Appraisement of Landslide Propensity Mapping by Machine Learning Algorithm in Indian Siwalik Himalayan Region, Mirik and Kurseong

Himalaya is the northern limit of the Indian subcontinent. Siwalik range is its southern margin. The hilly Siwalik range is considered as the landslide-prone region. Mirik and Kurseong, part of the Darjeeling Himalaya the examples of landslides of Siwalik Himalaya. The Chapter analyzes landslide propensity point (LPP) based on several influencing indicators in different aspect, viz. lithological (geological setup, nature and depth of soil and lineament), topographical (morphological furnish, slope and their direction of inclination, relief degree of land surface, curvature and topographic position index), hydrological (flow direction, drainage density, distance from river, topographical wetness index, stream power index, and sediment transport index), environmental (vegetation cover assessment by normalized difference vegetation index, landuse and landcover distribution, precipitation, magnitude of earthquake) and anthropogenic (distance from road and kernel density) factors. Based on 77 landslide inventories of the Siwalik Himalaya of Mirik and Kurseong, the frequency ratio model (FRM) and random forest model (RFM) are used for the perfection of computation in propensity mapping. Inventories are coined from field surveys (7), the NASA Global landslide points (18), and Google Earth Pro (53). Among the two models, landslide propensity mapping (LPM) by random forest model (accuracy level 90.32%) gained a higher accuracy level than the zonation mapping by frequency ratio model (FRM) (accuracy level 87.096%) as per confusion matrix and ROC-AUC. In addition, it also evaluates how artificial intelligence is better for susceptibility mapping and analysis. Through this chapter, it has been quantified that the Siwalik range in Mirik and Kurseong is facing vulnerable circumstances from landslides. It will be quite helpful to assess the management required point of the said study area.