Predicting mass transport in discrete fracture networks with the aid of geometrical field data

The importance of knowing the details of the fracture network geometry when predicting mass transport in discrete fractured media is investigated in this study. It is assumed that some information of the fractured medium is known, such as the location and direction of fractures that intersect drill cores in the investigated region and the overall conductivity of the fracture network in this region. In comparison, individual fracture properties are assumed to be unknown, apart from the assumption that they follow specified stochastic distributions. A two-dimensional test problem is analyzed with the Monte Carlo technique. Reality is there represented by a hypothetical fracture network. On this real network, cores are drilled and the thereby intersected fractures recorded. On the basis of the data obtained, realizations of fracture networks that might represent reality are generated. The data from the real network will limit the variability of the possible realizations of fracture networks. A problem of mass transport is solved on the real network and on each of the generated ones, using a particle tracing technique. By comparing the real transport time with the estimated transport time as is obtained from the Monte Carlo runs, the advantage of increasing the measurements (for example, drilling more holes) is analyzed. It is found that introducing more cores will decrease the estimated uncertainty but that the fracture lengths, the fracture line density, and the spatial correlation of the aperture along the fracture influence the value of the geometrical information. The estimated uncertainty is less in networks with long fractures combined with a high line density and little spatial correlation of the apertures. It is also noted that if the fracture apertures are adjusted so that all conditional networks have the same hydraulic conductivity, the estimated uncertainty in transport times is reduced substantially.