815-828 Gw N-d 04

Field-scale dispersion of a conservative tracer is primarily the result of smallto intermediate-scale advective processes resulting from natural heterogeneities rather than pore-scale hydrodynamic dispersion (Guven et al. 1992). Fully incorporating physical heterogeneity into a field-scale model is extremely challenging because subsurface variability is pervasive, being present from pore-scale through regional-scale, and characterization data can only define a small fraction of this variability, especially at smaller scales. Furthermore, computational resources usually limit the model grid resolution to a relatively coarse scale in practical applications. While high-resolution modeling can be technically effective, the large computational resources and sophisticated mathematical/statistical techniques used to generate the random synthetic hydraulic conductivity (K) fields are often not appropriate for routine applications. As a practical alternative, subgrid modeling (hydrodynamic dispersion and/or exchange between zones having large contrasts in K) is typically used to account for heterogeneity at scales smaller than those explicitly represented within the flow model. The dual-domain model (DDM), employing hydrodynamic dispersion and exchange, and the singledomain model (SDM), employing hydrodynamic dispersion only, are well-known examples of subgrid models. The DDM assumes that a porous continuum may be conceptually divided into two interacting subcontinua or domains—a mobile domain where transport is dominated by advection and an immobile domain where advective transport does not occur. Solute exchange between the two domains is characterized by a mass transfer relationship that is assumed to be driven by the concentration difference between the domains. The terms—dual-domain, dualmedia, and dual-porosity—associated with the concept Abstract Subgrid modeling of some type is typically used to account for heterogeneity at scales below the grid scale. The single-domain model (SDM), employing field-scale dispersion, and the dual-domain model (DDM), employing local hydrodynamic dispersion and exchange between domains having large hydraulic conductivity contrasts, are wellknown examples. In this paper, the two modeling approaches are applied to tritium migration from the H-area seepage basins to a nearby stream—Fourmile Branch—at the Savannah River Site. This location has been monitored since 1955, so an extensive dataset exists for formulating realistic simulations and comparing the results to data. It is concluded that the main parameters of both models are scale-dependent, and methods are discussed for making initial estimates of the DDM parameters, which include mobile/immobile porosities and the mass exchange coefficient. Both models were calibrated to produce the best fit to recorded tritium data. When various attributes of the dataset were considered, including cumulative tritium activity discharged to Fourmile Branch, plume arrival time, and plume attenuation due to closure of the seepage basins in 1988, the DDM produced results superior to the SDM, while causing no unrealistic upgradient dispersion. A sensitivity analysis showed that only the DDM was able to accurately produce both the instantaneous activity discharge and cumulative activity with a single parameter set. This is thought to be due to the advection-dominated nature of transport in natural porous media and the more realistic treatment of this type of transport in the DDM relative to the SDM.