Prediction of retrogressive landslide in sensitive clays by incorporating a novel strain softening law into the Material Point Method

Abstract

Sensitive clays, when subjected to large strains, exhibit a unique strain-softening behavior, transforming into a remolded liquid with remarkably low shear strength. When a slope fails, this behavior leads to the remolded clay moving away from its original position, triggering subsequent failures and catastrophic outcomes. To accurately predict such scenarios, it is crucial to incorporate realistic strain-softening characteristics into the constitutive soil model. This paper presents a novel yet practical strain-softening law developed by the authors that effectively captures the post-peak behavior of sensitive clays down to their remolded strength. The softening law is implemented in an elastoplastic Mohr-Coulomb model and incorporated into Anura3D, an open-source software that uses the Material Point Method to simulate large deformations. The constitutive model is calibrated by simulating the stress-strain behavior through direct shear tests conducted at three sensitive clay landslide locations. The accuracy of the overall numerical framework is assessed by predicting the post-failure movements of these landslides. Notably, two of the landslides, Sainte-Monique (1994) and Saint-Jude (2010), have previously been analyzed using other numerical tools, allowing for a comparative analysis with the method presented here. The third landslide, the Saint-Luc-de-Vincennes landslide, representing a composite flow slide and spread, has been numerically simulated for the first time. The post-failure behavior observed in the landslide events is compared with field observations and other numerical analyses. The results show that the MPM models with the proposed strain-softening law can reasonably predict post-failure retrogression and runout distance, which are crucial parameters for determining the risk of landslides in sensitive clays.