The Study of a Naturally Fractured Gas Reservoir Using Seismic Techniques 1

Transcription

1 The Study of a Naturally Fractured Gas Reservoir Using Seismic Techniques 1 C. R. Bates, 2 H. B. Lynn, 3 and M. Simon 3 ABSTRACT The upper Green River Formation at the Bluebell- Altamont field, Utah (Figure 1) is a tight gas sand reservoir where economic production can be sustained only in regions of high natural fracturing. In 1994, a demonstration seismic project was conducted at the field to show how exploration for, and the characterization of, naturally fractured gas reservoirs can be more effective through the integrated use of seismic techniques. Study of field exposures, well logs, and regional stress indicators prior to the seismic survey indicated a high degree of preferential orientation to the dominant fracture trend at the field. The seismic survey consisted of two crossing, ninecomponent surface seismic lines and a nine-component vertical seismic profile. The compression, and shearwave surface seismic both recorded anisotropies that were related to the presence and azimuth of the natural fracturing. The surface seismic results were supported by results from the nine-component vertical seismic profile. This program demonstrates the potential offered by the use of integrated seismic and geological techniques for the analysis of both land and marine naturally fractured reservoirs; furthermore, it demonstrates the possibilities of reviewing existing databases containing compression-wave surface seismic data for fracture information. NATURALLY FRACTURED GAS RESERVOIR PROGRAM This study formed part of the Department of Energy (DOE) program for the detection and analysis Copyright The American Association of Petroleum Geologists. All rights reserved. 1Manuscript received March 23, 1998; revised manuscript received January 18, 1999; final acceptance January 29, Sedimentary Systems Research Group, University of St. Andrews, Fife, Scotland; crb@st-and.ac.uk 3Lynn Inc., Houston, Texas. This work was conducted under Blackhawk Geosciences contract for the U.S. Department of Energy (DOE) Contract #DE-AC21-92MC Pennzoil E and P is thanked for support and commitment to the project. Dave Phillips, Stewart Squires, Mike Jones, and Wallace Beckham are all thanked for their help during the project and for many invaluable discussions. of naturally fractured gas reservoirs. The principal goal of the research program was to expand current levels of industry development and production efficiency of natural gas from the extensive tight gas resource base of the United States where lowmatrix permeability reservoirs are expected to represent more than one-half of expected domestic gas production by 2030 (Watts, 1996). The aim of this project was the field demonstration of surface and downhole seismic techniques for predicting areas of fracturing that can be linked to potentially enhanced gas production in reservoirs with lowmatrix porosity and permeability. For the productivity to be enhanced in lowporosity and low-permeability reservoirs, the reservoir must be extensively fractured (Szpakiewicz et al., 1986; Lorenz and Finley, 1991). The fracture networks, however, commonly are not uniformly distributed throughout a reservoir (Lorenz et al., 1996). Locating zones with high fracture density prior to a drilling program is desirable so that these zones can be specifically targeted. High fracture density zones of interconnected fractures with sufficient aperture to allow the enhanced flow of hydrocarbons have been linked to zones of anomalous, high azimuthal seismic anisotropy (Lynn, 1995); furthermore, these zones are characterized by a dominant open-fracture direction, the orientation of which is controlled by the deformation history of the rock and the current in situ regional stress regime. To take advantage of a natural fracture pattern, a horizontal well would be located perpendicular to the dominant fracture direction to provide maximum drainage potential of the reservoir. This project compares seismic azimuthal anisotropy data recorded over a naturally fractured gas reservoir to the geological information about fracturing within and around the reservoir. SEISMIC TECHNIQUES Most open, natural fractures in the subsurface are near vertical and are preferentially oriented. The influence of vertically aligned fractures on seismic waves has been documented in the literature for 1392 AAPG Bulletin, V. 83, No. 9 (September 1999), P

2 Bates et al U S A Uinta Mts D.O.E..E. Project Uinta Basin Roosevelt Figure 1 Regional site location map for the Department of Energy project and generalized stratigraphic column for the Eocene upper Green River Formation. Project Location Piceance Basin Q M - P OLIG Q QT T5 T4 T3 Browns Park FM Bishop Conglomerate Duchesne River FM Uinta FM Upper Green River Mahogony Bench ECOCENE T2 Green River FM Eastern Uinta Basin Members Evacuation Creek Mbr Parchute Creek Mbr (Oil shale) Garden Gulch Mbr Douglas Creek Mbr T1 Wasatch/Colton FM Gas Oil many years (Nur and Simmons, 1969; Nur, 1971; Crampin, 1985; Lynn, 1986; Crampin et al., 1989); furthermore, it has been demonstrated that the presence and properties of these fractures, especially when they are open and filled with fluid or gas, can be mapped using shear-wave seismic techniques, such as shear-wave birefringence measurements (Mueller, 1991). When a shear wave enters an anisotropic region, such as a fractured reservoir, it undergoes a phenomena known as shear-wave birefringence or shear-wave splitting (Figure 2) (Crampin, 1985). The shear wave splits into two vertically propagating shear waves with the fast shear wave polarized parallel to the fractures (S1) and the slow shear wave polarized perpendicular to the fractures (S2). To a first-order approximation, the S1 travels at the uncracked shear-wave rock velocity, whereas the velocity of S2 is a function of fracture density (Lynn and Thomsen, 1990). These two directions, parallel and perpendicular to the fractures, are referred to as the principal directions or axes of anisotropy. The magnitude of shear-wave birefringence is determined from traveltime differences between the fast (S1) and slow (S2) waves and the direction of polarization of S1 is determined by the azimuth of open fracturing and orientation of maximum and minimum in situ horizontal stress (Crampin, 1985). Whether the differences between in situ horizontal stress or the fractures related to this stress are the primary cause of shear-wave splitting currently is a subject under intensive research and is not addressed in this paper. Numerous seismic studies have recorded shearand compression-wave signals using multicomponent vertical seismic profiles (VSP) and multicomponent surface seismic techniques (Lynn and Thomsen, 1990; Queen and Rizer, 1990; Mueller, 1991). These techniques seek to record the full shear- and compression-wave signature by using multicomponent recording. Multicomponent recording typically involves using receivers with

