Multi-Source Spatio-Temporal Data to Assess Landslide Hazard Evolution in Alpine Environments under Climate Change Scenarios

Landslides are a widespread geomorphological process in alpine environments, where steep relief, complex litho-structural conditions, and active surface processes interact with hydrometeorological forcing to produce highly dynamic patterns of slope instability. Hillslope behavior in these settings reflects the combined influence of long-term terrain conditioning factors and short-term hydrological perturbations. Ongoing climate change is expected to further modify these dynamics through changes in precipitation regimes, snowmelt timing, and cryosphere processes, with important implications for slope stability and landscape evolution. Despite significant advances in landslide susceptibility modelling, monitoring technologies, and climate impact assessment, many existing approaches remain static, event-focused, or weakly connected to geomorphological process understanding, limiting their ability to represent the evolving nature of slope instability over meaningful temporal scales. This thesis aims to improve the geomorphological interpretation, assessment, and monitoring of landslide processes in alpine environments through the development of an integrated, data-driven approach that explicitly accounts for temporal variability and climate-driven change. The research is conducted in the Cordevole Basin (Belluno Province, northeastern Italy), a climatically and geomorphologically complex Alpine catchment characterized by frequent rainfall-induced landslides and pronounced spatial heterogeneity in slope processes. This research starts with the development of a comprehensive multi-temporal landslide geodatabase spanning more than three decades. National-scale landslide records are combined with newly digitized inventories derived from high-resolution orthophotos, allowing the reconstruction of landslide occurrence, reactivation, and spatial patterns of hillslope response through time. The resulting inventories are analyzed together with daily rainfall observations to characterize both short-term triggering conditions and longer-term hydrological forcing acting on slopes. Dynamic landslide susceptibility maps are then produced using a slope-unit-based spatial–temporal Generalized Additive Model (GAM) that integrates static geomorphological attributes with time-dependent rainfall variables. In parallel, a multi-platform framework for long-term landslide monitoring is developed by integrating multi-temporal landslide inventories with ground deformation measurements derived from multiple Interferometric Synthetic Aperture Radar (InSAR) datasets. This framework supports a slope-unit-based interpretation of landslide activity state and its temporal evolution, enabling the identification of persistent deformation patterns and transitions between stable, dormant, reactivating, and active conditions. Finally, the dynamic susceptibility mapping framework is combined with downscaled climate projections to evaluate potential future changes in landslide susceptibility between 2030 and 2100 under two emission scenarios (SSP2-4.5 and SSP5-8.5). The results suggest relatively stable conditions under the moderate-emission scenario, whereas the high-emission scenario indicates a progressive expansion of high-susceptibility areas toward mid- and late-century, driven by intensified rainfall and altered snowmelt regimes.