2 N.R. WARPINSKI, P.T. BRANAGAN, R.E. PETERSON, S.L. WOLHART, J.E. UHL SPE background, the task of developing analysis and processing techniques

Transcription

1 SPE Mapping Hydraulic Fracture Growth and Geometry Using Microseismic Events Detected by a Wireline Retrievable Accelerometer Array N.R. Warpinski, SPE Sandia Natl. Labs, P.T. Branagan, SPE, Branagan & Assoc., R.E. Peterson, SPE, Branagan & Assoc., S.L. Wolhart, SPE, Gas Research Institute, and J.E. Uhl, Sandia Natl. Labs Copyright 1998, Society of Petroleum Engineers, Inc. This paper was prepared for presentation at the 1998 SPE Gas Technology Symposium held in Calgary, Alberta, Canada, March This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box , Richardson, TX , U.S.A., fax Abstract Technology has advanced to the point where microseismic monitoring of hydraulic fractures can provide critical information for fracture optimization. Important elements of a monitoring system include the receivers, the telemetry system, and automatic processing of the vast amounts of data. Procedures and additional data requirements are discussed and examples of the important results which can be obtained are illustrated. Introduction Hydraulic fracturing is a critical technology for the exploitation of natural gas and oil resources, but its optimization has been impeded by an inability to observe how the fracture propagates and what its overall dimensions are. Recent field experiments in which fractures have been exposed through coring 1-4 or mineback 5,6 have demonstrated that hydraulic fractures are not the ideal, symmetric, planar features that are currently envisioned. Instead, they appear to commonly have multiple strands, secondary fractures, height and length asymmetries, and other complexities which make a priori predictions difficult. It is clear that model validation, fluid selection, proppant loadings, problem identification and solution, field development, and many other aspects of fracture optimization have been encumbered by the absence of groundtruth information on fracture behavior in normal field settings. Technology is now becoming available, however, to provide extensive diagnostic information on fracture growth, final size and geometry. Multi-level wireline receiver arrays for downhole passive imaging of fracture behavior have become viable and are demonstrating that hydraulic fractures can be imaged, assessed, and eventually controlled. These receiver arrays require high-quality transducers, well-designed clamping systems, high-speed telemetry, real-time processing capabilities, and careful procedures to be effectively used. This paper discusses this technology, its application and validation, and examples of the value of fracture imaging. It concentrates on a 5-level, accelerometer-based, fiber-optictelemetry system currently being used for microseismic mapping. Background Microseismic theory is firmly rooted in earthquake seismology and requires little additional work for extension to hydraulic fracturing applications. During a hydraulic fracture, the formation undergoes significant stressing in proportion to the net treatment pressure and large changes in the pore pressure in proportion to the difference between the treatment pressure and the reservoir pressure. Both of these changes affect the stability of planes of weakness (such as natural fractures and bedding planes) adjacent to the hydraulic fracture and allow them to undergo shear slippage. 7 The shear slippages are analogous to earthquakes along faults (just much lower amplitude) and, hence, the name microseism, or microearthquake, has been used to describe them. As with earthquakes, microseisms emit elastic waves, but they occur at much higher frequencies and generally fall within the acoustic frequency range. These elastic-wave signals can be detected using an appropriate transducer and analyzed for information regarding the source. Microseismic measurements resulting in images of fracture behavior have been performed in several large-scale field experiments These tests have demonstrated that the technology can be applied under field conditions and can provide significant benefits. With this comprehensive Page 1 of 12

2 2 N.R. WARPINSKI, P.T. BRANAGAN, R.E. PETERSON, S.L. WOLHART, J.E. UHL SPE background, the task of developing analysis and processing techniques is reduced, as there is considerable experience to draw on. The microseismic data can be analyzed using two possible approaches. If receivers are located in several wells, then the microseismic locations can be triangulated upon and accurate locations calculated, just as would be done for an earthquake. The triangulation is accomplished by determining the arrival times of the various waves (usually just the compressional, p, wave and the shear, s, wave) and, knowing the formation velocities, finding the best-fit location to match the arrival times. Unfortunately, multiple, nearby offset wells will only be available in a few comprehensive field experiments and some other isolated cases and would not lend itself to a general field service that could be applied widely. The