A probabilistic approach to risk assessment of slow slope movements

Many regions are getting more and more vulnerable from a hydro-geological point of view. The causes are due to both the fragility of the territory and to the anthropic influence on its continuous modifications. To avoid or reduce the human life and property economic losses, the quantitative prediction of landslide occurrence and the estimation of the amount of consequent slope movements are necessary. In this paper a specific type of landslide, characterized by a viscous behaviour of soil, is discussed. To catch the displacements evolution of this type of landslide, an analytical dynamic-viscous model was set up. In order to develop an advanced analysis and prediction procedure of the behaviour of such creep landslides, a probabilistic approach, based on Bayesian theorem, is presented. The method is validated making use of a well established and highly reliable monitoring database (Alverà landslide), located in Cortina d'Ampezzo (Dolimites, Italy). The model calibration was then probabilistically performed solving the inverse problem. The solution is then characterized by probability density functions of model parameters, including their corresponding correlation structure. Furthermore, this analysis enables to describe statistically the model error, associating a degree of uncertainty to the predictions.