Landslide susceptibility zonation in the Nepalese Siwaliks using GIS and the frequency ratio model

Abstract

Landslides are recurrent and destructive hazard in Nepal’s geologically young and fragile Siwalik (Chure) Hills, where the lack of systematic susceptibility assessments limits effective land-use planning and disaster risk reduction. This study aimed to evaluate and map landslide susceptibility in Rajpur Rural Municipality, Dang District, using Geographic Information System (GIS) and the bivariate Frequency Ratio (FR) model. A spatially explicit landslide inventory of 151 historical events was prepared from high-resolution Google Earth imagery. Nine landslide conditioning factors-slope, aspect, elevation, plan curvature, land use/land cover (LULC), geology, soil type, mean annual precipitation, and proximity to rivers—were processed and analyzed within a GIS framework. Class-wise FR values were calculated to quantify the relationship between each factor and past landslide occurrences, and these values were integrated to produce the Landslide Susceptibility Map (LSM) that classified the area into five susceptibility zones. The FR analysis revealed a strong association between landslide occurrence and water bodies within the LULC parameter (FR=1.52), highlighting the role of fluvial erosion and slope undercutting. Soil emerged as the most influential factor, exhibiting the highest prediction rate (PR = 2.57). Model validation using the Area Under the Curve (AUC) method yielded an AUC value of 0.545, indicating a marginal but positive predictive performance above random classification. The resulting LSM provides a scientifically grounded decision-support tool for local authorities, planners, and disaster management agencies to identify priority areas for targeted mitigation measures, including bio-engineering, community-based afforestation, and risk-sensitive infrastructure development, demonstrating the practical utility of the FR model in data-scarce mountainous environments of Nepal.