∑ wk,l − wk−dx=1,l−dy=1 backscatter measurements to biomass and allowing ing and soil moisture remote sensing at L-band have n ( ) The NISAR L2 science requirement for above ground g k=dx l=dy coverage of more than 50% of the global forests and shown that the surface scattering can be expressed biomass (AGB) is expressed as: (6.1-11) the entire area of other woody vegetation (FRA, 2010). in terms of soil dielectric constant at the top 5 cm This sensitivity will allow NISAR to quantify the carbon and the surface roughness characteristics in terms of The NISAR project shall measure above ground where dx=[1:nx], dy=[1:ny] are the sampling intervals stocks and changes of the most dynamic and variable RMS (Root Mean Square) roughness height and spatial woody vegetation biomass annually at the hectare of w in the two geographic directions, nx and ny are the component of global vegetation and to provide signifi- correlation length (Fung et al., 1992). In most SAR-re- scale (1 ha) to an RMS accuracy of 20 Mg/ha for 80% maximum distances covered by the matrix w in x and y, cant contribution to the global carbon cycle and climate lated models for the remote sensing of soil surfaces, of areas of biomass less than 100 Mg/ha. and ng is the number of valid values within the overlap- science (Houghton et al., 2009; Saatchi et al. ,2011; it is assumed that the effect of the spatial correlation

ping region at each shift (dx,dy). ng is not necessarily Harris et al., 2012). is reduced significantly during the SAR azimuthal Above ground biomass is a fundamental parameter equivalent to nx times ny, due to water bodies and other processing and multi-looking, and that the sensitivity of characterizing the spatial distribution of carbon in the regions that are decorrelated in most interferograms. 6.2.1 Theoretical Basis of Algorithm the radar signature to soil surface RMS height variation biosphere and defined as the total mass of living matter remains as the dominant surface structure influencing within a given unit of environmental area. Biomass is of Synthetic Aperture Radar (SAR) backscatter measure- While, in general, noise in w is anisotropic, here we the surface scattering (Oh et al., 1992; Shi et al., 1997; interest for a number of reasons. It is the raw material ments are sensitive to vegetation AGB. SAR obser- neglect this anisotropy and assume that the directional Dubois et al., 1995; Bagdadi et al. 2002; Bryant et al., of food, fiber, and fuelwood. It is important for soil, fire vations from a spaceborne SAR can thus be used for average of Sw versus distance is a good approximation 2007). Other landscape features such as directional and water management. It is also related to the veg- mapping AGB on a global basis. However, the radar of Cn(r). Given Sw, values at scales L=[5,10,20,30,40,50] row or tillage may impact radar cross sections at etation structure, which, in turn, influences biological sensitivity to AGB values changes depending on the km are extracted from Sw and compared to the L2 100-m spatial resolution but are assumed irrelevant in diversity of the planet (Bergen et al., 2009; Saatchi et wavelength and geometry of radar measurements, and requirements at these scales for validation. natural vegetation such as forests and shrublands. al.,, 2007; Frolking et al. 2009). Biomass density (the is influenced by the surface topography, structure of

quantity of biomass per unit area, or Mg dry weight per vegetation, and environmental condition such as soil 6.1.3 Planned Output Products A variety of approaches exist for describing vegeta- ha) is used to determine the amount of carbon released moisture and vegetation phenology or moisture. The tion media, including characterization of vegetation NISAR L3 Solid Earth products will include: to the atmosphere (as CO , CO, and CH through NISAR algorithm will make use of high-resolution and 2 4 structure such as stalks, trunks, and leaves in terms of burning and decay) when ecosystems are disturbed time series backscatter observations at dual-polariza- • Maps of locations where the InSAR and GNSS data canonical cylindrical or disk shapes with specified size and is a strong indicator of the ecosystem function in tions (HH and HV) to estimate AGB by compensating for are being compared and orientation distributions in a set of vegetation lay- terms of carbon sequestration through photosynthesis the effects of environmental changes (soil and vegeta- • LOS displacement vs. time plots, showing: ers, and with dielectric constants similar to live wood and primary production. Above ground carbon densi- tion moisture and phenology) and structure (vegetation of trees and leaf material (Saatchi et al., 1994; Saatchi • InSAR time series using a standard SBAS ty of woody vegetation is approximately 50% of the and surface topography). and McDonald, 1997; Saatchi and Moghaddam, 2000; approach (Berardino et al., 2002, Hooper, biomass with small variations depending on forest type Yueh et al., 1992; Lang et al. 1983; Karam et al. ,1992; 2006) and composition (IPCC, 2006). The current knowledge Radar observations from vegetation have been studied Ulaby et al., 1990). The total L-band backscatter from of the distribution and amount of terrestrial biomass is for more than four decades both theoretically and • The parameterized LSQ solution to the vegetation arises from a combination of scattering and based almost entirely on ground measurements over experimentally (Ulaby et al., 1984; Tsang et al., 1985; InSAR data attenuation of individual canopy components (trunk, an extremely small, and possibly biased sample, with Ulaby and Dobson, 1989; Cloude, 2012). At L-band • The corresponding time series of the LOS branch, and leaf) that can be represented as a sparse almost no measurements in southern hemisphere and frequencies, these studies have shown that the radar component of the GNSS time series scattering medium (Lang, 1981; Chauhan et al., 1994). equatorial regions (Schimel et al, 2015). measurements depend strongly on the structure, This approach requires knowledge of tree structure • The corresponding LSQ solution to the LOS dielectric properties of vegetation components and un- (size, orientation, and density; or equivalently species component of the GNSS time series derlying soil surface (Saatchi et al., 1994; Saatchi and McDonald,1997; Ulaby et al., 1990). The soil is most 64 NISAR Science Users’ Handbook 65 NISAR Science Users’ Handbook

We have expressed these terms in a closed and For the scattering from the rough soil surface Spq(,s), semi-empirical form as: there are several models that can be adopted at the

β ⎡ ⎛ pq ⎞ ⎤ L-band frequency. The semi-empirical model developed α B b 0 pq ⎢ pq ⎥ σ pq-vol = Apqb cosθ 1− exp⎜ − ⎟ by Oh et al. (1992) is used in the NISAR algorithm. This Volume Surface Volume-Surface ⎢ ⎜ cosθ ⎟ ⎥ ⎣ ⎝ ⎠ ⎦ model is derived from a set of radar polarimetric mea-

[ Figure 6-2] 0 0 (6.2-2) surements at multiple frequencies (L-, C-, and X-bands) s , , hh shv Dominant scattering and incidence angles (10°–70°) over rough soil surface β Forest ⎛ pq ⎞ mechanisms of 0 γ Bpqb with a variety of moisture (dielectric constants) and Height pq σ pq-vol-surf = CpqΓ pq (ε,s)b exp⎜ − ⎟ L-band SAR (h) ⎝⎜ cosθ ⎠⎟ roughness (RMS height). The model provides good measurements of agreements with radar backscatter measurements in forest ecosystems (6.2-3) the field at L-band frequency and can be summarized contributing to as: Biomass (b) β 2 2 NISAR dual-pol ⎛ pq ⎞ S = g v cos3θ R + R 0 Bpqb HH H V backscatter σ = S (ε,s)exp⎜ − ⎟ ( ) pq-surf pq ⎜ cosθ ⎟ observations. ⎝ ⎠ g 3 2 2 SVV = cos θ RH + RV v ( ) (6.2-4) SHV = uSVV β ⎛ pq ⎞ 0 Bpqb σ pqwhere-surf = S p q ( ε , s ) eisx pthe⎜ −scattering ⎟from rough soil (6.2-7) Soil Surface ⎝⎜ cosθ ⎠⎟ RMS Height (s) surface and can be represented by the semi-empir- where Soil Dielectric Constant ( ) e ical model of Oh et al. 1992. The surface reflectivity 1.8 g = 0.7 ⎡1− exp −0.65 ks ⎤, ⎢ ( ( )) ⎥ pq(e,s) in the vol-surf backscatter is given by: ⎣ ⎦ u = 0.23 Γ ⎡1− exp −ks ⎤, 0 ⎣ ( )⎦ * 2 2 2 Γ = R (ε)R (ε) exp −k s cosθ 1 pq p q ( ) 2 3Γ and biome), dielectric constant, and ground charac- These are: 1.) volume (vol) scattering, 2.) volume and ⎛ 2θ ⎞ 0 1− ε = − − Γ = v 1 ⎜ ⎟ exp( ks), and 0 teristics (RMS height, correlation length, and dielectric surface interaction (vol-surf) and 3.) surface scattering (6.2-5) ⎝ π ⎠ and 1+ ε constant). (surf) where R () is the Fresnel reflection coefficient of is the Fresnel reflectivity of the surface at nadir. 0 0 0 0 p Simpler approaches only use the vegetation water con- σ pq = σ pq-vol +σ pq-vol-surf +σ pq-surf semi-infinite soil medium at polarization p with the tent (VWC) to provide analytical forms for attenuation (6.2-1) dielectric constant of  and exp (-k2s2 cos q2) rep- We use the above model because it provides simple and scattering effects. The most common model used resents the Kirchhoff’s damping factor associated with expression as a function of soil dielectric constant and in microwave frequencies is the water cloud model that where p and q denote polarization of transmitted and the RMS height (s) of the surface (Fung et al., 1981), k the surface RMS height. Other rough surface scatter- includes two scattering components from vegetation received radar signals, respectively. These can be is the wavenumber, q is the local incidence angle, and ing models can also be used. Some examples are the volume and its underlying ground but ignores the either vertical (v) or horizontal (h) in a linear polarization b is the above ground biomass density in the unit of Integral Equation Method (IEM) model, small perturba- volume-ground interaction (Attema and Ulaby, 1978). radar system. The three dominant scattering terms are Mg ha-1. The Fresnel reflection coefficients in terms of tion method and Kirchhoff approximation. These models This model is mainly applicable for higher frequency derived from basic electromagnetic theory by solving complex dielectric constant () are: have been compared and tested over study sites with (C-band and above) characterization of the vegetation Maxwell’s equations in a discrete random media (Saat- detail ground measurements to suggest that 1) The cosθ − ε − sin2 θ backscatter (Matzler, 1994; Ulaby and El-rayes, 1987). chi and Lang, 1989; Lang, 1981; Tsang et al., 1985; R = contribution from rough surface scattering is compar- H 2 In this work, and for the fuller scattering model, the Saatchi and McDonald, 1997; Chahan et al., 1991). cosθ + ε − sin θ atively smaller than the volume and volume-surface 2 backscattering coefficient is expressed as the com- ε cosθ − ε − sin θ contributions, particularly in the forested environments. RV = bination of three scattering components (Figure 6-2). ε cosθ + ε − sin2 θ Therefore, the residual effects of the uncertainty of (6.2-6) surface scattering characterization is small. 2) the Oh et al. (1992) model is preferred over other models because of its simplicity (based only on two parameters) 66 NISAR Science Users’ Handbook 67 NISAR Science Users’ Handbook

and its direct link to backscattering coefficients of the branches, and leaves and may vary depending on the –12 –12 10.00

soil dielectric constant instead of soil moisture. 3) Other vegetation type. The volume surface interaction term –14 8.75 gpq –14 models such as the small perturbation method has Cpqb represent the strength of the specular reflection –16 7.50 no cross polarized (HV) term and underestimates the and include the scattering from both trunk and crown –16 measurements of radar backscatter over bare layers reflected from ground surface. Similarly, the co- –18 6.25 soil surfaces. efficients A B C also depend on the forest structure pq pq pq –18 –20 5.00 and the SAR backscatter radiometric calibration (e.g., 2 –22 3.75 This model above is characterized by a set of coeffi- terrain correction) but are independent of aboveground –20 R = 0.71

Mbam Djarem, Cameroon (dB) HV BACKSCATTER cients (A B C and a b d ) that depend on the biomass. The model therefore, has three unknown –24 2.50 (dB) (H, V) ATTENUATION pq pq pq pq pq pq Barro Colorado, Panama ALOS L-HV BACKSCATTER (dB) ALOS L-HV BACKSCATTER –22 HV backscatter without attenuation polarization of the observation but are independent biophysical variables (b, , s), and six polarization Pearl River, Louisiana –26 HV backscatter with attenuation 1.25 La Selva, Costa Rica Attenuation at horizontal polarization of the vegetation aboveground biomass (b), and soil dependent coefficients (Apq Bpq Cpq and apq bqpdpq) that Attenuation at vertical polarization –24 –28 0.00 dielectric constant () and surface roughness (s). These must be determined for different forest types. 0 100 200 300 400 500 50 100 150 200 250 300 350 400 450 coefficients represent weighting factors for scattering ABOVEGROUND BIOMASS (Mg/ha) BIOMASS (t.ha–1) and attenuation of vegetation through its various com- The overall sensitivity of the model at the L-band fre- ponents (trunks, branches, leaves) that depend on their quency is shown in terms of the biomass by using data orientation and configurations (arrangements) within from SAR measurements from the ALOS PALSAR satel- [ Figure 6-3] Based on the empirical/theoretical experience outlined 3. The model has a simple analytical formulation the forest canopy. The semi-empirical model separates lite and model simulations (Figure 6-3). The sensitivity Sensitivity of L-band above, NISAR will generate biomass estimates of allowing sensitivity analysis and error propagation the ground and vegetation parameters. The vegetation of backscatter measurements to AGB depends on the HV backscatter to woody vegetation up to 100 Mg/ha using high-resolu- (Hensley et al., 2010). parameters are all combined into aboveground biomass wavelength, with longer wavelengths allowing better vegetation biomass. tion multi-temporal NISAR L-band SAR backscatter im- (b) and ground parameters represented by surface penetration of the microwave signal into the canopy Both ALOS PALSAR agery, and the above semi-empirical algorithmic model All vegetation and biome (i.e., coniferous, deciduous, dielectric constant (soil moisture) and roughness. and scattering from the tree trunks that contain most of satellite L-band obser- (Saatchi and Moghaddam, 2000; Hensley et al., 2014). mixed, tropical evergreen, and shrubland savanna as the tree biomass. At shorter wavelengths, the attenua- vations (left) and model The target area of the NISAR biomass product will be shown in Figure 6-5) specific structural and calibration As discussed earlier, the algorithm model coefficients tion of the signal limits the penetration and reduces the simulations (right) show all forests and shrublands across different ecoregions, coefficients of the model will be derived for the NISAR apq, bqp, and dpq are considered the allometric or effect of the scattering from tree components causing the effect of vegetation distributed globally (Figure 6-4). Even in regions where mission: (See Fig. 6-4). structure-related parameters and depend only the a loss of sensitivity to biomass at some threshold AGB. attenuation on the radar forest biomass is larger than 100 Mg/ha, there are orientation or arrangement of scatterers in the vegeta- At the NISAR L-band frequency (~24-cm wavelength), saturation level. significant areas with degraded or naturally heteroge- To use the semi-empirical model as an algorithm to tion but are independent of biomass. Similarly, Apq, Bpq the biomass sensitivity threshold also depends on the neous vegetation that the biomass may remain below estimate the forest or vegetation AGB requires a priori and Cpq are considered the radiometric coefficients of vegetation structure (configuration and size of scatter- the NISAR sensitivity limit. The low biomass regions are quantification of the model coefficients for different the algorithm that depend on the radiometric correction ing elements), the dielectric constant (water content in considered among the most dynamic regions due to forest types and the number of observations to account of radar due to the terrain correction and heterogeneity the vegetation components), soil moisture, topography various management and human land use activities, or for the soil moisture () and surface roughness (s) of vegetation structure. and surface roughness. It has been established that frequency of natural disturbance such as drought, fire, variations. To meet this challenge, the model must be the upper limit of L-band radar sensitivity to biomass and storms. apq bpq developed through a process of calibration and valida- In the volume term, Apqb and Bpqb control the rela- is approximately 100 Mg/ha (Mitchard et al., 2009; tion (CAL/VAL) approach over different forest types or tionship between biomass and the backscatter power Robinson et al., 2013; Saatchi et al., 2007; Saatchi et The semi-empirical algorithm has several advantages ecoregions before the launch of the NISAR. The model of and the attenuation respectively. These terms are al., 2011; Mermoz et al., 2015). In regions with forest over fully empirical regression models. These advan- coefficients are quantified over a series of study sites represented in the form of a power-law derived from biomass > 100 Mg/ha, it is considered best to use oth- tages are: (CAL/VAL sites) that includes ground measurements of a series of allometric models combining size, growth er sensors such as ESA’s P-band SAR mission named vegetation structure, and airborne or satellite L-band rate, and their metabolic characteristics (Sarabandi and BIOMASS (Le Toan et al., 2011), and/or a combination 1. The model is physically based and captures the observations that can simulate the NISAR observations Lin, 2000; Enquist et al., 2009; Smith and Heath, 2002). of SAR interferometry and backscatter at L-band, and behavior of radar measurements over complex vegetation structures. (see Section 8.0 for ecosystem CAL/VAL plan). The model parameters apq and bpq are independent of lidar sensors, such as will be available from NASA’s vegetation biomass and depend on geometry of tree GEDI mission (Saatchi et al., 2011; Shugart et al., 2011; 2. The model includes surface moisture variables as canopies in terms of size and orientation of trunks, Hall et al., 2011). the key variable impacting the temporal observations of radar backscatter, and 68 NISAR Science Users’ Handbook 69 NISAR Science Users’ Handbook

[ Figure 6-4]

Global distribution of above ground biomass. Map is stratified in categories to demonstrate areas in green and yellow where NISAR above ground biomass products will be of low uncertainty Regions with Regions with Regions with Regions with Open AGB <100 Mg/Ha AGB >100 Mg/Ha AGB <20 Mg/Ha No Woody Water 50% of Area 50% of Area 50% of Area Vegetation

6.2.2 Implementation Approach for the exponents of the model fits and are used as

Algorithm initial conditions in retrieving these coefficients 1. TROPICAL/SUBTROPICAL BIOMES 2. TEMPERATE BIOMES 4. POLAR/MONTANE BIOMES over CAL/VAL study sites in step 2. During the pre-launch CAL/VAL activities, the science Flooded Grasslands and Broadleaf and Mixed Forests Boreal Forests/Taiga team determines the initialization of the algorithm and Savannas 2. It is assumed that the structural parameters of the evaluates its performance to meet the science require- Coniferous Forests Montane Grasslands and algorithm will remained fixed and will not change ments. The algorithm depends on a number of model Coniferous Forests Shrublands spatially or temporally within each ecoregion coefficients that are expected to vary as a function of Grasslands, Savannas and and during the NISAR time series observation. To Dry Broadleaf Forests Shrublands Rock and Ice biome and be subjected to a natural variability of ob- determine  ,  , and  for different ecoregions served radar backscatter with changes in soil moisture pq pq pq and for two polarizations of HH and HV, the Grasslands, Savannas, Tundra and season. The following describes the approach that 3. DRY BIOMES CAL/VAL study sites within each ecoregion will be and Shrublands is used for determining these model parameters. used. At these sites, multi-temporal (3-5 images Deserts and Xeric Moist Broadleaf Forests Shrublands capturing seasonal variations) radar backscatter 5. AQUATIC BIOMES QUANTIFICATION OF MODEL COEFFI- measurements, ground vegetation biomass, soil CIENTS  ,  , AND  Mediterranean Forests, Lakes pq pq pq moisture and surface roughness measurements Woodlands and Scrub are available or estimated from radar measure- The model coefficients related to vegetation structure Mangroves ments directly over bare surfaces within the study can be determined in two steps: site. The coefficients are determined using the [ Figure 6-5] 1. A three-dimensional forward scattering mod- Levenberg-Marquardt Approach (LMA) for non- Distribution of global ecoregions and biomes for the development of the vegetation biomass algorithm. el (Saatchi and McDonald, 1997; Saatchi and linear least square estimation (Marquardt, 2009). The ecoregions are derived from a combination of climate, topography, soil and vegetation data (Olson Moghaddam, 2000) has been used over the key The method is used in many software applications et al., 2001). The focus of the CAL/VAL plan and algorithm development would be on biomes that have CAL/VAL study sites with ground measurements for solving generic curve-fitting problems and has distinct differences in the model. of tree structure to fit a power law function to the already been applied in several SAR estimation scattering and attenuation terms of the scattering approaches (Troung-Loi et al., 2015). model to vegetation biomass. The coefficients are 70 NISAR Science Users’ Handbook 71 NISAR Science Users’ Handbook

3. If the soil moisture and roughness data are not 1. Coefficients Apq, Bpq, Cpq are assumed to vary tem- POST-LAUNCH CALIBRATION OF MODEL surements. The algorithm, shown in Figure 6-6, uses available from ground measurements in the CAL/ porally due to changes in vegetation water content COEFFICIENTS a Bayesian approach to estimate AGB. The estimation VAL study area, these variables are estimated and phenology. This assumption can be verified approach enables the use of multi-temporal back- The post-launch calibration is mainly focused on from areas of low vegetation or bare fields within over different ecoregions to relax the temporal scatter measurements to quantify all variables while assessing the assumption of spatial heterogeneity as the study area or the SAR image scene. A crude variations to monthly or seasonal. accounting for measurement uncertainty. observed by NISAR large incidence angle variations and low vegetation or non-forest mask is generated for 2. The radiometric coefficients are assumed to therefore larger topographical variations. the time series data stack. This mask is obtained remain constant spatially within a local moving The implementation includes the following steps: by thresholding the HV SAR image scene available [ Figure 6-6] window (3×3 or larger) to allow for spatial stability APPLICATION OF THE BIOMASS 1. The time series (t1, t2, … tn) of radiometric terrain over the CAL/VAL site. A threshold of -13 dB has Flow chart showing of the algorithm. This assumption depends on the ALGORITHM TO THE NISAR TIME-SERIES corrected (RTC) HH and HV polarized images are been used to generate such a forest mask on the implementation spatial heterogeneity of vegetation structure (e.g., IMAGE STACK fed into the algorithm as they become available ALOS-1 and -2 data sets by JAXA. By assuming of the NISAR canopy gaps) that influences the magnitude of from the NISAR processor. NISAR dual-pol obser- 0 > forest_threshold, the non-forest or low-veg- algorithm for AGB The AGB (b), soil dielectric constant () and roughness s hv volume and volume-surface interactions. 0 0 vations are collected for NISAR ecosystem science etation areas are separated. A similar approach is estimation globally. (s) are estimated from dual-pol (s H, and s HV) mea- 3. The coefficients can be determined over CAL/ used in the NISAR algorithm for disturbance and VAL study sites where biomass, soil moisture and will be the same for both algorithms. Once the roughness are available or determined as dis- mask is developed, the soil dielectric constant and cussed in 6.2.1.2 to allow for testing the validity Time Series NISAR Observations RMS height of the surface roughness is deter- of moving window size for each vegetation type t1 mined by inverting the Oh et al., (1992) model t2 or ecoregion. Using a minimum of 3×3 moving described above in equation 6.2-7. These values . Radiometric NISAR window will allow the algorithm to have different . Terrain Validation are used as the initial condition of the estimation . Calibration Data coefficients for each local area. The alternative of the structural variables for all areas considered . approach is to use A , B , C derived over the . forest or vegetation using a nearest neighbor pq pq pq tn CAL/VAL sites within each ecoregion as the fixed AGB (Mg/ha) interpolation approach (Troung-Loi et al., 2015). coefficients for the entire ecoregion as shown in Time Series NISAR Table 6.2 for the five dominant ecoregions globally. Multi-Look (s , s ) 4. For cases where pq, pq and pq cannot be esti- hh Hv Observations (100m) mated unambiguously using the LMA curve-fitting PRE-LAUNCH CALIBRATION OF MODEL or estimation approach, the theoretical values SRTM Topography derived from the forward model simulations and COEFFICIENTS power-law model fits will be used. Estimates of The pre-launch calibration of the algorithm model Bayesian Uncertainty AGB Uncertainty coefficients related to vegetation structure will Estimator Analysis will apply to the structural coefficients, pq, pq and (Mg/ha) also include uncertainty associated with the LMA pq, that remain constant for each ecoregion globally least-squared approach. The uncertainty can be throughout the NISAR mission. Using ALOS PALSAR or used within a Bayesian approach to account for UAVSAR data that simulates the NISAR observations uncertainty in the algorithm and estimation of the Algorithm can be used to estimate these coefficients. The require- Initialization biomass. ment for pre-launch calibration is the selection of the study sites that represent the variability in structure of QUANTIFICATION OF MODEL COEFFI- the dominant vegetation types.

CIENTS Apq, Bpq, Cpq

The radiometric coefficients of the algorithm can be determined simultaneously with those related to vegetation structure. The estimation of these coefficients is Global Terrestrial Algorithm CAL/VAL Ecoregion Map Data based on the following assumptions: 72 NISAR Science Users’ Handbook 73 NISAR Science Users’ Handbook

in the background land mode every 12 days on 8. When the NISAR image t2 becomes available, 16. Annually, the updated biomass values will be values at the pixel level. The detection of disturbance descending orbits and every other on ascending repeat step 2 to develop a new forest/non-forest reported as a map with 100 m × 100 m (1 ha) can be the simple band threshold as determined for the

orbits. This is approximately 45 observations per mask. Compare the mask derived from t1 with the spatial grid cells globally. However, the algorithm forest/non-forest mask or the use of disturbance algo-

year. mask from t2. Develop a new mask to update the will provide estimates of all three variables, and rithm. Here, a similar approach as in the disturbance forest/non-forest mask by adding the non-forest the uncertainty, every time the RTC NISAR images algorithm will be implemented to report the vegetation 2. Use t1 data at HV polarization and a simple threshold to develop a mask of forest/non-forest pixels. become available throughout the year. biomass before the disturbance and detection the over the entire NISAR image scene. This threshold 9. Repeat steps 3 and 4 for all new non-forest pixels. 17. The algorithm assumes that during a one-year post-disturbance accumulation of the biomass.

is initially set to -13 dB (as derived from ALOS 10. Update (b ,  , s ) maps with the all new forest/ period of multi-temporal observations, the soil 1 1 1 6.2.3 Planned Output Products PALSAR data). It will be adjusted, as necessary, non-forest pixels. dielectric constant will vary, while AGB and rough- when NISAR data become available. ness, s, are treated as constant except in cases of The L2 above ground biomass product is a raster im- 11. Repeat steps 5 and 6 by using the updated values 3. Use the Oh et al. (1992) model to estimate soil land cover change (e.g., deforestation or distur- age at 100-m spatial resolution produced over the CAL/ for (b1, 1, s1) as the new a priori information in the dielectric constant and roughness for all bance; see forest disturbance product description). VAL sites. The raster product is in one-byte format with NISAR retrieval algorithm and produce (b2, 2, s2) pixel values representing AGB as an integer number non-forest pixels identified by the mask. Use the and the uncertainty. 18. The algorithm performance using the Bayesian estimates of ( ,s ) as the initial conditions for approach is evaluated at 100-m spatial resolution from 0 to 100 Mg/ha, and a fixed value for biomass 0 0 12. If no more NISAR imagery is available, iterate all pixels in the NISAR image by using a nearest products for all areas with AGB > 0. All areas with greater than 100 Mg/ha. The product will be generated steps 9 and 10 by using the average of b and b neighbor interpolation of a simple kriging ap- 1 2 AGB > 100 Mg/ha will be identified and aggregat- every year using observations collected during the year. as the new a priori information for b and the av- proach. The interpolation provides initial conditions ed into one class, and areas with AGB = 0 will be The input product is multi-look L2, 25 m, radiometrical- erage of s and s as the new a priori information and bounds for all pixels with forests or vegetation 1 2 another class. ly terrain-corrected imagery. Also required for generat- for s. This step is designed to make sure that the cover that will be used in the NISAR biomass ing the biomass products are ancillary data of a global biomass and roughness remain constant for the retrieval algorithm. land ecoregion map to select the algorithm coefficients, NISAR observations, while soil dielectric constant By following this method, the algorithmic model is used surface digital elevation model to improve the inversion 4. Use a simple model based on HH and HV polar- is updated. The iteration will continue to provide to estimate AGB from radar backscatter observations. model with local incidence angle, a soil moisture map ization, derived from ALOS PALSAR data, and ad- stable values of (b ,  , s ) and improved estimates The effects of other variables associated with soil 2 2 2 (derived from SMAP or SMOS) and in situ and Lidar justed with local incidence angle (Yu and Saatchi, of the uncertainty. moisture and surface roughness on the radar back- data for calibration and validation of the model. The 2016) to estimate forest biomass (b0) for all forest scatter measurements are also taken into account. 13. If more NISAR imagery is available, repeat steps Bayesian methodology will also provide uncertainty and non-forest pixels in the NISAR image scene. Currently, the algorithm is calibrated over different 7 to 11. estimates at the pixel level. Initial values for surface Use b0 and its distribution as the initial condition ecoregions using the CAL/VAL data. Locations of forests roughness, s, are obtained for the CAL/VAL sites during and the bounds for the bioboth for the retrieval 14. If a disturbance has been detected using the used to derive parameters used by the algorithm, as pre-launch activities and determined post-launch by algorithm. disturbance algorithm during the time series developed over five key forest and woodland biomes is analysis, reset the all three variables for the pixel the closest CAL/VAL site within the same ecoregion. 5. Include the NISAR HH, HV as the measurements given in Table 6-1. Parameters are given in Table 6-2. by repeating steps 2 to 4. data and the initial conditions (b ,  , s ), and A general flowchart, describing the current algorithm 0 0 0 6.3 ECOSYSTEMS PRODUCTS- implementation is shown in Figure 6-6. The number the joint probability distributions as the a priori 15. Forest biomass growth can be detected during DISTURBANCE information in the NISAR Bayesian-based retrieval the algorithm retrieval from the time series NISAR of ecoregions that require separate algorithms will be algorithm. data if a significant trend is observed in biomass finalized later after performing the algorithm CAL/VAL The NISAR L2 science requirement for above ground estimation after implementing step 10. The time activities across the global ecoregions. biomass (AGB) is expressed as: The NISAR project 6. For ecosystems with a strong phenological series estimates of the biomass, b, can be used to shall measure global areas of vegetation disturbance signature in the L-band radar cross section, the study or report the trend in biomass from the first IDENTIFICATION OF BIOMASS at 1 hectare resolution annually for areas losing at algorithm uses a global land cover or ecoregion DISTURBANCE NISAR imagery. least 50% canopy cover with a classification accura- map to set the geographically and temporally ap- cy of 80%. propriate coefficients for the inversion algorithm. The implementation of the biomass algorithm using the time series stack of dual-pol NISAR imagery requires 7. The algorithm will provide the first estimates of the detection of disturbance to reset the biomass the biophysical variables (b1, 1, s1) from the first NISAR image along with the uncertainty of the estimates. 74 NISAR Science Users’ Handbook 75 NISAR Science Users’ Handbook

TABLE 6-1. STUDY SITES USED TO DEVELOP REPRESENTATIVE MODELS FOR GLOBAL Accurate annual measurements of the global area of L-band calibrated backscatter measurements, foremost ESTIMATION OF BIOMASS forest disturbance from NISAR will be a significant con- using cross-polarized observations (L-HV). At its core, FOREST RADAR LOCATION DATE IN SITU REFERENCE tribution to the global accounting of carbon emissions the algorithm is comparing backscatter from a set TYPE OBSERVATION DATA from land cover change (van der Werf et al., 2009). of two time-series of equal observation length from

Needleleaf AIRSAR Boreal Forest 1993–1996 18 Sites,64 Saatchi and The NISAR disturbance product complements the subsequent years. Annual observation time series may of Canada plots Moghaddam, NISAR biomass, inundation, and active agricultural area be temporally segmented (e.g., freeze/thaw, wet/dry 2000 products to jointly improve our understanding of carbon season observations only), determined spatially based Broadleaf UAVSAR Howland 2009–2010 32 Robinson et emissions from land cover change and the success on the observational data to account for ecosystem Deciduous Forest, Maine, 1-ha plots al. 2013 of mitigation strategies like avoided deforestation. As specific seasonality (see Figure 6-5). A simple, yet USA Lidar data such, the disturbance product from NISAR will consti- robust approach for detecting disturbance in these time Mixed broad- AIRSAR/ Maine, Duke, 2004/2009 78 plots/Lidar Robinson et tute an invaluable contribution to the accounting needs series of backscatter images is based on change point leaf/Needle UAVSAR Harvard, etc. data al. 2013 Leaf for the United Nations negotiated policy mechanism detection with cumulative sums analysis which have for Reducing Emissions from Deforestation and Forest been employed in many sectors such as statistical Broadleaf AIRSAR/ Sites distrib- 2004–2015 Combined Saatchi et al. Degradation (REDD) as well as emerging bi- and multi- control, financial trends and meteorological analysis. Evergreen UAVSAR/ uted in Costa plots/Lidar 2011 ALOS Rica, Peru, data lateral carbon treaties involving forest carbon account- During the NISAR mission, time series-based cumu- PALSAR Gabon ing (Romijn et al., 2012; Baker et al., 2010). To date, lative sums are calculated for each 25 m pixel, either

Savanna/ ALOS/ Uganda/ 2007–Present 160 plots Mitchard et forest carbon accounting with remote sensing methods from the full year observation period, or from season- Dry Forest PALSAR Cameroon/ 0.4–1.0 ha al. 2009 has made significant progress and international efforts ally segmented subsets, which is adequate in complex Mozambique/ by the Committee on Earth Observing Systems (CEOS) Gabon biomes like the boreal region with strong seasonality. and the GEO Global Forest Observing Initiative (GFOI) strive to improve national efforts on Forest carbon The corresponding cumulative sum curves from a full TABLE 6-2. MODEL PARAMETERS DERIVED FOR DIFFERENT VEGETATION TYPES DURING THE Monitoring, Reporting and Verification (MRV, GFOI, year of data to initialize the algorithm. Subsequent NISAR PHASE A STUDY OVER EXISTING CAL/VAL SITES 2013). As cloud-cover is a major impediment to reliably observations will be classified based on threshold cri-

MODEL BROADLEAF BROADLEAF NEEDLELEAF MIXED DRY FOREST acquire high-spatial resolution annual data sets with teria to identify timing of change detected. Backscatter PARAMETERS EVERGREEN DECIDUOUS BROADLEAF WOODLAND optical sensors over tropical, sub-tropical and boreal means before and after the detected change points are & NEEDLELEAF SAVANNA regions, SAR observations have emerged as a critical calculated. Thresholds are established from backscat- complement (Walker et al., 2010; Kellndorfer et al., ter-canopy density curves which are established after AHH 0.229 0.241 0.189 0.211 0.11 2014; Reiche et al., 2016). Achieving reliable and timely year one of the mission. Backscatter change for 50% A HV 0.0867 0.0683 0.013 0.0365 0.03 accounting of forest disturbance globally at consistent canopy density change can thus be determined. The annual or even sub-annual intervals will improve our backscatter-canopy density curves are generated on BHH 0.0108 0.0944 0.00211 0.0789 0.00908 scientific understanding of global carbon emission a per-scene basis based on ancillary canopy density B 0.012 HV 0.0148 0.0165 0.00195 0.0855 dynamics from forests, both from natural and anthropo- observations from global MODIS or Landsat products genic disturbances. and can further be stratified by unsupervised clustering CHH 0.005 0.008 0.0076 0.0083 0.009 of the NISAR time series data set. With this approach, C 0.002 0.0062 0.0047 0.0053 0.007 HV 6.3.1 Theoretical Basis of Algorithm ecosystem specific variations on backscatter-canopy

a HH 1.1 1.1 0.19 0.96 0.20 density relationships are accounted for. If within a 100 The NISAR L2 science requirement for forest distur- × 100 m resolution cell, 8 or more pixels are flagged as

a HV 0.2 0.3 0.11 0.27 0.18 bance is expressed as: The NISAR project shall mea- disturbed, either from the entire time series or seasonal sure global areas of vegetation disturbance at 1 hectare b 1.1 1.1 0.89 0.96 1.0 subsets of the time series, the entire cell is flagged as HH resolution annually for areas losing at least 50% disturbed. 1.1 1.0 0.9 0.89 1.0 bHV canopy cover with a classification accuracy of 80%.

1.1 1.1 0.89 0.96 1.3 bHH The NISAR measurement metric for disturbance deter- The NISAR disturbance detection algorithm is based 0.5 0.9 0.23 0.27 1.1 mination relies on the measurement of cross-polarized bHV on time series analysis techniques of observed NISAR 76 NISAR Science Users’ Handbook 77 NISAR Science Users’ Handbook

L-band backscatter change with forest fractional which is an “infinite resolution” model borrowed from optical techniques, where the contribution of the ground sur-

canopy cover loss of 50% or more as observed and face °ground is combined with the average return from a layer of vegetation, °veg, weighted by the fraction of vegeta- compared over annual timeframes. At its core, L-band tion canopy cover, h. In the above, the two-way loss of signal energy as it passes through the canopy is accounted cross-polarized backscatter exhibits a significant vari- for by a, the extinction, and a vegetation height (h) estimate. a is normally given in units of dB/m. ation (several dB), depending on initial state of canopy density and forest structure, when forest fractional The above equation can be rearranged so as to separate the ground and the vegetation scattering returns, as in canopy cover is reduced by 50% or more. σ ! = σ ! ⎡1−η 1− e−αh ⎤ +σ ! ⎡η 1− e−αh ⎤ forest ground ⎣ ( )⎦ veg ⎣ ( )⎦ In order to provide a more theoretical foundation for (6.3-3) the use of time series analysis of backscatter change based on the target scattering physics, a theoretical When multiple observations are made, (6.3-1) through (6.3-3) can be combined to relate the vector of observations scattering model has been developed and described to the spatially varying values and the set of constants that describe the mean radar cross section of the ground (Cartus et al., 2018). This model includes the scattering and vegetation, as in model and an observational error model, in order to show the separation between simulated natural and ⎡ ⎤ ⎡ −αh −αh ⎤ σ ! 1−η 1− e η 1− e disturbed forest canopies. A summary of this simple ⎢ obs-1 ⎥ ⎢ 1 ( ) 1 ( ) ⎥ ⎡ ! ⎤ observational model tailored for disturbance (i.e., ignor- ⎢ ! ⎥ ⎢ −αh −αh ⎥ σ σ 1−η2 1− e η2 1− e ⎢ ground ⎥ ⎢ obs-2 ⎥ = ⎢ ( ) ( ) ⎥ ing double bounce) using cross-polarized observations ⎢ ⎥ ⎢ ⎥ " ⎢ ! ! ⎥⎢ σ ! ⎥ is given here. ⎢ ⎥ ⎢ ⎥⎣ veg ⎦ ⎢ ! ⎥ −αh −αh σ ⎢ 1−ηN 1− e ηN 1− e ⎥ obs-N ⎣ ( ) ( ) ⎦ ⎣⎢ ⎦⎥ (6.3-4) In a relationship between radar observation and classification accuracy, an error model is needed for the which, for a given number of observations, N, can be and scene stratification from image unsupervised clus- observations and those components that contribute to inverted to estimate the RCS of the ground and vege- tering and are employed to mask non-forest areas such the target radar cross section (RCS). The observational tation returns. Through simulations with real ALOS-1 as wetlands and agricultural areas, although masks for error model that relates the observed radar cross-sec- ⎡ ⎤ ⎡ −αh −αh ⎤ σ ! L-band1−η measurements1− e withη estimates1− e for h and h from these areas will be available from the NISAR Level-2 tion, for each polarization pq, written here as ° for ⎢ obs-1 ⎥ ⎢ 1 ( ) 1 ( ) ⎥ obs ancillary data sources the validity of backscatter-based⎡ ⎤ wetlands and agriculture products. The time-series RTC ⎢ ⎥ ⎢ −αh −αh ⎥ σ ! simplicity, to the observation error sources, °obs-error and ! −η − η − ⎢ ground ⎥ ⎢ σ ⎥ change1 2detection(1 e of )50% canopy2 (1 densitye ) loss was products are subjected to multi-temporal speckle noise obs-2 = ⎢ ⎥⎢ ⎥ the radar cross-section of a forest canopy, °forest is ⎢ ⎥ " ⎢ demonstrated! for the project in a! memorandum⎥⎢ byσ ! ⎥ reduction according to Quegan et al., 2001. A diagram ⎢ ⎥ ⎢ ⎥⎣ veg ⎦ ⎢ ! ⎥ Siqueira−η and− others−αh in 2014.η Time− series−αh analysis al- giving the processing flow for the disturbance algo- ! ! ! σ ⎢ 1 N (1 e ) N (1 e ) ⎥ σ obs = σ obs-error +σ forest . ⎢ obs-N ⎥ ⎣ lows for the minimization of error sources ⎦from soil and rithm is shown in Figure 6-7. (6.3-1) ⎣ ⎦ vegetation moisture as well as speckle noise variations.

Mark Prytherch/Shutterstock SEASONAL SUB-SETTING OF TIME SERIES The observational errors consist of instrumental effects, 6.3.2 Implementation Approach for DATA STACK such as calibration and quantization errors, obser- THE NISAR PROJECT SHALL MEASURE GLOBAL AREAS OF Algorithm vational ambiguities, and speckle noise. With these For many biomes, seasonal stratification of time VEGETATION DISTURBANCE AT 1 HECTARE RESOLUTION factors taken into account, the radar cross-section of Pre-requisite for the disturbance detection algorithm is series will improve detection of disturbance events, the forest can be written as fully calibrated, radiometrically terrain corrected (RTC) e.g., where freeze/thaw or dry/wet season conditions ANNUALLY FOR AREAS LOSING AT LEAST 50% CANOPY backscatter time series where pq=HV. When pq=HH ! ! ⎡ ! −αh ! −αh ⎤ σ = 1 − η σ + η σ e + σ 1 − e data are useful for resolution of potential ambiguities COVER WITH A CLASSIFICATION ACCURACY OF 80%. forest ( ) ground ⎣⎢ ground veg ( )⎦⎥

(6.3-2) 78 NISAR Science Users’ Handbook 79 NISAR Science Users’ Handbook

introduce significant backscatter changes. Thus, the Both the forest mask and change threshold can be Nisar Subsetting By Relative Calibration Time Series Seasonality Within Stack first step in the disturbance detection algorithm is the estimated per ecosystem from statistical analysis with sub-setting of time-series data stacks and selection canopy density masks. During the NISAR mission we of scenes to minimize gross environmental effects will generate a lookup table for biomes and ecoregions on backscatter levels. Selection of the scenes can be for these thresholds. performed with a global scene means comparison and threshold approach as follows: RELATIVE CALIBRATION OF SUBSETTED DATA STACK 1. A crude forest/non-forest mask is generated for a Mean Scene Backscatter time series data stack. This mask is obtained from For improved results, the time-series stacks are cali- Mean Scene Backscatter JFMAMJJASOND JFMAMJJASOND JFMAMJJASOND JFMAMJJASOND ancillary existing land cover classifications (e.g., brated relative to each other to a higher precision than Year 1 Year 2 Year 1 Year 2 from MODIS, Landsat, ALOS-1), or by thresholding perhaps required through routine standard calibration

an early HV SAR image from typical seasons of of the NISAR imagery. This calibration step examines Unfrozen interest (e.g., non-frozen, dry season). A threshold distributed targets that are expected to be unchanged (e.g., Boreal Biome)

of -13 dB has been used to generate such a forest or minimally changed in brightness over a set time Change mask on ALOS-1 and 2 data sets by JAXA: span of images. With NISAR’s 240 km swath width, it is Point Analysis reasonably assumed that a statistically large area, Ani, 2. For all pixels ti under the mask, the mean (on the power scaled data) at each time step i = 1, n is will not be disturbed (or otherwise changing) during any of the observations in the subsetted time-series obser- Disturbance Mean Before Mean After generated to produce a time series of means as: Mapped Change Point Change Point vations. These areas will be identified partly through 3. t ={t , t , t ,…, t } mean 1 2 3 n use of the threshold-based forest mask from one scene (B-A) >50%? 4. Here, t mean is a collection of mean values, where and applied again through all images. } ti indicates the mean pixel value for the forested pixels in the image. This is a large-scale assess- The calibration correction for image n, fn, for each ment of the seasonal effects within the image. polarization channel pq, is Backscatter L-HV 5. t mean is sorted from low to high values. Disturbance 0 A 50 B 100 (%) σ A Example from ALOS Path/Row 170/610 n ( pq ( ni ))t 6. The gradient for the sorted t mean is computed as f pq = Regrowth Canopy Density . σ A from MOIDS VCF ∇t mean ( n,pq ( ni )) 6.3-5 7. A threshold for significant major backscatter change is applied to the gradient of the sorted where is the average  over the area A time series means such that σ pq ( Ani ) ni ( )t [ Figure 6-7] of image n for each polarization channel will be given image, n, will then be the result of the cumulative sum for all images over the timespan t corresponding to the 8. subset (∇ t ) = ∇t > change_threshold by c , analysis performed for each pixel. mean mean selected images according to the procedure above, and Algorithmic flow of n,pq 9. NISAR images that correspond to time steps in the disturbance detection σ c = f n σ n,pq pq n,pq Cumulative sum analysis of time series is the basis for subset from step 5 (or the complement of subsets) with NISAR time σ A is the average  over the area Ani for ( n,pq ( ni )) series data based on 6.3-6 classical change point detection that investigates the are selected to form the time series for change the image n. Image values for the refined calibration change point analysis change in mean before and after a change in a time detection analysis. CHANGE POINT DETECTION WITH and canopy density – series (Schweder et al., 1976). It is a distribution-free CUMULATIVE SUM ANALYSIS backscatter curves. approach, applicable to short, irregular time series for Disturbance detection for each calibrated pixel x, y detecting gradual and sudden changes. A graphical (or segment, k) of the image for disturbed forests for example of this process is shown in Figure 6-8. 80 NISAR Science Users’ Handbook 81 NISAR Science Users’ Handbook

Let C be the time series of the subset of n selected Also cross-checking will be performed to determine [ Figure 6-8] NO CHANGE CHANGE scenes as 0 0 whether all observations shall be part of the cumulative Example of ALOS-1 Upland Forest Deforestation C = C , C , …, C sum calculation in (3), i.e. ,whether scene subsetting Riparian Forest Degradation 1 2 n time series (top) for –5 Agriculture –5 Regrowth was indeed appropriate or if further pruning might be various unchanged (left) The residuals against the mean of the time series is necessary. Once a threshold value is determined from and changed (right) –10 –10 computed (in power units) as calibration efforts, change can be flagged based on the land cover types. –15 –15 Bottom figures show the C = C1, C2, …, Cn cumulative sum values. corresponding cumula- –20 –20 tive sum curves. The cumulative sums, Si, are defined as the sum of the A candidate change point is identified from the S curve Black lines=L-HH, blue –25 –25 residuals, R, at each time step such that (dB) L-HH/L-HV BACKSCATTER i at the time where SMAX is found: TCPbefore = T(Si = SMAX) lines=L-HV backscatter/ Si = Si-1 + Ri with –30 –30 cumulative sum curves. 6.3-7 • TCPbefore, timestamp of the last observation before change

With I = 1, …, n and S0 = 0. 40 40 • Si, cumulative sum of R with i = 1, ... n

In the cumulative sum, the slope of S is indicative of • n, number of observations in the time series 20 20 change in a time series: The first observation after change occurred (TCPafter) is 0 0 • Upward slope: Values are above global mean then found as the first observation in the time series following T . • Downward slope: Values are below global mean CPbefore –20 –20

• Change in slope direction: Indication of change NORMALIZED CUMSUM A possible method to define thresholdS is based on a DIFF point location –40 –40 standard deviation of all SDIFF observations in the image The magnitude of change is calculated as stack. A suitable value found from experimental ALOS-1 Feb Aug Feb Aug Feb Aug Feb Aug Feb Feb Aug Feb Aug Feb Aug Feb Aug Feb 2007 2007 2008 2008 2009 2009 2010 2010 2011 2007 2007 2008 2008 2009 2009 2010 2010 2011 SDIFF = max(S) – min(S) data analysis is: 6.3-8 thresholdS = MEAN SDIFF +1.5∗ STDDEV SDIFF DIFF ( i ) ( i ) Larger S values are indicative of greater change. A DIFF 6.3-9 change point can be defined with two criteria: dence in a change point. Count can be expressed as a applied to the raster data set where change points are a) A clear change in slope is detected in the This threshold will also vary with ecosystem and forest percentage confidence level (CL): identified set. This is not necessary if the analysis was cumulative sum curve, with upward and structural types. performed on image segments (to be tested). downward slopes exceeding a gradient C L = count S < S / N ( DIFF-Bootstrapped DIFF ) threshold. For assessment of the robustness of detected change 6.3-10 RULE-BASED CLASSIFICATION OF points, CUMSUM change point detection can be DISTURBANCE b) SDIFF exceeds a threshold for change combined with bootstrapping (random reordering of with N = number of bootstrapped samples. labeling. After change point detection a classification rule set is the observation dates in the time series) to measure applied based on threshold curves of mean L-HV back- confidence in marking a change point. The latter computation also makes change point de- Criteria a) and b) are values to be determined empiri- tection in time series somewhat robust against outliers scatter values from the of the time series segments cally from calibration activities as they can be expected in a time series as their importance in a bootstrapped before γ ! and after ! the detected change First, a confidence interval is computed from “how before γ after to be different for different forest structural types and points for each 25-m pixel. many times a bootstrapped SDIFF is less than the analysis decreases. environments with varying soil moisture conditions. original SDIFF. High count corresponds to higher confi- After applying a confidence level filter to further clean Threshold curves are derived empirically for each out spurious single 25-m pixels, a 2×2 sieve filter is time-series frame from corresponding MODIS or Landsat-based canopy density layers and stratified 82 NISAR Science Users’ Handbook 83 NISAR Science Users’ Handbook

based on NISAR time-series based unsupervised disturbance product will be issued as a 25-m binary sonal inundation patterns, allows for partially wetted or shorter multi-temporal average. This would improve image clustering, e.g. via the Isocluster algorithm. For base product with pixels flagged as disturbed (1) or not temporarily wet pixels to be identified as wetlands. the robustness of the classification by increasing the each cluster, canopy density – backscatter curves are (0), and a 100-m vector product with numbers from 0 effective number of looks within 1 ha pixels at the generated based on linear-least squares regression or to 16 for the count of marked disturbed pixels. Error For classification of open water, the backscatter is expense of temporal resolution. If multi-temporal av- other appropriate statistical models. From the curves, metrics for detection will also be made available for the generally significantly lower than non-inundated eraging were required to meet classification accuracy thresholds for expected L-HV backscatter change in dB 25-m and 1-ha products. landscapes with three possible exceptions: 1) depend- requirements, the multi-temporal averaging would be for 50% change before and after disturbance observa- ing on the noise equivalent o of NISAR, distinguishing accomplished as a separate pre-processing step and 6.4 ECOSYSTEMS PRODUCTS – tions can thus be established (Threshold50). This allows open water from other low backscatter targets such as implemented as rolling averages to maintain the 12- the final labeling of a pixel as disturbed (with greater INUNDATION bare ground or mudflats may be difficult; 2) at steep day interval for the output of classification results. As than 50% loss in fractional forest canopy cover): incidence angles, wind roughening can make open such, pixels transitioning between different inundation The NISAR L2 science requirement for wetlands water brighter than typical open water values and; 3) states during the temporal averaging period could still inundation is expressed as: The NISAR project shall open water with floating vegetation is not categorized be captured. ⎧ measure inundation extent within inland and coast- ⎪ disturbed, ! ! Threshold50 separately. In the first case, since open water does not pixel = ⎨ γ before −γ after → al wetlands areas at a resolution of 1 hectare every ⎩⎪ undisturbed, other generally change quickly or to a large degree compared 6.4.1 Implementation Approach for 12 days with a classification accuracy of 80%. 6.3-11 to the extent of inundated vegetation, averaging the Algorithm data over time or over area can be helpful for reducing A review of publications quantifying the accuracy of The detection of inundation in NISAR images follow the The resulting image is vectorized to a fixed one-hectare the noise over bare ground areas, which are generally mapping wetlands with L-band SAR was completed processing flow shown in Figure 6-9 and is described grid, and all polygons containing disturbance flagged brighter than the expected noise equivalent o. In in October 2014 (after 2014). The review concluded here. The images of the multi-temporal sequence pixels are retained for a final output product. If in any the second case, we will make use of images taken the wetlands accuracy requirement could be achieved are radiometrically calibrated relative to each other one-hectare cell, eight or more 25-m pixels are labeled with different viewing geometry and time to identify by NISAR. Methods to classify, radar images, ranged to a higher precision than perhaps required through disturbed, the 1-ha cell is flagged as disturbed. open water surface. Indeed, observing the response of from utilizing simple thresholds to machine learning routine standard calibration of the NISAR imagery. This Retaining only one-hectare cells as vector layers with open water and land surfaces viewed from a differ- approaches, sometimes in combination with image calibration step examines distributed targets that are attributes for number of detected disturbed pixels, error ent perspective differs and could be discovered from segmentation. Inundated vegetation can be observed expected to be unchanged or minimally changed in metrics and trends (retaining values of subsequent comparing ascending and descending passes. Another by L-band SAR when woody vegetation vertically brightness over a set timespan of the set. With NISAR’s years), will result in a vast reduction of the image raster method to distinguish open water from bare ground emerges from the water surface enhancing the double 240 km swath width, it is reasonably assumed that layers as only 3%–5% of any given area on average is through examination the interferometric coherence. bounce scattering mechanism which is especially ap- a statistically large area, Ani, will not be inundated (or can be expected to be disturbed. In order to monitor The observed repeat-pass coherence over bare ground parent in the HH channel. Wetlands are often adjacent otherwise changing) during any of the 2n observations disturbance trends, it is suggested that any 1-ha cell is is typically higher than that of open water. For the to open water or senesce into open water surfaces surrounding the image to be calibrated and classified. retained where two or more 25-m pixels were flagged third case, a subcategory of floating vegetation is not that provide significant contrast landscapes, facilitating These areas will be identified through use of a priori as disturbed. classified, but these areas will often be identified as detection and mapping of inundation regions using the wetlands mask and partly through image segmentation inundated areas and therefore meet the objectives of NISAR mission. or other methods over the 2n images. 6.3.3 Planned Output Products the requirement.

The NISAR mission L2 science requirement for distur- The NISAR baseline algorithm uses the most common The calibration correction for image n (the middle As a preliminary step, a baseline classification will be bance detection defines disturbance as “50% or more method to identify inundated vegetation or open water: image of the 2n multi-temporal sequence), fn, for each generated from the multi-temporal radar backscatter fractional forest canopy cover lost in a one-hectare detection thresholds. A recent example that describes a polarization channel pq, is average of an images sequence to represent the initial (100×100 m2) resolution cell”. The mission shall procedure similar to that which would satisfy the NISAR σ A inundation state representative of the time period of n ( pq ( ni ))t measure disturbance annually with an error rate of less requirement can be found in Chapman et al, 2015. The f p q = the images. The wetland classification generated for σ A than 20% globally. The first NISAR disturbance product algorithm benefits from the double bounce scattering ( n , p q ( n i ) ) 6.4-1 each orbit cycle is obtained through change detection will be issued for the second year of the mission. The effect that occurs in inundated vegetation, and particu- of the images within this orbit cycle relative to this σ σσA AA algorithm specified in this document is designed to larly strong in the HH channel. Meanwhile, the observed where( p q((( ppnqqi()()tnni i)))t tis the average backscatter, , baseline. The accuracy of the subsequent classifica- produce products meeting this mission requirement HV backscatter changes remains relatively small. To over the area A for all images and all polarizations, pq, tions generated for each orbit cycle could potentially be by quantifying annually disturbed forested areas. The refine wetland classification, a change detection algo- improved through comparison with an additional but rithm over a sequence of radar images covering sea- 84 NISAR Science Users’ Handbook 85 NISAR Science Users’ Handbook

IMAGE TIME SERIES CALIBRATED RELATIVE TO where the thresholds s r_iv(q ) and s hh_iv(q ) may be a function of the incidence angle, q , and are determined MULTI-TEMPORAL AVERAGE T inc T inc inc r_iv through a pre-launch and post-launch calibration process. sT (qinc) is the threshold value for classification of Find multi-temporal σ HH (θinc ) average; calibrate indi- hh_iv inundated vegetation from the ratio of the polarization channels HH and HV, and sT (qinc) is the vidual images relative σ HV (θinc ) to the multi-temporal threshold value for classification of inundated vegetation from the HH backscatter given by σ HH (θinc ) . average using data outside of wetlands mask. Classification C for each pixel x, y (or segment k) of the image for open water for the multi-temporal average image is:

σ HH (θinc ) r _ ow Cx,y = 2 or Ck = 2 if > σ T (θinc ) and From the multi-tempo- σ HV (θinc ) ral average image, use h h _ o w [ Figure 6-9] σ HH (θinc ) > σ T (θinc ) calibrated classifi- 6.4-4 cation thresholds to Algorithm flow for an determine “typical” ex- example multi-temporal tent of open water and r_ow hh_ow where the thresholds sT (qinc) and sT (qinc) may be a function of the incidence angle, qinc, and are determined inundated vegetation sequence of 3 images. r_ow over the time period of through a pre-launch and post-launch calibration process. sT (qinc) is the threshold value for classification of the image sequence. σ θ HH ( inc ) hh_ow open water from the ratio of the polarization channels HH and HV, , and sT (qinc) is the threshold σ HV (θinc ) 1 2 3 value for classification of open water from the HH backscatter given by σ HH (θinc ) .