second approach uses a single, multi-level, vertical array of receivers which can be used to back-locate the microseismic source from a single, nearby offset well, as shown in Figure 1. This approach is more readily accomplished in normal field settings, but requires higher-level technology in the form of a multi-level array and higher sensitivity transducers to capture additional information not needed for triangulation but required for single-well analyses. To extract meaningful microseismic information using this approach, it is necessary to detect the arrival times of the p waves and s waves on as many levels as possible and to accurately record the particle motion of the p wave. In a homogeneous medium, the direction of the p-wave particle motion is the direction of travel of the wave and, thus, points back to the origin point. Detecting the particle motion is much more demanding than just the arrival time, as it requires that good signal-to-noise ratios be obtained and that data be obtained across the full spectral bandwidth of the signal. The application of this second approach is the subject of this paper. Included in the discussion are the technology needs, the theoretical bases for the analyses, validation of the microseismic results, and field examples. A glossary of instrumentation terms is included in an appendix for those readers unfamiliar with this technology. Hardware Receivers. The key to performing microseismic surveys is having the best possible seismic receivers for this application. Normal surface seismic surveys and vertical seismic profiles, which are characteristically at lower frequencies, can acquire excellent data with low-resonance geophones that work optimally in the seismic range (<100 Hz). Microseismic applications, however, require receivers that can detect much higher frequencies. Figure 2 shows a plot of typical frequency ranges for various sonic/seismic technologies where it can be seen that microseisms fall in the middle of the frequency range encountered in oil-field applications, but considerably higher than the other seismic techniques. In addition to the frequency issue, many seismic surveys use stacking to enhance the signalto-noise ratio, but microseismic signals are one-shot events in which the lowest-possible electrical noise floor will provide important benefits. These two considerations drive the receiver requirements for optimal system performance. Frequency Response. Based on data obtained at the GRI/DOE Multi-Site Hydraulic Fracture Diagnostic Project (M-Site), 18 it is clear that microseisms have their energy spread out over a wide frequency range, generally from 200 Hz to greater than 2000 Hz. If the signal-to-noise ratio is sufficiently large, the receiver bandwidth will not be critical for arrivaltime information, as the waves can usually be identified on any number of sensors having a wide variety of characteristics. However, the bandwidth is important for particle-motion information, since a high-fidelity recording of the ground motion is needed to extract the correct directional information. Furthermore, accurate selection of the s wave may require more exact information regarding particle motion to determine phase-shift characteristics. Accurate frequency response is dependent on three issues: transducer response, receiver resonances, and sampling rate. While geophones have adequate response at low frequencies (<200 Hz), accelerometers have far superior characteristics at high frequencies (>200 Hz) and are the transducer of choice for this application. Accelerometers are rapidly becoming smaller, more shock resistant, and better-suited for high temperature applications, i.e., just the qualities required for wellbore applications. Receiver resonances are the key difficulty in pursuing highfidelity recording of the wave motion. To assure highfrequency capabilities with no resonances in the microseism bandwidth, the receiver must be firmly clamped in the wellbore. Unfortunately, the standard swing-arm clamp generally has a resonance around Hz and may have additional resonances at higher frequencies depending on internal design. The receiver used in this advanced application has been designed using modal-analysis techniques to avoid resonances below 2000 Hz. It achieves this capability by using a different clamp-arm design - a right-angle piston drive. However, this system cannot be used in open holes and it will not be suitable for some wellbore configurations, so there are tradeoffs to having better performance. Other designs may also provide good response (e.g., hydraulic or magnetic), but each type of receiver needs to be thoroughly checked. As is the case with the right-angle piston drive, each of these designs carries some penalty associated with its use. The final issue is the sampling rate, which is controlled by the A/D system, telemetry limitations, the data acquisition system, and the data backup system. Experience with M-Site applications have shown that 1/8 msec sampling intervals are needed and faster sampling would be advantageous. The reason for the high sampling-rate requirement is that the microseisms have frequency content up to at least 2000 Hz, so a 1/8 msec sampling rate would provide 4 points per cycle. Four points are adequate to describe a sine wave, but these transient signals are much more complex than a simple sine wave and any improved definition yields benefits in resolution and accuracy. Page 2 of 12

