PNNL

Abstract

To test and evaluate the ability of commonly used constitutive realtions to predict multi-fluid flow, predictions for a numerical flow and transport simulator are compared to experimental data. Three quantitative experiments were conducted in one meter-long vertical columns. The columns were filled with either a uniform sand, a sand with a broad particle-size distribution, or with a layered system where a layer of a coarse-textured uniform sand was placed between two layers of a finer-textured uniform sand. After establishing a variably water-saturated condition, a slug of a light nonaqueous-phase liquid (LNAPL) was injected uniformly at a constant rate. Water and LNAPL saturations were measured as a function of time and elevation with a dual energy gamma-radiation system. The infiltration and redistribution of the LNAPL were simulated with nonhysteretic and hysteretic parametric relative permeability saturation-pressure (k-S-P ) models. The models were calibrated using two-phase air water retention data and an established scaling theory. The nonhysteretic Brooks Corey k-S-P model, which utilizes the Burdine relative permeability model, yielded predictions that closely matched the experimental data. Use of the nonhysteretic and hysteretic k-S-P models, based on the van Genuchten S-P relations and k-S relations derived from the Mualem relative permeability model, did not agree as well with the experimental data as those obtained with the Brooks-Corey k-S-P model. Explanations for the differences in performance of the three tested parametric k-S-P models are proposed.