Effect of soil spatial variability on simplified seismic slope displacement assessments

Simplified procedures for evaluating seismic displacements of engineered and natural slope systems depend on the dynamic resistance and seismic loading of the potential sliding mass. Typically, the dynamic resistance is proxied by a seismic yield coefficient, and the seismic loading is proxied in part by the spectral acceleration at the degraded fundamental period of the slide mass. However, inherent soil spatial variability within slope systems can significantly contribute to uncertainties in these variables, which can propagate to uncertainties in the evaluated seismic displacements. This study investigates the influence of subsurface soil variability on the resulting uncertainty in the seismic yield coefficient, slide mass period, and estimated displacements for a series of hypothetical clay slopes. Pseudostatic stability analyses were performed using the finite difference program FLAC for approximating the seismic yield coefficient, the associated slide mass geometry, and slide mass period. Seismic displacements were determined using a state-of-practice simplified procedure for 40 ground motions. The soil was modeled using lognormal spatial random fields of the undrained shear strength, with alternative assumptions for the mean, coefficient of variation, and horizontal correlation range. Other parametric variations included alternative assumptions for the shear wave velocity, soil shear strength characterization, and slope geometry. The results demonstrate important considerations for interpreting the combined uncertainty (i.e., inherent soil variability and associated biases, compounded with modeling uncertainty) for practical evaluations of seismic displacements. Implications for incorporating ground motion uncertainty and using these findings within performance-based models are discussed. • Spatial variability can largely influence probabilistic seismic slope displacements. • Uncertainty metrics and workflow are proposed for probabilistic methods. • Representative percentiles suggested for uniform slopes to match variable slopes. • Uncertainty increases for individual response spectra relative to smoothed spectra. • Slide mass period to yield coefficient correlation is sensitive to many parameters.