3 SPE MAPPING HYDRAULIC FRACTURE GROWTH AND GEOMETRY USING MICROSEISMIC EVENTS DETECTED 3 Noise Characteristics. Since microseisms are isolated small-magnitude events which cannot be stacked as typical seismic-source signals are, it is critical that the signal-to-noise ratio (SNR) be as large as possible. Obviously the source strength cannot be changed, so it is imperative that the noise be minimized. Two types of noise exist which can detrimentally affect the SNR - cultural and electrical. Cultural noise is often impossible to modify, but occasionally some remedial action in the wellbore or regulation of surface activity helps to mitigate cultural noise levels. For example, if gas is bubbling through open perforations from a zone below the receivers, a bridge plug will often produce significant improvement in the wellbore noise levels. Electrical noises are usually due to the transducer, the electronic components, noise in the supply power, and pickup over unshielded components and wires. It is usually possible to minimize power supply noise (e.g., use batteries), and components with acceptable common-mode-rejection characteristics are available, so the noise issue generally revolves around transducer selection and pickup of stray electromagnetic energy. In this application, the pickup issue has been minimized by digitizing data downhole so that the low-amplitude signals are not distorted during telemetry. Downhole amplification could provide similar benefits by creating a signal that is so large that any added noise is immaterial, however, downhole amplification must be done carefully to avoid introducing additional noise from the power supply, cross talk, or the amplification circuit. Downhole digitization is more complicated and expensive, but is generally a better choice for retrievable instruments. The transducer electrical-noise floor of accelerometers is superior to geophones at the higher frequencies required for microseismic applications. 19 Sleefe et al. 20 have characterized example boreholes and transducer systems, as shown in Figure 3. They found that accelerometers have decreasing noise floors with increasing frequency, which matches the typical borehole characteristics. From a noise perspective, this behavior makes accelerometers a better choice for recording low-amplitude, high-frequency events, assuming that other cultural and electrical noise factors have been optimized. A/D. An effective A/D system for this application must confront three issues: sampling rate, dynamic range, and timing. The sampling rate (1/8 msec) has already been specified by the requirement of measuring frequencies up to at least 2000 Hz. The timing issue, which is insignificant for low frequency applications, requires that all channels on the system be sampled at a fast enough rate that there are no analysis complications due to the time lag between sample points. Example concerns are the arrival-time differences between different levels (usually a minimal problem) and particlemotion analyses at the highest frequencies (at 2000 Hz, this can be significant). This problem can be solved by either a simultaneous sample-and-hold or by an A/D which samples much faster than any of the data analysis requirements. The dynamic range (and bit resolution) is a complicated issue, as the effective bit rate of A/D systems depends on the frequency. In general, at least 16 bit digitization is needed for maximum effectiveness. While A/D devices with digitization rates as large as 24 bits are advertised, it should be remembered that the effective bit rate of such devices at 2000 Hz may be as low as 18 bits. Other standard data acquisition issues, such as anti-aliasing filters, amplification, and input impedence, need to be considered prior to the A/D. Thus, a careful understanding of the A/D system is required to achieve the best results. Telemetry. With multi-level receiver systems, telemetry is a major design issue. Assuming that downhole digitization is employed, a five-level system will have each of 15 channels sending bit words per second. This is a data rate of 1.9 megabits/sec without any control characters, checksums, or other information. This data rate surpasses that of a 7- conductor wireline, as currently used, and thus requires a specialized wireline system. The current 5-level system uses a fiber-optic wireline to handle the data rates, but other specialty multi-conductor cables could be built and used. Fiber optics, with its better communications capabilities, is more attractive for future system enhancements and enlargements (e.g., faster sampling, wider dynamic range, more receivers, etc.), so it was picked as a superior technology. As currently constructed, the system being used sends about 3.1 megabits/sec of data, but the total telemetry rate