Propagation analysis and risk assessment of an active complex landslide using a Monte Carlo statistical approach

The risk assessment of a rapid landslide is a difficult topic, even if based on the results of numerical analyses. The hypotheses on which every model is developed, the choice of rheological laws to be adopted, and the selection of soil parameters make the simulation results highly dependent on the user. This is particularly evident when there is no model calibration for the specific site or reliable information on soil properties. The paper presents a forecasting process obtained using a Monte Carlo approach in coupling with a propagation model developed with the SPH integration technique. The Monte Carlo analysis allows automatically carrying out a large number of simulations, each performed using an independent parameter set randomly selected within a priori assigned statistical distributions. The numerical results are then analysed with statistical tools to create a risk map based of the frequency of the unstable mass runouts. In this way, it is possible to reduce the user dependence of results and increase the examined potential scenarios. The procedure is here applied to the case study of the Sant'Andrea landslide, a slope movement active since several decades in the municipality of Perarolo di Cadore (Belluno, Italy). This complex slide involves an about 30 m-thick deposit of calcareous debris overlying anhydrite-gypsum rocks. Depending on the intensity and duration of rain, the slope alternates phases characterized by slow displacements and significant accelerations, then followed by a long relaxation period in which the displacement rate slowly regresses, without returning to the previous condition of movement. In recent years, the landslide activity has caused a progressive enlargement of the unstable area and a gradual increase of the basal rate, thus increasing the risk that the landslide may suddenly undergo to the collapse. Moving from the knowledge of the unstable volume, an SPH propagation model is used to study the area affected by the debris-flow runout. In particular, the analysis aims to define a statistical strategy to perform and interpret a large number of simulations and to create the consequent risk map. The analyses carried out lead to a satisfactory interpretation of the spatial variability of the deposit heights referred to the post-failure conditions, useful for the development of a risk analysis, from which a site risk map can be obtained.