NASA Community Snow Meeting at CU Boulder, Colorado, August 2024

Getting The Snow Community Ready for The NASA-ISRO SAR Mission Shadi Oveisgharan, NASA L-Band InSAR SWE Working Group, NISAR Project Team Jet Propulsion Laboratory California Institute of Technology 2024 NASA Snow Community Meeting Boulder, CO August 14, 2024 Copyright 2023 California Institute of Technology. Government sponsorship acknowledged. SWE Retrieval Using Zero baseline repeat-pass interferometry •2017 Decadal Survey calls for Snow Water Equivalent (SWE) measurement with 3cm accuracy, using active sensor (radar) as part of the DS Earth Explorer mission category. •Leinss et al, TGRS 2015 (First Experimental demo) : Tower based repeat-pass interferometry at X-band •successful with very fast (every 10 minutes) revisit time •Ruiz et al, TGRS 2022: Multi channel (1-18 GHz) SodSAR tower radar in northern Finland. • L-, SC and X-bands over open and clear area. •all possible pairs with 12 days temporal baseline (NISAR replica) •Provides robustness against decorrelation and can solve for lost phase cycles on high frequency bands. Guneriussen et al., TGRS 2001 SWE from scale (mm) SWE from scale (mm)SWE from scale (mm) RMSE (mm) L - 6.63 S - 12.63 C - 21.75 L+S 4.21 L+C 4.42 All 3.97 •SnowEX campaign was focused on UAVSAR repeat-pass interferometry. •winter 2020: Biweekly spread over different US sites (6 sorties over 13 sites, across 5 states) •winter 2021: Weekly with fewer regional sites and more frequent temporal sampling (10 sorties over 6 sites, across 4 states) •Local experienced field teams deployed on the date of each over pass (Source of validation) •Airborne LIDAR over subdomain at a subset sites (Great source for validation) •coordinated with Sentinel-1 team for 6 days revisit for some of the SnowEx sites Repeat-pass Interferometry; SnowEx UAVSAR Time Series •Eli Deeb, HP Marshall, Jack Tarricone organized monthly zoom meetings for InSAR SWE retrieval since 2021 (30+ people, 12 students/recent grads) •In person meetings at different conferences (WSC, NISAR Workshop, IGARSS) •Some recent InSAR SWE/depth publications using SnowEx data: •Zachary Hoppinen, Shadi Oveisgharan, Hans-Peter Marshall, Ross Mower, Kelly Elder, and Carrie Vuyovich, Snow water equivalent retrieval over Idaho – Part 2: Using Lband UAVSAR repeat-pass interferometry, The Cryosphere, 18, 575– 592, https://doi.org/10.5194/tc-18-575-2024, 2024 •Shadi Oveisgharan, Robert Zinke, Zachary Hoppinen, and Hans Peter Marshall, Snow water equivalent retrieval over Idaho – Part 1: Using Sentinel-1 repeat-pass interferometry, The Cryosphere, 18, 559–574, https://doi.org/10.5194/tc-18-5592024, 2024 •Jack Tarricone, Ryan W. Webb, Hans-Peter Marshall, Anne W. Nolin, and Franz J. Meyer, Estimating snow accumulation and ablation with L-band interferometric synthetic aperture radar (InSAR), The Cryosphere, 17, 1997– 2019, https://doi.org/10.5194/tc-17-1997-2023, 2023 •Palomaki, R. T. and Sproles, E. A., Assessment of l-band insar snow estimation techniques over a shallow, heterogeneous prairie snowpack, Remote Sensing of Environment, 296:113744, 2023. •Randall Bonnell, Daniel McGrath, Jack Tarricone, Hans-Peter Marshall, Ella Bump, Caroline Duncan, Stephanie Kampf, Yunling Lou, Alex Olsen-Mikitowicz, Megan Sears, Keith Williams, Lucas Zeller, and Yang Zheng, Evaluating L-band InSAR Snow Water Equivalent Retrievals with Repeat Ground-Penetrating Radar and Terrestrial Lidar Surveys in Northern Colorado, EGUsphere, https://doi.org/10.5194/egusphere-2024236, 2024 NASA L-Band InSAR SWE Working Group •Retrieved snow depth change using UAVSAR shows a very good agreement with LIDAR over flat 5km x 5 km region on Grand Mesa with dry snow •Motivated the time series SnowEx campaign 2021 •Airborne Lidar flights aligned with UAVSAR overflights, allowing direct comparison of depth change •Marshall, H.P., Deeb, E., Forster, R., Vuyovich, C., Elder, K., Hiemstra, C., and Lund, J. (2021). L-band InSAR depth retrieval during