is 6.2 megabits/sec because of self-clocking (Manchester encoding). Data Management. Once at the surface, large quantities of data must be handled, stored, displayed, and processed. Personal computers (PC s) are now sufficiently powerful to handle this load, providing a portable, versatile, and universally used platform for all aspects of data management. PC s can also be easily networked to provide considerable low-cost computer power. Full backup can be handled on any of various tape or hard disk configurations. A standard file format needs to be specified for portability. For this application the SEG2 format, which was designed specifically as a disk file format, is used. Event Processing. Event processing has three major parts. First, all of the receiver data must be thoroughly searched for events. Second, specific attributes of the event that are required for further analysis must be resolved. Third, the event attributes are used to calculate the microseism locations. Event Detection. Event detection is a relatively straightforward process for which there are several adequate algorithms. One could probably use a simple amplitude trigger with reasonable success, but the likelihood of changes in the background cultural noise due to pumping or other influences make it worthwhile to use an adaptive procedure. McEvilly and Major 21 give an example of a simple and effective procedure using short-term and long-term averages to declare an amplitude-based event. This technique, or adaptations of it, Page 3 of 12

4 4 N.R. WARPINSKI, P.T. BRANAGAN, R.E. PETERSON, S.L. WOLHART, J.E. UHL SPE are commonly used; a modified version of this technique is used for initial event screening with this multi-level system. Event Analysis. The most difficult part of event processing is developing automatic techniques for ascertaining whether the event is a microseism and not noise, and then making an accurate determination of the important attributes: p-wave arrival point, s-wave arrival point, and p-wave particle-motion directionality. On multi-level systems, effective algorithms can be written to comb through both time and space to assure that the microseism is recorded on several levels at the appropriate time. Analysis accuracy is aided by finding methods that pick the p-wave arrival as close as possible to the point where it rises above the noise. S-wave determinations are more difficult, as the shear arrival may be weak or absent depending on where the receiver array is located relative to the radiation pattern of the s wave, or it might be veiled by reflected waves or other interfering waves. Effective s-wave detection can make use of three characteristics that often differentiate it from the p wave: lower frequency, an amplitude shift, and an orthogonal particle motion shift. While amplitude changes can be seen on nearly any receiver system, resonances can often obscure both frequency shifts and changes in polarization, thus, emphasizing again the importance of high-quality receivers. Microseism Location. Microseisms can be located by a wide number of techniques involving some combination of directionality, triangulation, distance calculations, ray tracing, and others. Having applied many different approaches, a twolevel methodology has been adopted. For all initial calculations, including real-time ones, a homogeneous velocity field is assumed for both the p and s waves. A joint regression on both the p and s wave distance equations is used to calculate the elevation of and distance to the microseism. The equation which is minimized is ( ) ( ) ( ) 2 F = wp Vpi tpi to ri ro zi zo n + w s Vsi( tsi to) ( ri ro) ( zi zo) 2, n where the V pi and V si are the p and s velocities, r refers to the horizontal distance, z refers to the elevation, t is the time, the subscript o denotes the origin location and the subscript i denotes the i th receiver. Variables w p and w s are weighting functions which can be used if one of the phases has less certainty than the other. The result of the minimization is the location coordinate, (r o, z o, t o ), which places the microseism in a two-dimensional vertical plane at the time the microseism occurred. A regression of this nature has the added advantage that uncertainties can be calculated directly from regression parameters. To complete the three-dimensional location, the final parameter necessary is the azimuthal direction, θ (i.e., the orientation in the horizontal plane). This parameter is determined by examining the first 1-2 cycles of the p wave for its particle motion. Since the p-wave vibrational vector points in the direction of travel, then it also defines the direction back to the source, assuming that the formation is homogeneous and isotropic. Obviously, this is not the case vertically because of the layered rock structure, but horizontally is appears to be a valid approximation over the relatively short distances at which microseismic mapping is conducted. Thus, knowing the horizontal distance (r o ), the horizontal azimuth (θ), and the elevation (z o ), the microseism source is fully located in space. Accurate time stamping