Analysis of rainfall-induced slope instability using a field of local factor of safety

Abstract

Slope-stability analyses are mostly conducted by identifying or assuming a potential failure surface and assessing the factor of safety (FS) of that surface. This approach of assigning a single FS to a potentially unstable slope provides little insight on where the failure initiates or the ultimate geometry and location of a landslide rupture surface. We describe a method to quantify a scalar field of FS based on the concept of the Coulomb stress and the shift in the state of stress toward failure that results from rainfall infiltration. The FS at each point within a hillslope is called the local factor of safety (LFS) and is defined as the ratio of the Coulomb stress at the current state of stress to the Coulomb stress of the potential failure state under the Mohr-Coulomb criterion. Comparative assessment with limit-equilibrium and hybrid finite element limit-equilibrium methods show that the proposed LFS is consistent with these approaches and yields additional insight into the geometry and location of the potential failure surface and how instability may initiate and evolve with changes in pore water conditions. Quantitative assessments applying the new LFS field method to slopes under infiltration conditions demonstrate that the LFS has the potential to overcome several major limitations in the classical FS methodologies such as the shape of the failure surface and the inherent underestimation of slope instability. Comparison with infinite-slope methods, including a recent extension to variably saturated conditions, shows further enhancement in assessing shallow landslide occurrence using the LFS methodology. Although we use only a linear elastic solution for the state of stress with no post-failure analysis that require more sophisticated elastoplastic or other theories, the LFS provides a new means to quantify the potential instability zones in hillslopes under variably saturated conditions using stress-field based methods.