Abstract
Alpine glaciers retreat induces slope debuttressing which cause destabilization of lateral rock walls, landslides and debris discharge onto the glacial surface. Rocks accumulated on the ice modify the glacial thermal regime, debris disposal and nonetheless shift of climbing routes. Identifying the location and occurrence of these events is thus crucial for assessing areas susceptible to destabilization and linking their occurrence with external triggers.
Here, we propose a tool implemented in Google Earth Engine that integrates spaceborne radar and multispectral information to detect and classify glacial changes to minimize the detection limitations of the two sensors. Our workflow, tested over Vedretta della Miniera (Val Zebrù- IT) and Tscherva glacier (Val Roseg- CH) exploits backscatter changes of the surface, leveraging the decibel (dB) information from Sentinel-1, and integrates the Normalized Difference Snow Index (NDSI) derived from Sentinel-2 to validate the effectiveness of the detected changes.
Starting from Sentinel-1 intensity images (Mulissa et al., 2021) selected within a user-defined timespan, we compute the backscattering backward difference (∆). Using a size-dependent filter, we outline discrete pixel clusters and compute the mean ∆VH within each. Clusters whose mean ∆VH falls outside the interquartile range (IQR3) of the cluster values in the image pinpoint the dates when most prominent changes occurred on the glacier. For these dates, we consider n subsequent Sentinel-2 images and evaluate the modal NDSI value over each pixel in the clusters. Based on NDSI, we attribute the backscatter change to snow, rock and debris exposure or indicate glacial areas affected by modification if persistent cloud coverage prevents specific classification.
The tool successfully identified the dates of two major landslide events in different seasons and snow conditions over the areas of interest. For Vedretta della Miniera, where manual mapping validation was feasible, it detected 41% (0.015 km2) of the 0.035 km2 landslide deposit.