Organic liquid saturation distributions resulting from a simulated tetrachloroethene (PCE) spill were generated with alternative models of spatially varying aquifer properties for a statistically homogeneous, nonuniform sand aquifer. The distributions were analyzed to quantify DNAPL source zone characteristics and then incorporated as initial conditions for simulated PCE recovery using surfactant-enhanced aquifer remediation (SEAR). The predicted evolution of the spatial distribution of DNAPL saturations or source zone architectures and associated remediation efficiencies are strongly influenced by the spatial correlation of aquifer parameters and multiphase flow constitutive relationships. Model predictions suggest that removal of 60 to 99% of entrapped PCE can reduce dissolved contaminant concentration and mass flux under natural gradient conditions by approximately two orders of magnitude. Aqueous phase contaminant flux, however, does not vary consistently as a function of the percentage of DNAPL removed, and notable differences in flux evolution were observed for models incorporating correlated versus uncorrelated capillary entry pressure and permeability fields. Simulation results demonstrate that the application of alternative models of aquifer parameter spatial variability can influence predicted DNAPL infiltration, entrapment, and recovery, even for relatively homogeneous aquifers of the type investigated here. Results also demonstrate potential benefits, in the form of reduced mass flux, accruing from partial mass removal that may not be readily predicted from analyses relying on simplified conceptual models for DNAPL source zone architecture or aquifer flow fields.