All other image pixels have Cx,y=3 or Ck=3. This generally encompasses areas outside the wetlands mask and may include areas within the wetlands mask, such as those areas determined as not inundated by exclusion FINAL CLASSIFICATION FOR EACH from the open water and inundated vegetation classes. Examine change in backscatter be- IMAGE DATE tween the multi-temporal average to each individual image (or image There may be a small number of pixels where the classification is indeterminate. For example, at the given sequence) to derive the classifi- 1 2 3 cation result for each image date incidence angle, it is not possible to conclusively classify the data. For those cases, the classification would be or date range, based on calibrated Cx,y=4 or Ck=4. thresholds of backscatter change that indicate a refinement to the σ θ classification of inundation state. r _ in HH ( inc ) r _ in Cx,y = 4 or Ck = 4 if > σ T (θinc ) > > σ T (θinc ) and σ θ HV ( inc )

hhin hh_ ind σ T (θinc ) > σ HH (θinc ) > σ T (θinc ) 6.4-5

over the timespan t corresponding to the 2n images, and σ A is the average  over the area A for the ( n,pq ( ni )) Cx′,y = 31 or Ck′ = 31 if image n. Backscatter values for the refined calibration of image n for each polarization channel is given by: C = 3 and

C hhveg σ n,HH (θinc ) > σ T (θinc ) and c n 6.4-2 C σ n,pq = f pqσ n,pq σ θ n,HH ( inc ) trmin_ iv > σ T (θinc ) and σ HH (θinc ) The multi-temporal average of the 2n images is given by spq. Classification C for each calibrated pixel x, y (or σ C θ segment k) is: n,HV ( inc ) trHVmax > σ T (θinc ) σ HV (θinc ) 6.4-6 σ HH (θinc ) r _ iv σ HH (θinc ) r _ iv Cx,y = 1 or Ck = 1 if > σ T (θinc ) > σ T (θinc )and σ HV (θinc ) σ HV (θinc ) hhveg trmin_iv hh_ iv where sT (qinc) is the minimum HH threshold backscatter value for vegetation covered terrain; sT (qinc) is σ HH (θinc ) > σ T (θinc ) 6.4-3 the minimum change threshold ratio relative to the previous observation for HH backscatter indicating a transi- 86 NISAR Science Users’ Handbook 87 NISAR Science Users’ Handbook

tion from non-inundated to inundated vegetation ; and Looking for areas that may be decreasing in open Also required for generating the classification product use such data for allocating resources, projecting trHVmax( ) is the maximum HV change threshold ratio water extent: is an a priori wetlands mask where inundation could agricultural outlook, publishing market projections, sT qinc relative to the previous observation indicating that the occur and excluding confounding landscape types allocating resources and assessing food security often

vegetation characteristics did not otherwise change. Cx′,y = 23 or Ck′ = 23 if such as urban areas and agricultural areas, as well as on a month-to-month basis. While current remote C = 2 and terrain slopes, volcanic terrains and deserts. The output sensing inputs for crop-area identification methods rely hh Looking for areas that may be decreasing in inundation C veg resolution of the product will be 1 ha. The value of the primarily on reflectance spectra from optical data, radar σ n,HH (θinc ) < σ T (θinc ) and extent from inundated vegetation to not-inundated: σ C θ 1-ha cell will be either through direct classification of has the potential for making a great impact because n,HH ( inc ) trmin_ ow > σ T (θinc ) the average of the input 25-m SAR data product or by of its sensitivity to the structure of ground-cover and σ HH (θinc ) majority vote among the of the classes of the 16 input its insensitivity to cloud cover and lighting conditions. Cx′,y = 13 or Ck′ = 13 if C = 1 and 6.4-9 pixels and the direct classification of the 1 ha SAR. Through its global observing strategy and 12-day revisit hh hh C veg C veg period, the NISAR mission has the capacity for col- σ n,HH (θinc ) > σ T (θinc )σ n,HH (θinc ) > σ T (θinc ) and trmin_ow C where s (q ) is the minimum change threshold An error probability layer for the classification will be lecting data that is relevant to the societally important σ θ T inc n,HH ( inc ) trmax_ iv < σ T (θinc ) and ratio relative to the multi-temporal average for HH provided, based on a statistical analysis of the ob- applications of monitoring and measuring global food σ HH (θinc ) backscatter indicating a transition from open water to served backscatter distributions versus the backscatter production. This is reflected in the mission’s crop area C σ n,HV (θinc ) trHV min non-inundated terrain without vegetation. thresholds used in the classification. requirement. > σ T (θinc ) σ HV (θinc ) 6.4-7 Similar tests for indeterminate areas where C=4 for 6.5 ECOSYSTEMS PRODUCTS – 6.5.1 Theoretical Basis of Algorithm Equations 6.4-6 and 6.4-8 would be made. CROP MONITORING The NISAR L2 science requirement for wetlands where strmax_iv(q ) is the maximum change threshold For all pixels or regions where C′ is nonzero, C is T inc The NISAR L2 science requirement for agricultural crop inundation is expressed as: The NISAR project shall ratio relative to the multi-temporal average observation replaced with C′. area is expressed as: The NISAR project shall measure measure crop area at 1 hectare resolution every 3 for HH backscatter, indicating a transition from inun- crop area at 1 hectare resolution every 3 months months with a classification accuracy of 80%. dated to non-inundated vegetation; and strHVmin(q ) is An error layer will be generated utilizing the observed T inc with a classification accuracy of 80%. the minimum HV change threshold ratio relative to the probability distribution function of inundated vegetation, The Crop Area algorithm is based on the coefficient multi-temporal average, indicating that the vegetation open water, and non-inundated backscatter values To feed a growing population of more than 8 billion, of variation (CV), which is the ratio of the standard characteristics did not otherwise change. compared with the calibrated threshold values. food production and supply occur on a global basis. deviation over the mean for a time series of orthorec- In order to better guide policy and decision making, tified radar cross-section data (Whelen and Siqueira, Similarly, for open water we may examine cases where 6.4.2 Planned Output Products national and international organizations work to trans- 2018). Here, the coefficient of variation is computed open water extent is increasing: The specified product for validation of the L2 require- parently monitor trends and conditions of agriculture for both the co- and cross-polarized data (HH and VH) ment to measure inundation extent is a raster classi- on a timely basis. Because of the variable nature of averaged to a hectare-scale, and where the time series Cx′,y = 32 or Ck′ = 32 if fication of inundated extent at a spatial resolution of 1 planting and harvesting practices, efforts such as this are collected quarterly. These time periods cover typical C = 3 and hectare. The pixel values have the following values: 1) are manpower intensive and time-consuming tasks. growing seasons of crops and make best use of the hh σ C θ < σ veg θ and n,HH ( inc ) T ( inc ) inundated vegetation; 2) open water; 3) not inundated; Organizations such as the USDA, World Bank, and background (i.e., HH ± HV) land observations planned σ C θ n,HH ( inc ) trmax_ ow 4) indeterminate. Categories 1, 2, 3 and 4 may have FAO, publish statistics on crop area, type and yield by NISAR. < σ T (θinc ) σ HH (θinc ) subcategories for pixels in transition. The resolution of on a regular basis. Much of this data is derived from

the product will be 1 ha. The product will be generated in-country surveys, augmented by what are, to date, There are two principal advantages that are offered 6.4-8 every 12 days after the first 6 (approximate) months of limited remote sensing components. by NISAR over existing technical approaches for crop observations are completed, and assumes that area estimation. These are: 1. an effectively all-weather where strmax_ow(q ) is the maximum change threshold T inc 20 MHz dual polarization HH, HV data are acquired Recent efforts to increase the use and accuracy of observing strategy that will provide observations of a ratio relative to the multi-temporal average for HH every 12 days for both ascending and descending orbit remote sensing data for agriculture applications have given area every 12 days (every 6 days if we include backscatter, indicating a transition from non-inundated directions. The input product is the L2, 25 m, radio- been led by the Group on Earth Observations (GEO) ascending and descending passes), and 2. the measure terrain without vegetation to open water. metric and terrain corrected, multi-look imagery. The under the GEOGLAM initiative (GEO Global Agriculture of radar cross section, which is dependent on contribu- incidence angle for each image pixel should also be Monitoring). This initiative consists of governmental tions of volume and surface scattering which are likely provided. education and non-governmental organizations that to change dramatically for actively managed agricul- 88 NISAR Science Users’ Handbook 89 NISAR Science Users’ Handbook

tural landscapes. Measures of radar cross section are will have less of an effect on the overall signature, even In order to provide a more theoretical foundation for the ground, volume and double-bounce). With these two more robust than interferometric measures of change, for regions which are left fallow, and hence classified use of the CV based on the target scattering physics, simulated time series, the observing period of NISAR such as through the decorrelation signature, which as non-active crop regions. The changes in the radar a theoretical scattering model was developed and is included to determine the season during which the may be an appealing alternative or augmentation to the cross-section for barren ground is only on the order described in the above-mentioned NISAR memorandum; two target types are being observed, and a time series base-algorithm that will be used for estimating active of a few dB or less, much smaller when compared to one that includes the scattering model and an obser- of NISAR observations simulated and the CV computed. crop-area. the many dB that the RCS undergoes throughout the vational error model, in order to show the separation Once done, a threshold classifier is employed based on growing season. between simulated natural and actively managed land- the CV and a hypothesis test applied to the resulting There are two types of error that can affect the active scapes. A summary of this model is given here. classification. Given that the CV pdfs have thus been crop area estimates mentioned here: those associated Errors induced by a disturbance event as opposed determined for the two different landcover types with the instrument and those related to the region to actively managed land may result in a degree of A relationship between radar observation and classi- (managed versus unmanaged) the hypothesis test and being observed. misclassification for agricultural area. This effect fication accuracy is needed for the observations and probabilities of correct and incorrect classification is is expected to be small however, especially when those components that contribute to the target RCS. determined as a function of the choice of threshold. For sources of error related to the instrument, mea- comparing regions from one year to the next. For this The observational error model is the same model as surement stability and cross-track variability in the reason, estimates of active agricultural crop area are used by the forest disturbance algorithm (Cartus et al., 6.5.2 Implementation Approach for signal to noise ratio (SNR) will be the dominant factors. not planned to meet the full requirement accuracy of 2018). It relates the observed radar cross-section for Algorithm Simulations and ALOS-1 observations for India have 80% until after the first year of NISAR observations. each polarization pq, written here as o for simplicity, obs The algorithm flow is presented in Figure 6-10. Time shown that the coefficient of variation will be robust for to the observation error sources, o and the radar obs-error series are assembled every 3 months after the first relative calibration errors up to 1 dB (ALOS-1 is quoted Lastly, misclassified regions such as open water and cross-section of an agricultural field, o is field year of data collection, and from that, the CV is com- to have a calibration accuracy of some 10ths of a dB). urban regions included in the mask for NISAR assess- ! ! ! puted for each available polarization. Minimally, this Variation in the SNR will occur as a function of the tar- ment of agricultural area are a potential source of error. σ obs = σ obs-error +σ field 6.5-1 would be HH and HV polarized fields; however, in the get brightness and the incidence angle. Normalization Use of a simple threshold classifier on the RCS mean U.S. and in India, it is expected that fully polarimetric of the RCS standard deviation by the RCS mean will however has shown to be an effective method for It can be shown that the radar cross-section of the field data will be available. For each of the computed coef- remove much of this variability. Regions with very low removing open water regions and those with low SNR; can be broken down into components of the return from ficients of variation, a determination will be made via a SNR, close to the noise floor of NISAR, will be removed while urban regions with a bright RCS and proportion- the ground, o ,volume,  , and double-bounce, ground ovol pre-determined threshold, on a per hectare basis, if the through a simple threshold classifier based on the ally small variation in the RCS as a function of time,  , returns from the vegetation components. Additional odb CV indicates that the area is actively being managed brightness of the mean RCS. have been shown thus far to be correctly identified by parameters that govern the model are the fractional or not. Results for each polarization will be compared the CV-based classifier of crop area used here, and canopy cover, , the attenuation of the signal as it pass- with the other polarization results, as well as combined Sources of error in active crop area that are associ- appear to be more successful and detailed than those es through the vegetation layer, , and the height of the with ancillary data that may be available from ESA’s ated to the target can be principally assigned to three classifiers that depend on optical data alone. vegetation layer, h. The net model, which is based on CCI (esa-landcover-cci.org) and the SRTM (or better) sources. These are: 1. weather induced changes to the (Dobson and Ulaby, 1986; Askne et al., 1997) is given as topographic map. Based on the limited set of inputs, radar cross section, 2. disturbance events mistaken The NISAR measurement metric for Crop Area deter- and comparison to the previous quarter’s results, a defor crop regions, and 3. misclassification of regions as mination relies on the CV, which is a measure of the σ ! = σ ! ⎡1−η 1− e−ah ⎤ + field ground ⎣ ( )⎦ termination will be made for which of the four classes potentially belonging to agricultural landscapes (e.g., degree of change (normalized with respect to the mean ! −ah ! −ah that each 1-ha region should be classified: 1) active urban areas and open water). backscatter) as a function of time. This metric makes σ ⎡η 1− e ⎤ +σ ηhe vol ⎣ ( )⎦ db crop area, 2) newly active crop area, 3) inactive crop use of the fact that agricultural landscapes are heavily 6.5-2 area, and 4) not crop. Errors that are related to weather induced changes managed, and hence, the scattering physics of agricul- in the RCS will manifest themselves as short-term tural crops change more than other landcover types. A By creating a time series model for the inputs of (6- The observing strategy for the determination of crop variations in the trajectory of the RCS as the agricultur- full treatment and analysis using the CV for crop crea 5.2) (e.g., in terms of how , h, are changing over the area is broken down into two time periods: 1) during al region changes from barren land, emergent plants, determination using ALOS-1 observations can be found growing season) it is possible to create a time series for the first year of observations, there is no planned full grown crops to harvested land. As the plants above in Whelen and Siqueira (2018). the radar cross-section observed for a field. A similar delivery of crop area determination from NISAR, and 2) the soil grow and mature, changes in the soil moisture time series is created for a landcover type that is not during successive years, a 1 ha-resolution raster image changing over time (e.g., given values for the RCS of 90 NISAR Science Users’ Handbook 91 NISAR Science Users’ Handbook

s 0 TIME SERIES COEFFICIENT OF VARIATION CLASSIFICATION of the crop area classification will be generated every N 2 HV ⎛ ! ⎞ 1 ⎛ ! ! ⎞ (e.g., Chv) (e.g., Areahv) three months. stddev σ t = σ nΔt −σ ⎝ pq ( )⎠ N −1∑⎝⎜ pq ( ) pq ⎠⎟ n=1 [ Figure 6-10] 6.5-5 The first year of NISAR observations are used for the Data and algorithm determination of baseline thresholds for crop area clas- flow diagram. In the above, N is the total number of observations sification based on the CV metric determined for both in the observing period, and t is the time period } HH and HV polarizations (CVHH and CVHV), computed between NISAR passes (expected to be 12 days). separately for the ascending and descending passes Calculations of the CV are made on a per-pixel (25 m) of NISAR. These thresholds are determined through basis and aggregated after computation into 1 hectare the NISAR post-launch calibration period, where pixels where both the mean and standard deviation of histograms of CVHH and CVHV for the ascending and Co-Pol 0 Area the CV determination are kept at the 1 ha-resolution Time s Chh hh Voting descending passes are created for crop and non-crop C Areaw Classifier for each polarization and orbital direction of the satel- Series w regions using pre-determined validation resources from lite. Hence, each 1 ha-resolution element will consist ESA’s CCI and USDA’s CropScape data layers. of eight values, four for the mean of the CV, and four for the standard deviations, as in: Values of CVHH and CVHV are determined for each 3-month period, post launch, bracketed by the dates: Globcover and Topographic Map CVHH-asc, CVHV-asc, CVHH-desc, CVHV-desc, January 1 – March 31, April 1 – June 30, July 1 – Sep- (Optional) stddev(CV ), stddev(CV ), stddev(CV ), tember 30, and October 1 – December 31, as well as HH-asc HV-asc HH-desc stddev(CV ) for the entire one-year period. HV-desc

Crop Area Of the group of four estimates of CV, the estimate with Map The CV for each polarization is determined by the the highest standard deviation will be eliminated and standard deviation of the radar cross-section divided the remaining three will be used in a threshold classifi- by the mean of the radar cross-section, collected as a cation whose value of threshold is determined through function of time (e.g., the 3-month period), at the 25-m 6.5.3 Planned Output Products Mosaics of the data are not desired. In order to make the NISAR calibration phase. An additional thresh- resolution of the input data product. Units of the input the co-registration and geolocation of images a trivial old will be used for detecting water bodies, which, The specified product for validation of the L2 require- radar cross-sections should be in m2/m2 and not dB. process, pixel locations should be quantized onto a because of their small values for radar cross-section, ment to measure crop area is a raster classification. That is: pre-determined geographic grid. will also display large values for CV. In the last step, a The pixel values based on prior and current determi- voting classifier is used to determine if the region is an nation of active crop area are given in Table 6-3: 1) stddev⎛σ ! t ⎞ Also required for generating the classification product ⎝ pq ( )⎠ active crop region or not. not a crop, 2) newly active crop area; 3) inactive crop CVpq = is a landcover mask that indicates those regions where mean⎛σ ! t ⎞ area (fallow); 4) active crop area; and 5) not evaluated ⎝ pq ( )⎠ agricultural monitoring is intended to be performed. Determination of which pixels change classes over (class 0). The resolution of the product will be 1 ha. The Additional layers would prove useful in increasing the 6.5-3 time can be determined by comparing one time period product is intended to be generated every 3 months accuracy of the agricultural area classification and are classification versus a previous, thus providing useful after the first year of observations are completed, and being investigated as part of the NISAR phase C efforts. where information about planting and harvesting periods. assumes that 20-MHz dual polarization HH, HV data These are as follows: More accurate determination of the start of the are acquired every 12 days for both ascending and 1 N 1. An up-to-date version of ESA’s CCI landcover mean⎛σ ! t ⎞ = σ ! = σ ! nΔt planting and harvesting seasons can be determined descending orbit directions. ⎝ pq ( )⎠ pq ∑ pq ( ) N n=1 on a per-pixel basis by going back to the original radar map (esa-landcover-cci.org). This data is useful 6.5-4 cross-section data and using a running window to The input product is the L2, 25-m, radiometric and ter- for defining limits of urban areas, inside of which and determine when the CV statistics changed. rain corrected, calibrated multi-look imagery for each formal agricultural practices are excluded from the of the polarizations, for each period of data collection. analysis. 92 NISAR Science Users’ Handbook 93 NISAR Science Users’ Handbook

2. Topographic and look angle maps that are co-reg- that flow is parallel to the known surface (Joughin et interferometric phase is ideal for the slow-flowing 2010; Rignot et al., 2010a; Mouginot et al., 2017), istered to the NISAR data grid. Such maps need al., 1998). An advantage of this technique is that the interior. Although the 50 m/yr distinction between slow with recent work demonstrating the range-range only be supplied once using the expected viewing data are relatively high-resolution (<100 m) and the and fast flow in the requirements is aimed at sepa- offsets combination (Joughin et al., 2018). This latter geometry of the sensor. phase noise is low (<~2 cm). A major disadvantage is rating the areas where each technique should work combination is more attractive for NISAR because a) that for fast moving areas it is difficult or impossible to the best, in many cases interferometric phase will still the 80-MHz mode provides considerably finer range 6.6 CRYOSPHERE PRODUCTS- unwrap the phase. Another issue is regions where there work in considerably faster-flowing areas (up to about (~2.5 m) than azimuth (~7 m) resolution, b) there ICE SHEETS is significant ionospheric activity such that the spatially 500 m/yr for NISAR). Thus, no single velocity threshold will be ample ascending/descending coverage, and variable path delay introduces large interferometric can cleanly separate the regions where phase fails c) relative to azimuth offsets, the range offsets are less A major objective for NISAR is to collect data to phase errors (several m/yr errors). For NISAR, these and speckle-tracking must be used. For example, with affected by ionospheric distortion. Hence, the range- measure velocity over the Greenland and Antarctic ice errors will largely be removed using split-spectrum RADARSAT phase can sometimes be unwrapped on range offsets combination likely will be the dominant sheets through time. These same data will be used to processing applied to the 80-MHz-bandwidth data. smooth flowing ice shelves at speeds approaching contributor to velocity estimates in regions of fast flow. determine the time-varying position of the grounding 1000 m/yr. By contrast, for some regions on the ice All of the above methods will be implemented, but any line around Antarctica and on floating ice tongues in In areas where the motion is too fast for interferometric sheets where the speeds are less than 100 m/yr, there of them can be selectively turned off (e.g., methods 1 Greenland. The ice sheet related science requirements phase measurements, velocity will be determined using are strong phase gradients where ice flows over bumps and 2 where azimuth offsets add no improvement to call for measurements of ice sheet velocity derived the azimuth and range offsets derived by cross-cor- that make phase unwrapping difficult or impossible. As the derived estimates.) using a combination of interferometric phase data and relating patches from pairs of image to determine a result, for all of the velocity related requirements, at offsets from speckle tracking. Near ice sheet grounding displacements between image acquisitions (Gray et each point on the ice sheet, the corresponding require- All ice-sheet velocity maps will be produced on polar lines, time series of interferometric phase will be dif- al., 1998; Michel and Rignot, 1999). Although image ment will be met using the best available combination stereographic grids at a posting of 100 m (actual reso- ferenced to estimate relative tidal displacement, which features can improve correlation, this technique works of interferometric phase and speckle-tracked offsets. lution in faster-moving regions will be 250 m or better.) helps grounding line position. This section describes best when the speckle patterns are well correlated; The different temporal and spatial resolutions specified Consistent with the existing products, the Greenland the algorithms needed to generate these products. hence this technique is often called speckle tracking. in the requirements reflects the amount of spatial map-projection will use a standard latitude of 70°N and Advantages to this method are that velocity estimates and temporal averaging necessary to meet each a central meridian of 45°E and the Antarctica projec- 6.6.1 Theoretical Basis of Algorithm can be derived from a pair of images collected along requirement. tion will use a standard latitude of 71°S and a central For slow-moving areas (<50 m/yr) and some fast a single orbit track (i.e., ascending or descending only meridian of 0°. Glacier products outside of Greenland moving areas where the data are conducive to such orbits) and it can be used to measure extreme motion The basic algorithmic approach will follow earlier and Antarctica will use region-dependent projection measurement, horizontal velocity will be measured (>10 km/yr). Because the technique uses image chips approaches (Joughin, 2002). Specifically, at each point (e.g., UTM). using radar-line-of sight determined from the interfer- several 10s of pixels in dimension, the spatial resolu- in the output grid, the algorithm will cycle through the ometric phase from at least two crossing orbit tracks tion is much poorer (>~200 m) than phase estimates. various options: TIDAL DISPLACEMENT (i.e., ascending/descending) under the assumption Since displacement is resolved with to a fraction (i.e., 1. Range-azimuth offsets from a single orbit track, Differential tidal displacement products will be ~1/20 of a several-meter pixel) of a range or azimuth 2. Unwrapped phase (for range) with azimuth offsets produced by differencing pairs of interferograms over TABLE 6-3. PIXEL CLASSES FOR ACTIVE CROP AREA BASED pixel, accuracy also is much less than phase estimates, from a single orbit track, the ice shelves and grounding lines. This differencing ON PRIOR AND CURRENT DETERMINATION which resolve motion to a fraction of a wavelength. approach cancels the horizontal motion (assumed PRIOR DETERMINATION 3. Range-range offsets from crossing orbit tracks, constant) common to both interferograms, leaving In polar regions, ionospheric distortion can be severe, and 0 1 only the double-differenced, time-varying, vertical particularly for the azimuth offsets. This distortion 4. Unwrapped phase-phase data from crossing tidal displacement, which indicates the location of the Current 0 Not a crop (class 1) Fallow (class 2) can produce errors of more than 100 m/yr in some orbits. grounding line/zone, i.e., the place where ice detaches determination locations. This problem can be mitigated by using 1 Newly active (class 3) Active (class 4) from the bed and starts becoming afloat in the ocean range-only offsets from crossing orbits as described At each point in the output, all of the viable combina- waters (Rignot et al., 2011b; Scheuchl et al., 2016). below. tions will be calculated. Estimated errors for each type Although this technique is generally applied to of velocity estimate will be used to weight the results phase-only data, in the presence of very high strain The requirements for fast and slow motion reflect to produce an optimal inverse-error weighted average rates it is possible to apply the technique on range the fact that lower resolution speckle tracking is best for the horizontal components of velocity. All of these offsets with a reduced level of precision in determining suited to measuring fast-flowing outlet glaciers, while combinations have been widely used (Joughin et al., 94 NISAR Science Users’ Handbook 95 NISAR Science Users’ Handbook

the grounding line position and in detecting vertical RAW SPECKLE TRACKED OFFSETS define terms rather than to provide the details of the ICE VELOCITY DERIVED FROM SPECKLE displacements (Joughin et al., 2016). corrections, which are provided elsewhere, here we TRACKING ALONG A SINGLE ORBIT TRACK At a given set of range-azimuth coordinates, (ρ ,s ), in 1 1 have bundled the geometry, baseline, and elevation the reference SLC (first image acquired), cross correla- Speckle tracking provides two components of the GLACIER ESTIMATES dependent corrections in a scalar function f . Similarly, tion is used to locate same point, (ρ ,s ) in the second ρ three-component velocity vector: the along-track 2 2 we can correct the azimuth displacements as The mission will collect an unprecedented volume of SLC, which is in non-integer values. The raw range and horizontal component and the line-of-sight component, data to measure glacier velocities in regions outside ! azimuth offsets, (δ ,δs), given by, δ s = δ − f o ,o which mixes vertical and horizontal motion (Figure ρ s s ( 1 2 ) 6.6-4 of Greenland and Antarctica. The steep terrain where 6-11). Although there is a component of the vertical many of these glaciers exist, however, present chal- δ = ρ − ρ and δ = s − s velocity directed toward or away from the ice-sheet ρ 2 1 ρ 2 1 The unwrapped interferometric phase, f, requires lenges (e.g., glaciers lying in radar-shadowed regions) 6.6-1 surface, this motion generally is small enough similar correction such that that make it difficult to quantify what fraction of gla- (<1 m/yr) that it can be ignored or estimated inde- RAW INTERFEROMETRIC PHASE ! ciers can be successfully mapped; prior measurements φ = φ − fφ (o1 ,o2 ,z) pendently. Instead, much of the vertical motion is 6.6-5 indicate a relatively high likelihood of success for many assumed to be due to surface-parallel flow (i.e., a Given to co-registered SLCs, _ and _ , the phase of regions (Burgess et al., 2013). As a result, glaciers are I 1 I 2 particle on the surface flowing along the surface gra- the interferogram is given by: Note this correction for phase assumes that at least a mission goal rather than a requirement, requiring dient) (Joughin et al., 1996). If the slope is known and one point of known speed is used as control point to no formal validation. Thus, the focus of this document φ = ∗ surface-parallel flow can be assumed, the line-of-sight w A r g ( I 1 I 2 ) 6.6-2 determine the unknown 2-π ambiguity associated with is on producing ice-sheet velocity measurements. component can be resolved into horizontal and vertical phase unwrapping. Such control points are routinely These algorithms, however, are directly applicable to components. which is only known modulo 2π. Thus, a phase used in ice-sheet velocity mapping (Joughin et al., mapping glaciers (actual implementations of production unwrapping algorithm is applied to determine the 2010; Rignot et al., 2011a). processors might require some modification for specific The line of sight displacement is given by unwrapped phase, f. projections and other region dependent data.) Hence, VELOCITY ESTIMATES AT A POINT ! throughout the remainder of the document where δ ρ = Δg sin ψ − Δzcos ψ 6.6-6 CALIBRATED OFFSETS AND PHASE ( ) ( ) ice-sheet velocity mapping is referred to it is with the The following subsection describes how velocity is es- understanding that the text is equally applicable to The interferograms and range offsets also contain timated at each point. Note all equations are computed where ψ is the local incidence angle (with respect an glaciers (any place where this might not be the case information about the topography, with sensitivity assuming the look vector lies in a plane orthogonal to ellipsoidal Earth), and ∆z and ∆g are the vertical and will be so noted.) determined by the baseline. The imaging geometry will the satellite track (e.g., small squint). These equations ground-range displacements, respectively. Solving for introduce additional displacements unrelated to surface have been widely used with data from a variety of the horizontal displacement yields 6.6.2 Implementation Approach for motion. These differences can be corrected using the sensors, with no issues thus far. The imaging geometry ! Algorithm orbit and timing information. Here we encapsulate for NISAR has low squint, so these equations should be δ ρ Δg = + Δzcot(ψ ) this information (i.e., state vectors, range delays, and similarly valid. As synthetic products become available, sin ψ 6.6-7 The implementation approach for estimating ice veloci- ( ) any other ancillary information) into vectors, o and we will examine the validity of this assumption given ties uses speckle tracking and interferometric phase to 1 o , for the first and second images, respectively. With the more rigorous NISAR error requirements (< 1 m/yr). Assuming surface parallel flow, the vertical displace- compute the velocities. 2 this information, then signals other than those related Should this assumption not hold, then a transformation ment is given by to surface motion can be removed to produce the to a squinted coordinate system will be applied to the QUANTITIES USED IN VELOCITY surface-displacement only component of the range equations below. Such a transformation, however, does ! ∂z ∂z ESTIMATION Δz = δ s + Δg offset as not change any of the underlying principles described ∂s ∂g 6.6-8 Velocity estimates are derived using either interfero- below, nor have an impact on the viability of the algo- metric phase or speckle-tracked matches as described ! rithms, which are all well tested. Combining these two equations yields δ ρ = δ ρ − fρ (o1 ,o2 ,z) below. Here we define the notation used for the quanti- 6.6-3 ties that go into the velocity estimation equations. ! Note here we assume the offsets have been scaled δ ρ ∂z +δ s cot(ψ ) from pixels to meters. As our purpose here is to sin(ψ ) ∂s Δg = ∂z 1− cot ψ ( ) ∂g 6.6-9 96 NISAR Science Users’ Handbook 97 NISAR Science Users’ Handbook

Using this equation and the azimuth-offset estimate the NORTH [ Figure 6-12] NORTH

velocities in the radar-determined horizontal coordi- y Definition of angles used in yPS g nates are given by g the computation of horizontal asc a r velocities from ascending and ! sasc Δg δ s s descending orbits. The angle vg = and vs = ΔT ΔT between the polar stereographic 6.6-10 a x-axis and the local along-track aps Equation 6-6.13 gives the horizontal ice velocity in the direction is denoted by α and xPS x the angle between the ascend- radar-determined coordinates, but the final estimate is gdesc ing and descending along-track b produced in the projection-determined xy-coordinate [ Figure 6-11] Radar- and projection-determined directions by β. system (Figure 6-11). The rotation angle of the radar coordinate systems and their rota- coordinates with respect to North is given by tion angles relative to North.

s ⎛ ds dg ⎞ desc αa = aatan2tan2 , r ⎝⎜ dlat dlat ⎠⎟ ICE VELOCITY DERIVED FROM SPECKLE 6.6-11 TRACKING AND INTERFEROMETRY ALONG ICE VELOCITY DERIVED FROM INTERFER- 2 SINGLE ORBIT TRACK ⎛ 1 ⎞ ⎡ 1 −cosα ⎤ B = ⎢ ⎥, and The rotation angle relative to north for a point (x ,y ) in OMETRY FROM CROSSING ORBITS WITH ⎜ α ⎟ PS PS ⎝ sin ⎠ ⎣ −cosα 1 ⎦ SURFACE-PARALLEL FLOW 6.6-17 polar stereographic coordinates is given by In areas where interferometric fringes are noisy or aliased so they cannot be unwrapped, speckle tracking When data from crossing ascending/descending orbits provides a reasonable estimate. If data are available αaPS == aatan2tan2( yPS ,xPS ) are available the surface-parallel flow assumption can PS ⎡ ∂z ∂z ⎤ 6.6-12 only along a single orbit track and the phase can be ⎢ cot ψ cot ψ ⎥ be used to estimate horizontal components of velocity ∂x ( a ) ∂y ( a ) unwrapped, then a hybrid estimate can be derived (Fig- C = ⎢ ⎥. (Joughin et al., 1998; Mohr et al., 1998). Geometrically, ⎢ ∂z ∂z ⎥ ure 6-12). In this case, substituting the range displace- cot ψ cot ψ Horizontal velocities are then determined by rotating to this makes this 3-D problem a 2-D problem by assum- ⎢ ( d ) ( d ) ⎥ ~ ⎢ ∂x ∂y ⎥ 6.6-18 the polar stereographic system as ment given by the offsets (δ ∗ ∆ ) for the equivalent ⎣ ⎦ ~ ρ ρ ing the velocity vector lies in tangent-plain to the ice displacement in phase (λf/4π) in Equation 6.6-9 yields surface. In this case, using phase from ascending and ⎡ ⎤ ⎡ ⎤⎡ ⎤ the surface-parallel-flow approximated ground range In the equations, quantities are as defined above with vx cos(α PS −α r ) sin(α PS −α r ) vg descending passes, the horizontal components of the ⎢ ⎥ = ⎢ ⎥⎢ ⎥ subscripts a and d to indicate whether they are from an ⎢ v ⎥ ⎢ ⎥⎢ ⎥ displacement as y −sin(α PS −α r ) cos(α PS −α r ) vs velocity vector are given by ⎣ ⎦ ⎣ ⎦⎣ ⎦ ascending or descending pass, respectively. The angles ! λφ ∂z and are defined in Figure 6-13. The incidence +δ Δ cot ψ ⎡ ⎤ α β 6.6-13 s s ( ) λ φ! 4π sin(ψ ) ∂s ⎢ a a ⎥ angles ψ are ψ are defined relative to an ellipsoidal Δg = ⎡ ⎤ ⎢ 4πΔT sin ψ ⎥ a d Note the polar-stereographic coordinate system ∂z vx −1 a ( a ) 1− cot ψ ⎢ ⎥ = I − ABC ⎢ ⎥ Earth. A detailed derivation of these equations is given ( ) ( ) ! preserves angles but has a latitude-dependent scale ∂g ⎢ vy ⎥ ⎢ λ φ ⎥ 6.6-14 ⎣ ⎦ d d by Joughin et al. (1998). distortion. While locations are posted in polar-stereo- ⎢ 4πΔT sin ψ ⎥ ⎣⎢ d ( d ) ⎦⎥ graphic coordinates, which are subject to this distor- ICE VELOCITY DERIVED FROM SPECKLE Substituting this quantity into Equations 6.6-10 and tion, velocity vectors are posted in meters/year with no 6.6-15 TRACKING FROM CROSSING ORBITS WITH 6.6-13 yields the horizontal velocity vector in polar- scale distortion. SURFACE-PARALLEL FLOW stereographic coordinates. Where As described above, the ionosphere may introduce ⎡ cosβ cos(α + β ) ⎤ unacceptably large errors in some azimuth offset A = ⎢ ⎥, ⎢ sinβ sin α + β ⎥ estimates. Range offsets are much less sensitive to ⎣ ( ) ⎦ ionospheric errors, so when range offsets are combined 6.6-16 from crossing orbits, they can produce far less noisy 98 NISAR Science Users’ Handbook 99 NISAR Science Users’ Handbook

In producing such a mosaic, the algorithm proceeds single estimate for most points in the output grid. By by looping over the images to be mosaicked. If an es- contrast, for annual velocities, thirty or more indepen- timate is being derived using data from along a single dent estimates may be averaged for each point in the track (i.e., azimuth offsets are used), the algorithm next final output. Thus, as each point estimate is derived identifies where the corresponding region in the output using image pairs, the individual estimates are weight-

grid lies, and loops over the corresponding points in ed by wx and wy, summed in an output buffer. The final the output grid. It then interpolates the relevant offset velocity estimate in the x-direction is derived from N or phase data from the source image, which is in radar individual estimates is given by coordinates. Where this interpolation is successful and N there are valid data, the velocity components, vx and vx = ∑ fi wx,ivx,i vy , are determined using Equations 6.6-10, 6.6-13, and i=1 6.6-19 6.6-14. At each point, the algorithm uses phase data

if available for the range component, and if not, then and a similar expression applies for vy . Note fi is an [ ] Figure 6-13 range-offset data. After looping through all points in the additional feathering weight as described below. If we Example of the types of Level 3 velocity estimates. Such horizontal velocity estimates (i.e., it took 20 years of the data from several SARs sub-region of output grid, the algorithm proceeds to the assume the errors are unbiased (zero mean), then the products that will be produced can be determined from range-offsets with the same to produce a gap-free Greenland mosaic). Thus, by next image. weights must sum to one. In this case and assuming for the cryosphere using the methods as for interferometric phase. This measure- routinely imaging the ice sheets NISAR will greatly the individual estimates are independent, the minimum algorithms described in this ~ ~ 2 ment is made by replacing (λa f a)/4π and (λd f d)/4π improve the coverage, accuracy, and spatio-temporal Where crossing orbits are used, the algorithm cycles error (σx ) is given by fi=1 and weights ~ ~ document. Left) Greenland in Equation 6.6-15 with and , where the a and through all of the descending (arbitrarily decided; δρ, a δρ, d resolution of ice velocity estimates to help improve our example with slow moving interior −1 d subscripts indicate the offsets for the ascending and understanding of how the ice sheets will contribute to ascending first will work just as well) images. For each N velocity and fast moving glaciers 1 ⎛ 1 ⎞ descending orbits, respectively. sea level change. of these descending images, the program then loops wx,i = 2 ⎜ ∑ 2 ⎟ derived from ALOS PALSAR and σ ⎝ i=1 σ ⎠ over all of the ascending images to determine if there x,i x,i 6.6-20 RADARSAT-1 tracks. Middle) ICE VELOCITY MOSAICKING COMBINED ESTIMATE is overlap. If there is overlap, then the code identifies Antarctic with ice sheet interior where the region of overlap falls in the output grid. If feathering (see below) is applied (f ≠1) then and fast ice streams example The sections above describe how to measure velocity As described above, there are multiple methods for i from RADARSAT-2, ALOS PALSAR, Next, the algorithm loops over these output points at a point given the relevant phase or offset data. Rath- determining velocity at each point using phase or off- −1 and computes the surface-parallel-flow approximated f ⎛ N f ⎞ Envisat ASAR and ERS-1/2 tracks er than point measurements, what is required for ice sets from single or crossing orbit tracks. To apply these i i fi wx,i = 2 ⎜ ∑ 2 ⎟ using speckle tracking. Latitudes velocities using Equation 6.6-15, using either phase or σ σ sheets are continental-scale mosaics stitched together methods, a mosaicking algorithm is needed to produce x,i ⎝ i=1 x,i ⎠ from the pole to red line in each range-offset data. Above we have assumed that where 6.6-21 from data derived from hundreds to thousands of SAR a large-scale mosaic, using the best data available at image will not be observed with phase data are available, they are available for both image pairs. Such algorithms are relatively mature and each point. Here we describe an approach to mosa- the left-looking mission. Errors ascending and descending passes and, if not, then In practice, the mosaicking algorithm doesn’t know ice-sheet wide mosaics have already been produced icking the data based on a specific implementation of in these maps do not meet range-offset data are available for both passes. There how many estimates are available at any given point from earlier sensors (e.g., Figure 6-13; Joughin et al., a processor, which implements all of the algorithms NISAR requirements. With 30+ can be cases, however, where range-offset data only in the output grid. As a result, it weights each estimate 2010; Rignot et al., 2011a). While providing a major described above to produce a mosaic (Joughin, 2002). acquisitions a year along each are available from one track direction and phase data v by f /σ2 and sums the result in the output buffer. leap forward in our understanding of ice sheet behavior, Variations on this approach exist (Mouginot et al., x,i i x,i track and phase data, NISAR will from the other. In this case, there is nothing to preclude At the same time, a separate buffer is maintained and products from existing sensors are limited in accuracy 2017). meet its stated requirements. ⎛ N f ⎞ using Equation 6.6-15 with range-offset from one track i the weights are summed ⎜ 2 ⎟ . When all data are Right) Coverage at the north pole. by insufficient data collection from instruments not ∑σ direction and phase data from the other. ⎝ i=1 x,i ⎠ Because NISAR will be left-look- optimized for this type of mapping. Temporal resolution included in the mosaic, the weighted average is coming the northernmost part of of these products is also limited by a dearth of data pleted by normalizing the final result by the summed As just described, for each pair or crossing pair the Greenland will not be imaged. Sea weights. algorithm estimates, v and v , at each point in the ice outside the red circle will be x y imaged, but not inside. output grid. For coastal velocities, there may only be a 100 NISAR Science Users’ Handbook 101 NISAR Science Users’ Handbook

The error estimate for the weighted average is given by weights in the temporary buffers, and the results are materially considered to be a brittle solid with some discontinuous nature of sea ice. Thorndike et al. (1975) added to the weight buffers as indicated in Equation plasticity and its motion is spatially discontinuous. It and Haas (2010) conceived of the temporal develop- −2 N ⎛ N f ⎞ N f 6.6-19. is forced by winds and currents, which results in both ment of ice thickness distribution, sg/st, which can be σ 2 = f w σ 2 = i i . x ∑ i x,i x,i ⎜ ∑ 2 ⎟ ∑ 2 lead formation where new ice is formed and defor- written as: i=1 ⎝ i=1 σ x,i ⎠ i=1 σ x,i 6.6.3 Planned Output Products mation that produces ridges and complex motions dssgg −–−dss((ffgg)) 6.6-22 including rotation, shear, and vorticity. == ++ddiivv((vvgg))++ΦΦ The Science Team shall produce the following ice-sheet dsstt dsshh demonstration/validation products: As a result, error estimates are cumulated by summing 6.7.1 Theoretical Basis of Algorithm Three terms contribute to this equation (Figure 6-14): f /σ2 in error buffer, and the results are normalized as • Ice sheet velocity products at time scales of 12 f(h, x,t) = dh/dt is the thermodynamic growth or melt i x,i In this section, we describe a fundamental concept of indicated in Equation 6.6-22. days to a year for validation purposes. Examples of rate of ice of thickness h at a location x and time t, v is sea ice parameters that explains the role of sea ice such products are velocity maps covering the GPS the ice drift velocity vector, and  is the so-called re- motion and deformation from SAR. FEATHERING validation sites, areas that overlap with coverage distribution function. In general, the thinner and thicker provided by other sensors, and regions of ice-free components of the thickness distribution arise from dy- While the weighting method described above is SEA ICE THICKNESS DISTRIBUTION stationary areas (e.g., bedrock outcrops). namics and the median values from thermodynamics. designed to achieve a minimum variance estimate, it After Haas (2010), the sea ice thickness distribution In Figure 6-14, thermodynamic growth is faster for thin may be sub-optimal with respect to other factors. In • Differential tidal displacement maps to validate is probably the single most important parameter of ice than thick due to steeper temperature gradients particular, a discontinuity at a data-take boundary is grounding line requirements. sea ice and its role in the global climate system. The between the ocean and atmosphere. The presence of a non-physical result and can lead to problems when • Velocity estimates to validate the mountain glacier distribution is comprised of both dynamic and ther- snow reduces ice growth and pressure ridges (keels) attempting model inversions. As a result, additional measurement goals. modynamic processes and presents the aggregated, weighting is employed to “feather” the data and redis- • A limited set of demonstration products within tribute local errors over a wider range. This additional budgetary limitations. weighting function is used to apply a linear taper from the edge of the data to some distance from the edge. [ Figure 6-14] SEA ICE THICKNESS DISTRIBUTION These products are designed to validate the L2 require- This is accomplished by applying a distance transform Illustration of the ments, but not to completely fulfill them. While data will g(h) that gives the distance, d, at any point in the interior contribution of the be collected to meet the requirements throughout the to the nearest point on the image edge. The feathering different terms and Mode mission, the bulk of the processing to fully meet the function is then given by processes in the equation requirements will be carried out by the scientific com- above to the ice munity at large, with funding external to the project. ⎧ d thickness distribution. Mean ⎪ 0 ≤ f ≤ 1 f l f (d) = ⎨ l . After Haas (2010) and ⎪ 1 f ≥ 1 6.7 CRYOSPHERE PRODUCTS- ⎪ l Thorndike et al. (1975). ⎩ SEA ICE 6.6-23 The basic concepts of sea ice motion are position, h h + dh This is similar to the feathering scheme using for the displacement, and velocity. Displacement is the differ- PROCESSES THAT ALTER THE THICKNESS DISTRIBUTION SRTM mosaicking. Note the distance transform is ence in position over time of an ice feature. Velocity is applied to the source data, so the feather length, f, is in derived from displacement during that the measured l f(h) g div n f (h) n radar coordinates. This function is applied as indicated time interval. Sea ice moves within the general ocean by Equation 6.6-19. circulation forced by winds and currents but also at the smaller scales of individual floes, aggregates of

As an example, if the feather length is 20, then pixels floes, and the formation of leads (or open water). Ice h h h on the edge are weighted by 0, pixels within 20 pixels motion controls the abundance of thin ice and surface of the edge are weighted linearly with distance from exchange processes including heat flux between the

the edge over a range from 0 to 1, and interior pixels by ocean and atmosphere and ice production. Sea ice is THERMODYNAMICS DIVERGENCE DEFORMATION 1. The feathering weights are used to update the initial 102 NISAR Science Users’ Handbook 103 NISAR Science Users’ Handbook

may exceed a depth that will lead to melt if the depth it depends on fracture mechanics and other factors er-latitude oceans. This is also called the advective part deformation, which impact the strength of the ice and extends down into the warmer ocean layers. including small-scale ice properties, friction between of the sea ice momentum balance, which represents its thermal properties over a wide range of temporal ice blocks, snow and ice interfaces and deformation the velocity boundary condition of the ice-ocean sur- and spatial scales. Accurate quantification of sea ice The second term in the equation above represents energy and length scales. Thinner ice will deform more face (Figure 6-15). Included in the response of the ice dynamics, such as can be derived from basin-scale, divergence and advection due to ice motion, as forced readily than thicker ice. Within the redistribution term, cover to large-scale circulation is mechanical defor- rapidly repeating SAR observations obtained over by winds and currents. Away from the coast or even ice strength and rheology are of great importance. mation, which describes ice floe to ice floe interactions multiple years, as well as thickness, currently derived from the ice pack itself, ice will drift freely, and drift Models were derived that consider ice rheology as a (convergence and divergence (Figure 6-15). by measurements of sea ice freeboard from altim- direction and speed are closely related to geostrophic plastic or viscous-plastic (Hibler, 1979 and Coon 1980). etric observations, are crucial for understanding the wind (outlined below). Divergence generates cracks, The rheology describes a viscous flow of an ice field, From deformation, individual floes may aggregate into behavior and the vulnerability of the Arctic ice cover in leads, or polynyas, all areas of open water where new with plastic deformation once ice concentration and larger floes, areas of open water may form as floes a warming climate. ice will form. Divergence changes the sea ice fraction internal ice forces exceed a certain threshold, driven move apart which then become locations of intense of an area or grid cell, either removing ice of finite by winds and currents. Contemporary models include heat flux between the warmer ocean and the relatively SEA ICE DEFORMATION thickness and causes a delta signal at zero thickness in coupled atmosphere-ice-ocean conditions (e.g. Zhang cooler atmosphere where new ice rapidly grows, and Sea ice motion and deformation have multiple forcing the thickness distribution (Figure 6-14 middle). et al., 2000; Holland et al. 2006). ridges may form as floes are forced together, where the functions including the large-scale Coriolis force, air thinner floe breaks-up into smaller pieces and blocks and water drag arising from the variable top and bot- The last term in the above equation is the redistri- ICE MOTION AND OBSERVATIONAL BASIS which pile up around the perimeter of encountered tom ice surface topographies, the ocean gravity field, bution function describing how thin ice is deformed thicker floe (Figure 6-15). A key aspect of the mechan- Winds and ocean currents continually move the sea ice and gradients of ice stress resulting from floe-to-floe and transformed into thicker ice classes from both ical response to larger-scale forcing is that the ice cover at multiple scales. The larger geostrophic scale interactions and even within individual floes (Holt et convergence and deformation. It is the most critical response (opening and closing, shear) is concentrated of motion describes the regional transport of sea ice al., 1992; Kwok, 2001). These forces vary by ocean term to understand the temporal development of the along fractures which are tens of meters to kilometers across the ocean basins as well as the export to low- basin as well. The sea ice cover within the Arctic Ocean thickness distribution and also most unknown, since wide, and lengths that can span thousands of kilo- is contained within surrounding continental bound- meters. This mechanical response that ranges from aries, with only outlets in the Fram/Greenland and floe-floe interactions along extensive, narrow fractures Bering Straits, resulting in more convergent circulation have been shown to be optimally observed by rapid- Winds flows and resulting deformation (Wadhams, 2000). ly-repeating SAR observations acquired at ocean basin In contrast, the sea ice cover in the Southern Ocean scales on a continuing, seasonal basis. surrounds the Antarctic continent and has a more divergent, northerly drift circulation patterns towards the DIVERGENCE CONVERGENCE At the end of summer melt, the sea ice cover has [ Figure 6-15] surrounding ocean with reduced deformation impacts thinned and reduced in extent, exposing more open Illustration of the Ridge Sails compared to the Arctic (Kottmeier et al., 1992). water. As temperatures fall below freezing, the ice processes that dynamically cover will cease melting and new ice growth will (by divergent or convergent Mechanical deformation results in divergence and con- commence when the summer-warmed ocean water ice motion and defor- Ice Floe vergence motion (openings and closings) that produces Ice Floe Ice Floe sufficiently cools below freezing. Sea ice grows rapidly mation) modify the ice area changes and shear of the ice pack (represented in but the still thicker ice cover may move without much thickness distribution GENERATION OF OPEN WATER a simple schematic in Figure 6-16). Ice floes also resist (After Haas, 2010). –> NEW ICE GROWTH resistance and break-up the thinner ice. This is contin- motion, termed the strain rate, which is the spatial uous process of opening and closing of floes along with variation in ice velocity. These relative motions occur new ice growth and fracturing gradually slows over on both small floe-to-floe scales but may also occur the winter months as the ice continues to expand and Ridge Keels along extensive fracture zones. When combined with thickens. Over time, the volume of sea ice is continu- ice growth thermodynamics, a complex interplay of Currents PRESSURE RIDGE FORMATION ously altered through this ongoing redistribution of sea ice between thermodynamic growth and dynamics of 104 NISAR Science Users’ Handbook 105 NISAR Science Users’ Handbook

ICE MOTION, DEFORMATION Divergence, ∇ ⋅ u, is a measure of area change. detect ice deformation, which can range from sub-daily area that form ridges. However, Lagrangian tracking Vorticity, ζ, is the principle measure of rotation. Shear, time scales associated with tidal forcing or inertial adds complexity to the ice-motion tracking process and A A’ Velocity e, is the scalar magnitude of shear. Figure 6-16 is a scales (Kwok et al., 2003). Presently, basin-scale fields quality checking, since the connectivity of each array [ Figure 6-16] schematic of motion and deformation concepts from a of sea ice motion at different spatial resolutions can must be maintained along each time step. Both Euleri-

Illustration of the taxon- sequential image pair. be derived by tracking similar ice features with polar an and Lagrangian type products have been generated orbiting sensors with moderate (1-4 km) or coarse previously using Radarsat-1 and Envisat ASAR for the omy of ice motion and A deformation of interest When combined with thermodynamics, i.e. ice growth (> 10 km) resolutions (Emery et al., 1997; Spreen et Arctic and limited portions and times of the Southern Shear in atmosphere-ice-ocean B and melt, how do measurements of dynamics contrib- al., 2009). As shown by numerous studies (Kwok et al., Ocean sea ice cover and are available at the Alaska interaction modeling. ute to the sea ice thickness distribution, given that ice 1990; Holt et al., 1992; Kwok et al., 1995; Kwok and Satellite Facility. For NISAR, the focus for development exists as an aggregate composed of multiple forms of Cunningham, 2002; Kwok, 2001), satellite SAR data and post-mission validation will be the use of Eulerian Divergence -> Area Change, A OR B sea ice in terms of thickness and age? Motion or ve- with resolutions on the order of tens of meters along tracking. Opening Or locity, distance traveled over time, simply indicates that with its capabilities to operate without sunlight and Closing ice is forced by wind and currents and is not stationary through clouds, and good radiometric performance, are For a pair of SAR images of sea ice, grid cells with ini- in space. When two pieces of ice move apart from each well suited for observations of both drift and defor- tial dimensions of 5 of 10 km on a side are commonly other, an opening or lead is formed, exposing the ocean mation, with the proviso that rapid repeat sampling is used to sample the motion and deformation of the ice A A’ Vorticity directly to the atmosphere. In winter, ice grows rapidly routinely obtained. cover. Past results show that basin-scale deformation -Rotation within the lead while In summer, the open water will be of the divergence, vorticity, and shear of the ice cover warmed preferentially to the ice by solar radiation, en- EULERIAN AND LAGRANGIAN ICE MOTION may extend across a significant distance of the sea hancing ice melt. Openings can occur within a defined ice cover across the Arctic (Kwok, 2001; Kwok and There are two general ways of sampling a sea ice mo- area or cell or between adjacent cells and is defined Cunningham, 2002). The high-resolution ice motion motion and growth produces and maintains the thick- tion field: Eulerian and Lagrangian. Eulerian sampling as a fractional increase in area. When two pieces of ice drift vectors derived from this approach have a data ness distribution of both polar sea ice covers. Sea ice observes the movement or offset of an individual sea are forced together, this represents a loss in area within quality comparable to that from buoy drifts (~0.1 cm/s) circulation models generally contain only large-scale ice floe or ice area/grid between a pair of images (A a cell or between adjacent cells which usually indicates and have provided an unprecedented level of spatial sea ice motion fields along with ice growth models. and B) over time. This observation is then repeated by ridging process, where the thinnest components of and temporal detail of deformational features. Narrow Sea ice dynamic models include both thermodynamics resetting the grid or individual ice particles in image sea ice will preferentially break-up and be piled up fracture zones have been shown to extend for thou- and multiple scales of sea ice motion and deformation B and then deriving motion in the next image C. The into pieces which stack up both on top (ridge sails) sands of kilometers and these fracture patterns appear fields to capture overall sea ice thickness distributions. Eulerian approach provides velocity and deformation and bottom (ridge keels) of the remaining, thicker ice more linearly oriented rather than as random patterns. SAR-derived ice motion and deformation have been for each pair of images through time. By contrast, floe (Figure 6-15). Shear and vorticity represent other Most of the SAR-derived ice motion products to date proven to be fundamental in providing dynamic fields Lagrangian sampling involves deriving the motion field components of deformation (Figure 6-16) are closely have focused on the western Arctic Ocean, where the necessary measurements of spatial distribution over time through multiple observations without reset- related to adjacent divergence but have less effect on continuous SAR coverage of the sea ice has been and temporal development of these processes (Holt et ting the initial grid. This approach has the advantage of the overall sea ice thickness distribution. more frequently and continuously available. Providing al., 1992; Kwok, 2001). providing motion and deformation over time as well as complete maps of the Arctic Ocean as well as totally other aspects including relating area changes to either As discussed in Holt et al. (1992), the varying spatial new motion/deformation mapping of the dynamic sea The basic forms of sea ice deformation are divergence, opening and closing of the floe field (Kwok et al., 1995; and temporal scales of small-scale ice motion are ice cover surrounding Antarctica will provide complete vorticity, and shear, as below: Stern et al., 1995). Since sea ice is a brittle solid, it difficult to observe without adequate imaging or arrays information on the motion and deformation of the entire u , u , v , v are the spatial gradients in ice motion does not deform continuously throughout the ice cover x y x y of buoys placed on the ice. The relative motion between sea ice cover for the first time. Continuing coverage of but along cracks and fractures which indicates varying computed using a contour integral around the bound- ice floes along narrow (meters to kilometers) fractures the Arctic by the Sentinel-1 series and the upcoming patterns of ice strength and weakness. From the de- ary of an area of ice, or in terms of SAR, a grid cell requires imaging sensors with high spatial resolution launch of the Radarsat constellation mission (RCM) tailed mapping of area change, areas of opening can be (~10 km on a side). The boundaries are defined by the but also obtained at sampling intervals sufficient to opens the possibility of obtaining daily observations of used to estimate new ice growth over time while areas line segments connecting the four vertices of a cell. SAR-derived ice motion and deformation, a product that of closing indicates a loss of the thinnest ice within an lends itself to inclusion within circulation models. 106 NISAR Science Users’ Handbook 107 NISAR Science Users’ Handbook

6.7.2 Implementation Approach for respectively (Kottmeier et al., 1992). While the physical Spatial differences in displacement between two fea- The Lagrangian products are fundamentally equivalent Algorithm processes are many, on the time scales of days, more tures (6.7-4 and 6.7-5) are in terms of the ice displacements but ,because these than 70% of the variance of the ice motion is explained are generated based on observing and maintaining The derivation of sea ice motion and deformation by the geostrophic wind alone. As will be described in a Δu = ⎡ x + e − x + e + e ⎤ the original grid area over time even when the ice employs the use of feature tracking using cross cor- ⎣( b2 gb2 ) ( a2 ga2 ) f 2 ⎦ later section, this relationship of ice motion and geos- undergoes deformation, additional valuable products relation. Basin-wide seasonal sea ice motion fields are − ⎡ x + e − x + e + e ⎤ trophic wind are used in the ice motion algorithm to do ⎣( b1 gb1 ) ( a1 ga1 ) f 1 ⎦ are generated. Also, previous results using Radarsat-1 obtained using repeating image pairs obtained over a the initial identification of the second of the image pair 6.7-6 imagery performed Lagrangian processing, so much time separation of a few days. to be used for tracking, guided by weather data. of the documentation and subsequent literature is The error in u has zero mean. Its variance contains a based on these multiple products. The general algo- BASIC SEA ICE MOTION CONCEPT contribution from each of the tracking errors e and e GEOLOCATION ERRORS f2 f1 rithm for both types of processing are discussed in that are independent. If the geolocation errors are all As described in Holt et al. (1992), the fundamental this document. The two primary sources of error in measuring independent then their variances all add and the vari- concepts of ice motion are position, displacement, and ice motion with satellite imagery are absolute geo- ance of u is 22. This quantity is an upper bound on velocity. With an initial ice field at position X at time u There are several fundamentally key components of graphic position (e ) of each image pixel and a tracking this error when the points are separated by hundreds t=0, the ice field has moved to a new position x(t: X) g the NISAR mission that make it particularly valuable for error (e ), which is the uncertainty in identifying of kilometers. However, if the two features are close, over some time interval. The measured displacement is f deriving sea ice motion and deformation, that will lead common features from one image to the next (Holt et the geolocation errors are no longer independent and in the difference in the positions of an ice field or floe at to the derivation of uniquely valuable sea ice products. al., 1992; Kwok and Cunningham, 2002). The position fact tend to cancel; the error variance of u tends to- two different times. First, the longer frequency of L-band has been shown error applies independently to each position in each wards the lower bound 22. This means that even if the f to highlight deformed ice preferentially compared to the image, i.e., a position is the true position plus an error geolocation errors are large, differential motion or de- u = ⎡x t +1 − x t ⎤ long and extensive C-band SAR record. This is expected ⎣ ( i ) ( i )⎦ x=constant* of x + e . The tracking error ef applies to a displace- formation can be estimated well, even if displacement 6.7-1 g to provide improved and more accurate sea ice feature ment observed between two images. If it is assumed cannot. The mean spatial gradient over the distance tracking in the winter and spring and importantly for that eg and ef are each normally distributed with zero between two features (x) can be calculated by u /x. The average velocity over the intervening time interval, a longer and more continuous period into the summer bias, have standard deviations s and s and are Lindsay and Stern (2003) found displacement errors T = t − t , is g f months, where ice surface features on C-band become i+1 i uncorrelated between two time-separated images between SAR and ice drifting buoys to be on the order u less distinct due to surface melt. Next, the synoptic = A and B, the two errors can be treated separately. of 0.2–0.3 km, derived from sensor geolocation errors V 6.7-2 coverage of the entire Arctic and Southern Ocean sea T Including errors, an estimate of the displacement of an of 0.1 km, tracking errors from 0.1–0.3 km, and pixel ice covers during the entire duration of the mission ice feature is given by resolutions from 0.05–0.1 km. This is a measure of su As described in Kwok et al., (1990) and Holt et al. will provide unprecedented SAR coverage of both polar assuming the buoy positions are correct. In the Lind- (1992), the linear model of ice motion relates the mean regions that can be used for ice motion. Radarsat-1 say and Stern study (2003), buoy location errors were ice velocity, v, of an ice field to the geostrophic wind G u = x + e − x + e + e provided annual ice motion mappings of much of the ( b gB ) ( a gA ) f 6.7-4 on the order of 0.3 km. Current buoys using GPS have (Thorndike and Colony, 1982; Colony and Thorndike, western Arctic but never complete and continuous reduced location errors down to about 10 m or less. 1984), where v, the ice velocity, and G, the geostrophic coverage for multiple years over the entire Arctic. Ice The error in velocity is su divided by the displacement wind, are vectors and consequently treated as complex motion of Antarctic sea ice from SAR has been limited time interval which usually has only small errors. NISAR SEA ICE MOTION RETRIEVAL numbers. The term A is a scaling factor |A| giving the v = AG ALGORITHM to date to only one to two mappings for periods of a ratio of ice speed to wind speed and an ageostrophic few month from the Ross or Weddell Seas. The sea ice σ 2 = 2σ 2 +σ 2 6.7-5 drift angle q (positive counterclockwise) from the u g f The NISAR science team requirement is to produce motion mapping of the Southern Ocean from NISAR will wind vector to the ice vector. Typical values for the Eulerian sea ice motion products due to its expediency be unprecedented and will enable a thorough derivation The error in velocity is  divided by the time interval Arctic of (|A|, q) range from (0.011, -18°) in summer u in production, which requires a minimum of operator of the different ice dynamics from both polar seas. of displacement. Errors in the time interval are usually to (0.008, -5°) in winter in relation to the mean wind quality assurance and correction. It is expected at some negligible. speed (Thorndike and Colony, 1992). For the Weddell point that Lagrangian products will also be produced Sea, these numbers are 1.6% and 10–15° to the left, when supported by additional funding, such as from NASA MEaSUREs opportunities. 108 NISAR Science Users’ Handbook

The algorithm to be used has been described in 6.7.3 Planned Output Products multiple publications based on the use of ERS-1 and NISAR will produce ice motion products for the Arctic primarily Radarsat-1 SAR imagery (Kwok et al., 1990; and Southern Oceans. It will also produce demonstra- Holt et al., 1992; Kwok et al., 1995; Kwok and Baltzer, tion products of seasonal maps of sea-ice motion for 1995; Kwok and Cunningham, 2002) and will be the Arctic Ocean and Weddell Sea and export for the modified to incorporate the NISAR image format and Arctic Ocean. metadata. Summarizing, the input images are first mapped to a polar stereographic projection. From there, EULERIAN ICE MOTION PRODUCTS the algorithm has an ice motion estimator to locate processed image pairs using a linear ice drift model a. Displacement (x, y, km)

based on geostrophic wind fields. Following that, the b. Ice motion vector (direction, deg) search identifies image pairs with maximum overlap c. Deformation (spatial variation of velocity): and minimum time spacing between acquisitions. A shear, divergence, vorticity uniform 5 km grid is then overlaid on the source image. The initial tracking for both the central and marginal ice zones uses both areal correlation and feature matching. Seasonal examination of Arctic region where there is The algorithm uses multiple filters at several stages a large density of sea ice drift buoys will be done each of the tracking process to remove low-quality vectors, year. This will be performed over a region of Antarctic based on correlation statistics. Dominant motion sea ice cover if sufficient number of sea ice drift buoys clusters are located in the sampled output, after which are present. additional filtering discards erroneous vectors by identifying cluster centroids that are inconsistent with the DEMONSTRATION PRODUCTS dominant clusters. A smoothness filter is then applied a. First product (Year1-Year2). Map of one full to ensure the spatial consistency of the displacement season of sea ice motion for the Weddell and field. A flag is assigned to each vector to give a value Arctic sea ice covers. indicating the quality of each derived vector. The whole b. Second product (Year2-Year3): Map of one filtering process in the algorithm provides output such full season of sea-ice motion for the Southern that 95% of the motion vectors are accurate based on Ocean and map of one full season of sea ice the derived displacement error. motion export from the Arctic Ocean. NISAR Science Users’ Handbook 111 NISAR Science Users’ Handbook