of the microseism also allows them to be temporally located. Although a homogeneous velocity field is a good approximation if there are many receivers, the limited number of receivers in a wireline system may not be sufficient to provide adequate results if the velocity structure varies significantly. For such situations, a higher-level processing methodology is applied. The technique applied here, that of Vidale and Nelson, 22 is ideal for field applications. The algorithm starts by breaking the formation up into a gridded space. For a single multi-level receiver system, only a two-dimensional grid is required (e.g., the vertical plane joining the microseism and the receiver array). Then the minimum travel time from every grid point to every receiver is calculated and stored in a file. This task is relatively time consuming, but it can be performed prior the fracture treatment (once the receiver locations and velocity structure are known) and can be performed on a PC platform. Once microseismic events are processed, the actual locations are found by determining the optimum location to minimize the residuals of the arrival times of all phases at all stations. Any number of functions can be used to perform the minimization. As with the joint p-s regression used for a homogeneous medium, the Vidale-Nelson algorithm for this 2- D plane only provides the horizontal distance and elevation. The azimuth needs to be determined from the p-wave particle motion. Interpretation of Microseismic Maps Once the microseisms are located, they can be plotted on a map to produce a microseismic image of the fracture. There remains the question of how well the microseismic image matches the actual fracture. Some validation experiments will be shown later, but initial theoretical considerations are important for understanding when and where the application is valid. Conceptually, one can envision that a hydraulic fracture causes two significant changes in the reservoir. Because of the dilation of the fracture, there are significant solid stresses which develop around the fracture (on the order of the net pressure) and because of the leakoff, there are even larger changes in the pore pressure surrounding the fracture, where the change is equal to the difference between the total fracture pressure and the reservoir pressure. Considering the solid effects first, the largest stresses that exist around a fracture are those in the near-tip region, as very Page 4 of 12

5 SPE MAPPING HYDRAULIC FRACTURE GROWTH AND GEOMETRY USING MICROSEISMIC EVENTS DETECTED 5 large tensile stresses form ahead of the crack. If a stability analysis is performed in this region, it is found that large amounts of shear stress are generated, and pre-existing planes of weakness, such as natural fractures and bedding planes, are unstable and will undergo shear slippage. As noted before, it is the shear slippages that emit the microseisms. These shear slippages can occur off to the side (usually less than a few tens of feet) and slightly in front of the tip, i.e., the hydraulic fracture is embedded within an envelope of microseismic events. Since the fracture has a tip which circumscribes the entire hydraulic fracture, it is clear that stress-generated microseisms are generated continuously as the tip expands, as long as there are favorable weakness planes and sufficiently high pressures in the fracture. This behavior is favorable for microseismic imaging, as it suggests that microseisms will be good markers of the crack-tip location. The leakoff mechanism is an even more efficient means of creating microseisms, as the changes in pore pressure can be much larger than the changes in stress (except very close to the crack tip). As fluid leaks off into the formation and pore pressure increases, the net normal stress on any plane of weakness decreases, decreasing the frictional force. Thus, leakoff causes a destabilization of any planes of weakness in areas where leakoff has occurred, essentially around the body of the fracture. This destabilization allows shear slippage to occur if favorably oriented planes of weakness are present and if the destabilization is sufficient. However, these microseisms should occur later in time and back toward the central regions of the fracture. Based on theoretical considerations, it is therefore expected that microseisms will be of two types, those that occur around the tip due to stress and those that occur wellbehind the tip due to leakoff. The orientation of the weakness planes, the original in situ stresses, the net pressure, the reservoir pressure, the size of the fracture, the leakoff characteristics of the reservoir, and the strength characteristics of the weakness planes and the rock itself will all have an impact on the generation and expanse of microseisms generated by a fracture treatment. For example, in hard sandstones with low permeability, microseisms will generally fall within a narrow band surrounding the fracture and provide clear image information. On the other hand, fractures in high permeability rocks, poorly consolidated rocks, and fully water or oil saturated