Applicazione di un modello previsionale di tipo bayesiano per la valutazione della pericolosità connessa a movimenti di versante

A quantitative prediction model for landslide hazard assessment is presented here. It is applied in the Versa river basin (Oltrepo Pavese, north-western Apennines - northern Italy), an area characterised by a complex geological and geomorphological setting with high frequency and density of landslide phenomena. The applied prediction model (Weight of Evidence, BONHAM-CARTER ET AL. 1989) is the log-linear form of the bayesian model and is based on a spatial database, consisting of several layers of digital maps representing the causal factors of the occurrence of landslides. This model gives a quantitative measure of future landslide hazard and is based on two basic assumptions: 1 - future landslides will occur under circumstances similar to the ones of past landslides; 2 - the spatial data representing the causal factors contained in the GIS database can be used to formulate the future landslide hazard. The study is subdivided in nine steps: 1) definition of the conceptual model and map scale; 2) collection of available data; 3) construction of GIS database containing all causal factors (themes and other supporting information); 4 - 5) pre-processing of selected data in order to obtain the classified themes and the digital elevation model of the study area; 6) application of the prediction model and estimation of probability values; 7) display of the preliminary hazard map; 8) validation of the prediction results through sensitivity and comparative analysis (the availability of landslide inventories referring to different periods (1978, 1992, 1999) gives consistency to this phase of validation of prediction results); 9) construction of the final hazard map.