7 ERROR SOURCES

This section describes errors in the measurements noise is by definition target-dependent, and therefore that impact science performance. Understanding these can be correlated with the signal of interest. sources of errors will help users interpret NISAR data. These errors can be related to instrument noise, geo- In polarimetry, we observe the covariances of like- metric considerations, scattering behavior, or propaga- and cross-polarization images and use these to tion effects, to name a few. infer properties of the surface. To reduce the natural variance of the covariance estimates, we typically must 7.1 POLARIMETRIC ERROR SOURCES average data over a local region. Thus, to achieve good estimation performance at a desired resolution, the The radiometric properties of a surface, represented observations must be acquired at finer resolution to by the observed backscatter amplitude and phase, are allow for such averaging. The number of independent characterized through the radar backscattering cross resolution elements averaged is generally referred to as section. The amount of energy scattered back to the the number of looks. radar depends on the detailed arrangements of scatterers within a resolution element and their electrical Besides errors related to element-by-element random properties, so in general the cross section is dependent noise, the other major sources of error in polarimetry on the observation angle and environmental conditions. are systematic amplitude and phase fluctuations that Since radar images are coherent, they exhibit “speckle” vary over the image and potentially over time. These properties: even in a region with multiple distribut- arise from uncertainties in the knowledge of the ed scatterers with uniformly constant radar cross radar’s system delays and losses, its antenna pattern section, each resolution element will exhibit amplitude or the pointing of the antenna pattern. To derive the and phase variations that differ wildly, such that the radar cross section, the total “link budget” from signal images appear to be spatially random from element to transmission to backscatter to its reception must be element. This natural variance, coupled with random quantified according to the radar equation, which in- noise sources in the radar system, requires describing volves these quantities. Thus, the radiometric accuracy radar cross section as a statistical process, using the requirements arising from the science requirements covariances of the observed quantities. imply knowledge and stability requirements throughout the radar system. The element-to-element random error from the radar system includes additive thermal noise in the radar Overall, the error in the radar backscatter measure- system, and multiplicative noise from quantization, like- ments, Δσ (pq=hh,hv,vv), is a function of speckle, and cross-channel signal leakage, and ambiguities, pq thermal noise, temporal variability of the backscatter, which are ghosts of pulse echoes taken at a different calibration errors (which in turn depend on pointing place and time but show up in the data. Multiplicative DEM errors) and area projection correction terms. An 112 NISAR Science Users’ Handbook 113 NISAR Science Users’ Handbook

expression for this error (in terms of the measurement parameters) is given by Hensley et al., 2013: correlated but random component to the differential sidual noise term that includes scattering variability and signal that is one of the chief limiting noise sources. thermal noise. All of the noise terms contribute to the Speckle Thermal Backscatter Temporal Calibration Area Projection Noise Noise Variability Errors Errors The wide bandwidth of the radar data can be exploited signal quality, as quantified by correlation, at various to estimate signal dispersion due to the ionosphere spatial scales. The correlation can be expressed as: such that this dispersion can be mitigated (Meyer et al., ⎛ 1 1 1 1 1 ⎞ 1 1 A 1 1 Δσ = ⎜ + ⎟σ + Δσ b + Δσ + dem Δσ 2011). We can mitigate the effects of atmosphere prop- pq pq pqt ( ) c a γ e = γ SNRγ Bγ Vγ φγ T ⎝ N N N N SNR⎠ N N Apix N N os ot ot ot os agation noise through a combination of modeling using

independent estimate of the state of the atmosphere where the correlation terms are defined in Table 7-2. where Table 7-1 defines the symbols: and through an averaging or filtering process that The table provides formulas illustrating the dependency assumes a spatially correlated but temporally uncor- of the various correlation terms in terms of system TABLE 7-1. BACKSCATTER ERROR MODEL DEFINITIONS related random process, as distinct from the ground parameters. Symbol Definition motion which is generally both spatially and temporally N Number of spatial looks per observation correlated. The displacement noise corresponding to this correla- Δϕ = Δϕ + Δϕ def err tion is given by Not Total number of observations = Δϕdef + Δϕatmos + Δϕorb + Δϕtopo + Δϕn 2 λ 1 1−γ e Nos Observations with non-correlated speckle σ d = 2 where ∆φdef is the phase due to the true ground defor- 4π 2N γ e mation in the LOS direction, ∆φatmos is the phase due Noi Observations with correlated speckle When the correlation is low, the displacement noise is to the tropospheric and ionospheric delays, ∆φorb is the high and vice versa. N is the number of pixels that can SNR Signal-to-noise ratio phase due to satellite orbit errors, ∆φtopo is the phase be averaged to reduce the noise level. due to error in the surface topography, and ∆φn is a re- Δσpqt Backscatter temporal variability

Δσc Backscatter calibration error TABLE 7-2. ELEMENTS OF THE INTERFEROMETRIC CORRELATION Correlation Term Expression System Dependence Δσa Backscatter area projection error

Total SNR γ e = γ SNRγ Bγ Vγ φγ T Signal-to-noise-ratio Adem Area of a pixel in DEM used for slope computations

Apix Area in a radar pixel SNR γ = SNR Short baseline , fine resolution δ , and long wavelength λ SNR +1 B g Geometric Baseline from maximize correlation. Look angle θl and range ρ are relatively 2Bcosθ 2δ 7.2 INTERFEROMETRIC antenna pattern will cancel in the phase difference. l g fixed in useful orbits with low drag (above ~ 600 km) γ B = 1− ERROR SOURCES In practice the system will not be perfectly pointed, or λρ the antenna patterns and system phases will vary over As with polarimetry, random resolution element-to- ⎛ kz hc ⎞ time. These differential phase effects typically have a γ = sinc Short baseline , and long wavelength maximize correlation. V ⎝⎜ 2 ⎠⎟ B λ element noise is introduced from speckle, thermal different nature from those due to ground motions and Geometric Volume from Look angle θl and range ρ are relatively fixed in useful orbits effects, and multiplicative noise sources. These are ⎛ B ⎞ are tied to the geometry to the spacecraft orbit, so they k = 4π with low drag (above ~ 600 km) z ⎜ λρ sinθ ⎟ quantified by the interferometric correlation, which can often be mitigated in scientific data reduction. ⎝ l ⎠ is the amplitude-normalized cross covariance of the

Small pointing rotation f rot, fine along-track resolution δaz, interferometric observations. As with polarimetry, local Another effect of importance is the phase delay 2sinθ φ δ and long wavelength λ maximize correlation. Look angle θl averaging reduces this noise component. experienced by the electromagnetic wave propagating Geometric Rotation from γ = 1− l rot az B λ and range ρ are relatively fixed in useful orbits with low drag through the ionosphere and the neutral atmosphere. (above ~ 600 km) The broader systematic effects on the phase difference The state of these media changes rapidly over time, so are important in interferometry. Since it is a differential 2 every time an observation is made (i.e. every 12 days ⎛ 4π ⎞ 2 Depends on natural targets. Longer wavelengths decorrelate −⎜ ⎟ σ los measurement, if the system is stable and the pointing from a given vantage point), the phase delay across Temporal ⎝ λ ⎠ less for a given surface change, proportional to wavelength γ T = e is perfect over time, phase due to system delays or the image will be different. These introduce a spatially squared in general 114 NISAR Science Users’ Handbook 115 NISAR Science Users’ Handbook

Interferometric performance depends critically on how Imaging Spectroradiometer (MODIS) and GPS data to cluding SBAS, NSBAS, MInTS and various permutations mation. Such algorithms to mitigate tropospheric noise can estimate the water vapor field in order to correct inter- of these approaches. These algorithms can filter out have been shown to be very effective, even in reducing well the total interferometric phase difference jint be measured, which in turn depends on the SNR. We ferograms that are corrupted by atmospheric artifacts. tropospheric delays in InSAR data assuming that errors the very short wavelength TerraSAR-X data. Thus, with can relate SNR to the phase-difference measurement Foster et al. (2006) employed a high-resolution weather in InSAR deformation estimates are primarily due to long and dense time series, we can address many state uncertainty  . The variance of the measured phase model to predict tropospheric delays for the acquisition tropospheric noise that are uncorrelated in time. These of the art problems and applications. φint difference, 2 , is due to the random phase compo- times of SAR images. However, estimating tropospheric methods require many observations at frequencies φint nent introduced by the noise accompanying the signal, delays using auxiliary data such as GNSS, MERIS/MO- much greater than the expected time scale of defor- and it is approximately proportional to the inverse of DIS or weather model usually produces a tropospheric the SNR, noise model with resolution much coarser than InSAR 1 image resolution, and the model uncertainty can be σ 2 ≈ ϕ int SNR relatively large for studying centimeter-level crustal deformations. So, for example, to secure the single-look value  = jint 0.1 rad, it is necessary that SNR = 100, or, equivalently, Many have proposed algorithms to estimate tropo- 20 dB. spheric delays during SAR data acquisition times directly from InSAR data. Emardson et al. (2003) Phase artifacts in InSAR images are often attributed mitigated tropospheric effects by averaging N inde- to neutral tropospheric delays (Zebker et al., 1997; pendent interferograms because the wet component of Hanssen et al., 1998). Because the Earth’s troposphere the neutral atmosphere is uncorrelated at time scales is non-dispersive at appropriate frequencies, radar longer than 1 day. This stacking approach is limited by signals that operate at different frequencies are subject the number of interferograms that are available over to the same tropospheric delays. For a typical X-band the time of interest. Lin et al. (2010), Lauknes (2011), interferogram (such as TerraSAR-X), a phase cycle of and Hooper et al. (2012) assumed that tropospheric 2π in the interferogram corresponds to λ/2 = 1.55 cm delays in InSAR data are topographically correlated deformation, where λ is the radar signal wavelength. and can be partially removed by knowledge of the In a typical radar scene, tropospheric noise occurs with local elevation changes. However, the assumption that variation on the order of centimeters or even greater tropospheric delay is proportional to surface elevation across the interferogram. As a result, any expected may not be valid for turbulent tropospheric processes. centimeter-level crustal deformation signature is Use of globally available weather reanalysis models obscured by tropospheric noise. (e.g., ECMWF and NARR) has also shown considerable ability to mitigate topographically correlated phase In order to obtain accurate InSAR deformation mea- errors – with the advantage of not absorbing potential surements, some effort is needed to handle or suppress geophysical signals into empirical corrections (e.g., the atmospheric noise signature. Onn and Zebker Jolivet et al., 2014a). (2006) introduced a method to correct for atmospheric phase artifacts in a radar interferogram using spatially Since many of the problems proposed by the science interpolated zenith wet delay data obtained from a team for this mission require correction at the mm to network of GNSS receivers in the region imaged by the cm level, a more complex approach will be required. radar. Li et al. (2006a, 2006b) used Medium Resolution A variety of InSAR time series algorithms now exist in- Imaging Spectrometer (MERIS), Moderate Resolution NISAR Science Users’ Handbook 117 NISAR Science Users’ Handbook

8 CALIBRATION AND VALIDATION

Calibration and Validation for NISAR comprises in- A useful reference in developing a validation plan is the strument calibration, image calibration, calibration of CEOS Hierarchy of Validation:

algorithms used to derive higher level science products • Stage 1: Product accuracy has been estimated us- such as biomass or glacier velocities, as well as demon- ing a small number of independent measurements stration (validation) that the data acquired, when flowed obtained from selected locations and time periods through the science processing algorithms, create and ground-truth/field program effort. products that meet the mission’s science requirements. • Stage 2: Product accuracy has been assessed Instrument calibration is generally addressed in the pre- over a widely distributed set of locations and time launch period through measurements made in a relevant periods via several ground-truth and validation simulated space-like environment. This section address- efforts. es the other elements of Cal/Val mentioned above. • Stage 3: Product accuracy has been assessed, and 8.1 BACKGROUND the uncertainties in the product well-established via independent measurements made in a system- In developing the Calibration/Validation plan for NISAR, atic and statistically robust way that represents there are precedents and experiences that can be global conditions. utilized. The Committee on Earth Observation Satellites (CEOS) Working Group on Calibration and Validation A validation program would be expected to transition (WGCV; http://calvalportal.ceos.org/CalValPortal/ through these stages over the mission life span. welcome.do) has established standards that may be The NISAR mission collaborates with the NASA Global used as a starting point for NISAR. The Land Products Ecosystem Dynamics Investigation Lidar (GEDI) mission Sub-Group (http://lpvs.gsfc.nasa.gov/) has expressed and the ESA BIOMASS mission due to complementary the perspective that “A common approach to validation science requirements for measuring above ground would encourage widespread use of validation data, biomass. It is likely that science operations for all three and thus help toward standardized approaches to missions will partly overlap in time. Therefore, joint global product validation. With the high cost of in-situ validation of biomass requirements may be possible data collection, the potential benefits from international and desirable. cooperation are considerable and obvious”.

DEFINITIONS Cal/Val has become synonymous in the context of remote sensing with verifying to be self-consistent the In order for the Calibration/Validation Plan to effectively suite of processing algorithms that convert raw data address the achievement of mission requirements, a into accurate and useful geophysical or biophysical unified definition base has to be developed. The NISAR quantities. This can include vicarious calibration, which Cal/Val Plan uses the same source of terms and defini- refers to techniques that make use of natural or artificial tions as the NISAR Level 1 and Level 2 requirements. sites on the surface of the Earth for the post-launch calibration of sensors, which is typically called “image NISAR Calibration and Validation are defined as follows: calibration” for SAR systems. • Calibration: The set of operations that establish, under specified conditions, the relationship between sets of values of quantities indicated by 118 NISAR Science Users’ Handbook 119 NISAR Science Users’ Handbook

a measuring instrument or measuring system and For calibration of radar specific-parameters, the Project instrument calibration and, 2) Validation of higher-level America and in other parts of the world, and these the corresponding values realized by standards. will either deploy or employ existing corner reflector data products (L3) with ground truth at selected vali- serve as natural validation sites. GNSS sites and arrays

• Validation: The process of assessing by inde- arrays, for example the array at Rosamond dry lake on dation sites. Instrument calibration stability is verified have been used for a number of years for this purpose. pendent means the quality of the data products Edwards Air Force Base, currently used for calibration by continuing to collect calibration data over sites used The density of GNSS stations is on the order of one per derived from the system outputs. of NASA’s airborne L-band radar instrument UAVSAR during instrument checkout, using same radar modes 10-20 km, which will allow validation of the upper end (Uninhabited Aerial Vehicle Synthetic Aperture Radar). as in nominal science operations (this is different from of the accuracy length scale. The shorter scales will For some science requirements, ground instrumenta- instrument checkout, when multiple modes are used be validated by analysis – examining areas known to 8.2 CALIBRATION AND VALIDATION tion will be deployed prior to launch to Cal/Val sites and for various calibrations). Table 8-1 lists the instrument be stable over a period of time and comparing the ex- ACTIVITIES verified with contemporary data sources. Contemporary parameters that will be calibrated post-launch by the pected noise performance to that measured. Since the Calibration and validation is divided into pre- and post- and historical data sets, especially L-band SAR and instrument and science team. errors tend to be dominated by environmental effects launch activities. Pre-launch activities focus on instru- time series data from Sentinel-1, will be compiled for like water vapor and surface decorrelation, what is ment calibration. Post launch calibration and validation Cal/Val sites; demonstration products will be developed The objective of science data product validation is to most important is to validate that the contributions of activities focus on the data products. for algorithm testing and verification. validate that global data yielded by NISAR will meet the instrument noise are within acceptable values allowing project’s L2 science requirements. L3 products will be the overall accuracies to be met. PRE-LAUNCH POST-LAUNCH generated by the science team at the selected validation sites. Validation of the L3 science products will be For ice sheets and glaciers, the higher level products During the pre-launch period, there are a variety of In the post-launch period, the calibration and valida- carried out by a combination of fieldwork and analysis. will be validated in the relevant environment of Green- activities that fall under calibration and validation. tion activities will address directly the measurement land and Antarctica. The plan calls for the deployment These mainly involve on-ground instrument calibration, requirements for the L1-L3 data products. Each data For solid Earth deformation, there are a number of of arrays of GNSS ground stations on a divide-to-coast algorithm development and evaluation, and establish- product has quantifiable performance specifications to natural validation sites in the world that can be used: flow line, through a variety of ice types to which the ice ing the infrastructure and methodologies for post- be met over the mission lifetime, with calibration and GNSS networked arrays exist throughout western North velocity products will be compared. launch validation. Requirements for Cal/Val related to validation requirements addressed in their respective specific NISAR data products will be identified by the ATBDs. TABLE 8-1. POST-LAUNCH CALIBRATION PARAMETERS AND METHODS respective science algorithm teams in their Algorithm Parameters Methods Theoretical Basis Documents (ATBDs). The production Post-launch calibration and validation activities are processing algorithms in the ATBDs will be coded and divided into three main parts after launch: Antenna Pattern and Use Amazon to compare residual brightness variations relative Beamforming to ideal tested in later phases of the project (prior to launch). 1. Three-month instrument checkout phase, after Pre-launch activities will include development of the Impulse Response which delivery of validated L1 products to the Measure 3-dB resolution, ISLR, PLSR of corner reflector arrays calibration procedures and algorithms for the NISAR public archive will begin. Multiplicative Noise Use a radar-opaque fence to measure total MNR plus thermal radar (L1 products), higher level image products (L2) Characterization noise 2. Five-month geophysical product Cal/Val phase, (incorporating such characteristics as geocoding and/or after which delivery of validated L3 products to Thermal Noise Use sniffer pulses to measure thermal noise levels multilooking), and the L3 products (which will be used Characterization the public archive will begin. to validate the NISAR science requirements). 3. Periodic Cal/Val performed annually. During Common Time Delay Compare range measurement on surveyed corner reflectors

Pre-launch instrument calibration will include modeling, this period, additional algorithm upgrades and Differential Time Delay Cross-correlate data between polarimetric channels to analysis, simulations, and laboratory and test-facility reprocessing of data products can be implement- measure channel misregistration ed if found necessary (e.g., as a result of drifts measurements. Algorithm development for all products Time Tag Compare along-track position measurement on surveyed will include testbed simulations, laboratory and test-fa- or anomalies discovered during analysis of the corner reflectors cility data, field campaigns, exploitation of existing science products), as well as validation of those Pointing Angle Biases Use bright homogeneous backscatter region to compare in-situ and satellite data, and utilization of instrument science requirements that require a year’s worth measured Doppler centroid to expected Doppler centroid and and geophysical models. of data or more. measure angle biases

Polarimetric Balance Use combination of polarization targets and distributed targets The science team will identify calibration and validation The main objectives of post-launch calibration/valida- to estimate polarimetric calibration sites and resources needed for post-launch calibration. tion activities are two-fold: 1) Monitoring the stability of 120 NISAR Science Users’ Handbook 121 NISAR Science Users’ Handbook

For sea-ice, the project will exploit buoy data in the ters, inundation threshold values, etc.) are to be TABLE 8-2. SUMMARY OF NISAR CAL/VAL VALIDATION SITES

arctic, comparing buoy velocities to measured ice calibrated and updated by Joint Science Team Measurement Validation Site Comment velocity vectors from the data. as validation data becomes available or at TBD Instrument Corner reflector arrays such as the Absolute radiometric calibration, intervals. Frequency of updates is TBD and may calibration Rosamond Corner Reflector Array, relative calibration, instrument For biomass and disturbance, there is a worldwide net- depend on the sensitivity of the algorithm and the California; Delta Junction, Alaska; performance, geolocation, beam work of managed and measured forests and fields that timing of the field campaigns Surat Basin, Australia; formation

provide a natural in situ data set against which to vali- • The L2 science requirements will be validated by Instrument Distributed targets in non-flooded, Cross-talk calibration, antenna date the biomass and disturbance products. The Project calibration non-deforested tropical forest loca- pattern, channel imbalance, rela- the joint science team. tions in South America and Africa tive calibration will support fieldwork at these and any supplemental 2-D and 3-D 10 GNSS receivers along a divide-to- Also, could use wider area data sites needed to acquire enough forest types to validate A variety of field experiments/campaigns to validate the velocity coast flow line in Greenland. 4 GNSS such as Ice Bridge contemporane- the product over the range of biomass validity. L3 science products that will be used to validate the time-series devices on Antarctic Ice Shelf. ISRO ous data sets, should they exist For permafrost deformation, wetlands inundation, and L2 science requirements will be organized by the Joint of ice sheet and independently funded investi- crop area requirements, the science team will compare gators may have GNSS devices at science team during this sub-phase. Possible cam- additional locations NISAR products to those derived from a combination paigns include, but are not limited to: of proven remote sensing techniques using other data Sea-ice West Arctic, Southern Ocean Using available buoy data from the • Deployment, inspection and maintenance of velocities International Arctic Buoy Program sets, such as optical imagery, and through the collec- trihedral corner reflectors at selected PBO stations (IABP) and International Pro- tion of field measurements. gramme for Antarctic Buoys (IPAB) and/or Surat Basin site in Australia. 2-D deformation U.S. Network of the Americas (NOTA), Other similar size and scale Table 8-2 shows the nominal list of global sites at • Used for instrument calibration and time-series of Hawaii Volcano Observatory (HVO), geodetic ground networks may be which L3 data products for all science disciplines will performance solid Earth GEONET-Japan, GEONET-New Zea- available as well land, AGOS, ISRO network, CALM be generated and validated. • Reflectors are to be deployed prior to launch, network and inspected and maintained during the A number of teams will be performing various functions Biomass Five canonical biomes with field mea- Use existing and heritage Cal/Val cal/val phase and once every year of science surements of biomass: Needleleaf, locations (roughly 30 sites distrib- during the calibration/validation sub-phase. operations thereafter Broadleaf Deciduous, Mixed Broad- uted globally). Collaboration with leaf/Needleleaf, Broadleaf Evergreen, BIOMASS and GEDI validation • The joint science team, which will be composed of • Biomass estimated from airborne and/or field Savanna/Dry Forest campaigns the NISAR science team and the Project science measurements for globally representative forest team at JPL, will plan and organize field campaign Disturbance Known areas of forest management Forest management plans for areas (e.g. Southeastern U.S.) clearcutting and selective logging. support (e.g., corner reflectors, GNSS stations, in Fire databases Use of high resolution optical • Used for calibration of biomass algorithm situ campaigns) Targets of opportunity (determined data to determine canopy fraction parameters, and validation of science require- after disturbance events) change. Use of active fire data- • The necessary NISAR observations for Cal/Val ment bases activities will be planned by the MOS team • Field validation of inundation extent for boreal, Crop area U.S. and India agricultural areas Local assessment surveys and imaged with quad pol mode, and se- cropscape, JECAM data • The instrument health and performance will be temperate, and tropical wetlands lected JECAM sites. evaluated with auxiliary measurements on the • Used for calibration of inundation threshold spacecraft and instrument by the Radar instru- Inundation area Wetland sites with NASA funded stud- Other international sites as well if values and validation of inundation science ies in Alaska (ABoVe); South America field data is available. Five types ment team requirement (Pacaya-Samiria, or Pantanal); AfriS- of validation data may be used • The SAR image data will be processed by the SDS AR (Gabon) site; Florida everglades; depending on location • Field validation of active crop area Louisiana Delta; coastal lagoon sites team in India; Sudd, South Sudan. • Used for calibration of crop area threshold • All image calibration parameters will be evaluated values and validation of the active crop area and validated by the algorithm development team. science requirement • Algorithm parameters needed for generating L3 data products (e.g., biomass algorithm parame- 122 NISAR Science Users’ Handbook 123 NISAR Science Users’ Handbook

• Field validation of surface deformation in perma- ground systems, mission operations, ground field TABLE 8-4. POSSIBLE FIELD EXPERIMENTS FOR NISAR CAL/VAL. frost areas campaigns and supporting infrastructure including Field Experiments/ Number of Planned Experiments corner reflectors, GNSS stations, etc.) Airborne Data/ • Used for validation of the permafrost defor- Objectives Science Science Science Satellite Observ. CAL/VAL Pre-Launch Operations Operations Operations Observations Checkout Phase mation science requirement • L3 data products over Cal/Val sites have been Year 1 Year 2 Year 3 validated via a mix of ground truth and remote • Installation of 10 GNSS receivers along a di- Instrument Deployment of 50 1 vide-to-coast flow line in Greenland and 4 GNSS sensing data (this only refers to the initial valida- corner reflectors (CR) calibration receivers on an ice shelf in Antarctica tion; these products will be validated periodically Inspection and main- Instrument 1 1 1 over the course of the mission) tenance of 50 CRs calibration • GNSS receivers will be deployed after launch • The flight systems (spacecraft, engineering Deployment of one Validation of • Used for validating observations for all snow 1- 3 payload, RBA), payloads (L-SAR and S-SAR in- passive receiver antenna pattern and types and melt states of glacier velocities digital beamforming struments) and ground systems (GDS, SDS, MOS) parameters biases are well characterized, so that calibrations Calibration of Members of the joint science team will also utilize data Biomass from field TBD 6 6 6 can be routinely applied and incorporated to adjust measurements/air- biomass algorithm from various resource networks for validating the L3 or remove biases to generate calibrated L1/L2 borne lidar parameters, and data products, e.g., NASA ABoVe (Arctic-Boreal Vul- validation of data products science requirement nerability Experiment), NEON (NSF National Ecological • The algorithms and retrieval of geophysical pa- Calibration of Observatory Network), NOTA (Network of the Americas), Field validation of 1 2 2 2 2 rameters (L3 data products) from L1/L2 products inundation extent for inundation Corner reflector arrays and GNSS station networks that boreal, temperate, threshold values are distributed globally. are validated, and any biases can be sufficiently and tropical wetlands and validation of characterized and removed inundation science requirement The exit criteria/final conditions to be satisfied for Calibration of crop Some validation campaigns will involve comparisons Field validation of 2 2 2 2 2 ending calibration/validation sub-phase are: crop area area threshold with datasets from airborne sensors (e.g., NASA values and valida- • L-SAR and S-SAR instrument calibration stability UAVSAR, DLR F-SAR, LVIS; see Table 8-3) and other tion of inundation has been demonstrated and verified. Appropri- science requirement contemporary spaceborne sensors (e.g., NASA GEDI, ate adjustments have been proposed, verified Validate surface ICESat-2, ESA Biomass, World View-3, Landsat 8, Field validation 2 2 2 2 2 and processed (revisions resultant from Cal/Val of permafrost deformation in Sentinel-1 A/B, Sentinel 2). Possible field campaigns deformation permafrost areas could affect mission timeline, radar modes, Cal/ are noted in Table 8-4. Velocity measure- Val process, SDS processing and data analysis, 10 GNSS receivers 1 1 Greenland ments for all snow facies and melt TABLE 8-3. EXISTING OR NEAR-TERM L-BAND AIRCRAFT-BASED SENSORS. states.

Airborne Systems Sensor Maintain 10 GNSS Validate observa- 1 1 1 receivers Greenland tions for all snow NASA UAVSAR L-band quad-pol repeat pass InSAR, P-band quad-pol SAR, facies and melt Ka-band single pass InSAR states

Validate velocity DLR FSAR X-band through P-band quad pol repeat pass InSAR 4 GNSS receivers on 1 1 ice shelf in Antarctica measurements JAXA Pi-SAR L-band quad-pol SAR Validate velocity Maintain 4 GNSS 1 1 1 1 1 receivers on ice shelf measurements LVIS Scanning laser altimeter in Antarctica

G-LiHt Scanning lidar, profiling lidar, VNIR imaging spectrometer, thermal imager

ISRO L/S airborne radar S-band and L-band SAR

UAS Lidar, thermal IR, and/or multispectral instruments 124 NISAR Science Users’ Handbook 125 NISAR Science Users’ Handbook

8.3 CALIBRATION/VALIDATION ROLES The NISAR Cal/Val Plan is developed and implemented TABLE 8-5. CAL/VAL ROLES AND RESPONSIBILITIES AND RESPONSIBILITIES by the NISAR Cal/Val team, which includes members Project Joint Science Radar of the joint science team, the ISRO Cal/Val team, and Science Science Data Sys- Instrument The NISAR joint science team (consisting of scientists Team Team tems Team Team members of the project science and science data selected by NASA and ISRO), along with the supporting system staff at JPL. The NISAR Cal/Val Plan will be Validation Algorithms Project Science Team (PST), will plan and organize field developed taking into consideration a broad range of campaign support (e.g. corner reflectors, GNSS sta- L0a-L0b X X inputs and contributions from the U.S. and international tions, in situ campaigns). The SDS will nominally collect communities, including Cal/Val plans of other Synthetic L0-L1 X X and process the radar data. The NISAR SDS and radar Aperture Radar (SAR) missions related to the NISAR instrument team will work together to regularly update L1-L2 X X X science disciplines. Detailed roles and responsibilities instrument calibration parameters for generating L1 L2-L3 X X for specific tasks are shown in Table 8-5. and L2 products. The instrument team will work with Calibration Algorithms the mission planning team to ensure appropriate cal- COMMUNITY INVOLVEMENT ibration data are acquired. The joint science team will Point target analysis X X analyze and evaluate imagery data processed by the The NISAR project welcomes high-quality in situ data Doppler analysis X X SDS, interpret results and generate L3 data products that can be used for calibrating or validating NISAR over selected science validation sites. They will cali- images, algorithms, and products. A formal mechanism GNSS network comparisons X X

brate and update algorithm parameters (e.g., biomass organized through the NISAR Project Cal/Val lead will Tropospheric Phase Calibration X X algorithm parameters, inundation threshold values, be established and described on the NISAR web site. etc.) regularly in their calculations of L3 products. They Ionosphere (absolute delay/ X X relative split spectrum delay) will also verify the end to end acquisition, calibration, and processing of the imagery. Lastly, the joint science Soil moisture X X

team will validate that the science requirements have Others been achieved by the mission. Calibration Activities

Work associated with X X X X calibration clgorithms

Coding of algorithms X X X (Phase C/D)

Acquisition of test data - X X scoped by each discipline

Testing of Calibration Tools X

Field Work - Scoped by X X Each Discipline

Validation Activities

Validation field work X X

Processing test data X X X

Processing of mission data X X

Comparison of results to X X X requirements NISAR Science Users’ Handbook 127 NISAR Science Users’ Handbook

9 CONCLUSIONS

Earth’s surface and vegetation cover are constantly as the applications communities. Training programs changing on a wide range of time scales. Measuring focused on radar data analysis and processing will be these changes globally from NISAR will enable break- provided to foster integration of NISAR data into Earth through science with important applications to society. science studies by future generations of scientists, NISAR will significantly expand the value of NASA’s geologists, and engineers. Workshops and conferences missions from being purely science-driven to also will be organized to develop detailed plans for calibra- encompassing informed decision support across a wide tion and validation, as well as other science activities. range of applications. The potential for synergistic satellite observations, The baseline requirements for the NISAR mission complementary to the NISAR science objectives, is also express challenging and exciting goals, to measure the quite exciting. The ESA Sentinel-1 satellites are already deforming land and ice surfaces to accuracies and spa- providing regular global sampling at C-band. The CSA tial extents that go well beyond what past and current RADARSAT constellation mission will provide similar international missions have accomplished and what C-band measurements in and around Canada system- future missions plan. These requirements are met by atically, and elsewhere around the world with more the system described in this document: first-of-a-kind limited sampling. NASA’s Global Ecosystem Dynamics technology for wide-area mapping, with a regular and Investigation (GEDI) Lidar is expected to launch to the uniform observation strategy. International Space Station in the year 2019. This lidar mission has a biomass measurement goal which is The NISAR mission will be the first NASA radar mission relevant to NISAR, in addition to measuring forest struc- to systematically and globally study the solid Earth, ture. The European Space Agency’s BIOMASS mission the ice masses, and ecosystems by regularly sampling is also expected to launch prior to NISAR. BIOMASS Earth’s land and ice covered surfaces from ascending is a fully polarimetric P-band SAR whose main goal is and descending orbit vantage points every 12 days. to measure above ground biomass. In a complemen- As an all-weather, day/night imaging system with an tary fashion, the primary biomass objectives of the exceedingly rich and far-reaching set of science objec- BIOMASS mission (biomass measured in areas over tives, NISAR is arguably the most likely Decadal Survey 100 Mg/ha) is complementary to the NISAR biomass mission to fulfill the call from the committee to expand science requirement (biomass measured in areas under the value of NASA’s missions from purely scientifically 100 Mg/ha). The Argentine Space Agency’s (CONAE) driven to encompass applications for societal benefit. SAOCOM satellite constellation, which will use L-band Many of the examples shown in this document demon- SAR for disaster monitoring, is also expected to launch strate the potential of SAR missions for applications. in the years preceding NISAR’s launch. Studies based NISAR will add a tremendous new data set to create on combining datasets from these complementary new-and greatly improve upon existing-applications. sources will not only assist in verification and validation As the mission progresses to launch in the next few but will also yield new insights for investigations of years, one of the goals of the project will be broader Earth surface processes (which were previously impos- community engagement involving the scientific as well sible due to the lack of such overlapping datasets). Taras Vyshnya/Shutterstock Taras 128 NISAR Science Users’ Handbook 129 NISAR Science Users’ Handbook

10 ACKNOWLEDGMENTS

TRAINING PROGRAMS Over 20 years in the making, NISAR represents the of the techniques developed with SEASAT and SIR-C, This document was produced at Jet Propulsion Labora- Margaret Glasscoe have also contributed to editing of hopes and aspirations of a generation of scientists both short-lived missions flown two decades ago, are tory, California Institute of Technology, under a contract the handbook. The project engineering team has ded- FOCUSED ON RADAR awaiting the data they need to perform broad-area as relevant today as they were then. The international with the National Aeronautics and Space Administra- icated countless hours to developing a first-of-a-kind DATA ANALYSIS AND Earth system studies in their disciplines, using the SAR sensors that blossomed after these US missions tion. It was initially generated with the assistance of radar system with unprecedented capabilities and has uniquely sampled data from this mission. The NISAR flew have indeed led to new and exciting discoveries. the 2012-2016 NISAR Science Definition Team and provided content to demonstrate that in this document. PROCESSING WILL BE ST comprises many scientists who have exploited SAR Yet the vision of ubiquitous SAR data for the research Project Science Team as a means to report on science In particular, Josh Doubleday as mission planner, Scott data from many sources, some from as early as SEA- community articulated in numerous scientific reports progress through the formulation and early develop- Shaffer as instrument system engineer, Sara Hatch PROVIDED TO FOSTER SAT in 1978. These scientists have tremendous depth year upon year has not yet come to pass. The examples ment phases of the NISAR project. as mission design and navigation engineer, and Peter INTEGRATION OF of experience in what SAR can and cannot do. They are shown in this document can only hint at the orders of Xaypraseuth as flight system engineer, have provided of like mind in both frustration with the lack of available magnitude improvement in our understanding of Earth Since entering Phase C in August 2016, Priyanka key content. The current science team has supplied NISAR DATA INTO science-grade SAR data available to the research that the NISAR mission will contribute. Sharma has taken over as chief editor and custodian continuing updated science results and feedback on EARTH SCIENCE community, and excitement about the opportunities of this document, which has evolved into a science the document. NISAR will provide to scientists over the world. Many users’ handbook. Paul Rosen, Andrea Donnellan, and STUDIES BY FUTURE

GENERATIONS Cost estimates stated herein are as of publication and do not necessarily represent cost at final mission. OF SCIENTISTS,

GEOLOGISTS, AND CONTRIBUTORS

2012-2016 NISAR Science Susan Callery, JPL Paul Rosen, JPL ENGINEERS. Definition Team Andrea Donnellan, JPL Priyanka Sharma, JPL 2016-2019 NISAR Science Team Margaret Glasscoe, JPL

Sean Buckley, JPL Scott Hensley, JPL NISAR Science Users’ Handbook 131 NISAR Science Users’ Handbook

11 REFERENCES

ACIA, 2004. Impacts of a Warming Arctic: Arctic IPCC, 2007: Climate Change 2007: The Physical Alley, W.M., Healy, R.W., LaBaugh, J.W. and Reilly, Climate Impact Assessment. ACIA Overview report. Science Basis. Contribution of Working Group I to T.E., 2002. Flow and storage in groundwater sys- Cambridge University Press. 140 pp. the Fourth Assessment. Report of the Intergovern- tems. science, 296(5575), pp.1985-1990. mental Panel on Climate Change, 996 pp., Cam- CEOS, C., 2014. Strategy for Carbon Observations bridge University Press, Cambridge and New York. Alpers, W., 1992, May. SAR imaging of ocean from Space. The Committee on Earth Observation waves-A review. In Geoscience and Remote Sens- Satellites (CEOS) Response to the Group on Earth JECAM, 2014, JECAM Guidelines for cropland and ing Symposium, 1992. IGARSS’92. International Observations (GEO) Carbon Strategy., April 2014 crop type definition and field data collection, Group (Vol. 1, pp. 451-454). IEEE. (issued date: September 30, 2014). on Earth Observations, 7 pp. downloaded from http://jecam.org/documents/ December 3, 2018. Alsdorf, D.E., Melack, J.M., Dunne, T., Mertes, L.A., DESDynI Applications Workshop report (2008). Avail- Hess, L.L. and Smith, L.C., 2000. Interferometric able at: http://appliedsciences.nasa.gov/pdf/09DES- Millennium Ecosystem Assessment, 2005. Ecosys- radar measurements of water level changes on the DynI_Applications_Report_V1.0.pdf tems and Human Well-being: Synthesis. Amazon flood plain. Nature, 404(6774), p.174. Island Press, Washington, DC. GCOS. 2010. Implementation Plan for the Global Ob- Amelung, F., Galloway, D.L., Bell, J.W., Zebker, H.A. and Laczniak, R.J., 1999. Sensing the ups serving System for Climate in Support of the UNFCCC Responding to the Challenge of Climate and Envi- and downs of Las Vegas: InSAR reveals structural (2010 Update), GCOS-138. Secretariat of the World ronmental Change: NASA’s Plan for a Climate-Cen- control of land subsidence and aquifer-system Meteorological Organization: Geneva, Switzerland, tric Architecture for Earth Observations and Appli- deformation. Geology, 27(6), pp.483-486. p 180. cations from Space (2010). Available at: https:// smd-prod.s3.amazonaws.com/science-green/ “GEOSS 10-year Implementation Plan,” Group on Arendt, A.A., Echelmeyer, K.A., Harrison, W.D., Lin- s3fs-public/atoms/files/Climate_Architecture_ Earth Observations (GEO) Implementation Plan Task gle, C.S. and Valentine, V.B., 2002. Rapid wastage Final.pdf Team (IPTT), 2005. of Alaska glaciers and their contribution to rising sea level. Science, 297(5580), pp.382-386. A Global Carbon Project, http://www.globalcarbonproj- ect.org/carbonbudget/ Agnew, D.C., 2012. SPOTL: Some programs for Askne, J.I., Dammert, P.B., Ulander, L.M. and Smith, ocean-tide loading. SIO Technical Report, Scripps G., 1997. C-band repeat-pass interferometric SAR FRA, 2010,The Global Forest Resources Assessment Institution of Oceanography, https://escholarship. observations of the forest. IEEE Transactions on 2010, Wilkie, M.L., Cumani, R., Martucci, A. and org/uc/item/954322pg Geoscience and Remote Sensing, 35(1), pp.25-35. Latham, J., 2007. The Global Forest Resources As- sessment 2010. Forest. Tsang et al., 1985: Tsang, L., Agram, P.S., Jolivet, R., Riel, B., Lin, Y.N., Simons, Aster, R.C., Borchers, B. and Thurber, C.H., 2018. Kong, J.A. and Shin, R.T., 1985. Theory of microwave M., Hetland, E., Doin, M.P. and Lasserre, C., 2013. Parameter estimation and inverse problems. remote sensing. New radar interferometric time series analysis Elsevier. toolbox released. Eos, Transactions American IPCC 2006, 2006 IPCC Guidelines for National Green- Geophysical Union, 94(7), pp.69-70. house Gas Inventories, Prepared by the National Attema, E.P.W. and Ulaby, F.T., 1978. Vegetation Greenhouse Gas Inventories Programme, Eggleston Ahlstrøm, A., Andersen, S.B., Nick, F.M., Reijmer, modeled as a water cloud. Radio science, 13(2), H.S., Buendia L., Miwa K., Ngara T. and Tanabe K. C.H., van de Wal, R.S.W. and Hubbard, A., 2013. pp.357-364. (eds). Published: IGES, Japan. Seasonal velocities of eight major marine-terminat- ing outlet glaciers of the Greenland ice sheet from B IPCC, 2007: Climate Change 2007: Synthesis Report. continuous in situ GPS instruments. Earth system Contribution of Working Groups I, II and III to the science data discussions, 6(1), pp.27-57. Baccini, A.G.S.J., Goetz, S.J., Walker, W.S., Laporte, Fourth Assessment Report of the Intergovernmen- N.T., Sun, M., Sulla-Menashe, D., Hackler, J., Beck, tal Panel on Climate Change [Core Writing Team, Alley, R.B., Clark, P.U., Huybrechts, P. and Joughin, P.S.A., Dubayah, R., Friedl, M.A. and Samanta, S., Pachauri, R.K and Reisinger, A. (eds.)]. IPCC, Geneva, I., 2005. Ice-sheet and sea-level changes. science, 2012. Estimated carbon dioxide emissions from Switzerland, 104 pp. 310(5747), pp.456-460. tropical deforestation improved by carbon-density maps. Nature climate change, 2(3), p.182. 132 NISAR Science Users’ Handbook 133 NISAR Science Users’ Handbook

Baghdadi, N., King, C., Chanzy, A. and Wigneron, Blewitt, G., GPS and space based geodetic meth- Capps, D.M., Rabus, B., Clague, J.J. and Shugar, Chatterjee, R.S., Sinha, A., Mahato, A.B., Champati- Cloude, S.R., Goodenough, D.G. and Chen, H., 2012. Davis, J.L., Kellogg, L.H., Arrowsmith, J.R., Buffett, J.P., 2002. An empirical calibration of the integral ods, in Treatise on Geophysics, vol. 3, edited by D.H., 2010. Identification and characterization ray, P.K., Lakhera, R.C., Singh, K.B., Varuna Kumar, Compact decomposition theory. IEEE Geoscience B.A., Constable, C.G., Donnellan, A., Ivins, E.R., equation model based on SAR data, soil moisture T. Herring, pp. 351– 390, Academic, Oxford, U.K., of alpine subglacial lakes using interferometric syn- G., Sengupta, S., Raju, E.V.R., Sharma, P.D., 2007b. and Remote Sensing Letters, 9(1), pp.28-32. Mattioli, G.S., Owen, S.E., Pritchard, M.E. and and surface roughness measurement over bare 2007. thetic aperture radar (InSAR): Brady Glacier, Alaska, Geoenvironmental Mapping of Jharia Coalfield from Purucker, M.E., 2016. Challenges and Opportunities soils. International Journal of Remote Sens- USA. Journal of Glaciology, 56(199), pp.861-870. Multi-polarization and Interferometric Synthetic Coe, J.A., 2016. Landslide hazards and climate for Research in ESI (CORE): Report from the NASA ing, 23(20), pp.4325-4340. Bonan, G.B., 2008. Forests and climate change: Aperture Radar Data, Conference of JEP-Micro- change: A perspective from the United States. Slope Earth Surface and Interior (ESI) Focus Area Work- forcings, feedbacks, and the climate benefits of Cartus, O., Siqueira, P. and Kellndorfer, J., 2018. An wave held at Space Application Centre, Ahmedabad safety preparedness for impact of climate change, shop, November 2-3, 2015, Arlington, VA. Arlington, Baghdadi, N., Gauthier, Y. and Bernier, M., 1997. forests. science, 320(5882), pp.1444-1449. Error Model for Mapping Forest Cover and Forest during May 15-16, 2007, SAC/RSMET/JEPMW/ pp.479-523. Virginia, xx pp., doi: TBD. Capability of multitemporal ERS-1 SAR data for Cover Change Using L-Band SAR. IEEE Geoscience CP/03/2007, 5-58 – 5-83 wet-snow mapping. Remote sensing of environ- Bourne, S., K. Maron, S. Oates, and G. Mueller, and Remote Sensing Letters, 15(1), pp.107-111. Cohen, W.B., Fiorella, M., Gray, J., Helmer, E. and Delbridge, B.G., Bürgmann, R., Fielding, E., Hensley, ment, 60(2), pp.174-186. 2006. Monitoring deformation of a carbonate field Chauhan, N.S., Lang, R.H. and Ranson, K.J., 1991. Anderson, K., 1998. An efficient and accurate S. and Schulz, W.H., 2016. Three-dimensional in Oman: Evidence for large scale fault re-activation Castañeda, C., Gutiérrez, F., Manunta, M. and Galve, Radar modeling of a boreal forest. IEEE Transac- method for mapping forest clearcuts in the Pacific surface deformation derived from airborne inter- Bally, P., 2012. Scientific and technical memoran- from microseismic, InSAR, and GPS. Proceedings of J.P., 2009. DInSAR measurements of ground defor- tions on Geoscience and Remote Sensing, 29(4), Northwest using Landsat imagery. Photogram- ferometric UAVSAR: Application to the Slumgullion dum of the international forum on satellite EO and 68th EAGE Annual Conference and Exhibition/SPE mation by sinkholes, mining subsidence, and land- pp.627-638. metric engineering and remote sensing, 64(4), Landslide. Journal of Geophysical Research: Solid geohazards, 21–23 May 2012. Santorini, Greece. Europec, June 12–15, 2006. EAGE Publications BV, slides, Ebro River, Spain. Earth Surface Processes pp.293-299. Earth, 121(5), pp.3951-3977. doi, 10, p.5270. Austria, Vienna. and Landforms, 34(11), pp.1562-1574. Chen, C.W. and Zebker, H.A., 2002. Phase unwrapping for large SAR interferograms: Statistical seg- Colony, R. and Thorndike, A.S., 1984. An estimate of Deverel, S.J., Bachand, S., Brandenberg, S.J., Bamber, J.L., Alley, R.B. and Joughin, I., 2007. Bryant, R., Moran, M.S., Thoma, D.P., Collins, Chambers, J.Q., Negron-Juarez, R.I., Marra, D.M., mentation and generalized network models. IEEE the mean field of Arctic sea ice motion. Journal of Jones, C.E., Stewart, J.P. and Zimmaro, P., 2016. Rapid response of modern day ice sheets to exter- C.H., Skirvin, S., Rahman, M., Slocum, K., Starks, Di Vittorio, A., Tews, J., Roberts, D., Ribeiro, G.H., Transactions on Geoscience and Remote Sensing, Geophysical Research: Oceans, 89(C6), pp.10623- Factors and processes affecting delta levee system nal forcing. Earth and Planetary Science Letters, P., Bosch, D. and Dugo, M.G., 2007. Measuring Trumbore, S.E. and Higuchi, N., 2013. The steady- 40(8), pp.1709-1719. 10629. vulnerability. San Francisco Estuary and Watershed 257(1-2), pp.1-13. surface roughness height to parameterize radar state mosaic of disturbance and succession across Science, 14(4). backscatter models for retrieval of surface soil an old-growth Central Amazon forest landscape. Chen, J., 2014a. InSAR Time Series Analysis of Coon, M.D., 1980. A review of AIDJEX modeling. Bawden, G.W., W. Thatcher, R.S. Stein, K.W. Hudnut, moisture. IEEE Geoscience and Remote Sensing Proceedings of the National Academy of Sciences, Subtle Transient Crustal Deformation Signals As- Sea Ice Processes and Models, pp.12-27. Dixon, T.H., Amelung, F., Ferretti, A., Novali, F., and G. Pletzer, 2001: Tectonic contraction across Letters, 4(1), pp.137-141. 110(10), pp.3949-3954. sociated with the 2010 Slow Slip Event at Kilauea, Rocca, F., Dokka, R., Sella, G., Kim, S.W., Wdow- Los Angeles after removal of groundwater pumping Hawaii (Doctoral dissertation, Stanford University). D inski, S. and Whitman, D., 2006. Space geodesy: effects, Nature, 412, 812-815. Burgess, E.W., Forster, R.R. and Larsen, C.F., 2013. Chapman, B.D., Celi, J.E., Hamilton, S.K. and Subsidence and flooding in New Orleans. Nature, Dale, V.H., Joyce, L.A., McNulty, S., Neilson, R.P., Flow velocities of Alaskan glaciers. Nature commu- McDonald, K.C., 2014, December. Detection and Chen, J., Zebker, H.A., Segall, P. and Miklius, A., 441(7093), p.587. Ayres, M.P., Flannigan, M.D., Hanson, P.J., Irland, Berardino, P., Fornaro, G., Lanari, R. and Sansosti, nications, 4, p.2146. Monitoring of Inundation with Polarimetric L-Band 2014b. The 2010 slow slip event and secular mo- L.C., Lugo, A.E., Peterson, C.J. and Simberloff, D., E., 2002. A new algorithm for surface deformation SAR. In AGU Fall Meeting Abstracts. tion at Kı-lauea, Hawai-i, inferred from TerraSAR-X Dobson, M.C. and Ulaby, F.T., 1986. Active micro- 2001. Climate change and forest disturbances: monitoring based on small baseline differential SAR C InSAR data. Journal of Geophysical Research: Solid wave soil moisture research. IEEE Transactions on climate change can affect forests by altering the interferograms. IEEE Transactions on Geoscience Chapman, B., McDonald, K., Shimada, M., Rosen- Earth, 119(8), pp.6667-6683. Geoscience and Remote Sensing, (1), pp.23-36. Cairns, M.A., Brown, S., Helmer, E.H. and frequency, intensity, duration, and timing of fire, and Remote Sensing, 40(11), pp.2375-2383. qvist, A., Schroeder, R. and Hess, L., 2015. Mapping Baumgardner, G.A., 1997. Root biomass allocation drought, introduced species, insect and pathogen regional inundation with spaceborne L-Band SAR. Chubey, M.S., Franklin, S.E. and Wulder, M.A., 2006. Doin, M.P., Lasserre, C., Peltzer, G., Cavalié, O. and in the world’s upland forests. Oecologia, 111(1), outbreaks, hurricanes, windstorms, ice storms, or Bergen, K.M., Goetz, S.J., Dubayah, R.O., Henebry, Remote Sensing, 7(5), pp.5440-5470. Object-based analysis of Ikonos-2 imagery for Doubre, C., 2009. Corrections of stratified tropo- pp.1-11. landslides. AIBS Bulletin, 51(9), pp.723-734. G.M., Hunsaker, C.T., Imhoff, M.L., Nelson, R.F., extraction of forest inventory parameters. Photo- spheric delays in SAR interferometry: Validation Parker, G.G. and Radeloff, V.C., 2009. Remote Chatterjee, R.S., Roy, P.S., Dadhwal, V.K., Lakhera, grammetric Engineering & Remote Sensing, 72(4), with global atmospheric models. Journal of Applied Canadell, J.G., Le Quéré, C., Raupach, M.R., Field, Darras S, Michou M, Sarrat C. 1998. IGBP-DIS Wet- sensing of vegetation 3-D structure for biodiversity R.C., Quang, T.X. and Saha, R., 2007a. Assessment pp.383-394. Geophysics, 69(1), pp.35-50. C.B., Buitenhuis, E.T., Ciais, P., Conway, T.J., Gillett, land Data Initiative: a first step towards identifying and habitat: Review and implications for lidar and of land subsidence phenomenon in Kolkata city, N.P., Houghton, R.A. and Marland, G., 2007. Con- a global delineation of wetlands. IGBP-DIS Office, radar spaceborne missions. Journal of Geophysical India using satellite-based D-InSAR technique. Clark, D.B., Read, J.M., Clark, M.L., Cruz, A.M., Doin, M.P., Guillaso, S., Jolivet, R., Lasserre, tributions to accelerating atmospheric CO2 growth Toulouse. Research: Biogeosciences, 114(G2). Current Science, pp.85-90. Dotti, M.F. and Clark, D.A., 2004. Application of 1-M C., Lodge, F., Ducret, G. and Grandin, R., 2011, from economic activity, carbon intensity, and effi- and 4-M resolution satellite data to ecological stud- September. Presentation of the small baseline ciency of natural sinks. Proceedings of the national Berardino, P., Fornaro, G., Lanari, R. and Sansosti, ies of tropical rain forests. Ecological Applications, NSBAS processing chain on a case example: The academy of sciences, 104(47), pp.18866-18870. E., 2002. A new algorithm for surface deformation 14(1), pp.61-74. Etna deformation monitoring from 2003 to 2010 monitoring based on small baseline differential SAR using Envisat data. In Proceedings of the Fringe interferograms. IEEE Transactions on Geoscience Symposium (pp. 3434-3437). ES. and Remote Sensing, 40(11), pp.2375-2383. 134 NISAR Science Users’ Handbook 135 NISAR Science Users’ Handbook

Donnellan, A., Rosen, P., Graf, J., Loverro, A., Emery, W.J., Fowler, C.W. and Maslanik, J.A., 1997. Fournier, T.J., Pritchard, M.E. and Riddick, S.N., Galloway, D.L., Hudnut, K.W., Ingebritsen, S.E., Grenerczy, G. and Wegmüller, U., 2011. Persistent Hanson, R.T., 1988. Aquifer-system compaction, Freeman, A., Treuhaft, R., Oberto, R., Simard, Satellite-derived maps of Arctic and Antarctic sea 2010. Duration, magnitude, and frequency of Phillips, S.P., Peltzer, G., Rogez, F. and Rosen, P.A., scatterer interferometry analysis of the embank- Tucson Basin and Avra Valley, Arizona. U.S. Geolog- M., Rignot, E., Kwok, R. and Pi, X., 2008, March. ice motion: 1988 to 1994. Geophysical Research subaerial volcano deformation events: New results 1998. Detection of aquifer system compaction and ment failure of a red mud reservoir using ENVISAT ical Survey Water-Resources Investigations Report Deformation, ecosystem structure, and dynamics of Letters, 24(8), pp.897-900. from Latin America using InSAR and a global land subsidence using interferometric synthetic ASAR data. Natural hazards, 59(2), p.1047. 88–4172, 69 p. ice (DESDynI). In 2008 IEEE Aerospace Conference synthesis. Geochemistry, Geophysics, Geosystems, aperture radar, Antelope Valley, Mojave Desert, (pp. 1-13). IEEE. Englewood Cliffs, New Jersey. Mette, T., 2007. 11(1). California. Water Resources Research, 34(10), Grosse, G., Harden, J., Turetsky, M., McGuire, A.D., Hanssen, R.F., 2001. Radar interferometry: data Forest biomass estimation from polarimetric SAR pp.2573-2585. Camill, P., Tarnocai, C., Frolking, S., Schuur, E.A., interpretation and error analysis (Vol. 2). Springer Donnellan, A., Parker, J., Heflin, M., Lyzenga, interferometry (Doctoral dissertation, Technische Freeman, A., Krieger, G., Rosen, P., Younis, M., Jorgenson, T., Marchenko, S. and Romanovsky, Science & Business Media. G., Moore, A., Ludwig, L.G., Rundle, J., Wang, J. Universität München). Johnson, W.T., Huber, S., Jordan, R. and Moreira, A., Galloway, D.L., Jones, D.R. and Ingebritsen, S.E., V., 2011. Vulnerability of high-latitude soil organic and Pierce, M., 2018. Fracture Advancing Step 2009, May. SweepSAR: Beam-forming on receive 1999. Land subsidence in the United States (Vol. carbon in North America to disturbance. Journal of Harris, N.L., Brown, S., Hagen, S.C., Saatchi, S.S., Tectonics Observed in the Yuha Desert and Ocotillo, F using a reflector-phased array feed combination for 1182). US Geological Survey. Geophysical Research: Biogeosciences, 116(G4). Petrova, S., Salas, W., Hansen, M.C., Potapov, P.V. CA, Following the 2010 Mw7. 2 El Mayor-Cucapah spaceborne SAR. In Radar Conference, 2009 IEEE and Lotsch, A., 2012. Baseline map of carbon Falkowski, M.J., Evans, J.S., Martinuzzi, S., Gessler, Earthquake. Earth and Space Science. (pp. 1-9). IEEE. Gardner, A.S., Moholdt, G., Cogley, J.G., Wouters, Gupta, M., Kamini Kanta Mohanty and Tapan Jyoti emissions from deforestation in tropical regions. P.E. and Hudak, A.T., 2009. Characterizing forest B., Arendt, A.A., Wahr, J., Berthier, E., Hock, R., Majumdar., 2007. Land Subsidence Mapping in Science, 336(6088), pp.1573-1576. succession with lidar data: An evaluation for the Dubois, P.C., Van Zyl, J. and Engman, T., 1995. Frolking, S., Palace, M.W., Clark, D.B., Chambers, Pfeffer, W.T., Kaser, G. and Ligtenberg, S.R., 2013. Jharia Coalfield using InSAR and DGPS Measure- Inland Northwest, USA. Remote Sensing of Environ- Measuring soil moisture with imaging radars. IEEE J.Q., Shugart, H.H. and Hurtt, G.C., 2009. Forest A reconciled estimate of glacier contributions to ments, In proceedings of Conference of JEP to- Hasager, C.B., Vincent, P., Badger, J., Badger, M., ment, 113(5), pp.946-956. Transactions on Geoscience and Remote Sens- disturbance and recovery: A general review in the sea level rise: 2003 to 2009. science, 340(6134), wards Microwave Remote Sensing Data Utilization Di Bella, A., Peña, A., Husson, R. and Volker, P.J., ing, 33(4), pp.915-926. context of spaceborne remote sensing of impacts pp.852-857. (JEP-MW), SAC, Ahmedabad, 7-21-7-30 2015. Using satellite SAR to characterize the wind Fattahi, H. and Amelung, F., 2013. DEM error cor- on aboveground biomass and canopy structure. flow around offshore wind farms. Energies, 8(6), rection in InSAR time series. IEEE Transactions on Duputel, Z., Jiang, J., Jolivet, R., Simons, M., Rivera, Journal of Geophysical Research: Biogeosciences, GFOI, 2013, Integrating remote-sensing and H pp.5413-5439. Geoscience and Remote Sensing, 51(7), pp.4249- L., Ampuero, J.P., Riel, B., Owen, S.E., Moore, A.W., 114(G2). ground-based observations for estimation of 4259. Haas, C., 2010. Dynamics versus thermodynamics: Samsonov, S.V. and Ortega Culaciati, F., 2015. emissions and removals of greenhouse gases in Hensley, S., Chen, C., Erkmen, B., Haddad, Z., The sea ice thickness distribution. Sea Ice, 82, The Iquique earthquake sequence of April 2014: Fung, A. and Eom, H., 1981. Multiple scattering forests: Methods and Guidance from the Global Harcke, L., Rosen, P., Saatchi, S., Oveisgharan, S., Ferretti, A., Prati, C. & Rocca, F., 2001. Permanent pp.113-152. Bayesian modeling accounting for prediction and depolarization by a randomly rough Kirchhoff Forest Observations Initiative: Pub: Group on Earth Shaffer, S., Simard, M. and Papathanassiou, K., Scatterers in SAR Interferometry, IEEE Trans. Geos- uncertainty. Geophysical Research Letters, 42(19), surface. IEEE Transactions on Antennas and Propa- Observations, Geneva, Switzerland, 2014. 2010, June. A Performance Estimation Methodology ci. Remote Sens., 39, 8–20 Hall, F.G., Bergen, K., Blair, J.B., Dubayah, R., pp.7949-7957. gation, 29(3), pp.463-471. for DESDynI. In Synthetic Aperture Radar (EUSAR), Houghton, R., Hurtt, G., Kellndorfer, J., Lefsky, M., Girard-Ardhuin, F., Mercier, G., Collard, F. and Garel- 2010 8th European Conference on (pp. 1-4). VDE. Finlayson, C.M., Davidson, N.C., Spiers, A.G. and Ranson, J., Saatchi, S. and Shugart, H.H., 2011. Dzurisin, D., 2007, Volcano deformation—geodetic Fung, A.K., Li, Z. and Chen, K.S., 1992. Backscatter- lo, R., 2005. Operational oil-slick characterization Stevenson, N.J., 1999. Global wetland invento- Characterizing 3D vegetation structure from space: monitoring techniques: Springer-Praxis Books in ing from a randomly rough dielectric surface. IEEE by SAR imagery and synergistic data. IEEE Journal Hensley, S., P. Agram, S. Buckley, H. Ghaemi, E. Gur- ry–current status and future priorities. Marine and Mission requirements. Remote Sensing of Environ- Geophysical Sciences, Springer, Berlin, 441 p. Transactions on Geoscience and remote sens- of Oceanic Engineering, 30(3), pp.487-495. rola, L. Harcke, C. Veeramachaneni, S. Yun 2014. Freshwater Research, 50(8), pp.717-727. ment, 115(11), pp.2753-2775. ing, 30(2), pp.356-369. NISAR Performance Model and Error Budget, NISAR E Goodison, B.E., Waterman, S.E. and Langham, E.J., Project Document. Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Hallikainen, M.T., Ulaby, F.T., Dobson, M.C., El- G 1981. Application of synthetic aperture radar data Eiken, O., Ringrose, P., Hermanrud, C., Nazarian, Bonan, G., Carpenter, S.R., Chapin, F.S., Coe, Rayes, M.A. and Wu, L.K., 1985. Microwave dielec- to snow cover monitoring. In Canadian Symposium Hess, L.L., Novo, E.M., Slaymaker, D.M., Holt, J., B. & Torp, T., 2011. Lessons learned from 14 years M.T., Daily, G.C., Gibbs, H.K. and Helkowski, J.H., Gabriel, A.K., Goldstein, R.M. and Zebker, H.A., tric behavior of wet soil-part 1: Empirical models on Remote Sensing, 6 th, Halifax, Canada (pp. Steffen, C., Valeriano, D.M., Mertes, L.A., Krug, of CCS Operations: Sleipner, In Salah and Snøhvit. 2005. Global consequences of land use. science, 1989. Mapping small elevation changes over large and experimental observations. IEEE Transactions 263-271). T., Melack, J.M., Gastil, M. and Holmes, C., 2002. Energy Procedia, Volume 4, 5541-5548 309(5734), pp.570-574. areas: differential radar interferometry. Journal on Geoscience and Remote Sensing, (1), pp.25-34. Geocoded digital videography for validation of land of Geophysical Research: Solid Earth, 94(B7), Gray, A. L., K. E. Mattar and P. W. Vachon. 1998. In- cover mapping in the Amazon basin. International Emardson, T.R., Simons, M. and Webb, F.H., 2003. Foster, J., Brooks, B., Cherubini, T., Shacat, C., pp.9183-9191. Hansen, J., Sato, M., Ruedy, R., Lacis, A. and SAR results from the RADARSAT Antarctic mapping Journal of Remote Sensing, 23(7), pp.1527-1555. Neutral atmospheric delay in interferometric Businger, S. and Werner, C.L., 2006. Mitigating Oinas, V., 2000. Global warming in the twenty-first mission data: estimation of data using a simple synthetic aperture radar applications: Statistical atmospheric noise for InSAR using a high resolution century: An alternative scenario. Proceedings of the registration procedure. In Stein, T., ed. IGARSS ’98. Hetland, E.A., Musé, P., Simons, M., Lin, Y.N., description and mitigation. Journal of Geophysical weather model. Geophysical Research Letters, National Academy of Sciences, 97(18), pp.9875- 18th International Geoscience and Remote Sensing Agram, P.S. and DiCaprio, C.J.,` 2012. Multiscale Research: Solid Earth, 108(B5). 33(16). 9880. Symposium, 6 10 July 1998, Seattle, Washington. InSAR time series (MInTS) analysis of surface de- Proceedings. Piscataway, NJ, Institute of Electrical formation. Journal of Geophysical Research: Solid and Electronics Engineers, 1638-1640. Earth, 117(B2). 136 NISAR Science Users’ Handbook 137 NISAR Science Users’ Handbook

