Identification of landslide susceptible zones in Idukki district (Southern Western Ghats) employing the REPTree model and geospatial techniques

Abstract

Landslides, being the most frequent natural catastrophe in the Western Ghats of India, need immediate attention and further research to minimize their impacts. This research aimed at identifying landslide susceptible zones in Idukki district, situated in the Southern Western Ghats, one of the most impacted districts. For the analysis, a machine learning ensemble model called REPTree (Reduced Error Pruning Tree) has been employed, and the map has been created using geospatial techniques. The conditioning factors selected for the analysis include slope, distance from the road, soil texture, curvature, lineament density, aspect, topographic position index (TPI), lithology, land use/land cover (LULC), stream power index (SPI), elevation, and rainfall. According to this modeling, 13.30% of the district is very highly susceptible; 17.00% is highly susceptible to sliding. The validation of the created map employing the ROC curve techniques proved that the map has good predictive capacity. This ascertained the efficacy of the REPTree model in identifying susceptible zones, which therefore can be successfully applied in other regions of similar geomorphological and climatic settings.