CryoNet: New AI model maps debris-covered glaciers with 90% IoU accuracy

Climate scientists from Graz University of Technology and the University of Sannio have developed CryoNet, a deep learning framework that automatically maps debris-covered glaciers—a notoriously difficult task due to spectral similarity between ice and surrounding rock. The model uses an encoder-decoder CNN with nested skip connections and spatial-channel Squeeze-and-Excitation (scSE) attention, built on a ResNet101 backbone. It fuses a rich multi-modal dataset: Sentinel-2 optical imagery, DEM-derived topographic variables, spectral indices, PCA features, InSAR coherence and phase, tasseled-cap components, and GLCM texture. This allows CryoNet to distinguish clean-ice glaciers, debris-covered glaciers, and glacial lakes in complex high-mountain terrain.

In a case study of the Poiqu Basin in the Central Himalaya, CryoNet achieved an overall IoU of 90.52%, mean recall of 98.08%, and mean precision of 92.26%. For debris-covered glaciers specifically, it reached 90.46% IoU, 95.79% recall, and 94.21% precision, outperforming state-of-the-art models like DeepLabV3+, SegFormer, and U-Net. The framework also demonstrated transferability by applying the trained model to the Mont Blanc Massif in the Alps. The authors analyzed the importance of each data layer, noting that multi-modal fusion was critical for the performance boost. CryoNet provides a robust tool for automated glacier monitoring, aiding climate change research and water resource management.