3 1394 Study of a Gas Reservoir Figure 2 Schematic representation of shear-wave splitting through an anisotropic material and compression-wave conversion to shear waves from reflection at an isotropic-anisotropic boundary. Ray paths for compression waves are shown both perpendicular and parallel to the dominant open-fracture orientation. K HMAX = maximum horizontal stress direction (modified from Crampin, 1985; Lynn, 1996). K HMAX Compression-Wave Source P P Shear-Wave Source SV(S2) SV(S1) K HMAX Open Fractures Open Fractures S2: Polarized fractures S1: Polarized fractures recording elements oriented in three mutually perpendicular axes (3 component or 3C) and either one component vertically oriented compressionwave sources or with additional compression-wave sources oriented in orthogonal horizontal directions. When three component sources are used together with three component receivers, a ninecomponent (9C) survey can be acquired. Using 3C receivers allows the split shear waves to be deconvolved, thus providing the faster (S1) direction and the magnitude of shear-wave splitting or time delay between the split waves in addition to standard compression wave results and amplitude variation with offset (AVO) data. The magnitude of the time delay between S1 and S2 can be calculated by comparing the arrival times to the same reflectors on separate S1 and S2 migrated stacked surface-reflection data at common mid-point locations (Mueller, 1991). The magnitude of splitting can be calculated in VSP data in a similar manner by recording the time delay between S1 and S2 to the same horizon. The arrival time delay between S1 and S2 reflections to the top of a target zone of interest documents the magnitude of splitting in the formations overlying the target, whereas time delays through the target zone of interest give splitting values for the target. The polarization directions determined either at the surface or in a borehole will give the splitting direction of the last event through which the shear waves traversed. Lateral differences in S1-S2 time delay along a seismic line reveal the lateral changes in relative fracture density. Although shear-wave studies can be used to measure significant properties of fractured reservoirs, the acquisition and processing of the data are complex and thus costly. Historically, the cost of shear-wave surveys has prohibited their routine use because typical surveys can be three to eight times more expensive than compressionwave surveys (Lynn et al, 1996). The magnitude of compression-wave anisotropy in fractured rock is less than that for shear waves; however, it has been demonstrated that a small amount of gas over water in the open fractures will produce a compression-wave AVO anomaly when the compression-wave ray paths are oriented perpendicular to the fractures (Crampin, 1985; Thomsen, 1986). Figure 2 also shows part of the mode conversion energy of compression-wave reflection at an isotropic-anisotropic boundary for reflections parallel and perpendicular to an open-fracture zone.

4 Bates et al When the orientation of the compression wave is parallel to the fractures, the wave mode converts to the fast shear direction S1, but when the compressionwave orientation is perpendicular to the fractures, the wave converts to the slow shear direction S2. Modeling studies by Allen and Peddy (1993) clearly demonstrated the variation of amplitudes due to gas-filled fractures in the Austin Chalk with overlying shale. In this Austin Chalk example, the relative fracture density governs the compression-wave reflection coefficients on middle to far offsets on reflection profiles where ray paths are perpendicular to the fractures. In a study by Johnson (1995), the attenuation of compression waves propagating perpendicular to the fractures (thus crossing the fractures) was found to be greater than the attenuation for propagation directions parallel to the fractures; therefore, it is necessary to obtain data for ray paths both parallel and perpendicular to the fractures to determine the seismic anisotropy using compression waves. Unlike shear-wave anisotropy studies where a single wave path can be used to determine the magnitude of splitting, the data acquisition for compression waves requires travel paths through different rock sections, which can introduce other potential causes of amplitude variations such as heterogeneity in the rock or pore-fill type. Despite these limitations, a major advantage of using compression waves for fracture analysis is that the cost is much reduced compared to the shear-wave techniques. For this demonstration project, both shear-wave and compression-wave techniques were tested for surface seismic and vertical seismic profiling. The surface shear-wave seismic yielded splitting variations for every point along the surface lines no matter what the orientation of the lines with respect to the anisotropy (fractures); however, we wanted to orient the compression-wave seismic nearly parallel and perpendicular to the fractures to measure the largest AVO signatures using this technique. Because compression-wave anisotropy studies require ray paths from more than one azimuth, it is only at crossing points in lines that appropriate data are available. In typical two-dimensional (2-D) surface seismic surveys, this information is present at the tie points of lines. In three-dimensional (3-D) surveys, this information is available at every bin location provided that the survey is collected with a sufficient range of azimuths. GEOLOGICAL SETTING The Bluebell-Altamont field is located in the northern part of the Uinta basin, Utah (Figure 1), and has been classified as a fractured, but otherwise tight, gas sand play (Spencer, 1989; Fouch et al., 1992). The field lies within an asymmetric east-west trending basin south of the Uinta Mountains with a steep thrust-bounded north flank and a gently sloping (1 2 ) south flank. The Uinta basin was the focus of lacustrine, fluvial, and alluvial deposition from the Late Cretaceous to the late Eocene (Johnson, 1985). The Tertiary upper Green River Formation represents the last major lacustrine deposit within the basin before renewed uplift of the Laramide structures, such as the Uinta Mountains to the north. Lacustrine intervals are represented by lake deltas, beaches, offshore bars, nearshore swamps, and sediments deposited on oxygenated and anoxic lake bottoms. The lake was large and deep with a stratified water column located relatively close to a source area characterized by active volcanism. Potentially productive sandstone sequences represent deposition in delta channels and stream mouth bars that were associated with tidal flat/shoreline fine-grained deposits. The sandstones are classed as quartzarenites with detrital constituents principally of subangular to subrounded grains of quartz. The fine-grained deposits are dominantly clastic shale with kerogen (plant fragments) and limestone (micrite) with fragments of thin-shelled pelecypods. The youngest sediments present are the terrigenous rocks of the lower Oligocene Duchesne River Formation. Significant erosion of these sediments (approximately 4500 m) has created the highly developed badlands terrain at the surface today. Widespread fracturing has been observed in the Bluebell-Altamont field (Lucas and Drexler, 1976). Narr and Currie (1982) noted orthogonal joint sets at outcrops in the central and southern parts of the field, together with monodirectional fractures in cores from the field. Using these observations, they developed a stress history model of fracture formation by failure during extension after burial and subsequent uplift. This model supports the proposal by Osmond (1965) that the Tertiary rocks of the Uinta basin have undergone only a single cycle of burial, diagenesis, uplift, and denudation. Although various origins and mechanisms have been proposed for fracture development (Stone, 1969; Gries, 1983), their presence is undisputed. The gas intervals are trapped by a combination of structural and stratigraphic factors. Situated on the gently dipping (1 2 ) southern limb of a large regional basin, the Bluebell-Altamont field exists as a small east-west oriented anticline with approximately 15 m of closure. The gas is trapped in structurally updip pinch-outs of the prograding lake margins. Producing intervals consist of fractured lake-margin sandstones encased by tight shales and carbonate of the lacustrine deposits. Individual sandstone units range in thickness from about 1.5 to 6 m thick and can occur as composite units up

