GIS-Based Deterministic and Statistical Modelling of Rainfall-Induced Landslides: A Comparative Study

Abstract

In this paper three different approaches for landslide susceptibility modeling—Shallow Landslide Stability model (SHALSTAB), Likelihood Ratio (LR) and Generalized Additive Model (GAM)—are compared. They are based on deterministic and statistical methods, respectively. These methods were tested in the Pogliaschina catchment (25 km2 wide; Northern Apennines, Eastern Liguria, Italy), heavily hit by an intense rainfall on 25 October 2011, that caused hundreds of shallow landslides, human losses and severe damage to infrastructure and buildings. The paper focuses on the assessment of the predictive performance of the three methods through a two-fold cross-validation technique and prediction rate curves (PRCs) analysis. The preliminary results have revealed that statistical methods have a higher predictive capability than the deterministic one.