ABSTRACT: The Beauregard Landslide is a deep-seated gravitational slope deformation located in the Aosta Valley (Dora di Valgrisenche river) in northwestern Italy. Numerical simulations of the Beauregard Landslide use three-dimensional mixed continuum-discontinuum methods to explore the role and importance of sophisticated geometric interpretations in analyzing landslide mechanics and to test model sensitivity to shear zone strength parameters. 3DEC (3-Dimensional Distinct Element Code) has been used to generate complex threedimensional landslide geometries. The landslide and surrounding, undisturbed, rockmass are defined as distinct continuum blocks which interact along discrete discontinuities representing landslide shear surfaces. The full three-dimensional geometries of these shear surfaces are interpreted from geological and morphological data using a rigorous statistical interpolation approach. This study aims to improve landslide hazard management by recreating observed slope deformations which vary across the landslide footprint. The simulated deformations from models are compared to observed deformations from real slope monitoring data to assess the validity of modelled slope behaviour.
1 INTRODUCTION To effectively manage hazards associated with massive, slow moving landslides, it is necessary to understand geomechanical factors controlling slope kinematics. These factors, including material strength, slope geometry and groundwater conditions, are rarely homogeneous across the extent of a landslide mass and usually change over time. Detailed site investigation is required for thorough landslide analysis. This should include studies of site specific geology, geomorphology and hydrogeology, as well as slope monitoring to assess howdifferent regions of a massive landslide exhibit spatially discriminated magnitude and direction of deformation, as well as modes of instability. Based on detailed interpretations of site specific conditions sophisticated three-dimensional numerical models can be developed, and then trained to reproduce observed slope behaviour. Once models are calibrated to reproduce observed slope deformations, mitigation techniques such as slope drainage can then be numerically tested.