Assessing Landslide Susceptibility in Indian Himalayas: Comparing Polygon and Point-Based Inventories with Modified Frequency Ratio Approach

Abstract

Abstract This study examines the effects of using point and polygon-based landslide inventory on the process of mapping landslide susceptibility in the Northwestern Indian Himalayas. The modified frequency ratio method was utilized to generate the landslide susceptibility map, applying classification through the define, equal, geometric, natural break, and quantile reclassification procedures. Comparative analyses were performed to compare the polygon-based and point-based landslide susceptibility maps using different reclassification methods. The polygon-based methodology achieved success rates/prediction rates of 75.0%/75.4%, 76.1%/76.4%, 77.9%/78.4%, 77.9%/78.4%, and 78.1%/78.6% for the define, equal, geometric, natural break, and quantile classification methods, respectively. On the other hand, the point-based strategy resulted in success rates/prediction rates of 81.8%/82.1%, 83.0%/83.2%, 84.2%/84.6%, 84.3%/84.6%, and 83.5%/83.7% for the respective categorization techniques. The results showed that the point-based landslide susceptibility map had a higher performance in terms of AUC values, but the polygon-based map was better at portraying ground conditions. Geometric, natural break, and quantile reclassification methods consistently shown superior performance compared to define and equal methods in both point and polygon-based approaches. Although both point and polygon-based inventories showed acceptable levels of accuracy, it is advisable to use the polygon-based technique, provided that the necessary data and computer resources are available. This research provides useful insights into the selection of inventory types and classification methods for the accurate mapping of landslide susceptibility in the rugged terrain of the Northwestern Indian Himalayas.