SAU-Net: A Deep Learning Approach for Glacier Mapping Based on Multisource Remote Sensing Data

Glaciers are vital indicators of climate change, particularly in the Tibetan Plateau, which is experiencing unprecedented glacial retreat due to rising temperatures. This study presents a novel deep learning model, SAU-Net, for automatic glacier segmentation using multisource remote sensing data. SAU-Net, built upon the U-2-Net architecture, incorporates a two-layer nested U-shaped structure and a Simple Attention Module (SimAM) to enhance feature extraction and segmentation accuracy. The model is applied to the Himalayas and Karakoram regions, capturing diverse glacier types and environmental conditions. Results demonstrate that SAU-Net achieved an impressive accuracy of 94.6% and F-beta score of 0.835, significantly outperforming the conventional U-Net model in glacier identification tasks. The successful application of SAU-Net not only addresses challenges associated with glacier mapping in complex terrains but also provides a scalable method for generating regional glacier inventories, contributing to improved monitoring and understanding of glacial dynamics under climate change. This research underscores the transformative potential of deep learning in environmental monitoring and provides a framework for future studies in cryospheric science.