the NASA SnowEx 2020 Campaign: Grand Mesa, Colorado. IGARSS, DOI:10.1109/IGARSS47720.2021.9553852 InSAR Working Group Findings: UAVSAR over Grand Mesa HP Marshall et al, IGARSS 2021 TLS ΔSWE retrievals (first column) and UAVSAR ΔSWE retrievals (second column) overlaid on a shaded relief raster, and pixel-wise comparison (third column). Rows organized by date/field site. •Right figure; UAVSAR vs TLS (Terrestrial Lidar Scan) •UAVSAR and TLS ΔSWE retrievals exhibit similar spatial patterns and ΔSWE magnitudes •Correlation: 0.72, RMSE = 19 mm •Poor UAVSAR coherence for 3–23 February 2021 – unwrapping issues •Below figure; UAVSAR vs GPR (Ground penetrating Radar) •ΔSWE retrieval comparison: Large spread, but absolute difference in medians = 2 mm •ΔSWE retrieval accuracy degrades at lower coherence Figure: (a) UAVSAR ΔSWE retrievals vs. GPR ΔSWE retrievals. Box plot distributions of UAVSAR and GPR ΔSWE retrievals for the (b) 2020 and (c) 2021 seasons. InSAR Working Group Findings: UAVSAR over Northern Colorado Randall Bonnell, EGUsphere 2024 (a) InSAR SWE between 19 and 26 February aggregated to the 30 m Landsat resolution. (b) The Landsat fSCA between 18 February and 5 March. The color scale for panel (a) was changed to −5 to 5 cm to exemplify the patterns. InSAR Working Group Findings: UAVSAR over Jemez Mountains, NM Jack Tarricone et al, The Cryosphere 2023 •SWE loss and gain can be estimated within the same radar swath •Morning (~9 AM) acquisitions held coherence and provided reasonable ΔSWE results •InSAR SWE loss visually similar to Landsat fSCA loss patterns •UAVSAR retrieved SWE vs. SNOWMODEL (RMSD = 0.023m, r= 0.60, n = 3e6) •Best performance in high elevation, dry snow regions •UAVSAR retrieved snow depths vs in situ (RMSE = 0.1m, r = 0.80, n = 64) InSAR Working Group Findings: UAVSAR over Idaho Zach Hoppinen et al, The Cryosphere 2024 •SWE = SWE at LIDAR dates •Very good resemblance and correlation between LIDAR snow depth and Sentinel SWE. Site LIDAR Date Correlation Basin Summit (BS) 02-19-2020 0.51 Basin Summit (BS) 03-15-2021 0.47 Mores Creek (MC) 02-09-2020 0.66 Mores Creek (MC) 03-15-2021 0.59 Dry Creek (DC) 02-19-2020 0.56 •SWE = SWE for in situ stations •9 stations, with total SWE error < 2 cm for the entire time series •15 stations, with total SWE error > 2 cm but similar pattern •SnowEx Coordination •Sentinel-1 6-days revisit over Idaho •Closest to NISAR time series •240km covers 32 in situ stations •Continuous time series for the entire winter (12/1 to 3/31) •The correlation between Sentinel-1 and In Situ SWE is 0.8 and the RMSE error is 0.93cm. InSAR Working Group Findings: Sentinel-1 C-band Time Series over Idaho Shadi Oveisgharan et al, The Cryosphere 2024 InSAR products from NISAR Product Grid Posting Spatial scope GSLC Geocoded 5m x (40/10/5/2.5)m Target dependent GUNW Geocoded 80 m Global •All InSAR products will be produced for nearest neighbor pairs •12 days global interferograms •Global spatial scope for Geocoded Unwrapped Interferogram (GUNW) •Correction layers in GUNW products •Ionospheric phase screen •Tropospheric delay •Generating product with better resolution (spatial correlation of. Snow: 50m) or longer time sampling (24, 36, … ) needs tools for snow community Three sample GUNW products produced with NISAR processor (ALOS-1 raw data > RSLC > GUNW) NISAR for Snow Community ●NISAR launch is close – global coverage every 12 days, 80m resolution ●It will be an incredible time series data for retrieving SWE ○UAVSAR: shown the capability of L-band InSAR to retrieve changes in SWE and depth ○Sentinel-1 data analysis have shown the capability of InSAR time series to retrieve total SWE ●Advantages ○Global data at 80m resolution, every 12 days ○The large frame covers large number of in situ stations and SWE spatial variability ○Temporal coherence is higher at L-band compared to Sentinel-1 C-band ○Ambiguous SWE is higher at L-band compared to Sentinel-1 C-band ○It propagates through the vegetation more at