Assessing the present and future landslide susceptibility in Indian Himalayan Region due to climate variability

Landslides pose tremendous challenges to the livelihood of local communities, especially in the Indian Himalayan Region (IHR), where around 75% of the world’s landslides occur annually. The present study demonstrates the applicability of the maximum entropy (MaxEnt) model in accurately predicting the landslide-susceptible areas in the present and future scenarios using the historical landslide occurrence data. Eleven states and two union territories of India were considered in the study. The landslide occurrence data were used to train and test the MaxEnt models with acceptable prediction accuracy (Area Under Curve = 0.65–0.99). The static spatial drivers include elevation, aspect, slope, road network, soil type, and normalized difference vegetation index (NDVI), whereas the dynamic climate variability drivers include bioclimatic data (bio1 to bio19) derived from precipitation and temperature values. The uncertainty in the model outcome is captured by taking the ensemble weighted average of tenfold cross-validated probability layers. Results infer that Jammu & Kashmir, Mizoram, and Sikkim States have more than 50% landslide susceptibility in the present scenario. The remaining states have 30–50% of areas susceptible to landslides, except in Ladakh and West Bengal. Due to the future climate change scenarios under Shared Socio-economic Pathways (SSPs)-126 and -585, an additional area in the range of 0.04–32.34% is estimated to fall under landslide susceptibility, with Tripura and Arunachal Pradesh States having the maximum increase. The results are more sensitive towards the drivers—slope, NDVI, bio12, bio4, bio19, distance to road, and soil type, which vary according to the study area. The study characterizes the landslide susceptibility with landscape metrics and analyses the shift in centroids to understand better the nature of susceptibility in future climate change scenarios.