A 60-year-old closed tailings storage facility (TSF) in a semi-arid climate was re-evaluated for embankment slope stability. The TSF has not completely drained down, and contains perched saturation of low-lying area of the pre-TSF topography, which may impact slope stability. Compounding the problem is the formation of a large lake at the beach of the TSF following flood events, and occasional flooding of a creek near the toe of the TSF embankment. These flooding events lead to temporary recharge of the aquifer beneath the TSF and potential rise of the water table to the bottom of the TSF.
Geotechnical design of buttress support for the TSF by another consultant assumed a worst-case scenario of 20-foot rise in water levels within the TSF near the embankment, an assumption that was challenged by the TSF owners. In addition, fluxes that needed to be accommodated by buttress underdrains were uncertain. However, it was difficult to objectively come up with a reasonably-conservative model to guide the design, as opposed to an unrealistically-conservative model that would suggest need for further buttress reinforcement and investment based on an unlikely scenario.
To aid in this decision, SRK constructed a cross-sectional model of the TSF and underlying aquifer, deliberately designed to be a light-weight model that would accommodate uncertainty analysis. The model was run thousands of times in order to explore a wide range of hydraulic parameter values. Acknowledging that a model is not an exact replica of reality, each model run was ranked based on its ability to match historical data (previous flooding events), thus allowing to constrain the range of reasonable hydraulic parameter values. Then, the model was run again multiple times in predictive mode, this time adding duration of flooding events to the parameter space. The probability of future flood events to cause a water level rise of 20 ft was then explored in context of the best-performing models in terms of history matching, as opposed to arbitrary assumption of “worst case” around a single “calibrated” model. Considering parameter configurations that reasonably matched historical data, only one predictive scenario resulted in exceedance of the threshold 20 ft. water level rise. Under the predictive model assumptions, buttress underdrain fluxes of up to 800 gpm were calculated.
