Monitoring Snowmelt in Mountainous Areas by Considering SAR Geometric Distortion From Ascending and Descending Orbits

Abstract

Synthetic aperture radar (SAR) multitemporal and multipolarization change detection method has been widely utilized for wet snow extraction. Although SAR terrain correction can alleviate the geometric distortion caused by foreshortening and layover, it fails to fully eliminate terrain shadows and great layover effects on steep slopes, presenting significant challenges for high-precision monitoring of snowmelt. This study proposes a method for identifying snowmelt in mountainous areas by combining Sentinel-1 ascending and descending orbits data, taking advantage of the significant differences in geometric distortion areas between two types of orbits that transit on the same day. First, geometric distortion areas and reliable areas are identified on the Sentinel-1 ascending orbit images (evening) and descending orbit images (morning) using radar and local incidence angles, respectively. Wet snow is extracted from each reliable area using SAR multitemporal and multipolarization change detection algorithm. Then, the Sentinel-1 images of the two orbits are combined to obtain the reliable areas with minimal geometric distortion, and wet snow information is obtained through wet snow priority method and reliable areas priority method. Using the Babao River Basin in the Qilian Mountains as the study area, the data were accessed through the Google Earth Engine (GEE) and analysis was run on GEE. At the same time, in order to accurately extract the snow cover range, the spatiotemporal data fusion model was employed to simulate Sentinel-2 images that transit on the same day as Sentinel-1. Finally, the extracted dry/wet snow was corrected using a digital elevation model to determine the distribution of dry/wet snow in the Babao River Basin for a hydrological year from September 2021 to August 2022. Using 71 field measurements of snow water content for accuracy verification, the results indicated that the overall accuracy of using combined ascending and descending orbits data increased from 71.2% for descending orbit alone and 72.4% for ascending orbit alone to 81.7%.