Probabilistic TSF stability assessment using a NorSand-based random-field framework

Tailings materials are inherently heterogeneous systems where deposition processes produce layering, segregation, and spatial variability in density and structure. Conventional analyses often neglect this spatial variability and instead rely on a single deterministic representation of in-situ state. This study presents a practical framework to assess tailings storage facility (TSF) stability by integrating random field theory with the NorSand model, which explicitly links strength, dilatancy, and static liquefaction susceptibility to the state parameter (ψ). Unlike prior works that approximate state dependence indirectly by randomizing shear strength, the present work models the initial state parameter (ψ_0) itself as a spatially correlated random field with a depth-dependent mean profile, reflecting depositional variability while keeping the NorSand constitutive parameters fixed. Random-field correlation lengths and anisotropy are adopted from CPTu-based spatial variability studies on tailings deposits while the NorSand parameters are calibrated against published experimental response data using an element-level implementation developed in MATLAB. The ψ_0 field is generated through Karhunen-Loève expansion and propagated through Monte Carlo effective-stress TSF stability analyses in PLAXIS 2D. Results show that ψ_0 heterogeneity creates zones with different degrees of contractive response, which leads to localized pore-pressure build-up and deformation. As a result, excess pore-pressure response and the predicted failure mechanism vary across realizations, rather than remaining confined to a single deterministic prediction. The probabilistic workflow provides a robust, data-driven pathway toward performance-based TSF assessment and strengthens mine-closure decisions for long-term stability under depositional uncertainty.