Parameter Sensitivity in the Predication of DNAPL Infiltration and Redistribution in Heterogeneous Porous Media

Subsurface heterogeneity has a significant influence on infiltration and distribution of DNAPLs. Numerical models of multiphase flow are limited by their ability to predict behavior of DNAPLs in the field due to lack of accurate data on characterization of heterogeneity. However, these models help to obtain insights into the complex behavior that will be useful in designing effective remediation schemes. A number of multiphase flow codes that have the ability to capture the fundamental processes that control the migration of NAPLs exist. However, the issue of which constitutive model and parameters need to be used in simulating DNAPL behavior at various scales of interest is not studied satisfactorily. With the goal of contributing to the knowledge needed to address this issue, an experimental study was conducted to generate an accurate data set on the migration of DNAPLs in a synthetic aquifer with various combinations of heterogeneous/ homogenous configurations. This paper presents the results of a study where the data from these experiments was used in combination with an existing multiphase code UTCHEM9 to evaluate constitutive model sensitivity to prediction of both infiltration and redistribution. In the experiment, a tank with dimensions 0.71 m x 0.53 m x 0.04 m was packed with two layers. The layer in which the spill occurs was packed using five well-characterized sands to represent a spatially correlated random field with geostatistical parameters; Lnk of 22.5 (k in m2) and variance of 1. Second layer was homogenously packed with a single sand. The interface between the two layers had a slope of 3.5º. NAPL was injected as point source and saturation distribution was accurately monitored using an automated x-ray attenuation system. Modifications to the code were made to incorporate hysteresis and accommodate layer inclinations. The model simulations were conducted for two phases of the experiment; during NAPL injection and re-distribution for the following cases: (1) Brooks Corey drainage model during the injection and redistribution, without hysteresis and entrapment, (2) Brooks and Corey drainage model during injection and imbibition during redistribution, without hysteresis and entrapment, (3) Parker and Lenhard model with hysteresis, (4) Parker and Lenhard model without hysteresis, and (5) Brooks and Corey drainage model with trapping during injection and redistribution. The analysis shows during injection period, all of the above models were able to match the experimental observations reasonably well. The models without trapping and hysteresis effect were also able to match the experimental results. During the redistribution period, none of the five models were able to reproduce the experimental results. This preliminary analysis suggests that further detailed study is needed to determine which parameters control redistribution of NAPL. Also it is observed that entry pressure is the most important parameter of the constitutive relations; residual saturations of wetting and non-wetting fluids have minor effects during infiltration and redistribution.