Reconstructing rainfall-induced landslides at the global scale

Rainfall-induced landslides are among the most widespread and destructive natural hazards, yet their physical reconstruction has rarely been explored beyond local or regional scales. We present a simplified slope-stability framework driven entirely by globally available rainfall, soil, and topographic datasets, and demonstrate its ability to reproduce thousands of rainfall-triggered landslides documented in the Global Landslide Catalog (GLC).By avoiding computationally intensive hydrological simulations while retaining physical interpretability, the proposed approach enables large-scale reconstruction of rainfall-induced slope failures across diverse environmental settings. Sensitivity analyses indicate that slope geometry and rainfall forcing primarily control proximity to failure and its timing, whereas soil bulk density exerts a disproportionate influence on model uncertainty due to its structural role in both mechanical resistance and hydrological response.Model performance is strongest in tropical and temperate regions, while reduced skill is observed in arid and cold climates, where failures tend to be conservatively predicted, favouring early-warning applications. Under scenarios characterised by intensified extreme rainfall, the framework suggests an overall increase in global slope instability. These results demonstrate the feasibility of reconstructing rainfall-induced landslides at the global scale using simplified physical representations, and highlight key directions for further improvement, including vegetation effects, subsurface heterogeneity, and hydrological process representation.