Mirza et al. (2025) Evaluating the utility of Sentinel-1 in a Data Assimilation System for estimating snow depth in a mountainous basin

Identification

• Journal: ˜The œcryosphere
• Year: 2025
• Date: 2025-12-09
• Authors: Bareera N. Mirza, Eric E. Small, Mark S. Raleigh
• DOI: 10.5194/tc-19-6691-2025

Research Groups

• College of Earth, Ocean, and Atmospheric Sciences, Oregon State University, Corvallis, OR, USA
• Geological Sciences, University of Colorado, Boulder, CO, USA

Short Summary

This study evaluates the temporal and spatial accuracy of Sentinel-1 (S1) C-band radar snow depth retrievals and their utility within a data assimilation (DA) system for mountain snowpack characterization in the East River Basin, Colorado, finding significant inconsistencies and limited potential for S1 to improve snow DA in this region.

Objective

• To understand how Sentinel-1 snow depth errors vary across space and time.
• To determine the relative value of using full-season Sentinel-1 data compared to early-season data in a data assimilation framework.
• To assess whether the joint assimilation of Sentinel-1 snow depth and MODIS snow disappearance date can enhance DA accuracy.

Study Configuration

• Spatial Scale: East River Basin, Colorado, USA (~748 km² area); daily snow depth maps at 500 m resolution.
• Temporal Scale: Water years 2018–2021; S1 data evaluated from 2017 to 2021 with daily to weekly temporal resolution; assimilation windows included full season (15 November to 30 April) and early season (15 November to 15 January).

Methodology and Data

• Models used:
  • Flexible Snow Model version 2 (FSM2) (physics-based snow model).
  • Particle Batch Smoother (PBS) (data assimilation algorithm).
  • Multiple Snow Data Assimilation System (MuSA) (open-source toolbox for FSM2 ensemble generation).
• Data sources:
  • Satellite/Remote Sensing:
    • Sentinel-1 (S1) C-band Synthetic Aperture Radar (SAR) snow depth product (500 m resolution).
    • MODIS Snow Disappearance Date (SDD) derived from MOD10A1 product.
    • Airborne Snow Observatory (ASO) LiDAR snow depth surveys (50 m resolution, resampled to 500 m).
  • Observation/Ground-based:
    • 12 ground-based stations (11 NRCS SNOTEL stations, 1 independent site at Snodgrass Mountain) for daily snow depth.
    • Monthly in situ snow pit measurements from Snodgrass Mountain.
  • Reanalysis:
    • ECMWF ERA5-Land reanalysis data (0.1° / 9 km resolution, hourly) for meteorological forcing (incoming shortwave/longwave radiation, total precipitation, surface atmospheric pressure, 2 m air temperature, 2 m relative humidity, 10 m wind speed).

Main Results

• Sentinel-1 (S1) snow depth data exhibited significant inconsistencies between temporal and spatial errors.
• Temporally, S1 showed an average root mean square error (RMSE) of 0.40 m and coefficient of determination (R²) of 0.74 against ground stations, with errors increasing during ablation.
• Spatially, S1 had higher errors (average RMSE 0.82 m, R² 0.23) and poor agreement with Airborne Snow Observatory (ASO) LiDAR data.
• Data assimilation (DA) experiments with full-season S1 data (Hs-F) and early-season S1 data (Hs-E) showed minimal temporal performance differences, but Hs-F generally outperformed Hs-E spatially.
• Joint assimilation of S1 snow depth with MODIS Snow Disappearance Date (SDD) did not enhance performance; SDD assimilation alone yielded the best spatial accuracy (average RMSE 0.41 m, R² 0.72).
• The study concludes that S1 has limited potential to improve snow DA in the East River Basin, Colorado.

Contributions

• Provided a comprehensive evaluation of Sentinel-1 C-band SAR snow depth data's utility within a data assimilation framework in a complex mountainous basin (East River Basin, Colorado).
• Systematically quantified the spatiotemporal error discrepancies of S1 snow depth, highlighting its limitations, particularly in spatial accuracy and during ablation periods.
• Assessed the impact of different assimilation window sizes (full-season vs. early-season) for S1 data, finding minimal differences in temporal performance but better spatial results with full-season data.
• Investigated the value of joint assimilation of S1 snow depth with MODIS Snow Disappearance Date, demonstrating that SDD alone provided superior spatial accuracy.
• Contributed to understanding the challenges and limitations of current S1 snow depth retrieval algorithms for improving snow water equivalent (SWE) mapping in the Western US.

Funding

• National Aeronautics and Space Administration (NASA) Terrestrial Hydrology Program (Award No. 80NSSC22K0685).

Citation

@article{Mirza2025Evaluating,
  author = {Mirza, Bareera N. and Small, Eric E. and Raleigh, Mark S.},
  title = {Evaluating the utility of Sentinel-1 in a Data Assimilation System for estimating snow depth in a mountainous basin},
  journal = {˜The œcryosphere},
  year = {2025},
  doi = {10.5194/tc-19-6691-2025},
  url = {https://doi.org/10.5194/tc-19-6691-2025}
}