Evaluating landslide hazard, vulnerability, and risk using machine learning; A case study from the Alaknanda Valley, NW Himalaya

Landslides are one of the most destructive geological hazards in the Himalaya, and their frequency has increased in the last few decades. Therefore, it is essential to evaluate the landslide risk in such mountains. The present study provides a comprehensive landslide hazard, vulnerability, and risk assessment along the Alaknanda valley, NW Himalaya, India. The investigated area is selected because of its importance for holy pilgrimage sites in India. A machine learning technique, the MLP (Multilayer Perceptron) approach, has been used for preparing the landslide susceptibility map, which has provided efficient results with SRC (success rate curve) of 0.95 and PRC (prediction rate curve) of 0.87. The landslide hazard map was prepared using the maximum rainfall intensity from the last two decades of precipitation data. The economic vulnerability was derived by using the the monetary value of the components of LULC (Land Use Land Cover). It has been observed that dam sites, settlements, and roads are highly vulnerable to landslide hazards. Finally, the risk map was generated by integrating the landslide hazard map with the vulnerability map that covered the 4% high to very high risk, 16% moderate risk, and 80% low to very low risk areas. Results reveal that high-risk areas were mainly concentrated in the southern parts, and dam sites in the northern parts were categorized as high-risk. About 42% of the population lives in high risk areas. The results of this research may be helpful for disaster management administrators in future planning and development of areas with the least risk.