L-band compared to Sentinel-1 C-band ●Limitation ○C-band is more sensitive to DSWE than L-band ○NISAR collects InSAR data every 12 days, compared to 6-days for Sentinel-1. It reduces the temporal coherence. ●We need to develop software workflows to generate InSAR data with desired spatial resolution and time sampling ○UAVSAR_pytools, an open source library led by students (now postdocs) started at SnowEx Hackweek ●LIDAR data is the best resource to quantify the sources of error ●Targeted field observations to identify melt, temperature, interval boards, density are very helpful to identify the limitation of this method ●Huge opportunity for snow community – but lots of work to do for a global snow retrieval NISAR is almost here… For more information: https://nisar.jpl.nasa.gov Back Up L-band InSAR ΔSWE retrievals in shallow, variable prairie snow (Palomaki and Sproles, 2023, RSE ) 99% conf. 95% conf. 90% conf. 90% conf. 95% conf. 99% conf. None High variability Low variability good absolute agreement poor absolute agreement Subgrid spatial variability within InSAR pixels contributes to poor absolute agreement. 1 0 Normalized LiDAR SWE 1 0 Normalized InSAR SWE 0.57 -0.93 Difference (LiDAR –InSAR) 0.58 0 Local standard deviation 0 (a) (b) (c) (d) 1 0 Normalized LiDAR SWE 1 0 Normalized InSAR SWE 0.57 -0.93 Difference (LiDAR –InSAR) 0.58 0 Local standard deviation 0 (a) (b) (c) (d) Lidar and L-band InSAR show similar spatial patterns in shallow snow. average snow depth = 12 cm Shown is about 100 Tb of raw SAR data that will be downlinked over each 12 day orbit cycle. Many of NISAR’s science Cal/Val sites are jointly imaged by L-band and S-band Targets include: •India/ISRO (land, ice, ocean) •Global distribution of Ecosystem Cal/Val sites •Arctic/Antarctic ice sheets •Deformation–Western U.S. •Reflector arrays –India, Alaska, California Planned joint L/S band observations for one 12-day orbit cycle NISAR Observation plan Dual-pol 40+5 MHz Dual-pol 20+5 MHz Dual-pol 40+5 MHz Dual-pol 20+5 MHz Ascending Descending ●Currently most science observations over the globe are dominated by 20+5 MHz Dual-Pol Htransmit ●North America is mostly covered by 40+5 MHz Dual-Pol H-transmit acquisitions ●The observation plan is diverse in space but consistent in time (few exceptions) Geocoded Single Look Complex from NISAR Range Azimuth Easting Northing RSLC GSLC ●GSLC products are Single Look Complex (SLC) SAR imagery geocoded on a map coordinate system ●The GSLC products are provided on NISAR frames and will be produced globally ●The grid of the products are in UTM projection at mid latitudes and in Polar stereographic in polar regions ●The grid of GSLC products from different frames have an integer spacing difference allowing to mosaic the products without interpolation. Range bandwidth North spacing East spacing 5 MHz 5 m 40 m 20 MHz 5 m 10 m 40 MHz 5 m 5 m 80 MHz 5 m 2.5 m ●RSLC represents simulated NISAR product from UAVSAR ●GSLC is derived by geocoding the RSLC •The RSLC products will be ~1.2x oversampled in both range and azimuth directions •Azimuth spacing: 1520 Hz (equivalent to ~5 m pixel size) •Slant range spacing: mode dependent •A Kaiser(1.6) window to weight the Range spectrum •The azimuth weighting is the antenna pattern Range bandwidth Azimuth resolution Slant range resolution Azimuth posting Slant range posting 5 MHz ~6 m 30 m ~5 m 25 m 20 MHz ~6 m 7.5 m ~5 m 6.25 m 40 MHz ~6 m 3.75 m ~5 m 3.12 m 80 MHz ~6 m 1.95 ~5 m 1.56 m RSLC resolution and posting RSLC Specification 40 MHz 80 MHz 5 Range frequency [MHz] Range spectrum of different radar modes 20 MHz 5 40 MHz 20 MHz 5 1257.5 The variation of the wavelength (center frequency) across all different L-band modes is less than 3%. NASA’s DAAC (ASF DAAC for NISAR L1L2 products) InSAR Scientific Computing Environment, Enhanced Edition (ISCE3) ISCE3 is the core software to produce the NASA’s SAR products from NISAR (and Sentinel-1) https://github.com/isce-framework/isce3 ISCE3 HySDS Users