Geostatistical and Stochastic Study of Flow and Transport in the Unsaturated Zone at Yucca Mountain

Abstract

Yucca Mountain has been proposed by the U.S. Department of Energy as the nation’s long-term, permanent geologic repository for spent nuclear fuel or high-level radioactive waste. The potential repository would be located in Yucca Mountain’s unsaturated zone (UZ), which acts as a critical natural barrier delaying arrival of radionuclides to the water table. Since radionuclide transport in groundwater can pose serious threats to human health and the environment, it is important to understand how much and how fast water and radionuclides travel through the UZ to groundwater. The UZ system consists of multiple hydrogeologic units whose hydraulic and geochemical properties exhibit systematic and random spatial variation, or heterogeneity, at multiple scales. Predictions of radionuclide transport under such complicated conditions are uncertain, and the uncertainty complicates decision making and risk analysis. This project aims at using geostatistical and stochastic methods to assess uncertainty of unsaturated flow and radionuclide transport in the UZ at Yucca Mountain. Focus of this study is parameter uncertainty of hydraulic and transport properties of the UZ. The parametric uncertainty arises since limited parameter measurements are unable to deterministically describe spatial variability of the parameters. In this project, matrix porosity, permeability and sorption coefficient of the reactive tracer (neptunium) of the UZ are treated as random variables. Corresponding propagation of parametric uncertainty is quantitatively measured using mean, variance, 5th and 95th percentiles of simulated state variables (e.g., saturation, capillary pressure, percolation flux, and travel time). These statistics are evaluated using a Monte Carlo method, in which a three-dimensional flow and transport model implemented using the TOUGH2 code is executed with multiple parameter realizations of the random model parameters. The project specifically studies uncertainty of unsaturated flow and radionuclide transport caused by multi-scale heterogeneity at the layer and local scales. Typically, in studies of Yucca Mountain, the layer scale refers to hydrogeologic layers with layer-wise average properties, and the local scale refers to the spatial variation of hydraulic properties within a layer. While most studies of radionuclide transport in the UZ have been conducted at the layer scale, the uncertainty at the local scale within a layer is also important, since it affects flow path, velocity, and travel time of radionuclide. This report first presents the uncertainty caused by layer-scale heterogeneity of matrix permeability, porosity, and sorption coefficients of reactive tracers. Homogeneous fields of the parameters are generated at each hydrogeologic layer for Monte Carlo simulations. This study is referred to as the homogeneous case. To assess the uncertainty caused by local-scale heterogeneity, the sequential Gaussian simulator (SGSIM) of GSLIB (Deutsch and Journel, 1998) is used to generate heterogeneous parameter fields within each layer, and Monte Carlo simulations are conducted. This study is referred to as the heterogeneous cases. For the homogeneous and heterogeneous cases, the mean, variance, 5th and 95th percentiles of simulated state variables are estimated for uncertainty assessment. In addition, the statistics of the two cases are compared to investigate effect of local-scale heterogeneity on the unsaturated flow and radionuclide transport. It is found that the local-scale heterogeneity increased the predictive uncertainty of percolation flux and cumulative mass arrival for computational blocks below the footprint of proposed repository, whereas mean predictions are hardly affected. The local-scale heterogeneity significantly affects travel times to the water table for both conservative and reactive tracers. In the early simulation period, tracer mean travel times are delayed, whereas the influence of local-scale heterogeneity diminishes during the late simulation period. Simulated state variables in this project are more realistic than those of using one- or two dimensional models, due to a three-dimensional numerical model used in the project to characterize hydrological conditions at the UZ. Therefore, we expect that results of this project can be used directly to facilitate DOE site performance analysis and decision making.