Copula-based conditional reliability analysis of slopes in spatially variable soils

Cross-correlated conditional random fields (CRFs) can better characterize soil spatial variability by incorporating multivariate site investigation data. However, existing studies often assume a Gaussian copula dependence structure for strength parameters in CRFs, which may not hold in practice. This study advances the understanding of how non-Gaussian copula dependence structures between cross-correlated CRFs influence slope reliability analysis. A combined approach of the multiple response surface method and subset simulation is employed to efficiently compute the probability of failure (Pf) of slope stability. Using a typical cohesive-frictional soil slope as an illustrative example, parametric studies are conducted to investigate the effects of non-Gaussian copula dependence structures on the factor of safety (FS), critical slip surface (CSS) and Pf. The results show that the commonly used Gaussian copula-based CRFs may underestimate the Pf, and such situation becomes the most severe when the underlying dependence structure follows the No.16 copula. Compared with copula-based unconditional random fields, the copula-based CRFs can more efficiently reduce the uncertainty of the FS and spatial distribution of CSS. This study provides practical guidance for copula selection in conditional slope reliability analysis leveraging multivariate site data, addressing a critical gap in geotechnical reliability analysis and risk assessment.