The influence of dimensionality on simulations of mass recovery from nonuniform dense non-aqueous phase liquid (DNAPL) source zones

The influence of model dimensionality on predictions of mass recovery from dense non-aqueous phase liquid (DNAPL) source zones in nonuniform permeability fields was investigated using a modified version of the modular three-dimensional transport simulator (MT3DMS). Thirty-two initial two- (2D) and three-dimensional (3D) tetrachloroethene-DNAPL source zone architectures, taken from a recent modeling study, were used as initial conditions for this analysis. Commonly employed source zone metrics were analyzed to determine differences between 2D and 3D predictions: (i) down-gradient flux-averaged contaminant concentration, (ii) reductions in contaminant mass flux through a down-gradient boundary, (iii) source zone ganglia-to-pool (GTP) ratio, and (iv) time required to achieve a remediation objective. 3D flux-averaged contaminant concentrations were approximately 3.5 times lower than concentrations simulated in 2D. This difference was attributed to dilution of the contaminant concentrations down gradient of the source zone. Contaminant flux reduction predictions for a given mass recovery were generally 5% higher in 3D simulations than in 2D simulations. The GTP ratio declined over time as mass was recovered in both 2D and 3D simulations. Although the source longevity (i.e., time required to achieve 99.99% mass recovery) differed between individual 2D and 3D realizations, the mean source longevity for the 2D and 3D simulation ensembles was within 2%. 2D simulations tended to over-predict the time required to achieve lower mass recovery levels (e.g. 50% mass recovery) due to a smaller contaminated area exposed to uncontaminated water. These findings suggest that ensemble averages of 2D numerical simulations of DNAPL migration, entrapment, dissolution, and mass recovery in statistically homogenous, nonuniform media may provide reasonable approximations to average behavior obtained using simulations conducted in fully three-dimensional domains.