Increased demand for quantitative answers to ground‐water problems, particularly associated with the use of numerical models, has increased the need to accurately determine the distribution of hydraulic parameters. Researchers have attempted to find correlations between electrical resistivity and the permeability of fresh‐water aquifers since 1951. Several recent studies report either direct or inverse relations between apparent formation factor and aquifer permeability. The basis for these relations is a direct or inverse relation between porosity and permeability and, as matrix conduction effects are not taken into account, constant fluid conductivity is either implicitly or explicitly assumed. Laboratory experiments conducted on granular materials suggest that matrix conduction (surface conduction) effects are either as important as, or dominant over, porosity‐permeability relations. Our experiments on granular materials show only weak relations between true formation factor and permeability. Relations between apparent formation factor and permeability are good only for constant fluid conductivity. Most importantly, the strongest relationship found was that between permeability and matrix conductivity. These data suggest either that (1) relations between permeability and apparent formation factor must be applied in very restricted geologic environments and only where fluid conductivity remains relatively constant, or (2) more fundamental relations between matrix conductivity and aquifer permeability should be applied.

Hydrocarbon baildown tests involve the rapid removal of floating hydrocarbon from an observation or production well, followed by monitoring the rate of recovery of both the oil/air and oil/water interfaces. This test has been used erroneously for several years to calculate the “true thickness” of hydrocarbon in the adjacent formation. More recent analysis of hydrocarbon distribution by Farr et al. (1990), Lenhard and Parker (1990), Huntley et al. (1994), and others have shown that, under vertical equilibrium conditions, there is no thickness exaggeration of hydrocarbon in a monitoring well, though there is a significant volume exaggeration. This body of work can be used to demonstrate that the calculation of a “true hydrocarbon thickness” using a baildown test has no basis in theory. The same body of work, however, also demonstrates that hydrocarbon saturations are typically much less than one, and are often below 0.5. Because the relative permeability decreases as hydrocarbon saturation decreases, the effective conductivity and mobility of the hydrocarbon is much less than that of water, even ignoring the effects of increased viscosity and decreased density. It is important to evaluate this decreased mobility of hydrocarbon due to partial pore saturation, as it has substantial impacts on both risk and remediation. This paper presents two analytic approaches to the analysis of hydrocarbon baildown test results to determine hydrocarbon transmissivity. The first approach is based on a modification of the Bouwer and Rice (1976) analysis of slug withdrawal test data. The second approach is based on a modification of Jacob and Lohman's (1952) constant drawdown—variable discharge aquifer test approach. The first approach can be applied only when the effective water transmissivity across the screened interval to water is much greater than the effective hydrocarbon transmissivity. When this condition is met, the two approaches give effectively identical results.

Transmissivity is often estimated from specific capacity data because of the expense of conducting standard aquifer tests to obtain transmissivity and the relative availability of specific capacity data. Most often, analytic expressions relating specific capacity to transmissivity derived by Thomasson and others (1960), Theis (1963), or Brown (1963) are used in this analysis. This paper focuses on a test of these relations using a large (215 pairs) data set from a heterogeneous aquifer. The analytic solutions predicting transmissivity from specific capacity do not agree well with the measured transmissivities, apparently due to turbulent well loss within the production wells, which is not taken into account by any of the analytic solutions. Empirical relations are better than the theoretical relations. Log‐log functions have greater correlation coefficients than linear functions. The best relation found for the data set chosen for this study has a correlation coefficient of 0.63, but the prediction interval was about 1.2 log cycles, indicating that the range of probable transmissivities corresponding to a single specific capacity was more than one order of magnitude. Tests with smaller subsets of data suggest that correlations based on data sets of 10 points or less are of limited value.

No abstract

No abstract