An inverse approach to the construction of fracture hydrology models conditioned by geophysical data. An example from the validation exercises at the Stripa Mine

One approach for the construction of fracture flow models is to collect statistical data about the geometry and hydraulic apertures of the fractures and use this data to construct statistically identical realizations of the fracture network for fluid flow analysis. We have found that this approach has two major problems. One is that an extremely small percentage of visible fractures may be hydrologically active. The other is that on any scale you are interested in characterizing usually a small number of large features dominate the behaviour ([1] Transport Processes in Porous Media. Kluwer Academic, The Netherlands, 1989). To overcome these problems we are proposing an approach in which the model is strongly conditioned by geology and geophysics. Tomography is used to identify the large features. The hydraulic behaviour of these features is then obtained using an inverse technique called "simulated annealing." The first application of this approach has been at the Stripa mine in Sweden as part of the Stripa Project. Within this effort, we built a model to predict the inflow to the Simulated Drift Experiment (SDE), i.e. inflow to six parallel, closely-spaced holes, the N- and W-holes. We predict a mean total flow of approx. 3.1 (l/min) into the six holes (two-holes) with a coefficient of variation near unity and a prediction error of about 4.6l/min. The actual measured inflow is close to 2l/min.