Snow water equivalent retrieval and analysis over Altay using 12 d repeat-pass Sentinel-1 interferometry

Abstract

Accurate Snow Water Equivalent (SWE) estimation is significant for understanding global climate change, surface energy balance, and regional water cycles. However, although many studies have examined the inversion of SWE using active and passive microwave remote sensing, it remains challenging to assess its global distribution with sufficient temporal and spatial resolution and accuracy. Interferometric Synthetic Aperture Radar (InSAR) has become a promising technique for SWE change estimation, which is limited by the optimal radar frequencies and revisit intervals that have not been available until recently. In this study, 12 d Sentinel-1 C-band InSAR data from 2019 to 2021 are used to retrieve 1SWE (SWE change in one InSAR pair) and cumulative SWE in the Altay region of Xinjiang, China. The correlation between the retrieved 1SWE and in-situ observations reaches R = 0.56, with a low RMSE of 9.54 mm (n = 152) throughout the two whole snow seasons, with values of R = 0.58 and RMSE of 10.1 mm for 2019–2020, and R = 0.48 and RMSE of 8.6 mm for 2020–2021, respectively. These results are obtained by filtering wet snow. Heavy snowfall leads to decorrelation and phase unwrapping errors, which affect 1SWE retrieval and are propagated into cumulative SWE. Validation of the cumulative SWE after removing wet snow yields an RMSE of 40.9 mm, which improves to 28.3 mm when high-elevation stations with unwrapping errors due to heavy snowfall are also excluded. InSAR-derived cumulative SWE time series show consistency with ground observations at some stations, though slight overestimations and underestimations are observed due to error accumulation. Various factors combined with validation results show that higher coherence, lower air temperature, and reliable snow density improve the retrieval accuracy. The proposed coherence-weighted least squares phase calibration method demonstrates that selecting at least half of the available in-situ 1SWE stations for calibration yields reliable 1SWE estimates, although including more points can further improve the robustness. Calibrating only the integer multiples of 2π provides reasonable accuracy, but is still inferior to the full calibration method, indicating that residual modulo 2π phase has a noticeable contribution to the final inversion accuracy, which highlights that phase calibration plays a key role in the accurate 1SWE retrieval. This study provides a valuable reference and processing prototype for applying 12 d revisit Sentinel-1 and future NISAR InSAR data to SWE monitoring.