Statistical and machine intelligence based model for landslide susceptibility mapping of Nilgiri district in India

Abstract

The Nilgiri district is the mountainous landscape which is frequently facing landslides that leads to severe economic and human loss. So, it is most important to inspecting and exploring the degree or magnitude of landslide hazard in Nilgiri district. The current research presents a detailed mapping on vulnerability of landslide for the future with the assist of frequency ratio (FR) and binary logistic regression (LR) by the support of ArcGIS and SPSS modeler finally assessed the performance and accuracy of model has carried out with the aid of Area under the curve (AUC) for the receiver operating characteristics (ROC). For this study ten important conditioning factors has been used, they are: Lithology, NDVI, Rainfall, Lineament density, Lineament buffer, Slope, Soil, Depth of the soil, Aspect and Land use Land cover. The most significant part of the research has made an attempt to use statistical package (SPSS modeler) for geospatial analysis particularly landslide study. The major finding of this study is that FR is good as compared to LR when using training dataset whereas LR is performed as good as compared to FR when using testing dataset. The resultant Landslide susceptibility map (LSM) from the FR and LR using SPSS modeler is significant units not dependent of governmental boundaries, this can be also advantageous to planners or law executer and policymakers to mitigate landslide.