Exploring the stability of unsaturated soil slope under rainfall infiltration conditions: a study based on multivariate interrelated random fields using R-vine copula

• Published in: Catena (Giessen) 2024-01, Vol.234, p.107587, Article 107587
• Main Authors: Xu, Binni, Pei, Xiangjun, Li, Jingji, Yang, Hailong, Wang, Xinqing
• Format: Article
• Language: English
• Subjects: Cross-related random fields; Probability of failure; R-vine copula; Rainfall-induced soil slope instability; Safety factor
• ISSN: 0341-8162; 1872-6887
• DOI: 10.1016/j.catena.2023.107587

[Display omitted] •A general framework for coupled hydro-mechanical modeling of rainfall-induced instability in unsaturated slopes with multivariate random fields is developed. And the R-vine copula is introduced to simulate the intricate non-Gaussian dependencies between soil parameters;•The Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction technique is employed to visualize the fitted dependency structures, and LOWESS Regression is used to compare the fitting performances of R-vine and Gaussian copulas;•The R-vine copula model proficiently captures the non-Gaussian dependencies in soil data, demonstrating superior fitting performance over the Gaussian copula.•The deterministic model for the slope stability evaluation may underestimate the failure possibility;•Groundwater is the primary driver for slope instability under rainfall infiltration conditions;•As rainfall continues, the overall re-balancing triggered by the displacement of the sliding body enhances the contribution of the sliding volume to the stability of the spoil heap slope. Estimating the instability of soil slopes due to rainfall in highly heterogeneous materials poses a considerable challenge. In this study, a general framework for coupled hydro-mechanical modeling of rainfall-induced instability in unsaturated slopes with multivariate random fields is developed. The R-vine copula is introduced to simulate the intricate non-Gaussian dependencies between soil parameters. UMAP is utilized for visualizing these dependencies. The study compares the fitting performances of R-vine and Gaussian copulas using LOWESS Regression. Subsequently, deterministic model computations are conducted in Abaqus, along with batch random field model analysis based on the R-vine copula. The instability probability of soil slopes under rainfall infiltration conditions is evaluated through direct Monte Carlo simulations, and statistically investigate the relationships between groundwater level, sliding volume, plastic zone volume, and safety factor. The findings indicate that: 1) The R-vine copula model proficiently captures the non-Gaussian dependencies in soil data, demonstrating superior fitting performance over the Gaussian copula; 2) deterministic simulations might overestimate the safety factor at certain instances; 3) as rainfall progresses, a growing negative correlation is observed between groundwater levels and slope instability; 4) continuous rainfall leads to a re-equilibrati