Bob Parney1 and Paul LaPointe2
Search and Discovery Article #40098 (2003)
*Adapted for online presentation from the Geophysical Corner column in AAPG Explorer October and November, 2002, entitled “Fractures Can Come Into Focus”and “Simple Seismic, Complex Fractures,” respectively, and prepared by the authors. Appreciation is expressed to the authors, to R. Randy Ray, Chairman of the AAPG Geophysical Integration Committee, and to Larry Nation, AAPG Communications Director, for their support of this online version.
1Axis Geophysics, Denver, Colorado ([email protected])
2Golder Associates, Redmond, Washington ([email protected])
It has long been recognized that the presence of naturally occurring fracture networks can lead to unpredictable heterogeneity within many reservoirs. Conversely, fractures provide high permeability pathways that can be exploited to extract reserves stored in otherwise low permeability matrix rock. One of the primary difficulties in managing fracture heterogeneity and the consequent uncertainty is that production rates and volumes are controlled by fracture network connectivity between the producing wells, while the primary sources of data on fracture properties are measured only in the vicinity of wells. In some ways this is like trying to predict the size of a schoolyard by close examination of a single link in the surrounding fence.
Recent advances in the processing of 3-D seismic data, however, are providing valuable new tools for the imaging of fracture properties between wells. Those tools are the analysis of seismic velocities as affected by raypath direction and offset distance. Specifically, adjusting velocities as a function of azimuth (velocity anisotropy) to improve reflection imaging has produced by-product data volumes of seismic velocity anisotropy (ANMO) and improved data volumes of azimuthal changes in amplitude as a function of offset (AVAZ).
These seismic advances raise the following questions:
How do fractures influence these data?
Geologically, what should this newly imageable level of fracture heterogeneity look like?
How do we interpret this new data for fracture properties?
How do we then make the link between fracture properties and reservoir performance?
Figure 1. A maximum and minimum direction of seismic anisotropy are used to estimate fracture orientation and intensity. The geophysical model assumes a set of vertical fractures with constant strike orientation. In this figure the velocity is slowed by crossing the fractures so that the maximum velocity is parallel to the fracture strike.
Figure 3. Outline of structural geometry for Circle Ridge Field looking from north to south. Major fault surfaces are color-coded and labeled. Overthrust block appears on the top surface and is bounded by the Red Gully Fault. Compressionial shortening is higher in the north and decreases to the south. As a result the northern part of the field is fragmented into several imbricates.
Figure 4. Magnitude and orientation of fracturing in the overthrust block based on extensional strain calculated through a three dimensional Palinspastic reconstruction. Warmer colors indicate increased extensional strain and increased fracture intensity. Fault block is bounded on southwest edge by Red Gully Fault. Dot indicates location of well data.
Figure 5. Outcrop of Jurassic Nugget Sandstone, Circle Ridge Field, Wind River Reservation, Wyoming, looking to the north. Lines above photograph are fracture traces from outcrop, color-coded by set. Black fracture set is completely absent on east side (right) of outcrop.
Figure 7. Examples of fracture networks that would have similar seismic attributes over the volume delineated by the square, but very different network permeability values. The pattern in Figure 7a is unconnected; fracture permeability would be zero. On the other hand, the network shown in Figure 7b is well-connected, leading to a permeable fracture network.
Figure 8. Conversion of a DFN model of fracturing into a finite element mesh for use in simulating flow and transport through the fracture network. DFN models make it possible to calculate the network permeability at any scale, and thus provide the link between seismic attribute data and permeability values.
Figure 9. Snapshot of pressure in the fractures after injection. The colors indicate the pressure variations in the network (Blue colors indicate high pressure, orange indicates low pressure). Orange arrows show direction of flow out from the injector into the fracture network.
Figure 10. A DFN model with fractures curling around the structure of a plunging anticline. The cyan and blue colors indicate higher permeability, and the magenta cells with lower fracture permeability. Note the high permeability corridor set up along the crest of the anticline.
The underlying theory behind the ANMO and AVAZ processing is quite simple: Most geophysical processing algorithms assume that all fractures are approximately vertical, and are locally oriented in a single dominant direction (Figure 1). The maximum detectable seismic effect is when the seismic raypath travels perpendicular to the open fractures, crossing the slow velocity, possibly fluid-filled, open fracture. A maximum and minimum direction of fracture influence on P-wave and S-wave velocity can be determined and used to indicate the dominant fracture orientation.
The difference between the maximum and minimum effect gives some measure of the fracture intensity. This same process can be applied in a number of data volumes where the change in Vp or Vs as a function of azimuth is measured by the change in stacking velocities (azimuthal NMO) or the change in reflection coefficients (azimuthal AVO).
A critical feature of recently processed AVAZ and ANMO data volumes has been that the dominant fracture orientation can change dramatically over short distances. Recent work on a project sponsored by the U.S. Department of Energy (www.fracturedreservoirs.com) shows that these changes are not only possible, but also highly likely in a Rocky Mountain compressional setting where the stress field is complex.
The Circle Ridge Field, in Wyoming’s Wind River Reservation, Wind River Basin (Figure 2), was characterized through a combination of 2-D cross-sections and 3-D structural reconstructions based on well and surface data, and fracture data from surface outcrops and subsurface image logs. The fracture and structural data were supplemented with data from several transient well tests, a bromide tracer test and a nitrogen injection test.
