Abstract
In this paper, an integration landslide susceptibility model by combining expert-based and
bivariate statistical analysis (Landslide Susceptibility Index—LSI) approaches is presented. Factors
related with the occurrence of landslides—such as elevation, slope angle, slope aspect, lithology, land
cover, Mean Annual Precipitation (MAP) and Peak Ground Acceleration (PGA)—were analyzed
within a GIS environment. This integrated model produced a landslide susceptibility map which
categorized the study area according to the probability level of landslide occurrence. The accuracy of
the final map was evaluated by Receiver Operating Characteristics (ROC) analysis depending on
an independent (validation) dataset of landslide events. The prediction ability was found to be 76%
revealing that the integration of statistical analysis with human expertise can provide an acceptable
landslide susceptibility assessment at regional scale.