The potential of predictive methods for landslide occurrences in a GIS environment and on statistical basis has been dealt with by the paper for a wide area across the Southern Italian Apennines (sector of Daunia, Puglia). These methods require a comprehensive knowledge of geographical variables, mostly but not entirely related to morphology, and are able to tie such variables with the occurrence/not occurrence of landslides phenomena within the area of interest. In particular, this study introduces the determination of relevant variables (causal factors) in the study of slope instability and failure that must necessarily be derived from reliable Digital Elevation Model (DEM) and analyzes the characteristics of accuracy and resolution that such models should exhibit in accordance to the scale of investigation and data collecting and/or mapping. The implementation of these and other factors in a Logistic Regression Model (LRM) allowed to calculate the proneness to failure (landslide susceptibility) on a statistical basis through the establishment of the existing relationship between causal factors and presence or absence of landslide phenomena. Therefore, for each of the model resolution cells the likely of investigated phenomena are being quantitatively assessed and, hence, a landslide susceptibility map created. Result are finally validated by comparing the theoretical proneness to landslides with data included in the available inventory