Geostatistical analysis of the Borden aquifer hydraulic conductivity field

A total of 11 vertical cores spaced 1 m apart along a single transect were extracted from the Borden aquifer at a location 60 m north-northeast from the site studied by Sudicky previously. Examination of these cores shows the upper 1.5-2.0 m of the aquifer to be comprised of relatively unstructured, fine to medium sands, underlain by a 0.1-0.8 m thick peaty layer and a 0.1 m thick organic-rich horizon. Between depths of 3.05 and 6.10 m the aquifer is comprised of clean, horizontally bedded fine to medium sands similar to those examined by Sudicky previously. Statistical analysis of 642 subsamples of 5 cm length from this interval suggests that the permeability data may follow a log-normal distribution, with a mean In-transformed permeability (m(2)) of -25.67 and a variance of 0.585. No consistent trend with depth was detected for ln-k, or for the mean and variance of ln-k on a core by core basis. Exponential, spherical, and Gaussian models all provided excellent fits to the vertical experimental variogram calculated using the classical estimator without the removal of outliers. These fits produced correlation scales ranging from 0.16 to 0.23 m. Of the three theoretical models, only the Gaussian model produced a reasonable fit to the horizontal experimental variogram with a correlation scale of 4.8 m. A linear model was also found to fit the experimental variogram data well. In comparison with earlier results of Sudicky, and Woodbury and Sudicky, it can be concluded that similar deposits were examined in this study, particularly with respect to correlation structure. On the basis of statistical tests, however, it cannot be concluded that the mean and variance of ln-k are stationary over the 60 m distance between the cores examined in this study and those of Sudicky. The results of this study indicate that the Borden hydraulic conductivity held is relatively uniform over the scale examined, but that it cannot be characterized by a single, spatially invariant statistical distribution.