Stability analysis of reservoir slope under water-level drawdown considering stratigraphic uncertainty and spatial variability of soil property

Time-dependent stability analysis of reservoir slopes during water level drawdown is crucial. Due to the sparse investigation, the stability analysis often contains geological uncertainty (e.g., stratigraphic uncertainty) and geotechnical uncertainty (e.g., spatially varying soil properties). Nevertheless, their coupling effect on the reservoir slopes has not been well considered in the past. A data-driven probabilistic framework, capable of simulating stratigraphic uncertainty and spatial variability of soil property directly from limited investigation, is proposed to incorporate the uncoupled finite element method and Markov Chain Monte Carlo simulation for slope analysis under rapid drawdown. A typical reservoir slope with a real-world piezocone penetration test database is adopted to illustrate the framework's effectiveness. The applications showcase the framework's capability in accurately simulating geological cross-sections and spatially varying soil properties, eliminating the need for parametric estimation. The coupling effects of both uncertainties and drawdown velocity are investigated through comparative analyses. It's found that neglecting stratigraphic uncertainty and spatial variability of soil properties leads to considerable discrepancies in time-dependent stability, as it overlooks possible sliding surfaces and underestimates pore pressure accumulation. This highlights the importance of analyzing reservoir slope stability using soil profiles with quantified uncertainty as close as possible to the actual ones.