Probabilistic vs. Deterministic Approach in Landslide Triggering Prediction at Large–scale

Abstract

Reliability in the prediction of rainfall-induced shallow landslides at large scale has constituted a great challenge in the last decades. Different approaches have been adopted to include in the forecasts both the geometric, mechanical and climatic factors that affect the triggering phase of the process. A quite promising one is based on the probabilistic physically–based model implemented in the code PG_TRIGRS, which takes into account the uncertainty in soil spatial variability and characterization. The model uses the Kriging technique to assess the spatial distribution of soil properties for the study areas, starting from available georeferenced measurements, alongwith their probability distribution functions. The Point Estimate Method (PEM) is then used to evaluate the Probability of Failure (PoF) within the study area, where PoF is defined as the probability that the Factor of Safety is less or equal than 1. This version is an extension of the original TRIGRS code, which combines a 1D hydrologic model with a stability analysis to assess the safety level of a given slope in a deterministic manner. In this work we compare the results provided by PG_TRIGRS versus the original version of the code, by applying both of them to the same study area in Central Italy.