Performance Assessment of a Low-Level Radioactive Waste Disposal Facility in Texas

Performance assessment of radioactive waste repositories requires simulation of radionuclide transport through all potential pathways to demonstrate safety to the public and compliance with regulations. The performance period often involves predictions thousands of years into the future. Given the uncertain nature of surface and subsurface mass transport processes and future conditions, a typical performance assessment process must evaluate a large number of scenarios and quantify the impact of various model structural and parametric uncertainties in a Monte Carlo framework. The computational cost associated with such uncertainty analyses can be prohibitive if complex, process-level models are directly used in all evaluations. Therefore, various model reduction methodologies have been developed and applied in system performance assessment, including model abstraction and response surface methods. The response surface methods aim to find a surrogate and, hopefully, simpler relationship between system inputs and responses. In this work, we use the probabilistic collocation method (PCM) to construct response surfaces for performance assessment of a low-level radioactive waste facility located in semi-arid west Texas. PCM is a nonintrusive, stochastic spectral method that finds the unknown coefficients of a response surface via a set of model simulations. The resulting stochastic response surface can then be used in uncertainty quantification of solute breakthrough fluxes, which in turn, are used for dose estimates. The number of runs required by PCM can be much smaller than that required for Latin Hypercube Sampling, an uncertainty quantification technique often used in performance assessments. In this study, we consider a number of uncertain parameters in the coupled, unsaturated flow and transport models. Because of dry conditions and low permeability values, the flow and transport problems are highly nonlinear and require extensive simulation time. The PCM thus offers an efficient alternative to Monte Carlo simulation in our performance assessment.