Comparison of two data fusion methods from Sentinel-3 and Himawari-9 data for snow cover monitoring in mountainous areas

Abstract

Snow cover in mountainous areas is characterized by high reflectivity, strong spatial heterogeneity, rapid changes, and susceptibility to cloud interference. However, due to the limitations of a single sensor, it is challenging to obtain high-resolution satellite remote sensing data for monitoring the dynamic changes of snow cover within a day. This study focuses on two typical data fusion methods for polar-orbiting satellites (Sentinel-3 SLSTR) and geostationary satellites (Himawari-9 AHI), and explores the snow cover detection accuracy of a multi-temporal cloud-gap snow cover identification model (Loose data fusion) and the ESTARFM (Spatiotemporal data fusion). Taking the Qilian Mountains as the research area, the accuracy of two data fusion results was verified using the snow cover extracted from Landsat-8 SR products. The results showed that both data fusion models could effectively capture the spatiotemporal variations of snow cover, but the ESTARFM demonstrated superior performance. It not only obtained fusion images at any target time, but also extracted snow cover that was closer to the spatial distribution of real satellite images. Therefore, the ESTARFM was utilized to fuse images for hourly reconstruction of the snow cover on February 14–15, 2023. It was found that the maximum snow cover area of this snowfall reached 83.84% of the Qilian Mountains area, and the melting rate of the snow was extremely rapid, with a change of up to 4.30% per hour of the study area. This study offers reliable high spatiotemporal resolution satellite remote sensing data for monitoring snow cover changes in mountainous areas, contributing to more accurate and timely assessments.