LAUSR

Seasonal snow cover is a critical component of the energy and water budgets of mountainous watersheds. Capturing the snow cover in complex environments is crucial for monitoring and understanding the temporal and spatial effects of climate change on alpine snow cover. The normalized difference snow index (NDSI) can be used to effectively and accurately estimate snow cover information from satellite images. However, the NDSI has limited utility for estimating the snow cover in heavily forested areas and relating this information to snowmelt-based runoff. In this study, a new algorithm based on a multi-index technique is proposed. The technique combines the NDSI, the normalized difference forest snow index, and the normalized difference vegetation index, and decision rules are established to increase the accuracy of snow mapping in forested areas. The new algorithm based on a multi-index technique is tested in the mountainous forested areas of North Xinjiang, China. In a winter image with full snow and a spring image with patchy snow, most of the forest snow, which is underestimated by the NDSI, is recognized by the multi-index technique. The accuracy of snow detection in forested areas is more than 90%. Additionally, in an experiment using a summer image without snow in forested areas no commission errors were detected. The snow detection algorithm based on a multi-index technique uses a simple set of decision rules for snow and can be run automatically without a priori knowledge of the surface characteristics.