Fractures present in a rock mass directly control fluid flow. However, only a small percentage of fractures can represent true flow paths, and characterizing the fracture geometry might not lead to a complete representation of the flow field. Other factors such as permeability and fracture orientation with respect to the actual stress field must be evaluated.

A 3D fracture characterization within a rock mass has been performed to provide a more realistic and representative geological fracture network as a basis for numerical fluid flow simulations. The studied rock mass is a 12x45m mine pillar in the CANMET Experimental Mine at Val d'Or, Quebec, located within the Bourlamaque granodioritic batholith. All four vertical faces of the pillar were mapped and fractures were located and measured along the walls by multidirectional sampling. Structural analysis of the fractures shows two main conjugate sets, which are compatible with the Cadillac-Larder Lake fault system and parallel to the regional Archean structural trend.

 Ground Penetrative Radar (GPR) profiles were performed along two 20 meters vertical faces to image penetration of the fractures within the rock mass. Linear anomalies, interpreted as fractures, were identified on the profiles and correlations were established between the measured fractures and fractures detected on GPR profiles. Some fractures, present on both sides of the pillar, were visible throughout the entire penetration depth of the profile so that face to face correlations within the rock mass were possible.

 Combined detailed mapping, structural characterization and GPR work helped reconstruct a geologically sound 3D representation of the fracture network by:
1) recognizing physical fractures that cut through the entire rock mass and that probably correspond to major flow paths,
2) discriminating between wet fractures that actively participate in flow and
non penetrative dry fractures,
3) demonstrating that fractures that actively participate in flow are 2 Ga old structures, favorably oriented within present day stress conditions,
4) documenting that younger, much smaller, non penetrative fractures related to erosional uplift or excavation do not seem to play an important role in fluid flow.

Active fractures can be incorporated in a discrete-fracture numerical flow model. Fluid flow can be studied in a probabilistic framework, which accounts for the uncertainty in the rock mass description. Furthermore, testing of the applicability of dual continuum or equivalent porous medium representation of the rock mass can be done with the discrete fracture model.