Application of a statistical approach to landslide susceptibility map generation in urban settings

Landslide susceptibility maps are effective tools for the mitigation of risks caused by such geological events. In line with recent scientific trends and thanks to the availability of detailed geological data, landslide susceptibility modeling, by means of statistical methodologies, has gained increasing consideration. The present work is based on a methodology widely employed in the field of ecology to draw prediction maps for the occurrence probability of certain species (MaxEnt). The study area is located in Palma Campania, a town sited in the peri-vesuvian area (in the province of Napoli, southern Italy) and characterized by a significant presence of pyroclastic soils, affected by several landslide events, one of which killed eight people in 1986. In this work, eleven geomorphological and geological predisposing factors were selected, based on previous experiences of landslides in peri-vesuvian areas and following several field surveys. Results were critically evaluated using a validation dataset (Receiver Operating Characteristic—ROC curves), by means of Sensitivity-Specificity graphs estimating Area Under Curve (AUC), and other tests such as the Jackknife and response curves, which highlighted the major role played by a number of factors. The consistent agreement between our results and the existing official map demonstrates the validity of the adopted procedure for emergency and land planning