APPLICATION OF DATA DRIVEN AND KNOWLEDGE DRIVEN METHODS FOR LANDSLIDE SUSCEPTIBILITY MAPPING

Creating landslide susceptibility map is one of the important steps to understand the probable spatial occurrence of landslides in near future. Different methods can be used to make the landslide susceptibility maps, which are broadly classified into two types – data-driven and knowledge-driven. The main objective of this work was to compare the efficacy of the Weighted Overlay Method (knowledge-driven) and Information Value Method (data-driven) to produce the landslide susceptibility models in GIS platform. The proposed work was performed in the Upper West Prong Little Pigeon River watershed, in Great Smoky Mountain National Park that is characterized by high debris flow and active landslide scars. In order to create the susceptibility models five geo-factors were considered including slope, TWI (Topographic Flow Index), soil type, lithology and annual rainfall. Landslide location data were collected from field surveys, analyzing Satellite imageries and Google Earth. In Weighted Overlay Method (WOM), these individual geo-factors were assigned weightage based on our understanding about their influence on landslides and the susceptibility map was prepared. In Information Value Method (IVM), using bivariate statistics numerical weightage of individual classes of each geo-factors were assigned based on the presence of landslide cells in the particular class. Subsequently, Landslide Potential Index (LPI) value was calculated for each of the cells of the geo-factors by performing the arithmetic overlay operation in order to create the susceptibility map. The susceptibility maps were classified into four categories very high, high, moderate and low. The results were validated with available landslide records, which revealed that IVM seems to have better prediction accuracy than the WOM and also exhibits the correlation between landslides and individual classes of geo-factors. However, this research suggests to further evaluate additional geo-factors and statistical models to understand the factors responsible for landslides in the study area.