Sensitivity analysis for shallow Landsliding susceptibility assessment in northern Tuscany

In two areas located in the north-western part of Tuscany, central Italy, Lunigiana and Garfagnana, noticeable heavy rainfall events occurred in the last years. During these events, the rainfall amounts and intensities triggered a great number of shallow landslides, causing damages, injuries and human losses. Steep slopes and deep valleys induced a persistently high relief of energy and high shallow landsliding susceptibility. In the present paper, the authors considered 4 heavy rainfall events that affected the area in 2009–2011. They carried out an analysis including a statistical modelling of spatial landslide occurrence by using Random Forest classifiers (RFc) after model selection by means of a stepwise AIC (Akaike Information Criterion) procedure. Event landslides occurrences permitted to build four event-specific RFc training sets, considering a large number of predictors reliable to characterize landslide susceptibility. Furthermore, the analysis took into account some relevant meteorological variables directly linked to the events themselves. An exploratory evaluation of the skills of a numerical weather prediction (NWP) model was conducted, to give a reliable supply to the RFc framework by using its weather forecast. For one selected event, a shallow landslide hazard model with meteorological inputs was validated. The preliminary results are shown and discussed.