Landslide Susceptibility Modelling and Infrastructure Exposure in the Zillertal, Austrian Alps

Anna Lisa Marie Meiffert , Noah Schreiber

Landslides are a major threat in alpine regions, as steep terrain, complex geology and locally concentrated infrastructure increase societal exposure to landslide hazards. This study develops a landslide susceptibility model for the Zillertal (Tyrol, Austria), utilizing a Random Forest modelling approach. A high-resolution landslide susceptibility map was created using seven geomorphometric, geological and structural conditioning factors. Despite varying positional accuracy of the landslide inventory, the model achieved a respectable performance and predictive capacity (AUC = 0.874). To improve practical usability, susceptibility classes were overlaid with transport infrastructure and built-up areas. The results indicate significant exposure of roads, railways and settlements within high-susceptibility zones. Additionally, a 500 m buffer was applied to assess potential exposure in the surrounding areas and provisionally account for runout. Overall, the study demonstrates that machine-learning-based susceptibility modelling can produce robust and planning-relevant results even where landslide inventories have varying positional accuracy, highlighting its potential for hazard assessment in data-constrained alpine environments.