Accurate Computation of Fracture Density Variations: A New Approach Tested on Fracture Corridors

Fracture density is an important parameter for characterizing fractured reservoirs. Stochastic object-based simulation algorithms that generate fracture networks commonly rely on a fracture density to populate the reservoir zones with individual fracture surfaces. Reservoirs, including fracture corridors, represent particular challenges in petroleum reservoir studies. Indeed, it is difficult to identify fracture corridor zones objectively and precisely along one-dimensional well data, which are characterized by high fracture densities compared to diffuse fractures. To estimate fracture density, a common practice is to graphically depict only fracture corridors on fracture cumulative intensity curves. In this paper, an approach is proposed to formalize this technique using hypothesis testing. This method precisely compartmentalizes the well data into several zones having specific fracture densities. The method consists of the following steps: (i) dividing the diagram into zones depending on a priori drastic changes in density, (ii) computing the local accurate fracture density for each zone and (iii) clustering the zones characterized by similar densities statistically. The key point is to couple regression and hypothesis testing. The regression aims at computing local average fracture density and the hypothesis testing aims at clustering zones for which the densities are statistically the most similar. The proposed approach is dedicated to one-dimensional fracture surveys, such as well data and outcrop scanlines. First, a synthetic case study is presented to prove the ability to highlight changes in fracture density. Second, the procedure is applied on a scanline dataset collected in a quarry (Calvisson, SE France) to show the usefulness of characterizing fracture corridors.