Application of Frequency Ratio and Logistic Regression Model in the Assessment of Landslide Susceptibility Mapping for Nilgiris District, Tamilnadu, India

Abstract

This study aims to develop landslide susceptibility mapping for Nilgiris district of Tamil Nadu, India using GIS and remote sensing. The Northeast monsoons often trigger numerous landslides in the Nilgiris district of Tami Nadu, resulting in a high death toll and considerable property damage. The landslide causative factors namely elevation, slope, aspect, land use/land cover, rainfall, geology, geomorphology, soil, distance to lineament, distance to road, distance to stream, stream power index and topographic wetness index extracted from the spatial database were incorporated in the analysis, where the correlation between each parameter and landslides were computed using the frequency ratio model and logistic regression model. Finally, the resultant landslide susceptibility zonation maps were obtained from both the models, which were classified as very high, high, moderate, low and very low landslide susceptible zones. The results convey that about 8.78% and 23.22% of the study area were found to be prone to very high landslide susceptibility zones based on the frequency ratio and logistic regression model, respectively. The model validation was carried out using the Receiver Operating Characteristic (ROC) curve by correlating the information obtained from the field verification. The obtained Area Under the Curve (AUC) value of the frequency ratio and logistic regression model was 82.3% and 84.2%, respectively. Therefore, the landslide susceptibility index map of Nilgiris district is claimed to be valuable for the decision-makers to reduce the susceptibility associated with landslide hazard.