Comparison of two‐dimensional and three‐dimensional simulations of dense nonaqueous phase liquids (DNAPLs): Migration and entrapment in a nonuniform permeability field

Abstract

The influence of reduced dimensionality (two‐dimensional (2‐D) versus 3‐D) on predictions of dense nonaqueous phase liquid (DNAPL) infiltration and entrapment in statistically homogeneous, nonuniform permeability fields was investigated using the University of Texas Chemical Compositional Simulator (UTCHEM), a 3‐D numerical multiphase simulator. Hysteretic capillary pressure–saturation and relative permeability relationships implemented in UTCHEM were benchmarked against those of another lab‐tested simulator, the Michigan‐Vertical and Lateral Organic Redistribution (M‐VALOR). Simulation of a tetrachloroethene spill in 16 field‐scale aquifer realizations generated DNAPL saturation distributions with approximately equivalent distribution metrics in two and three dimensions, with 2‐D simulations generally resulting in slightly higher maximum saturations and increased vertical spreading. Variability in 2‐D and 3‐D distribution metrics across the set of realizations was shown to be correlated at a significance level of 95–99%. Neither spill volume nor release rate appeared to affect these conclusions. Variability in the permeability field did affect spreading metrics by increasing the horizontal spreading in 3‐D more than in 2‐D in more heterogeneous media simulations. The assumption of isotropic horizontal spatial statistics resulted, on average, in symmetric 3‐D saturation distribution metrics in the horizontal directions. The practical implication of this study is that for statistically homogeneous, nonuniform aquifers, 2‐D simulations of saturation distributions are good approximations to those obtained in 3‐D. However, additional work will be needed to explore the influence of dimensionality on simulated DNAPL dissolution.