COMBINING STRUCTURAL ANALYSIS AND BAYESIAN STATISTICS TO QUANTIFY THE IMPORTANCE OF MULTIPLE FRACTURE CAUSING AGENTS: APPLIED TECHNOLOGY ILLUSTRATED WITH A WYOMING STRUCTURE

Characterizing fractured reservoirs for flow modeling often requires interpolating between areas of limited geologic control. A common approach is to perform 3D structural analysis on a reservoir model to generate attributes relating to fracture formation. Structural attributes can be generalized into classes: morphologic – relating to the present-day shape of the reservoir, kinematic – relating qualitatively to incremental or cumulative deformation, and mechanistic – relating quantitatively to incremental or cumulative deformation. Typically, a combination of attributes best explains patterns and intensity of fracturing. However, attributes are usually calibrated to fracture occurrence one at a time and then used to infer the distribution between control points. To provide the best possible model describing fracturing, multiple attributes should be combined simultaneously in a mathematically unbiased and consistent manner.

Bayesian Updating is a statistical theory relating conditional probabilities through multivariate correlations. In considering the prediction of a primary event A (fracture intensity) with two secondary events B and C (structural attributes), Bayesian Updating provides the conditional probability of A given events B and C. This is done by combining B and C into a likelihood that updates A. In application we assessed the distribution of fractures in the Frontier Sandstone flanking Oil Mountain Anticline in central Wyoming. The intensity of folding-related fractures, which are easily separable from those that predate folding, were obtained by scan-line measurement of fracture spacing along a 5 km outcrop extending around the flanks of the anticline. We compared fracture intensity against 13 morphologic and kinematic structural attributes and found that flexural-folding strain, as a single attribute, best explains fracture intensity. Convolving that attribute with a properly filtered curvature attribute improves the correlation. We used the combined attribute to populate the entire anticline with tectonic fractures and consider the result to be a viable analog describing fracture development in analogous cases.