Application of frequency ratio, information value, and weights-of-evidence models and their comparison in landslide susceptibility mapping in Murree region, Sub-Himalayas

Abstract Landslide is a chronic geohazard in hilly regions which affects the socioeconomic trends globally. Current study is conducted to apply three models including Information Value (IV), Frequency Ratio (FR) and Weights of Evidence (WoE) for Landslide Susceptibility Mapping (LSM) of Murree. Sentinel-2, Google Earth, and field surveys data were used to generate an inventory map of 102 landslides and these events were divided into two subsets i.e., 70% and 30% for LSM and model validation respectively. Eleven causative factors including soil erosion, elevation, slope, aspect, curvature, drainage, fault, road, precipitation, Land Use Land Cover (LULC), and lithology maps were prepared using Google Earth Engine (GEE). The final susceptibility maps were produced with the training datasets of landslide events and causative factors using IV, FR and WoE, whereas these maps were validated using the Receiver Operating Characteristic (ROC) technique. The Area Under Curve (AUC) illustrated the Success Rate Curve (SRC) of 69%, 78% and 79% for the IV, FR and WoE models, respectively, while Predicted Rate Curve (PRC) were 80%, 95% and 87% for the IV, FR and WoE models, respectively. The results of this study can be used by policymakers to plan some mitigation regarding soil erosion and landslides-prone region.