Predicting Natural Fracture Distribution in Reservoirs From 3D Seismic Estimates of Structural Curvature

Abstract

Abstract Natural fracturing is a manifestation of brittle strain imparted to the rock mechanical units during deformation. It seems logical that attempts to predict distribution of fractures in reservoirs should then use strain-based measures as a guide. One way of estimating relative strain is with structural curvature. The approach taken here is to examine curvature as a function of scale. The reason for the scaling approach is to examine folds of different wavelengths for fracture potential. With this technique, principal curvature magnitude and direction can be determined at any point on the seismic grid and at any scale. From the principal curvatures an estimate of preferred fracture orientation can be made. Comparison of preferred orientation estimates made by this technique with borehole measurements of preferred orientation using imaging tools show very good correlation. The correlation holds even when fracture orientations are oblique to the maximum in-situ stress. When relative strain measures derived from the principal curvatures are compared to well-productivity indices, a positive correlation is found. This suggests that the technique is able to both identify areas of increased fracture potential and determine preferred orientation prior to drilling. Introduction One goal in the characterization of naturally fractured reservoirs is to predict the distribution of fractures and their properties in the inter-well region. Models that relate fracturing to brittle strain distribution form a basis for understanding the scale-independent spatial distribution of fractures in the reservoir1. If some measure of strain can be extracted from seismic data, then predictions of fracture spatial distribution are possible with the caveat that the strain distribution, may not by itself, say anything about the permeability of the fracture system. However, when integrated with production data or seismic attributes, it may be possible to accurately predict inter-well properties. The usual approach of discrete fracture modeling to characterize the inter-well region of naturally fractured reservoirs is subject to much uncertainty, primarily in the spatial distribution of fractures. If some information about the strain distribution of the reservoir were known, it would make fracture spatial modeling easier, as shown for 1-d examples1. However, discrete fracture modeling is often not necessary to answer many of the important questions needed for exploration or development of naturally fractured reservoirs2. The premise explored here is that structural curvature can be used as a relative measure of strain and that this measure is capable of predicting fracture distribution in reservoirs. An early attempt at using curvature as an indicator of fracturing was based on the idea that fracture intensity should be proportional to the derivative of structural dip3. High productivity wells were found to be coincident with the areas of high curvature. More typically, curvature is used as a seismic attribute to aid in defining faults 4. This paper examines several different curvature measures that can serve as a proxy for strain. The possibilty of using curvature information to predict fracture preferred orientation at any point in the reservoir is also tested. Finally, the relationship between well productivity and curvature is explored. Structural Curvature and Scale A description of the cross-sectional topography of a structure depends on the scale at which it is viewed. If viewed from a distance it may be described as a broad symmetric fold. Closer to the main structure smaller scale folds on the flanks or crest become apparent. These may be the result of faulting or bedding plane slip processes. Thus, folds of all wavelengths exist on structures and the fracturing associated with individual beds or groups of beds may be related to folds of different wavelengths5. This suggests that strain depends on scale and similar behavior may be expected for curvature.