From Hazard Assessment to Action: National Landslide Susceptibility to Rainfall and Early Warning in Nepal

Abstract

Nepal experiences frequent rainfall-induced landslides due to steep terrain, fragile geology, and intense monsoon storms. This study develops a national-scale landslide susceptibility and early warning framework by integrating statistical modeling, machine learning, and satellite-based rainfall analysis. Slope-unit-based susceptibility was mapped using GAMI-Net with terrain and environmental factors, yielding 84 percent accuracy and demonstrating strong spatial consistency with observed landslides. Susceptibility outputs were coupled with rainfall thresholds derived from IMERG satellite data and DHM gauges to support near-real-time monitoring. The system was deployed in Google Earth Engine as LhamNepal, providing dynamic hazard updates for decision makers. Results confirm the value of localized thresholds and interpretable models for advancing operational landslide early warning in Nepal.