5 1396 Study of a Gas Reservoir Figure 3 Core from the upper Green River Formation, Bluebell-Altamont field. (A) Fine-grained poorly sorted sandstone with no significant porosity (1995 m depth); (B) medium-grained poorly sorted calcareous sandstone, minor porosity (blue); (C and D) fine-grained poorly sorted sandstone with a large micrite clast seen as a large dark area in (C) and in the top half of (D). The clast is crossed by an open fracture and a partially open fracture filled with calcite. Core from 2248 m depth. to 30 m thick. The sands have limited extent in the east-west or strike direction. The relatively clean sandstones result in fast sonic velocities ( m/s), with strong reflections that can be easily correlated between wells using the surface seismic data. Gas is currently being produced from the upper Green River Formation at depths from 1980 to 2590 m. At the field, the gas production is interpreted to be from fractures in sandstones with a matrix porosity that is typically less than 8% and with permeabilities of less than 1 md. The low-matrix porosity is a function of the cementation history of the rock. Most of the sands are strongly cemented with fine-grained dolomite as an alteration of micrite, calcite, and quartz (Figure 3A, B, depth 1995 m). Calcite cement generally is poikilotopic and occurs either in association with the calcite grains or without calcite grains. Quartz occurs as a cement overgrowth. Fractures observed in cores commonly are open but can also be filled with calcite cement (Figure 3C, D). Figure 3C and D is from core at a depth of 2248 m, where the calcite crystal growth is into the open fractures. The

6 Bates et al Table 1. Fracture Azimuth Database from Surface Geological Measurements, Seismic Measurements, and Regional Information Azimuth of Azimuth of Location or Information Type Dominant Fracture Direction Location (Depth of Information) Other Fracture Directions Depth of Information Field exposure mapping N20 40W 0 m (surface) N60-70E Surface of near vertical fractures Well log, FMS* N20 30W m East-west >3320 m Well log, Northeast, minimum m N30W-N10E >3320 m borehole breakout horizontal stress N45 30W Regional stress, 0 m (surface) Gilsonite veins Regional stress N30W, maximum 0 m (surface) (Zobak and Zobak, 1991) horizontal compression Regional stress; N10 20W, 20,000 ft focus earthquake focus 35 mi west maximum horizontal compression *FMS = Formation Microscanner. presence of the calcite crystals growing into the open fractures suggests that the fractures are natural and not drilling induced. The open fractures have an aperture of at least 10 µm as determined by the size of calcite crystals, and thus the fractures themselves are estimated to be at least this width in the subsurface. Some of the fractures have greater widths in the cores, but it is difficult to estimate if this would have been the true width of the fracture in situ because there could have been subsequent stress relief. Gas production rates range from uneconomic rates of 100 MCFGD from unfractured reservoirs to economic rates of more than 5000 MCFGD in fractured reservoirs. Production has been enhanced in several wells with hydraulic fracturing of the reservoir. FRACTURE STUDIES To obtain optimal seismic data for fracture characterization, a database of geological fracture information was collected prior to the seismic survey design. These data are summarized in Table 1 and consist of fracture information from core, well logs, outcrop fracture mapping, and a literature review of the current in situ stress for the region. The orientations of near vertical fractures were mapped on surface exposures of the Duchesne River Formation at the field site, but no attempt was made to record fracture length because the deeply eroded badlands topography inhibited this type of study; however, a fracture was noted in the field only if its length was greater than 1 m. Fracture information from field mapping showed two dominant trends approximately N30W and N65E (Table 1, Figure 4). Fracture information from well logs was in the form of Formation Microscanner (FMS) images and well breakout information from caliper logs. Analysis of the FMS images showed a fracture trend in the northwest-southeast direction for the upper Green River Formation with an east-west trend in the lower Green River Formation. Well breakout data have been used in many studies to infer the directions of current in situ regional stress (Zobak et al., 1985), where the elongation of the borehole is measured by four- or six-arm caliper logs. The long axis of the ellipse is interpreted to show an alignment to the minimum horizontal stress direction with the potential for open natural fractures perpendicular to this. The FMS images for this study showed that from depths of m, there was an elongation in the northeast direction, indicating that fractures with a northwest azimuth would be preferentially open; furthermore, the orientation of regional maximum horizontal stress was determined using deep earthquake focus and other stress indicators by Zobak and Zobak (1991) to be in a direction of N30W. The presence of gilsonite veins or dykes in the basin gives other regional information concerning the trend of natural fracturing in the area. Gilsonite or Uintahite is a black to green shiny asphalite found almost uniquely in veins in Utah. The veins are from 30 cm to 6 m wide and can be many kilometers long. Fouch et al. (1992) postulated that the gilsonite is injected as veins subsequent to hydrofracture of the rock perpendicular to the minimum horizontal stress. This natural hydrofracturing is thought to have occurred due to formation water expulsion from the lower Green River Formation. Within the project area, the veins are

