Bayesian identification of the optimal soil-water characteristic curve (SWCC) model and reliability analysis of unsaturated loess slope from extremely sparse measurements

Abstract

Soil-water characteristic curve (SWCCs) are crucial in engineering geology and geotechnical engineering for understanding the behavior of unsaturated soils, such as loess, which directly affects permeability, shear strength, and volume change-key factors in slope stability and soil-structure interactions. Conventionally, SWCC estimation relies on multiple (saying approximately ten) measurements fitted to parameterized models. However, in practical applications, especially for medium- or small-scale projects, the availability of SWCC measurements is often extremely limited (e.g., one or two measurements) due to the time-intensive nature of the experiments. This presents significant challenges in accurately identifying suitable SWCC models and performing reliable stability analyses for unsaturated soil slopes. To address these challenges, this study employs a hierarchical Bayesian framework that integrates information from similar geotechnical site, enabling robust SWCC estimation and model selection from minimal measurements with the aid of Markov Chain Monte Carlo (MCMC) sampling, thereby quantifying model uncertainties and providing more scientifically informed decision-making for construction in engineering geology. MCMC samples obtained further facilitate both the identification of the most suitable SWCC model and the quantification of associated uncertainties. Then, a reliability-based stability analysis of an unsaturated loess slope is conducted, using the optimal SWCC model and its quantified uncertainty. The proposed methodology is validated through a real-world case study, demonstrating its effectiveness in deriving reliable SWCC models and performing stability analyses under conditions of extremely sparse data. The results highlight the potential of this method as a practical tool for advancing reliability assessments of unsaturated soil slopes in engineering geology.