A new methodology is presented to determine 3-D hydraulic conductivity tensor for a fractured rock mass through 3-D discrete fracture fluid flow modeling. First, a 3-D stochastic fracture network model was built and validated for a gneissic rock mass based on the fracture data mapped from scanline surveys at the site. This validated fracture network model was combined with the fracture data observed on a borehole to generate a stochastic-deterministic fracture network system in a cubic block around each packer test conducted in the same borehole. Each packer test was simulated numerically applying a developed discrete fracture fluid flow model to estimate the influenced region or effective range for the packer test. A block size of 60 feet with the packer interval located at the center of this block was estimated for the influenced region. Using this block size, the average flow rate per unit hydraulic gradient (defined as the transmissivity multiplied by mean width of flow paths) field for fractures was calibrated at different depth regions around the borehole by numerically simulating the packer tests conducted at different depth regions. The average flow rate per unit hydraulic gradient of the fractures in the immediate vicinity of the borehole was considered to be quite different to the average flow rate per unit hydraulic gradient of the fractures at a significant distance away from the borehole. A relation was developed to quantify the ratio between these two parameters. Representative Elementary Volume (REV) for the hydraulic behavior of the rock mass was then estimated to be a block size of 50 feet. Finally, the hydraulic conductivity tensor in 3-D was obtained. The principal directions of hydraulic conductivity were found to be consistent with the existed fracture system. Further, the geometric hydraulic conductivity calculated was found to be comparable to the hydraulic conductivity estimated through the radial flow assumption in continuum porous media.
Currently used methods for estimating threedimensional (3-D) hydraulic conductivity tensor using either aquifer pumping test or packer test data are based on the assumption that the groundwater is flowing through a geologic continuum (Bear, 1972; Marsily, 1985; Hsieh and Neuman, 1985). These methods can generally be applied when wells penetrate the porous geological media such as alluvial deposits, but may have limited applicability when the geological medium is dominated by a fracture system that has well defined fracture sets. The porous, continuum media assumption is based on an average flow within a Representative Elementary Volume [REV] (Bear, 1972; Marsily, 1985). Even though the REV is small for porous media, for fractured rock masses it can be very large or in some cases may not exist (Kulatilake and Panda, 2000). If the REV does not exist, or is larger than the distance between the pumping and observation wells or the packer test hole and observation wells, it is not appropriate to use the equivalent continuum approach to analyze the aquifer pumping or packer test data for fractured rock masses. Before applying an equivalent continuum approach for a rock mass, one should investigate the REV for the hydraulic behavior of the rock mass. The REV for the hydraulic behavior is defined as the size beyond which the rock mass hydraulic conductivity tensor remains the same. For sizes larger than or equal to REV size, the equivalent continuum approach can be applied for the rock mass with a second-rank, symmetric, positive-definite hydraulic conductivity tensor. Estimation of the REV for hydraulic behavior of a highly heterogeneous, anisotropic rock mass requires monitoring of groundwater level of a significant number of obs