--> Abstract: An Overview of Seismic Fracture Prediction, by Enru Liu, Rishi Bansal, Xin Zhan, Alex Martinez, and Chris Harris; #120034 (2012)

Datapages, Inc.Print this page

An Overview of Seismic Fracture Prediction

Enru Liu, Rishi Bansal, Xin Zhan, Alex Martinez, and Chris Harris
ExxonMobil Upstream Research Company, Houston, Texas, USA

Natural fractures are the most abundant visible structural features in the Earth’s crust, and they are probably more common than we think. They are evident at most outcrops and in core samples. It is likely that most reservoirs contain some natural fractures, particularly in carbonates and unconventional reservoirs (tight gas, tight oil and shale gas formations). Fractures are mechanical discontinuities or partings caused by brittle failure due to an imposed stress field. They exist across a vast range of scales, from nearly ubiquitous microcracks to macroscopic killometer-long features, and they often form the so-called fracture clusters (swarm) or corridors (a fracture corridor is a zone where huge numbers of sub-parallel fractures are densely packed together to form a volume that is typically a few meters wide, a few tens of meters high, several hundred meters long and with a permeability well above 10 Darcy). Fractures can be open, permeable pathways, or they can be sealed as permeability baffles resulting from the presence of secondary mineralization or other fine-grained materials filling their aperture (e.g. quartz-formed bridges). Fractures can have a significant impact on reservoir management and field economics, from drilling, well completion, data collection, well placement, EOR strategy to estimation of ultimate recoverable reserves. Therefore, the earlier the influence of fractures is determined, the better.

Fracture information can be obtained from a range of measurements including outcrops, cores, image logs, cross-dipole log data, well testing, etc. We believe azimuthal seismic anisotropy measurements provide the best opportunity to identify sub-seismic fractures and their spatial variation (anisotropy measurements include shear-wave splitting and azimuthal P-wave attribute analysis and their various extensions). Although fractures are inheritably assumed to be smaller than a seismic wavelength, and individual fractures are not directly observed, we do get an average response. This average leads to a directional dependence, i.e. seismic velocities (hence amplitudes and other seismic attributes) are azimuthally anisotropic. Therefore the use of seismic anisotropy is arguably filling the gaps between the characterization of microscopic and macroscopic fractures.

Despite over two decades of industrial and academic effort, the reliability of seismic fracture detection technology is still constantly being questioned, and many geoscientists (including some geophysicists) still think that this technology is in its infancy or at least in its adolescence. To most engineers, our claim that we can extract fracture parameters from seismic data is still a myth. The main criticism results from the difficulty of cross-validation of seismically-derived fracture attributes with geological fractures (inferred from cores and borehole data) due to sampling, scaling and resolution issues, and is partly attributed to the misunderstanding of seismic fracture prediction concepts. Other issues include how to separate the artifacts caused by the acquisition footprint and near-surface or overburden anisotropy/structural variations from the anomalies caused by the presence of fractures. In this talk, we provide a review of this technology and summarize the key issues that need to be addressed using examples from carbonate and tight gas reservoirs.

The physical basis for the seismic prediction of fractures are the so-called equivalent medium theories, which are used to forward model the seismic response in fractured rock. Seismic velocity varies with fracture orientation (azimuthal anisotropy) and shear-waves birefringence occurs in fractured media. To model seismic wave response in fractured media, we have to make some assumptions about fractures (shape, aperture, length, surface roughness, spatial distribution, connectivity, etc). There are many models that have been proposed. Fractures may be simply as ellipsoidal cracks or as slip interfaces or displacement discontinuities, or as complex fractal fracture network. While mathematicians and physicists are still debating the details and accuracy of different fracture models, the fact that we have found that the seven most popular models are shown to match equally well the same laboratory data in terms of fracture-induced seismic velocity variations indicates that there is no model that can claim to be superior to other models. The differences between those models are mainly in the ways that fluid flow is treated at microscopic scales, resulting in differences in seismic attenuation. The same equivalent medium concept is also used to model hydraulic response in fractured media. Though the theories for modeling both elastic and hydraulic responses are similar (and even the equations look similar), there are fundamental differences between these two kinds of modeling. For example, fracture aperture is a key parameter controlling fluid flow in fractured rock, but it is insensitive to seismic response. For a fracture modeled as a planar surface with aperture distribution, the equivalent hydraulic aperture can be proven to be significantly smaller than the equivalent mechanical (elastic) aperture. Therefore, we should not expect to infer fracture aperture from seismic data.