Hibler III, W.D., 1979. A dynamic thermodynamic Hooper, A., Bekaert, D., Spaans, K. and Arıkan, M., I Jolivet, R., Duputel, Z., Riel, B., Simons, M., Rivera, Joughin, I., Smith, B.E. and Howat, I., 2018. Green- Ke, Y., & Quackenbush, L. J., 2011, A review of sea ice model. Journal of Physical Oceanography, 2012. Recent advances in SAR interferometry time L., Minson, S. E., Zhang, H., Aivazis, M. A. G., Ayoub, land Ice Mapping Project: ice flow velocity variation methods for automatic individual tree-crown detec- Ireland, R. L., J. F. Poland, and F. S. Riley, Land 9(4), pp.815-846. series analysis for measuring crustal deformation. F., Leprince, S., Samsonov, S., Motagh, M., and at sub-monthly to decadal timescales. Cryosphere, tion and delineation from passive remote sensing. subsidence in the San Joaquin Valley, California, as Tectonophysics, 514, pp.1-13. Fielding, E. J., 2014b, The 2013 Mw 7.7 Baloch- 12(7). International Journal of Remote Sensing. https:// of 1980, U.S. Geol. Surv. Prof. Pap., 437-I, 93 pp., Hilley, G. E., R. Bürgmann, A. Ferretti, F. Novali, istan Earthquake: Seismic Potential of an Accretion- doi.org/10.1080/01431161.2010.494184 1984. and F. Rocca, 2004, Dynamics of Slow-Moving Hosonuma, N., Herold, M., De Sy, V., De Fries, R.S., ary Wedge: Bulletin of the Seismological Society of Joughin, I., Shean, D.E., Smith, B.E. and Dutrieux, Landslides from Permanent Scatterer Analysis. Brockhaus, M., Verchot, L., Angelsen, A. and Romijn, America, v. 104, no. 2, p. 1020-1030. P., 2016. Grounding line variability and subglacial Kheshgi, H.S., Jain, A.K., Kotamarthi, V.R. and Ismaya, F. and Donovan, J., 2012. Applications of Science, v. 304, pp. 1952-1955. Doi: 10.1126/ E., 2012. An assessment of deforestation and forest lake drainage on Pine Island Glacier, Antarctica. Wuebbles, D.J., 1999. Future atmospheric methane DInSAR for measuring mine-induced subsidence science.1098821 degradation drivers in developing countries. Envi- Jolivet, R., Simons, M., Agram, P.S., Duputel, Z. and Geophysical Research Letters, 43(17), pp.9093- concentrations in the context of the stabilization and constraining ground deformation model. In ronmental Research Letters, 7(4), p.044009. Shen, Z.K., 2015. Aseismic slip and seismogenic 9102. of greenhouse gas concentrations. Journal of GeoCongress 2012: State of the Art and Practice in Hock, R., 2005. Glacier melt: a review of processes coupling along the central San Andreas Fault. Geo- Geophysical Research: Atmospheres, 104(D16), Geotechnical Engineering (pp. 3001-3010). and their modelling. Progress in physical geogra- Houghton, R.A., Lawrence, K.T., Hackler, J.L. and physical Research Letters, 42(2), pp.297-306. Joughin, I., Kwok, R. and Fahnestock, M., 1996. pp.19183-19190. phy, 29(3), pp.362-391. Brown, S., 2001. The spatial distribution of forest Estimation of ice-sheet motion using satellite radar Iwahana, G., Uchida, M., Liu, L., Gong, W., Meyer, biomass in the Brazilian Amazon: a comparison of Jones, C.E. and Blom, R.G., 2014. Bayou Corne, interferometry: method and error analysis with Kienholz, C., 2010. Shrinkage of Selected F.J., Guritz, R., Yamanokuchi, T. and Hinzman, Hock, R. and Jansson, P., 2006. Modeling glacier estimates. Global Change Biology, 7(7), pp.731- Louisiana, sinkhole: Precursory deformation application to Humboldt Glacier, Greenland. Journal South-Central Alaskan Glaciers AD 1900-2010: A L., 2016. InSAR detection and field evidence hydrology. In: Anderson, M.G. and J. McDonnell 746. measured by radar interferometry. Geology, 42(2), of Glaciology, 42(142), pp.564-575. Spatio-temporal Analysis Applying Photogrammet- for thermokarst after a tundra wildfire, Using (Eds.). Encyclopedia of hydrological sciences. John pp.111-114. ric, GIS-based and Historical Methods (Doctoral ALOS-PALSAR. Remote Sensing, 8(3), p.218. Wiley & Sons, Ltd, Chichester, 4, 2647-2655, 2005 Houghton, R.A., 2005. Aboveground forest biomass Joughin, I.R., Kwok, R. and Fahnestock, M.A., 1998. dissertation, Verlag nicht ermittelbar). and the global carbon balance. Global Change Joughin, I., 2002. Ice-sheet velocity mapping: a Interferometric estimation of three-dimensional ice- J Hock, R., de Woul, M., Radic, V. and Dyurgerov, Biology, 11(6), pp.945-958. combined interferometric and speckle-tracking flow using ascending and descending passes. IEEE Kindermann, G., McCallum, I., Fritz, S. and Ober- M., 2009. Mountain glaciers and ice caps around Jenkins, A. and Doake, C.S.M., 1991. Ice-ocean approach. Annals of Glaciology, 34, pp.195-201. Transactions on Geoscience and Remote Sensing, steiner, M., 2008. A global forest growing stock, Antarctica make a large sea-level rise contribution. Houghton, R.A., 2005. Tropical deforestation as interaction on Ronne Ice Shelf, Antarctica. Journal 36(1), pp.25-37. biomass and carbon map based on FAO statistics. Geophysical Research Letters, 36(7). a source of greenhouse gas emissions. Tropical of Geophysical Research: Oceans, 96(C1), pp.791- Joughin, I. and Padman, L., 2003. Melting and Silva Fennica, 42(3), pp.387-396. deforestation and climate change, 13. 813. freezing beneath Filchner-Ronne Ice Shelf, Antarcti- Joughin, I., Smith, B.E. and Abdalati, W., 2010. Hoffmann, J., Zebker, H.A., Galloway, D.L. and ca. Geophysical Research Letters, 30(9). Glaciological advances made with interferometric King, N.E., Argus, D., Langbein, J., Agnew, D.C., Amelung, F., 2001. Seasonal subsidence and Houghton, R.A., Hall, F. and Goetz, S.J., 2009. Jolivet, R. and P. S. Agram, Python-based atmo- synthetic aperture radar. Journal of Glaciology, Bawden, G., Dollar, R.S., Liu, Z., Galloway, D., rebound in Las Vegas Valley, Nevada, observed Importance of biomass in the global carbon cycle. spheric phase screen mitigation using atmospheric Joughin, I., Abdalati, W. and Fahnestock, M., 2004a. 56(200), pp.1026-1042. Reichard, E., Yong, A. and Webb, F.H., 2007. Space by synthetic aperture radar interferometry. Water Journal of Geophysical Research: Biogeosciences, re-analysis, 2012. URL: http://pyaps.googlecode. Large fluctuations in speed on Greenland’s Jakob- geodetic observation of expansion of the San Ga- Resources Research, 37(6), pp.1551-1566. 114(G2). com. shavn Isbrae glacier. Nature, 432(7017), p.608. K briel Valley, California, aquifer system, during heavy rainfall in winter 2004–2005. Journal of Geophysi- Kääb, A., 2008. Remote sensing of permafrost Holt, B., Rothrock, D.A. and Kwok, R., 1992. Deter- Howat, I.M., Joughin, I. and Scambos, T.A., 2007. Jolivet, R., Grandin, R., Lasserre, C., Doin, M.P. and Joughin, I., MacAyeal, D.R. and Tulaczyk, S., 2004b. cal Research: Solid Earth, 112(B3). related problems and hazards: Permafrost and mination of sea ice motion from satellite images. Rapid changes in ice discharge from Greenland Peltzer, G., 2011. Systematic InSAR tropospheric Basal shear stress of the Ross ice streams from Periglacial Processes, v. 19, p. 107-136 Microwave Remote Sensing of Sea Ice, pp.343- outlet glaciers. Science, 315(5818), pp.1559-1561. phase delay corrections from global meteorological control method inversions. Journal of Geophysical Klein L. Taaheri A., 2016, Data Model, File Format 354. reanalysis data. Geophysical Research Letters, Research: Solid Earth, 109(B9). and Library. HDF-EOS, NASA ESDS-RFC-008. v1. 1. Kääb, A., Huggel, C., Fischer, L., Guex, S., Paul, F., Howat, I.M., Smith, B.E., Joughin, I. and Scam- 38(17). Roer, I., Salzmann, N., Schlaefli, S., Schmutz, K., Hooper, A., Zebker, H., Segall, P. and Kampes, B., bos, T.A., 2008. Rates of southeast Greenland ice Joughin, I., Smith, B.E., Howat, I.M., Scambos, T. Kottmeier, C., Olf, J., Frieden, W. and Roth, R., 1992. Schneider, D. and Strozzi, T., 2005. Remote sensing 2004. A new method for measuring deformation on volume loss from combined ICESat and ASTER Jolivet, R., P. S. Agram, N. Y. Lin, M. Simons, M.-P. and Moon, T., 2010a. Greenland flow variability Wind forcing and ice motion in the Weddell Sea of glacier-and permafrost-related hazards in high volcanoes and other natural terrains using InSAR observations. Geophysical Research Letters, 35(17). Doin, G. Peltzer, and Z. Li, 2014a, Improving InSAR from ice-sheet-wide velocity mapping. Journal of region. Journal of Geophysical Research: Atmo- mountains: an overview. Natural Hazards and Earth persistent scatterers. Geophysical research letters, geodesy using Global Atmospheric Models, Journal Glaciology, 56(197), pp.415-430. spheres, 97(D18), pp.20373-20383. System Science, 5(4), pp.527-554. 31(23). Hurtt, G.C., Fisk, J., Thomas, R.Q., Dubayah, R., of Geophysical Research: Solid Earth, 119(3), 2324- Moorcroft, P.R. and Shugart, H.H., 2010. Linking 2341, doi:10.1002/2013JB010588. Joughin, I., Smith, B.E. and Holland, D.M., 2010b. Kwok, R., Curlander, J.C., McConnell, R. and Pang, Kaser, G., Cogley, J.G., Dyurgerov, M.B., Meier, M.F. Hooper, A.J., 2006. Persistent scatter radar models and data on vegetation structure. Journal of Sensitivity of 21st century sea level to ocean-in- S.S., 1990. An ice-motion tracking system at the and Ohmura, A., 2006. Mass balance of glaciers interferometry for crustal deformation studies and Geophysical Research: Biogeosciences, 115(G2). duced thinning of Pine Island Glacier, Antarctica. Alaska SAR facility. IEEE Journal of Oceanic Engi- and ice caps: consensus estimates for 1961–2004. modeling of volcanic deformation. Geophysical Research Letters, 37(20). neering, 15(1), pp.44-54. Geophysical Research Letters, 33(19). 138 NISAR Science Users’ Handbook 139 NISAR Science Users’ Handbook

Kwok, R. and Baltzer, T., 1995. The geophysical Le Toan, T., Quegan, S., Davidson, M.W.J., Balzter, Liu, L., Jafarov, E.E., Schaefer, K.M., Jones, B.M., M Meyer, F.J., 2011. Performance requirements for Moon, T., Joughin, I., Smith, B. and Howat, I., 2012. processor system at the Alaska SAR Facility. H., Paillou, P., Papathanassiou, K., Plummer, S., Zebker, H.A., Williams, C.A., Rogan, J. and Zhang, T., ionospheric correction of low-frequency SAR data. 21st-century evolution of Greenland outlet glacier Photogrammetric engineering and remote sensing, Rocca, F., Saatchi, S., Shugart, H. and Ulander, L., 2014. InSAR detects increase in surface subsid- MacAyeal, D.R., 1993. A tutorial on the use of IEEE transactions on geoscience and remote sens- velocities. Science, 336(6081), pp.576-578. 61(12), pp.1445-1453. 2011. The BIOMASS mission: Mapping global forest ence caused by an Arctic tundra fire. Geophysical control methods in ice-sheet modeling. Journal of ing, 49(10), pp.3694-3702. biomass to better understand the terrestrial carbon Research Letters, 41(11), pp.3906-3913. Glaciology, 39(131), pp.91-98. Moorcroft, P.R., Hurtt, G.C. and Pacala, S.W., 2001. Kwok, R., 2001. Deformation of the Arctic ocean cycle. Remote sensing of environment, 115(11), Meyer, F.J., 2014, June. Quantifying ionosphere-in- A method for scaling vegetation dynamics: the sea ice cover between November 1996 and April pp.2850-2860. Lohman, R.B. and Simons, M., 2005. Locations of Malinowski, R., Groom, G., Schwanghart, W. and duced image distortions in L-band SAR data using ecosystem demography model (ED). Ecological 1997: a qualitative survey. In IUTAM symposium on selected small earthquakes in the Zagros moun- Heckrath, G., 2015. Detection and delineation of the ionospheric scintillation model WBMOD. In EU- monographs, 71(4), pp.557-586. scaling laws in ice mechanics and ice dynamics Li, Z. H., Fielding, E. J., Cross, P., & Muller, J. P. tains. Geochemistry, Geophysics, Geosystems, 6(3). localized flooding from WorldView-2 multispectral SAR 2014; 10th European Conference on Synthetic (pp. 315-322). Springer, Dordrecht. 2006a. Interferometric synthetic aperture radar at- data. Remote Sensing, 7(11), pp.14853-14875. Aperture Radar; Proceedings of (pp. 1-4). VDE. Moore, R.K., Claassen, J.P. and Lin, Y.H., 1981. mospheric correction: GPS topography-dependent Lohman, R.B., Simons, M. and Savage, B., 2002. Scanning spaceborne synthetic aperture radar with Kwok, R. and Cunningham, G.F., 2002. Seasonal ice turbulence model, 111, B02404. Location and mechanism of the Little Skull Moun- Malvern, L. E. (1969), Introduction to Mechanics of Meyer, 2014 Michel, R. and E. Rignot. 1999. Flow integrated radiometer. IEEE Transactions on Aero- area and volume production of the Arctic Ocean: tain earthquake as constrained by satellite radar a Continuous Medium, Prentice-Hall, 713 pp. of Glaciar Moreno, Argentina, from repeat-pass space and Electronic Systems, (3), pp.410-421. November 1996 through April 1997. Journal of Li, Z., Muller, J. P., Cross, P., Albert, P., Fischer, J., interferometry and seismic waveform modeling. Shuttle Imaging Radar images: comparison of the Geophysical Research: Oceans, 107(C10). & Bennartz, R. 2006b. Assessment of the potential Journal of Geophysical Research: Solid Earth, Marquardt, F., 2009. F. Marquardt and SM Girvin, phase correlation method with radar interferometry. Morlighem, M., Seroussi, H., Larour, E. and Rignot, of MERIS near-infrared water vapour products to 107(B6), pp.ETG-7. Physics 2, 40 (2009). Physics, 2, p.40. J. Glaciol., 45(149), 93-100. E., 2013. Inversion of basal friction in Antarctica Kwok, R., G. F. Cunningham, and W. D. Hibler, 2003. correct ASAR interferometric measurements. Int. J. using exact and incomplete adjoints of a higher-or- Sub-daily ice motion and deformation from Radar- Rem. Sens., 27, 349–365. Lu, Z., and Dzurisin, D., 2014. InSAR Imaging of Massonnet, D., Rossi, M., Carmona, C., Adragna, F., Mironov, V.L., Dobson, M.C., Kaupp, V.H., Komarov, der model. Journal of Geophysical Research: Earth sat observations, Geophysical Research Letters, 30 Aleutian Volcanoes: Monitoring a Volcanic Arc from Peltzer, G., Feigl, K. and Rabaute, T., 1993. The dis- S.A. and Kleshchenko, V.N., 2004. Generalized re- Surface, 118(3), pp.1746-1753. (23), doi:10.1029/2003GL018723. Lin, Y.N.N., Simons, M., Hetland, E.A., Muse, P. Space, Springer Praxis Books, Geophysical Scienc- placement field of the Landers earthquake mapped fractive mixing dielectric model for moist soils. IEEE and DiCaprio, C., 2010. A multiscale approach to es, ISBN 978-3-642-00347-9, 388 pp. by radar interferometry. Nature, 364(6433), p.138. Transactions on Geoscience and Remote Sens- Mouginot, J., Rignot, E., Scheuchl, B. and Millan, Kwok, R., 2010. Satellite remote sensing of sea-ice estimating topographically correlated propagation ing, 42(4), pp.773-785. R., 2017. Comprehensive annual ice sheet velocity thickness and kinematics: a review. Journal of delays in radar interferograms. Geochemistry, Lu, Z., and Wicks, C., 2010, Study of the 6 August Matthews, E., 2000: 12. Wetlands. In Atmospheric mapping using Landsat-8, Sentinel-1, and RADAR- Glaciology, 56(200), pp.1129-1140. Geophysics, Geosystems, 11(9). 2007 Crandall Canyon mine (Utah, USA) collapse Methane: Its Role in the Global Environment. and Mitchard, E.T., Saatchi, S.S., Woodhouse, I.H., Nan- SAT-2 data. Remote Sensing, 9(4), p.364. from ALOS PALSAR InSAR: Geomatics, Natural M.A.K. Khalil, Ed. Springer Verlag, 202-233. gendo, G., Ribeiro, N.S., Williams, M., Ryan, C.M., L Lindsay, R.W. and Stern, H.L., 2003. The RADAR- Hazards and Risk, 1, 85-93. Lewis, S.L., Feldpausch, T.R. and Meir, P., 2009. Myneni, R.B., Dong, J., Tucker, C.J., Kaufmann, SAT geophysical processor system: Quality of sea Matzler, C., 1994. Microwave (1-100 GHz) dielectric Using satellite radar backscatter to predict above- R.K., Kauppi, P.E., Liski, J., Zhou, L., Alexeyev, V. Lang, R.H., 1981. Electromagnetic backscattering ice trajectory and deformation estimates. Journal Lu, Z., Dzurisin, D., Biggs, J., Wicks, C., Jr, and model of leaves. IEEE Transactions on Geoscience ground woody biomass: A consistent relationship and Hughes, M.K., 2001. A large carbon sink in from a sparse distribution of lossy dielectric scat- of Atmospheric and Oceanic Technology, 20(9), McNutt, S., 2010, Ground surface deformation and Remote Sensing, 32(4), pp.947-949. across four different African landscapes. Geophysi- the woody biomass of northern forests. Proceed- terers. Radio Science, 16(1), pp.15-30. pp.1333-1347. patterns, magma supply, and magma storage at cal Research Letters, 36(23). ings of the National Academy of Sciences, 98(26), Okmok volcano, Alaska, from InSAR analysis: 1. Meier, M.F., Dyurgerov, M.B., Rick, U.K., O’neel, pp.14784-14789. Larour, E., Rignot, E., Joughin, I. and Aubry, D., Liu, L., Zhang, T., and Wahr, J., 2010, InSAR mea- Intereruption deformation, 1997–2008: Journal of S., Pfeffer, W.T., Anderson, R.S., Anderson, S.P. Mohr, J.J., Reeh, N. and Madsen, S.N., 1998. 2005. Rheology of the Ronne Ice Shelf, Antarctica, and Glazovsky, A.F., 2007. Glaciers dominate surements of surface deformation over permafrost Geophysical Research, 115, B00B02, doi:10.1029/ Three-dimensional glacial flow and surface eleva- N inferred from satellite radar interferometry data on the North Slope of Alaska: Journal of Geophysi- 2009JB006969. eustatic sea-level rise in the 21st century. Science, tion measured with radar interferometry. Nature, using an inverse control method. Geophysical Neumann, M., Saatchi, S.S., Ulander, L.M. and cal Research, v. 115, p. F03023 317(5841), pp.1064-1067. 391(6664), p.273. Research Letters, 32(5). Fransson, J.E., 2012. Assessing performance of Lucas, R., Bunting, P., Paterson, M., & Chisholm, L., L-and P-band polarimetric interferometric SAR data Liu, L., Schaefer, K., Zhang, T. and Wahr, J., 2012. 2008, Classification of Australian forest commu- Meinzer, O.E., 1928. Compressibility and elasticity Monaldo, F. and Kerbaol, V., 2004, June. The SAR Lauknes, T.R., Zebker, H.A. and Larsen, Y., 2011. in estimating boreal forest above-ground biomass. Estimating 1992–2000 average active layer thick- nities using aerial photography, CASI and HyMap of artesian aquifers. Economic Geology, 23(3), measurement of ocean surface winds: an overview InSAR deformation time series using an L1-norm IEEE Transactions on Geoscience and Remote ness on the Alaskan North Slope from remotely data. Remote Sensing of Environment, 112(5), pp.263-291. for the 2nd workshop on coastal and marine ap- small-baseline approach. IEEE Transactions on Sensing, 50(3), pp.714-726. sensed surface subsidence. Journal of Geophysical 2088–2103. plications of SAR. In Proc. of the 2nd workshop on Geoscience and Remote Sensing, 49(1), pp.536- Research: Earth Surface, 117(F1). Mermoz, S., Réjou-Méchain, M., Villard, L., Le Toan, SAR coastal and marine applications (pp. 8-12). 546. Lundgren, P., F. Casu, M. Manzo, A. Pepe, P. Berar- T., Rossi, V. and Gourlet-Fleury, S., 2015. Decrease Liu, M., Tian, H., Yang, Q., Yang, J., Song, X., dino, E. Sansosti, and R. Lanari, 2004, Gravity and of L-band SAR backscatter with biomass of dense Monaldo, F., Jackson, C., Li, X. and Pichel, W.G., Le Quéré, C., Andrew, R.M., Friedlingstein, P., Sitch, Lohrenz, S.E. and Cai, W.J., 2013. Long-term trends magma spreading of Mount Etna volcano revealed forests. Remote Sensing of Environment, 159, 2016. Preliminary evaluation of Sentinel-1A wind S., Pongratz, J., Manning, A.C., Korsbakken, J.I., Pe- in evapotranspiration and runoff over the drainage by radar interferometry, Geophys. Res. Lett, pp.307-317. speed retrievals. IEEE Journal of Selected Topics in ters, G.P., Canadell, J.G., Jackson, R.B. and Boden, basins of the Gulf of Mexico during 1901–2008. L04602, doi:10.1029/2003GL018736. Applied Earth Observations and Remote Sensing, T.A., 2017. Global carbon budget 2017. Earth Water Resources Research, 49(4), pp.1988-2012. 9(6), pp.2638-2642. System Science Data Discussions, pp.1-79. 140 NISAR Science Users’ Handbook 141 NISAR Science Users’ Handbook

O Paterson, W. S. B. (1994), The Physics of Glaciers, Pritchard, M.E., Norabuena, E.O., Ji, C., Boroschek, Rignot, E., Jacobs, S., Mouginot, J. and Scheuchl, Romeiser, R., 2013. The Future of SAR-Based S 3rd Edition, 3rd ed., 480 pp., Pergamon, Oxford. R., Comte, D., Simons, M., Dixon, T.H. and Rosen, B., 2013. Ice-shelf melting around Antarctica. Oceanography: High-Resolution Current Measure- Oh, Y., Sarabandi, K. and Ulaby, F.T., 1992. An em- Saatchi, S.S. and Lang, R.H., 1989. Self Consistent P.A., 2007. Geodetic, teleseismic, and strong motion Science, 341(6143), pp.266-270. ments by Along-Track Interferometry. Oceanogra- pirical model and an inversion technique for radar Approach to Average Waves in a One Dimensional Paulson, A., Zhong, S. and Wahr, J., 2007. Inference constraints on slip from recent southern Peru sub- phy, 26(2), pp.92-99. scattering from bare soil surfaces. IEEE transac- Discrete Random Medium. In PIERS: Progress in of mantle viscosity from GRACE and relative sea duction zone earthquakes. Journal of Geophysical Rignot, E., J. Mouginot, and B. Scheuchl 2011a, Ice tions on Geoscience and Remote Sensing, 30(2), Electromagnetics Research Symposium (p. 277). level data. Geophysical Journal International, Research: Solid Earth, 112(B3). Flow of the Antarctic Ice Sheet, Science, 333(6048), Rosen, P.A., Hensley, S., Joughin, I.R., Li, F.K., Mad- pp.370-381. 171(2), pp.497-508. 1427–1430, doi:10.1126/science.1208336. sen, S.N., Rodriguez, E. and Goldstein, R.M., 2000. Saatchi, S.S., Le Vine, D.M. and Lang, R.H., 1994. Q Synthetic aperture radar interferometry. Proceed- Olofsson, P., Foody, G.M., Herold, M., Stehman, S.V., Microwave backscattering and emission model for Paulson, A., Zhong, S. and Wahr, J., 2007. Inference Rignot, E., Mouginot, J. And Scheuchl, B., 2011b. ings of the IEEE, 88(3), pp.333-382. Woodcock, C.E. and Wulder, M.A., 2014. Good prac- Qu-lin, T., Qing-chao, W., and Song-lin, Y., 2010, grass canopies. IEEE Transactions on Geoscience of mantle viscosity from GRACE and relative sea Antarctic grounding line mapping from differential tices for estimating area and assessing accuracy of Discussion on Monitoring the Subsidence of Sub- and Remote Sensing, 32(1), pp.177-186. level data. Geophysical Journal International, satellite radar interferometry. Geo. Phys. Res. Lett., Rosen, P.A., S. Hensley, C. Chen 2010. Measure- land change. Remote Sensing of Environment, 148, grade in Permafrost Region With Satellite D-InSAR 171(2), pp.497-508. 38: L10504, doi:10.1029/2011GL047109 ment and mitigation of the ionosphere in L-band pp.42-57. Technology: Journal of Railway Engineering Society Saatchi, S.S. and McDonald, K.C., 1997. Coherent Interferometric SAR data. Proc. IEEE Radar Concer- effects in microwave backscattering models for Payne, A.J., Vieli, A., Shepherd, A.P., Wingham, D.J. Rignot, E., Velicogna, I., van den Broeke, M.R., ence 2010, Washington DC. Olson, D.M., Dinerstein, E., Wikramanayake, E.D., R forest canopies. IEEE Transactions on Geoscience and Rignot, E., 2004. Recent dramatic thinning Monaghan, A. and Lenaerts, J.T., 2011c. Accel- Burgess, N.D., Powell, G.V., Underwood, E.C., and Remote Sensing, 35(4), pp.1032-1044. of largest West Antarctic ice stream triggered by Ramaswamy, V., Boucher, O., Haigh, J., Hauglustine, eration of the contribution of the Greenland and Rosen, P.A., Gurrola, E., Sacco, G.F. and Zebker, H., D’amico, J.A., Itoua, I., Strand, H.E., Morrison, J.C. oceans. Geophysical Research Letters, 31(23). D., Haywood, J., Myhre, G., Nakajima, T., Shi, G.Y. Antarctic ice sheets to sea level rise. Geophysical 2012, April. The InSAR scientific computing envi- and Loucks, C.J., 2001. Terrestrial Ecoregions of Saatchi, S.S. and Moghaddam, M., 2000. Estima- and Solomon, S., 2001. Radiative forcing of climate. Research Letters, 38(5). ronment. In Synthetic Aperture Radar, 2012. EUSAR. the World: A New Map of Life on EarthA new global tion of crown and stem water content and biomass Peplinski, N.R., Ulaby, F.T. and Dobson, M.C., 1995. Climate change, 349. 9th European Conference on (pp. 730-733). VDE. map of terrestrial ecoregions provides an innova- of boreal forest using polarimetric SAR imag- Dielectric properties of soils in the 0.3-1.3-GHz Rigor, I.G., Wallace, J.M. and Colony, R.L., 2002. Re- tive tool for conserving biodiversity. BioScience, ery. IEEE Transactions on Geoscience and Remote range. IEEE Transactions on Geoscience and Re- Reiche, J., Lucas, R., Mitchell, A.L., Verbesselt, J., sponse of sea ice to the Arctic Oscillation. Journal Rosen, P.A., Hensley, S., Shaffer, S., Veilleux, L., 51(11), pp.933-938. Sensing mote Sensing, 33(3), pp.803-807. Hoekman, D.H., Haarpaintner, J., Kellndorfer, J.M., of Climate, 15(18), pp.2648-2663. Chakraborty, M., Misra, T., Bhan, R., Sagi, V.R. and Rosenqvist, A., Lehmann, E.A., Woodcock, C.E. and Satish, R., 2015, May. The nasa-isro sar mission-an Onn, F. and Zebker, H.A., 2006. Correction for Saatchi, S.S., Houghton, R.A., Dos Santos Alvala, Peters, G.P., Marland, G., Le Quéré, C., Boden, T., Seifert, F.M., 2016. Combining satellite data for Riley, F.S., 1969. Analysis of borehole extensometer international space partnership for science and interferometric synthetic aperture radar atmospher- R.C., Soares, J.V. and Yu, Y., 2007. Distribution of Canadell, J.G. and Raupach, M.R., 2012. Rapid better tropical forest monitoring. Nature Climate data from central California. Land subsidence, 2, societal benefit. In Radar Conference (RadarCon), ic phase artifacts using time series of zenith wet aboveground live biomass in the Amazon basin. growth in CO 2 emissions after the 2008–2009 Change, 6(2), p.120. pp.423-431. 2015 IEEE (pp. 1610-1613). IEEE. delay observations from a GPS network. Journal of Global Change Biology, 13(4), pp.816-837. global financial crisis. Nature Climate Change, 2(1), Geophysical Research: Solid Earth, 111(B9). p.2. Richards, J. A. (2009), Remote Sensing with Imag- Ringrose, P., et al., 2009, Plume development Rosenqvist, Å., Milne, A., Lucas, R., Imhoff, M. and Saatchi, S.S., Harris, N.L., Brown, S., Lefsky, M., ing Radar, Springer Publishers. ISBN 978-3-642- around well KB-502 at the In Salah CO2 storage Dobson, C., 2003. A review of remote sensing tech- Onuma, T. and Ohkawa, S., 2009. Detection of Mitchard, E.T., Salas, W., Zutta, B.R., Buermann, Pritchard, M.E. and Simons, M., 2002. A satellite 02020-9 site, First Break, v. 27, pp 85-89. www.firstbreak. nology in support of the Kyoto Protocol. Environ- surface deformation related with CO2 injection by W., Lewis, S.L., Hagen, S. and Petrova, S., 2011. geodetic survey of large-scale deformation of org mental Science & Policy, 6(5), pp.441-455. DInSAR at In Salah, Algeria. Energy Procedia, 1(1), Benchmark map of forest carbon stocks in tropical volcanic centres in the central Andes. Nature, Rignot, E. and Kanagaratnam, P., 2006. Changes in pp.2177-2184. regions across three continents. Proceedings of the 418(6894), p.167. the velocity structure of the Greenland Ice Sheet. Robinson, C., Saatchi, S., Neumann, M. and Rosenqvist, A., Shimada, M., Ito, N. and Watanabe, National Academy of Sciences, 108(24), pp.9899- Science, 311(5763), pp.986-990. Gillespie, T., 2013. Impacts of spatial variability M., 2007. ALOS PALSAR: A pathfinder mission for P 9904. Pritchard, M.E. and Simons, M., 2004. An In- on aboveground biomass estimation from L-band global-scale monitoring of the environment. IEEE Paine, J.G., Buckley, S.M., Collins, E.W. and Wilson, SAR-based survey of volcanic deformation in the Rignot, E. and Jacobs, S.S., 2002. Rapid bottom radar in a temperate forest. Remote Sensing, 5(3), Transactions on Geoscience and Remote Sensing, Saatchi, S., Halligan, K., Despain, D.G. and Crabtree, C.R., 2012. Assessing collapse risk in evaporite central Andes. Geochemistry, Geophysics, Geosys- melting widespread near Antarctic ice sheet pp.1001-1023. 45(11), pp.3307-3316. R.L., 2007. Estimation of forest fuel load from radar sinkhole-prone areas using microgravimetry and tems, 5(2). grounding lines. Science, 296(5575), pp.2020- remote sensing. IEEE Transactions on Geoscience radar interferometry. Journal of Environmental and 2023. Romanovsky, V.E., Smith, S.L., and Christiansen, Rott, H. and Davis, R.E., 1993. Multifrequency and and Remote Sensing, 45(6), pp.1726-1740. Engineering Geophysics, 17(2), pp.75-87. Pritchard, M.E., Ji, C. and Simons, M., 2006. H.H., 2010, Permafrost thermal state in the polar polarimetric SAR observations on alpine glaciers. Distribution of slip from 11 Mw> 6 earthquakes Rignot, E., Casassa, G., Gogineni, P., Krabill, W., Northern Hemisphere during the international polar Annals of glaciology, 17, pp.98-104. Saatchi, S., Marlier, M., Chazdon, R.L., Clark, D.B. Pan, Y., Birdsey, R.A., Fang, J., Houghton, R., in the northern Chile subduction zone. Journal of Rivera, A.U. and Thomas, R., 2004. Accelerated ice year 2007–2009: a synthesis: Permafrost and and Russell, A.E., 2011. Impact of spatial variability Kauppi, P.E., Kurz, W.A., Phillips, O.L., Shvidenko, Geophysical Research: Solid Earth, 111(B10). discharge from the Antarctic Peninsula following Periglacial Processes, v. 21, p. 106-116 Rykhus, R., and Lu, Z., 2011, Monitoring a 2003, of tropical forest structure on radar estimation of A., Lewis, S.L., Canadell, J.G. and Ciais, P., 2011. the collapse of Larsen B ice shelf. Geophysical Yukon Flats, Alaska wildfire using multi-temporal aboveground biomass. Remote Sensing of Environ- A large and persistent carbon sink in the world’s Research Letters, 31(18). Radarsat-1 intensity and interferometric coherence ment, 115(11), pp.2836-2849. forests. Science, p.1201609. images: Geomatics, Natural Hazards and Risk, Vol. 2, No. 1, 15–32. 142 NISAR Science Users’ Handbook 143 NISAR Science Users’ Handbook

Saatchi, S., Asefi-Najafabady, S., Malhi, Y., Aragão, Schweder, T., 1976. Some “optimal” methods to Stern, H.L., Rothrock, D.A. and Kwok, R., 1995. Tsang, L., Kong, J.A. and Shin, R.T., 1985. Theory of Vesecky, J.F., Samadani, R., Smith, M.P., Daida, J.M. Y L.E., Anderson, L.O., Myneni, R.B. and Nemani, R., detect structural shift or outliers in regression. Open water production in Arctic sea ice: Satellite microwave remote sensing. and Bracewell, R.N., 1988. Observation of sea-ice Yu, Y. and Saatchi, S., 2016. Sensitivity of L-band 2013. Persistent effects of a severe drought on Am- Journal of the American Statistical Association, measurements and model parameterizations. Jour- dynamics using synthetic aperture radar images: SAR backscatter to aboveground biomass of global azonian forest canopy. Proceedings of the National 71(354), pp.491-501. nal of Geophysical Research: Oceans, 100(C10), U Automated analysis. IEEE transactions on geosci- forests. Remote Sensing, 8(6), p.522. Academy of Sciences, 110(2), pp.565-570. pp.20601-20612. ence and remote sensing, 26(1), pp.38-48. Ulaby, F.T., Stiles, W.H. and AbdelRazik, M., 1984. Segall, P., 2013. Volcano deformation and eruption Snowcover influence on backscattering from ter- Z Sahagian, D., Melack, J., Birkett, C., Chanton, J., forecasting, Geological Society, London, Special Strozzi, T., Kouraev, A., Wiesmann, A., Wegmüller, W rain. IEEE Transactions on geoscience and remote Dunne, T., Estes, J., Finlayson, M., Fresco, L., Gopal, Publications, V. 380, doi:10.1144/SP380.4, 2013. U., Sharov, A. and Werner, C., 2008. Estimation of Zebker, H.A. and Goldstein, R.M., 1986. Topographic sensing, (2), pp.126-133. Wadhams, P., 2000. Ice in the Ocean, Gordon and B., Hess, L. and Hollis, T., 1997. Global wetland Arctic glacier motion with satellite L-band SAR mapping from interferometric synthetic aperture Breach Publishers, 351 pp. distribution and functional characterizaton: Trace Shimada, T., Kawamura, H. and Shimada, M., 2003. data. Remote Sensing of Environment, 112(3), radar observations. Journal of Geophysical Re- Ulaby, F.T. and Dobson, M.C., 1989. Handbook of gases and the hydrologic cycle. In Joint IGBP GAIM- An L-band geophysical model function for SAR pp.636-645. search: Solid Earth, 91(B5), pp.4993-4999. radar scattering statistics for terrain (Artech House Walker, W. S., Stickler, C. M., Kellndorfer, J. M., DIS-BAHC-IGAC-LUCC workshop, IGBP GAIM/IGBP wind retrieval using JERS-1 SAR. IEEE transactions Remote Sensing Library). Norwood, MA, USA: Kirsch, K. M., & Nepstad, D. C., 2010, Large-area Global Change, Santa Barbara CA, USA, 1996 (pp. on geoscience and remote sensing, 41(3), pp.518- T Zebker, H.A., Rosen, P.A. and Hensley, S., 1997. Artech House. classification and mapping of forest and land cover 45-45). 531. Atmospheric effects in interferometric synthetic Tello, M., López-Martínez, C. and Mallorqui, J.J., in the Brazilian Amazon: a comparative analysis of aperture radar surface deformation and topograph- 2005. A novel algorithm for ship detection in SAR Ulaby, F.T. and El-Rayes, M.A., 1987. Microwave ALOS/PALSAR and Landsat TM data sources. Jour- Sandwell, D. T., and E. J. Price, 1996. Multiple pass Snedecor, G.W. and Cochran, W.G., 1989. Statistical ic maps. Journal of Geophysical Research: Solid imagery based on the wavelet transform. IEEE dielectric spectrum of vegetation-Part II: Dual-dis- nal of Selected Topics in Applied Earth Observations INSAR processing for geophysical applications: methods, 8thEdn. Ames: Iowa State Univ. Press Earth, 102(B4), pp.7547-7563. Geoscience and remote sensing letters, 2(2), persion model. IEEE Transactions on Geoscience and Remote Sensing, 3(4), 594-604. stack phase gradient then unwrap, in Eos Trans. Iowa. pp.201-205. and Remote Sensing, (5), pp.550-557. Amer. Geophys. Un. Meeting Suppl., p. F52. Zhang, J., Rothrock, D. and Steele, M., 2000. Whelen, T. and Siqueira, P., 2018. Coefficient of Sneed, M. and Galloway, D.L., 2000. Aquifer-system Recent changes in Arctic sea ice: The interplay be- Terzaghi, K., 1925. Principles of soil mechanics, Ulaby, F.T., Sarabandi, K., Mcdonald, K.Y.L.E., Whitt, variation for use in crop area classification across Scambos, T.A., Bohlander, J.A., Shuman, C.U. and compaction and land subsidence: measurements, tween ice dynamics and thermodynamics. Journal IV—Settlement and consolidation of clay. Engineer- M. and Dobson, M.C., 1990. Michigan microwave multiple climates. International Journal of Applied Skvarca, P., 2004. Glacier acceleration and thinning analyses, and simulations-the Holly Site, Edwards of Climate, 13(17), pp.3099-3114. ing News-Record, 95(3), pp.874-878. canopy scattering model. International Journal of Earth Observation and Geoinformation, 67, pp.114- after ice shelf collapse in the Larsen B embayment, Air Force Base, Antelope Valley, California (No. Remote Sensing, 11(7), pp.1223-1253. 122. Antarctica. Geophysical Research Letters, 31(18). 2000-4015). Geological Survey (US). Zhang, Y., Wang, C., Wu, J., Qi, J. and Salas, Thorndike, A.S., and R. Colony, 1982, Sea ice mo- W.A., 2009. Mapping paddy rice with multitem- tion in response to geostrophic winds, J. Geophysi- V Williams, S., Bock, Y. and Fang, P., 1998. Integrated Schaefer, K., Liu, L., Parsekian, A., Jafarov, E., Chen, Sneed, M., Brandt, J., Solt, M., 2013, Land poral ALOS/PALSAR imagery in southeast China. cal Res., 80(33), 4501-4513. satellite interferometry: Tropospheric noise, GPS A., Zhang, T., Gusmeroli, A., Panda, S., Zebker, H.A. Subsidence along the Delta-Mendota Canal in the van der Werf, G. R., Morton, D. C., DeFries, R. S., International Journal of Remote Sensing, 30(23), estimates and implications for interferometric and Schaefer, T., 2015. Remotely sensed active Northern Part of the San Joaquin Valley, California, Olivier, J. G. J., Kasibhatla, P. S., Jackson, R. B., … pp.6301-6315. Thorndike, A.S., Rothrock, D.A., Maykut, G.A. and synthetic aperture radar products. Journal of layer thickness (ReSALT) at Barrow, Alaska using 2003-10: U.S. Geological Survey Scientific Investi- Randerson, J. T.,2009, CO2 emissions from forest Colony, R., 1975. The thickness distribution of Geophysical Research: Solid Earth, 103(B11), interferometric synthetic aperture radar. Remote gations Report 2013-5142, 87 p loss. Nature Geoscience, 2, 737. Zhao, W., Amelung, F., Dixon, T.H., Wdowinski, S. sea ice. Journal of Geophysical Research, 80(33), pp.27051-27067. Sensing, 7(4), pp.3735-3759. and Malservisi, R., 2014. A method for estimating pp.4501-4513. Spreen, G., S. Kern, D. Stammer, and E. Hansen, van Oort, P.A., 2006. Spatial data quality: from ice mass loss from relative InSAR observations: Woodcock, C.E., Macomber, S.A., Pax-Lenney, M. Scheuchl, B., Mouginot, J., Rignot, E., Morlighem, 2009. Fram Strait sea ice volume export esti- description to application. Wageningen Universiteit. Application to the Vatnajökull ice cap, Iceland. Toda, S., Lin, J. and Stein, R.S., 2011. Using the and Cohen, W.B., 2001. Monitoring large areas for M. and Khazendar, A., 2016. Grounding line retreat mated between 2003 and 2008 from satellite Geochemistry, Geophysics, Geosystems, 15(1), 2011 M w 9.0 off the Pacific coast of Tohoku forest change using Landsat: Generalization across of Pope, Smith, and Kohler Glaciers, West Antarc- data, Geophysical Research Letters, 36, L19502, van Zyl, J., Y. Kim, 2011. Synthetic Aper- pp.108-120. Earthquake to test the Coulomb stress triggering space, time and Landsat sensors. Remote sensing tica, measured with Sentinel-1a radar interferom- doi:10.1029/2009GL039591. ture Radar Polarimetry, John Wiley and Sons, hypothesis and to calculate faults brought closer to of environment, 78(1-2), pp.194-203. etry data. Geophysical Research Letters, 43(16), DOI:10.1002/9781118116104 Zhen, L., and Huadong, G., 2000, Permafrost failure. Earth, planets and space, 63(7), p.39. pp.8572-8579. Stehman, S.V., 2005. Comparing estimators of mapping in the Tibet plateau using polarimetric gross change derived from complete coverage Vasco, D. W., Rucci, A., Ferretti, A., Novali, F., Bissell, SAR, Geoscience and Remote Sensing Symposium, Truong-Loï, M.L., Saatchi, S. and Jaruwatanadilok, Schimel, D., Pavlick, R., Fisher, J.B., Asner, G.P., mapping versus statistical sampling of remotely R. C., Ringrose, P. S. Mathieson, A. S. and Wright, I. 2000. Proceedings. IGARSS 2000. IEEE 2000 Inter- S., 2015. Soil moisture estimation under tropical Saatchi, S., Townsend, P., Miller, C., Frankenberg, sensed data. Remote Sensing of Environment, W., 2010. Satellite-based measurements of surface national, Volume 5, p. 2024-2026 vol.5 forests using UHF radar polarimetry. IEEE Transac- C., Hibbard, K. and Cox, P., 2015. Observing terres- 96(3-4), pp.466-474. deformation reveal fluid flow associated with the tions on Geoscience and Remote Sensing, 53(4), trial ecosystems and the carbon cycle from space. geological storage of carbon dioxide. Geophysical pp.1718-1727. Global change biology, 21(5), pp.1762-1776. Research Letters,Vol. 37, L03303 144 NISAR Science Users’ Handbook 145 NISAR Science Users’ Handbook

12 GLOSSARY

Acronym Definition

AOCS Attitude and Orbit Control System

DAAC Distributed Active Archive Center DESDynI Deformation, Ecosystem Structure and Dynamics of Ice DESDynI-R DESDynI Radar DSSG DESDynI Science Study Group

ISAC ISRO Satellite Centre ISLR Integrated Side Lobe Ratio ISRO Indian Space Research Organisation

LIDAR Light Detection and Ranging

MCR Mission Concept Review MNR Multiplicative Noise Ratio

NASA National Aeronautics and Space Administration NISAR NASA-ISRO SAR

PRF Pulse Repetition Frequency PSInSAR Persistent Scatterer Interferometic SAR PSLR Peak Side Lobe Ratio

SAC Space Applications Centre (ISRO) SAR Synthetic Aperture Radar SDD Science Definition Document SDT Science Definition Team NISAR Science Users’ Handbook 147 NISAR Science Users’ Handbook

13 APPENDIX A HISTORICAL BACKGROUND FOR NISAR

The first civilian SAR satellite in history, called SEASAT, of an L-band polarimetric SAR designed to operate as was launched by NASA in 1978. SEASAT’s L-band a repeat-pass interferometric SAR (InSAR) and a multi- (24 cm wavelength) SAR operated for three months beam lidar. before the failure of the spacecraft’s power system. SEASAT led to a series of NASA space shuttle-based In 2008, NASA appointed a DESDynI Science Study radar missions and inspired the development of space- Group (DSSG) to articulate specific science require- borne SAR systems worldwide. Launching another ments for the DESDynI mission and established free-flying scientific SAR in the U.S. has proven elusive, a pre-formulation project team at Jet Propulsion despite strong demand from the science and applica- Laboratory (JPL) and Goddard Space Flight Center tions community. (GSFC) to flow these requirements down to a specific mission implementation. JPL was overall project lead In 2007, the National Research Council Committee on and responsible for the SAR project element; GSFC was SEASAT LED TO A Earth Science and Applications from Space recom- responsible for the lidar project element. The DSSG mended a mission to measure changes in land, ice, wrote a Science Definition Document (SDD) describing SERIES OF NASA and vegetation structure, called DESDynI (Deformation, in great detail the science behind the mission and de- SPACE SHUTTLE- Ecosystem Structure, and Dynamics of Ice) as one of veloped a set of Level 1 and Level 2 requirements and the first in a series of Decadal Survey missions to carry preliminary science targets, including observing attri- BASED RADAR forward the nation’s spaceborne observation program. butes such as radar mode, sampling strategy, pointing MISSIONS AND The objective for DESDynI was to address the critical diversity, etc., which guided the project work. needs of three major science disciplines – Solid Earth, INSPIRED THE Ecosystems, and Cryospheric sciences – plus provide A complete mission concept was developed for DES- DEVELOPMENT data important for many applications. The primary DynI. The pre-formulation team conducted a Mission mission objectives for DESDynI were to: 1) Determine Concept Review (MCR) successfully in January 2011. OF SPACEBORNE the likelihood of earthquakes, volcanic eruptions, and After the MCR, NASA received direction from the US SAR SYSTEMS landslides through surface deformation monitoring; Administration (Office of Management and Budget) to 2) Characterize the global distribution and changes reformulate the concept to reduce its scope. The lidar WORLDWIDE. of vegetation aboveground biomass and ecosystem was to be removed as a component of the DESDynI structure related to the global carbon cycle, climate and program, and the cost of the radar project element was biodiversity; and 3) Project the response of ice masses to be reduced significantly. to climate change and impact on sea level. In addition, NISAR will provide observations that will greatly im- In May 2012, NASA competed and selected a DESDynI prove our monitoring of groundwater, hydrocarbon, and radar (DESDynI-R) Science Definition Team (SDT) to sequestered CO2 reservoirs. The Decadal Survey noted redefine DESDynI science to flow to an affordable, ra- that these surface processes can be characterized and dar-only NASA mission. Past and current SDT members monitored from space using SAR and Light Detection are listed in Tables 14-1 and 14-2 in Appendix B. At the and Ranging (lidar). Initial designs of DESDynI consisted same time, the JPL project team studied a number of 148 NISAR Science Users’ Handbook 149 NISAR Science Users’ Handbook

options to reduce cost and/or scope including partner- sensitivity at L-band to lower values while decreasing ships with other space agencies. sensitivity to ionospheric and soil moisture effects. 14 APPENDIX B

Through discussions between NASA and ISRO on the Since January 2012 when the initial L- and S-band SAR NISAR SCIENCE TEAM possibility of a joint radar mission, it became clear that mission concept was put forward as a partnership, JPL the goals originally identified for DESDynI-R were of and ISRO teams have been attempting to refine the sci- The NISAR science team members are drawn from the 14.1 NASA SCIENCE DEFINITION TEAM great interest to the ISRO science community. In Janu- ence plan and its implications for the mission. In Sep- science disciplines related to the mission. The science NASA selected a science definition team (SDT) known ary 2012, ISRO identified targeted science and applica- tember 2013, ISRO received initial approval from the team is renewed at three-year intervals and the as the “DESDynI-R SDT” in May 2012. The expertise of tions that were complementary to the primary mission Government of India for jointly developing with NASA makeup of the team evolves to fit the NISAR science the members spans the science disciplines identified objectives, agricultural monitoring and characterization, the L- and S-band SAR mission. A Technical Assistance and mission needs. An ISRO science team addresses in the 2007 NRC Decadal Survey of Earth Science landslide studies, Himalayan glacier studies, soil mois- Agreement (TAA) between ISRO and Caltech/JPL was the ISRO objectives. JPL also has an internal project and Applications for the DESDynI mission concept. ture, coastal processes, coastal winds and monitoring enacted on September 30, 2013. NASA Administrator science team to coordinate technical activities. hazards. For many of these objectives, the addition of Charles Bolden and K. Radhakrishnan, Chairman of TABLE 14-1. NASA SAR MISSION SCIENCE DEFINITION TEAM (2012-2015) an S-band polarimetric capability will add consider- ISRO, signed the NISAR Implementing Arrangement (IA) ably to the measurement, extending the measurement on September 30, 2014. SDT Member Institutional Affiliation Areas of Interest

Bradford Hager Massachusetts Institute Solid Earth Deformation Lead of Technology .

Ralph Dubayah University of Maryland Ecosystems Ecosystems Lead

Ian Joughin University of Washington Cryosphere Cryosphere Lead Applied Physics Lab

Gerald Bawden* US Geological Survey/ Hazards, hydrology, NASA HQ applications

Kurt Feigl University of Wisconsin Solid Earth

Benjamin Holt Jet Propulsion Laboratory Sea ice

Josef Kellndorfer Woods Hole Research Ecosystems, carbon policy Center

Zhong Lu Southern Methodist University Volcanoes

Franz Meyer University of Alaska, Applications, techniques, Fairbanks deformation

Matthew Pritchard Cornell University Solid Earth and cryosphere

Eric Rignot University of California, Cryosphere Irvine

Sassan Saatchi Jet Propulsion Laboratory Ecosystems

Paul Siqueira University of Massachusetts, Ecosystems, techniques Amherst

Mark Simons California Institute of Technology Solid Earth, hazards, cryosphere

Howard Zebker Stanford University Solid Earth, applications, techniques

*Transitioned from SDT member to NASA HQ after selection 150 NISAR Science Users’ Handbook 151 NISAR Science Users’ Handbook

“DESDynI-R” refers to the radar component of the 14.2 ISRO SCIENCE TEAM TABLE 14-2. NASA SAR MISSION SCIENCE DEFINITION TEAM (2016-2019) contd DESDynI concept. In addition to discipline scientists, the ISRO forms a science team once a project is approved, SDT Member Institutional Affiliation Areas of Interest team comprises applications and radar phenomenology which is the equivalent of entering formulation. Though experts. Table 14-1 lists the 2012-2015 SDT members, the project has been approved, a science team has not Sassan Saatchi Jet Propulsion Laboratory Ecosystems their affiliations and areas of interest. A solicitation was yet been formed. To date, ISRO science formulation Marc Simard Jet Propulsion Laboratory Ecosystems, techniques issued in 2015 to recompete the SDT. Selections were has been conducted by an ad hoc team of ISRO staff made in April 2016, and the new team has been in Howard Zebker Stanford University Solid Earth, applications, scientists. The ISRO scientists involved in defining the place since May 2016. The new team has 11 returning techniques ISRO specific science requirements through KDB-B are members and 9 new team members. Table 14-2 lists given in Table 14-3. the new science definition team. In April 2015, Dr. Chakraborty retired and Tapan Misra the project manager to coordinate the science and became the director of the Space Applications Centre. technical developments and calls on JPL staff scientists TABLE 14-2. NASA SAR MISSION SCIENCE DEFINITION TEAM (2016-2019) For most of Phase B, the science element at ISRO was to perform analysis in support of the SDT activities. SDT Member Institutional Affiliation Areas of Interest led by Dr. Raj Kumar. The project scientist conducts weekly coordination Mark Simons California Institute of Technology Solid Earth Deformation Lead 14.3 PROJECT SCIENCE TEAM teleconferences with the SDT leads, and alternate fortnightly teleconferences with the full SDT to maintain Paul Siqueira University of Maryland Ecosystems JPL maintains a project science team distinct from information flow and coordinate analysis, requirements Ecosystems Lead the SDT, headed by a project scientist, currently Paul definition, and documentation. Ian Joughin University of Washington Cryosphere Rosen. The project scientist works side-by-side with Cryosphere Lead Applied Physics Lab

TABLE 14-3. ISRO PRE-FORMULATION SCIENCE TEAM (PRIOR TO KDP-B) Cathleen Jones Jet Propulsion Laboratory Applications Applications Lead SDT Member Institutional Affiliation Areas of Interest

Falk Amelung University of Miami Solid Earth, atmospheres Tapan Misra* Space Applications Centre, Radar phenomenology, Ahmedabad lead prior to KDP-B Adrian Borsa Scripps Institution of Solid Earth Oceanography Manab Chakraborty** Space Applications Centre, Agriculture Raj Kumar (after KDP-B) Ahmedabad oceans, lead after KDP-B Bruce Chapman Jet Propulsion Laboratory Wetlands

Anup Das Space Applications Centre, Ecosystems Eric Fielding Jet Propulsion Laboratory Solid Earth Ahmedabad Richard Forster University of Utah Cryosphere Sandip Oza Space Applications Centre, Cryosphere Ahmedabad Bradford Hager Massachusetts Institute of Solid Earth Technology *Became SAC director in April 2016 ** Retired in April 2016 Benjamin Holt Jet Propulsion Laboratory Sea ice

Josef Kellndorfer Earth Big Data, LLC. Ecosystems, carbon policy

Rowena Lohman Cornell University Solid Earth and cryosphere

Zhong Lu Southern Methodist University Volcanoes

Franz Meyer University of Alaska, Fairbanks Applications, techniques, deformation

Frank Monaldo National Oceanic and Atmospheric Oceans, sea ice Administration

Eric Rignot University of California, Cryosphere Irvine NISAR Science Users’ Handbook 153 NISAR Science Users’ Handbook

15 APPENDIX C KEY CONCEPTS

This appendix covers the key concepts for the NISAR SAR techniques exploit the motion of the radar in orbit radar mission. The key concepts include an overview to synthesize an aperture (antenna), which typically of the radar imaging and the basic related science will be about 10-km long in the flight direction. This products that the mission will produce. principle is illustrated in Figure 15-1. While the radar is traveling along its path, it is sweeping the antenna’s 15.1 BASIC RADAR CONCEPTS: RADAR footprint across the ground while it is continuously IMAGING, POLARIMETRY, AND transmitting and receiving radar pulses. In this scenar- INTERFEROMETRY io, every given point in the radar swath is imaged many THIS APPENDIX For those unfamiliar with the NISAR mission, this times by the moving radar platform under constantly COVERS AN section gives a brief introduction to key concepts and changing yet predictable observation geometries. In terms that are central to NISAR science and mea- SAR systems, this change in observation geometry, OVERVIEW OF THE surements. These include radar imaging, polarimetry, resulting in a constant change of the distance from the RADAR IMAGING and interferometry concepts. There are a number of radar to the point on the ground, is precisely encod- excellent introductory books (Richards, 2009; van Zyl ed in the phase of the observed radar response. The AND THE BASIC and Kim, 2011; Hanssen, 2001) and book chapters) phase history for any point on the ground located at a constant distance parallel to the flight track is unique RELATED SCIENCE Simons and Rosen, 2007; Burgmann et al., 2000) on these subjects. to that point. By compensating the phase history of PRODUCTS THAT each pulse that is affecting a particular point on the ground, it is possible to focus the energy across the 10 THE MISSION WILL 15.1.1 Synthetic Aperture Radar km synthetic aperture and create an image of vastly PRODUCE. Synthetic Aperture Radar (SAR) refers to a technique improved resolution. The theoretically achievable for producing fine resolution images from an intrinsi- synthetic aperture resolution can be calculated from cally resolution-limited radar system. The wavelengths, D/2, is independent of the range or wavelength, and , that are used for radar remote sensing of the earth’s corresponds to D/2=5 m for the previously outlined surface are typically in the range of a few to tens of spaceborne scenario. centimeters. At these wavelengths, the energy radiated from a radar antenna of dimension D fans out over Through the outlined principles, SAR defeats the intrin- an angular range that is equivalent to the beam width sic resolution limits of radar antennas in the along- D of the antenna. For a typical spaceborne SAR track direction. In the cross-track or range direction, configuration with wavelengths of ~10 cm and an an- orthogonal to the satellite path, the resolution is not de- tenna of 10 m size, this beam width is 1/100 radians, fined by the antenna beam width, but rather the width or about 0.6 degrees. For a radar in space observing of the transmitted pulse. Referring to Figure 15-1, this the Earth 1000 km below, the beam size on the ground is because the transmitted pulse intersects the imaged is then 1000 D = 10 km. This intrinsic resolution of surface as it propagates in the beam. After a two-way the radar system is insufficient for many applications trip of a transmitted pulse from sensor to the ground and practical solutions for improving the resolution and back, two objects can be distinguished if they are needed to be found. spatially separated by more than half the pulse width. Hence, range resolution is controlled by the transmit- 154 NISAR Science Users’ Handbook 155 NISAR Science Users’ Handbook

[ Figure 15-1] A polarimetric radar can be designed to operate as a and random surfaces, while better balancing the power Configuration of a radar in single-pol system, where there is a single polarization between the receive channels. motion to enable synthetic transmitted and a single polarization received. A typical aperture radar imaging. Radar single-pol system will transmit horizontally or vertically Classical radar polarimetry focuses on relating the antenna illuminates an area polarized waveforms and receive the same (giving HH complex backscatter observed in various polarimetric on the ground determined by Radar Antenna Radar Flight Path or VV imagery). A dual-pol system might transmit a combinations to the electrical and geometric properties its wavelength and antenna horizontally or vertically polarized waveform and mea- of the observed surfaces in order to extract meaningful dimension. Pulses are sent Radar Pulse In Air sure signals in both polarizations in receive (resulting information. Observation-based empirical work, as well and received continuously in HH and HV imagery). A quad-pol or full-pol system as theoretical modeling, helps establish these relation- such that any point on the will alternate between transmitting H-and V-polarized ships. For example, over soils, surface roughness and ground is sampled often. The waveforms and receive both H and V (giving HH, HV, VH, moisture both contribute to the backscattered ampli- range/phase history of each VV imagery). To operate in quad-pol mode requires a tude, but it can be shown that HH and VV images have point is compensated to focus pulsing of the radar at twice the rate of a single- or similar responses to roughness, such that the ratio HH/ energy acquired over the dual-pol system since the transmit polarization has VV is primarily an indicator of moisture content. As an- synthetic aperture time to fine resolution. In range, resolution to be alternated between H and V in a pulse-by- other example, bare surfaces have a weak depolarizing is achieved by coding the pulse manner to enable coherent full-polarized data effect, while vegetation canopies generally are highly pulse with a wide bandwidth acquisitions. Since this type of operation can cause depolarizing. So, a joint examination of the dual-pol interference between the received echoes, a variant channels HH and HV can distinguish these surface signal waveform. Radar Pulse Sweeping Across of quad-pol known as quasi-quad-pol can be used, types. Path whereby two dual-pol modes are operated simultane-

