A geostatistical method has been developed for creating 3D fracture network models that honor information typically available at the time of preliminary site characterization. The method produces results that are visually realistic and structurally complex, preserving the detail of original surface traces and honoring down-dip behavior of fracture surfaces. Such models are useful as input to many types of subsequent engineering analysis, from mechanical stability to groundwater flow and contaminant transport. Field data from a high-resolution mapping of fractures have been used to verify that this method does produce results that reflect geometric and statistical characteristics of actual fracture networks. Fractures mapped on the front face of a 220×35×25m volume were used as conditioning data to create 100 equally likely models of the 3D fracture network in the entire volume. Fractures mapped on ten other subparallel faces across the volume were used as verification data, with the geometric and statistical characteristics of the actual fractures being compared to corresponding measurements from the fracture network simulations. The simulated fracture networks honor simple geometrical characteristics, such as fracture length and orientation distributions, and also correctly portray more complex characteristics such as the size distribution of unfractured regions.


In its ongoing efforts to improve geological supportof the safety case for waste repositories, Ontario Power Generation's Deep Geologic Repository Technology Program (DGRTP) has developed a geostatistical method [1] for creating 3D fracture network models (FNMs) that honor information typically available at the time of preliminary site characterization:

· Detailed information on the locations of surface lineaments, which is typically

available from aerial photography, remote sensing or ground reconnaissance;

· Regional tectonic information on stress, which can often help to constrain the style of

fracturing: dominantly low-angle features in compressional regimes, dominantly subvertical in tensional regimes;

· Geomechanical and structural geology principles, which often assist the modeling of the down-dip behavior of fractures; and,

· Field data gathered from geologically analogous sites, which may provide supporting information on fracture density and orientation, as well as on the way that fractures truncate against one another.

Early applications of this method [1,2] have been encouraging for several reasons:

· The simulated fracture networks are visually realistic, maintaining much of the detail that had to be sacrificed in previous attempts at fracture network modeling;

· The workflow leads to a transparent and auditable model whose parameters and assumptions are easily identified and traced;

· The probabilistic approach produces a family of equally likely scenarios that can be used to support the analysis of uncertainty and risk;

· The approach fosters interdisciplinary collaboration between specialists in structural geology, in rock mechanics and in geostatistics.

With promising results from initial studies, the research focus has moved to verification: how well do the simulated fractures mimic features observed in actual fracture networks?

This content is only available via PDF.
You can access this article if you purchase or spend a download.