7 1398 Study of a Gas Reservoir Figure 4 Surface seismic locations and rose diagram showing outcrop fracture orientations, maximum horizontal stress directions from borehole elongation and fracture orientations measured from core, and the orientation of gilsonite dykes in outcrop. The exact location for the seismic lines cannot be shown due to proprietary concerns. Line 2 B maximum horizontal stress from borehole elongation gilsonite dykes fracture from cores J N C K D E fractures from outcrops F L G M N O H XX - 9CVSP - 9C VSP wells used in synthetic compressionwave study multicomponent vertical seismic profile m N W E P Line 1 oriented dominantly northwest-southeast (Narr and Currie, 1982). In summary, the geological and stress information indicates that there are two major trends of fracturing for the upper Green River in the field, namely northwest-southeast and northeast-southwest. Of these two directions, we postulate from the stress data that the northwest trend is more likely to be the azimuth of open fractures. SEISMIC PROGRAM The optimal location for the surface seismic lines and VSP in the field was critical to the success of the project. The orientation of the surface seismic was based on the best estimate of the open natural fracture trend in a northwest direction with seismic lines laid out parallel (line 1) and perpendicular (line 2) to this (Figure 4). The final orientation of the two crossing lines, northwest-southeast and northeast-southwest, also facilitated the acquisition effort because a major drainage crossed the area with a northwest-southeast azimuth. A more detailed location map has not been provided due to proprietary concerns of the industrial partner. The final acquisition parameters for the surface seismic were determined during a wave test program using impact sources to test receiver polarities and both compression-wave and shear-wave Vibroseis for variations in sweep length and frequency. Compression-wave acquisition for AVO requires long offsets to record the variations in amplitude due to fracture content. For this project the target reservoir was at a depth of m, and therefore offsets of at least this range were necessary. With shear wave data, near-vertical travel paths are considered more useful for recording split shear waves. Thus, for ideal survey conditions, one would like to be able to record full fold data for both compression waves and shear waves. Unfortunately, this would have required more recording channels than were available at the time in the field. A compromise was made and a reduction in fold of the shear-wave data was necessary to accommodate the m offset range for the compression-wave data. In areas where low shear-wave signalto-noise ratios are experienced, this compromise could have had a major impact on the study; however, the wave tests indicated that the signal-to-noise ratio was high for the shear-wave data, and thus it was better to record more data at far offset for the compression-wave AVO than to concentrate on the near-offset shear waves. The 9C VSP was acquired prior to the 9C surface seismic survey at the eastern end of line 2. The field results from this also were used in the planning of the surface seismic acquisition parameters. Table 2 summarizes the 9C VSP and surface seismic acquisition parameters. 9C VSP The objectives of the 9C VSP were to determine the time-depth-velocity relationship for split shear