Radar Swath ously: an HH/HV mode is placed in the lower portion For this mission, quantifying biomass is an important Azimuth Direction of the allowable transmit frequency band and a VH/VV measurement objective. Empirical relationships have Continuous Stream Range Direction of Pulses mode is operated in the upper portion. Being disjoint been developed that allow mapping of radar back- in frequency, the modes do not interfere with each scatter amplitude to the amount of biomass present in other. However, the observed HH/HV and VH/VV data are an image resolution cell. The relationship varies with mutually incoherent. vegetation type and environmental conditions (e.g. soil moisture and roughness), but with multiple polariza- ted waveform that is generated by the radar and not 15.1.2 Polarimetry While most spaceborne systems are linearly polarized, tions and repeated measurements, the biomass can be the size of the antenna footprint on the ground. Wider it is also possible to create a circularly polarized signal determined with sufficient accuracy. A radar antenna can be designed to transmit and bandwidth signals generate finer resolution images in on transmit, whereby the tip of the electric field vector receive electromagnetic waves with a well-defined range. is rotating in a circle as it propagates. This is typically 15.1.3 Interferometric Synthetic Aperture polarization, which is defined as the orientation of implemented by simultaneously transmitting equal Radar (InSAR) the electric field vector in the plane orthogonal to the For most purposes, the transmitted signal can be amplitude H and V signals that are phase shifted by wave propagation direction. By varying the polar- As noted above, each resolution element encodes the thought of as a single frequency sinusoid with a 90 degrees. Various combinations of right-circular and ization of the transmitted signal, SAR systems can phase related to the propagation distance from the well-defined amplitude and phase. Thus, the image left-circular polarization configurations on transmit and provide information on the polarimetric properties of radar to the ground as well as the intrinsic phase of the constructed from the SAR processing is a complex receive allow synthesizing single-, dual-, and quad-pol the observed surface. These polarimetric properties backscattering process. The resolution element com- image – each resolution element, or pixel, has an am- mode data from circular-polarized observations. are indicative of the structure of the surface elements prises an arrangement of scatterers – trees, buildings, plitude and phase associated with it. Once calibrated, Circular polarization is relevant to NISAR as recent work within a resolution element. Oriented structures such people, etc. – that is spatially random from element the amplitude is proportional to the reflectance of the has emphasized the benefits of hybrid polarization, as buildings or naturally aligned features (e.g. sand to element and leads to a spatially random pattern surface. The phase is proportional to the distance the where a circularly polarized wave is transmitted and ripples) respond preferentially to similarly oriented of backscatter phase in an image. As such, since we wave traveled between the radar and the ground, any H and V signals are received. The dual-pol instance polarizations and tend to preserve polarimetric coher- can only measure the phase in an image within one propagation phase delays due to the atmosphere or of this mode is known as compact-pol. Compact-pol ence, whereas randomly oriented structures lead to cycle (i.e. we do not measure the absolute phase), it is ionosphere, and any phase contribution imparted by the captures many of the desirable scattering properties of depolarization of the scattered signals. reflectance from the surface. a dual-pol system, e.g. discriminating between oriented 156 NISAR Science Users’ Handbook 157 NISAR Science Users’ Handbook

not possible to observe the deterministic propagation topography, the unwrapped (module 2π) InSAR phase S2 4πΔd S1 component directly. Δϕ = ϕ −ϕ −ϕ′ =is proportional to the ground [ Figure 15-2] 2 1 λ deformation ∆d between t1 and t2 along the satellite Illustration of InSAR imaging geometry. Interferometric synthetic aperture radar (InSAR) (Rosen LOS ground direction as: The distance between the satellite at et al., 2000; Hanssen, 2001) techniques use two or 4πΔd S1 and a ground pixel A is r1 and the r2 Δϕ = ϕ −ϕ −ϕ′ = more SAR images over the same region to obtain sur- 2 1 λ distance between the satellite at S2 and Z face topography or surface motion. In this section, we where λ is the radar wavelength. In this equation, the ground pixel A is r2. The topographic r1 explain how an InSAR phase measurement relates to we assume that there is no error in the InSAR phase height of the pixel A is z. Here we assume |r1 − r2| << |r1| (the parallel-ray approxi- actual ground deformation (see Massonnet et al. 1993; measurement. Below we discuss in depth various error A mation) and no ground deformation a tutorial summary is found in Chen 2014a). sources in InSAR deformation measurements and their occurs at pixel A between the two SAR impact on InSAR image quality. data acquisition times. Ground Figure 15-2 illustrates InSAR imaging geometry. At time t1, a radar satellite emits a pulse at S1, then receives Note that InSAR techniques only measure one-dimen- an echo reflected from a ground pixel, A, and measures sional LOS motion. However, deformation is better the phase φ1 of the received echo. All scatterers within characterized in three dimensions: east, north and up. the associated ground resolution element contribute Given an LOS direction unit vector e = [e1, e2, e3], to φ1. As a result, the phase φ1 is a statistical quantity we can project the deformation in east, north and up [ ] that is uniformly distributed over interval (0, 2π) so coordinates along the LOS direction as: Figure 15-3 that we cannot directly use φ1 to infer the distance r1 InSAR deformation geometry. Δd = e Δd + e Δd + e Δd between S1 and A. Later at time t2, the satellite emits 1 east 2 north 3 up At time t1, a ground pixel of another pulse at S2 and makes a phase measurement Because radar satellites are usually polar orbiting, the interest is at point A and a radar satellite measures the phase 1 φ2. If the scattering property of the ground resolution north component of the LOS unit vector e2 is often φ between the satellite and the ground LOS Direction element has not changed since t1, all scatterers within negligible relative to the east and vertical components. pixel along the LOS direction. Later the resolution element contribute to φ2 the same way When InSAR measurements along two or more LOS at time t2, the ground pixel moves to as they contribute to φ1. Under the assumption that directions are available, we can combine multiple LOS A′ and the satellite makes another f1 |r1 − r2| << |r1| (the parallel-ray approximation), the deformation measurements over the same region to measurement φ2 between the satel- phase difference between φ1 and φ2 can be used to separate the east and vertical ground motions, given lite and the ground pixel. The phase f2 infer the topographic height z of the pixel A (Hanssen, that the term e2 ∆d is negligible. north difference ∆φ is proportional A 2001, Section 3.2). to the ground deformation between For this mission, interferometric observations of any t1 and t2 along the LOS direction. If we know the topographic height z, we can further given point on the ground are acquired every 12 days. Ground at t1 Df measure any small ground deformation occurring at To the extent that the ground does not change appre- pixel A between t1 and t2. Figure 15-3 illustrates the ciably in most places over 12 days, ground motion can A' InSAR imaging geometry in this case. At time t1, a be measured. Ground at t'2 radar satellite measures the phase φ1 between the satellite and a ground pixel A along the LOS direction. 15.2 DEFORMATION-RELATED Later at time t2, the ground pixel A moves to A′ and the TERMINOLOGY satellite makes another phase measurement φ2 between the satellite and the ground pixel. Chen (2014b) The Earth’s crust and cryosphere deform due to differ- After removing the known phase φ′ due to the surface ence forces acting on them. Deformation may be linear, episodic, or transient. 158 NISAR Science Users’ Handbook 159 NISAR Science Users’ Handbook

15.2.1 Deformation and Displacement To quantify deformation using radar, the mission offers albeit with more noise and poorer resolution. Where For example, a difference of 0.1 cycles in phase or at least two approaches: available, the less noisy phase data can be combined 2.8 mm in range change over the 100 m distance In general, the term deformation refers to the change in with the azimuth offsets to produce a less noisy vector between adjacent pixels in a C-band interferogram shape of a solid or quasi-solid object. In the context of 1. Degrees or cycles. One fringe in an interferogram estimate of displacement (Joughin 2002). corresponds to a range gradient of ψ ~ 2.8 x 10–5. this mission, surface deformation refers to the change corresponds to one cycle (= 2π radians) of phase While range change is one component of the displace- in shape as observed on the Earth’s free surface, i.e., change or half a wavelength in range change. 15.2.2 Strain, Gradients, and Rotation ment vector (measured in millimeters), its (dimension- the interface separating the atmosphere from the Phase is ambiguous because it is defined as an less) gradient is one component of the “deformation uppermost layer of solid Earth, whether rock, soil, ice or For a one-dimensional element, the strain ε is ex- angle on the unit circle such that π ≤ Δφ ≤ π. gradient” tensor F (Malvern, 1969). Unlike wrapped a combination thereof. pressed as the dimensionless ratio of its change length ij Since the phase change is known only to within phase change, the range change gradient is continuous ΔL to its original length L, such that ε = ΔL/L. If one an integer number of cycles (i.e. modulo 2π), and differentiable (Sandwell and Price, 1996), offering a Displacement u is a vector quantity defined as the end of the element is held fixed, then the change in it is called “wrapped”. Converting ambiguous, number of advantages for streamlining data analysis. change in a particle’s position X between one instant length is equal to the displacement of the other end, wrapped phase change Δρ in radians to range in time (epoch) t1 and a later epoch t2, such that u = such that ΔL = u. For small strains, we can think of change Δρ in millimeters requires unwrapping 15.2.3 Stress and Rheology X(t2) – X(t1). Typically, displacement is calculated with the strain as the gradient of the displacement, i.e., the algorithms (e.g., Chen and Zebker, 2002; Hooper respect to the particle’s initial, fixed positionX (t0) at partial derivative of displacement u with respect to the Stress is the force applied to a body per unit area. It and Zebker, 2007). some reference time t0. position coordinate x. Generalizing to three dimensions also can imply the resistance a solid body offers to an 2. The techniques called speckle tracking and yields a second-order tensor called the deformation applied force. In mechanics, constitutive relationships feature tracking estimate the shift of an image gradient tensor F = ∂ /∂ (Malvern, 1969). The de- describe how stress and strain depend on each other. The change in range Δρ is a scalar quantity equal ij ui xj to the change in the (1-way) distance from the radar patch relative to its neighbors by cross-correlating formation tensor can be decomposed into a symmetric The study of constitutive relationships and the relevant sensor to the target pixel on the ground. Range change the amplitudes or complex values of two images part, called the strain tensor, and an anti-symmetric material properties is called rheology. For example, is a particular component of the displacement vector. covering the same location at two different times part called a rotation or spin. The temporal derivatives during an earthquake, the Earth’s crust deforms with To calculate the range change, we project the displace- (e.g., Vesecky et al., 1988). To do so, the technique of these quantities are called the velocity gradient an elastic rheology according to Hooke’s Law. Under- · generates “normalized cross-correlation” of image tensor Lij, the strain rate tensor E , and the spin rate standing Earth’s rheology is one of the primary goals of ment vector u onto the line of sight using the scalar ij · patches of complex or detected real-valued SAR tensor Ω respectively (Malvern, 1969). the mission. (“dot”) product such that Δρ = – u • ^s, where ^s is ij a unit vector pointing from the target on the ground images. The location of the peak of the two-di- toward the radar sensor in orbit. If the target moves mensional cross-correlation function yields the SAR interferometry is especially sensitive to gradients 15.3 ECOSYSTEMS-RELATED toward the sensor, then the distance between them image offset (displacement). of the displacement field. For example, if a rock outcrop TERMINOLOGY decreases and the range decreases such that Δ < 0. 10 meters in width stretches by 10 mm, then the strain ρ NISAR addresses the amount living material in The successful estimation of the local image offsets will be = 0.001. Similarly, if the same outcrop tilts ε ecosystems as well as the disturbance and recovery Line-of-sight (LOS) displacement u is a scalar quan- depends on having correlated speckle patterns (speck- by 10 mm (about a horizontal axis) or spins (about a LOS of ecosystems. tity that is equal in absolute value to the range change le tracking) and/or the presence of nearly identical vertical axis), then the angle of rotation will be approx- features (feature tracking) in the two SAR images at the imately 1 milliradian. Such behavior was observed in Δρ. Most, but not all, authors reckon upward motion of 15.3.1 Biomass the target (toward the sensor) to be a positive value of scale of the employed patches. If speckle correlation interferogram of the deformation field produced by the is retained and/or there are well-defined features, the Landers earthquake in California (Peltzer et al., 1994). Biomass is defined as the total mass of living matter LOS displacement, such that uLOS > 0. tracking with image patches of 10s to 100s of meters within a given unit of environmental area, usually in size can be performed to a tenth-of- a-pixel or better To quantify the deformation gradient tensor F , we can measured as mass or mass per unit area of dry weight. Velocity v is a vector quantity defined as the derivative ij of displacement with respect to time t such that v = accuracy with improved accuracy at the expense of differentiate the wrapped phase in an interferogram to Biomass is a fundamental parameter characterizing du/dt. In discussing velocity fields, it is important to resolution by averaging adjacent estimates (Gray et al., find the range gradientψ . Following Sandwell and Price the spatial distribution of carbon in the biosphere. The define the reference frame. The relative velocity of a 1998; Michel and Rignot, 1999; Strozzi et al., 2008). (1996), Ali and Feigl (2014) take the discrete derivative NISAR mission will focus on the above-ground biomass particle j with respect to particle i is vj,i = vj – vi. A The result yields two horizontal components of the of range change Δρ with respect to a horizontal coordi- of woody plants and forests, comprising about 80% so-called absolute velocity is taken with respect to a displacement vector. Of these, the component that is nate in position X to define the observable quantity for of terrestrial total biomass in vegetation (Houghton, fixed origin located at positionX 0 that is assumed to be parallel to the ground track of the satellite is also called the kth pixel as: 2005; Cairns et al., 1997). Half of all biomass in woody an azimuth offset. The other effectively measures the Δρ (k+1) − Δρ (k−1) stationary, such that v0 = 0. ψ k = same range displacement as the interferometric phase, X (k+1) − X (k−1) 160 NISAR Science Users’ Handbook

vegetation is carbon equivalent to approximately particular disturbance. The nature and rate of recovery

3.67 units of CO2 that directly links biomass to the depend on the size and severity of disturbance and the terrestrial carbon cycle and climate change (IPCC Good pre-disturbance state of the ecosystem (Frokling et Practice Guide, 2003). al., 2009; Chazdon et al., 2001). Recovery can follow a prescribed trajectory to meet certain production goals 15.3.2 Disturbance in managed ecosystems and or a natural trajectory depending on environmental conditions in the case Disturbance is defined as a discrete event that involves of unmanaged ecosystems. We focus on recovery the removal of biomass, mortality, or change in the as a process or trajectory defined by the area of the structure and is considered the major agent in deter- post-disturbance growth of biomass at scales of dis- mining the heterogeneity of forest ecosystems across turbance (> 100 m) and measured in the units of area/ a broad range of scales in space and time. Forest year or mass/area/year. disturbance can be abrupt (e.g., hurricanes) or chronic (e.g., acid rain); stand-replacing (e.g., clear-cut logging) 15.3.4 Classification or not (e.g., selective logging); complete (e.g., landslides) or incomplete (e.g., insect defoliation); natural Classification is the problem of identifying to which (e.g., tornados) or anthropogenic (e.g., land conver- set of categories a set of objects or new observations sion); widespread (e.g., fire) or geographically restricted belongs on the basis of their relationships or charac- (e.g., avalanches); temporary (e.g., blow downs) or teristics. This includes the classification of observations permanent (deforestation and land use conversion) into events, processes, or thematic categories that (Frolking et al., 2009; Chambers et al., 2013). We focus impact the vegetation structure, biomass, cover, and on disturbances as abrupt events that cause changes characteristics. Classification involves definition of in forest biomass and are at the scale detectable by class boundaries based on objective criteria including spaceborne remote sensing (> 100 m). Disturbance is type, scale, and source used in a diagnostic system as measured as the area and/or the intensity of biomass a classifier. We focus on classification of SAR imagery changes in units of area/year or mass/area/year. as the new observations, at landscape scale (> 100 m) into disturbance (e.g. deforestation, degradation), re- 15.3.3 Recovery covery (e.g. biomass regrowth) or change of vegetation status (e.g. inundation). Recovery of forests and woody vegetation refers to the reestablishment or redevelopment of above ground biomass and structure characteristics after the impact of a NISAR Science Users’ Handbook 163 NISAR Science Users’ Handbook

16 APPENDIX D BASELINE LEVEL 2 REQUIREMENTS

The Level 1 requirements capture the essential ele- accurate measurements preceding and soon after the ments of the measurements by discipline and expand event are needed. to greater detail at Level 2. The baseline L 2 requirements capture the specific measurements that will Resolutions vary depending on science focus. Interseis- be validated by research area and product type. Table mic deformation is generally broad except at aseismi- 16-1 shows the high-level mapping from L1 to L2. The cally creeping faults, so low resolution is adequate. For colors codify the relationships at Levels 1 and 2. The deformation associated with earthquakes, volcanoes, tabs on the corners of the requirements boxes indicate subsidence and landslides, the spatial patterns of the radar technique used to make the measurements. It deformation are finer, so finer resolution is required. should be clear that there are a limited number of tech- The coseismic deformation requirement for large earth- niques used to support a multiplicity of requirements, quakes specifies the entire land surface. Unlike the which should help reduce the amount of validation interseismic and targeted requirements, this require- required. ment ensures that there will be observations sufficient to capture events outside of the areas that are known to be deforming rapidly. 16.1 LEVEL 2 SOLID EARTH

Table 16-2 itemizes the solid Earth L2 require- The validation requirement is included to specify that ments, which comprise interseismic, co-seismic, and the validation program is limited in scope to fixed post-seismic deformation, and a set of additional areas, employing analysis to extrapolate performance deforming sites on land that include volcanoes, land- to the globe. slide-prone areas, aquifers, and areas of increasing relevance such as hydrocarbon reservoirs and seques- 16.2 LEVEL 2 CRYOSPHERE tration sites. Table 16-3 identifies the cryosphere L2 requirements. These requirements have a very similar form to the The requirements for deformation are specified in solid earth requirements, as they too map to geodetic terms of accuracy over relevant relative length scale, imaging methods. Ice sheets and glaciers move quickly, and at a particular resolution, and vary depending on so the faster the sampling rate for measurements, the the style of deformation and its expected temporal greater the coverage of dynamic processes and the variability. Interseismic deformation is specified in more interesting the science. Sea-ice moves even fast- terms of a relative velocity over a given length scale. er. Thus, the cryosphere L2 requirements are explicit Interseismic deformation is slow, on the order of cm/yr. in terms of the required sampling as well as regionally To adequately model this deformation, accuracies far differentiated velocity accuracy and resolution require- better, on the order of mm/yr, are required. To achieve ments. good accuracy, often many measurements are needed over time to reduce noise via averaging. Co-seismic de- The permafrost requirement is explicitly a deformation formation is extremely rapid, on the order of seconds, requirement similar to the solid Earth requirement. It with post-seismic deformation occurring thereafter, will be considered one of the targeted sites in the solid and generally larger in magnitude, so frequent, less Earth L2 requirements but is included here to call out demamiel/Shutterstock 164 NISAR Science Users’ Handbook 165 NISAR Science Users’ Handbook

TABLE 16-1. LEVEL 2 REQUIREMENTS FOR SOLID EARTH TABLE 16-2. SOLID EARTH LEVEL 1 AND 2 REQUIREMENTS SUMMARY

LEVEL 1 LEVEL 2 Attribute Secular Co-seismic Transient Deformation Deformation Deformation I S (658) (660) (663) I S I S 2-D Solid Earth Secular Coseismic Transient Measurement Spatially averaged Point-to-point rela- Point-to-point relative Displacement Deformation Deformation Deformation relative tive displacements displacements in two velocities in two in two dimensions dimensions dimensions I S I S I Method Interferometry, Interferometry, Interferometry, 2-D Ice Sheet & Slow Ice Sheet Fast Ice Sheet Permafrost speckle tracking speckle tracking speckle tracking Glacier & Glacier Velocity Velocity Deformation Displacement Duration 3 years 3 years Episodic over mission, depending on science target I S I

Product 100 m; smoothed 100 m 100 m Sea Ice Sea Ice Time-Variable Vertical Resolution to according Velocity Velocity Velocity Motion to distance scale L F

Accuracy 2 mm/yr or better, 4 (1+L1/2) mm or 3 (1+L1/2) mm or better, 0.1 km < L better, 0.1 km < L 0.1 km < L < 50 km, over I C Technique < 50 km, over < 50 km, over > 70% > 70% of ~ 2,000 target- Biomass & Biomass Disturbance > 70% of of coverage areas ed sites Disturbance Estimation Classification I = Interferometry coverage areas P P S = Speckle Tracking Sampling One estimate over 4 times per year to Every 12 days, two C = Coherence 3 years, two direc- guarantee capture directions P = Polarimetry C C tions of any earthquake F = Feature Tracking on land before Cropland, Cropland Inundation surface changes too Inundation Area Area Area = Experimental greatly P P

Coverage Land areas pred- All land, as earth- Post-seismic events, icated to move quake locations are volcanoes, ground-wa- faster than 1 mm/ known a priori ter, gas, hydrocarbon yr reservoirs, land- the explicit cryosphere focus. The validation require- cation accuracy must be met for all biomass. This key slide-prone ment is included to specify that the validation program implication of this is that the observing strategy must Response N/A 24-hour tasking, 24/5 is limited in scope to fixed areas, employing analysis to include sufficient global observations of biomass to 5-hour data delivery. Latency Best effort basis on extrapolate performance to the globe. enable this classification. Best effort basis on event event 16.3 LEVEL 2 ECOSYSTEMS The requirements related to wetlands, areas of inundation and agriculture are “globally distributed”, implying Table 16-4 identifies the ecosystems L2 requirements. regional measurements, as specified in the science These parallel but give greater specificity to the L1 re- implementation plan target suite. quirements. The woody biomass accuracy requirement is the same at Levels 1 and 2. In addition, the details of In addition to the requirements, the ecosystems sub- the requirement for classification are spelled out. While group has identified a goal to determine the ability of biomass is required to meet its accuracy requirements NISAR to estimate vertical canopy structure. Current only where biomass is below 100 Mg/ha, the classifi- research in polarimetric interferometry shows that 166 NISAR Science Users’ Handbook 167 NISAR Science Users’ Handbook

TABLE 16 3. LEVEL 2 BASELINE REQUIREMENTS FOR CRYOSPHERE TABLE 16-4. LEVEL 2 BASELINE REQUIREMENTS FOR ECOSYSTEMS

Ice Sheets & Glaciers Ice Sheets & Glaciers Ice Sheet Attribute Biomass (673) Disturbance (675) Inundation (677) Crop Area (679) Attribute Velocity Slow Velocity Fast Time-Varying Deformation (667) Deformation (668) Velocity (738) Measurement Biomass Areal extent Areal extent Areal extent

Point-to-point displacements in Point-to-point displacements in Point-to-point displacements in Measurement Method Polarimetric Polarimetric Polarimetric Polarimetric back- two dimensions two dimensions two dimensions backscatter backscatter backscatter scatter contrast and to biomass temporal change contrast temporal change Method Interferometry, speckle tracking Interferometry, speckle tracking Interferometry, speckle tracking Duration 3 years 3 years 3 years 3 years Duration 3 years 3 years 3 years Product 100 m 100 m 100 m 100 m 100 m 250 m 500 m Product Resolution Resolution Accuracy 20 Mg/ha or better 80% or better classifi- 80% or better 80% or better 3% of the horizontal velocity 3% of the horizontal velocity 3% of the horizontal velocity mag- Accuracy where biomass is cation accuracy where classification accuracy classification accuracy magnitude plus 1m/yr or better, magnitude plus 5 m/yr or better, nitude plus 10 m/yr or better, over < 100 Mg/ha, over canopy cover changes over > 90% of coverage areas over > 90% of coverage areas > 80% of coverage areas 80% of coverage by >50% areas Sampling Each cold season, two directions Each cold season, two directions Each 12 days, two directions Sampling Annual Annual Seasonal, sampled Quarterly; sampled Areas moving slower than Areas moving slower than Outlet glaciers, or other areas of Coverage every 12 days to track every 12 days to track 50 m/yr of both poles and 50 m/yr of both poles seasonal change beginning and end of beginning and end of glaciers and ice caps flooding events growing season

N/A N/A 24/5 Response Coverage Global areas of Global areas of woody Global inland and Global agricultural Best effort basis on event Latency woody biomass biomass coastal wetlands areas

Grounding Line Sea Ice Response N/A 24/5 24/5 N/A Attribute Permafrost Vertical Velocity (670) Latency Best effort basis on Best effort basis on Displacement (671) Displacement (445) event event

Measurement Spatially averaged in two Point-to-point displacements in Point-to-point relative horizontal dimensions two dimensions displacements when temporal decorrelation is small and the inter- 16.4 LEVEL 2 URGENT RESPONSE Method Interferometry Interferometry, speckle tracking Backscatter image feature tracking ferometric baselines are large (but not too large) it is There is no urgent response L2 science requirement. possible to retrieve canopy structure from the data. Duration 3 years 3 years 3 years While the mission envisioned by the community – one As a repeat-pass interferometer designed for two with relatively fast revisit and a capacity for acquiring Product 100 m 100 m Gridded at 5 km disciplines to have small baselines, NISAR is not ideally data over the globe – can serve an operational need for Resolution suited to this technique. However, the dense interfer- reliable, all-weather, day/night imaging in the event of Accuracy 4*(1+L^1/2) mm or better, over 100 mm or better, over 95% of 100 m/day, over 70% of coverage ometric time-series may lead to new innovations that a disaster, the project has been sensitive to the costs length scales 0.1 km < L < 50 km, coverage areas annually, over area allow structure estimates of value, and the team would > 80% of coverage areas 50% of coverage areas monthly associated with operational systems that must deliver like to explore these possibilities. such data. However, demonstrating the utility of such Sampling In snow-free months sufficient to Monthly Every 3 days meet accuracy (semi-monthly) data for urgent response for the benefit of society is The validation requirement is included to specify that important and in keeping with the recommendations Coverage Targeted priority regions in Greenland and Antarctic coastal Seasonally-adjusted Artic and the validation program is limited in scope to fixed of the 2007 Decadal Survey. To that end, the NISAR Alaska and Canada zones Antartic sea ice cover areas, employing analysis to extrapolate performance L1urgent response requirement has been written with Response N/A 24 hour tasking, 5 hour data 24/5 to the globe. Latency delivery. Best effort on event Best effort basis on event a focus on targeting and delivery latency, as previously described. The L1 urgent response requirement flows to other L2 mission requirements, but not to science directly. NISAR Science Users’ Handbook 169 NISAR Science Users’ Handbook

17 APPENDIX E NISAR SCIENCE FOCUS AREAS

This appendix provides additional background and • Characterize aquifer physical and mechanical rationale for the science objectives to be addressed properties affecting groundwater flow, storage, by NISAR. Each section describes the 2007 Decadal and management.

Survey objectives that guided the development of the • Map and model subsurface reservoirs for efficient requirements for NISAR in each major science focus hydrocarbon extraction and CO2 sequestration. area, amplifying their importance through examples • Determine the changes in the near surface stress in the literature that were generated from existing field and geometry of active fault systems over data – something that can only loosely approximate the major seismically active regions in India. richness of the results that will be derived from NISAR’s dense spatial and temporal data set. • Determine land subsidence rates of major reported land subsidence areas (due to mining and/or 17.1 SOLID EARTH groundwater induced) in India. • Map major landslide prone areas in the hilly The 2007 Decadal Survey identified the following regions of India. overarching science goals and related questions for solid Earth: • Determine the likelihood of earthquakes, volcanic These objectives require dense spatial coverage of eruptions, landslides and land subsidence. How Earth, and dense temporal sampling to measure, char- can observations of surface deformation phenom- acterize and understand these often unpredictable and ena lead to more complete process models for dynamic phenomena. earthquakes, volcanoes, landslides and land subsidence and better hazard mitigation strategies? In situ GNSS arrays constrain the large-scale motions of Earth’s surface where the arrays exist. In particular, • Understand the behavior of subsurface reservoirs. these GNSS data can provide temporally continuous • Observe secular and local surface deformation on point observations that are best exploited when com- active faults to model earthquakes and earthquake bined with the spatially continuous coverage provided potential. by the InSAR imaging that NISAR will provide. U.S. Geological Survey scientists • Catalog and model aseismic deformation in Randy Jibson and Jon Godt regions of high hazard risk. With NISAR, scientists will be able to comprehensively investigate the Seaside landslide generate time-series of Earth’s deforming regions. that was triggered by the 2016 • Observe volcanic deformation to model the volca- magnitude 7.8 Kaikoura, New no interior and forecast eruptions. When combined with other sources of geodetic imaging – optical satellite imagery when daytime, cloud-free Zealand, earthquake, along a • Map pyroclastic and lahar flows on erupting vol- observations are available and when expected ground fault that experienced significant canoes to estimate damage and model potential surface offsets. displacements are large; international SAR imagery future risk. when data are available and of suitable quality – an • Map fine-scale potential and extant landslides to even more complete picture of Earth’s 3-D motions can assess and model hazard risk. be constructed. USGS/Kate Allstadt. 170 NISAR Science Users’ Handbook 171 NISAR Science Users’ Handbook

17.1.1 Earthquakes and Seismic Hazards elastic strain (i.e., the spatial gradient in displacements) [ Figure 17-1] –60 is accumulating most rapidly (not shown in the figure) NISAR data will address several aspects of earthquake Daily position time series at are those where earthquakes are most likely. Temporal –60 physics and seismic hazards including: GNSS site CARH for north changes in the elastic stressing rate such as occurred 1. Determine crustal strains across the different (blue) and east (red) compo- at CARH in March 2004, are associated with temporal –60 phases of the seismic cycle. Because Earth’s nents, illustrating the coseis- changes in the probability of earthquake occurrence. mic offset and postseismic upper crust is elastic, inter-seismic deformation deformation associated North rates can be mapped to stressing rates, which –60 Mw 6.0 Parkfield Understanding coseismic fault slip magnitude and with the Mw 6.0 Parkfield East in turn are used to guide assessments of future Equation geometry, as well as regional local deformation signals earthquake (September 27, earthquake occurrence. –60 such as triggered slip can lead to understanding of 2004, dashed line). A linear 2. Derive physics-based models of faulting and changes in surface deformation on nearby (and distant) trend has been subtracted Displacement (Detrended, Mm) Displacement (Detrended, from each component using –60 crustal rheology consistent with multi-compo- faults. For example, CARH lurched about 15 mm to both the time period January nent displacement maps across all phases of the north and west at the time of the Mw 6.0 Parkfield 2008-January 2018. –60 the seismic cycle, complementing conventional earthquake. These detailed displacement measure- land-based seismological and geodetic measurements allow inference of the magnitude and sense of JAN 04 APR 04 JUL 04 OCT 04 JAN 05 APR 05 JUL 05 OCT 05 JAN 06 ments. Estimates of rheological parameters are slip on the fault plane during the earthquake. Changes essential to understand transfer of stress within in deformation rates on distant faults are then mon- fault systems. itored for evaluating any “linkages”. Because of the which are in turn used to understand strong ground 3. Assimilate vector maps of surface deformations sharp discontinuity in surface displacement and imme- seismic tomographic models of Earth’s interior struc- motions that impact the built environment. These through various stages of the earthquake cycle in diate postseismic deformation, a rapid repeat sampling ture by reducing the tradeoff between seismic wave kinematic models are among the few constraints we large-scale simulations of interacting fault sys- strategy permits accurate determination of coseismic velocities and source locations. Detailed understanding have on the underlying physics that shape our under- tems, currently a “data starved” discipline. displacements, which can otherwise be obfuscated by of the location and mechanism of small earthquakes standing of earthquake rupture mechanics. Such well- postseismic deformation occurring between the time of are also essential to illuminate important faults. These earthquake parameters provide important ingredients constrained co-seismic earthquake source models are THE EARTHQUAKE CYCLE the earthquake and the time of the first observation. also routinely compared with inferences of earth- The post-seismic deformation field immediately fol- when estimating the state of stress and changes in the quake magnitudes from geological field observations, Deformation of Earth’s crust in tectonically active lowing an earthquake can be significant, with deceler- state of stress in the crust, as well as indicators of the providing a needed calibration of paleo-seismological regions occurs on a rich variety of spatial and temporal ating surface displacements in the following week to boundaries between creeping and non-creeping fault inferences of historic earthquakes. scales. To date, the best temporal sampling is obtained months and possibly years for larger earthquakes. Such segments of a given fault. from continuously operating GNSS sites. Figure 17-1 post-seismic displacements as a function of time are The fusion of multiple imagery sources illustrates the shows the evolution of displacements at the GNSS site frequently characterized by a logarithmic dependence Larger earthquakes: NISAR will provide maps of power of geodetic imaging (here a combination of Carr Hill (CARH), near Parkfield, CA. The Mw 6.0 Park- on time consistent with a frictionally controlled fault surface faulting complexity and will constrain first radar and optical geodetic imaging) to constrain the field earthquake of September 27, 2004, corresponds slip process (as opposed to viscous processes). Within order geometric variability of the coseismic rupture at complex curved surface trace of the 2013 Mw 7.7 to the discrete jump midway through the time series. the time interval shown in the figure, rates have not yet depth. Spatially continuous maps (combined with GNSS Pakistan earthquake (Figure 17-2). The geodetic data The average secular motion (i.e., linear trend) has been returned to those observed preceding the earthquake. data when available) of surface displacements provide subtracted from each component. These time series critical constraints on models of coseismic fault rupture also require the dip of the fault to approximately 45 degrees from vertical, thereby documenting this event can be divided conceptually into three parts: (1) The COSEISMIC DEFORMATION for both small and large earthquakes. The geodetic im- “interseismic” part that occurs in the interval several aging data of the kind that will be routinely provided by as the first example of a large strike slip event on a years after the previous earthquake until just before Small earthquakes: NISAR will provide unique NISAR has already been shown to be crucial in estimat- non-vertical fault – well outside the expectations from the most recent earthquake; (2) the coseismic step at observations of ground displacement that will improve ing the distribution of co-seismic slip on the subsurface conventional faulting theory. Such “surprise” events the time of the earthquake; and (3) the “postseismic” location accuracy of such events by an order of mag- fault and earthquake-induced changes in crustal stress. that challenge conventional wisdom frequently occur period, occurring in the days to years immediately nitude (e.g., Lohman et al., 2002; Lohman and Simons, Elastic models of the lithosphere and geodetic data, outside the scope of existing ground-based geodetic following the earthquake, after which it merges con- 2005). Such improved locations can be used to improve combined with seismic data, reveal temporal evolution networks and thus underscore the need for the global tinuously into the interseismic phase. Regions where (i.e., kinematic models) of slip during an earthquake, access provided by NISAR. 172 NISAR Science Users’ Handbook 173 NISAR Science Users’ Handbook

[ ] 30° Figure 17-2 Eurasia environments. Such events have been documented in ing regional earthquake events. Thus, NISAR will allow 28° Mexico, Japan, Chile, the Pacific Northwest, Alaska, a systematic assessment of the relationship between Surface displacements and 27°30’ 26° India 24° Arabia New Zealand and Southern California. These events seismicity and fault slip across the different phases of co-seismic slip (Top) surface 22° 60° 65° 70° displacements for the 2013 Mw 7.7 are sometimes quasi-periodic, they are frequently the seismic cycle. Balochistan (Pakistan) earthquake associated with increased seismic tremor, and remain 27°00’ derived from cross-correlation of enigmatic as to their origin. Of great interest is the ex- Viewed from above, the Earth’s outer rock layers are di- Landsat 8 images (black arrows tent to which such aseismic fault slip transients occur vided into multiple tectonic plates. The slow movement indicate the displacement direction at different time scales (days, weeks, years) and the of each plate results in concentrated zones of defor- 26°30’ and amplitude). (Below) Derived degree to which large seismic earthquakes are more or mation in the Earth’s crust – zones that are frequently distribution of co-seismic slip less likely in periods of these aseismic transient events. found at the boundaries between the plates and are the on the subsurface fault. The surface Thus far, the existence of these events has been limited locus of large destructive seismic events on interacting rupture of the fault, the first order 26°00’ to regions with pre-existing ground-based networks systems of faults. The next leap in our understanding of subsurface geometry of the fault, and we have no knowledge of the occurrence, or lack earthquakes and our ability to minimize their asso- and the distribution of slip are all thereof, on most of the world’s major faults. The global ciated hazards requires us to (1) detect regions that derived using a combination of 25°30’ coverage, frequent repeats, and high correlation geo- are undergoing slow elastic loading of seismogenic available optical and radar geodetic 5 m imaging data. Figure modified from detic imaging provided by NISAR will enable a complete faults, (2) understand what controls the distribution of Kilometers 0 2 4 6 8 10 Coseismic Displacement (m) Jolivet et al., 2014b. 0 50 inventory of shallow aseismic fault slip and thereby subsurface fault slip during individual large events, (3) 64°00’ 64°30’ 65°00’ 65°30’ 66°00’ allow us to begin to understand the underlying causes quantify the Earth’s response to large earthquakes (es- of these events. sentially using these events as probes of the mechan- South East ical nature of faults and the surrounding crust), and Fault slip events result in a redistribution of stress in (4) understand the role played by major earthquakes the crust and thus may be important in triggering seis- on changing the likelihood of future seismic events in mic activity. Current research is elucidating the nature neighboring regions. North of earthquake-to-earthquake interactions, quantifying Slip (m) 0 0 5 10 15 the statistical likelihood of linkages, and elucidating Earthquakes are part of a cycle commonly divided into Ratio/Slip 5 0% 25% time-dependent processes (e.g., postseismic relaxation, periods associated with elastic stress accumulation, 10 50% state and rate of fault friction) that influence triggered release of elastic stress during an earthquake, and a 15 75% 100% 20 activity. For instance, Figure 17-3 compares the cumu- period associated with rapid readjustment of the fault Depth (km) 25 lative rate of aftershock production after the 2005 Mw system and surrounding crust following a large tremor. 30 8.8 Nias earthquake with the rate of post-seismic dis- In some faults, there are periods of very slow transient placement observed at one of a few sparse GNSS sites. fault slip events that are so slow they do not cause Note that seismicity represents only a few percent of significant ground shaking. Models currently used to POST-SEISMIC DEFORMATION determine the extent to which such behavior is ubiqui- the total slip required to explain the GNSS data. The understand the earthquake system explore the fric- tous for large earthquakes and if so, raises the question observed behavior suggests that the temporal behavior tional properties of faults (which fault segments creep Important constraints on fault behavior are also gleaned of what controls seismogenic behavior. Only with data of displacement and seismicity is nearly identical and aseismically versus which segments fail in a stick-slip from comparisons of the distribution of co-seismic fault from many additional events will we be able to address that post-seismic fault slip processes control the rate of fashion) across all phases of the seismic cycle. These slip with estimates of inter-seismic and post-seismic this fundamental question. earthquake production. models also incorporate both elastic and inelastic fault slip. For instance, in Figure 17-3, we see the behavior of the crust in which earthquake faults are im- spatially complimentary distribution of co-seismic and Beyond seismic and post-seismic fault slip, recent Existing observations of seismicity and fault slip also bedded. The aim of these models is to rigorously simu- post-seismic fault slip associated with the 2005 Mw results document an exciting range of aseismic fault suggest longer-range interactions that are not fully late observations over short time scales (e.g. ,a single 8.8 Nias earthquake, with little overlap between the two slip events (fault slip not associated with a preceding understood. Such interactions should have detectable earthquake or a short period of time before and after an phases of fault slip. Future geodetic study will need to large earthquake) in both strike slip and thrust faulting deformation signatures (Toda et al., 2011). Synoptic earthquake) in a way that is consistent with observa- space-based imaging offers a new and promising tions of longer time scale deformation as inferred from means to identify deformation causes and effects link- geology. There are many proposed models designed to 174 NISAR Science Users’ Handbook 175 NISAR Science Users’ Handbook

explain existing observations of deformation in different be used during or after an eruption to determine 100 km 4°N phases of the earthquake cycle but we lack sufficient the locations of lahars or landslides. 27 km observations to test these reliably. The dedicated 4. Map localized deformation associated with volca- observing schedule of NISAR will increase the number, Sumatra nic flows that can persist for decades to under- 5 km spatial coverage, temporal resolution, and accuracy of [ Figure 17-3] 3°N stand physical property of volcanic flows, guide Nias-Simeulue earthquake data. observations sufficiently to allow us to systematically ground-based geodetic benchmarks, and help (Top) Coseismic (2-m interval white test, reject and/or constrain the competing models of avoid misinterpretations caused by unrecognized contours) and postseismic slip earthquake forcing. 2°N deformation sources. (color), from the 2005 Mw 8.8 aftershock of the 2004 Mw 9.1 17.1.2 Volcano Hazards Sumatra earthquake. Arrows indicate Deformation data are the primary observables in un- 1°N Earthquakes Improving volcano hazard prediction requires deter- observed (black) and predicted (red) Before 03/28/05 derstanding the movement of magma within volcanoes. GNSS observations, stars show Earthquakes mining the location, size and composition of magma Although uplift from the ascent of magma into the shal- After 03/28/05 epicenters of 2004 (white) and 2005 reservoirs via geodetic, seismic, geochemical, and low crust has been observed prior to some eruptions, 0° (red) earthquakes. Dots denote Postseismic other observations. We must also identify the type of particularly on basaltic shield volcanoes, the spa- Obs. earthquakes before (pink) and after 0.3 m magmatic unrest associated with eruptions, char- tio-temporal character of such transient deformation is Model (green) the 2005 event. (Bottom left) acterize deformation prior to volcanic eruptions, and poorly known, especially at the locations of the largest Postseismic Slip (m) Observed and modeled post-seismic 1°S 0.0 0.5 1.0 predict the type and size of impending eruptive events. explosive eruptions. Little is known about deformation displacements at one continuous High-quality geodetic observations are necessary in on most of the world’s volcanoes because only a small GNSS site. Black solid lines are 94°E 95°E 96°E 97°E 98°E 99°E 100°E order to constrain timescales and mechanisms of these fraction of them are monitored. Even the incomplete estimated from a 1D spring-slider processes. surveys to date from previous satellites have dis- model in which afterslip obeys a covered many newly active volcanoes (Pritchard and velocity-strengthening friction law. Volcanic hazard science flows from the same crustal Simons, 2004; Fournier et al., 2010). Data shown for vertical (U), east (E), LHWA deformation data used to study the seismic cycle. and north (N) displacements. (Bottom 0 Deformation data allow us to: right) The relationship between Detection and modeling of deformation can provide 0 post-seismic ground displacement U U warning of impending eruptions, reducing loss of and the cumulative number of 1. Identify and monitor surface deformation at qui- life, and mitigating impact on property. Even remote –20 escent and active volcanoes: Only InSAR has the regional aftershocks. –20 volcanoes are important to monitor as large eruptions Figure from Hsu et al., 2006. capability for monitoring virtually all of the world’s can have a global impact through ash ejected into E potentially active volcanoes on land (approximate- the stratosphere that can affect air travel (e.g., the E –40 –40 Displacement (cm) ly 1400 volcanoes). billion-dollar impact of the 2010 Eyjafjallajokull, Iceland Displacement (cm) eruption) and climate (like the 1991 Mt. Pinatubo, Phil- N 2. Derive models of magma migration consistent –60 with surface deformation preceding, accompa- ippines eruption). In addition, InSAR data provide de- –60 N nying, and following eruptions to constrain the tailed spatial information not available from GNSS and other available geodetic data, allowing us to explore 0 100 200 300 0 400 800 nature of deformation sources (e.g., subsurface magma accumulation, hydrothermal-system models to reveal complex geometries of intrusions and Days After Cumulative Number their interactions with regional crustal stress regimes. Nias-Simeulue Earthquake of Aftershocks depressurization resulting from cooling or volatile escape). Furthermore, higher temporal resolution deformation imagery combined with other geophysical and geo- 3. Monitor and characterize volcanic processes such chemical observations will make it possible to advance as lava-dome growth and map the extent of erup- volcano forecasting from empirical pattern recognition tive products (lava and pyroclastic flows and ash to one based on deterministic physical–chemical mod- deposits) from SAR backscattering and coherence els of the underlying dynamics (Segall, 2013). imagery during an eruption, an important diagnostic of the eruption process. Similar methods can 176 NISAR Science Users’ Handbook 177 NISAR Science Users’ Handbook

An example of the potential of frequent InSAR obser- from past satellite missions has been characterized by [ Figure 17-4] vations to monitor the temporal evolution of a volcano comparatively poor coherence and temporal resolution, Volcano monitoring. (Top) Average through an eruption cycle is illustrated (Figure 17-4) by restricting the application of those data to simple kine- deformation of Mt. Okmok volcano in the work of Lu et al. (2010). Mt. Okmok in the Aleutian matic models of magma storage and transport—espe- the Aleutian volcanic arc are related to arc erupted during February–April 1997 and again cially location, geometry, and volume change. A better magma movement from 1997. Each during July–August 2008. The inter-eruption deforma- understanding of volcanic activity requires models that fringe (full color cycle) represents tion interferograms suggest that Okmok began to re-in- are based on the underlying physics of magma ascent 2.83-cm range change between the

flate soon after its 1997 eruption, but the inflation rate and eruption. As input, such models require a variety of ground and satellite along the satellite 2 km generally varied with time during 1997–2008. Modeling geochemical and geophysical data, including, critically, line-of-sight direction. Areas that N these interferograms suggests that a magma storage deformation measurements with improved spatial and lack interferometric coherence are zone centered about ~3.5 km beneath the center of the temporal resolution. NISAR will provide 2-D vector de- uncolored. (Bottom) Estimated volume of magma accumulation beneath Mt. 10-km-diameter caldera floor was responsible for the formation measurements at higher temporal resolution 0.06 ) Okmok as a function of time based on 3 observed deformation at Okmok. Multi-temporal InSAR and better coherence than any past or present satellite multi-temporal InSar (Lu et al., 2010) 0.05 1997 ERUPTION VOLUME deformation images can be used to track the accumu- InSAR sensor, making it possible to explore volcano error bars represent one-sigma uncer- 0.04 lation of magma beneath Okmok as a function of time: models with complex source geometries in hetero- tainties. The shaded zone represents the total volume of magma added to the shallow stor- geneous media. When combined with GNSS, seismic, 0.03 the source volume decrease associated 1997 ERUPTION 2008 ERUPTION age zone from the end of the 1997 eruption to a few gas emissions, and other measurements of volcanic 0.02 with the 1997 eruption, as inferred days before the 2008 eruption was 85–100% of the activity, NISAR will facilitate the development of more from a co-eruption interferogram. 0.01 amount that was extruded during the 1997 eruption. realistic models that estimate, for example, absolute 0 While the eruptive cycle from Okmok shows a pattern magma storage volume, reservoir overpressure, volatile Change (km Cumulative Volume of deformation that may be diagnostic of impending concentrations, and other parameters. These results 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 Time eruption, only a fraction of the potentially active volca- are critical for deterministic eruption foresting that can noes have frequent enough observations from available be updated as new data are acquired, which represents GNSS or existing SAR satellites to detect such patterns. a fundamental advance over empirical forecasting that

Furthermore, even from limited observations, it seems is based primarily on past experience—a common [ Figure 17-5] AGUNG VOLCANO MAY 4 —NOVEMBER 16, 2017 that other volcanoes show different and sometimes practice presently at most volcanoes worldwide (Segall, High-resolution Cosmo-Skymed more complex patterns of deformation before eruption 2013). interferogram (top right) and – in some cases, no deformation is observed before backscatter SAR images (bottom) eruptions (Pritchard and Simons, 2004). The obser- 17.1.3 Landslide Hazards of Agung volcano, Bali, Indonesia of vations from NISAR will allow us to make dense time Landslides threaten property and life in many parts of the 2017-2018 unrest and eruption. series observations at nearly all the world’s subaerial the world. Steep slopes, rock types and soil conditions The interferogram shows about volcanoes to better understand the relation between are key underlying causes of landslides, which are 15 cm of inflation within the summit deformation and eruption. crater that occurred prior to the typically (but not always) triggered by rainfall events, eruptions of November 2017. The earthquakes, or by thawing in arctic regions. Improved NOVEMBER 20, 2017 NOVEMBER 28, 2017 Among the most important parameters needed to backscatter image of November 28 knowledge of surface composition and topography are assess short-term volcanic hazards and better under- clearly shows the accumulation of important for characterizing landslide risk. Prediction of stand volcanic processes are the location, volume, and new lava within the crater. Imagery landslide movement is aided significantly by spatially composition of potentially eruptible magma (Figure from the Italian Space Agency (ASI) and temporally detailed observations of down-slope 17-5). Together with seismology, continuous ground through the CEOS volcano pilot. motion at the millimeter to centimeter level. Such deformation measurements (like GNSS), and gas geo- observations, possible with InSAR measurements such chemistry observations, the spatially dense, InSAR-de- as NISAR, can identify unstable areas. Similar to the Mt rived deformation field can play a pivotal role in con- Okmok volcano, studies in areas that can be monitored straining these unknowns (Pritchard and Simons, 2002; with current InSAR capable satellites have shown the Dzurisin, 2007; Lu and Dzurisin, 2014). InSAR data 178 NISAR Science Users’ Handbook 179 NISAR Science Users’ Handbook

potential for observations at critical times. One example activity associated with production at Groningen has 2.7 United is in the Berkeley Hills region in Northern California, caused great public concern and in response, the Dutch States 2.4 where interferometric analysis reveals the timing, government decided to cut the production cap for this spatial distribution, and downslope motion on several reservoir in half in January 2014 (van Daalen, Wall Lake City Colorado 2.1 Slumgullion landslides that had damaged homes and infrastructure Street Journal, 01/17/2014). The financial cost to the Lake Fork Landslide 1.8 Henson Gunnison River (Hilley et al., 2004). A more active example, shown Dutch government in 2014 is 700 million euros ($1.14 Creek Boundary of the active Slumgullion landslide 1.5 in Figure 17-6, is the Slumgullion landslide in south- billion). Observations such as those to be provided by Boundary of the inactive landslide

western Colorado, which is moving at 1-3 cm day, as NISAR will provide a comprehensive geodetic dataset deposit 1.2 cm/day determined using L-band UAVSAR observations. that will inform such billion-dollar decisions. 0.9

17.1. 4 Induced Seismicity In addition to earthquake activity associated with the N 0.6 IN ADDITION TO production of hydrocarbons, there is now evidence Landslide monitoring site Management of subsurface fluid reservoirs is an 0 1 2 km 0.3 that production of water from aquifers can trigger EARTHQUAKE ACTIVITY Lake San Cristobal economically and environmentally important task. 0.0 earthquakes. On May 11, 2012, an Mw 5.1 earthquake ASSOCIATED WITH THE Obtaining observations to better manage subsurface struck the town of Lorca, Spain, resulting in 9 fatalities. [ Figure 17-6] Creeping Slumgullion landslide in southwestern Colorado (Left, at the Slumgullion reservoirs can have substantial benefits. In addition, Despite its relatively small magnitude, the quake was Natural Laboratory). Figure on the right shows a velocity map derived from UAVSAR PRODUCTION OF the past decade has seen a substantial increase in the shallow enough that InSAR observations of surface repeat pass observations separated by 7 days in April 2012. number of earthquakes triggered by both injection and HYDROCARBONS, THERE deformation allowed inversion for the distribution of slip production of subsurface fluids (Figure 17-7), leading at depth (Gonzalez et al., 2012). Most slip occurred at IS NOW EVIDENCE THAT to a review of the situation by the National Research a depth of 2–4 km, with a second slip patch shallow- PRODUCTION OF WATER Council (2013). InSAR provides an important tool for er than 1 km depth – both very shallow hypocentral [ ] 1200 understanding and managing the risks. Figure 17-7 depths for this region. Over 250 m of water had been FROM AQUIFERS CAN Cumulative number of earthquakes pumped from a shallow aquifer, with subsidence of An early investigation into understanding the geo- with a magnitude of 3.0 or larger in TRIGGER EARTHQUAKES. up to 160 mm/yr observed by InSAR (Figure 17-8). mechanical response to hydrocarbon production and the central and eastern United States, 1000 Gonzalez et al. (2012) hypothesize that stress changes 1970–2016. The long-term rate of induced seismicity at a hydrocarbon field in Oman Central U.S. from depletion of the aquifer triggered this unusually approximately 29 earthquakes per (Bourne et al., 2006) utilized InSAR. An oil field is over- Earthquakes shallow event. year increased sharply starting around 1973–2016 lain by a gas reservoir, both producing from carbonate 800 2009. The increase has been attribut- layers. InSAR, and microseismic data were acquired InSAR measurements of surface deformation can also ed to induced seismicity. to monitor the reservoirs’ responses to changes in provide a powerful tool for short-term risk assessment fluid pressure. The changes in stress associated with 600 associated with production of unconventional reser- Source - https://en.wikipedia.org/wiki/ differential compaction resulted in fault reactivation. As voirs. For example, a recent major bitumen leak from Induced_seismicity#/media/File:Cumula- hypothesized for tectonic earthquakes, there is a strong tive_induced_seismicity.png cyclic steam injection in Alberta, Canada, in June 2013 relationship between stressing rates and seismicity, 400 was associated with substantial precursory surface NUMBER OF M3+ EARTHQUAKES with the rate of seismic activity proportional to both the This image is in the public domain in the deformation would have placed valuable constraints rate of pressure change and the rate of surface defor- United States because it only contains ma- on the physics of this unusual sequence, unfortunately, 855 M ≥ 3 EARTHQUAKES 1973–2008 mation. Based on these observations, geomechanical terials that originally came from the United there appears to be no existing InSAR coverage. States Geological Survey, an agency of the 200 models can be built to enable accurate prediction of the 2897 M ≥ 3 EARTHQUAKES 2009–2016 United States Department of the Interior. risk for well-bore failure due to fault reactivation. Finally, injection of CO2 into the crust is expected to be- For more information, see the official USGS Understanding the relationship between production of copyright policy. come an increasingly important means for sequestering 0 hydrocarbons and induced seismicity is a problem of this greenhouse gas from the atmosphere. Monitoring 1975 1980 1985 1990 1995 2000 2005 2010 2015 tremendous economic importance. For example, the the surface deformation caused by fluid injection will vast gas reservoir in Groningen province of the Nether- likely become an important technique for understand- lands provides almost 60 percent of the gas production in the Netherlands. A recent increase in earthquake 180 NISAR Science Users’ Handbook 181 NISAR Science Users’ Handbook

ing reservoir behavior and monitoring its integrity. The 17.1.5 Aquifer Systems SUBSIDENCE (cm/yr) [ Figure 17-8] In Salah field in Algeria is in a favorable environment Natural and human-induced land-surface subsidence 0 5 10 15 for monitoring by InSAR. ENVISAT C-band InSAR studies Relationship between across the United States has affected more than of deformation associated with CO2 injection show subsidence associated 44,000 square kilometers in 45 states and is estimated that the field response can indeed be monitored in this with groundwater with- to cost $168 million annually in flooding and structural way (Ringrose et al., 2009). In particular, as shown drawal (red shading) and LORCA damage, with the actual cost significantly higher due in Figure 17-9. The surface deformation observed by the May 11, 2012, Mw to unquantifiable ’hidden costs’ (National Research InSAR shows a two-lobed pattern near well KB-502, a 5.1 earthquake near the Council, 1991). More than 80 percent of the identified horizontal well injecting CO into a 20-m thick saline city of Lorca, Spain (after 2 subsidence in the United States is a consequence of aquifer at 1.8 km depth. Such a two-lobed pattern Gonzalez et al., 2012). the exploitation of underground water. The increasing indicates that, in addition to a component of isotropic development of land and water resources threaten to volume expansion, a vertical fracture has opened, ap- exacerbate existing land subsidence problems and parently extending into the caprock above the aquifer initiate new ones (Figure 17-10) (Galloway et al., 1999). (Vasco et al., 2010). This fracture explains the early Temporal and spatial changes in the surface elevation breakthrough of CO into observing well KB-5 along 5 km 2 above aquifers measured with geodetic techniques strike to the northwest. In response to the confirmation provide important insights about the hydrodynamic of the fracturing of the caprock, the injection of CO2 at this site has been suspended.

b. NEVADA IDAHO COLORADO NEW JERSEY [ Figure 17-9] Las Vegas Raft River Denver 1.72 Atlantic City-Oceanside Area Valley Area Area In Salah oil field deforma- 1.76 Barnegat Bay-New York Bay Coastal Area KB-5 1.80 tion. Interferogram of the KB-503 CALIFORNIA In Salah oil field in Algeria 1.84 Antelope Valley Depth (km)

N29°10’ KB-502 San Benito Valley ,showing deformation 1.88 Coachella Valley –6 –4 –2 0 2 4 6 Elsinore Valley San Bernardino Area associated with CO2 Distance to the Northwest La Verne Area San Gabriel Valley injection over the period Lucerne Valley San Jacinto Basin Mojave River Basin San Joaquin Valley DELAWARE a. –0 Aperture Change (cm) 8 from March 2003 to San Luis Obispo Area Oxnard Plain Bowers Area 10 Santa Clara Valley December 2007, (Onuma Vertical Pomona Basin Dover Area Displacement Sacramento Valley Temecula Valley & Ohkawa, 2009). 9 Rate Salinas Valley Wolf Valley 8 8 VIRGINIA 7 7 SOUTH CENTRAL ARIZONA Franklin-Suffolk Area 6 6 Avra Valley Williamsburg- 5 KB-501 Major alluvial KB-502 5 East Salt River Valley West Point Area

N29°5 ' aquifer systems 4 4 Eloy Basin Distance North (km) in the conterminous LOUISIANA 3 Gila Bend Area NEW MEXICO N29°5 ' 3 United States Harquahala Plain Baton Rouge Area 2 2 Albuquerque Basin San Simon Valley New Orleans Area GEORGIA 1 Mimbres Basin 1 Stanfield Basin Savannah Area 1 2 3 4 5 6 7 8 9 10 Tucson Basin TEXAS 0 Distance East (km) West Salt River Valley Houston-Galveston –1 Wilcox Basin Hueco Bolson-El Paso, Juarez 0 Fractional Volume Change (%) 0.6 –2 KB-CE –3 [ Figure 17-10] U.S. Subsidence areas. Areas where subsidence has been attributed to groundwater pumping (Galloway, KB-1 –4 D.L., Jones, D.R., and Ingebritsen, S.E., Galloway et al, 1999, Land Subsidence in the United States: U.S. –5 2 0 2 4 Kilometers –6 Geological Survey Circular 1182, 175 p.) mm/yr E2°10' E2°15' 182 NISAR Science Users’ Handbook 183 NISAR Science Users’ Handbook

properties of the underground reservoirs, the hy- thickness and the vertical hydraulic diffusivity of the Hoffman et al., 2001), Los Angeles (Bawden, et al., of land subsidence have been documented throughout dro-geologic structure of the aquifer, the potential fine-grained strata (aquitards) within an aquifer, the flu- 2001), Las Vegas (Amelung et al; 1999), and Phoenix the world during the past few years. Globally, InSAR infrastructure hazards associated with pumping, id-pressure change within those layers will lag behind (Casu et al., 2005). PSInSAR and related processing has measured and tracked subsidence in areas across and effective ways to manage limited groundwater the pressure/hydraulic head change from pumping. approaches allow InSAR to measure subsidence in Europe, the Middle East, China, Japan, and Thailand. resources. The pressure gradient between the pumped (usually agriculture and heavily vegetated regions such as New The extent of the InSAR imagery allows hydrologists to coarse-grained) strata and the center of the fine- Orleans (Dixon et al., 2006) and the California Central model spatially varied skeletal storage aquifer param- Groundwater extraction from aquifers with unconsol- grained strata takes time to re-equalize. In practice, Valley (Sneed et al., 2013). More than 200 occurrences eters as they change seasonally and annually. Before idated fine-grained sediments like clay can lead to land subsidence can continue for decades or centuries irreversible land subsidence (Poland, 1984). A sedimen- after cessation of groundwater pumping, whatever Subsidence measured along Hyw 152, 1972–2004 (32 years) tary aquifer system consists of a granular skeleton with time is required for balance to be restored between the 0

interstices (pores) that can hold groundwater and the pore pressure within and outside the fine-grained units –0.2

pore fluid pressure is maintained by the groundwater (Sneed et al., 2013). The time constant of an aquitard –0.4 Subsidence data along the Delta-Mendota Canal that fills the intergranular spaces (Meinzer, 1928). is defined as the time following an instantaneous de- –0.6 0 –0.8 Under conditions of constant total stress, the pore fluid crease in stress that is required for 93% of the excess –0.6 1972–1988 –1.0 pressure decreases as groundwater is withdrawn, pore pressure to dissipate, i.e., the time at which 93% 1972–2004 –1.2 1935–1953 Subsidence (m) –1.2 –1.8 1935–1953 leading to an increase in the intergranular stress, or of the maximum compaction has occurred. The time 1935–1953 –1.4 1935–1953

Subsidence (m) –2.4 effective stress, on the granular skeleton. The principle constant is directly proportional to the inverse of the 1935–1953 –1.6 of effective stress (Terzaghi, 1925) relates changes vertical hydraulic diffusivity and, for a confined aquitard –3.0 –1.8 0 10 20 30 40 50 60 70 80 in pore fluid pressure and compression of the aquifer (draining both above and below), to the square of the Distance (km) system as half-thickness of the layer:

