Geostatistical simulation of discrete fracture network models

A new geostatistical procedure has been developed that enables probabilistic modelling of fracture geometry that honours detailed and complex information on the locations of known fractures. The method produces realistic fracture geometries, and does not simplify the fractures by treating them as planar features. The procedure uses sequential Gaussian simulation to produce equiprobable realisations of 3D fracture geometry that honour: i) the surface expression of known fractures (i.e. their curvilinear traces in map view and, where available, their measured dips at the surface), ii) the fracture length distribution (and, as a result, the fracture density), iii) the fracture orientation distribution, iv) truncation probabilities between the various directional sets of fractures, v) the spatial clustering of fractures, and vi) "soft" information derived from geologic analogues and regional tectonic considerations. The procedure has been developed as part of Ontario Power Generation's Deep Geologic Repository Technical Program for the purpose of creating models of subsurface fracture networks for use in site characterisation, and as input into flow and transport models. This paper presents the procedure, shows its use in a case study based on actual field data from a chalk quarry in northern Germany, and illustrates some of its capabilities with a verification study of the shape factor reproduction in the simulated fracture networks.