reservoirs will likely induce a broad spread of microseisms. An understanding of the reservoir which is imaged will help in the interpretation process. Additional Required Data The extra information required for an accurate imaging test are good formation velocities, accurate location data, and some type of source shots in the treatment well to check orientations, check velocities, and test capabilities. Without this type of information, the image will be little better than a guess. While the optimum velocity information is that derived from a cross-well tomogram, a high-quality sonic log that provides both compressional and shear velocities is usually perfectly adequate. Velocity data must be available throughout the entire interval over which receivers are placed, not just the zone being stimulated. Surface well-location surveys and downhole deviation surveys are needed to adequately locate the receivers relative to the fracture interval. Since wells can deviate many tens of feet under typical situations, deviation surveys are mandatory or fracture azimuths and other parameters may be seriously in error. Most receivers require external orientation of the transducer horizontal axes, and this is usually done by providing source shots in the treatment well. The source shots are usually perforations, and the actual perforation of the fracture interval can be used, but other sources such as airguns are adequate and often provide other useful information. Source shots not only orient the receivers, they can also provide a check on the log-derived velocity and the capability of the receiver array to monitor for microseisms. The advantage of using an airgun is that a complete scan of the test interval can be performed. Examination of the data from this scan provides information on vertical velocities, head waves, blind spots, and effects due to other formation heterogeneities and anisotropies. If pre-fracture orientation fails for any reason (e.g., the receiver must be unclamped and moved after the orientation was performed), there is also a fall-back orientation procedure that involves using the initial few microseisms to locate the wellbore, under the assumption that initial microseisms are generally scattered about the treatment wellbore. However, there are times when microseisms are not observed until the fracture becomes sufficiently energetic, and this approach may produce poor results under such a situation. Procedure The procedure for conducting a microseismic survey is generally straightforward. Given the conditions of the treatment and monitor well and the velocity structure, the spacing of the receivers is selected and their location in the monitor well is chosen. The sonde spacing is usually chosen so that the total aperture of the array is about half the distance between the two wells (for acceptable triangulation purposes), but it is also felt that apertures (total distance between top and bottom sonde) greater than about 800 ft (240 m) do not provide much additional advantage because the waves must travel through so many layers and layer interfaces that they become distorted or attenuated. The location of the receivers is also chosen so that there is at least one receiver in the formation to be fractured and one above and below this zone. Receivers should not be placed close to bed boundaries with strong velocity contrasts (e.g., at shale/sandstone interfaces). On site, the receivers are checked out and tap tests are performed to ascertain that all wiring is right and the receiver axes are correctly located. Receivers are installed in the well, positioned, and clamped in place. Page 5 of 12

6 6 N.R. WARPINSKI, P.T. BRANAGAN, R.E. PETERSON, S.L. WOLHART, J.E. UHL SPE Connections (usually a fiber-optic land line) between the source shots and the receiver array are made up in order to provide accurate timing of the source for velocity checks. Source shots are performed, usually the day before the treatment, and the data gathered. Analysis of the orientation data needs to be performed immediately to assess whether the orientation shot was acceptable. After orientation, the receiver system should be turned on for a considerable period to survey background activity that may be confusing to the subsequent survey. Oftentimes, background seismic activity has considerably different characteristics than that associated with a hydraulic fracture. On the day of the fracture treatment, the array should run automatically: declaring events, processing data, and generating images. If necessary, the treatment can be replayed from the full-backup system and data reassessed under different event conditions. The last step is the generation of the final maps, which may include effects of the velocity structure. Microseismic Viewing Distance One of the key unknowns for microseismic imaging at this time is the distance at which microseisms can be detected and located. This question cannot be answered directly because it depends on the source strength of the microseisms, which is a function of both treatment and reservoir conditions, and the attenuation characteristics of the