THREE-DIMENSIONAL ANALYSIS OF HYDRAULIC AND TRACER RESPONSE TESTS WITHIN MODERATELY FRACTURED ROCK BASED ON A TRANSITION PROBABILITY GEOSTATISTICAL CHARACTERIZATION

Stochastic permeability fields produced by traditional geostatistical approaches are generally smooth and continuous. Thus, the discreteness of hydrostratigraphic units is not well represented and this deficiency is exacerbated in the context of fractured rocks where the bulk of the flow occurs within individual fractures or thin fracture zones that form an interconnected network. While indicator geostatistical procedures have the ability to represent discrete features, geological datasets rarely provide the necessary detail to implement a comprehensive variogram analysis. In this paper, we apply an alternative transition probability-based geostatistical approach combined with a three-dimensional spatial Markov chain analysis to synthesize the discrete permeability structure of moderately-fractured rock. This is done in a conditional stochastic framework which honors the data and can infuse either hard or subjective categorical information that is consistent with geological interpretations. The methodology is being tested using data collected from the Moderately Fractured Rock (MFR) experiment area, a 100,000m3 block of granitic rock located at the 240-m level of the Underground Research Laboratory (URL) in southeastern Manitoba, Canada. Characterization activities within the MFR block have lead to the creation of an extensive geologic, structural, geochemical and hydrogeological database. In particular, attributes pertaining to fracture location, frequency and orientation along an array of boreholes intersecting the MFR experiment area, taken together with results from numerous hydraulic response tests within packed-off intervals along the boreholes, are used to produce conditional stochastic realizations of permeability in three dimensions. Using the generated permeability realizations as input to a transient, three-dimensional flow and transport model, we also compare predicted pumping test responses and tracer concentrations to the results of head and concentration breakthrough data obtained from a series of experiments conducted within the MFR block.