8 Bates et al Table 2. Acquisition Parameters for 9C VSP* and Surface Seismic Survey Receiver Recording Source Type Source Spacing Parameters and Spacing Parameters 9C VSP* 3C** phones in Mertz M18, five stages at near and far 15 m separation ( m) offset Surface 10 Hz 3C phones, 2 msec sample, Mertz, Hz Shear Wave 12 per station, 6 sec record 91 m interval nonlinear sweep 46 m interval Surface 10 Hz 3C phones, 2 msec sample, Mertz M Hz Compression Wave 12 per station, 46 m interval 6 sec record vibrators, 91 m interval linear sweep *9C VSP = Nine-component vertical seismic profile. **3C = Three component. waves and to produce a corridor stack for compression wave, S1, and S2 for comparison with the surface seismic. The 9C VSP was acquired using both compression- and shear-wave vibrators with 3C receivers magnetically clamped to the well at 16 m intervals over the target reservoir from depths of m. The compression data followed standard VSP processing to produce a corridor stack for comparison with the surface seismic data. The shear-wave data were first deconvolved using the Alford rotation (Alford, 1986) to allow S1-S2 separation. The separated S1 and S2 data were then individually processed and stacked in a manner similar to that of the compression data to give corridor stacks. Comparison of the S1 and S2 stacked data gave values of time difference between the fast and slow waves that indicated the degree of shear-wave splitting. The time delays are shown in Figure 5. The azimuth of the faster S1 orientation is shown in hodograms on Figure 5 for three depth intervals in the well over a 60 ms time interval around the major downgoing shear-wave event. All depths displayed similar results to the three levels shown; however, these are not displayed due to space restrictions on the figure. A hodogram is a particle motion diagram that shows the direction of first motion of the shear wave across the 3C receivers. The hodogram plots the arriving energy of one receiver against another over a predetermined time interval around a major event, such as a reflection or first arrival. Only hodograms are shown here for the horizontal geophones rather than full 3-D hodograms for clarity of interpretation. These plots show horizontal linear first motion, indicated by the arrow on the figure, in a northwest-southeast direction (N43W ±10 ), thus indicating that the first arriving wave is polarized in this direction. That is, the polarization direction of the fast shear wave is in a northwest-southeast direction within a few wavelengths of the receivers at this location. Depth Below KB (m) Average time delay Time Delay S1-S2 (msec) N y x 1-2% Top of Upper Green River 5-12% TN1 Mahogony Bench Analysis of the hodograms from all the 3C receiver levels showed a consistent orientation of split shear-wave data, indicating that the azimuth of fast N y Actual delay for individual levels Figure 5 Multicomponent vertical seismic profile (VSP). Time delays (points) between fast shear wave (S1) and slow shear wave (S2) shows shear-wave splitting of an average maximum value of 12% through the top of upper Green River Formation. Hodograms (particle motion diagrams) for three depth zones are shown in inserts with first motion to the northwest highlighted by the arrows. Depth below KB = depth below kelly bushing. N y x x 8% Anisotropy (%)

9 1400 Study of a Gas Reservoir S1 direction did not change with depth from 868 to 2637 m. From the uniform azimuth of fast S1 direction, we postulate that the dominant open-fracture azimuth is also uniform over this depth range around the well. The difference between arrival times of S1 and S2 indicates a variation in amount or degree of splitting with depth. The top 868 m exhibit a 3% shearwave splitting. Between 868 and 2011 m, background values of splitting of 1% were recorded. At 2011 m (the top of upper Green River), an interval of anomalous birefringence is marked. This increase, to a maximum of 11% splitting, is consistent with an interpretation of a high open-fracture density region aligned at an azimuth of N43W through the upper Green River Formation. CROSS SECTIONAL SYNTHETIC SEISMIC MODELS Before the influence of fractures is considered as causing variations in the compression-wave surface seismic data, variations in the reflection data due to stratigraphy must be accounted for. For this purpose, we created compression-wave zero-offset synthetic seismogram models from the sonic logs of 17 wells projected along the seismic lines (Figure 4). Each synthetic model top was 107 m above the top of the upper Green River to ensure that a full seismic wavelength above the section of interest was used. The models were generated first by creating sonic log cross sections (for line 1 see Figure 6; for line 2 see Figure 7). These cross sections were tied for unit tops and a smooth interval velocity model was interpolated between each log. The compression-wave reflection coefficients were calculated from the sonic logs with a constant density assumption (a recognized limitation in the modeling), and the models were converted to twoway traveltime and filtered to match the field data (Figure 6C). The compression-wave zero-offset synthetic models were then compared to the compressionwave migrated near-offset stack field data. These field data, which contain only near-offset stacked data, should be the most similar to that from the model, which is only zero offsets and has no AVO effects from the far offsets. From a comparison of the zero-offset synthetic models and near-offset stack field data (Figure 6C, D), we concluded that lateral stratigraphic changes are faithfully represented by the reflections in the near-offset ( m) field stack data. That is, near-offset stack amplitude changes on both lines showed first-order influence of stratigraphy and lithology on the amplitude responses that were predictable from the sonic logs and geological history of basin development; furthermore, because variations in lithology and stratigraphy can account for the amplitude changes, no influence of fracturing is manifest on the near-offset stack data. This conclusion can be made for both line 1, parallel to the open-fracture trend, and line 2, perpendicular to the predicted open-fracture trend. SURFACE SEISMIC: SHEAR WAVE In a similar manner to the VSP data, the surface seismic reflection data were analyzed using the Alford rotation to separate the split shear waves before separate processing of the fast and slow shear-wave data to yield S1 and S2 stacked sections. From the analysis of the S1 and S2 directions using the Alford rotation, an average azimuth of the fast shear wave at N30W was measured along both of the seismic lines. By comparing the traveltimes to reflectors in the S1 and S2 stacked sections, the magnitude of shear-wave splitting in the subsurface was calculated. An example of the shear-wave stacked sections is shown in Figure 8 where both S1 and S2 sections are spliced together at well G, a gas-producing well. With the S1 and S2 sections time-aligned at the top of the upper Green River marker, an incremental shift with time (and depth) is seen between the two time sections such that the S1 section has the least traveltime (fastest velocity) within each interval below. From this time shift the magnitude of traveltime splitting can be calculated at 12% through the gas-producing interval within the top of the upper Green River. Along both surface seismic lines a maximum shear-wave traveltime splitting of 18% was recorded. Typically, shear-wave splitting of below 5% is regarded as a background value (Crampin, 1994). The amplitude variation between the S1 and S2 stacked sections most likely is a function of lithology or heterogeneity, but its analysis was beyond the scope of this project. SURFACE SEISMIC: COMPRESSION WAVE The near-offset stack data have been discussed and are interpreted as a reliable representation of the stratigraphic variations across the field. To record a manifestation of the fractures in the compression-wave data, it is necessary to analyze the far-offset data for amplitude variations with offset (AVO). This is done at a location where there is information from more than one azimuth, parallel and perpendicular to the dominant fracture trend. It is assumed that changes in amplitude due to stratigraphy and lithology would give an equal response to ray paths from both azimuths, and thus their affect can be accounted for. In this survey,

