Random forest based spatial modeling of landslide susceptibility and land use exposure for sustainable land management in rural area

Landslides pose a severe environmental and socioeconomic threat in mountainous regions, yet existing spatial risk assessments often isolate hazard prediction from actual land-use exposure.This study develops an integrated geospatial framework to evaluate landslide susceptibility and quantify anthropogenic exposure in Tinggimoncong District, Indonesia, to support sustainable land management.A Random Forest algorithm was utilized within Google Earth Engine to model landslide susceptibility using an inventory of 198 landslide and non-landslide points and 11 parameter predictors.Concurrently, a land-use classification was performed using Sentinel-2A imagery.The susceptibility model achieved robust predictive performance with an area under the curve (AUC) of 0.843 and an overall accuracy of 80.3%, identifying slope gradient and distance to roads as the most dominant landslide-driving factors.The validated land-use map with overall accuracy of 85.6% was spatially intersected with high-susceptibility zones (>0.7), revealing that 16.97% of the road infrastructure and 4.90% of settlements are highly exposed to landslide hazards, contrasting with a minimal exposure 0.22% for agricultural lands.These findings provide a robust scientific foundation for sustainable land management, recommending targeted interventions such as a moratorium on new construction in red zones, structural geotechnical engineering on vulnerable road-cut slopes, and deep-rooted vegetative conservation on exposed agricultural lands.