Develop predictive model for contaminant transport that incorporates complex biology, hydrology, and chemistry of the subsurface. Validate model through field
New information on biogeochemistry, groundwater and subsurface media from the Oak Ridge Field Research Center (FRC) was utilized to update and run large-scale 3-D flow and chemical transport models. The FRC is a focal point for ERSD (now CESD) program field research on natural and stimulated biologically-mediated attenuation of metals and radionuclides. Information from these studies on the interactions among biological, chemical and physical processes in a real-world setting with complex hydrogeology and contaminant characteristics promises to significantly improve DOE’s ability to effectively manage legacy waste sites. Computer models are important tools for interpreting field and laboratory data from this complex system to understand the nature of process interactions, to help guide ongoing and future research efforts, and to enable predictions of plume-scale behavior in response to various remediation/management strategies.
Field-scale modeling efforts are being conducted by ORNL, PNNL, Oregon State University and Stanford University researchers on experimental plots within the FRC and at a larger scale that encompasses all of the field plots to better understand their interactions and large-scale behavior. These efforts have utilized the code HYDROGEOCHEM Version 5, which simulates three-dimensional transient density-dependent, fully-anisotropic saturated and unsaturated water flow, dissolved transport, and complex equilibrium- kinetically-limited biogeochemical reactions. Work was undertaken to develop a site-wide 3D flow and transport model encompassing an area of approximately 280 acres that includes FRC Areas 1, 2 and 3, the former S3 Ponds and the Bear Creek watershed from its head waters to the NT2 tributary. Refinements in the model have been completed to incorporate new data and improvements in the conceptual model of the site. The refinements include:
The refined site-wide FRC model is discretized into 8 layers of 23,967 elements and 10 layers of 13,680 nodes. Four types of bedrock, including a “potential high permeability zone” identified by geophysical testing, are overlain by saprolite, gravel fill, a permeable trench and a rock saprolite “transition” zone (Figure 1). Nonlinear optimization using the PEST code was performed to recalibrate a steady-state groundwater flow model to stream gauging data and water levels from 74 wells. Rock and saprolite were modeled as anisotropic media with a maximum permeability along strike, minimum permeability in the cross-bed direction and intermediate permeability along the dip direction. Field-scale dispersivities and effective porosities were manually calibrated to measured nitrate concentrations from 19 wells within and near the dissolved plume by simulating non-reactive nitrate plume evolution from the S3 Ponds considering density-dependent flow effects. Dissolved nitrate plume predictions are shown in Figure 2. Sensitivity analyses were performed. A higher resolution model was developed for Area 3 to facilitate interpretation of in situ Area 3 flow and tracer studies in the vicinity and to design subsequent experiments (Figure 3). A sub-model domain was delineated and a numerical mesh developed to accommodate the finer resolution needed to simulate flow and transport within Area 3. Boundary conditions on the sub-model domain were obtained from the site-wide model. Material properties were initially mapped from the site-wide model, and then refined to account for finer level details relevant to the experimental data interpretation. The fine-scale model will be utilized to assess potential interactions between current experiments in Area 3 and proposed new studies and to adjust experimental plans if necessary to avoid adverse effects on neighboring field plots. Simulated breakthrough curves for a tracer test in FW106 are shown in Figure 4.A high resolution model was also developed for Area 2 to help design and interpret field-scale experiments in this area (Figure 5). The 800 m3 model domain encompasses the disturbed fill, coarse gravel, and intact saprolite zones. The model considers pulsed injection of tracers and ethanol in three wells (FW213, FW212, FW214) and simulates 94 chemical species, 8 biomass populations, 58 equilibrium reactions, 77 kinetic reactions, and 37 terminal electron accepting reactions. Reactions and reaction coefficients and aquifer properties were based on data from field and laboratory studies. Figure 6 shows comparisons of model predictions and observations at 2.5 m (MLS-A, MLS-B, FW216) and 7.5 m (FW202) downstream of the injection wells during the first week. Generally, the model predictions and observations match reasonably well given that no reaction parameter fitting was performed for the field observations. The timing and rates of utilization of the various terminal electron acceptors is approximately correct in the simulation results. The primary mismatches occur at MLSA-4 and MLSA-5, ports at which the bromide simulation very poorly matches the observed concentrations. This suggests that incorrect representation of the flow pathway, probably associated with uncharacterized physical heterogeneity, is the cause of the mismatch rather than improper modeling of the reaction system and rates.
Prepared by: Jack Parker (ORNL) with contributions from Fan Zhang (ORNL), Yilin Fang (PNNL), Dave Watson (ORNL), Tim Scheibe (PNNL), Wiwat Kamolpornwijit (PNNL), Eric Roden (Univ. Wisconsin), Scott Brooks (ORNL)
Figure 2. Simulated dissolved phase plume for site-wide model. Top: north-south transect through S3 Ponds; Middle: horizontal distribution within upper aquifer; Bottom: horizontal distribution within lower aquifer.
November 2019 [PDF]