10 Figure 6 Compression-wave zero-offset synthetic seismogram model for line 1 (northwest-southeast line). (A) sonic log cross section; (B) interpolated sonic log cross section; (C) zero-offset synthetic seismic section; (D) migrated near-offset seismic stack with synthetic sections spliced in below each well. T/GR = top of upper Green River Formation, Z and TN1 are marker horizons, MB = Mahogony Bench. Bates et al. 1401

11 Figure 7 Compression-wave zero-offset synthetic seismogram model for line 2, (northeast-southwest line). (A) Sonic log cross section; (B) interpolated sonic log cross section; (C) zero-offset synthetic seismic section; (D) migrated near-offset seismic stack with synthetic sections spliced in below each well. T/GR = top of upper Green River Formation, Z and TN1 are marker horizons, MB = Mahogony Bench Study of a Gas Reservoir

12 Bates et al Figure 8 Fast shear-wave (S1) stacked reflection section and slow shear-wave (S2) stacked reflection section spliced together at well G along line 1. Both sections thus are from northwest-southeast ray paths. information from more than one azimuth is available only at the tie point of the two lines. This limitation is quite common for reconnaissance data, but also demonstrates the utility of the technique if a grid of 2-D compression-wave surface reflection data is available. The AVO gradient (change in AVO) or linear variation in reflectivity is proportional to the change in Poisson s ratio across a reflecting interface (Shuey, 1985). Because the effect of gas in the pore space (or fracture space) (Nur, 1971) of a rock is to decrease the Poisson s ratio for the rock, AVO is an appropriate method for gas detection. A comparison of AVO at the line tie is shown in Figure 9A for near-offset summed supergather and in Figure 9B for far-offset summed supergather. A supergather is a sum of nine common depth-point locations on each line centered on the tie point. At near offsets ( m), a clear tie in time can be made between the two lines, and for the majority of reflectors a good comparison of amplitudes also can be made; however, at far offsets ( m) marked differences in amplitude are seen on the Z and Mahogony Bench marker horizons. For line 1, the amplitudes of these events are weak, as would Figure 9 Compression-wave stacked sections along lines 1 and 2 spliced at the well D, the line tie. (A) Comparison of compression-wave near-offset (<1826 m) stack for line 1 and line 2; (B) comparison of compressionwave far-offset (>1826 m) stack for line 1 and line 2. Note the amplitude differences between the far-offset comparison and the near-offset comparison. be expected for amplitudes dimming with offset in either unfractured rock or rock where there are no fractures filled with gas being crossed by the compression-wave ray paths; however, on line 2 there are large increases in amplitude, suggesting that the orientation of these ray paths crosses gas-filled fractures. The increase is counterintuitive because of the reverse display polarity in the data, but is consistent with models for open fractures provided by Crampin (1985), Thomsen (1988), and Allen and Peddy (1993). Note that the

13 1404 Study of a Gas Reservoir (cool, blue colors) and is interpreted as a potential depleted zone on line 2 at the gas blow-out well location 1.2 km west of the line tie. Because line 1 is oriented approximately parallel to the measured open-fracture direction, little influence on the AVO signatures of the gas-filled fractures is seen on this line The preferred flow direction or maximum horizontal permeability direction of a reservoir usually is parallel to the dominant open-fracture direction. Heffer and Dowokpor (1990) attempted to relate seismic anisotropy not only to the local fracturing, but also to the amount of permeability anisotropy that exists in naturally fractured reservoirs. Unfortunately at the Bluebell-Altamont field, no two wells were close enough to record interference effects between the wells, and no other information was available on flow anisotropy. COMPARISON OF SHEAR-WAVE AND COMPRESSION-WAVE RESULTS Figure 10 Comparison of AVOA (amplitude variation with offset and azimuth) at line tie (well D). Positive values (hot colors) show AVO anomalies on line 2 across the open gas-filled fractures. No anomalies are seen at these depths on line 1 parallel to the fractures. amount of gas in the fractures cannot be determined using this method because only a small amount of gas will cause an anomaly (Domenico, 1976; Ostrander, 1984). Castagna and Smith (1994) suggested an alternative display of AVO anomaly using the AVO gradient and AVO zero intercept. This display method has the advantage of suppressing variations in lithology and porosity while remaining consistent to pore fluid content variations and is shown for the tie point in Figure 10. Areas of anomalous amplitudes at the tie point on line 2 (interpreted as crossing the dominant open gas-filled fractures) are shown with hot colors (dominantly reds) at the Z marker horizon and at the Mahogony Bench. Thus, the compression-wave data can be said to show an amplitude variation with offset and azimuth (AVOA) at the tie point. The tie point is approximately 1.2 km west of the gas discovery well at this field where a closed-in well blew with dramatic effect in The horizon that was thought to have caused the gas buildup was the Z interval. In Figure 10, the Z interval shows diminished amplitudes A comparison of the surface shear-wave splitting data and surface compression-wave AVOA results can be made only at the line tie. The results for shear-wave traveltime splitting and compressionwave AVOA are plotted in Figure 11. A relationship exists between the percent shear-wave traveltime splitting and the AVOA anomalies. The highest AVO anomaly is observed at the Z-reflector (top of Z to TN1 interval), where an interval of high (12%) shear-wave traveltime splitting also begins. The implication from the compression-wave data is that if data for both azimuths (parallel and perpendicular to the dominant open and gas-filled fractures) existed at all points along the surface lines, similar results could be anticipated. This approach suggests that the natural extension for this work is in full 3-D seismic surveys where every bin has an equal distribution of azimuths or has data for at least two azimuths that are parallel and perpendicular to the dominant fracture orientations. This recommendation was made to the DOE and subsequent continuation projects seek to investigate the potential of 3-D seismic technologies. We postulate that in 3-D surveys over fractured gas reservoirs, the data quality for compression-wave images of structure and stratigraphy would be higher when the ray paths are parallel to the open fractures and, furthermore, that if a gas chimney effect were present in the data, it would not be seen for these fracture parallel azimuths. A further implication of the compression-wave results is that if existing 2-D data are available at line tie points and if the orientation of the seismic lines are parallel and perpendicular to a major open-fracture trend, then there is the possibility of

