A Novel Framework for Reliability Analysis of Rainfall‐Induced Slopes Failure Based on Site‐Specific Data

ABSTRACT

Rainfall‐induced slope failure is governed by complex multivariate interactions and inherent spatial variability of soil properties. However, conventional Gaussian‐based reliability models fail to capture the nonlinear dependencies between hydraulic and mechanical parameters and cannot fully utilize limited site‐specific data. This paper develops a Copula‐based conditional random field (Copula‐CRF) framework that explicitly incorporates multivariate non‐Gaussian dependencies and maximizes the use of site‐specific data in probabilistic slope stability evaluation. The framework first identifies optimal Copula structures for multivariate parameters based on site‐specific data and then embeds the identified dependence models into a conditional random field to generate spatially correlated realizations constrained on both hydraulic and mechanical observations. A representative slope case exhibiting failure is analyzed to evaluate the performance of the proposed approach. The Copula‐URF approach yields a failure probability of 0.22, whereas the Copula‐CRF approach yields 0.911, accompanied by a reduction in standard deviation from 0.207 to 0.023. The results indicate that the Copula‐CRF framework provides more realistic reliability estimates than Copula‐URF approaches by fully utilizing available site‐specific data. The performance of the proposed framework is evaluated under varying parameter uncertainties, spatial correlations, rainfall patterns, and site investigation schemes, confirming the robustness and applicability of the approach. Applications of the proposed Copula‐CRF framework demonstrate that it can reliably evaluate the stability of rainfall‐induced slopes with site‐specific data. The method also provides a general reference for reliability analysis and risk‐informed decision‐making in geotechnical engineering.