Sensitivity Analysis of Physically-based 3D Landslide Susceptibility Model from Variation of Input Parameters

Shallow landslides frequently occur on natural slopes and cause flow-like disasters. The authors have previously developed 3-Dimensional Translational Slide (3DTS), a physically-based 3D shallow landslide susceptibility model accounting for side resistance and vegetation effects, to efficiently evaluate the slope stability in terms of the factor of safety (FS) over a regional scale. Traditionally, a deterministic slope stability analysis was performed by assigning representative values to rainfall history, soil layers, and soil properties; however, new design standards demand reliability-based analyses that account for the uncertainty and variation in precipitation, subsurface conditions, soil hydro-geotechnical properties, and vegetation root reinforcement. Therefore, this research proposes extending the developed model into a 3-Dimensional Translational Slide-Probabilistic (3DTSP) model to enable reliability-based landslide susceptibility assessment. The developed 3DTSP model combines the generalized Green-Ampt infiltration model and the 3D Janbu simplified slope stability model. The 3D slope stability analysis accounts for additional soil frictional resistance at the side regions in translational slides and additional reinforcements from tree roots. The 3DTSP model uses a Monte Carlo simulation with a random-field approach to determine the FS statistical distribution from variations in the following input parameters: soil thickness, hydraulic properties, Mohr-Coulomb criterion-based shear strength properties, unsaturated soil strength properties, and vegetation resistance properties. Based on the statistical distribution and characteristic length, the model generates a random field of input parameters that accounts for spatial variation in the horizontal direction. For each Monte Carlo simulation iteration, a new random input field is generated to compute FS. The performance and applicability of the developed 3DTSP for probabilistic assessment of landslide susceptibility over regional scales were demonstrated by analyzing landslide case studies. A sensitivity study was conducted to assess the sensitivity of FS to variations in soil thickness, soil properties, and vegetation properties.