Conditional random field-based stochastic analysis of unsaturated slope stability combining Hoffman method and Bayesian updating

This paper proposes a new framework for discretizing the conditional random fields that combines the Hoffman method and Bayesian updating, where multiple types of geotechnical data can be utilized, such as the field monitoring data and the site investigation geological borehole data. The feasibility of the proposed method is demonstrated using a rainfall-induced slope example in spatially varying soils. Results showed that the proposed framework of incorporating Bayesian updating to Hoffman method can effectively reduce the uncertainty in site characterization and guarantee precise risk assessment of unsaturated slope stability. Otherwise, the slope failure probability and landslide sliding volume would be overestimated when only the Hoffman method is adopted for site characterization. Subsequently, the developed method is further used to estimate the probabilistic stability of Baishuihe landslide in the presence of rainfall infiltration and reservoir water level fluctuation. It is hoped that this study can provide a reference for risk assessment of the unsaturated slope stability in reservoir areas.