Predictive Power Evaluation of a Physically Based Model for Shallow Landslides in the Area of Oltrepò Pavese, Northern Italy.

The use of real-time landslide early warning systems is attracting the attention of the scientific community, since it allows to assess "where" and "when" a shallow rainfall-induced landslide might occur by coupling rainfall amounts, hydrological models and slope-stability analysis. The paper deals with the main results of a back analysis, which refers to the application of a physically based stability model [Shallow Landslides Instability Prediction (SLIP)] on regional scale. The analysis concerns the occurrence of some recent rainfall-induced shallow landslides in the municipal territory of Broni, in the area of Oltrepò Pavese (Northern Italy). The study area is a hilly region 2.4 km2 wide, where more than 40 % of the territory has slopes steeper than 15° and altitudes are between 90 and 250 m a.s.l. As regards the geologic setting, clayey-silty shallow colluvial deposits, with a maximum thickness of about 3 m, overlap a bedrock made of clayey shales, calcareous flysch and marls. The SLIP model is based on the limit equilibrium method applied to an infinite slope and on the Mohr-Coulomb strength criterion for the soil. By assuming that the main hydro-geotechnical process that leads to failure is the saturation of parts of the soil, the model allows to take into account the condition of partial saturation of the soil. The safety factor (FS) of a slope is also function of previous rainfalls. After the implementation of the model at territory scale, the input data have been introduced through a geographic information systems platform. In the current paper we mainly intend to evaluate the performance of SLIP at catchment scale, by comparison to (1) observed landslide events and (2) another well-established physically based model (TRIGRS). Further, we want to assess the suitability of the model as early warning tool. The results produced by the model are analyzed both in terms of safety factor maps, corresponding to some particular rainfall events, and in terms of the time-varying percentage of unstable areas over a 2-year span period. The paper shows the comparison between observed landslide localizations and model predictions. A quantitative comparison between the SLIP model and TRIGRS is presented, only for the most important event that occurred during the analyzed period. Overall, the results of the stability analyses based on observed rainfalls show the capability of the SLIP model to predict, in real-time and on a wide area, the occurrence of the analyzed phenomena.