Snow dynamics in the Alps: long-term trends and remote sensing-driven hydrological modeling

Snow represents a key component of the Alpine hydrological cycle, strongly influencing water availability, ecosystems, and natural resource management. In the context of rapid climate change, understanding the evolution of snow conditions and developing reliable tools for snow monitoring have become increasingly important, particularly in mountain regions. This PhD thesis addresses these issues by focusing on the Central Alps region, considering long-term observational analyses and innovative monitoring approaches integrating satellite remote sensing, hydrological modeling and machine learning. In the first part of the work, long term changes in snow conditions are investigated using in-situ snow and meteorological observations. The results reveal a pronounced decrease in seasonal snow accumulation, with statistically significant negative trends at the regional (Trentino – South Tyrol) and Alpine scales. These changes are particularly evident at mid- to low elevations and are consistent with increasing winter air temperatures, leading to changes in snow/rain partitioning. These findings highlight the need for effective tools to monitor the spatial and temporal variability of snow and snowmelt dynamics. In this context, Sentinel satellite missions provide valuable information due to their high spatial and temporal resolution. However, satellite observations alone are insufficient to fully characterize key snow variables such as SWE, which is essential for quantifying the amount of water stored in the snowpack. The final part of this thesis presents an innovative framework for integrating remote sensing products into a semi-distributed hydrological model with the aim of improving spatial estimates of SWE. The proposed method enables the downscaling of the semi-distributed hydrological model GEOframe with high-spatially distributed satellite-derived snow cover duration information, using a machine learning approach. In this way, the catchment-scale mass balance of the hydrological model is preserved, addressing one of the main limitations of traditional data assimilation techniques. Results demonstrate that this integration leads to a more realistic and spatially consistent representation of SWE and snowmelt processes, enhancing the model’s ability to capture snow dynamics in complex Alpine environments, using limited observations for model training and minimal additional computational load. Overall, this thesis advances the understanding of climate change impacts on Alpine snow and proposes a novel methodological strategy for snowmelt modeling, emphasizing the value of collecting long-term ground observations and integrating satellite remote sensing with hydrological models for hydrological applications and water resource management in mountain regions.