Landslide susceptibility mapping of Kalimpong in Eastern Himalayan Region using a Rprop ANN approach

Kalimpong district, a part of the Darjeeling Himalaya, exhibits a variety of factors that are ideal for the occurrence of landslides. Therefore, it is imperative to demarcate the zones that are highly susceptible to landslide phenomena in advance, so that the risk, and hence the damage can be reduced to a significant extent through proper land-use planning. The factors that have been considered for this study are: (1) elevation, (2) slope, (3) aspect, (4) curvature, (5) distance to drainage, (6) soil type, (7) rainfall, (8) distance to lineaments, (9) landuse, (10) distance to road, (11) TWI, and (12) NDVI. For landslide susceptibility mapping of Kalimpong district, a resilient back propagation (Rprop) artificial neural networks (ANN) approach was used in this study. The results of the Rprop ANN model were validated using the AUC of the ROC Curves. The prediction rate AUC value was found to be 84.35% which showed that this combination of factors with the Rprop ANN model gave satisfactory accuracy in agreement with past landslide phenomena. The derived landslide susceptibility map was categorized in extremely low, low, moderate, high, and very high susceptibility zones covering 610, 272, 83, 61, and 66.7 km 2 of Kalimpong’s area, respectively. Research highlights Landslide Susceptibility Mapping (LSM) of Kalimpong, Eastern Himalaya. A new improvised approach to LSM using Resilient Back Propagation ANN. Different combination of 12 landslide factors with Rprop ANN. High Prediction accuracy of Rprop ANN Model at 84.35%. Quick and reliable LSM method for susceptible zone management.