Example data of Trans-CAUNet model for automated extraction of mountain glaciers

Accurate and automated glacier extraction is pivotal for water resource management and ecological protection in cryospheric systems. Given the expanding access to remote sensing imagery from considerable satellite platforms, there is an urgent requirement to develop efficient data mining methods that allow for the rapid identification of mountain glaciers in High Mountain Asia (HMA). Proposed here, a novel model based on the UNet framework that embeds Swin transformer block and channel attention mechanism (CAM) parallel modules for glacier extraction. The primary advantage of the model lies in its ability to capture global-local and contextual information from the feature maps based on multisource remote sensing imageries, further enhancing spectral features. This model was performed on a combined dataset from two prototypical regions, namely Qilian Mountains and Western Kunlun Mountains. This dataset is an example data set.