Assessment of Slope Instability in Darjeeling Himalayan Region: Comparative Evaluation of Bi-Variate Statistical Methodologies

Landslides, which can occur naturally or as a result of human activity, are one of the primary natural disasters that occur on Earth. Landslides, which can be fatal, are more common in mountainous and hilly terrain. Assessing the vulnerability and identifying the landslide sensitive areas becomes crucial in order to preserve human lives and reduce economic damage, which are the two main effects of the fatal calamity. In order to identify the places in the Darjeeling Himalayan region that are prone to landslides, the current study compares and analyses two bi-variate statistical approaches, namely weight-of-evidence and frequency ratio. With utilization of Google Earth imagery, the Geological Survey of India's standard repository, and numerous field investigations, a total of 728 landslide sites have been identified. A random selection of 70% of the detected landslide areas are utilized as training data, while the rest of 30% are treated as validation data. Topographic, climatic, lithological, hydro-morphological, anthological, and land-use factors are the main categories for the influencing factors for slope instabilities, and these are further divided into sixteen sub-categories. Five classes of susceptibility for slope instabilities were established with the methodologies. The analysis exhibits 85% and 77% specificity in terms of training data; and 80% and 74
^%
sensitivity in terms of validation data. The substantiation of the constructed maps was performed using the AUC-ROC curve.