A Method for Mapping Snowmelt Extent with Multitemporal Radar Data

Snow cover and its melting have significance implications on simulating hydrological and climatological process. Due to the radar technique is not affected by weather conditions, it has a unique advantage in obtaining snow coverage and its bulk parameters. Numerous studies utilized the multitemporal algorithm to identify wet snow coverage, but it suffers from a simple equation and a fixed threshold, resulting in the obvious underestimation at the end of ablation period. As such, this study firstly explores the cause of the significant underestimation through sensitivity analysis (SA), and then presents a wet snow detection approach based on this analysis. The novel approach implemented the Otsu thresholding method to obtain the adaptive thresholds for various land categories, and further refined the wet snow recognition by using terrain information. To verify the feasibility of the method, one Sentinel-1 dual-polarized synthetic aperture radar (SAR) image acquired from March 23, 2019 was selected as the wet snow data. The extent of wet snow was delineated by multitemporal algorithm, combining with land cover type and topographic data. Compared with previous methods (77.2%), the accuracy of the proposed algorithm is up to 94.9%. The results show that the method has a superior recognition on wet snow at the end of snow melting period, and notably reduces the underestimation.