U.S. Geological Survey Toxic Substances Hydrology Program--Proceedings of the Technical Meeting, Colorado Springs, Colorado, September 20-24, 1993, Water-Resources Investigations Report 94-4015

Multiphase-Flow Modeling of the Bemidji, Minnesota Crude-Oil-Spill Site with Geostatistical Simulation of Hydraulic Properties

Abstract

This paper describes a study aimed at improving the understanding of multiphase flow in a heterogeneous aquifer by using data collected at the north pool sub-region of a crude-oil-spill site near Bemidji, Minnesota. Data collected include grain-size distributions, oil and water saturations (fluid saturation refers to the volume of the fluid phase relative to the volume of pores in the sample), and porosities of samples collected at the north pool site. These data are used to estimate permeability (k), retention curves, and the geometric mean and standard deviation of the sampled grain-size distributions. The k estimates are statistically distributed in a bimodal lognormal fashion, whereby the two k population distributions correspond to the two predominant aquifer lithologies at the north pool: a coarse glacial-outwash deposit and interbedded fine-silt lenses. A dual geostatistical approach involving sequential indicator and Gaussian simulation is used to characterize the bimodal aquifer heterogeneity present at the north pool by creating one realization of a three-dimensional grid of log(k) values conditioned on the original data.

Field oil-saturation data indicate that the spatial distribution of oil in the subsurface has been affected by the bimodal aquifer heterogeneity present at the north pool. The shape of the subsurface oil-saturation distribution is highly irregular; more specifically, the center of the oil lens is actually depressed below the water table in some locations, and oil saturations almost as high as 0.3 occur in the unsaturated zone. A cross-sectional numerical model is used to simulate the flow of oil and water at the north pool, assuming the air phase is at atmospheric pressure. A slice of the three-dimensional grid of log(k) values produced through geostatistical simulation is used as input in the multiphase model to investigate the effect of aquifer heterogeneity on multiphase flow. Flow simulations using the bimodal representation of heterogeneity produce an oil-saturation distribution that is similar to that found in the field. Other factors affecting multiphase flow, such as hysteresis and three-phase oil relative permeability, also are investigated. The effects of hysteresis are not as significant when heterogeneity is adequately characterized.