Evaluating risk in reservoir accumulation landslide: the role of instability pattern uncertainty

Reservoir accumulation landslides pose a significant challenge to hydropower projects, with their instability risks exacerbated by uncertainties in failure patterns due to the spatial variability of geo-material properties and dynamic environmental conditions. This study pioneers an in-depth investigation into the instability pattern uncertainty of reservoir accumulation landslides, a critical yet understudied aspect of landslide risk, exemplified by the Baijiabao landslide in the Three Gorges Reservoir (TGR) area. A stochastic finite element limit analysis (FELA), integrating comprehensive field investigations and laboratory experiments, was conducted to model the spatial variable shear strength in the sliding mass and slip zone. Employing the Gaussian Mixture Model-Expectation Maximization (GMM-EM) algorithm, the risk associated with instability patterns was identified and quantified, revealing a significant influence of seismic activity over water level on failure probabilities. The findings delineate two primary instability patterns of reservoir accumulation landslides: deep-seated instability along the slip zone and localized shallow instability within the sliding mass, all of which initiate at the leading edge, manifesting a sequence of retrogressive and progressive failures. This study offers a novel approach to understanding instability pattern uncertainty and establishes a feasible framework for risk assessment of reservoir landslides.