Formatted Title

Efficient Reactive Transport Simulations of Tracer Studies at a Former Uranium Mill Site Using PHT-USG

Background/Objectives

The U.S. Department of Energy Office of Legacy Management (LM) oversees long-term monitoring at former uranium mill sites across the United States. Despite removal and disposal of the original tailings pile contaminant sources, persistent uranium plumes remain in contaminated aquifers at many of these sites. Improved predictions of plume attenuation and remedial time frames at these sites requires more accurate simulations of the processes controlling uranium behavior in groundwater. Estimation of the governing reactive transport parameters requires geochemical simulation capabilities that go beyond the traditional linear sorption isotherm commonly applied in solute transport models. Reactive transport models (RTMs) can explicitly account for nonlinear sorption caused by shifts in geochemical conditions, which can occur during flooding events or injection of remedial fluids; however, RTMs are notorious for lengthy run times that can become unmanageable in a parameter estimation context, where hundreds or thousands of model runs are required. LM recently funded development of PHT-USG to address this challenge by taking advantage of efficient unstructured grids, which can be designed to reduce model run times by limiting the number of computational nodes outside of reactive zones. PHT-USG combines the popular geochemical simulation code PHREEQC with the 3D groundwater flow and solute transport capabilities of MODFLOW-USG-Transport for unstructured grids. In this study, PHT-USG was applied to simulate tracer testing that was performed at a former uranium mill site in Grand Junction, Colorado, to estimate RTM parameters for a contaminated aquifer.

Approach/Activities

Simulations of four push-pull tracer studies, two cross-hole tracer tests, and one infiltration test were conducted using PHT-USG. A separate RTM was set up to simulate each test. Nested grids were used to refine node spacing and layering in the test area, while a single-layer coarse grid was used throughout the rest of the model domain. The models were calibrated in two steps. First, physical parameters, including hydraulic conductivity, dispersivity, and effective porosity, were estimated by matching added nonreactive tracer curves. During the second step, the physical parameters were held constant while reactive transport parameters were adjusted to match reactive ions and uranium data. A generalized surface complexation modeling approach was used to simulate uranium (de)sorption and estimate sorption site densities, the primary parameters of interest. Ion exchange site densities, calcite saturation index, and gypsum saturation index were also estimated. The PEST suite of software was used to automate the parameter estimation process. For the multi-well tests, the iterative ensemble smoother implemented in PEST++ was used for parameter estimation and uncertainty analysis to better account for aquifer heterogeneity.

Results/Lessons Learned

Estimated parameter values from the RTMs were similar to prior estimates from laboratory column testing, except for a lower uranium sorption potential in one cross-hole test. This difference is likely due to the larger scale of the cross-hole testing including permeable pathways with a lower uranium sorption potential. Uranium sorption parameters were sensitive to uranium concentrations during the drift phase of the push-pull tests, with possible kinetic rate limitations. The infiltration test provided estimates of uranium release characteristics from the vadose zone to the saturated zone during river flooding. The RTM parameters estimated through this effort will provide initial parameter values for future sitewide modeling.