Comparison of Three Numerical Models for Chain-Decay Transport Simulation at a Closed AFB in Texas

Numerical schemes of three codes, MODFLOW-SURFACT, MT3D99, and RT3D, were compared for simulation of chain decay of PCE at a decommissioned Air Force Base in central Texas, to explore the implications of simulation code and solution scheme selection. The flow field generated by MODFLOWSURFACT was used by MT3D99 and RT3D for transport simulation. Nine different transport models were developed incorporating three solution methods in combination with different time-stepping schemes. The solution methods examined within MODFLOW-SURFACT were TVD/Crank-Nicolson, upstream finitedifference/Crank-Nicolson and upstream finite-difference/fully implicit-in-time weighting. Upstream weighting and fully implicit finite-difference schemes were used for inter-code comparisons. The results obtained with MT3D99 were closer to MODFLOW-SURFACT than to RT3D. RT3D incorporates the implicit assumption that decay reactions occur only in the dissolved phase so it predicts slower overall degradation. The results of the sensitivity analysis confirm that MODFLOW-SURFACT is capable of reproducing results obtained with public-domain transport simulators, and offers increased flexibility in the use of solution methods and time-stepping schemes. INTRODUCTION Biodegradation modeling is a useful tool for predicting the attenuation of organic contaminants in groundwater. Currently available MODFLOW-format numerical models for the simulation of a straight chain biodegradation of chlorinated ethenes (PCE→TCE→DCE→Vinyl Chloride) include MODFLOWSURFACT (HydroGeoLogic, 1998), MT3D99 (SSPA, 2000), and RT3D (Clement, 1997). The unsaturated zone over the study area of the decommissioned base is highly complex; therefore, MODFLOW-SURFACT was used to simulate groundwater flow because it can obtain physically realistic solutions without encountering numerical convergence problems with drying and re-wetting. MODFLOWSURFACT is a fully integrated groundwater flow and solute transport code developed as an extension to MODFLOW (McDonald and Harbaugh, 1988). Since the source code of MODFLOW-SURFACT is not publicly available, it is valuable to examine its performance on real site simulations against other transport simulators that are in the public domain. For this study, transport results obtained with MODFLOWSURFACT were compared with the results from MT3D99 and RT3D (version 2.5), using a flow-field generated by MODFLOW-SURFACT. This study is based on the Zone 4 off-base groundwater flow and transport zoom model for a decommissioned AFB developed by HydroGeoLogic, Inc. The model was extracted from the March 2001 basewide flow model and used to evaluate different remediation alternatives for the removal of the chlorinated solvents tetrachloroethene (PCE), trichloroethene (TCE), 1,2-dichloroethene (DCE) and vinyl chloride (VC). The objective of the analyses was to estimate the time required to meet the maximum contaminant levels (MCLs) mandated under federal drinking water standards. The MCLs for PCE, TCE, 1,2-cis DCE and VC are 5 parts per billion (ppb), 5 ppb, 70 ppb and 2 ppb, respectively. MODFLOW 2003 Conference, Golden, Colorado, September 17-19, 2003 Page 2 PROBLEM DESCRIPTION Figure 1 displays the numerical grid for the zoom model overlain on the basewide flow model. The surface area of the zoom model is approximately 28.6 square miles. The zoom model has a minimum grid spacing of 50 feet, occurring in the vicinity of the remediation system (pumping wells) and consists of 220 rows and 241 columns in each of 4 model layers. A period of 15 years is considered in the transport simulations. Prior to conducting simulations with MT3D99 and RT3D, several simulations were conducted with the various transport options within MODFLOW-SURFACT to better understand the factors controlling model performance. For spatial discretization of the advection term, upstream weighting, midpoint weighting, or the Total Variation Diminishing (TVD) method may be used. These schemes behave as follows: upstream weighting schemes tend to disperse sharp advective concentration fronts; midpoint weighting may create unphysical oscillations; and recently developed TVD schemes provide improvements to both problems but are computationally more demanding. For the temporal discretization of the transport equations, explicit, Crank-Nicolson (also called centeredin-time weighting) and implicit schemes can be used. Explicit time-weighting schemes are subject to stringent stability criterion that must be satisfied to ensure stable solutions. Although both Crank-Nicolson and fully implicit schemes are not constrained by stability criteria, the selection of time-step size may affect the accuracy of the simulations. In the current evaluation, TVD and upstream weighting schemes were selected for spatial discretization, and Crank-Nicolson and fully implicit schemes were chosen for temporal discretization. Nine different transport models were developed incorporating three solution methods in combination with time-step sizes of 2, 20, and 200 days. The solution methods examined were TVD/Crank-Nicolson, upstream finitedifference/Crank-Nicolson and upstream finite-difference/fully implicit-in-time weighting. EASTING (ft) N O R TH IN G (ft ) 213000