2 σ e = σ T − ρ ⎛ b′⎞ S′ s ⎝⎜ 2 ⎠⎟ where effective or intergranular stress (σe) is the dif- τ = K′ ference between total stress or geostatic load (σ_T) and v the pore-fluid pressure ρ( ). Changes in effective stress where S′s is the specific storage of the aquitard, ′b cause deformation of the aquifer granular skeleton. is the aquitard thickness, K′v is the vertical hydraulic Aquifer systems consisting primarily of fine-grained conductivity of the aquitard, and S′s / K′v is the inverse sediments, e.g., silt and clay, are significantly more of the vertical hydraulic diffusivity (Riley, 1969). Ireland ALOS Interferogram: compressible than those primarily composed of coarse- et al. (1984) estimated time constants of 5-1350 years 8 Jan 2008 – 13 Jan 2010 grained sediments, e.g., sand and gravel, and hence for aquifer systems at fifteen sites in the San Joaquin Check Station (with number)

experience negligible inelastic compaction (Ireland et Valley. The scale of groundwater pumping currently Continuous GPS Station Approximate InSAR detected subsidence al., 1984; Hanson, 1989; Sneed and Galloway, 2000). underway in many areas makes this a global issue maximum (actual maximum obscured) PS InSAR Contours 540 mm subsidence (minimum) Groundwater pumping can cause short- or long-term (Alley et al., 2002). 9 March 2006 – 22 May 2008 10 mm Contour Interval recoverable (elastic) or non-recoverable (inelastic) 0 5 10 Kilometers –12 cm 0 +12 cm N Relative compaction that reduces aquifer storage capacity Repeat-pass interferometric SAR has become an Subsidence Uplift (Sneed et al., 2013). invaluable tool for hydrologists to resolve spatially and temporal varying aquifer properties and model [ Figure 17-11] Subsidence measured with C-Band PSI and L-band InSAR. Ascending and descending ENVISAT PSI subsidence in the San Joaquin Valley, California measured between March 9, 2006 and May 22, 2008 shows a maximum subsidence rate Permanent (irrecoverable) compaction can occur if an parameters that are impractical to obtain with any of 30 mm/yr (white contours). ALOS interferogram (color image) shows a large subsidence feature in the region north aquifer is pumped below its pre-consolidation head other technology. Numerous studies have exploited of where the C-band ENVISAT data acquisition ended (red boundary) (January 2008-January 2010). Insets show sub- (lowest prior pressure) (Phillips et al., 2003), resulting InSAR imagery to assess land subsidence globally sidence computed from repeat leveling surveys along Highway 152 for 1972-2004 and along the Delta-Mendota Canal in collapse of the skeleton, decreased pore space, and (Figure 17-11). Early research in the United States for 1935-1996, subsidence computed from GNSS surveys at selected check stations for 1997-2001. Contours show permanent loss of storage capacity. Over-development focused on the deserts and major cities in the Western subsidence measured using PS InSAR during March 9, 2006-May 22, 2008 (Sneed et al., 2013). of aquifers can induce long-term elastic subsidence U.S. including the Mojave Desert (Galloway et al., 1998; that lasts for decades to centuries. Depending on the 184 NISAR Science Users’ Handbook 185 NISAR Science Users’ Handbook

the advent of InSAR, it was not possible to know the The improved temporal coherence achieved by L-band [ Figure 17-12]

boundary conditions of a pumped aquifer; subsidence imagery in agriculture and heavily vegetation regions Kolkata city land subsid- gradients are used to understand the margin locations (see, e.g. Figure 17-12) is one of the key motivations ence. Deformation maps and aquifer interactions. for India’s interest in a long-wavelength radar mission, showing average rate particularly coupled with more densely sampled data to of land subsidence in Therefore, InSAR’s ability to measure the spatial and reduce tropospheric noise and other effects. The Kolkata city obtained from temporal changes associated with aquifer system C-band subsidence map in and around the city of differential interferogram compaction/land subsidence provides a direct meth- Kolkata in Figure 17-12 shows coherence only in the of ERS-1 data. The rate odology for determining the hydrologic properties that urbanized areas (Chatterjee et al, 2006; 2007a; 2007b; of land subsidence was are unique to each aquifer system, thereby providing Gupta et al., 2007). With longer wavelength radar to found to be 6.55 mm/year (Chatterjee et al., 2006). fundamental geophysical constraints needed to un- improve coherence everywhere, however, subsidence derstand and model the extent, magnitude, and timing measurements can be extended to much broader areas of subsidence. Furthermore, water agencies can take in places like India. advantage of these geophysical and hydrodynamic DEFORMATION MAP DEFORMATION RATE parameters to optimize water production while mini- It can be difficult to resolve small-scale surface mizing subsidence and mitigating the permanent loss deformation associated with slip at depth on faults of aquifer storage. and the migration of magmatic fluids from the large

CIT1ground-surface deformations caused by anthropo- One of the greatest challenges for measuring land genic fluid withdrawal and injection. However, on their [ Figure 17-13] subsidence is the loss of interferometric correlation in own, the large signals provide information to better heavily vegetated regions and in areas with extensive understand managed groundwater, hydrocarbon, and Example of non-tectonic deforma- 5 km C 0 5 agricultural production. Persistent Scatterer Interferom- geothermal resources, aiding in the characterization tion. Unwrapped ENVISAT interferogram (January 2005 to July 2005) of etry (PSI) based DInSAR techniques (e.g. Ferretti et al., and modeling of reservoir dynamics. Pumping of un- –10 0 10 20 30 40 mm the San Gabriel Valley (CA) showing 2001; Hooper et al., 2004) have greatly expanded the confined aquifers can lead to elastic deformation (e.g., surface deformation over an area SUBSIDENCE UPLIFT efficacy of C-band SAR investigations in challenging seasonal uplift/subsidence) if fluid extraction/recharge CIT1 40 x 40 km associated with natural agricultural areas but are limited to the temporal sam- is well balanced, but when net fluid production is un- aquifer recharge during a record pling density of the SAR archive. Sneed at al. (2013) balanced, it can induce long-term surface deformation rainfall during the winter of 2005. LONG combined PSI C-band and differential L-band InSAR or permanent (inelastic) deformation. Surface change AZU1 The land surface uplifted 40 mm 10 mm to capture the full extent of the subsidence (Figure specific to fluid management is observed as horizontal pushing GNSS sites on the margins SGHS 17-11). The PSI approach, shown as contours in the and vertical deformation signals in GNSS and InSAR of the basin radially outward in CVHS figure, involved a long time-series of C-band images, time series data that correlate with the production excess of 10 mm (labeled vectors). and resolved a maximum subsidence rate of 30 mm/ activities. The timing, location and spatial extent of the This groundwater hydrology tran- WCHS yr. Only 2 ALOS L-band images spanning 2 years were InSAR signals is key to isolating individual processes sient was initially interpreted as an GVRS available, from which a subsidence rate of 54 cm in 2 particularly in situations where quasi-steady state fluid aseismic earthquake in an active WNRA tectonic environment; combined VYAS years was derived. pumping/injection mimics or masks tectonic/magmatic LPHS signal in GPS time series (Figure 17-13). InSAR imagery and GNSS time- RHCL series along with water levels were needed to resolve its genesis (King et al., 2007) 186 NISAR Science Users’ Handbook 187 NISAR Science Users’ Handbook

NISAR imagery can be used to help isolate, model, 17.2 ECOSYSTEMS priority to significantly reduce large remaining uncer- variability of atmospheric CO2 being strongly controlled and remove the effects of fluid extraction on tectonic/ tainties in global carbon cycle and climate prediction by the terrestrial biosphere, while the coupling between The 2007 Decadal Survey identified that a key goal for volcanic GNSS time-series data. Future GNSS networks and ecosystem models (CEOS 2014). Therefore, a the terrestrial biosphere and climate was identified ecosystems sciences is to characterize the effects of can be optimized to avoid anthropogenic and natural spaceborne mission meant to address the needs of as one of the major areas of uncertainty in predicting changing climate and land use on the terrestrial carbon surface deformation associated with the pumping of the link between ecosystems and the climate will have climate change over decadal to century time scales. cycle, atmospheric CO levels, and ecosystem services. fluids and natural groundwater recharge processes. 2 the following scientific objectives ranked amongst the Spatially, large uncertainties exist in the distribution of Human induced disturbances have dramatically altered GNSS sites placed on the margins of active aquifer/res- highest priority: carbon stocks and exchanges, in estimates of carbon the terrestrial ecosystems directly by widespread ervoir will have horizontal motion that can mask and at • Quantify and evaluate changes in the Earth’s emissions from forest disturbance and the uptake land use changes, converting old-growth and car- times mimic the tectonic signal (Bawden et al., 2001). carbon cycle and ecosystems and consequences through forest growth. bon-rich forests into permanent croplands and urban GNSS sites placed near the center of the subsidence for ecosystem sustainability and services. landscapes. Disturbances have also led to extensive will have high vertical signal with nominal horizontal A fundamental parameter characterizing the spatial losses of wetlands of up to 50 percent and increased • Determine effects of changes in climate and land displacements therefore improve the ability to resolve distribution of carbon in the biosphere is biomass, the probability of natural disturbances such as fire, use on the carbon cycle, agricultural systems and tectonic deformation in an active groundwater basin. which is the amount of living organic matter in a droughts, hurricanes, and storms due to fundamental biodiversity. given space, usually measured as mass or mass per shifts in the climate and atmospheric CO concen- • Investigate management opportunities for mini- 17.1.6 Glacial Isostatic Adjustment 2 unit area, with half of all dry biomass being carbon trations (Foley et al., 2005; Dale et al., 2001; IPCC, mizing disruption in the carbon cycle (ISRO). (Figure 17-14). Therefore, biomass represents a basic In areas of present or past glaciation, surface defor- 2007). In recent years, the critical ecosystem services • Determine the changes in carbon storage and accounting unit for terrestrial carbon stock, and its mation can be caused by a solid earth response to provided by mangroves have been particularly hard hit uptake resulting from disturbance and subsequent temporal changes from disturbance and recovery, play current glacier advance or retreat as well as a delayed by warming ocean temperatures, rising sea levels, and regrowth of woody vegetation. a major role in controlling the biosphere interaction response to changes centuries or millennia ago. The population pressures. Shifts in vegetation are occurring, • Determine the area and crop aboveground bio- with climate. Estimates of the amount of biomass in the magnitude and spatial patterns of ground deformation especially in high altitude regions where alpine tree mass of rapidly changing agricultural systems. world’s terrestrial ecosystems range from 385 to 650 can be used to infer changes in the ice load and the lines are advancing. PgC (Houghton et al., 2009). Forests contain more than rheology of the Earth’s crust and upper mantle. InSAR • Determine the extent of wetlands and characterize 80 percent of the above ground carbon stock and are has been used to measure the elastic response of ice While these changes have important implications for the dynamics of flooded areas. thus a dominant component of the global carbon cycle mass loss in Iceland in recent decades (Zhao et al., the global carbon cycle and its climate feedback, there • Characterize freeze/thaw state, surface deforma- (Houghton, 2005). Because of its importance for cli- 2014) and NISAR has the potential to make similar remains large uncertainty in the global extent and mag- tion, and permafrost degradation. mate, forest biomass is identified by the United Nations measurements around most of the current ice-covered nitude of these changes in the terrestrial component. • Explore the effects of ecosystem structure and its Framework Convention on Climate Change (UNFCCC) areas. Furthermore, frequent, L-band (high coherence) The Decadal Survey highlights this shortcoming by dynamics on biodiversity and habitat. as an essential climate variable needed to reduce the measurements with good orbital baseline control pro- stating that, “there are no adequate spatially resolved uncertainties in our knowledge of the climate system vided by NISAR will open new possibilities to measure estimates of the planet’s biomass and primary produc- (Global Climate Observing System GCOS, 2010). deformation caused by ice load changes since the Little tion, and it is not known how they are changing and 17 2.1 Biomass Ice Age and the Last Glacial Maximum (called Glacial interacting climate variability and change.” In May 2013, atmospheric CO2concentrations passed Our current knowledge of the distribution and amount Isostatic Adjustment, GIA) that can better constrain both the 400 ppm, indicating an alarming rise of more than of terrestrial biomass is based almost entirely on the ice load history and the viscosity beneath areas like Dynamics of global vegetation with strong impacts on 30% over the past 50 years, caused by fossil fuel ground measurements over an extremely small, and Canada, Alaska, Patagonia and Scandinavia. Deforma- global carbon cycle are identified as changes of woody emissions (~75%) and land use change (~25%). There possibly biased sample, with many regions still unmea- tion measurements in some of these areas have been biomass from deforestation, degradation, and regrowth, is strong evidence that during this period the terrestrial sured. A global, detailed map of aboveground woody made by GNSS, but NISAR will add important spatial changes in the extent and biomass production of global biosphere has acted as a net carbon sink, removing biomass density will halve the uncertainty of estimated resolution. For example, NISAR observations can test crops, and the extent and inundation cycling of global from the atmosphere approximately one third of CO2 carbon emissions from land use change (Houghton et some of the predictions of GIA made by GRACE satellite wetlands (NRC, 2007). Quantifying these changes is emitted from fossil fuel combustions (Canadell et al., al., 2009; Saatchi et al., 2013) and will increase our gravity observations, such as large uplift rates in north- critical for understanding, predicting, and ultimately 2007). However, the status, dynamics, and evolution understanding of the carbon cycle, including better in- ern Canada (e.g., Paulson et al., 2007) that should be managing the consequences of global climate change. of the terrestrial biosphere are the least understood formation on the magnitude, location, and mechanisms detectable with NISAR. It is the consensus of the scientific community that and most uncertain element of the carbon cycle (IPCC, responsible for terrestrial sources and sinks of carbon. systematic observations from space with the aim of 2007). This uncertainty spans a wide range of temporal Biomass density varies spatially and temporally. Living monitoring ecosystem structure and dynamics are a and spatial scales. The IPCC has identified interannual biomass ranges over two to three orders of magnitude, 188 NISAR Science Users’ Handbook 189 NISAR Science Users’ Handbook

LAND/ATMOSPHERE FLUXES [ Figure 17-15] GLOBAL DISTRIBUTION OF WOODY BIOMASS Global distribution of woody biomass. Forest biomass density predicted from a combination of inventory data and available statistics

Combustion Photosynthesis Autotropic Heterotropic (Kinderman, et al., 2008). 2 Gt C/yr 122 + 3 Gt C/yr Respiration Respiration 60 Gt C/yr 60 Gt C/yr Percentages refer to the percent of area DISTURBANCE Sediment REGIMES ECOSYSTEM DYNAMICS ECOSYSTEM DYNAMICS Transport for that class of biomass Oceans in each grid cell.

LU/LUC Insect/Disease Fire Human Weather Events Consumption Soil Warming REGIONS WITH REGIONS WITH REGIONS WITH REGIONS WITH OPEN Live Biomass Dead wood EXPORT FLUXES AGB < 100 Mg/ha AGB > 100 Mg/ha AGB < 20 Mg/ha NO WOODY WATER (400 Gt C) (100 Gt C) 50% OF AREA 50% OF AREA 50% OF AREA VEGETATION

Organic Soil Carbon (680 Gt C) These forests are also considered a major carbon sink United States (1.5 PgC per year) and China (2.5 PgC per Mineral Soil from long periods of management (Woodall et al. 2010) year), the top two carbon-emitting nations (Peters et al., Carbon (2400 Gt C) and increasing length of growing season from climate 2012; Global Carbon Project, 2012).

CARBON POOLS change (Myneni et al., 2001). Other low biomass density regions are savanna woodlands and dry forests, The location of the land carbon sinks and sources [ Figure 17-14] Major elements of the terrestrial carbon cycle: (1) Disturbance regimes; (2) Land/Atmosphere distributed globally, in temperate and tropical regions. are unknown, as well as the reasons for their annual fluxes; (3) Ecosystem dynamics; (4) Terrestrial carbon pools, and; (5) Export fluxes. NISAR These regions cover more than 50 percent of the area swings in strength, on occasion as much as 100% makes key observations in each element. (CEOS 2014). of forest cover globally and are considered highly het- (Canadell et al. 2007). To what degree are these large erogeneous spatially, and dynamic temporally. shifts a result of climate variability, or disturbance? Even where estimates of mean forest biomass are 17.2.2 Biomass Disturbance and Recovery known with confidence, as in most developed coun- from less than 5 MgC/ha in treeless grasslands, crop- accumulate biomass indefinitely, however, because tries, the spatial distribution of biomass is not, and the lands, and deserts to more than 300 MgC/ha in some stand-replacing disturbances keep turning old forests Perhaps more important than biomass distribution to possibility that deforestation occurs in forests with bio- tropical forests and forests in the Pacific Northwest of into young ones. However, most forest stands are in the the global carbon cycle, is the quantification of biomass mass systematically different from the mean, suggests North America. process of recovering from natural or human-induced change and its associated carbon flux (Houghton et that this potential bias may also contribute to errors in disturbances and, thus, are accumulating carbon, albeit al., 2009). The magnitude of the uncertainty in the flux estimates (Houghton et al. 2001, Houghton, 2005). Biomass density also varies considerably within eco- generally at lower rates as they age. global carbon flux is particularly large in the tropics. To address the uncertainty in carbon fluxes and the system types. This variability results, in part, from lim- Recent calculations estimate a net positive flux from terrestrial carbon sinks and sources, a series of accu- itations of the environment (for example, soil nutrients Forests in temperate and boreal regions have low the tropics of between 0.84 and 2.15 PgC per year rate, annual global maps of disturbance and recovery or the seasonal distribution of precipitation and tem- biomass density (< 100 Mg/ha) but are extensive in (Harris et al., 2012; Baccini et al., 2012; Pan et al., will significantly improve estimates of emissions to the perature), and in part from disturbance and recovery. area and are subject to climate change and variability 2011; Le Quéré et al, 2017). In the context of global atmosphere and quantification of the large proportion The aboveground living biomass density of a recently causing widespread disturbance (e.g. fire, hurricanes, climate mitigation approaches (UNFCC 2006) and the of the residual terrestrial sink attributable to biomass burned forest may be nearly zero, but it increases droughts), and human land use change (Bonan, 2008). relevant calculations of national carbon emissions, the recovery from such disturbances. The spatial and as the forest recovers (Figure 17-15). Forests do not difference between these two estimates (1.3 PgC per temporal distribution of disturbance events, many of year) lies between the total carbon emissions of the 190 NISAR Science Users’ Handbook 191 NISAR Science Users’ Handbook

which occur at one hectare or below must be observed observations, weather data, ground information and ANNUAL DEFORESTATIONMAP AT 100-m RESOLUTION at fine spatial resolution. Examples of such events economic reporting. All of these are used to inform [ Figure 17-16] include clear cutting, selective logging, fire, hurricanes, government and commodities markets that direct the Annual deforestation map at floods, disease and insect infestation. Figure 17-16 allocation of resources and predict nutrient availability. 100-m resolution. The use of bio- shows a typical landscape mosaic of disturbance and Identified in the 2010 GEO Carbon Strategy is the mass spatial distribution instead recovery following disturbance in the Amazon basin monitoring and measurement of agriculture biomass of a regional average can impact and the emissions calculated from the use of forest and areal extent, which are important components of the assessment of the carbon flux biomass maps (Saatchi et al., 2007; Harris et al., 2012). the global carbon budget and in the understanding of from deforestation by a factor of By developing an annual disturbance and recovery map the effects of policy and climate on land management two (bottom right). The annual at the same spatial resolution of the biomass map, we and crop yields. In the two-decade period from 1990- deforestation over the Amazon can radically improve the estimates of emissions and 2010, large-scale clearing and conversion of forests basin (top) is occurring at small 2000 2001 2002 scales (1 ha) (PRODOS, 2007). removals (Houghton et al., 2009). to agriculture has resulted in an average flux of 1.3 AGLB Mg/ha CORRECTION TO ANNUAL FLUX ESTIMATE FROM USING BIOMASS DISTRIBUTION MAP to 1.6 GtC/y since the 1990’s (Pan et al., 2011). While Having a biomass distribution at 0–25 the same spatial resolution of the 25–50 0.6 Because of various environmental and climate vari- the gross distribution of growing regions worldwide BIOMASS DISTRIBUTION disturbance can provide accurate 50–75 0.5 ables, forest ecosystems are heterogeneous in their is generally well known (Figure 17-17), it is not at 75–100 AVERAGE BIOMASS estimates of gross emissions from 0.4 cover, structure, and biomass distribution because resolutions required for carbon assessment and for 100–150 deforestation and other distur- 150–200 0.3 of various environmental and climate variables. The generating reliable, accurate, timely and sustained crop 200–250 bances (Saatchi et al; 2011; PGC/YR 0.2 heterogeneity of ecosystems occurs at different scales monitoring information and yield forecasts. The role of 250–300 Harris et al., 2012). 300–350 and has been studied extensively in ecological theory agriculture in the GEO System of Systems (GEOSS) has 0.1 350–400 0.0 and landscape dynamics. These studies recommend given rise to the Joint Experiment for Crop Assessment >400 2000 2001 2002 2003 2004 2005 detection and classification of disturbance and recovery and Monitoring initiative (JECAM), created by GEO YEAR events at one-hectare spatial scales to reduce the Agricultural Monitoring community, which has iden- uncertainty of carbon fluxes (Hurtt et al., 2010). NISAR tified high resolution SAR and optical remote sensing will provide a means to reliably generate annual capabilities as the necessary sensor platforms for crop disturbance and recovery estimates at hectare-scale monitoring and agricultural risk management (GEOSS resolution for the duration of the mission and thus Tasks AG0703a, b). help reduce the uncertainties in carbon emissions and sequestration estimates. With biomass levels in agriculture crops typically less than 50 t/ha, SAR backscatter observations provide 17.2.3 Agricultural Monitoring Growing Season an observational approach for the estimation of crop Dates biomass (Figure 17-18). By making short-revisit Apr – Dec Since the beginning of the agricultural revolution and observations throughout the growing season, additional Apr – Nov followed by the industrial revolution, agriculture has Aug – May information is obtained that will help refine these bio- been a driver and early adopter of technology for the Feb – Oct mass estimates as well as provide timely and sustained efficient production of crops. As populations have Mar – Dec crop monitoring information that will inform yield Mar – Oct grown and moved into urban centers, governmental forecasts and help evaluate agriculture management May – Nov organization have had an interest in food security and May – Sep practices in response to weather and governmental in assessing their availability and impact on world Oct – May policy initiatives. markets. Sep – Jun Year Round Because crop yield and resource planning are depen- Crop assessment depends on multiple sources of data dent, in part, on soil moisture, an L-band SAR can play [ Figure 17-17] Global image of agriculture crop areas and growing seasons. Crop types used for the assessment are that are used for determining crop condition and area, an important role in the planning and projecting of based on basic grains and economically significant crops of world agriculture (e.g. rice, wheat, soybean, often relying on inputs from previous year’s production. agricultural output. The longer wavelengths of NISAR’s maize, etc.). Data source: earthstat.org. The various sources of inputs include satellite-based 192 NISAR Science Users’ Handbook 193 NISAR Science Users’ Handbook

[ Figure 17-18] 500 6 NN Emissions from wetlands contribute about around distinguish between wet and dry land areas (regardless Heading Calibration 450 Canopy Height ~100–230 Tg/yr (Matthews, 2000) of methane to the of vegetation cover) and the degree of soil saturation

Examples of agricultural ) Stem Density 5

3 400 LAI atmosphere and represent 20–45% of total emissions change. Seasonal change of 350 (Sahagian et al, 1997). In addition, attributes for which Harvesting 4

Flooding LAI m plant variables determined by 300 Transplanting (~ 500 Tg/yr). Thus, changes in wetland emissions can data can be collected through an inventory and later destructive sampling on the 250 3 2 significantly impact future methane levels. Methane used to determine when change has occurred, are 200 Drainage cm weekly time-cycle and inter- 2 2 150 3 increases have contributed about 0.7 W/m to global particularly valuable, as wetlands are subject both to polated on a daily time series 100 2 STEM DENSITY (m 1 radiative forcing since preindustrial times (0.5 W/m natural change and, increasingly, to destruction and CANOPY HEIGHT (cm) 50 (top). Time course change of ChinaChina directly, plus an additional roughly one half the forcing 0 0 degradation associated with human activities. microwave backscattering 140 160 180 200 220 240 260 280 300 00 55 1010 2020 kmkm from CO2. This makes methane emissions the second coefficients at all combi- DOY most important greenhouse gas forcing (Hansen et al., Many wetlands are subject to seasonal or periodic nations of polarization and 2000; Ramaswamy et al., 2001). Therefore, controlling flooding, i.e. inundation, and knowledge of the spatial incident angle for the L-band methane emissions could mitigate global warming as and temporal characteristics of flooding patterns is over capturing the crop yield much as controlling CO2 over the next century (Kheshgi crucial to understand wetland biochemical processes, and biomass change. 500 Calibration L55HH et al., 1999), and might be a more practical way to including methane production. Furthermore, river inun-

) N (left) (Inoue et al, 2002) 450 3 L55Cross L55VV reduce near-term climate forcing, owing to methane’s dation represents a dominant mechanism in the lateral to after China. The L-band 400 Flooding shorter lifetime and the collateral economic benefits transport of sediments to the ocean basins, and thus ALOS PALSAR composite of 350 Transplanting of increased methane capture (Hansen et al., 2000). is a critical factor controlling the export flux of carbon three dates over Zhejiang 300 Harvesting Province in southeast China Heading Boundary Projections of future emissions are typically based only from terrestrial ecosystems (Figure 17-19). 250 Drainage Orchard

(right) and the crop classifi- STEM DENSITY (m Rice on potential changes in anthropogenic emissions. It is 200 Water cation from multi-temporal Dryland Crop possible, however, that natural emissions could also 17.3 CRYOSPHERE 0 Urban 0 5 10 20 km PALSAR imagery separating 140 160 180 200 220 240 260 280 300 change substantially. The cryosphere represents the Earth’s ice and dry and wet crops from DOY snow-covered areas. In particular, NISAR science ob- forests and urban areas (right) Globally, wetlands are also a critical habitat of numer- jectives will primarily focus on the ice sheets, glaciers, (Zhang et al., 2014). ous plants and animal species and play a major role in sea ice, and permafrost. Although these are the primary maintaining the biodiversity of the planet. Furthermore, focus, the mission will ultimately enhance science and natural wetlands and managed rice paddies are a major application studies aimed at many other elements of L- and S-band systems compared to that of C- and provide a base set of observations that will be combined source of food and fiber. These regions cover 5.7 x l06 cryosphere such as snow cover and lake and river ice. X-band systems (e.g., Radarsat 2, Sentinel-1 and with yield measures, weather records, and other remote km2 and 1.3 x 106 km2 with an estimated net primary The Decadal Survey articulated several overarching TerraSAR-X) make it uniquely capable of assessing this sensing resources to create predictive models that can production of 4-9 x 1015 and 1.4 x 1015 g dry matter per cryosphere-related objectives. Of these, NISAR will con- component of plant growth. be used from one season to the next. yea, respectively. The RAMSAR convention on wetlands tribute to addressing the following scientific objectives: has emphasized the role of remote sensing technology 17.2.4 Wetlands and Inundation The timeliness of observation for agriculture appli- in obtaining inventory information and monitoring the • Characterize and understand the processes that cations is also an important component. Because determine ice sheet and glacier sensitivity to Global wetlands and their hydrologic dynamics are of status and activity of wetlands globally (Rosenqvist et agricultural applications are a fundamental part of climate change. major concern with respect to their impact on climate al., 2007). A key challenge facing wetland researchers NISAR’s observing strategy, and NISAR’s polarimetric • Incorporate ice sheet and glacier displacement change. Wetlands are characterized by waterlogged and managers is in the development of techniques for capability, observations collected by the sensor in a information into coupled ice-sheet/climate models soils and distinctive communities of plant and animal assessing and monitoring the condition of wetlands consistent configuration will be collected throughout to understand the contribution of ice sheets to sea species that have evolved and adapted to the constant (Sahagian and Melack, 1996; Darras et al., 1998). the growing season and hence provide a resource that level change. presence of water. Due to this high level of water Parameters that have been used for these purposes will be of immediate use to global agricultural moni- saturation as well as warming weather in low and include the composition, location, areal extent, water • Understand the interaction between sea ice and toring efforts (e.g. GEOGLAM). Furthermore, NISAR will mid-latitudes and accelerated freeze/thaw cycles in status and productivity of wetlands over time (see climate. reviews in Finlayson et al., and Spiers (1999)). For wet- high latitudes, wetlands are one of the most significant • Characterize the short-term interactions between land inventory, techniques are sought that can reliably natural sources of increased atmospheric methane. the changing polar atmosphere and changes in sea ice, snow extent, and surface melting. 194 NISAR Science Users’ Handbook 195 NISAR Science Users’ Handbook

• Characterize freeze/thaw state, surface deforma- rates (up to 100 m/yr) (Howat et al., 2008). Therefore, [ Figure 17-19] tion, and permafrost degradation. special care must be taken in how such observations

L-band HH sensitivity in are evaluated, particularly when extrapolating to the wetlands. Sensitivity of 17.3.1 Ice Sheets future, since short-term spikes could yield erroneous L-band radar backscat- long-term trends. In addition to indicating a trend of Spaceborne InSAR and altimetry observations have ter at HH polarization for sustained speedup, recent results are significant in mapping the area and already made major changes to our perception of how that they show flow speed and mass balance can the cycle of inundation ice sheets evolve over time (Alley et al., 2005; Bamber fluctuate rapidly and unpredictably (Moon et. al., 2012). of the wetlands of the et al., 2007; Joughin et al., 2011), overturning the con- While existing sensors have revealed major changes, Amazon basin. (Rosen- ventional wisdom that ice sheets respond sluggishly to these observations, cobbled together from a variety of qvist et al., 2003). climate change at centennial to millennial time scales sensors, are far from systematic. Prior to 2015, there JERS-1 DRY SEASON JERS-1 WET SEASON (e.g., Paterson, 1994). Numerous observations have were no systematic observations by existing or future shown that large Greenlandic and Antarctic glaciers sensors with which to characterize ice-sheet flow vari- and ice streams can vary their flow speed dramatically ability and with which to develop the required modeling over periods of seconds to years (e.g., Bindschadler et capability to accurately project sea level trends. While al., 2003; Joughin et al., 2004a; Rignot and Kanaga- such observations have begun with the launch of the ratnam, 2006; Rignot 2011c). It was this unanticipated Copernicus Sentinel 1A/B SARs and the USGS/NASA variability that prompted the Intergovernmental Panel Landsat 8 optical instrument, existing coverage does on Climate Change (IPCC, 2007) to conclude: not meet community needs in terms of both resolution, coverage, and accuracy. Therefore, to accurately Dynamical processes related to ice flow not included determine ice discharge variability, to gain a firm in current models but suggested by recent observa- understanding of the dynamics that drive mass balance, and to avoid aliasing these rapidly changing VARZEA DRY SEASON VARZEA WET SEASON tions could increase the vulnerability of the ice sheets variables, NISAR will acquire annual-to-sub-an- to warming, increasing future sea level rise. Under- 1000 nual observations of outlet-glacier and ice-stream standing of these processes is limited and there is no Jaú River O AF (JERS-1 ESTIMATES) variability. {f(h)} consensus on their magnitude. 800 Ice-sheet velocity and surface elevation are two of the ) 2 Thus, NISAR’s major ice sheet goals are to provide most important observables for studying ice dynamics. 600 data critically needed to remove this gap in our While observations from space are largely limited to the understanding of the fundamental processes that ice sheet’s surface, when used in conjunction with ice Carbinani control ice-sheet flow. This knowledge is required to flow models, such data can be inverted to determine 400 {interp.} reliably model ice-sheet response to climate change basal and englacial properties (Joughin et al., 2004b;

FLOODED AREA (km and to project the resulting contribution to sea level Larour et al., 2005; MacAyeal, 1993; Morlighem et al., change over the coming decades to centuries. 200 2013). In particular, ice-flow velocity and accurate ice Interfl. {interpol.} topography (ICESAT-II) can constrain model inversions Because of the highly variable dynamics of outlet gla- for basal shear stress. Observations of changes in 0 ciers and ice streams, recent observations provide only ice-sheet geometry and the associated response also isolated snapshots of ice-sheet velocity (Figure 17-20) provide important information. For example, inversions 12 3 6 9 12 3 6 9 12 3 6 9 1995 1996 1997 1998 (Howat et al., 2007; Joughin et al., 2004a; Rignot and such as that shown in Figure 17-21 provide the magni- Kanagaratnam, 2006; Rignot et al., 2011a). Space- tude of the shear stress, but not the form of the sliding borne altimeters designed for mapping large-scale law. Observations of the spatio-temporal response to ice changes (e.g., ICESAT-I/II), under-sample many of an event such as the loss of ice-shelf buttressing can the narrow fast-moving glaciers with large thinning be used to derive parameters such as the exponent of a 196 NISAR Science Users’ Handbook 197 NISAR Science Users’ Handbook

power-sliding law or may indicate another type of slid- the models through massive data assimilation, and to 0 m/yr 11,000 0 m/yr 11,000 ing law is needed (e.g., plastic) (Joughin et al., 2010b). reduce uncertainties of sea level rise projections. Sustained and frequent sampling of rapidly changing areas by NISAR will provide the velocity observations 17.3.2 Glaciers and Mountain Snow necessary for such studies in place of the scattershot Glaciers and snow-covered regions are important for observations current systems provide, with ICESAT-II many applications such as melt runoff, hydropower providing complementary elevation data. stations and long-term climatic change studies. Be- cause they are often cloud-covered, microwave remote Antarctica has several large floating ice shelves that sensing is particularly useful for studying these areas extend over the ocean from the grounded ice sheet. due to its all-weather capability and ability to image A 15 km B In contact with the ocean and at low elevation, these through darkness. Radar backscatter is influenced elements of the coupled ice-sheet/ice-shelf system by material properties like surface roughness and [ Figure 17-20] Ice flow velocity at Jakobshavn Isbrae. Ice flow velocity determined from speckle-tracking as color over SAR are the most at risk in a warming climate (Rignot et al., dielectric constant and can therefore offer considerable amplitude imagery showing the rapid speed up of Jakobshavn Isbræ from a) February 1992 (ERS-1) to b) 2013). While the loss of floating ice has no direct impact information in relatively featureless snow-covered October 2000 (RADARSAT). In addition to color, speed is contoured with thin black lines at 1000-m/yr inter- on sea level, ice shelves buttress the flow of inland ice terrain. The potential of Synthetic Aperture Radar vals and with thin white lines at 200, 400, 600, and 800 m/yr (Joughin et al., 2004a). Over the last decade, and a reduction of this buttressing as ice shelves have (SAR) imagery for monitoring of snow cover was glaciers in Greenland have sped up on average by more than 30% (Rignot and Kanagaratnam, 2006; Moon et thinned or disintegrated is believed to be responsible discussed as early as in 1980 (Goddison et al., 1981). al., 2012). NISAR will provide continuous observations of such speedup to provide a better understanding the for the majority current mass loss in Antarctica (Payne processes governing such change. The attenuation length of microwave radiation in cold et al., 2004; Rignot et al., 2004; Scambos et al., 2004; dry snow is large and this kind of snow is transparent Joughin et al., 2012). Critical to studying ice shelves is and therefore invisible to radar (Rott and Davis, 1993) accurate knowledge of velocity and thickness, which unless the snow pack is very deep or at radar frequen- determine the mass flux distribution, the horizontal cies above about 10 GHz (i.e., a factor of 5-10 higher divergence of which can be used to infer basal melt than NISAR). However, when the liquid water content rates (Jenkins and Doake, 1991; Joughin and Padman, of snow exceeds about 1percent, the attenuation 2003; Rignot and Jacobs, 2002). Unlike sparse spatial length is reduced to a few centimeters, and the radar and temporal sampling from other missions, for the backscatter is usually dominated by surface scattering first time NISAR will provide comprehensive ice-shelf (Ulaby et al., 1984). The question of whether such snow velocity data. Similarly, ICESAT-II will provide compre- can be discriminated from snow-free terrain depends hensive measurements of ice shelf elevation, which can on the geometric and electromagnetic characteristics be used to determine thickness by assuming hydro- of snow cover and snow-free terrain. Various studies static equilibrium. Together, these data will provide the have shown that wet snow cover can be generally required observations to derive time series of ice-shelf distinguished from snow-free terrain using SAR data melting around Antarctica and areas of Greenland where (Baghdadi et al., 1997). Many studies have demon- ice shelves still exist. These observations of ice flow strated that use of multi-polarization SAR and InSAR on floating ice will serve multiple purposes. First, along techniques have substantially improved snow cover with observations of velocity for the grounded ice-sheet mapping and detection of dry and wet snow. Periodic periphery at high temporal resolution, NISAR will provide mapping of snow cover is important to estimate the 0 kPa ≥100 100 km an advanced warning system for rapid shifts in ice flow runoff and understand the effect of climate change on and the resulting contributions to sea level rise. Second, mountain ecosystem. [ Figure 17-21] Basal sheer stress for Foundation Ice Stream. Basal shear stress estimate for Foundation Ice Stream these observations will provide critical constraints to ocean models at the ice-ocean boundary, which are (left) and corresponding MOA image (right). Flow speed is shown with 100-m/yr contours (black lines). Himalayan glaciers feed into three major river systems needed to evaluate the skill of these models, to improve of India, and glaciers in many other parts of the world 198 NISAR Science Users’ Handbook 199 NISAR Science Users’ Handbook

are an important source of fresh water. Thus, runoff et al., 2005). For example, glacier advance can threaten al., 2012). Hence, any attempt to close the sea-level information on sea ice thickness, motion and defor- from changes in snow cover and glacier volume plays infrastructure while glacier-controlled dams (usually budget of the past and coming decades/centuries mation for the Southern Ocean and if any changes an essential role in long-term water resource man- below the surface) can fail catastrophically causing needs to include the contribution from glaciers as well to these parameters are occurring. With the recent agement and hydropower planning activities. Glaciers glacial lake outburst floods (GLOFs). While NISAR ob- as ice sheets. Therefore, just as for ice sheets, frequent and expected continuing increases in global ocean have generally been in retreat during the last century, servations are unlikely to be frequent enough to provide observations of glaciers by NISAR are needed to under- temperatures, wind speed, and wave height, what will with a marked acceleration in global mass-losses in immediate warning for GLOFs, subglacial lakes can stand glacier contributions to sea level rise. be the response and rate of impact on the contrasting recent years (Kaser et al., 2006; Meier et al., 2007). be detected and monitored by InSAR to provide some polar sea ice regimes? Will the response influence the For instance, glaciers in Alaska (see e.g. Figure 17-22) indication of the hazards they create (e.g., Capps et al., 17.3.3 Sea Ice ice albedo-temperature feedback, for example, with are currently retreating at some of the highest rates 2010). enhanced changes in the ice thickness distribution and Within the global climate system, sea ice is a primary on Earth (Arendt et al., 2002; Hock 2005). Such rapid motion/deformation? Such changes in thickness and indicator of climate change, primarily due to the pow- changes in glacier extent and volume will modify the The global distribution of glaciers, including those in deformation are not well captured in climate models, erful ‘ice albedo’ feedback mechanism that enhances quantity and timing of stream flow, even in basins with Greenland and Antarctica but not connected to the hence extending the observational record with NISAR climate response at high latitudes. The Arctic Ocean’s only minimal glacier cover (Hock and Jansson, 2005). ice sheets, contribute significantly to global sea-level in both polar regions will lead to improvements in sea ice cover is rapidly evolving in response to climate In highly glaciated regions, at times the increases rise and are sensitive indicators of climate change. A understanding of atmosphere-ice-ocean interactions change. Over the satellite period of observations, in runoff can exceed the runoff changes from other consensus estimate (Gardner et al., 2013) indicated and fluxes as well as in the short-term forecasting of Arctic sea ice has thinned, shifted from predominately components of the water budget. Thus, in glacier- that glaciers contributed 259 ± 28 Gt/yr (0.71 ± 0.08 changes in the sea ice cover. perennial ice to seasonal ice, and reduced in extent at ized drainage basins, accurate simulation of glacier mm Sea Level Equivalent (SLE)/yr) during the period the end of summer by nearly 30 percent. The result- response to climate change cannot be achieved without October 2003 to October 2009, even though they Away from the margins of the sea ice cover, the ing increase in open water, subsequent reduction in high-resolution observations, such as those NISAR make up less than 1 percent of the Earth’s global ice response of the ice cover to large-scale gradients in surface albedo, and increased absorption of incoming will acquire, of glacier dynamics, wastage, and retreat volume (roughly 0.5–0.6 m sea level equivalent, SLE). atmospheric and oceanic forcing is concentrated along radiation appears to be enhancing the strong ice-al- (Hock 2005). Thus, glaciers contribute to present sea level rise at a narrow zones of failure (up to tens of kilometers in bedo feedback mechanism. The expanded open water rate similar to the combined rate from the Greenland width) resulting in openings, closings or shears. In win- extent has also led to an increase in ocean surface In addition to influencing water resources, snow and and Antarctic ice sheets (289 ± 49 Gt/yr, Shepherd et ter, openings dominate the local brine production and temperatures, marine productivity and shifts in the glaciers pose hazards to nearby populations (e.g., Kääb heat exchange between the underlying ocean and the marine ecosystem composition, and an increase in atmosphere. Convergence of the pack ice forces the ice wave height that further impacts the margins of the sea to raft or pile up into pressure ridges and to be forced [ Figure 17-22] ice cover. By contrast, sea ice in the Southern Ocean, down into keels, increasing the ice-ocean and ice largely composed of thinner seasonal ice, may in fact Glacier change over time. From atmosphere drag. A combination of openings and clos- Kienholz (2010). Aerial imagery be increasing in extent, albeit slightly, largely due to 81 ings is typical when irregular boundaries are sheared showing the terminus of Valdez regional advances in ice advection from recent changes relative to one another. These processes shape the Glacier. (left) AHAP 1978 false in wind forcing. unique character of the thickness distribution of the ice color orthophoto. Historical cover and have profound impacts on the strength of the termini positions are indicated Sea ice thickness, a primary indicator of climate ice and its deformation properties over a wide range by yellow lines. The green line change in the polar oceans, is a time-integrated result 80 of temporal and spatial scales. A key observation for indicates the terminus position of both thermodynamic and dynamic processes. understanding the basin-scale mechanical character measured in summer 2008 As sea-ice is a solid, large-scale atmospheric and using GPS. (right) True color Ae- of the sea ice cover is how the ice moves at different oceanic forcing’s concentrate stress along quasi-linear ro-Metric orthoimage of 2007. length scales. These observations are of importance in fractures in the ice with widths that are typically less The rock covered terminus of quantifying and modeling sea ice behavior in a chang- 79 than several hundred meters, the dynamic process that the glacier can be seen at the ing climate and in facilitating operational applications. leads to ice motion and deformation and a resulting approximate position of the change in ice volume, heat exchange, and thickness green line marking the 2008 Systematic mapping of the sea ice with spaceborne distribution. As the Arctic sea ice has thinned, there position on the left image. 44 E 45 46 44 E 45 46 SAR has proven to be the ideal method to measure has been a subsequent increase in ice motion and small-scale detailed sea ice motion at the scales deformation. In comparison, there is a paucity of required to quantify sea ice deformation, based on 200 NISAR Science Users’ Handbook 201 NISAR Science Users’ Handbook

the fine resolution, increased temporal sampling in the deformation-induced ice production in the seasonal [ Figure 17-23] Siberia Siberia polar regions due to converging orbits, and operations ice zone is greater than 1.5 times that of the perennial Sea ice motion and defor- independent of cloud cover and daylight. Mapping is ice zone. The younger seasonal ice is mechanically mation in the Arctic. Sea ice Divergence X Motion required at regular intervals (3-6 days) with sufficient weaker; this points to a negative feedback mecha- deformation (left column) 0.04 10.0 resolution (50-100 m) to be able to identify morphologi- nism where higher deformation and ice production and motion (right column)

of the Arctic Ocean ice Alaska Alaska cal features of the sea ice cover such as ridges and the is expected as the ice cover thins. Such important 0.00 0.0 edges of floes. In the late 1980s and most of the 1990s, information is not available in the Southern Ocean, cover at a length scale of the availability of small volumes of ice motion data from where only limited SAR-derived ice motion maps have ~10 km derived from SAR the European SAR satellites (ERS-1, 2) allowed examina- been generated of the Ross Sea. data. (Deformation units: –0.04 –10.0 day-1; motion units: km/ 97315–97321 97315–97321 tion of sea ice strain rates at 5-10 km length scales and Nov 11 – Nov 17 Nov 11 – Nov 17 day). (Kwok, 2010). demonstrated the utility of these measurements for sea In the coastal margins of the Arctic, InSAR observa- ice studies. The most significant results to date were tions are useful for observation of land fast ice, which obtained with the systematic mapping of the western is sea ice that remains attached and grounded to Siberia Siberia Arctic Ocean obtained by RADARSAT-1, where the col- the coastal land margin, i.e. ice that is held fast to laborative mission between the Canadian Space Agency the land. Within the moving pack ice, the ice cover Vorticity Y Motion and NASA enabled systematic data collections during is changing too rapidly to derive coherence with 0.06 10.0 the winter months for the nearly all of the mission’s InSAR. However, land fast ice by definition remains

lifetime, from 1996 through 2008 (Figure 17-23). Using unchanged for long periods of time. Thus, InSAR Alaska Alaska 0.00 0.0 both Eulerian and Lagrangian tracking, which enables observations are useful for the automated detection the continuous tracking of grid elements over time, ice of the extent of land fast ice and information on the trajectory and detailed deformation of a grid element mechanisms by which such ice attaches or detaches –0.06 –10.0 97315–97321 97315–9732 were observed. In addition, these data were used to from the coast (Figure 17-24). There is an increasing Nov 11 – Nov 17 Nov 11 – Nov 17 derive the age of newly formed ice and the loss of ice presence of human activities in the Arctic coastal area due to ridging. margins due to ice retreat, related to oil exploration, increasing ship traffic, and the heightened need for a Siberia Siberia The decade-long ice-motion dataset from RADAR- military presence previously not required. Along with SAT-1 has been used to quantify the various measures the increasing potential of a hazardous event such Shear Motion Mag of opening, closing and shear, and to estimate ice as an oil spill, the need to improve the environmental 0.08 15.0 production and thickness. The data reveal an extent of understanding and monitoring of the dynamic coastal

the activity, persistence, orientation, and length scale margins is clear. Alaska Alaska 0.04 7.5 of the fracture patterns that are quite remarkable. The abundance of these quasi-linear fractures is correlated The NISAR mission enables the unprecedented capa-

to motion gradients and material strength, and they are bility to derive ice motion and deformation for system- 0.00 0.0 97315–97321 97315–9732 organized into coherent patterns that persist for days. atic mapping of both polar regions, to the extent pre- Nov 11 – Nov 17 Nov 11 – Nov 17 Contrast in the deformation shows that there are distinct viously not possible with international SAR missions. differences in the deformation-induced ice production, While Radarsat-1 provided excellent motion products and the density of these features in the seasonal and for the western Arctic, the dynamic eastern portion perennial ice zones. The long-time series of SAR ice of the Arctic was not mapped nor essentially was the motion were also used to determine the flux of ice out of Southern Ocean, so no detailed observations of those the Arctic Ocean on an annual basis. These were com- regions have ever been obtained. The Southern Ocean bined with SAR-derived deformation and ice production provides a challenging mapping scenario compared to estimates as well as independently derived sea ice the Arctic, due to its relatively lower latitude range and thickness measurements, to estimate annual changes that the ice motion is not constrained by land as in the in sea ice volume. RADARSAT observations show that Arctic. Ice motion results in reduced but still essen- 202 NISAR Science Users’ Handbook 203 NISAR Science Users’ Handbook

[ Figure 17-24] Convergence environments, changing permafrost conditions are a long-term permafrost dynamics (subsidence, lateral The NISAR mission, through the mission science Towards Shore Rotational major hazard to industrial installations, transportation requirements, has placed an emphasis on Disaster/ Small-scale motion of Motion movements) in sufficient spatial and temporal resolu- land fast ice. Physical corridors, and human settlements. Thawing permafrost tion as well as accuracy, to substantially transform our Hazard Response because of the unique value of interpretation of small- also affects other environmental conditions, including understanding of pan arctic active layer dynamics and frequent and regular observations of nearly all land scale motion within hydrological cycles, biochemical processes, and habitat permafrost thaw. Hence, NISAR will allow us to begin to across the globe. This focus acknowledges the value otherwise stationary (Grosse et al., 2011; Hinzman et al., 2005; Smith et perform a spatially explicit assessment of regional of the mission to both the U.S. and India (Figure land fast ice based on al., 2005; Walter et al., 2006). Spatially distributed to global impacts of permafrost dynamics on hydrology, 17-25). In most cases, the driving requirements for simple physical models. permafrost models, driven by global climate model carbon cycling, and northern ecosystem character and response are met by the science and applications Hypotheses for motion Thermal Contraction data, project strong degradation of permafrost by the functioning. communities’ needs, with the exception of rapid pro- regimes are indicated end of the century due to the Arctic warming at a rate cessing for response, which is incorporated into the by white arrows and more rapid than the global average (ACIA, 2004). These 17.4 APPLICATIONS mission system design. associated text boxes. results are consistent with a trend over the last several The same data that are used by the science disciplines decades of persistent rise in permafrost temperatures km Every year, natural and technological hazards in the to improve understanding of physical and ecological 2.5 2.5 2.5 10 measured throughout a pan-Arctic network of bore- processes can be used by the applications community United States cost an estimated $1 billion per week in Radial Divergence Lateral Divergence holes (Romanovsky et al., 2010). to inform decision making, improve risk management, the form of lives lost and public and private proper- Rotation Surface Tilt assess resource status, and respond to and recover ties destroyed. In 2004 alone, more than 60 major Shear Permafrost changes manifest themselves mainly from disasters. The involvement of applications com- as land surface elevation changes, which are well disasters, including floods, hurricanes, earthquakes, munity in the development of NASA’s Decadal Survey tially unknown deformation rates and faster velocities, suited to measurement by NISAR. The main drivers of tornadoes, and wildfires, struck the United States. mission requirements greatly expands the societal ben- for a thinner ice cover than in the Arctic. NISAR will elevation changes in permafrost regions are seasonal Subcommittee on Disaster Reduction (http://www.sdr.gov). efit and functionality of the NISAR mission to include achieve systematic and detailed mapping of the South- freeze-thaw cycles, short-term and long-term natural cross disciplinary and applied science research; opens ern Ocean sea ice cover for the first time. Mapping of disturbances caused by climate change, or anthropo- Between 2004 and 2013, there have been an average new collaborative opportunities between scientists, en- the Arctic Ocean will enable continuing deformation genic disturbances. In addition to surface displacement, of 147 disaster declarations per year in the United gineers, and policy makers; and significantly augments observations of the rapidly changing and likely still NISAR data will help detect changes in soil moisture, States (http://www.fema.gov/disasters/grid/year). In science unrelated to the primary mission goals. NISAR thinning sea ice cover. hydrology, and vegetation on various time scales. 2011, both India and the United States had more than will contribute to the following activities relevant to Observing permafrost and its change using NISAR will 10 major natural disasters, with an economic cost applications, among others: NISAR will enable precise motion and deformation help tie together sparse field measurements to help >$5 billion in the U.S. alone (Figure 17-25). • Improve hazard resilience by providing the measurements of the sea ice cover of both polar provide a comprehensive view of a problem that has observational foundation guiding future tasking, Regions, at an unprecedented detail and scope. These been a major challenge. In particular, recent research Historically, satellite imagery has been utilized on modeling, and forecasting strategies will be used to improve models of the sea ice circula- indicates that L-band SAR provides valuable large- an ad-hoc basis for disaster response and hazard • Detect early transients associated with natural, tion and energy fluxes within the global climate system. scale information on changes in land surface conditions science. Then in 1999, the establishment of the Inter- anthropogenic, and environmental and extreme When combined with thickness observations, such that can be used to directly infer permafrost and active national Charter (http://www.disasterscharter.org), an hazards as are planned to be obtained from IceSat-2, critical layer dynamics and their response to environmental agreement that now includes 14 of the world’s space time series of sea ice thickness distribution and mass change. Previous research has verified that perma- • Characterize evolving disasters in support of agencies and satellite management organizations, balance parameterization can be utilized within coupled frost-related surface deformation on Alaska’s North response and recovery efforts and better under- significantly advanced hazard science by providing a climate models to improve the prediction of sea-ice Slope can be derived from time series of L-band InSAR standing of fundamental science global mechanism to collect and distribute satellite data (Kääb, 2008; Liu et al., 2010; Qu-lin et al., 2010). changes and its role in the Earth’s climate system, • Support ecosystems applications in forestry and imagery in support of emergency response efforts Polarimetric data can also provide a means to identify based on both polar oceans (Kwok, 2010). agriculture during significant disasters with the objective to mini- regions with permafrost in Tibet (Zhen & Huadong, mize the loss of life and property. As of January 2014, • Determine environmental factors influence the 17.3.4 Permafrost 2000). the International Charter has been activated 474 times coastal processes such as erosion/deposition and since its inception, providing satellite imagery for Many permafrost regions contain large amounts of coastal land use/land cover change Data from the NISAR instrument will provide simulta- global disasters regardless of geopolitical boundaries ground ice that cause significant surface disturbance neous observations of several key parameters for sea- • Contribute to India’s science, applications, and for a wide range of disasters, including earthquakes, upon thaw. As a result, in many Arctic and sub-Arctic sonal active layer (freeze-thaw, subsidence, heave) and disaster response 204 NISAR Science Users’ Handbook 205 NISAR Science Users’ Handbook

floods, volcanic unrest, cyclones, fires, landslides/debris to develop the situational awareness of the primary flows, and anthropogenic disasters (Figure 17-26). hazard and the ability to recognize and characterize secondary hazards associated with the primary event. The Subcommittee on Disaster Reduction (SDR) iden- Societal impact, requires rapid damage assessment for tified four key factors for successful hazard mitigation emergency rescue efforts, system integrity assess- and developed six ‘Grand Challenges’ to provide a ment of lifelines, infrastructure (i.e. pipelines, levees, framework for sustained federal investment in science dams, urban corridors, factories), and environmentally and technology related to disaster reduction; the SDR sensitive regions, as well as, long-term facilitation of vision has been incorporated into the science and ap- societal/environment recovery efforts. Finally, societal plications objectives of NISAR for 10 of the 15 hazards integration combines the hazard response and hazard (Figure 17-27) recognized by the SDR. Specifically, science with societal needs to improve hazard mitiga- NISAR hazard response applications objectives are tion efforts by enhancing hazard resilience science by part of the traceability matrix for the mission and cover providing the observational foundation guiding future distinct areas of the hazard cycle: hazard detection, tasking, modeling, and forecasting strategies. disaster characterization, societal impact, and societal integration. Hazard detection requires systematic col- The following sections present the range of NISAR lection of geodetic observations to detect, characterize applications, generally split between:

Show All Earthquakes Forest Fires Oil Spills Volcanic Eruptions and model potential hazards and disasters. Disaster • Ecosystem applications Cyclone Floods Landslides Others characterization requires rapid disaster assessment • Geologic and land hazard monitoring [ Figure 17-25] International charter activations. Map of 474 international charter activations since the formation of the Charter in 1999 to February 2014. The color of the circle indicates the activation disaster type. Regions with multiple activations are shown as gray circles, with the number of activations in white text, and the circle [ Figure 17-27] sized to represent the number of activations (http://www.disasterscharter.org/) Landslides and SAR utility for disaster Earthquakes Debris Flows response. NISAR will provide response products for ten of the fifteen disasters identified by the Subcommittee on Disaster Volcanoes Hurricanes Reduction. NISAR data will directly contribute to disasters labeled in blue and will partially support disasters Anthropogenic- Technological Tsunamis labeled in yellow. Disasters

DWH Rig Site

Floods Wildland Fires

1 – 2 >10 Centre for Research on the Epidemiology of Disasters 3 – 5 Coastal No Reported Disasters Source: EM-DAT International Disaster Database Tornados 6 – 10 Inundation

[ Figure 17-26] Natural disaster occurrence statistics for 2011. The US and India have among the highest rate of natural disasters. 206 NISAR Science Users’ Handbook 207 NISAR Science Users’ Handbook

• Critical infrastructure monitoring production on a global basis is the United Nations Food launched on board the space shuttle in 1994. Data In the US, the Forest and Rangeland Renewable

• Maritime and coastal ocean applications and Agriculture Organization (FAO). According to FAO’s from areas such as the Dnieper River region of Ukraine Resources Planning Act of 1974 directed the Secretary 2015 statistics, over eleven percent of the Earth’s land were collected at study sites distributed throughout the of Agriculture to make and keep current a comprehen- • Hydrology and underground reservoirs surface (1.5 billion hectares) is used for farming. With globe and have been used by NISAR mission planners sive inventory for a prospective Renewable Resources an increasing population, after taking into account and other space agencies worldwide to understand Assessment of the forest and rangelands of the US. The NISAR applications topics include disaster/hazard expected improvements in land use efficiency, the how radar data can be used to improve our knowledge These assessments were focused on analysis of response, which cuts across all of the above catego- amount of land dedicated to food production is expect- of the world around us. Modern day SAR, such as the present and anticipated uses, demand for, and supply ries. In many cases, the difference between science ed to grow 7 percent by 2030 to keep up with demand. Canadian Space Agency’s Radarsat and the European of the renewable resources, with consideration of and applications is one of information usage, with This increase is equivalent to an additional 90 million Commission’s Sentinel satellite series, have benefited the international resource situation and with a strong applications end users interested in regularly available hectares, roughly the size of Texas and Oklahoma from the SIR-C mission and are being actively used emphasis of pertinent supply, demand and price observations or operationalizing product generation for combined. With the world’s population critically depen- today. trends. With increasing threats to forest resources from situational awareness, resource management, decision dent on the timely production of food and fresh water droughts, fire, and fragmentations, tracking the forest support, and event response. resources, the need is greater now than ever before for 17.4.2 Ecosystems: Forest Resource health, biomass stock, and tangible products such as the application of technology to assure that population Management timber has become a part of national security both at 17.4.1 Ecosystems: Food Security needs are met. Among the technical tools that are used home and globally. Forest ecosystems provide timber, fuel, and bioprod- to address these issues are the satellites that pro- To feed a growing population of more than 8 billion, ucts and a variety of services by sequestering carbon vide synoptic views of the globe from space. Satellite Moving toward an inclusive monitoring system, which food production and supply occur on a global basis. In from the atmosphere into the forest, purifying water sensors are routinely used to guide decision-makers can augment and enhance the national inventory data, order to better guide policy and decision making, na- and air, and maintaining wildlife habitats. One of the and commercial interests alike in scheduling future requires a departure from the past remote sensing of tional and international organizations work to transpar- greatest challenges facing forest managers in the US plantings and monitoring the effects of policy changes only the forest cover. New active remote sensing tech- ently monitor trends and conditions of agriculture in a and elsewhere in the world is to maintain the health and a dynamic global marketplace. niques using both Lidar and radar have the capability timely basis. Because of the variable nature of planting and resilience of the forest ecosystems. This requires to measure both forest height and biomass. This high and harvesting practices, efforts such as this are a coordinated effort for systematic monitoring of forest The NISAR mission will provide dependable observa- spatial resolution data on forest structure and biomass manpower intensive and time-consuming tasks. Among cover, volume, and productivity, to develop techniques tions throughout the growing season. Radar imagery density can be readily integrated into existing forest the organizations that track the trends in agricultural and policies for improving the stock and sustainable will provide near weekly observations of almost all land inventory systems. The NISAR mission will observe use of woody biomass. Optical satellite observations as areas that complement the optical data and provide in- forests weekly and collect the information needed to in Landsat have played a major role in monitoring the dependent information that is sensitive to the changing map global forests and shrub lands multiple times per forest cover and changes globally. But, with the advent [ ] structure and moisture conditions of the crops being year. Data products will be made available at inter- Figure 17-28 of modern radar techniques as in NISAR, frequent imaged. NISAR’s data products will be available open vals that are commensurate with the need of forestry Two-frequency radar and uninterrupted observations of forest volume can access. organizations and industry in the U.S. and around the image of the Dnieper River become a reliable data and tool for forest managers to world. NISAR images will provide near global infor- growing region collected assess forest status. The structures of different crop and land cover types mation sensitive to aboveground forest structure and in 1994 by NASA’s Shuttle provide a rich variety of responses to the radar illumi- biomass. The measurements can help monitor forest Imaging Radar program. Forest managers and the agroforestry industry are in nation in terms of varying polarization and frequency disturbance and recovery from both natural and human In this false color image, need of accurate and timely data over large areas to signatures. Because of the rapid, time-varying nature of sources allowing managers to improve forest health developing wheat fields assess forest development and prescribe actions nec- crop rotation, growth, and harvest, frequently repeated and products. show up as bright magenta essary to achieve regeneration objectives. Increasing radar observations can be used to determine both and forests as the bright emphasis on ecosystem management, escalating sil- white patches that follow the type of crop and its stage of growth. Information With increased urbanization in proximity to forests, vicultural (e.g. reforestation) treatment costs, evolving the river’s border. like this is used to predict the health of the region’s along with a growing variety of vegetation (fuel types) computer-based decision support tools, and demands crops and the planned agricultural output. Figure from changes in the landscape and management strat- for greater accountability have produced significant 17-28 shows data collected by SIR-C, a NASA mission egies, there is a pressing need for accurate, cost-effi- demands for spatial data on forest structure and productivity at national and subnational scales globally. 208 NISAR Science Users’ Handbook 209 NISAR Science Users’ Handbook

cient, large scale maps of forest biomass, fuel loads, of the dangerous nature of fires and their sometimes [ Figure 17-29] YELLOWSTONE NATIONAL PARK disturbance and recovery. Emerging remote sensing remote locations, remote sensing is a widely used and L-band airborne radar technologies can provide exactly the kind of large accepted tool used by national and international orga- data collected over the scale maps needed to more accurately predict forest nizations to detect active fires, monitor impacts from Yellowstone National biomass, fuel loads, fire risk, and fire behavior. fire and assess fire danger. For example, the National Park in 2003 was used Interagency Coordination Center (NICC), the US Forest to develop maps of The technology is now there to use radar to estimate Service Wildland Fire Assessment System (WFAS), the forest volume and fire Volume m3/ha forest fire fuel load (e.g., branch and stem biomass). US Forest Service Remote Sensing Application Center fuel load to help with park 0–50 management and fire 50–100 And the team recognized that a much more efficient, and the National Aeronautics and Space Agency (NASA) are among the key organizations in the U.S. providing suppression for improving 100–150 accurate, and cost-effective approach to sensing forest information to land and fire managers of daily and the recreational 150–200 structure and fuels—and then mapping them—might resources and revenues. seasonal projected fire danger, active fire detections, 200–300 (Saatchi et al. 2007). lie at the heart of radar remote sensing technology. area burnt, and fire severity. Fire danger is determined 300–400 400–500 said biologist Don Despain, now retired from the U.S. by current moisture conditions, duration of those condi- 500–750 tions, and vegetation water content, while fire severity Geological Survey in Montana, who was instrumental in >750 generating the fire management plans for the Yellow- refers to the total environmental change caused by fire. stone National Park. Figure 17-29 shows the derived Managers customize this information based on local forest volume from airborne radar data (AIRSAR) data expert knowledge of the total fuels available to burn over the Yellowstone National Park. NISAR will provide to provide public service announcements and develop similar measurements from a spaceborne platform to management strategies to mitigate potential impacts. enable monitoring changes of forest volume and fuel SAR CHANGE DETECTION loads across the park weekly. NISAR observations can be used for detecting vegetation and soil water content for assessing fire danger, Probability 17.4.3 Ecosystems: Wildland Fires and biomass that is used to quantify total available fuel 0 1 to burn and emissions lost to the atmosphere. Biomass, Unplanned wild land fires impact tens of millions of estimated from polarimetry data, is useful as input in acres annually around the world and cost billions of fire management models for quantifying total available dollars per year to manage and control. Although fires fuel to burn and emissions lost to the atmosphere. are crucial to many ecosystems, uncontrolled wildfires can burn homes, damage infrastructure and natural The structures of different land cover types provide a resources, kill and injure emergency responders, rich variety of responses to radar illumination through FIRE SEVERITY (dNBR) firefighters and the public, impact local/regional econ- time-varying polarization signatures. Because of rapid High omies, and adversely affect the global environment changes in structure and composition after distur- Low (http://www.sdr.gov). Categorizing fire danger, detecting bance like wildfire, information like this can be used to fires, identifying area burned and quantifying the sever- determine area burnt, even when traditional methods ity of fires is critical for mitigating the impacts of fire. do not work well. NISAR polarimetry data can be used [ Figure 17-30] Fire burn scar of the 2015 Lake Fire in San Bernardino National Forest, California. The left image shows to estimate the fuel load of unburned regions that can probability of change derived from interferometric radar (InSAR) using 14 pre-event images and one NISAR will provide a dependable observing strategy be used in fire management models during a wildfire, post-fire image from an overpass on June 29, 2015, using the airborne Jet Propulsion Laboratory (JPL) that will collect high-resolution SAR images over to map burn area perimeters, and to assess burn se- Uninhabited Aerial Vehicle (UAV) SAR radar instrument. Right image is the differenced Normalized Burn 90 percent of the Earth’s land surfaces throughout the verity post facto (Rykhus and Lu, 2011). Maps of InSAR Ratio (dNBR) fire severity map obtained from Landsat. year. NISAR imagery can provide observations that coherence change can be used to detect changes in complement optical data and independent information the land surface associated with wild land fires, thereby that is sensitive to the changing structure and moisture mapping fire perimeter. conditions of terrestrial (land) ecosystems. Because 210 NISAR Science Users’ Handbook 211 NISAR Science Users’ Handbook