The structure is primarily determined by NE-SW compression, which caused the formation of a series of imbricate fault blocks along the Red Gully Fault, including several imbricates to the north (Figure 3). The entire structure has been characterized as a fault-breached, fault-propagation fold. Development of the structure is likely to have produced the fracturing within the reservoir units. Fracture development was predicted using strain calculated through a 3-D palinspastic reconstruction of the field.
Figure 4 shows differences in extensional strain magnitude and orientation throughout a block of the Tensleep Formation in the hanging wall of the field’s Red Gully Fault. The contours and line lengths represent the magnitude of the maximum extensional strain due to the initial folding of the reservoir formations. The figure’s red lines represent the strike orientation of extensional fractures that would develop perpendicular to the local direction of maximum extensional strain. The red lines also show the dominant set; it is likely that a secondary joint set perpendicular to the set shown might also develop.
Ninety-degree changes in dominant fracture orientation across fracture fairways seen in Figure 4 are consistent with orientation patterns predicted by AVAZ data in nearby reservoirs. These orientation variations arise due to inhomogeneities in the stress field and the resulting fracture networks are consistent with well image log and tracer data.
Similar changes in fracture orientation occur in nearby outcrop at a much smaller scale (Figure 5). The black fractures occur only on the left portion of the outcrop, nowhere else. Red fractures dominate over blue fractures in the left portion, while blue fracture intensity increases markedly on the right hand side.
Since seismic anisotropy can be influenced by the presence of natural fractures – and that a high degree of variability in fracture orientation and intensity is to be expected in a Rocky Mountain compressional setting – interpretation of seismic data requires a sound link with knowledge of the fracture geology in a region.
The determination of fracture azimuth and intensity is usually based on the assumption that there is a single dominant fracture orientation, typically vertical. Frequently, fractures occur in several sets with cross-cutting orientations (Figure 6), and generally multiple sets are necessary in order to get well-connected plumbing for long-term productivity in the absence of high matrix permeability.
A number of attributes can be extracted from the seismic data. They can be grouped into two major categories:
Orientation attributes such as the fast P or S wave velocity azimuth were initially interpreted as the dominant fracture orientation. In the case of multiple fracture sets, the seismically sampled orientation is a function of the relative intensity of each fracture set. The net effect of multiple sets appears to be an average azimuth weighted toward the dominant set, although some data appear to show the seismic azimuth switching from one set orientation to another with no intermediate orientations apparent. For example, in an area characterized by a single dominant regional fracture trend orientation, any additional second fracture set may cause the attribute to appear to rotate away from regional trend, although there is no actual rotation of either of the fracture set orientations.
In the early development of anisotropic seismic analysis it was thought that high levels of anisotropy, as measured by the difference between the fast and slow P and S wave velocities, indicated a high level of fracturing. It is becoming clear that the influence of multiple fracture sets complicates the seismic intensity measurements. For example, where fracturing is intense, the seismic properties used to characterize orientation tend to become more isotropic. Small variations in any one set can produce apparent rotations of the interpreted fracture orientation. Isotropy in these seismic properties also exists when fracture intensity is very low.
Thus, the magnitude of the anisotropy does not in itself differentiate between regions of high fracture intensity and low fracture intensity. Other attributes such as interval velocity must be used to differentiate between an absence of fractures and an excess of fractures.
Once the attributes of the natural fracture system have been mapped, the next step is to take these attributes and use them as a predictive tool. This process, however, is not as simple as identifying fracture properties at a potential drilling location, as it is the connectivity between the well and the fracture network that is critical. Seismic attributes do not yet quantify any aspects of fracture network connectivity. For example, in Figure 7a the same five fractures occur in each of the two sample volumes, and would exhibit similar seismic attributes. However, only the network on the right (Figure 7b) would be conductive.
In order to assess the connectivity of a reservoir, the next step after obtaining the fracture attributes from the seismic data is to use DFN models to understand the consequences of fracture orientation and intensity on permeability. The DFN approach models fractures as two-dimensional polygonal planar objects, like playing cards, located in three-dimensional space (Figure 8a). Each fracture is characterized by its surface area and shape and has flow properties such as permeability, compressibility and aperture.
Network models can be formed based on an interpretation of seismic attribute data, engineering data, and image log data as available. Once fractures are generated, a finite element mesh can be constructed according to the fracture geometry (Figure 8b), and a flow solution can be obtained that takes into account the connectedness of the fracture system. Figure 9 shows an example of a pressure pulse spreading through a fractured reservoir in response to injection.
Discrete Fracture Network (DFN) models have provided an important tool to make the connection between seismic properties and reservoir. The DFN approach can be combined with seismic attribute mapping by first developing an interpretation of the link between attributes and fracturing. For example, the difference between the fast and slow P or S velocities can be used to control the fracture intensity of one fracture set within the DFN reservoir model, and the rotation of the P or S fast velocity azimuth can control the generation of the second fracture set.
Once the DFN model has been generated, a grid can be placed over the model and a finite element mesh used to calculate the potential volume of flow within each of the grid cells. In Figure 10, a DFN model is displayed with a grid populated by fractures, with the colors in each grid cell indicating the calculated permeability values. In this case, a high permeability pathway has evolved along the crest of the anticline due to the structural control of fracturing.
Recent advances in the processing of 3-D seismic data are providing valuable new tools for quantifying fracture properties between wells. In order to make use of this new information, it is necessary to:
Although uncertainties abound, these attributes provide new insight into notoriously difficult reservoirs, and promise to enhance recovery through focused engineering efforts.
Keefer, W.R., 1969, Geology of petroleum in Wind River Basin, Central Wyoming: AAPG Bulletin, v. 53, p. 1839-1865.