ABSTRACT Landslide susceptibility mapping (LSM) is a vital tool for managing natural hazards related to slope failures. In Himalaya, most of the LSM studies are confined to either regional scale or along road cut sections. The associated risk factor in such studies is less due to limited number of vulnerable elements at risk. This study focuses on slopes bordering the periphery of Bageshwar city, setting a paradigm for LSM in high‐risk settlements in the Himalaya. Frequency ratio (FR) and Random forest (RF) models were applied and compared. Fifteen conditioning factors pertaining landslides were utilised in the present susceptibility analysis, which incorporates elevation, plan curvature, slope, stream power index (SPI), profile curvature, slope aspect, topographic wetness index (TWI), distance to stream, lithology, land use land cover (LULC), Schmidt hammer rebound (SHR) values, distance to fault, rainfall, distance to road and normalised difference vegetation index (NDVI). The use of SHR values as one of the conditioning factors has been included as field‐generated parameter, which is important in geotechnical engineering. The predictive accuracy of the applied models was assessed by generating receiver operating characteristic (ROC) curves and computing the area under the curve (AUC), as the measures of performance of applied models. The analysis demonstrates that AUC for the success rate curve (SRC) is 90.3% for FR method and 94.8% for RF method. AUC for the prediction rate curve (PRC) is 92% for FR method and 98% for RF method. This depicts that the RF model is relatively better than FR model. This research significantly contributes to proactive disaster risk reduction and safely infrastructure development in high‐risk Himalayan cities by identifying critical susceptibility zones.
Ahamad et al. (Wed,) studied this question.