Landslide susceptibility zonation around Dharamshala, Himachal Pradesh, India: an artificial intelligence model–based assessment

Landslide susceptibility zonation (LSZ) is a prerequisite for sustainable development and disaster management, especially in mountainous settings. In recent landslide literature, due to the shortcomings of qualitative, statistical, and probabilistic approaches, artificial intelligence (AI) techniques are widely applied for geographic information system–based LSZ. Dharamshala, state capital of Himachal, is one of India’s fastest-growing tourist hotspots in Himalaya where developmental activities are taking place at a rapid pace. Increased pressure of urbanization and re-occurrence of slope instability problems demand systemic landslide hazard evaluation in this area. In this respect, GIS-based LSZ has been attempted for this area using different AI models: fuzzy set procedure (FSP), the fuzzy expert system (FES), and artificial neural network (ANN). To create the landslide susceptibility map (LSM), 9 causative factors and landslide inventory were defined using remote sensing and field data. A total of 12 LSMs were generated, then validated and compared statistically with the help of area under the receiver operating characteristic (ROC) curve and frequency ratio (FR) analysis, and also in terms of spatial distribution quality to depict an accurate map for the study area. The comparative analysis shows that ANN performs better than the other two models, and LSM-ANN-I is the best map for the research area. The northern section, made of metamorphic rocks like slate and phyllite, with a high slope, is more prone to landslides, whereas the southern part, comprised sandstone, shale, on gentle slopes, has minimal landslide susceptibility.