Typically we assume that most reservoirs contain one or more sets of nearly vertical fractures (and often bounded by layers), and fracture lengths are smaller than a seismic wavelength. For a single set of vertically aligned fractures, the resultant medium will show transverse isotropy with a horizontal axis of symmetry (HTI). For a medium with two orthogonal sets of vertical fractures, we have an orthorhombic anisotropic system. Seismic waves are slowed and attenuated when they cross fractures, and the fractures make the rock more compressible in the direction normal to the fractures. The observation of subtle discontinuity and scattering associated with sub-seismic faults and fracture corridors is very common. All these effects appear to be more pronounced if the fractures are gas-filled or filled with a compressible fluid. The current consensus is that fracture orientations and density may be reliably estimated from seismic data, and to a lesser degree, fluid properties. The estimations of other parameters from seismic data, particularly these controlling the hydraulic properties of rock such as fracture geometry, size, aperture, connectivity, permeability, etc., remain challenging issues for geophysicists.

We can use conventional 3D seismic data to extract fracture parameters. The azimuthal variations of P-wave attributes, such as travel-time, NMO velocity, reflected wave amplitudes, frequency, attenuation, etc. can all be approximately described by sinusoidal variation or an ellipse in fractured media, where one axis of this ellipse indicates the fracture orientation, and the ratio of two axes of this ellipse is proportional to the fracture density or intensity of the rock concerned. The direction of minimum travel-time or maximum NMO velocity is interpreted as the fracture orientation. For amplitudes, it is a little bit more complicated depending on the type of AVO, however through the analysis of azimuthal AVO response, we can obtain fracture orientation and density from 3D seismic data. Because P-wave azimuthal variations increase with offset, typically we require the offset-depth ratio to be larger than one.

Shear-wave data from multicomponent seismic data contain wealth of information about fractures (most people claim even more information in shear-wave data than in P-wave data). The idea is that when a shear-wave enters an anisotropic (fractured) medium, one shear-wave will split into two: one travels faster, the S1 or fast S-wave, and another one travels slower, the S2 or slow S-wave. This is known as shear-wave splitting or birefringence. The polarization (or vibration direction) of the fast S-wave is in the fracture plane, while the time-difference or time-lag between S1 and S2 is approximately proportional to the fracture density. This technology forms the fundamental basis for fracture prediction from cross-dipole data and has now been extended to PS-converted waves, and a range of tools have been developed to extract fracture parameters from multi-component data using PS waves.

The challenge of imaging through complex geology (such as folds and faults in overthrust geology) and extraction of anisotropy parameters through the use of seismic anisotropy to accurately determine reservoir subtleties requires high fidelity seismic data and careful processing. Thus the question arises – how to separate the artifacts caused by the acquisition footprint and near-surface or overburden anisotropy/structural variations from the anomalies caused by the presence of fractures. Over the last a few years, significant progress has been made to address this issue including the optimization of data processing and attribute extraction workflows, resulting in significant improvement in the geological consistency and reliability of seismically-derived fracture attributes. Nevertheless, there is no doubt that this technology is better suited to relatively ‘simple’ geological structures.

Most people would agree that the proper characterization of fractured reservoirs requires the integration of all available data: outcrops, borehole (cores, FMI and log data), geological models, geophysical and production data. However, how to integrate different information from different measurements remain a challenging issue. We know that different data that sample different rock-mass measure fractures at different scales: Using downhole instruments, the fractures intersecting boreholes can be detected and characterized; Cores and image logs typically provide the small-scale features of reservoirs near boreholes; Cross-dipole data sample rock only a few feet away from wells, and data from well tests and production logging tests are also essential to characterize the fluid flow around the wells. Further measurements that examine the inter-well space are required to understand how the fractures observed in wells extend across reservoirs – the use of seismic data fills this gap, and it provides an average of fracture information between wells in the order of 10s to 100s of feet from wells. We have to acknowledge that each tool yields only a portion of the total fracture network, and all these measurements are complementary. Therefore understanding the physics, limitations and complementary nature of these diverse measurements is crucial before we can even think about how to upscale and cross-validate different measurements for proper data integration and to build a representative and geological-geophysical consistent earth model for fluid simulation.

 

AAPG Search and Discovery Article #120034©2012 AAPG Hedberg Conference Fundamental Controls on Flow in Carbonates, Saint-Cyr Sur Mer, Provence, France, July 8-13, 2012