Time-series data following a major fire can be used to NISAR will provide bi-weekly observations that [ Figure 17-31] track the ecosystem response and recovery and char- complement optical data and provide independent Three-date (2007, 2008, 2009) acterize secondary hazards such as debris flows and information that is sensitive to the mapping of forest L-Band radar image (JAXA ALOS) landslides. NISAR alone cannot be used to track fires: disturbance, including below-canopy inundation from of timber production land in south- Since fires can travel 10s of kilometers per day, the natural and catastrophic flooding events. Observations ern Louisiana, one of the most in- imaging frequency (twice in 12 days) is not sufficient to of the Earth’s land surfaces from space using active tensive timber production areas of guide the hazard response community as the disaster microwave sensors allows for reliable and repeated the United States. Red and yellow unfolds, for which low-latency daily to sub-daily prod- measurements to be made even under dense cloud colors readily show various dates ucts are required. cover. When forests canopies are disturbed, such and intensities of forest management activities (clear-cut and that standing trees are partially or wholly felled or selective logging). Blue and purple Figure 17-30 shows the fire burn scar of the 2015 removed, or significant fractions of the upper canopy colors show areas and stage of Lake Fire in San Bernardino National Forest, California. are lost, e.g. in a forest fire, the changes are reflected re-growing forest plantations. The radar is able to identify the more severely burned in a rich variety of radar signals that can be measured. areas. Although it was possible in this particular fire to The time history of changes shows when, where, and use both Landsat and InSAR data, there are many areas by how much the woods were altered. Figure 17-31 (e.g., Alaska) where it is frequently too dark or overcast shows data collected by the JAXA ALOS L-band SAR Image Data © JAXA/METI 2007-2009. to produce maps from optical data on a regular basis. mission operating from 2006 to 2011. The image is a Image Processing: Earth Big Data, LLC. NISAR data can improve mapping capabilities across three-date color composite, where radar signatures many areas, times of year, and weather conditions. result in color combinations that are directly related to various types of forest disturbance and regrowth. 17.4.4 Ecosystems: Forest Disturbance densities. In the contiguous United States, 45 percent be useful for monitoring the changes in the geomor- The NISAR mission will provide data of similar quality, of the coast line is along the Atlantic or Gulf coast. The phological features and land use-land cover patterns. Optimal forest management requires knowledge of yet at greater observation frequency and with easy average erosion rate along the Gulf coast is nearly Time-series SAR data may also be used to monitor how forests change over time in response to natural data access by the U.S. timber industry, natural 2-meters a year and along the Atlantic is approaching shoreline changes. SAR data can be used to demarcate disturbances and management activities, including resource managers, natural disaster prevention and 1-meter a year. Coastal erosion is also significant in high tide lines (highest of high tides) along the coast invasive species; diseases; plant and animal pests; fire; response teams, researchers, and decision makers. Alaska where degradation of permafrost and reductions based on manifestation of the effect of seawater on changes in climate and serious weather events such as The data will be a critical complement to the U.S. glob- in coastal sea ice contribute to increased erosion rates coastal landforms and landward moisture content. tornadoes, hurricanes and ice storms; pollution of the al land observing system by providing routine, global, (e.g., Eicken et al., 2011). Extreme storms, sea-level air, water, and soil; real estate development of forest cloud-free forest monitoring capacity. rise, land subsidence, landslides, and flooding all 17.4.6 Geologic Hazards: Earthquakes lands; and timber harvest. With the world’s popula- accelerate costal erosion. Periodic observations of the tion critically dependent on sustainably managed and Earthquakes are amongst the deadliest natural hazards. 17.4.5 Ecosystems: Coastal Erosion coastline are necessary to characterize the dynamics utilized forest resources, the need is greater now than There have been 35 earthquakes since 1900 that have and Shoreline Change of coastal erosion and coastal accretion processes on ever before for the application of modern technology to killed more than 10,000 people, with seven since the coastal communities and infrastructure and begin to provide detailed and timely informational map data to Coastlines are continuously being reshaped by the year 2000 (Bally, 2012). The 2004 Mw 9.1 Indonesian develop models and coastal erosion/accretion scenari- the timber industry, resource managers and forest pol- interaction of strong wave action, storm surges, earthquake and tsunami, the 2010 Mw 7.0 Haiti earth- os for societal resiliency. icy makers. Satellite sensors provide synoptic views of flooding, currents, sea-level rise, river discharge, and quake, and the 2011 Mw9.0 Japan earthquake and the globe from space. This information is routinely used human activities with the local geology and mitigation tsunami combined killed more than 450,000 people. NISAR will collect systematic polarimetric SAR imagery to guide policy both decision-makers and commercial efforts designed to minimize the effects of shoreline The International Charter has been activated 70 times to directly measure positional changes to the global interests. Examples include planning forest manage- recession on coastal communities. Coastal erosion since 2000 to help the emergency response community coastline. The combined analysis of cross-polarized ment activities, supporting preparation of information in the US has increased over the past few decades, directly following a major earthquake by providing rapid SAR and like-polarized images will be used to uniquely for forest real estate transactions domestically and in and therefore represents a major coastline hazard to imagery to help develop the situational awareness demarcate coastlines. Changes in the coastline pattern foreign countries, and monitoring the effects of forest low-lying communities, infrastructure, and lifelines lo- necessary to respond to disaster. Furthermore, through on a half-yearly interval will address the coastal dy- policy changes, such as logging concessions or illegal cated near the coast, areas often with high population the globalization and interconnectedness of the world’s namics scenario. Like and cross-polarized images will logging activities. economy, earthquakes can have a negative worldwide 212 NISAR Science Users’ Handbook 213 NISAR Science Users’ Handbook

impact, e.g., the 2011 Mw 9.0 Tohoku earthquake in causes corrosion and harms vegetation and livestock. data from NISAR will facilitate the development of flows are a dangerous relative of landslides where Japan resulted in suspension of auto manufacturing Lava flows inundate property and destroy infrastruc- more realistic depictions of active volcanoes, which are slope material becomes saturated with water forming a in Detroit due to parts shortage (Wall Street Journal, ture. Volcanic mudflows have the potential to devastate critical for eruption foresting. slurry of rock and mud that moves rapidly down slope 2011) and elevated insurance rates globally. entire cities even far from the source volcano. Pyro- and along channels picking up trees, houses, and cars, clastic flows—high-speed avalanches of hot pumice, 17.4.8 Geologic Hazards: Sinkholes and thus, at times, blocking bridges and tributaries, causing The NISAR imagery collection requirements for pure ash, rock fragments, and gas—can move at speeds Mine Collapse flooding along its path. Landslide danger may continue research science and earthquake applications are in excess of 100 km/hr and destroy everything in their to be high even as emergency personnel are providing Sinkholes are formed either naturally in Karst regions effectively the same – collect SAR data on every path. rescue and recovery services. where carbonate rock is dissolved into groundwater possible orbit. The key difference is that the applica- or grow due to human activities such as oil extraction. tions community needs low-latency data to develop the Worldwide, it is estimated that less than 10% of active Both L and S band NISAR images have the potential Many sinkholes occur rapidly over a small spatial scale, situational awareness for the hazard response commu- volcanoes are monitored on an on-going basis, mean- to significantly advance research for landslide science so it is difficult to capture precursory deformation using nity. They seek to quickly understand the scope of the ing that about 90% of potential volcanic hazards do not and provide invaluable information to the broader remote sensing techniques. In some cases, howev- disaster and how to best allocate limited resources. Key have a dedicated observatory and are either monitored landslide science application communities. First, NIS- er, there may be slow deformation, before sinkholes questions include: What is the area affected? Where only occasionally, or not monitored at all (Bally, 2012). AR’s cloud penetrating imagery, coherency mapping, collapse catastrophically, indicating where a future have buildings been damaged? How many? Are there As for earthquakes, the NISAR imagery collection and rapid tasking capabilities will allow emergency collapse is possible (e.g., Castañeda et al., 2009; secondary hazards like landslides, damn collapse, fires, requirements for pure research science and volcanic responders to identify triggered landslides and assess Paine et al., 2012; Jones and Blom, 2014). In addition, etc.? Where are the safe places to evacuate people? applications are effectively the same – collect SAR data their societal impact. For example, the 2008 Mw 7.9 subsidence from mining activities and catastrophic on every possible orbit, but the applications community Great Wenchuan earthquake in China triggered more mine collapse can be measured by NISAR (e.g., Lu and What infrastructure and lifelines were damaged? Where also needs low latency data. than 60,000 landslides, blocking roads, impeding Wicks, 2010; Ismaya and Donovan, 2012). was the greatest shaking or liquefaction? NISAR imag- emergency response efforts, isolating and destroying ery will be used in a variety of ways (e.g., backscatter Many volcanic eruptions are preceded by surface villages, and damming rivers thereby creating addi- 17.4.9 Geologic Hazards: Landslides and or coherence change), and integrated infrastructure deformation induced by moving magma beneath the Debris Flows tional life threating conditions. The 1997-1998 El Niño and population density information and optical imagery ground. Measuring this deformation is key to un- rainstorms in the San Francisco produced thousands of where available to address these questions. derstanding the potential for future eruptions. Radar Landslides, debris flows, and other forms of ground landslides and caused over $150 million in direct public observations from NISAR and other satellite missions failure affect communities in every state of the United and private losses (http://www.sdr.gov). Secondly, given 17.4.7 Geologic Hazards: Volcanic Unrest can play a direct role in helping to monitor volcanoes States and result in the loss of life and cost billions of that two-thirds of the counties in the United States and assess associated hazards, both during periods dollars in property losses and environmental degrada- have moderate to high landslide risk, NISAR will be able Earth is home to about 1,500 volcanoes that have of unrest and during ensuing eruptions. Data from tion every year (http://www.sdr.gov). During a two-year to identify and track motion on landslides that pose a erupted in the past 10,000 years, and today volcanic NISAR and other radar missions allow us to identify and period between 2014 and 2016, 61 people were killed significant societal risk over wide areas. Time-series activity affects the lives and livelihoods of a rapidly monitor surface deformation at quiescent and active by landslides in the U.S., including 43 in the Oso land- analysis of these slides will detect transient changes in growing number of people around the globe. In the volcanoes through the use of radar interferometry slide in Washington State (Coe, 2016). Approximately deformation patterns that may represent an elevated United States, more than 50 volcanoes have erupted (InSAR). Only InSAR has the capability for monitoring two-thirds of the United States population lives in societal risk and provide early warning of imminent one or more times in the past 200 years. Volcanic deformation at virtually all of the world’s potentially counties where landslide susceptibility is moderate to catastrophic failure. Finally, time-series analysis of eruptions destroy cities and towns, eject ash clouds active volcanoes on land. InSAR observations allow high. Landslides are triggered by a number of mech- coherence images in recent burn areas can be used to that disrupt air travel, and impact regional agriculture. us to build models of subsurface magma movement anisms, including intense or long duration rainstorms, map the area of vegetation removal and then help iden- Explosive eruptions eject ballistic rock fragments preceding, accompanying, and following eruptions — earthquakes, volcanic activity, wild land fire, coastal tify and map subsequent debris flows and their spatial that can impact the surface up to several kilometers information that is critically important to understand the erosion, excavation for infrastructure, and the loss distribution with respect to lifelines, infrastructure and away from the vent. Smaller fragments are carried state of activity and anticipated hazards. Radar images of permafrost in arctic regions. Some landslides can residences. When combined with computer modeling, upward in eruption columns that sometimes reach that allow us to monitor and characterize volcanic remain active for years or even decades, and some of new debris flow hazard assessments can be made with the stratosphere, forming eruption clouds that pose a processes are also used to map the extent of eruptive these slowly moving landslides may transition to cat- the aim of improving societal resiliency. serious hazard to aircraft. Volcanic ash fall can collapse products, like lava and ash, during an eruption. When astrophic collapse. In areas of steep slopes, the debris buildings, and even minor amounts can impact electri- combined with other measurements of volcanic activity, cal systems and disrupt everyday life. Volcanic gases contribute to health problems, and also to acid rain that 214 NISAR Science Users’ Handbook 215 NISAR Science Users’ Handbook

17.4.10 Hazards: Anthropogenic When levees (or other water defense structures) [ Figure 17-32] Technological Disasters subside, there is a high risk of catastrophic flooding. Ground movement along one Such subsidence was observed by InSAR phase change Anthropogenic hazards, e.g., intentional attacks, indus- of the levees that prevents before the Hurricane Katrina floods in New Orleans flooding of an island in the trial explosion, dam failure, etc., are broadly distributed 38°2’10”N (e.g., Dixon et al., 2006). InSAR also detected motion Sacramento-San Juaquin across the globe and events that warrant monitoring of embankments before they failed catastrophically in Delta (Deverel 2016). Inset can occur with little to no a priori information on the Hungary, creating the worst environmental disaster in photo shows a view looking location and timing. The anthropogenic disasters can that country’s history (e.g., Grenerczy and Wegmüller, east towards the area of rapid also form as a result of a cascading natural disaster 2011). Both New Orleans and Hungary studies utilized deformation (red/orange color). (e.g. Fukushima, Japan), highlighting the importance high-resolution InSAR time-series analysis methods, The deformation signal is not of timely generation and delivery of disaster response obvious to the naked eye on which will benefit from the frequent and repeated 38°2’0”N data after the primary event. Anthropogenic-techno- high-resolution NISAR observations. It is important to the ground, but ground-based logical disasters that impact human populations often inspection revealed that cracks note that while these studies came after the disasters are related to critical infrastructure, such as bridges, had formed in the levee. took place, with the processing of higher level products InSAR Line-of-Sight dams, power plants, and industrial facilities, or involve Velocity mm/yr over hazard-prone areas, NISAR will allow the local, <–40 the release of material that can be distributed widely in state and federal agencies to switch from disaster –35 - –40 the environment by air or water. NISAR can be tasked in –30 - –35 response to disaster preparation and resilience. An –25 - –30 response to such events after an unforewarned disaster 38°1’50”N example of this is persistent monitoring of levees in –20 - –25 occurs or in advance if it is known to be likely to occur. –15 - –20 the Sacramento Delta, based on InSAR time series –10 - –15 methods, that revealed slow, steady, and damaging 17.4.11 Critical Infrastructure: Levees, –5 - –10 subsidence along the landslide slope of a levee 0 - –5 Dams, and Aqueducts N >0 (Figure 17-32). 0 0.05 0.1 0.2 Miles Water storage, conveyance, and defense structures are critical elements of a country’s infrastructure 17.4.12 Critical Infrastructure: 121°45’50”W 121°45’40”W 121°45’30”W that provide water and protection to businesses and Transportation communities. Levees and dams not only protect the Roads, bridges, railway tracks, and other transportation regional environment in the vicinity of the facility. For The U.S. arctic region is characterized by vast areas low-lying areas but also channel the water to commu- infrastructure or facilities require careful and continu- example, impending or actual water intrusion into with limited infrastructure. Furthermore, its land and nities and businesses where it is needed. Dams irrigate ous monitoring to maintain integrity. The regular time the facility could be identified during overbank flow ocean areas are increasingly affected by extreme at least 10 percent of U.S. cropland and help protect series of images from NISAR, analyzed to produce on nearby rivers, or changes in land use identified environmental conditions, threatening human live and more than 43% of the U.S. population from flooding, InSAR products, can be used to monitor the structures downwind from a facility. Slow creep landslides or damaging existing infrastructure installations. On land, while satisfying 60 percent of the electricity demand and the ground nearby for movement that could pres- fault slip could be identified that causes slow degra- the annual freeze-and-thaw cycles of thick soil layers in the Pacific Northwest (2017 NISAR Critical Infra- age damage or failure. dation of performance or stress on structures. lift surfaces several tens of centimeters every winter, structure Workshop). Monitoring of levees and dams in damaging roads and affecting the integrity of buildings the traditional manner through visual inspection and 17.4.13 Critical Infrastructure: Facility 17.4.14 Critical Infrastructure: Arctic and oil pipelines. An abundance of unstable slopes in situ instruments is time-consuming and personnel Situational Awareness Domain Awareness threatens some of the most sensitive transportation intensive, leading to infrequent monitoring of most corridors in Alaska, while regular earthquakes and vol- areas. NISAR will increase inspections as it can image In many cases critical infrastructure operators have a The high latitude regions of Earth are facing canic eruptions interfere with human life and endanger the entire U.S. several times a month regardless of good understanding of their facility through instruments increased challenges related to dynamic changes international air traffic. A recent increase in commercial cloud-cover. NISAR’s resolution of 6-to-12 m is a sig- deployed within its confines. NISAR can augment point of the arctic environment and modified land use activities on the opening U.S. arctic oceans have led to nificant improvement over the Sentinel 1a/b resolution measurements with extended spatial coverage, and patterns by polar communities and industry. SAR can rising risks of anthropogenic disasters such as oil spills of 20 m, particularly for monitoring mid-size structures NISAR can provide information about changes hap- provide important information to improve situational and ship wrecks that require regular large-scale remote like levees and aqueducts. pening outside the facility that could potentially impact awareness and crisis response capabilities related to sensing data to enable sufficient situational awareness. operation or safety. NISAR can augment their knowl- a range of these emerging issues including mari- NISAR will provide frequent, regular, and comprehen- edge by providing information in the neighborhood and time security, infrastructure health, natural disaster sive coverage of arctic land to identify and monitor resilience, and transportation. 216 NISAR Science Users’ Handbook 217 NISAR Science Users’ Handbook

surface deformation related to landslides, permafrost 17.4.16 Maritime: Sea Ice Monitoring 6. Providing sufficiently low latency is the primary L-band or S-band radar may be more attractive for change, and natural hazards such as active volcanoes challenge to operational usage of NISAR data for imaging of current boundaries, fronts, eddies and The U.S. National Ice Center (NIC) is a joint effort of the and earthquake zones through the use of InSAR. These sea-ice monitoring. internal waves, since the longer (compared to C-band) US Navy, NOAA, and the U.S. Coast Guard. The NIC is deformations are important for the assessment of wavelength is less sensitive to rapid variations in the also a part of the Northern American Ice Service (NAIS) hazards affecting infrastructure and people living in boundary layer wind speed and will therefore be more which is represents a joint with the Canadian Ice Ser- 17.4.17 Maritime: Coastal Ocean the Arctic. Only InSAR has the capability for monitoring Circulation Features modulated by varying surface currents. Dual-frequency vice. Their primary goal is to monitor sea-ice extent and deformation across the entire Arctic region to provide measurement capabilities will allow tailoring of obser- type, especially in the Arctic, for safety of navigation. NISAR will dominantly acquire data over land and the a synoptic picture of ongoing risks. Radar images have vations to different wind speed regimes. In addition to the important Cryospheric science goals cryosphere. To the extent that coastal regions are also the additional capability to detect changes in the north- Coastal upwelling processes and the formation of of NISAR, the routine imaging of the polar regions will imaged, NISAR data can be applied to a variety of ern coastlines, map flood extent and identify ice jams, coastal jets and fronts result in temperature gradients, potentially yield important benefits for operational mon- marine applications. Radar backscatter from the ocean monitor ship traffic (cooperative and non-cooperative) which may be detected by SAR due to reduced surface itoring. One of the primary observational instruments surface is directly dependent upon the roughness of the in Arctic waters, track the progression of oil spills and roughness over the colder water regions. Reduced sen- are available SAR data. Since SARs provide all-weather, ocean surface on scale of the radar. The rougher the sur- identify sea ice features that may threaten infrastruc- sible and latent fluxes over the colder water because day-night high-resolution, they represent the preferred face the larger the return from the surface. SAR marine ture installations and ship traffic. In concert with other of lowered air-sea temperature differences is accom- means of observation. The primary limitation of SARs applications are tied to how different ocean phenomena data, radar has shown to be an important tool in emer- panied by reduced atmospheric turbulence levels, and is coverage. At this point, C-band SAR imagery from affect surface roughness. NISAR will image long ocean gency response, which is important for remote areas thus less roughness in the regions of the cooler ocean Sentinel-1A and-1B are the NICs primary SAR data surface waves (>50 m wavelength). Although the exact where physical access is limited. contacting the atmosphere. At VV polarization, such source. The Canadian C-band Radar Satellite Constella- mechanisms for imaging of waves is more complicated, a pattern would appear similar to an HH-polarization tion to be launched in late 2018 will provide additional to first order the slope variations of the ocean surface 17.4.15 Maritime: Hurricanes and Wind image under stable air/sea conditions; however, under imagery. and the fact that surface roughness and hence radar Storms unstable conditions, simultaneous imagery at the two cross section at the crests of waves is higher than in polarizations will differ significantly. According to FEMA, hurricanes account for seven of the NISAR will not only represent an opportunity for sig- the wave troughs renders waves visible in SAR imagery. top ten most costly disasters in United States history. nificantly more polar coverage to support operational From this imagery, the two-dimensional ocean wave 17.4.18 Maritime: Ocean Surface The state of Florida was struck by four major hurri- monitoring, but L-band will augment the information spectrum can be computed (Alpers, 1992). There are Wind Speed canes in 2004 with losses totaling $42 billion (http:// available from C-band. Both frequencies are useful in other, more advanced, marine applications of SAR that www.sdr.gov). NISAR imagery can be used to estimate delineating areas of sea-ice coverage. However, the depend on taking advantage of direct velocity measure- The most direct SAR marine application is SAR wind wind speeds within the hurricane, show the shape and longer wavelength of L-band permits deeper ice pene- ments by SAR to map ocean currents (Romeiser, 2013). speed retrieval. Radar backscatter at off-nadir inci- structure of the hurricane eye, map the spatial extent tration and makes L-band more capable in discriminat- dence is proportional to surface roughness near the of the storm surge and flooding, detect coastal erosion, ing sea-ice type. and assess damage to buildings, infrastructure, and the ecosystem. Combining NISAR’s ascending and For operational sea ice uses, the priorities for NISAR descending repeat orbits provides two satellite images data acquisition in order are: [ Figure 17-33] 20.2 20.2 every 12-days which will provide the science and 1. Polar coverage, particularly in the Arctic where Sentinel-1A radar cross 20.0 operations communities detailed SAR imagery and geo- there are more frequent marine operations. 19.9 section image (left) scaled 19.8 detic measurements. Although this temporal frequency 2. The preferred data latency is 6 to 12 hours. After from -25 dB to 0 dB 19.6 19.6 is not sufficient to provide systematic coverage, NISAR 19.4 24 hours, the data are less useful for operations. off the coast of Mexico 19.3 will augment the global earth observation instrument acquired at 2018-01-25 19.2 network, and because of the global observations in 3. The preferred polarization is dual VV and HH. 19.0 19.0 00:32 UTC. Retrieved wind 18.8 some circumstances it may acquire the only pre-event 4. The next preferred polarization configuration is any speed (right) scaled from 18.7 18.6 images. like-polarization plus cross polarization. 0 to 25 m/s. 18.4 18.4 5. The least preferred polarization configuration, but –97.2 –96.7 –96.2 –95.7 –95.2 –94.7 –97.2 –96.8 –96.4 –96.0 –95.6 –95.2 –94.8 still very valuable, is any like-polarization. –25 –20 –15 –10 –5 0 0 5 10 15 20 25 0 10 20 30 40 50 dB Wind Speed (m/s) Wind Speed (knots) 218 NISAR Science Users’ Handbook 219 NISAR Science Users’ Handbook

scale of the radar wavelength. The higher the wind 17.4.19 Maritime: Iceberg and Ship regarding the potential of a hazardous oil spills is the response. Saving lives and property are the initial pri- speed, the rougher the ocean surface, and the higher Detection Arctic coastal zone, where the retreating and thinning orities, while later assessments are needed to evaluate the backscattered cross section. This principle is relied sea ice cover has increased interest in transportation the extent and severity of the disaster zone. One important marine application not directly related upon by conventional wind speed scatterometry. Radar and petroleum exploration. A hazardous spill in the Arc- to ocean surface roughness is to monitor ship traffic cross section is maximum looking in the wind direction tic presents an extremely challenging containment and Because satellite radar is a cloud penetrating technolo- and icebergs. At the high resolution of SARs, ships and and a minimum is the cross-wind direction. SAR cleanup response, where NISAR may play a critical role, gy, NISAR can acquire snapshots of the disaster extent icebergs are often visible (Tello et al, 2005). Imaging wind speed measurements generally have over and enhanced by the converging nature of polar observa- regardless of atmospheric conditions, help delineate the Atlantic shipping routes during seasons when order of magnitude finer resolution than conventional tions and the ability to image throughout the extensive flood hazard zones; measure water level changes, icebergs often move into the lanes will support the scatterometers. periods of darkness and cloud cover. Tasking of NISAR primarily in wetland environments; and measure flood NIC’s mandate to provide situational awareness data on in response to such disasters may be critical and will depth in areas where an accurate digital elevation that hazard. Identifying ships near coasts can also help Figure 17-33 shows a radar cross section image off the commence after the disaster occurs. model (DEM) is available. NISAR can be used to map locate and identify illegal dumping of material in coastal east coast of Mexico acquired by Sentinel-1 at 2018- flooding events on a global basis twice every 12 days. water. 01-25 00:32 UTC and the corresponding wind speed 17.4.21 Maritime/Hydrology: Floods Observations will be uninterrupted by clouds and will at resolution of 500 m. Sentinel operates at C-band Hazards provide timely information for flood responders. Even 17.4.20 Maritime: Oil Spills (5-cm wavelength). NISAR wind speed images will be Floods and other water related hazards are among flooding hidden beneath forest canopies will be visible Ocean surface roughness is suppressed by surfactants roughly similar though may be less sensitive at low the most common and destructive natural hazards, in many areas. Multiple types of NISAR measurements and oil slicks. In coastal regions, NISAR has the poten- wind speeds (5 m/s), but more accurate at higher wind causing extensive damage to infrastructure, public and will be useful for flood assessment: InSAR phase, tial to be used monitor oil spills from ships or oil-drilling speeds (>15 m/s; Monaldo et al., 2016; Monaldo and private services, the environment, the economy and coherence and backscatter change, including polarim- platforms (Girard-Ardhuin, 2005). Oil spills in oceans Kerbaol, 2003; Shimada and Shimada, 2003). Not only devastation to human settlements. Recurring hydro- etry, can be used to discern water flow direction, map and coastal waters have widespread impact to the can SAR wind speeds be used for weather forecast- logical disaster-related losses from floods caused by inundation extent and duration, estimate changes in environment, marine life, human health/safety, society, ing, but high-resolution wind speed climatologies can tsunamis, storm surge, and precipitation have handi- water level, as well as wind speed in open water that and regional economy. The 2010 Deepwater Horizon be used to help select sites for offshore wind power capped the economic advancement of both developed are not as readily or consistently available from optical oil spill caused a major economic disaster, spreading turbines (Hasager et al., 2015). Wide swath observa- and developing countries (e.g., https://www.gov.uk/ satellite sources. NISAR will be capable of monitor- oil from ~50 miles off the Louisiana coast throughout tions, similar to those for currents in the open ocean, government/uploads/system/uploads/attachment_data/ ing water level change in marsh areas, allowing for much of the Gulf of Mexico and to coastal areas in all are preferred for the winds, with resolutions being finer file/286966/12-1295-measuring-human-economic-im- prediction of downstream flooding. Permanent stream U.S. states bordering the Gulf (Figure 17-34). Smaller, than the scales of frontal systems and storm gradients pact-disasters.pdf). Most hazardous hydrologic events gauges are installed and monitored specifically for that yet significant, spills occur regularly, mainly in coastal at sea. These will produce even sub-kilometer scale are local and short-lived and can happen suddenly and purpose, but they are sparsely distributed. NISAR data areas around the globe. The hazard response com- wind estimates, which are important for coastal areas sometimes with little or no warning. Millions of people will augment these data and provide increased spatial munity and the International Charter have extensively with respect to, for example, building wind-turbines, can be impacted by major floods. U.S. insurance claims coverage, filling in the gaps between gauges. used SAR imagery to track oil spills and help guide coastal structures, shipping, and biological interactions. from floods total in the billions of dollars per year. In the mitigation efforts. A region of increasing concern 2015 and 2016, for example, 18 major flood events Among the organizations that respond to flooding di- hit Texas, Louisiana, Oklahoma, and Arkansas causing sasters are state and local agencies, as well as federal [ Figure 17-35] extensive damage (Figure 17-35). Timely evaluation agencies, such as Federal Emergency Management Left: Radar false color image product Agency (FEMA), the National Oceanic and Atmospheric [ Figure 17-34] of flooding conditions is crucial for effective disaster near Farmerville LA (March 13, 2016, UAVSAR image of oil from the by NASA’s UAVSAR) during a devas- Deepwater Horizon spill acquired tating flood. Orange and yellow areas over southeastern Louisiana. The are flooded forests, and black areas oil slick shows up as a dark area in are lakes and open floods. This type of this false color image, as it damp- information is invaluable for local, state, ens the capillary waves smoothing and federal agencies that provide assis- the surface, resulting in reduced tance. Right: Example of the immense backscatter energy. and costly flooding that occurred in the Farmerville region during this flood. (James Fountain, USGS) 220 NISAR Science Users’ Handbook 221 NISAR Science Users’ Handbook

Administration (NOAA), and the United States Geolog- Flooding, coastal inundation, and tsunami applications During natural disasters, first responders often look From these images, it was possible to measure centi- ical Survey (USGS). International aid in the event of will greatly benefit from the high frequency data collec- to NASA to provide timely and valuable information meter-level changes in water level during the 24 hours natural disasters caused by flooding often includes tion to assess flood duration, inundation zones, draining to assist their work to mitigate damage and assess that had elapsed between the observations. SIR-C data sharing arrangements to help our allies respond and habitat response. destruction by these common tragic events. Many demonstrated that radar could be used to make these to the humanitarian crises that flooding can cause. federal agencies and university researchers have types of measurements (Figure 17-36). During natural disasters, these first responders often 17.4.22 Hydrology: Flood Forecasting difficulty evaluating the health of our waterways and look to NASA to provide timely and valuable information wetlands due to lack of information regarding the ebb NISAR will function like a virtual stream gauge for Flood forecasting informs downstream communities to assist their work to mitigate damage and assess and flow of food waters during normal and extreme flooded conditions that occur along most of the world’s if a flood is coming and how much flooding to expect. destruction by these common tragic events. seasonal flooding. The data from NISAR over wetland major rivers, capable of precisely measuring change in Like a virtual stream gauge, synthetic aperture radar areas will be invaluable to management authorities, water level with every observation. is able to measure changing water levels in standing Surface water hydrology hazards have similar mission scientists, and local planning agencies. NISAR can meet vegetation as flood waters from heavy upriver rains requirements as the solid earth hazard applications, these diverse needs through its dependable observing 17.4.23 Hydrology: Coastal Inundation head downstream. where they need to have an up-to-date baseline data strategy that will collect high resolution data over 90 Monitoring inundation of marshes, swamps or other archive, rapid tasking to ensure that the satellite is percent of the Earth’s land surface. NISAR will provide Change in upstream water levels can be very useful flooded areas is difficult: on the ground, inundated collecting data on every possible orbit in case of an crucial information regarding flooding events, even in for predicting downstream flooding. Permanent stream areas can be treacherously difficult to navigate, while event, adequate spatial coverage of the target, and remote areas without stream gauges or other sources gauges are installed and monitored specifically for that from above, vegetation, clouds, and weather can make data quickly delivered in a georeferenced format that is of ground data measuring flood conditions. purpose, but they are sparsely distributed. Not only the water difficult to observe. Beyond the human im- easily disseminated to the emergency responders. The will NISAR be capable of augmenting this network of pact, the extent and duration of inundation has a heavy addition of polarimetric SAR capabilities provides im- InSAR can be used to precisely measure very small stream gauges with continuous maps of change in wa- influence on fish and other wildlife habitats, vegetation proved subcanopy imaging and characterization of the changes in water level in areas with standing vegeta- ter level in some areas, but NISAR will also be capable health, and other parameters of ecosystem health. flood extent and will likely provide better estimates of tion if repeated observations by radars like NISAR, are of monitoring the change in the level of floods far from NISAR will allow uniquely detailed monitoring of the the vegetative frictional contribution in the storm surge made from the same vantage point, i.e., from the same the main river channel, where water can increase in seasonal ebb and flow of flood waters in the Earth’s modeling (DESDynI Applications Workshop, 2008). Data orbit. This was first demonstrated with the NASA SIR-C level and subside at different rates. The same technolo- wetland areas, not just storm-related flooding. The frequency needs for the emergency responders are mission that flew on the NASA Space Shuttle in 1994. gy can provide information about soil moisture, another NISAR all-weather and forest-penetrating radar can daily with sub-daily optimum for hazard response, and SIR-C twice imaged the Purus River, a tributary of the parameter needed for flood forecast models. detect open water areas, and also the flooded areas thus NISAR will not fully meet these needs on its own. Amazon–Solimões River, during flooded conditions. below trees (Figure 17-37).

[ Figure 17-36] Flooded Forest [ Figure 17-37] “Change in water level” products in 0 Scale km 5 Dual polarization radar image flooded, vegetated areas were first Emergent Shrubs of the Maurepas Lake and demonstrated by the NASA SIR-C surrounding swamp in Louisiana. Synthetic Aperture Radar (SAR). In This image was acquired from this image, centimeter-level changes space by the Japanese ALOS-2 in water level were measured over L-band radar. In this false the Purus River in Brazil from two color image, yellow areas are the observations acquired just 24 hours flooded Cypress Tupelo swamp, apart. (Alsdorf et al., Nature, 2000). pink are unflooded areas, orange Colors indicate how much the water areas are degraded swamp level changed between the two ob- A A marshes, and dark areas are servations. Between transects A & B open water. Image © JAXA 2016. was a 1-5 cm change in water level. B B 222 NISAR Science Users’ Handbook 223 NISAR Science Users’ Handbook

Many federal agencies and researchers that study to surface water routing models that estimate rainfall Thus, the impact of soil moisture is more prominent in help estimate wildfire probability, constrain snowpack wetlands have difficulty evaluating their health status runoff for reservoirs, water conveyance systems, and the L-band signal as compared to C-band. water content (see Section 17.3.2), develop water rout- due to lack of information regarding the ebb and flow floods will benefit a wide societal cross-section. SAR ing and flooding numerical models, and detect spills. of flood waters during normal and extreme seasonal backscatter is directly related to near-surface moisture Potential applications of multi-frequency SAR data in inundation. NISAR imagery will provide near weekly content (volumetric) that changes the reflective target the field of soil moisture estimation were explored with 17.4.25 Underground Reservoirs: observations that complement optical data, imaging properties. At microwave frequencies, the dielectric the SIR-C/X-SAR mission over Bhavnagar, Gujarat. SIR- Groundwater Withdrawal through clouds and below the canopy. This capability constant of dry soil is around 3, while that of water C/X-SAR operated in the L-, and C- and X-bands. This Extraction of groundwater often causes the aquifers makes NISAR‘s imaging of wetland areas valuable to is around 80 and depends on salinity. The dielectric mission clearly showed (Figure 17-38) that L-band, and to contract at depth that in turn causes the overlying management authorities, scientists, and local plan- constant for a moist soil ranges between 3 and 30. presumably that S-band, is able to sense deeper layer ground surface to subside. The water in the aquifers, ning agencies. NISAR will provide invaluable new and As the dielectric constant of a material increases, soil moisture, whereas, C-band and X-band are not called groundwater, is an extremely valuable resource, independent information regarding flooding events in the Fresnel reflectivity also increases, resulting in an able to sense deeper layer soil moisture due to their like a water savings account that can be drawn on disaster scenarios, as well as data to develop unique increased backscatter. Radar wavelength determines low penetrability within the soil medium. when times are hard. The water in the aquifers origi- seasonal evaluations of wetland dynamics. the penetration depth. As longer wavelengths have nally was precipitation that made its way down through higher penetration depth within the soil medium, they A similar study conducted to understand the usefulness the soil and rock via cracks and pores. All aquifers 17.4.24 Hydrology: Soil moisture sense soil moisture from deeper layers as compared to of L-band in soil moisture estimation over agricultural are not created equal: aquifers can hold small or vast shorter wavelengths. The soil surface roughness is also terrain used DLR-ESAR data over an agricultural area Estimating spatial and temporal variability of soil mois- amounts of water and recharge quickly or slowly a function of the wavelength. For longer wavelengths having varying surface roughness, crop cover and ture globally at sufficient resolution to help manage depending upon the type of rock both in and above the the soil surface appears smoother, i.e., at L-band soil varying soil moisture content. This study also showed agriculture production, assess wildfire risks, track re- aquifer. The rate of groundwater extraction often in- surface appears smoother as compared to C-band. that L-band is able to capture the signature of soil gional drought conditions, detect spills, and contribute creases rapidly during droughts when surface water is moisture better than C-band. not available to supply demand for water. Groundwater is extracted from aquifers in every U.S. state and nearly An advantage of co-collecting L- and S-band radar im- L-BAND C-BAND X-BAND every country around the world, so this is a widespread [ Figure 17-38] agery will be to characterize soil moisture as a function issue. Multi-frequency SIR-C of penetration depth and to differentiate phase signa- response to deeper tures from soil penetration variations from deformation Ground subsidence due to groundwater withdrawal layer soil moisture from in repeat pass interferometry. While surface can have many effects on infrastructure and buildings. deformation is non-dispersive, one would expect the These can extend from cracks in roads and bridges phase signature from soil moisture variations to be to reduction in freeboard on levees, canal walls, and wavelength-dependent, so differences in S-band and 0 4 km dams to large-scale changes in runoff, river flow, 0 4 km 0 4 km L-band interferometric phase can be used to discrimi- or coastal flooding. Large amounts of contraction in nate deformation from moisture changes. ) aquifers can also damage the water extraction wells

6 0 themselves and require costly redrilling. The contrac- –5 NISAR has the potential to provide high-resolution tion of the aquifers when water is extracted usually –10 Lvv soil moisture variability products that will contribute L X includes both temporary elastic contraction that can be Cvv to each of the science components of the mission by –15 Xvv recovered when the water is replaced and permanent characterizing and removing soil-moisture-induced

SAR Backscatter ( a –20 deformation that cannot be recovered. When aquifers Dry Moist Deeper Wet noise within SAR imagery. The co-collection of L- and Layer are permanently deformed, they lose capacity to store S-band imagery will estimate depth dependent soil future water. C moisture variability and will help isolate and remove soil moisture phase noise in targeted deformation in- Sustainable, low impact groundwater extraction is terferograms. This combination will lower the detection possible, though, given information about the aquifer threshold for resolving subtle and transient deformation and the surface changes associated with pumping. This in NISAR imagery. The resolution of the NISAR imagery is where imaging by satellite radars capable of mea- will be at a level that can be used to manage crops, 224 NISAR Science Users’ Handbook 225 NISAR Science Users’ Handbook

suring changes in surface elevation, like NISAR, has than extraction, leading to uplift of the surface. In [ Figure 17-39] 400 immediate and practical value. NISAR will image global some cases, the oil and gas extracted includes a large Increasing rates of earthquakes in the central United SEISMIC SURGE IN OKLAHOMA amount of wastewater that requires disposal. In most Earthquakes/year land areas every 12 days, providing a time series of the States. Recent advances in the technology used in places wastewater is reinjected into rocks at depth, surface uplift and subsidence. This information shows hydrocarbon production, including enhanced oil recovery 300 both the long-term decline in surface elevation, which which can cause induced seismicity, which is described and increases in the volume of wastewater injected into corresponds to unrecoverable loss or slow recharge in another section. the subsurface, are associated with a dramatic increase of groundwater, and a seasonal cycle of uplift and in earthquakes felt in the central United States since the 200 subsidence that correlates to a sustainable balance The NISAR satellite mission will provide high-resolution mid-2000’s. Damaging earthquakes only appear to be between precipitation and withdrawals. Armed with this ground movement maps on a global basis with weekly related to a small fraction of wells, but there is not yet information, users can protect this valuable renewable sampling. Observations will be uninterrupted by weath- enough data to definitively determine in advance the 100 water asset over the long term, avoiding the terrible er and facilitate safe resource development by improv- safety of operations at a particular site. Seismology can consequences of permanent loss of water supply. ing understanding of processes that impact regions tell us the response of Earth’s crust to forcing by fluid undergoing active extraction or injection of subsurface injection and production. However, seismic data is blind to the slow, longer term deformation of the ground surface 17.4.26 Underground Reservoirs: Oil and fluids, including oil, gas, and geothermal power. The 0 associated with pumping, injection and even incipient 1980 1990 2000 2010 Gas Production observations made over the lifetime of NISAR will be a creep on faults that will eventually rupture in a more giant step forward in our understanding of subsurface Number of earthquake sequences each year that contain at least Efforts to utilize subsurface resources, including water, damaging earthquake. NISAR can provide the missing link fluid flow and associated seismicity and will inform one magnitude 3 or larger earthquake, since 1973, for California oil, gas, and geothermal power, necessarily involve the to this puzzle, complementing the available seismic data (light blue) and Oklahoma (dark blue). From McGarr et al., 2015, the next generation of methods for characterizing and in the journal Science. extraction and injection of large volumes of fluid from and helping to track how patterns of fluid flow beneath managing these resources. the ground – often in areas that also host valuable in- the surface relate to patterns of observed earthquakes. frastructure and large population centers. Groundwater 17.4.27 Underground Reservoirs: Induced effects were described in the previous section. Oil, gas,

Seismicity Displacement Towards Satellite (mm) and geothermal extraction operations affect a subset of 10 the United States and other countries around the world, Earth scientists have been investigating earthquakes

but new technologies, including hydraulic fracturing, of tectonic origin for more than a century, developing 36’30’ 9 15 have expanded the areas. Oil and gas are extracted significant insights and understanding about where 6 from a wide variety of rock reservoir types and depths, they occur, how frequently they occur, their links to 0 using a large range of methods. Geothermal power is geologic structures and processes, their magnitude 4 North (km) –15 often extracted from the Earth by pumping water out of distribution, and how frequently main shocks trigger 2 aftershocks. For the past 40 years, and particularly –2 0 2 hot rocks. Displacement (cm) 5 km over the past decade, a new class of earthquakes has 36’15’ 0 –97’15’ –97’00’ –96’45’ 0 2 4 6 8 10 12 14 Withdrawing fluid from rocks at shallow depths become increasingly important – earthquakes induced East (km)

without replacement will cause compaction within the or triggered by human activities. Human activities [ Figure 17-40] reservoirs and subsidence of the overlying land surface hypothesized to have caused earthquakes, in decreas- Left: NISAR data will permit systematic mapping and monitoring of effectively to determine the true ground movement. Right: Even tiny ing order of numbers of suggested instances (e.g., similar to the compaction of groundwater aquifers and earthquakes, even in agriculturally active areas. In this example, using earthquakes can be imaged when enough data is available. This image its associated surface subsidence. Extracting heat from www.inducedearthquakes.org) include mining, water data from the European Space Agency’s Sentinel-1a and -1b platforms, uses data from the European Space Agency’s TerraSAR-X platform geothermal reservoirs can also cause the rocks to con- reservoir impoundment, conventional hydrocarbon pro- we can see several centimeters of displacement over a 10 km x 10 km to constrain subsidence of the ground during a 2013 magnitude 3.2 tract and subsidence of the surface. Some advanced duction, fluid injection including disposal of wastewater region associated with the 2016 magnitude 5.8 Pawnee, Oklahoma, earthquake within the Chicago metropolitan area, triggered by a blast at methods of oil and gas extraction involve the injection associated with hydrocarbon production, geothermal earthquake. The main earthquake location (red star) and aftershocks a limestone quarry. Color indicates displacement of the ground surface of water or steam into the reservoirs to stimulate energy production, hydraulic fracturing, groundwater (black dots) outline a complicated pattern that provides insight into towards or away from the satellite, which was traveling to the northwest extraction. If injection volumes exceed extraction, then extraction, and carbon sequestration. The recognition the patterns of weakness in the subsurface. The red band on left is the in the direction of the blue arrow and aimed down towards the Earth in the ground surface above may move upward. Geo- that human activity can trigger earthquakes has led to signature of a large storm that was present during one of the image the direction of the red arrow. The observed displacement shown here thermal operations may also involve greater injection acquisitions. When many images are available, such as would be tightly constrains the depth of this earthquake to 700 meters, much provided by NISAR, such atmospheric effects can be averaged out more shallower than the zone where “natural” earthquakes occur. 226 NISAR Science Users’ Handbook

great concern among government, industry, and the to study because of the complete lack of data before public. the event. This situation should change in the future: NISAR data would be acquired regularly over the entire Some instances of earthquakes that may be triggered United States, allowing imaging of areas like Oklahoma, by human activity occur in regions that naturally experi- Texas and Kansas that have both active agriculture and ence frequent earthquakes (e. g., California, Italy, Spain, hydrocarbon/water resource development. Tibet). Others, such as those in the 21st century in the central United States and the Netherlands, represent a 17.4.28 Rapid Damage Assessment significant change: Oklahoma now experiences more A key need after disasters, with both natural and hu- earthquakes each year than California (Figure 17-39). man-induced causes, is a rapid assessment of damage Efforts to reduce our society’s reliance on fossil fuels to buildings and other infrastructure. Frequent coverage also compound the problem, with increases in earth- of the land areas by imaging radar satellites, including quake frequency associated with energy production at NISAR, enables all-weather assessment of damage geothermal power plants and at dams that contribute with measurements of coherency changes. Processing hydrothermal power to the grid. of a damage proxy map, or change detection map, can show areas of potential damage very quickly The increased frequency of triggered and induced after the radar data is received. This method has been earthquakes creates new challenges, particularly since demonstrated for a wide variety of disasters, including the energy and resource needs of our population are earthquakes, tsunamis, volcanic eruptions, hurricanes, likely to continue to grow. Even forecasting the expect- tornadoes, and landslides. ed damage from these new types of earthquakes is not just “business as usual.” Analyses of shaking reports One example of rapid damage proxy mapping was after from the central United States and the Netherlands a magnitude 7.8 earthquake hit central Nepal on April suggest that the distribution of damage from these 25, 2015. The quake killed nearly 9,000 people and earthquakes, which tend to be shallower, are different induced more than 4,000 landslides in the precipi- from the damage expected from “traditional” earth- tous valleys of the Himalayan Mountains. Widespread quakes, which often occur deep underground. building damage was rapidly mapped using radar data acquired by Italian COSMO-SkyMed and Japanese Satellite-based radar imagery, when available, can ALOS-2 satellites. The maps were quickly released be an extraordinary tool for characterizing how the to national and international responding agencies. Earth’s surface warps and deforms before, during, and Field crews were dispatched to damaged sites and after induced earthquakes. The examples of induced made ground observations guided by the maps, and a earthquakes in the central United States (Figure 17- satellite operating company used these maps to target 40) are cases where we were fortunate to have data areas for imaging with ultra-high resolution spaceborne both before and after the earthquake. Many other optical sensors. earthquakes in these regions have been impossible NISAR Science Users’ Handbook 229 NISAR Science Users’ Handbook

18 APPENDIX F IN SITU MEASUREMENTS FOR CAL/VAL

18.1 SOLID EARTH IN SITU MEASURE- This parameterization of ground deformation has a MENTS FOR CAL/VAL long heritage, both in the analysis of GNSS time series and more recently with InSAR data (e.g., Blewitt, 2007, 18.1.1 Co-seismic, Secular and Transient Deformation Rates Hetland et al., 2012, Agram et al., 2013). The project will have access to L2 position data for continuous The most direct validation of NISAR solid earth GNSS stations in third-party networks such NSF’s deformation measurements is with continuous GNSS Network of the Americas (NOTA), the HVO network for measurements of ground displacements. For individual Hawaii, GEONET-Japan, and GEONET-New Zealand, point locations, GNSS provides continuous time series which are located in target regions for NISAR solid of 3-component vector ground displacements that Earth calval. Station data are post-processed by anal- can be projected onto the SAR line-of-sight imaging ysis centers that include NSF’s GAGE Facility and the direction to allow direct comparison with InSAR-derived Nevada Geodetic Laboratory at the University of Nevada displacement/velocity observations. Validation will be Reno (UNR), are freely available, and have latencies of repeated annually in order to assess improvements several days to weeks. Current networks contain one or with increased numbers of image acquisitions and to more areas of high-density station coverage (2~20 km detect any potential degradation of the system. nominal station spacing over 100 x 100 km or more), which will support validation of L2 NISAR requirements Comparisons of GNSS and InSAR observations will be at a wide range of length scales. Future GNSS networks done in regions where many (5+) GNSS observations are likely to have even greater station density due to are available within the footprint of individual InSAR ongoing infrastructure investment at the federal and data products (e.g., coseismic displacement map, state levels. velocity map, etc.). GNSS secular velocities are now SECULAR, CO-SEISMIC, AND TRANSIENT routinely estimated at 1-σ levels of (0.2, 0.2, 0.6) mm/ DEFORMATION CAL/VAL SITES yr (east, north, up) – significantly better than NISAR’s L2 requirements (Figure 18-1). Similarly, coseismic After assessing the current national infrastructure offsets can be estimated at 1-σ (0.8, 0.5, 1.3) mm for GNSS processing and data availability, the NISAR (east, north, up) using 30-second position solutions Solid Earth Science team decided to use the GNSS (Liu et al., 2014), and significantly better using daily station displacement time series produced by the UNR solutions. Generally, validation will occur in locations Geodetic Laboratory for NISAR CalVal. The UNR dataset with stable, linear ground motion, i.e., with no events has global station coverage, uses the openly available generating transient displacements, by comparing GIPSY processing software written and maintained by background noise levels. Validating the ability to detect JPL, and has been produced continuously for over a transients will be done by assessing agreement of decade. UNR funds its operations via a mix of federal contemporaneous GNSS and InSAR measurements of (NASA, NSF, USGS) and state (Nevada Bureau of Mines seasonal quasi-periodic displacements where these are and Geology) support. Other processing centers in the known to occur (e.g., over shallow confined aquifers) U.S. (NSF’s GAGE facility, JPL’s Measures program) (e.g., Lanari et al., 2004). Different length scales will be currently process fewer stations globally and in North analyzed to validate performance over the length scales described in the L2 requirement. 230 NISAR Science Users’ Handbook 231 NISAR Science Users’ Handbook

50°N 700 Mean: 0.22 mm/yr 18.1.2 Permafrost Deformation In the past, the community has used the following data [ Figure 18-1] 600 Std: 0.06 mm/yr 500 types for calibration and validation of InSAR-based Location of 1,860 GPS 48°N As InSAR is inherently a relative measurement, the 400 permafrost measurements: sites in the Western 300 calibration and validation of permafrost deforma- 200 U.S. with histogram of 46°N tion measurements involves (1) the identification of a) Dry floodplain areas as no-deformation sites

Station Count 100 1-sigma errors of the 0 (Liu et al., 2014; Liu et al., 2010). 0.0 0.2 0.4 0.6 suitable reference points (calibration) to tie NISAR 44°N 3-component secular Sigma East (mm/yr) measurements to an absolute datum, as well as (2) the b) Dry margin of drained lake basins as no-deforma- 700 velocities estimated for 600 Mean: 0.22 mm/yr provision of a suitable number of validation points that tion sites (Liu et al., 2013). 42°N Std: 0.06 mm/yr the CPGS sites. The mean 500 can be used to analyze the permafrost deformation and standard deviation 400 c) Modelled seasonal subsidence at CALM grids 300 accuracy that could be achieved by the NISAR system. of the 1-sigma values for 40°N based on active layer thickness and assumed soil 200

the network are indicated Station Count 100 water content (Schaefer et al., 2015). 38°N 0 in each panel. Locations 0.0 0.2 0.4 0.6 and velocities down- Sigma North (mm/yr) 36°N 700 loaded from http://www. TABLE 18-1. TABLE OF SOLID EARTH SCIENCE CAL/VAL REGIONS (CHOSEN TO REPRESENT DIVERSITY 600 Mean: 0.72 mm/yr Std: 0.17 mm/yr OF TARGETS AND GPS COVERAGE). BHUJ, INDIA SITE IS TBD unavco.org/data/data. 500 34°N 400 html on Nov. 4, 2014. Desert, Scrub, Savanna 300 Additional dense CPGS is 200 32°N Central Valley, CA Csa Temperature/Dry/Hot Agriculture, soil moisture, no relief available in Hawaii. 0 200 400 km Station Count 100 0 126°W 123°W 120°W 117°W 114°W 111°W 108°W 105°W 0.2 0.6 1.0 1.4 LA Basin/Mojave Csa, Bsh Temperature/Dry/Hot, Urban, range of relief and decorrelation sources, Sigma Up (mm/yr) Arid/Steppe/Hot change in base elevation

Long Valley Caldera Bwk Arid/Steppe/Cold Variable relief, snow, ground type

America, but they make their data openly available and ments are modeled and removed using the FES04 Bhuj, India Bsh Arid/Steppe/Hot Climate, salt flats, seasonal flooding, agriculture would serve as a backup in the case of disruption to model, semiannual tidal loading is removed per IERS Mejillones, Chile Bwk Arid/Steppe/Cold Hyper-arid, ionosphere, large relief, change in ground UNR funding or operations, mitigating the risk to the 2010 conventions using the hardisp.f program, and all type project of using a single processing center. load calculations are made relative to Earth’s center Mixed Forest of mass. The SES team will facilitate the development Big trees, forestry, rain The GNSS displacement time series used for NISAR of the phase corrections needed to remove these tidal SW of Portland, OR Csb Temperature/Dry/Warm

CalVal will be consistently processed across all CalVal components in NISAR’s interferometric products. North Island, NZ Cfb Temperature/Wet/Warm Southern latitude sites. Additionally, the NISAR SES team will provide These networks will contain one or more areas of Cfa Temperature/Wet/Hot Southern U.S. climate, swamps, urban, no relief (either through its own work, or by linking to openly high-density station coverage (2~20 km nominal Houston/Galveston

available data) corrections for offsets in GNSS time station spacing over 100 x 100 km or more) to support Oklahoma Cfa Temperature/Wet/Hot Agriculture, strong atmosphere, no relief series due to GNSS-specific instrument changes that validation of L2 NISAR requirements at a wide range of Nepal Csa Temperature/Dry/Hot Monsoon, relief, arigculture, atmosphere would not appear in InSAR time series. GNSS and InSAR length scales. displacements also differ in their treatment of solid Maritime Tropical Earth and ocean tides, neither of which is currently The Cal/Val sites where the algorithms will be calibrat- Big Island, HI Af, Aw, As Tropical/Rainforest, Rain forest, relief, tropical climate, island, included in NISAR’s InSAR processing suite. In the case ed and the science requirements validated are listed in Tropical/Savanna lava flows of solid Earth tides, UNR corrects GNSS displacements Table 18-1. Maritime High Latitude using IERS 2010 conventions, although ir does not remove the permanent tide. Ocean tide load displace- Unimak Dfc Cold/Wet Artic, ocean island, unstable atmosphere, relief, snow

Permafrost

Dalton Highway, AK ET Polar/Tundra Permafrost, nothern climate, tundra, ionosphere 232 NISAR Science Users’ Handbook 233 NISAR Science Users’ Handbook

d) Bedrock outcrops as no-deformation sites. For accuracy assessment, we will use the following Sites should have historic records of repeated thaw- • Easy-to-maintain road accessible locations where e) Differential GNSS measurements (Iwahana et al., types of ground-truth information: depth measurements at fixed locations (several re- regular field measurements will be taken peated measurements per thaw season), soil moisture, 2016). 1. We will adopt the common assumption that dry Two general types of CAL/VAL sites will be used for this and galvanic electrical resistivity tomography mea- f) Thaw Tube measurements (Short et al., 2014). floodplain areas are free of seasonal surface effort, “passive” and “active”, depending on the efforts surements. Repeated airborne LiDAR measurements deformation (Liu et al., 2014; Liu et al., 2010). To needed for their maintenance: are also desirable for all proposed sites, providing a large extent, this assumption is based on the • Passive Cal/Val sites include gravely flood plains Traditionally, data types (a) – (d) were predominantly information on long-term surface elevations. Historic fact that low ice content sandy soils and coarse (sites of type (1)) as well as rock outcrops (sites used for algorithm calibration while (e) – (f) were used (5 years and counting) thermistor measurements are gravels present in floodplain deposits show very of type (2)). These sites do not to be maintained for measurement validation. available at all sites. little potential for settlement or upheaval (Pullman long term. Pre-launch tests at passive cal/val sites et al., 2007). Additionally, the heat transfer from should be conducted to verify their suitably for this In addition to validating deformation measurements Many potential passive calibration sites have been streamflow and spring flooding often causes the effort. directly, Schaefer et al. (2015) used a more indirect used in previous research studies either as reference permafrost surface to be several or even tens of approach and validated the InSAR-estimated active location or as a means for validation (i.e. Bartsch et al., • “active” calibration sites must be maintained (see meters under the riverbed and reduces preva- layer thickness (i.e., a higher level product) with the 2010; Liu et al., 2014; Liu et al., 2012). table 2.) lence of ice-rich permafrost, further contributing ALT measured from GPR and probing. to a reduction of long-term thaw settlement (Liu Active sites should have historic records of repeated For the active Cal/Val sites, historic records of repeated et al., 2010). Dry floodplain areas will be used thaw-depth measurements at fixed locations (sev- In this effort, we will use a combination of previous- thaw-depth (several repeated measurements per thaw both for calibration and validation. eral repeated measurements per thaw season), soil ly used methods for both algorithm calibration and season), soil moisture, and galvanic electrical resistivity moisture, and galvanic electrical resistivity tomography requirement validation. 2. Bedrock outcrops in the vicinity to target perma- tomography measurements should have been collected measurements. Repeated airborne LiDAR measure- frost regions will be used as both calibration and in the past. Repeated airborne Lidar measurements are ments are also desirable for all proposed sites, provid- Validating surface deformation estimates in permafrost validation points in similar ways. desirable for all proposed sites, providing information ing information on long-term surface elevations. His- regions is difficult due to the extreme seasonality and 3. In addition to these natural areas, regular field on long-term surface lowering. Historic thermistor toric (5 years and counting) thermistor measurements often remote regions covered by this requirement and measurements at a small set of easy-to-maintain measurements should be available at all sites. Future should be available at all active sites. Measurements due to the fact that in-situ measurements of per- road-accessible locations should be taken. field work will be required at some sites sites. Field should be acquired twice a year (at the beginning and mafrost deformation are difficult to conduct without Two general types of calibration and validation sites work measurements will include: the end of the thaw season) at these active sites, all disturbing the soil and vegetation. Since the ground will be used for this effort, including sites designated • Thaw-depth measurements along the transects currently maintained by the Cold Regions Research and thermal regime is largely controlled by the surface mat as “passive” and “active”, depending on the efforts (every 4 m) following measurement protocols Engineering Laboratory (CRREL) for many years. CRREL of organic soils, peats, and/or vegetation any major needed for their maintenance: established by the NASA ABoVE team is a part of the U.S. Army Corps of Engineers Engineer disturbances to the land cover can lead to subsequent • Passive Cal/Val sites include gravelly flood plains Research and Development Center. thaw and surface subsidence. To minimize disturbance, • Soil moisture measurements according to ABoVE (sites of type (1)) as well as rock outcrops (sites protocols our strategy for validation will include two compo- Site Center Site of type (2)). These sites do not to be maintained Coordinates Owner nents. First, we will use ground-truth data at sparse • Deformation measurements using differential Name long term. Pre-launch tests at passive Cal/Val locations with known surface deformation to assess GNSS equipment Permafrost 64°57’3.61”N CRREL sites should be conducted to verify their suitably 147°36’51.48”W the accuracy of NISAR-based permafrost deformation • Annual ground penetrating radar measurements at Tunnel for this effort measurements. Second, we will perform statistical the beginning and end of the thaw season Farmers 64°52’33.83”N CRREL analyses of selected NISAR observations to arrive at • Additionally, “active” calibration sites should be Loop West 147°40’47.83”W • Field work should be conducted twice per season, robust estimates of the achieved precision of NISAR maintained at the beginning (mid May) and the end (early Farmers 64°52’32.24”N CRREL products. Loop East 147°40’23.14”W October) of the thaw season

Cal/Val sites for permafrost deformation fall into three Creamers 64°52’3.53”N CRREL Field 147°44’17.72”W categories: Goldstream 64°54’41.80”N CRREL • Dry floodplain areas 147°50’59.24”W • Bedrock outcrops in the vicinity to target permafrost 234 NISAR Science Users’ Handbook 235 NISAR Science Users’ Handbook

TABLE 18-2. ACTIVE PERMAFROST SITES IN • As a result, GNSS data on moving ice will also be While most of the GNSS receiver locations will be primary functions beyond those receivers deployed in ALASKA used to help validate ice-sheet velocities placed on slower moving ice (<100 m/year). To the Greenland. Specifically, they will: 18.2 CRYOSPHERE IN SITU extent that there are no safety issues, some GNSS • Provide data to validate the vertical differential MEASUREMENTS FOR CAL/VAL Greenland has the full range of snow facies, ranging devices will be placed on faster moving locations to displacement measurement requirement, as they from wet snow through percolation to dry snow. Hence, validate the fast flow requirements. measure vertical motion due to tidal displacement 18.2.1 Fast/Slow Deformation of Ice the mission will install 10 GNSS receivers along a Sheets and Glacier Velocity • Provide data to validate velocity requirements in divide-to-coast line to validate observations for all snow 18.2.2 Vertical Displacement and Fast Ice-Shelf Flow regions that will rely on a tide model for correction, The main validation approach for the ice sheet and facies and melt states (Figure 18-2). The GNSS will • Provide information about the variability of the glacier requirements will be to compare NISAR-derived operate throughout the 3-year mission and will collect GNSS receivers will also be placed on an ice shelf in ionosphere in southern hemisphere velocity with points of known velocity. In particular, the data with at least daily frequency (e.g., daily 2-hour Antarctica to validate the Vertical Differential Dis- science team and project personal will use stationary segments), foregoing continuous (e.g., 15-s) sampling placement Measurement requirement. These mea- In addition, the mission also will piggyback on other points (exposed bedrock) and velocities measured with at least during the winter when power is limited. Daily surements also will contribute to validating the fast independently funded logistics (i.e., ongoing field GNSS on moving ice. sampling will allow estimation of velocity for any 12- deformation rates (ice shelves have large areas of projects). In particular, these measurements are better day interval, allowing validation of multiple overlapping fast flow with few crevasses, making them well suited suited to fast-flowing areas because the investigators Residuals on rock will provide hundreds to thousands tracks that cover the GNSS line (e.g., Figure 18-2). to GNSS deployment with little risk to the personnel are working in areas where they know the hazards and of zero-velocity validation points to allow monitoring installing them). Specifically, the project will deploy are doing only short term (a few weeks) deployments. In of several sources of error, particularly the ionosphere. These measurements will provide a consistent valida- 4 GPS receivers at TBD (exact siting will depend on any given year, several independently funded investi- While these points are extremely useful, other data are tion time series throughout the mission. These sites will whether the mission operates in a left/right or left-only gators have GNSS stations on the ice in Greenland and needed to supplement exposed bedrock because: be equipped with Iridium links to reduce data latency. mode) locations on an ice shelf near the grounding Antarctica, although several years out from launch we • Bedrock data have zero motion and provide no Such methods are used routinely and no new technol- lines of major ice streams. To minimize logistics costs, have no firm knowledge with regard to from whom, information about slope correction errors ogy development is required. Sites will be deployed this deployment likely can be carried out by UNAVCO when, and where the measurements will come. While near launch and maintained with an annual service personnel who staff McMurdo research station each such results won’t provide the sampling consistency • Scattering characteristics are different for rock visits. With at least 60 observations per year per site, Austral summer. These measurements will serve three of project-supported sites, they will greatly expand the and firn surfaces, resulting in generally lower 10 stations will provide a robust statistical sample for spatial coverage, particularly on fast moving ice. correlation over firn validation (10 x 60 x 3=1800 individual image pair • Bedrock points don’t provide information about comparisons). other ice-related effects (e.g. vertical motion asso- 2500 ciated with firn compaction) [ Figure 18-3] APL TERRASAR-X Example of validation GEUS ALOS/PALSAR of SAR-derived glacier 2000 APL TSX AGGREGATE MAP velocity data using GNSS [ Figure 18-2] (Ahlstrøm et al., 201).