formation, which are largely unknown. There are, however, a number of completed surveys that can provide some idea of the distance at which microseismic imaging is possible. Some of these surveys used old technology and others may not have had any events at distances greater than those listed, so these distances are probably underestimates. Maximum viewing distances are: Fenton Hill - granite ~5000 ft (~1500 m) Camborne - granite ~4500 ft (~1400 m) Austin chalk ft (760 m) Frio sandstone ft (455 m) North Slope sandstone ft (915 m) Barnett shale ft (915 m) Mesaverde sandstone ft (245 m) Frontier sandstone ft (305 m) Cotton Valley sandstone ft (455 m) From these results, one would speculate that a reasonable expectation is that microseisms can be observed within about 1500 ft (450 m) of the observation well (i.e., nominal 160-acre [65 ha] well spacing), and much farther in specific reservoirs. M-Site Diagnostic Tests The M-Site diagnostic experiments have been thoroughly documented elsewhere, so only a few pertinent results are provided here. These results include validation of the microseismic images and examples of unexpected behavior that could only be detected by imaging technology. M-Site Overview. M-Site was an experiment facility codeveloped by the Gas Research Institute and the US Dept. of Energy. 18 It was located in the Piceance basin of western Colorado with a target in sedimentary Mesaverde rocks. This site, which functioned as a hydraulic-fracture diagnostic laboratory from , consisted of a treatment well, MWX-2, flanked by an instrumented monitor well and a second observation well for wireline receiver arrays (MWX-3). Figures 4 and 5 show plan and side views of the site layout. As seen in Figure 5, the monitor well consisted of 30 tri-axial receiver and 6 bi-axial tiltmeter stations, all cemented in place across from two sandstone intervals which were hydraulically fractured. The observation well, MWX-3, was used for fielding wireline-run multi-level accelerometer arrays. From these two wells, hydraulic fractures in MWX-2 were monitored. Also, two deviated lateral wells, IW-1B and IW-1C, are shown in Figure 4. These wells were used for several aspects of the validation. The experiments conducted in the M-Site Project were designed around a series of hydraulic fracture treatments executed in three low-permeability sandstone intervals (termed the A, B, and C sands) in the uppermost 700 ft (213 m) of the Mesaverde Group. Conceptually, the M-Site plan was to perform research work from the deepest (A sand) to shallowest (C sand) interval, "using up" each interval before beginning testing in the next unfractured sand interval. The set of experiments conducted in each sand interval were designed to be increasingly more complex and build upon the results of the previous experiments. Overall, four injections/fracture treatments were performed in the A sand (1-A to 4-A), seven in the B sand (1-B to 7-B), and six in the C sand (1-C to 6-C). An intersection well (IW-1B) was cored through the B- sand fractures after all of the treatments were completed. This core provided unique information on the characteristics of the fracture system created by the injections. A second intersection well, IW-1C, was drilled through the C sandstone prior to any fracture injections and was used for various types of intersection tests. Microseismic Validation. The downhole tiltmeters and intersecting wells were specifically placed to provide validation of the microseismically derived geometry. Fracture height was independently determined by the tiltmeter array, which measured the deformation of the reservoir due to the inflated fracture. The azimuth of the fractures was independently measured by coring through the B-sand fractures. The length was validated in the C-sand tests by propagating a hydraulic fracture into the intersection well. Height Validation. The downhole tiltmeter array at M-Site measures the earth s mechanical deformation resulting from the inflated hydraulic fracture. As a result, the height and width of the fracture which produce this deformation can be calculated. The net pressure in the fracture (measured downhole in the treatment well) provides the other piece of information to accurately constrain the result. Five of the B-sand injections had both microseismic images and high-quality tiltmeter-derived heights. A comparison of these is given below: Page 6 of 12

