Building physically based models for assessing rainfall-induced shallow landslide hazard at catchment scale: case study of the Sorrento Peninsula (Italy)

The assessment of rainfall-induced shallow landslide hazards at the catchment scale poses a significant challenge. Traditional empirical approaches for landslide hazard assessment often assume that conditions having caused failure in the past will not change in the future. This assumption may not hold in a climate change scenario. Physically based models (PBMs) therefore represent the natural approach to include changing climate effects. PBMs would in principle require the combination of a three-dimensional (3-D) mechanical and water-flow model. However, a full 3-D finite element model at the catchment scale, with relatively small elements required to capture the pore-water pressure gradients, would have a significant computational cost. For this reason, simplifications to the mechanical (i.e., infinite slope) and water-flow models (i.e., one-dimensional or hybrid 3-D) are introduced, often based on a priori assumptions and not corroborated by experimental evidence. The paper presents a methodology to build a PBM in a bottom-up fashion based on geological surveys and geotechnical investigation. The PBM is initially set as simple as possible and then moved to a higher level of complexity if the model is not capable of simulating past landslide events. The approach is presented for the case study of Sorrento Peninsula and two main landslide events recorded during the winter of 1996-1997.