Preliminary locations of sites used for 1500 NISAR ice velocity validation. Final adjustment of points will occur just prior to deployment to take into acount and 1000 avoid hazards such as crevasses. The locations of the points are designed 1 7 to span a wide range of surface types 2 8

6 9 VELOCITY (m/yr) SATELLITE-DERIVED 3 5 500 and conditions, ranging from rapidly 4 10 melting bare ice, to radar bright percolation zone where there is strong refreezing of summer melt, to the radar-dark 0 0 500 1000 1500 2000 2500 interior of the ice sheet where GPS VELOCITY (m/yr) accumulation rates are high. 236 NISAR Science Users’ Handbook 237 NISAR Science Users’ Handbook

An example of a validation using velocities derived encountered in the outer ice zones where the ice is more GNSS buoys semi-randomly distributed across dependent on the geolocation error of the data and the freer to move and less encumbered by surrounding ice. the Arctic which could be sampled over a period of time 2 from TerraSAR-X and ALOS is shown in Figure 18-3. error tends to s f (Kwok and Cunningham, 2002). In addition to validating results, the GNSS data will Another source of error will come from the in situ drift (between 15-30 days, for example) with consistent be useful for determining and analyzing the impact of buoy data set used for SAR validation. SAR-derived motion fields. The buoy positions reflect The SAR-derived trajectories are derived from sequen- ionosphere’s total electron content (TEC) on velocity the continuous motion of the ice as well as provide indi- tial images obtained over a few days interval (approxi- measurements (Meyer, 2014). The most common approach to validate sea ice velocity cations of deformation events of the sea ice cover over mately 3-5 days with NISAR) based on 5-km grids, with (m/s) is by comparison of displacements derived from time. The accuracy of the trajectories derived from both four known grid corner points, using the Eulerian tech- Beyond GNSS, the mission will evaluate NISAR products the SAR imagery with those from drift buoys, which is the drift buoys and the SAR will be compared during se- nique. Using the 1-hourly buoy data, a buoy position is against those derived from other spaceborne sensors what will be done for NISAR. Sea ice drift buoy data lected winter and melt periods and in both polar regions linearly interpolated to the time of SAR image A. Then (other SARs and optical) by science team and members are available from multiple sources, supported by depending on buoy availability. the nearest grid point in Image A to the buoy position is of the larger community. This activity will help establish other research programs. The International Arctic Buoy determined. If the distance between the 2 points is less that there are no frequency, sweep-SAR or other sensor Program (IABP; http://iabp.apl.washington.edu/) began The two primary sources of error measuring ice motion than 3 km, this pair of points is retained for analysis, specific differences. measuring sea ice motion using drift buoys in the Arctic with tracking of image pairs are the absolute geographic Image A (x,y) and interpolated buoy (x,y). Then the buoy Ocean in 1979 and continues this effort to the present position of each image pixel and a tracking error, which position is interpolated to the time of Image B to obtain 18.2.3 Sea Ice Velocity day. This multi-country funded long-running program is the uncertainty in identifying common features from buoy (x’,y’), which is compared to the same grid point is expected to continue through the NISAR mission one image to the next image. Ice drift buoys are fixed from Image A now in Image B (x’,y’). The difference For NISAR, sea ice velocity products will be validated and well into the future. The position error for the in the ice upon which they are deployed. Buoy position in displacement D between the SAR-image pair (s) using displacement comparisons with drift buoys. The older buoys reported by the IABP using the Advanced errors depend on the positioning systems utilized (e.g., and the interpolated buoy (b) is then derived for each deformation-related output products generated by Research and Global Observation Satellite (ARGOS) po- GNSS), as discussed above. The comparison between comparison, the NISAR sea ice tracker (divergence, shear, rotation) sitioning system was ~0.3 km (Thorndike and Colony, SAR and buoy ice motion tracking then combines the er- u = ((s ’ – s )2 + (sy’ –sy) ) 1/2 will not be validated due to the significant expense of s x x 2 1982; Rigor et al., 2002). Current buoys using GNSS rors in SAR geolocation, tracking, and buoy positioning. u = ((b ’ – b )2 + (b ’ –b ) ) 1/2 mounting an appropriate field campaign and because b x x y y 2 have reduced this error to ~10 m or less and provide The buoy locations will be estimated for the SAR-derived D = (u -u ) these quantities are not included in the Level 1 or Level s b daily products at 1-hour intervals. Buoys are deployed positions and measurement times using the once hourly from which the mean, standard deviation and rms in m 2 requirements. Errors in measurements of sea ice vein the fall or winter on thick sea ice intended to last drift buoy data with linear interpolation. will be derived for multiple comparisons, by season and locity trajectories, derived from image pairs, come from through the summer, often remaining within the Arctic location. two primary error sources: errors in determining the Ocean more than one season before exiting out of the The errors in motion that will be derived include 1) location of ice in the second image that corresponds Fram Strait. Often up to 15-20 buoys may be present absolute geographic position error (provided by the The tracking error of the buoy/s is zero, since the buoy to the same ice in the initial image, and errors in the at any one time (Figure 18-4). Additionally, drift buoys project), 2) tracking errors between pairs of images, is stationary on the same piece of ice. The error will geolocation of either of the two images. The geolocation are being deployed that include ice and snow thick- and 3) the mean magnitude and standard deviation of then be based on geolocation errors associated with accuracy of NISAR is expected to be better than 10 m. ness measurements (http://imb.crrel.usace.army.mil/ the displacement differences between SAR-derived and the buoy location, the SAR grid point geolocation and Thus, the primary source of error is expected to come buoyinst.htm) as well as upper ocean properties (http:// buoy-derived displacements. the SAR grid point tracking error. Preliminary analysis from the first source. Ambiguities in identification of the www.whoi.edu/page.do?pid=20756; http://psc.apl. of recent data from 12 buoys gives worst case errors same ice in two images can arise from deformation and washington.edu/UpTempO/), which can be added to the The uncertainties in ice displacement, u, and spatial of 32 m/day in each component of 3-day velocity es- rotation of the ice field, SAR system noise, and variance analysis pool. There is a parallel program for the South- differences derived from SAR imagery are discussed by timates, which is of sufficient accuracy to validate the in backscatter due to environmental conditions that re- ern Ocean named International Programme for Antarctic Holt et al. (1992) and Kwok and Cunningham (2002). The displacement requirement. Prelaunch the displacement sults in reduced contrast of ice features such as ridges. Buoys (http://www.ipab.aq/), however the coverage is error in u has a zero mean and a variance of errors can be derived using image pairs from L-band One factor that can affect variations in ice ambiguities less dense due to deployment logistics and the typical s2 = 2s2 + s2 (ALOS-1, ALOS-2, and potentially SAOCOM) along with in SAR imagery is the repeat sampling interval of the u g f shorter buoy lifespans of <1 year due to the seasonal where sg and sf are uncertainties in the geolocation of C-band imagery from Sentinel-1 as a way to test the image pairs. In general, previous studies show that 3-4 nature of the ice cover in that region. the image data and the tracking of sea ice features from tracking algorithm. If there are significant differences day intervals are suitable for tracking sea ice within the one image to the next, respectively. Locally, where the between the image pairs, this may be due to either central pack. For sea ice within the marginal ice zones, Cal/Val data for sea ice velocity will be provided by geolocation errors between two images are correlated sea ice deformation including shear and divergence rel- shorter repeat intervals of 1-2 days enhance tracking non-project supported sea ice drift buoys deployed when the points are close together, the calculation of ative to the SAR and buoy locations, and difference in performance largely due to faster ice velocities often every year in sea ice regions of the Arctic and Antarctic spatial differences to determine velocity is no longer backscatter due to environmental conditions including oceans. A representative array would consist of 10 or warming and presence of melt ponds. 238 NISAR Science Users’ Handbook 239 NISAR Science Users’ Handbook

As mentioned previous evaluations of ice tracking to resolving the cell boundary with only four points. It forest canopy metrics to characterize the variations of The number of ground plots for each site must suffice errors using Radarsat-1 using 3-day image pairs was also found that the displacement errors be- AGB at the landscape scales (a minimum area of 100- to statistically develop the algorithmic model for and IABP buoys (3-hour data, Argos tracking) have tween buoys and SAR at the starting positions were 1000 ha depending on the vegetation and topography). achieving better than 20 percent uncertainty in AGB es- resulted in the following displacement errors: The significantly improved when the distance between a The landscape scale distribution of AGB in the form of a timation (NISAR requirement). This number is expected squared correlation coefficient for Radarsat-1 and buoy SAR image grid point and a buoy were <2 km (Figure map will be used to calibrate algorithms and to validate to be 20-30 plots depending on vegetation heteroge- displacements was 0.996 and the median magnitude 18-5a,b) compared to <5 km (Figure 18-5c, d), with the NISAR AGB product. neity. Ground measurements at each plot must include of the displacement differences was 323 m (Lindsay the latter results indicating a greater likelihood of tree size (diameter, height), wood specific gravity (by and Stern, 2003). The tracking errors gave rise to error deformation occurring over time and leading to greater The NISAR biomass algorithm depends upon parame- identifying plants), GNSS measurements to charac- standard deviations of 0.5% /day in the divergence, errors which were not included in the error tracking. ters that are a function of five global terrestrial biome terize the plot shape and size (< 5 m accuracy), and shear, and vorticity. The uncertainty in the area change types (broadleaf evergreen, broadleaf deciduous, other ancillary (optional) data such as soil moisture, of a cell is 1.4% due to tracking errors and 3.2% due 18.3 ECOSYSTEMS IN SITU MEASURE needleleaft, mixed broadleaf/needleleaf, and dry forest soil properties, phenology, etc. Ground-estimated AGB MENTS FOR CAL/VAL & woodland savanna). Biomes are referred to regions must use established local or global allometric models and must include any uncertainties associated with the 18.3.1 Above Ground Biomass (AGB) with similar climate and dominant plant or vegetation [ Figure 18-4] types that may be sub-divided into continents to ground-estimated AGB. Ground plot data may include A multiscale approach based on in situ and LIDAR all field measurements or only AGB estimates with ac- Representative map capture additional diversity in species and climate. data is necessary for validation of the NISAR bio- of Arctic drift buoys The NISAR Cal/Val sites are required to represent these curate location and size of plots if there are restrictions (from IABP). mass measurement science requirement and reduce biomes and span across their structural and topograph- on disseminating the tree level measurements. For uncertainty in AGB at the regional to continental scale. ical diversity to insure the algorithms meet the require- sites without the ALS data, the ground plot size must At the finest resolution, in situ field measurements ments for global estimation vegetation AGB. For each be > 1-ha (100 m x 100 m) to allow direct calibration of forest characteristics will be used to estimate AGB biome a minimum of two sub-regions for independent of the algorithmic model with radar imagery. For sites using allometric equations at the hectare or sub-hect- training and validation that include AGB range with ALS data, the plot size can vary from 0.1 ha to are scale. These in situ estimates of AGB will then be 0-100 Mg/ha are recommended.However, a larger num- 1.0 ha (plot shape variable) depending on vegetation upscaled with Airborne Scanning Laser (ALS) Lidar ber of Cal/Val sites will be selected for data sufficiency type and heterogeneity. and redundancy. The number and location of sites depend on three key requirements: 1) must represent ALS data must cover the minimum site area the landscape variability of vegetation, topography (100-1000 ha) with point density necessary to have 10 and moisture conditions within each biome, 2) must vegetation vertical structure and height, the digital ter- a) Buoy 7100 b) [ Figure 18-5] 400 8 be located in areas with existing data, infrastructure, rain model (DTM) with less than 1 m vertical resolution Examples of buoy (dots) and or programs to guarantee quality control and future and uncertainty at the plot size (2-4 points per m2 de- 6 Radarsat-1 (line) trajectories 350 data augmentation, and 3) must include a combination pending on vegetation type). The ALS data may include km after Lindsay and Stern (2003). 4 Separation Distance Y (km) 300 of ground plots, Airborne Laser Scanning (ALS), and the point density data (LAS files) or only the DTM and Note the very similar 2 airborne or satellite L-band SAR imagery. DSM (< 1-3 m horizontal resolution depending on veg- displacement differences 250 Displacement Difference 0 etation type) if there are restrictions on disseminating (<0.5 km) between buoys and –1550 –1500 –1450 –1400 310 320 330 340 350 The main objective of pre-launch Cal/Val activities will the point density data. SAR tracking over a 40-day X (km) Days of 1997 10 be the development of algorithms and validation of al- period in (a,b), while a shear- c) Buoy 1902 d) gorithm performance to meet the science requirements Ground plots will be used to derive lidar-AGB models to ing event occurred in days 840 8 using airborne and satellite L-band radar that simulates convert the ALS vegetation height metrics to develop 367-380, which resulted in 6 820 NISAR observations. Whereas, the post-launch Cal/Val maps over the Cal/Val sites and quantify the uncertainty

large displacement differences km 4 in (c, d) that were not suitable Y (km) 800 activities are designed to potentially adjust and verify at the 1-ha map grid cells. The AGB maps will be used 2 the performance of the algorithms when NISAR data are for calibration and validation SAR algorithm and prod- for error tracking. Similar 780 acquired. ucts including the propagation of uncertainty through results to (a, b) will be gener- 0 ated with NISAR imagery. –740 –720 –700 –680 –660 310 350 360 370 380 390 400 all steps of algorithm development and implementation. X (km) Days of 1997 240 NISAR Science Users’ Handbook 241 NISAR Science Users’ Handbook

The pre-launch NISAR biomass Cal/Val activities will Several data sources are available to obtain Forest TABLE 18-3. BIOMASS CAL/VAL SITES WITH CONTEMPORARY FIELD MEASUREMENTS AND LIDAR DATA ACQUISITIONS. focus on developing the algorithmic model parameters Fractional Canopy Cover (FFCC) estimates to support Site Name Country Site Name Country Site Name Country with existing ground, ALS or SAR data. The calibration the generation of a calibration/validation data set. or validation will be performed by available time series With the objective for consistency in the approach, the Mondah Gabon Ordway - OSBS USA (Florida) Germany - Germany Kljuntharandt SAR data (airborne or ALOS PALSAR) and simulations most suitable data sets are best chosen from global- Mouila Gabon Tomé-Açu Brazil of soil moisture and vegetation phenology. Once algo- ly acquired high-resolution optical imaging sensors Germany - Germany Cantareira 1 rithms are developed and tested on historical SAR data, for which NASA has data-buy agreements (e.g., as Lope Gabon Brazil Traunstein

they will be either directly applied to NISAR observa- currently established for WorldView satellite data), or Mai Ndombe DRC Cantareira 2 Brazil Italy - Trentino Italy tions or adjusted for NISAR radiometric calibration and can be purchased in pairs in regions where disturbance Lowveld, Kruger South Africa Goiás Brazil Netherlands - Netherlands configurations during the post-launch Cal/Val activities. is known to be occurring and where such data have National Park Loobos The validation of NISAR biomass products will be been collected by satellite resources. Given the current Bahía Brazil Amani Nature Tanzania Poland - Bialowieza Poland performed on selected test sites distributed across the availability of these data types, it is expected that they Reserve (ANR) Massaranduba Brazil terrestrial biomes. will continue to be available during the NISAR mission Spain - Soria_i Spain Miombo Tanzania Rondonia Brazil calibration and validation time frame. Spain - Soria_ii Spain The use of historical field measurements, ALS, and Bia Conservation Ghana Tapajos Brazil Area and Dadieso Spain - Valsaincircle Spain SAR data relies on international collaboration. Similarly, Some flexibility exists in the combining of observation Forest Reserve Roraima Brazil validation of the biomass is performed in collaboration pairs from different sources. In particular, if a visual Spain - Valsairect Spain West Africa (Gola Sierra Leone Litchfield Savanna Australia with the Cal/Val programs of the NASA Global Ecosys- interpretation approach for reference data generation Rainforest National /Liberia Switzerland - Switzerland tem Dynamics Investigation Lidar (GEDI) and the ESA (Cohen et al., 1998) is employed. Using heterogeneous Park) Harvard Forest USA Laegeren - HARV (Massachusetts) BIOMASS missions, as well as through partnerships data should be avoided, but targets of opportunity Tumbarumba Australia Doi Inthanon Thailand with resource networks and field locations where for validation after a large natural disturbance event Smithsonian USA Great Western Australia Environmental (Maryland) Palangkaraya Indonesia biomass is measured and monitored. (e.g., fire, tornado) may have a good mix of viable Woodlands Research Center Peat Central reference data. - SERC Kalimantan Mulga BIOMASS CAL/VAL SITES Australia Mai Ndombe DRC Sabah Forestry Malaysia Alternatives to the optical classification and measure- Karawatha Australia Research Center Biomass Cal/Val sites with contemporary field mea- San Joachin USA Area ment of of FFCC change are in the use of alternative surements and lidar data acquisitions will be selected Great Western Australia SJER (California) sources of multi- or hyper-spectral optical, radar, and Woodlands Sarawak Malaysia from the following study sites with historical measure- Laurentides LIDAR data from space and airborne resources from Quebec, ments (Table 18-3): Litchfield Savanna Australia Wildlife Reserve Canada Mudumalai India which viable data sets to determine FFCC. For optical observations (multi-spectral, hyper-spectral, and/or li- InJune Australia Sycamore USA (Arizona) Xishuangbanna China 18.3.2 Forest Disturbance Creek - SYCA dar) cloudy areas are masked from the data pairs. Field Ft. Liard NWT, Canada Changbaishan China The forest disturbance requirement is to detect a reference data collected by National Forest Services, Great Smoky USA Ft. Providence NWT, Canada Mountains National (Tennessee) Gutianshan China 50 percent area loss of canopy cover, taken over a research groups, or commercial timber management Park - GRSM 1-hectare region. This entails the detection of a ½ entities can serve as ancillary sources to provide Ft. Simpson NWT, Canada Dinghushan China Niwot Ridge USA hectare reduction in canopy cover. While the establish- geographically localized validation data. For a global Bartlett USA (New Mountain Research (Colorado) Donlingshan China ment of the calibration and validation of the accuracy comprehensive calibration and validation approach, Experimental Hampshire) Station - NIWO Forest -BART Heishiding China requirement of the forest disturbance algorithm can algorithms that have been published in the literature Howland Forest USA (Maine) take on a number of different forms, the primary can be used to determine FPCC and changes can be Healy - HEAL USA (Alaska) Hainan China Soaproot Saddle USA one used by NISAR will be through the analysis of applied to image pairs, collected one year apart, from Delta Junction USA (Alaska) - SOAP (California) Badagongshan China high-resoultion (5 m or better) pairs of multi-spectral these potential data sources (Cho et al., 2012, Chubey - DEJU SPER USA (Colorado) Baotianman China optical data, collected one year apart, in regions where et al., 2006 et al., Clark et al., 2004, Falkowski et al., Lower Teakettle USA disturbance is know to have occurred and where such 2009, Immitzer et al., 2012, Ke et al., 2011; Lucas et - TEAK (California) England United Daxinganling China Newforest Kingdom data exist. al., 2008). Lenoir Landing USA - LENO (Alabama) Estonia - Rami Estonia

242 NISAR Science Users’ Handbook 243 NISAR Science Users’ Handbook

Algorithm calibration will be performed prior to the With annual disturbance rates varying with disturbance [ Figure 18-6] launch of NISAR, over the cal/val regions, using a pair type (fires likelihoods, infestations, legal and illegal Target biomes for place- of optical images, collected roughly one year apart. logging operations), and with the fraction of annually ment of forest distur- Ideally these image pairs will be obtained from the disturbed areas tending to be relatively small, stratified bance validation sites. same sensor under similar acquisition conditions. sampling with sample sizes sufficiently large to cover During this calibration phase, manual interpretation of most disturbance scenarios should be used. Forest the high-resolution optical imagery will be coupled with management agencies will be consulted in advance supervised classification for determining percent of to identify areas of disturbance to insure bracketing of FFCC change derived from these optical resources for disturbance events with optical data. the NISAR disturbance cal/val sites. These data will be combined with available Sentinel-1 time series and/or Disturbance will be validated on at least 1000 1- ha ALOS-2 and SAOCOM data to create thresholds for the resolution cells for any given area, where each reso- cumulative sum algorithm that is being used by NISAR lution cell is fully mapped FFCC change values. NISAR Tropical and Subtropical Moist Broadleaf Flooded Grasslands and Savannas for detecting disturbance. disturbance detection results will be evaluated against Forests Montane Grasslands and Shrublands this full sample in order to capture errors of omission Tropical and Subtropical Dry Broadleaf Forests Tundra To reduce errors in the FFCC change detection of the (false negative) and commission (false positive). This Tropical and Subtropical Coniferous Forests Mediterranean Forests, Woodlands and pairwise analysis of NISAR datasets, only exact repeat approach follows guidelines for sample designs, which Temperate Broadleaf and Mixed Forests Scrub Temperate Coniferous Forests Deserts and Xeric Shrublands orbits and view angles are considered. Image pairs will have been established and discussed in the literature Boreal Forests/Taiga Mangroves be accurately co-registered, geocoded and geomet- (Olofsson et. al, 2014, Stehman 2005, van Oort, 2006, Tropical and Subtropical Grasslands Lakes rically matched to the NISAR time-series data stacks. Woodcock et al., 2001). Figure 18-6 shows two sites Savannas and Shrublands Rock and Ice Because of modern orbital control and the availability per continent/biome combination within the shown Temperate Grasslands, Savannas and Shrublands of DEMs for geocoding, co-registration errors are areas are proposed which results in a total of 96 vali- negligible for the purpose of establishing 1-ha FFCC dation sites. Map after Olson et al (2001). estimates. For a statistically viable validation approach, the reference image pairs need to be distributed in all FOREST DISTURBANCE CAL/VAL SITES observed biomes and image subsets of sufficient size Forest disturbance Cal/Val sites will be selected based need to be chosen to extract enough validation samples [ Figure 18-7] the availability of alternative measurements of ongoing to detect all sources of error (see below). To obtain disturbance such as from cloud-free optical imagery GFOI research and devel- 1000 1 ha samples from a high-resolution image pair bounding disturbance events or from information pro- opment study sites where requires approximately 3.2 x 3.2 km image subsets. a host of field and satellite vided by forest management agencies. The sites will Given potential cloud cover pixel elimination, 4 x 4 km data sets are being used be distributed globally and in every forest biome. subsets will be obtained within which 1 ha samples to study forest disturbance can be placed. With respect to biome sampling, the (http://www.gfoi.org/rd/ Areas known to undergo regular forest disturbance are stratification after the WWF biomes classification will be study-sites/). timber management sites. For example, large tracts in used (Olson et al., 2001). Of the 14 global biomes, eight the Southeastern United States have forest regrowth are critical for disturbance validation (Figure 18-6). Two cycles of 20 years where a stand replacement rate of validation sites will be chosen in each of the biomes 5 percent per year for forest land under timber man- with forest cover in each of the continents of North- agement. Forest management plans will be obtained and South America, Africa, Asia, Europe, Australia with for the year after NISAR launch from collaborators in forest cover (eliminating Antartica and Tundra regions). the USDA Forest Service and timer industry sector to This would bring the total number of validation sites to determine sites of forest disturbance a priori. Interna- 6 continents x 8 biomes x 2 sites each = 96 validation sites. 244 NISAR Science Users’ Handbook 245 NISAR Science Users’ Handbook

tional partnerships are established via collaboration in faces will depend on incidence angle and account for in the Co-Pol (HH) channel, which results from double in situ surveys with a Real Time Kinematic-GPS (RTK). the GEO Global Forest Observing Initative (GFOI) which the size of the water bodies. However, these threshold bounce reflections that occur when the radar-illumi- To capture inundation extent at the time of NISAR data operates a network of study regions of deforestation must remain biome independent. Inundation extent will nated area contains vegetation that is vertically emer- acquisition, gauges (e.g., pressure transducer) must be and forest degradation hotspots. Figure 18-7 illustrates not include snow-covered or frozen water. Desert areas gent from standing water. If the vegetation is small in recording water level continuously. where GFOI has established these study regions and will be excluded from analysis by the wetlands mask. stature and/or herbaceous, double bounce reflections constitutes a network of forest disturbance hotspots may be reduced, leading to specular reflection over 3. High-resolution optical data. This method utilizes and thus a set of first order targets of opportunities for Error rates for this requirement will only be evaluated the open water (i.e., low backscatter). As inundated spaceborne or airborne remote sensing instruments. post-launch disturbance monitoring. within a wetland mask. The initial wetland mask will vegetation transitions to non-inundated vegetation, HH For example, it was employed in 2012 using world- be determined prior to launch from ancillary sources radar brightness reduces to the level of the imaged view-2 multiband optical data. Malinowski et al., 2015 A resource to use for locating sites for fire based distur- of information on wetland extent but may be modified forest or marsh volume. L-band radar remote sensing found overall accuracy greater than 80 percent. The ad- bance is the Fire Information for Resource Management after launch if additional information warrants updates. is known to be a reliable tool for detection of inundated vantage of this method is that it is possible to efficiently System (FIRMS) (https://earthdata.nasa.gov/earth-ob- It should encompass an area greater than that which vegetation, and may overperform other remote sensing map large areas with good accuracy and low cost. The servation-data/near-real-time/firms). FIRMS distributes typically experience inundation but will exclude all measurement available for validation. There are four disadvantages are the reduced accuracy for detecting Near Real-Time (NRT) active fire data within 3 hours of urban areas, deserts and permanent open water potential methods that could be utilized for validation, inundation in areas with dense vegetation cover (>80%), satellite overpass from both the Moderate Resolution surfaces. The wetland mask will have a seasonal com- some shown in Figure 18-8. The combination of these and the non-simultaneous timing of data acquistion Imaging Spectroradiometer (MODIS) and the Visible ponent such that inundation is not evaluated during methodologies will be evaluated and selected prior to with NISAR. The latter effect can be alleviated with Infrared Imaging Radiometer Suite (VIIRS). In the U.S., frozen conditions or freeze/thaw transition periods. launch. The potential methods are: the extremely high resolution (cm scale) imagery from a resource for the estimation of burn severity is the The wetland mask will indicate those areas that are UAS-type aircraft overflights coordinated with the NISAR Burn Area and Severity mapping service by the USGS agricultural. 1. Ground transects. This method is the most acquisition times. Combined RGB plus IR cameras on (https://www.usgs.gov/apps/landcarbon/categories/ accurate and provides additional information such as UAS can be used to identify both vegetation extent (from burn-area/download/). This resource can be used to All Cal/Val sites must be located within the wetland inundation depth and vegetation characteristics. The RGB) and many cases of inundation (from IR) overlaid on support the targeting of high-resolution optical image mask. The Cal/Val sites must represent the varying disadvantages results from logistical considerations a high resolution digital surface model (DSM) and digital acquistion to estimate FFCC loss from fire disturbances. conditions present in different biomes, ranging from which will bias site selection and the likely provide in- terrain model (DTM) (though significant vegetation will boreal to temperate to tropical biomes with distinct complete sampling of the wetland extent. Ground tran- diminish the IR signature of water). This method is best 18.3.3 Inundation Area vegetation differences. The validation of open water sects performed by research partners could facilitate suited to boreal ecosystems where there is less obscu- is impacted by wind conditions, freeze/thaw state, the acquisition of validation ground transects. Time ration by woody vegetation, and for open water areas Inundation extent within wetland areas will be validated and the “radar darkness” of the surrounding environ- continuous measurement devices such as pressure where cloud free images can be obtained. for two conditions: near-shore open water (within ment, while the validation of inundated vegetation is transducers and soil moisture probes can be deployed 100 m of a shoreline), and standing water with emer- impacted by the structure and density of the emergent along transects traversing wetlands. This method is 4. High resolution quad-pol SAR data. While the gent vegetation. The requirement states that measure- vegetation. The distribution of inundation Cal/Val sites best suited to areas where remote observations are inundation extent algorithm for NISAR utilizes dual pol ments should be validated at 1-ha resolution. If a should sample the cross-track NISAR imaging swath expected to be less robust due to extensive canopy HH and HV data, validation could be done with enhanced 1-ha pixel is inundated, the predominate state will be due to expected sensitivity to incidence angle and the cover, such as tropical forests. quad-pol L-band or P-band SAR data (such as available validated. Inundation extent will only be validated when noise properties of the SAR data. on the airborne NASA UAVSAR platform) by polarimetric the water and surrounding landscape are not frozen or 2. 3D inundation extent model. This method relies decomposition. Polarimetric decomposition evaluates snow covered. The pre-launch and post-launch calibration of the on the knowledge of water level through time, and the contributions of the various scattering mechanisms algorithm thresholds, and the post-launch validation accurate knowledge of the wetlands digital elevation and can therefore be used to isolate the double bounce The measurement of near-shore open water extent by of the science requirement should be cost-effective. model (DEM). The inundation extent is determined by effect that occurs in inundated areas. However, the NISAR can often be validated with data from optical The planned launch of the NASA SWOT mission nearly numerically flooding the DEM given measurements methodology itself needs to be validated pre-launch. sensors, limited only by cloud cover. Open water sur- coincide with NISAR’s , thus coordination of cal/val recorded by an in situ water level gauge. This is the The advantage is that this approach provides wide area faces generally exhibit low radar backscatter similar to activities is recomended for the mutual benefit of these most efficient and reliable of all methods, but is limited mapping and on-demand timing with NISAR acquisi- bare ground and beaches. In larger open water sufaces, projects. by the sparse availability of DEMs. The latter can be tions. This method itself must be validated, such as by wind- induce roughening of the water increasing radar obtained from ALS during dry periods (lidar returns will methods described above. However, once validated, this brightness, especially at smaller incidence angles. Thus, The measurement of inundated vegetation by NISAR have no reflection where there is standing water, and spatially large product can be used to validate the large the selection of thresholds to identify open water sur- is enabled by the high-intensity backscatter observed therefore also has value during inundation periods) or scale NISAR products. 246 NISAR Science Users’ Handbook 247 NISAR Science Users’ Handbook

66 W 60 W TABLE 18-4. WETLAND INUNDATION NOMINAL CAL/VAL SITES 0 S [ Figure 18-8] Upland Site Vegetation Type Logistics and Methodologies Validation methods for inun- Braga Creek and Lagoon dation. a) measurements of Bonanza Creek/ Boreal wetland (marshes, Ground transects, high resolution tussocks, some forested) optical, and UAS data. inundation state along ground Yukon flats, Alaska (ABoVE site, possible SWOT site) transect of Napo River floodplain Braga Boreal wetland with classification based on Path Creek Scotty Creek and nearby Ground transects, high resolution Bordering sites, Canada optical, and UAS data. polarimetric decomposition Lagoon (ABoVE site, Ducks Unlimited) (Chapman et al., 2014). 4 S b) video survey of JERS-1 SAR Florida Everglades Freshwater marsh plus man- Pressure transducers combined a) b) classification, validation points grove area (shrubs and trees) with DEM. (possible SWOT site) indicated (Hess et al., 2002). N Louisiana Delta Fresh and brakish water, Pressure transducers combined c) WorldView-2 classification of cyprus, willows, marshes with DEM. (Possible SWOT site) inundation extent (Malinowski Pacaya-Samiria, Peru Tropical wetland (palms, etc.) Pressure transducer and ground et al, 2015). d) GoPro Cameras transects, lidar would be helpful onboard Solo Quadcopter 0 0.25 0.5 1 km (Schill, S, Personal Pantanal, Brazil/ UAS imagery. Measure DEM with Paraguay/Bolivia RTK in conjunction with pressure communication, 2016). transducers

Ogooue River, Gabon Freshwater marsh, tropical UAS imagery wetland palms, papyrus

Flooded Areas Non-Flooded Areas Bhitarkanika, India Mangrove ISRO monitoring site c) d) Chilika, India Lake near agriculture ISRO monitoring site

Nalsarovar, India Inland wetland ISRO monitoring site The ground-based in situ observations will be INUNDATION AREA CAL/VAL SITES Sud, South Sudan Marsh In coordination with current employed in validation of active crop area. Airborne studies Cal/Val sites for inundation extent will be selected from and spaceborne sensors will also be employed where Carpinteria Salt Marsh Marshes Pressure transducer combined sites listed in Table 18-4. Methodology at each site will possible to extend coverage from the plot level. Reserve with RTK DEM depend on vegetation and likely cloud cover. Establishing and maintaining a robust and globally distributed set of consistent in situ land cover data will Magdalena River, Mangroves RTK plus pressure transducer Colombia 18.3.4 Cropland area be essential to the success of the agricultural portion of the NISAR Cal/Val program. Recognizing the complex- Mamirauá Sustainable Tropical wetland In coordination with local experts Development Reserve Similar to the disturbance cal/val effort described ity of global agricultural systems, and the challenges above, calibration and calidation for NISAR’s crop- involved in acquiring these data, the NISAR project land area requirement will be principally based on will attempt to achieve its objective through partner- Another important consideration for implementation of tinent outside Antarctica, field size, crop type, climactic high-resolution optically based image classification ships with complementary operational and research the NISAR agricultural Cal/Val program (which will uti- conditions, and cropping practices can all vary greatly and informed through partners in the Group on programs around the globe. In addition to basic crop lize data from a variety of organizations worldwide) is between locations. NISAR will provide products around Earth Observations’ (GEO) Joint Experiment for Crop type measurements, additional information on cropping establishing global consistency in the correlative data. the world; therefore, validation data should be repre- Assessment and Monitoring (JECAM) or similar such practices are welcome, as this information can be use- While agriculture plays an important role on every con- sentative of a wide range of agricultural variation, and collaborations that are formed in NISAR’s pre-launch ful in refining the algorithms in use and understanding will require cooperation from a range of groups both Cal/Val period. NISAR’s sensitivity to other agricultural measurements. 248 NISAR Science Users’ Handbook 249 NISAR Science Users’ Handbook

within the U.S. and internationally. To assist in estab- (http://www.jecam.org/JECAM_Guidelines_for_Field_ • Planting and harvesting information provided parking lots, housing, and major power lines should lishing consistency between validation sites, a series of Data_Collection_v1_0.pdf). While the NISAR mission to within a week per field also be noted. guidelines have been developed to set expectations for is most interested in data for the major crops barley, • Cropping practices reported per field (i.e., Please note that exact survey methods are still being potential NISAR Cal/Val agricultural partners. cassava, groundnut, maize, millet, potato, rapeseed, tillage, residues, irrigation, etc. – see legend) rice, rye, sorghum, soybean, sugarbeet, sugarcane, developed, and that further specifications will likely be • Information reported once per season For consistency, the NISAR agricultural Cal/Val program sunflower, and wheat, any type of annual crop type added in order to ensure high-quality, consistent data is using the definition of annual cropland from a remote has value when making a crop/non-crop map. Minor • Tier II between the different partners. Information beyond the sensing perspective as defined by the Group on Earth crops should not be ignored or excluded from data • Active crop type data at the most detailed location of the crop/non-crop areas such as planting Observations’ (GEO) Joint Experiment for Crop Assess- collection, although they can be grouped into more level available (according to legend groupings) and harvest times and cropping practices might be ment and Monitoring (JECAM). The definition they use general categories (i.e. vegetables, pulses, etc.). If reported once per season obtained through farm-to-farm surveys or through col- is as follows, and will be used throughout this docu- fields are being actively managed off season, such laborations with agricultural collectives that might have The in situ data collection should consist of a large ment to define agricultural land: as planted with a green manure crop, this should be knowledge of this type of information. enough and spread out enough sample to adequately reported as an additional season of data. Similar- characterize the larger agricultural region. It is sug- The annual cropland from a remote sensing perspective ly, non-crop information is also vital to making an The in situ data would be provided once per growing gested that a “windshield survey” be conducted along is a piece of land of minimum 0.25 ha (min. width of accurate crop/non-crop map, so it is requested that season at planned intervals for at least two years before the main roads from a motorized vehicle, allowing the 30 m) that is sowed/planted and harvestable at least Cal/Val partners also provide information about the and after launch which, based on the current launch data collector to easily and rapidly capture basic crop once within the 12 months after the sowing/planting surrounding nonagricultural land covers. date, would be 2019-2023. information of all visible fields, and capture long tran- date. The annual cropland produces an herbaceous sects of data across the region under investigation. It is cover and is sometimes combined with some tree or Two types of in situ data will be acquired at NISAR Two methods of data collection will be utilized, poly- also recommended to complement the long transects woody vegetation.*° agricultural Cal/Val sites. Tier I sites will provide a gon-based or point-based, as described below. with regular additional transects throughout the study more comprehensive set of data about crop type, In order to categorize crop type consistently, this Cal/ area via secondary roads and tracks in order to reduce growing season, and cropping practices, whereas POLYGON METHOD Val program will be using the general legend developed the spatial bias brought about by roadside sampling. Tier II sites will only provide crop type information. by the JECAM project to define crop type. It follows a Several long transects running in various directions The first of the two preferred methods of data collection Tier I sites may receive priority over Tier II sites hierarchical grouping and has been adapted from the ensure coverage of the entire area, while the secondary is to mark field boundary polygons on high resolution during Cal/Val activities by the NISAR project. The tier Indicative Crop Classification (ICC) developed by the transects provide complimentary data that reduces the multispectral cloud free imagery prior to gathering designations are used to formalize what is expected Food and Agriculture Organization of the United Nations bias in the Cal/Val data set. field data, and then once in the field confirm that there from each site in terms of data they will provide. A (FAO) for use in agricultural censuses. This legend have been no major changes to field dimensions and summary of expectations for each tier are as follows: delineates only temporary crops falling under the afore- At least 20 fields with the minimum size described record crop type and any other crop information (such mentioned definition of agricultural land, with perennial • Tier I below should be recorded for each of the primary as planting and harvest dates, and cropping practices) crops listed at the end along with a few major nonag- • Crop type data at the most detailed level crops in the region, whether or not they are one of the being collected as attributes to each field polygon. ricultural land covers. See Section IIC of the JECAM available (according to legend groupings) crops NISAR is most interested in. The fields should This can be done with a variety of mobile device apps Guidelines for Field Data Collection for the full legend be spread over an area of at least 100 km2. A target and software, in order to mitigate potential error when sampling density would be about one observation / 5 transferring paper notes to a GIS system. This data km2, or within the range of one observation for every could be acquired during windshield survey described 1-10 km2, depending on the complexity of the cropping above. * The herbaceous vegetation expressed as fcover (fraction of soil background covered by the living vegetation) is expected systems and diversity of crop types. In addition to the to reach at least 30% while the tree or woody (height >2m) cover should typically not exceed an fcover of 20%. crops, 20 samples of each of the major nonagricul- POINT METHOD tural land covers should also be recorded, with each ° There are 3 known exceptions to this definition. The first concerns the sugarcane plantation and cassava crop which The second of the two preferred methods of data are included in the cropland class, although they have a longer vegetation cycle and are not yearly planted. Second, taken sample following the same minimum size restrictions collection is based on recording the GPS location of a individually, small plots such as legumes do not meet the minimum size criteria of the cropland definition. However, when as described above for crops. The locations of signifi- considered as a continuous heterogeneous field, they should be included in the cropland. The third case is the greenhouse specific in situ point and taking photos in each of the cant infrastructure, such as barns, processing facilities, crops that cannot be monitored by remote sensing and are thus excluded from the definition. cardinal directions (N, E, S, W) as a way of consistently 250 NISAR Science Users’ Handbook 251 NISAR Science Users’ Handbook

recording surrounding crops between different field CROPLAND AREA CAL/VAL SITES locations. Photos should be annotated with the in situ 19 APPENDIX G Cal/Val sites for validating active crop area will be location, crop type information, and any other cropping selected from among the 50 current JECAM study sites practice information being collected (tillage, irrigation, RADAR INSTRUMENT MODES shown in Figure 18-9. etc.) for Tier I sites. This data could be acquired during windshield survey described above. TABLE 19-1. OVERVIEW OF INSTRUMENT MODES BASED ON TARGET TYPES FOR NISAR L-SAR AND S-SAR INSTRUMENTS. (SP refers to single-polarization, DP is dual-polarization, and QP is quad-polarization, CP is compact-polarization)

Science Performance [ Figure 18-9] SWATH PRIMARY SCIENCE FREQ. POLARIZATION START JECAM sites BW PRF PW SWATH TARGET BAND LOOK (http://www.jecam.org/?/ ANGLE interactive-map). (MHZ) (HZ) (mSEC) (KM) (DEG)

Background Land L DP HH/HV 20+5 1650 25 242 30

Background Land Soil Moisture L QQ 20+5 1650 25 242 30

Background Land Soil Moisture L QQ 20+5 1650 20 242 30 High Power

Land Ice L SP HH 80 1650 40 121 30

Land Ice Low Res L SP HH 40+5 1650 45 242 30

Low Data Rate Study Mode L SP HH 20+5 1650 25 242 30 Single Pol

Sea Ice Dynamics L SP VV 5 1600 25 242 30

Open Ocean L QD HH/VV 5+5 1650 20 242 30

India Land Characterization L DP VV/VH 20+5 1650 25 242 30

Urban Areas, Himalayas L DP HH/HV 40+5 1650 45 242 30

Urban Areas, Himalayas SM L QQ 40+5 1650 45 242 30

Urban Areas, Himalayas SM L QQ 40+5 1650 40 242 30 High Power

US Agriculture, India L QP 40+5 1600 45 242 30 Agriculture HH/HV/VH/VV

US Agriculture, India L QP 20+5 1600 45 242 30 Agriculture Low Res HH/HV/VH/VV

Experimental CP mode L CP RH/RV 20+20 1650 40 242 30

Experimental QQ mode L QQ 20+20 1650 20 242 30

Experimental SP mode L SP HH 80 1650 20 242 30

ISRO Ice/sea-ice L DP VV/VH 5 1650 25 242 30 252 NISAR Science Users’ Handbook 253 NISAR Science Users’ Handbook

Science Performance Science Performance

SWATH SWATH PRIMARY SCIENCE FREQ. POLARIZATION START PRIMARY SCIENCE FREQ. POLARIZATION START BW PRF PW SWATH BW PRF PW SWATH TARGET BAND LOOK TARGET BAND LOOK ANGLE ANGLE (MHZ) (HZ) (mSEC) (KM) (DEG) (MHZ) (HZ) (mSEC) (KM) (DEG)

ISRO Ice/Sea-ice - Alternate L QD HH/VV 5 1650 25 242 30 L: QP 40+5 1550 45 242 30 HH/HV/VH/VV Solid Earth/Ice/Veg/Coast/ S Quasi-Quad 37.5 2200 10+10 244 30 Coastal - Land L+S Bathymetry S: DP HH/HV 37.5 3100 10 244 30

Ecosystem/Coastal Ocean/ S DP HH/HV 10 2200 25 244 30 L: QP 20+5 1550 45 242 30 Cryosphere HH/HV/VH/VV Coastal - X L+S

Agriculture/Sea Ice S CP RH/RV 25 2200 25 244 30 S: CP HH/HV 25 3100 10 244 30

Glacial Ice High Res S CP RH/RV 37.5 2200 25 244 30 L: QP 20+5 1550 45 242 30 HH/HV/VH/VV New Mode S DP HH/HV 37.5 2200 25 244 30 Coastal - X L+S S: DP HH/HV 3100 10 244 30 Deformation S SP HH (or SP VV) 25 2200 25 244 30 DP VV/VH 5 1910 25 242 30 L+S Deformation Max Res S SP HH (or SP VV) 75 2200 25 244 30 ISRO Ice/Sea-Ice CP RH/RV 25 25 244 30 DP HH/HV 20+5 1910 25 242 30 Systematic Coverage L+S DP VV/VH 5 1910 25 242 30 CP RH/RV 25 25 244 30 ISRO Ice/Sea-Ice - L+S Joint Alternate DP VV/VH 10 25 244 30 Systematic Coverage DP HH/HV 20+5 1910 25 242 30 & Deformation L+S DP HH/HV 37.5 25 244 30

Coastal Mudbank DP HH/HV 20+5 1910 25 242 30 L+S (Wet Soil) SP HH (or SP VV) 25 25 244 30

SP VV 5 1910 25 242 30 Ocean L+S DP VV/VH 10 25 244 30

L: DP VV/VH 20+5 1910 25 242 30 Sea Ice Type L+S S: CP RH/RV 25 25 244 30

L: DP HH/HV 40+5 1910 45 242 30 Glacial Ice Himalayas L+S S: CP/RH/RV 37.5 25 244 30

L: DP HH/HV 40+5 1910 45 242 30 High-Res Deformation L+S (Disaster/Urgent Response) S: SP HH (or 75 25 244 30 SP VV)

L: QP HH/HV/ 40+5 1550 45 242 30 VH/VV India Agriculture L+S S: CP RH/RV 25 3100 10 244 30 NISAR Science Users’ Handbook 255 NISAR Science Users’ Handbook

20 APPENDIX H SCIENCE TARGET MAPS

Table 3-1 summarized the Level 1 requirements that For example, in the cryosphere the requirements spec- NISAR must meet. The Level 1 requirements lead to ify coverage of ice sheets in Greenland and Antarctica, Level 2 measurement accuracy, sampling and coverage as well as polar sea ice, but do not specify all regions requirements for each of the scientific disciplines. The with mountain glaciers. Solid Earth deformation areas coverage requirements are globally distributed, but the are specified in terms of fast-moving plate boundaries areas over which requirements must be met are disci- and selected areas with transient deformation. The fig- pline-specific and are codified by the project in a set of ures below summarize the targets for each discipline. science target maps comprising geographical polygons.

[ Figure 20-1] Solid Earth discipline target map

SOLID EARTH TARGETS

Background Landslides Aquifers Permafrost Secular Volcanoes Oil and Land Deformation Gas

[ Figure 20-2] Ecosystems discipline target map

ECOSYSTEMS TARGETS

Biomass Disturbance Wetlands Agriculture (World) Agriculture (USA) 256 NISAR Science Users’ Handbook 257 NISAR Science Users’ Handbook

21 APPENDIX I [ Figure 20-3] Cryosphere discipline DATA PRODUCT LAYERS target map

TABLE 21-1. L1 SLC DATA LAYER DESCRIPTION

Number Name Data Spacing Description of Layers Type

Primary Data Layers

Complex back- Number of CInt16 Full Focused SLC image. scatter polarizations resolution All channels are regis- (primary mode) tered

Complex back- Number of CInt16 Full Focused SLC image. scatter (aux 5 polarizations resolution All channels are regis- CRYOSPHERE TARGETS MHz mode) tered

Q1Sea Ice Q2 Sea Ice Q3 Sea Ice Q4 Sea Ice Priority Glaciers Land Ice Ice Data quality 1 Byte Full Byte layer with flags for resolution various channels

Secondary Layers

Latitude 4 Float64 1 km Latitude polynomial grid

[ Figure 20-4] Longitude 4 Float64 1 km Longitude polynomial Applications discipline grid target map Incidence angle 4 Float32 1 km Incidence angle grid

Azimuth angle 4 Float32 1 km Azimuth angle grid

Elevation angle 4 Float32 1 km Elevation angle grid

Sigma0 Number of Float32 5 km az x 1 km rg LUT to convert DN to polarizations Sigma0 assuming constant ellipsoid height

Gamma0 Number of Float32 5 km az x 1 km rg LUT to convert DN to polarizations Gamma0 assuming constant ellipsoid height

APPLICATIONS TARGETS Thermal noise Number of Float32 5 km az x 1 km rg LUT for noise correction Urban Areas US Base Map Nuclear Power Plants Dams polarizations 258 NISAR Science Users’ Handbook 259 NISAR Science Users’ Handbook

TABLE 21-2. L1 MLD DATA LAYER DESCRIPTION TABLE 21-3. IFG DATA LAYER DESCRIPTION

Number Data Number Name Spacing Description Name Data Spacing Description of Layers Type of Layers Type

Primary Data Layers Primary Data Layers

Backscatter Number of Int16 25 m Focused SLC image. Complex Number of CInt16 25 m Focused SLC image. amplitude polarizations All channels are regis- Interferogram polarizations All channels are regis- (primary mode tered (primary mode tered. only) only)

Data quality 1 Byte 25 m Byte layer with flags for Mask 1 Byte 25 m Byte layer with flags for various channels various channels.

Secondary Layers Amplitude 2*Number of Int16 25 m co-polarizations

Latitude 4 Float64 1 km Latitude polynomial grid Secondary Layers

Longitude 4 Float64 1 km Longitude polynomial Latitude 4 Float64 1 km Latitude polynomial grid grid Longitude 4 Float64 1 km Longitude polynomial Ground range 1 Float64 1 km Ground range to slant grid to slant range range grid Incidence angle 4 Float32 1 km Incidence angle grid Incidence angle 4 Float32 1 km Incidence angle grid Azimuth angle 4 Float32 1 km Azimuth angle grid Azimuth angle 4 Float32 1 km Azimuth angle grid Elevation angle 4 Float32 1 km Elevation angle grid Elevation angle 4 Float32 1 km Elevation angle grid Baseline 4 Float32 1 km Parallel baseline grid Sigma0 Number of Float32 5 km az x 1 km rg LUT to convert DN to parallel polarizations Sigma0 assuming con- component stant ellipsoid height Baseline 4 Float32 1 km Perpendicular baseline Gamma0 Number of Float32 5 km az x 1 km rg LUT to convert DN to perpendicular grid polarizations Gamma0 assuming con- component stant ellipsoid height Range offsets 4 Float32 1 km Range offset grid Thermal noise Number of Float32 5 km az x 1 km rg LUT for noise correction polarizations Azimuth offsets 4 Float32 1 km Azimuth offset grid

LUT 2*Number of Float32 5 km az x 1 km rg To translate amplitude co-polarizations DN layers to backscatter 260 NISAR Science Users’ Handbook 261 NISAR Science Users’ Handbook

TABLE 21-4. L1 UNW DATA LAYER DESCRIPTION TABLE 21-5. L1 COV DATA LAYER DESCRIPTION

Number Data Number Data Name Spacing Description Name Spacing Description of Layers Type of Layers Type

Primary Data Layers Primary Data Layers

Unwrapped phase Number of CInt16 25 m Unwrapped phase in Complex 3 (Dual pol) CInt16 25 m Complex covariance matrix elements (primary mode polarizations radians correlation 6 (Quad pol) only) Mask 1 Byte 25 m Byte layer with flags for various channels Coherence Number of Byte 25 m Coherence range 0 – 1 (e.g., data quality and shadow-layover) polarizations Secondary Layers

Mask 1 Byte 25 m Byte layer with flags for various channels Latitude 4 Float64 1 km Latitude polynomial grid

Connected 1 Byte 25 m Connected components Longitude 4 Float64 1 km Longitude polynomial grid components flag for each pixel Incidence angle 4 Float32 1 km Incidence angle grid Amplitude 2*Number of Int16 1 km This is needed for full co- co-polarizations variance of interferograms Azimuth angle 4 Float32 1 km Azimuth angle grid

Secondary Layers Elevation angle 4 Float32 1 km Elevation angle grid

Latitude 4 Float64 1 km Latitude polynomial grid LUT 3 (Dual pol) Float32 5 km az x 1 km rg LUT to convert Beta0 to Sigma0 and Gam- 6 (Quad pol) ma0 assuming constant ellipsoid height Longitude 4 Float64 1 km Longitude polynomial grid

Incidence angle grid Incidence angle 4 Float32 1 km TABLE 21-6. L2 GSLC DATA LAYER DESCRIPTION

Azimuth angle 4 Float32 1 km Azimuth angle grid Number Data Name Spacing Description of Layers Type Elevation angle 4 Float32 1 km Elevation angle grid Primary Data Layers Baseline parallel 4 Float32 1 km Parallel baseline grid component Complex backscatter Number of CInt16 Full resolution Focused SLC image. All channels are (primary mode only) polarizations registered Baseline 4 Float32 1 km Perpendicular baseline perpendicular grid Mask 1 Byte Full resolution Byte layer with flags for various channels component Secondary Layers Range offsets 4 Float32 1 km Range offset grid Azimuth time 4 Float64 1 km Azimuth time polynomial grid Azimuth offsets 4 Float32 1 km Azimuth offset grid Slant range 4 Float64 1 km Slant range polynomial grid Solid Earth tides 4 Float32 10 km Solid Earth tide Incidence angle 4 Float32 1 km Incidence angle grid Tropospheric 40 Float32 25 km Dry delay estimated from dry delay ECMWF Azimuth angle 4 Float32 1 km Azimuth angle grid

Tropospheric 40 Float32 25 km Wet delay estimated from Elevation angle 4 Float32 1 km Elevation angle grid wet delay ECMWF Sigma0 Number of Float32 5 km az x 1 km rg LUT to convert DN to Sigma0 assuming Ionosphere phase 1 Float32 1 km Ionosphere phase screen polarizations constant ellipsoid height screen estimated from split spectrum method Gamma0 Number of Float32 5 km az x 1 km rg LUT to convert DN to Gamma0 assuming polarizations constant ellipsoid height LUT 2*Number of Float32 5 km az x To translate amplitude DN co-polarizations 1 km rg layers to backscatter Thermal noise Number of Float32 5 km az x 1 km rg LUT for noise correction polarizations

262 NISAR Science Users’ Handbook 263 NISAR Science Users’ Handbook

TABLE 21-7. L2 GUNW DATA LAYER DESCRIPTION. TABLE 21-8. L2 GCOV DATA LAYER DESCRIPTION.

Number Data Number Name Spacing Description Name Data Spacing Description of Layers Type of Layers Type

Primary Data Layers Primary Data Layers

Unwrapped phase Number of Float32 25 m Unwrapped phase in radians Complex 3 (Dual pol) CInt16 25 m Complex covariance matrix (primary mode only) polarizations correlation 6 (Quad pol) elements

Coherence Number of Byte 25 m Coherence range 0 – 1 Mask 1 Byte 25 m Byte layer with flags for polarizations various channels (e.g., data 1 quality and shadow-layover) Mask 1 Byte 25 m Byte layer with flags for various channels (e.g., data Projection Float32 25 m Projection angle grid quality and shadow-layover) angle

Connected 1 Byte 25 m Connected components flag Secondary Layers components for each pixel Azimuth time 4 Float64 1 km Azimuth time polynomial grid Amplitude 2*Number of Int16 25 m This is needed for full covari- co-polarizations ance of interferograms Slant range 4 Float64 1 km Slant range polynomial grid

Secondary Layers Incidence angle 4 Float32 1 km Incidence angle grid

Azimuth time 4 Float64 1 km Azimuth time polynomial grid Azimuth angle 4 Float32 1 km Azimuth angle grid

Slant range 4 Float64 1 km Slant range polynomial grid Elevation angle 4 Float32 1 km Elevation angle grid

Incidence angle 4 Float32 1 km Incidence angle grid

Azimuth angle 4 Float32 1 km Azimuth angle grid

Elevation angle 4 Float32 1 km Elevation angle grid

Baseline parallel 4 Float32 1 km Parallel baseline grid component

Baseline perpendicu- 4 Float32 1 km Perpendicular baseline grid lar component

Range offsets 4 Float32 1 km Range offset grid

Azimuth offsets 4 Float32 1 km Azimuth offset grid

Solid Earth tides 4 Float32 10 km Solid Earth tide

Tropospheric dry 40 Float32 25 km Dry delay estimated from delay ECMWF

Tropospheric wet 40 Float32 25 km Wet delay estimated from delay ECMWF

Ionosphere phase 1 Float32 1 km Ionosphere phase screen screen estimated using split spectrum method

LUT 2*Number of Float32 5 km az x To translate amplitude DN co-polarizations 1 km rg layers to backscatter