Landslide risk using Geospatial techniques and machine learning: Shimla district of Himachal pradesh, India

Abstract

The frequency of landslide occurrences has increased due to climate change and human-induced alterations in Shimla district of Himachal Pradesh in India. Occurrence of landslides has posed great risk to ecological, physical and social systems. Thus, assessing landslide risk is crucial for devising adaptation and mitigation strategies. This study makes concerted efforts to integrate hazard, vulnerability and element-at-risk for landslide risk assessment. The effectiveness of the multilayer perceptron model was assessed using key performance metrics. The validation of the map was carried out using confusion metrics. Landslide risk analysis revealed that most of the area falls in the very high risk followed by high, low and moderate risk. Mashobra, Basantpur, Narkanda and Rampur blocks (administrative divisions) experienced very high landslide risk. Rainfall, slope, wetness index, building density and extensive road network have been attributed to very high landslide risk. Theog, Chaupal and Rohru blocks experienced high landslide risk due to high temperature variability, high population density and low literacy rate. Stability measures, effective land use and instrument installation are suggested for enhancing adaptive capacity among communities. Thus, the comprehensive framework applied in this study may be used across other geographical regions to classify risk zones and recommend effective mitigation measures.