Predicting Upscaling Relationships for Heterogeneous Flow and Reactive Transport at the Savannah River Site

This study aims at understanding key hydrogeochemical processes dictating pH behavior and U transport at the Savannah River Site (SRS) F-Area, South Carolina, with particular focus on the impact of chemical and physical heterogeneities. Acidic waste solutions containing low level radioactivity from numerous isotopes were discharged to a series of unlined seepage basins at the F-Area, from 1955 through 1989, which resulted in a nearly 1 km long acidic uranium plume. Reactive facies is a new approach that spatially characterizes linked flow and geochemical properties over large domains, where it is typically challenging to obtain parameters with sufficient resolution for reactive transport modeling. This approach - based on the hypothesis that we can identify geological units that have unique distributions of reactive transport properties - allows us to integrate various types of datasets (e.g., historical data, laboratory analysis, crosshole and surface geophysics) for estimating heterogeneous reactive transport parameters. At the SRS F-Area, data mining and iteration with laboratory analysis identified two reactive facies coincident with the depositional facies, which have distinct distributions of reactive transport properties: %fines, permeability, and Al:Fe ratio (proxy for kaolinite:geothite ratio). The reactive facies over the plume-scale domain was estimated based on measured data (foot-by-foot core analysis, cone penetrometer, crosshole seismic and surface seismic data) and integrated using the Bayesian framework. In parallel, a numerical reactive transport model was developed including saturated and unsaturated flow, and complex geochemical processes such as U(VI) and H+ adsorption (surface complexation) onto sediments and dissolution and precipitation of Al and Fe minerals. By combining the developed reactive transport model with the estimated spatial distribution of reactive transport parameters, we perform stochastic simulations of U and pH plume evolution at the F-Area. The random reactive facies fields sampled from the estimated posterior distribution are used to distribute the reactive transport parameters (i.e., permeability, porosity, reactive surface area, mineral ratio). We explore the optimal level of representing heterogeneity needed for reliable predictions of contaminant mobility at plume scales over long time frames as well as the computational burden associated with it. We compare the results from different cases with various complexities such as (1) homogeneous properties in each formation, (2) heterogeneous mean field of each parameter, and (3) multiple random fields of either flow or geochemical parameters. The results show that it is necessary to have stochastic simulation using random fields (not just a mean field) for accounting the effect of heterogeneity, although it is computationally intensive, and also to have the linkage between flow and geochemical parameters. In the results, because sorption sites become quickly saturated by the massive H+ and U influx, heterogeneous surface areas (within a given facies) affect the predicted pH and U transport mainly at early times and at the plume edge. Heterogeneities are, therefore, expected to become relevant over the long term when contaminant concentrations have decreased below sorption saturation levels.