Uncertainty quantification in the calibration of a dynamic viscoplastic model of slow slope movements

Abstract

Most landslides occurring in Italy consist of shallow-translational movements, which involve fine, essentially clayey material. They are usually characterized by low velocities, typically of few centimeters per year. The main triggering factor is hydrologic, since movements are usually strictly connected to groundwater level fluctuations. This slow and periodical trend can be interpreted by a viscous soil response, and in order to catch the actual kinematics of the soil mass behavior, a dynamic analysis should be adopted. This paper discusses the case of the Alvera mudslide, located in the Northern Alps (Italy), for which a very detailed and almost 9-year-long monitoring database, including displacements and groundwater levels records, is available. A well-defined dynamic viscoplastic model, capable of returning a displacement prediction and a mobilized shear strength angle estimate from a groundwater level input, was considered. A first deterministic calibration proved the ability of the model to reproduce the mudslide overall displacements trend if a suitable reduction of the mobilized angle \(\varphi ^{\prime }_{0} \) is allowed. Then, an uncertainty quantification analysis was performed by measuring the model parameters variability, and all parameters could be represented using a probability density function and a correlation structure. As a consequence, it was possible to define a degree of uncertainty for model predictions, so that an assessment of the model reliability was obtained. The final outcome is believed to represent an important advancement in relation to hazard assessment and for future landslide risk management.