14 Bates et al Depth (m) AVO gradient difference Figure 11 Comparison of compression-wave AVOA (amplitude variation with offset and azimuth) values at reflection interfaces and shear-wave traveltime splitting through formations at the tie point (well D). reevaluating old data for amplitude anomalies. For such an evaluation, we recommend that the data be reprocessed to normalize parameters and properties other than the fracture elements at the line tie with careful attention paid to data quality. CONCLUSIONS S-wave Splitting (%) 1.0 1,000 P-wave AVOA (AVO gradient +intercept)/2 Z Marker The upper Green River Formation at the Bluebell- Altamont field, Utah, is a naturally fractured, but otherwise tight, gas sand reservoir. The natural fractures show a preferential azimuth in field exposure (N30W and N65E) and well logs (N25W), with an indication from regional stress data that the open fractures are aligned in the northwestsoutheast direction. The azimuth of open fracturing was coincident with that measured using surface and well shear-wave seismic. The surface shear-wave seismic mapped zones of high shearwave splitting from which zones of increased natural fracturing are inferred. The surface seismic results were confirmed by results obtained from a nine-component (9C) vertical seismic profile , ,000 Top of Upper Green River TN1 Mahogony Bench Compression-wave surface seismic results at the tie point of two lines indicated amplitude variation with offset and azimuth anomalies that were consistent with an interpretation of an interval that contained open fractures parallel to the northwestsoutheast line. The fracture content was most likely to be gas; furthermore, the zones of high shearwave traveltime splitting showed a relationship to the compression-wave AVO and azimuth anomalies recorded at the top of the fractured intervals. Open, aligned, vertical fractures, which are at least partially gas saturated, are interpreted to be the cause of both the compression-wave and shear-wave variations. These seismic techniques are highly recommended for reconnaissance work over naturally fractured reservoirs where new data are to be acquired. The techniques also are useful for reevaluating older two-dimensional compression-wave data where crossing lines or preferably a grid of lines is available; however, the full potential of using these seismic techniques to map and evaluate reservoirs with naturally fractured zones where there is likely enhanced production will be realized only in a three-dimensional (3-D) seismic survey where many subsurface locations have ray paths in many orientations. Such 3-D surveys will require full-offset and full-azimuth data. The utility of this technique will extend beyond that for land reservoirs to offshore marine reservoirs, in particular marine reservoirs that may have had imaging problems in the past because of naturally fractured gas chimneys when the gas cloud is primarily contained within vertically aligned fractures. REFERENCES CITED Alford, R. M., 1986, Shear data in the presence of azimuthal anistropy; Dilley, Texas: Abstracts of papers presented at the 56th annual international SEG meeting, Houston, Texas, p Allen, J. L., and C. P. Peddy, 1993, AVO frontiers, in Amplitude variation with offset: Gulf Coast Case Studies, Geophysical Development Series, v. 4., p Castagna, J. P., and S. W. Smith, Comparison of AVO indicators: a modeling study: Geophysics, v. 59, p Crampin, S., 1994, The fracture criticality of crustal rocks: Geophysical Journal International, v. 118, p Crampin, S., 1985, Evaluation of anisotropy by shear-wave splitting: Geophysics, v. 50, p Crampin, S., H. B. Lynn, and D. C. Booth, 1989, Shear-wave VSP s: a powerful new tool for fracture and reservoir description: Journal of Petroleum Technology, v. 41, p Domenico, S. N., 1976, Effect of brine-gas mixture on velocity in an unconsolidated sand reservoir: Geophysics, v. 41, p Fouch, T., V. Nuccio, J. Osmond, L. MacMillan, W. Cashion, and C. Wandrey, 1992, Oil and gas in the uppermost Cretaceous and Tertiary rock, Uinta basin, UT, in T. D. Fouch, V. Nuccio, and T. C. Chidsey, eds., Hydrocarbon and mineral resources of the Uinta basin, UT and CO: Utah Geological Association Guidebook 2, p Gries, R., 1983, North-south compression of Rocky Mountain foreland structures, in J. D. Lowell, ed., Rocky Mountain foreland

