DNAPL infiltration, redistribution, and immobilization in porous media

A constitutive model is developed that incorporates the critical capillary and relative permeability phenomena necessary for simulating the spatial and temporal migration of nonwetting fluid in saturated porous media. Bench-scale experiments involving the sequential infiltration, redistribution, and immobilization of dense, nonaqueous phase liquid (DNAPL) pools were conducted in a one-dimensional apparatus packed with water-saturated, unconsolidated porous media to provide model validation data. A light transmission/image analysis system, calibrated at the local scale, Captured the elevation of the connected-phase DNAPL/water interface as a function of time and provided the evolution of saturation profiles throughout the experiments. Hysteretic capillary pressure-saturation (P_C-S) and nonwetting phase relative permeability-saturation (k_rN-S) relationships were measured simultaneously at the local scale for the sands employed. Multiphase flow simulations of the bench-scale experiments, employing the developed constitutive model and using the independently measured porous media and fluid model parameters, predicted the observed DNAPL migration processes within measurement uncertainty in both space and time. This represents the first constitutive model validated against a one-dimensional experiment for DNAPL infiltration, redistribution, and immobilization processes below the watertable.; Predicting equilibrium DNAPL pool heights is demonstrated to depend on accounting for the non-zero capillary pressure across the fluid-fluid interface at the top of the pool. The terminal pressure is demonstrated to be the minimum sustainable capillary pressure in connected-phase nonwetting fluid experiencing imbibition, below which residual is formed. At the macroscopic scale, the terminal pressure corresponds to the extinction saturation (i.e., zero nonwetting phase flow) at the inflection point on the imbibition P_C-S curve. A ratio of terminal to displacement pressure of approximately 0.6 applies at both bench and macroscopic scales, and is independent of porous media and fluid properties. Constitutive models that do not incorporate both a displacement and a terminal pressure, such as those based upon the standard van Genuchten function, are demonstrated to be unable to predict the observed equilibrium DNAPL pool heights. Predicting the observed rates of nonwetting fluid redistribution and immobilization is demonstrated to require properly accounting for k_rN-S hysteresis, including imbibition k_rN-S curvature and the abrupt extinction of k_rN at the terminal pressure.; Ten simulations of a fixed-volume DNAPL release in a spatially correlated random permeability field demonstrate that none of the simpler or more typical constitutive models examined were able to reproduce the spatial or the temporal characteristics of DNAPL migration predicted by the validated model. Constitutive models that do not incorporate both a displacement and a terminal pressure, such as those based upon the standard van Genuchten function, severely over predict the spatial extent of nonwetting fluid advance during infiltration and redistribution. Failing to account for critical parameters of either P_C-S or k_rN-S hysteresis is demonstrated to result in over predicting the time required for DNAPL immobilization by a time range of months to thousands of years. The practical implication of this study is that, in addition to knowing fluid and porous media properties, all of the examined constitutive model hysteresis and trapping phenomena must be incorporated, and appropriate values for the corresponding parameters need to be known, in order to accurately simulate the redistribution and immobilization of a DNAPL release below the watertable in space an...