Algorithmic Insights into Mapping Landslide Susceptibility: A Random Forest Approach in High-Risk Zones of Uttarakhand

This study undertakes the mapping of landslide susceptibility in the vulnerable Tehri-Garhwal and Rudraprayag districts of Uttarakhand, India. Annually, these regions endure heavy precipitation resulting in devastating landslides leading to loss of life and widespread disruptions. To mitigate the effects of such calamities, there is a need for predictive approximations of susceptibility. The assessment of landslide susceptibility is done over 6,228,310 data points, and it includes the contribution of 13 diverse conditioning factors. Modeling for this study is done using the ensemble learning approach to handle the severe class imbalance problem that is present. The resultant model, a Random Forest, addresses complex terrain and environmental dynamics, achieving an overall accuracy of 92.23%.