15 1406 Study of a Gas Reservoir basins and uplifts: Denver, Colorado, Rocky Mountain Association Geologists, p Heffer, K. J., and A. B. Dowokpor, eds., 1990, Relationship between azimuths of flood anisotropy and local earth stresses in oil reservoirs, in A. T. Buller, E. Berg, O. Hjelmeland, J. Kleppe, O. Torsaeter, and J. O. Aaser, eds., North Sea oil and gas reservoirs II: The Norwegian Institute of Technology, London, Graham and Trotman, p Johnson, R. C., 1985, Early Cenozoic history of the Uinta and Piceance Creek basins, Utah and Colorado, in R. N. Flores and S. S. Kaplan, eds., Cenozoic paleogeography of west-central United States: Rocky Mountain Section, Society for Economic Paleontologists and Mineralogists, p Johnson, W. E., 1995, Direct detection of gas in Pre-Tertiary sediments?: Leading Edge, v. 14, p Lorenz, J. C., and S. J. Finley, 1991, Regional fractures II: fracturing of Mesaverde reservoirs in the Piceance basin, Colorado: AAPG Bulletin, v. 75, p Lorenz, J. C., N. R. Warpinski, and L. W. Teufel, 1996, Natural fracture characteristics and effects: The Leading Edge, v. 15, p Lucas, P. T., and J. M. Drexler, 1976, Altamont-Bluebell a major naturally fractured stratigraphic trap, Uinta basin, Utah, in North American oil and gas fields: AAPG Memoir 24, p Lynn, H. B., 1986, Seismic detection of oriented fractures: Oil & Gas Journal, v. 84, n. 31, p Lynn, H. B., 1996, Opening Address of 6th International Workshop on Seismic Anisotropy, in Seismic anistropy: Society of Exploration Geophysicists, p Lynn, H. B., and L. A. Thomsen, 1990, Reflection shear-wave data collected near the principal axes of azimuthal anisotropy: Geophysics, v. 55, p Lynn, H. B., K. M. Simon, and C. R. Bates, 1996, Correlation between p-wave AVOA and s-wave traveltime anisotropy in a naturally fractured gas reservoir: Leading Edge, v. 15, p Mueller, M. C., 1991, Prediction of lateral variability in fracture intensity using multi-component shear-wave surface seismic as a precursor to horizontal drilling in the Austin Chalk: Geophysical Journal International, v. 107, p Narr, W., and J. B. Currie, 1982, Origin of fracture porosity example from the Altamont field, Utah: AAPG Bulletin, v. 66, p Nur, A., 1971, Effects of stress on velocity anisotropy in rocks with cracks: Journal of Geophysical Research, v. 76, p Nur, A., and G. Simmons, 1969, Stress-induced velocity anisotropy in rock: an experimental study: Journal of Geophysical Research, v. 74, p Osmond, J. C., 1965, Geologic history of site of Uinta basin, Utah: AAPG Bulletin, v. 49, p Ostrander, W. J., 1984, Plane-wave reflection coefficients for gas sands at non-normal angles of incidence: Geophysics, v. 49, p Queen, J. H., and W. D. Rizer, 1990, An integrated study of seismic anisotropy and the natural fracture system at the Conoco borehole test facility, Kay County, Oklahoma: Journal of Geophysical Research, v. 95, no. B7, p Shuey, R. T., 1985, A simplification of the Zoeppritz equations: Geophysics, v. 50, p Spencer, C. W., 1989, Review of characteristics of low-permeability gas reservoirs in the western United States: AAPG Bulletin, v. 73, p Stone, D. S., 1969, Wrench faulting and Rocky Mountain tectonics: The Mountain Geologist, v. 6, no. 2, p Szpakiewicz, M. J., K. McGee, and B. Sharma, 1986, Geologic problems related to characterization of clastic reservoirs for enhanced oil recovery: Society of Petroleum Engineers, Proceedings of the SPE/ DOE Fifth Symposium on Enhanced Oil Recovery, v. 2, p Thomsen, L. A., 1986, Weak elastic anisotropy: Geophysics, v. 51, p Thomsen, L. A., 1988, Reflection seismology over azimuthally anisotropic media: Geophysics, v. 53, p Watts, R., 1996, Objectives of the U.S. DOE s research: Leading Edge, v. 15, p Zobak, M. L., and M. D. Zobak, 1991, Tectonic stress field of the continental United States, in L. Pakiser and W. Mooney, eds., Geophysical framework of the continental United States: Geological Society of America Memoir 172, p Zobak, M. D., S. Moos, and L. Martin, 1985, Well bore breakouts and in-situ stress: Journal of Geophysical Research, v. 90, p

16 Bates et al ABOUT THE AUTHORS Richard Bates Richard Bates received a B.Sc. degree in geology from the University of Edinburgh in 1986 and his doctorate in geophysics from the University of Wales in He then joined Blackhawk Geometrics as a geophysicist before becoming program manager for Department of Energy Contracts with a focus on multicomponent seismic techniques and hydrocarbon exploration. Currently, he is a lecturer in the sedimentary systems research group, University of St. Andrews. Michele Simon Michele Simon graduated in physics from the University of Houston in She worked for Gulf Oil for eight years and for Marathon Oil for 12 years before joining Lynn Inc. to work on multicomponent seismic for 4 years. Currently she is a reservoir geophysicist in the Permian basin group at Amerada Hess Corporation in Houston, Texas. Heloise Lynn Heloise Lynn holds a master s degree and doctorate in geophysics from Stanford University. She worked for Texaco and Amoco before forming Lynn Inc. in Lynn Inc. specializes in multicomponent seismic acquisition, processing, and interpretation. Heloise Lynn frequently teaches courses on fracture detection using geophysical measurements and on anisotropy.