7 SPE MAPPING HYDRAULIC FRACTURE GROWTH AND GEOMETRY USING MICROSEISMIC EVENTS DETECTED 7 Injection Microseismic Height Tilt Height 3-B 55 ft (17 m) 53 ft (16 m) 4-B 55 ft (17 m) 53 ft (16 m) 5-B 75 ft (23 m) 67 ft (20 m) 6-B 80 ft (24 m) 67 ft (20 m) 7-B 135 ft (41 m) 135 ft (41 m) The agreement between the two techniques is within the accuracy of locating the microseisms and shows that microseismic images provide an accurate estimate of the induced fracture height, even as fractures grow into other lithologies. Azimuth Validation. Wellbore IW-1B, which was cored through the B-sand fractures after completion of the testing, located multiple fractures over a 2-3 ft (~1 m) interval. A downhole deviation survey was used to find the exact location of the fracture intersection and the azimuth of the fracture was then projected to that point. The corehole azimuth projected out at N72 W, while the six treatments which could be imaged had azimuths ranging from N74 W to N77 W, with an average of N75 W. Azimuth agreement within 3 indicates a superior accuracy to any other technique for estimating stress orientation and fracture azimuth. Length Validation. Length validation was achieved using the special circumstance of a pre-drilled deviated well within the C sandstone which could be intersected by a propagating hydraulic fracture. During the second injection (2-C), the intersecting wellbore was instrumented with a downhole pressure gage and shut-in to observe the pressure transient that would occur when the high-pressure fracture intersected the well. The treatment consisted of a 40lb/1000 gal (4.8 kg/m 3 ) linear gel pumped at 20 bpm (3.2 m 3 /min) until the well was intersected, at which time pumping was quickly terminated. This intersection occurred after 132 bbl (21 m 3 ) and resulted in the microseismic image shown in Figure 6. Figure 6 is a plan view of the microseismic locations as taken from both monitoring wells. A segment of the deviatedwell trajectory through the fracture plane is also indicated. The microseisms accurately describe the fracture length within 25 ft (8 m), essentially the accuracy of their locations. These sets of experiments demonstrate that the microseismic method is an accurate technology for imaging fracture growth in this type of reservoir. Interpretation of the microseismic images in this fluvial reservoir does not appear to be an issue. M-Site Key Observations. The detailed testing at M-Site provided some surprising results that were not expected based on recent accepted fracture theory. A brief summary of these findings, with a few selected microseismic images, follows. Favorable height containment. Most fracture injections at M-Site (but not all) resulted in fracture heights that were considerably less than expected based upon the measured stress contrasts and net pressures. An example of one such case is injection 5-C, a 480-bbl (76 m 3 ) treatment using a 40lb/1000 gal (4.8 kg/m 3 ) cross-linked borate gel pumped at 30 bpm (5 m 3 /min). This test reached a net pressure of 1400 psi (9.7 MPa), while the confining stresses were less than 1000 psi (7 MPa) in the vicinity of the fracture intervals. As seen in side view in Figure 7, the fracture reached a length of nearly 500 ft (152 m), but the height growth was only about 30 ft (9 m) upward and 30 ft (9 m) downward. Current fracture models would predict almost as much height as length. Another example, which may be more informative as to mechanisms, is from injection 4-C. This test had 980 bbl (156 m 3 ) of the same cross-linked fluid injected at 40 bpm (6.4 m 3 /min). Possibly because of a shut-in period near the beginning of the treatment, this test resulted in highly asymmetric height growth, as shown in Figure 8. The left wing had some downward growth, while the right wing had moderate downward growth. The right wing had considerable upward growth, while upward growth on the left wing was small. The limited height growth seen in most tests and the asymmetric growth seen in many of them suggests that features other than stress are adding to fracture containment. The most likely feature is the complex lithology, which may make fracture growth difficult across such a layered medium, much like a composite material is designed to do. Rapid Lateral Fracture Extension with Thin Fluids. One of the features noted in all test zones was that fracture treatments with thin fluids (water, linear gel) had rapid length growth with minimal height extension in the early part of the treatment. The cross-linked fluids, on the other hand, showed a much more deliberate fracture advance which tended to be upward as well as outward. Figures 9 and 10 show a comparison of the fracture geometries achieved in injections 1- C and 2-C, respectively, after 95 bbl (15 m 3 ) of fluid, both pumped at the same rate of 20 bpm (3.2 m 3 /min), but injection 1-C having a 40lb/1000 gal (4.8 kg/m 3 ) borate cross-linked gel and injection 2-C having a 40lb/1000 gal (4.8 kg/m 3 ) linear gel. As seen in this example, there are considerable differences in the resultant fracture geometry, due to some combinations of viscosity, leakoff, elasticity, and possibly other effects. Complex Fracture Geometries. One of the more unexpected results was the development of secondary fractures during several of the cross-linked gel experiments. While most of the plan-view maps look similar to Figure 6, with a band of microseisms surrounding the fracture azimuth, two of the tests apparently reached conditions necessary to spawn secondary fractures. Figure 11 shows a plan view of the image of injection 6-C, a cross-linked gel injection, where two fracture strands have apparently initiated in directions oblique to the main fracture trend after only 15 minutes of injection. This particular test eventually resulted in spawning a horizontal fracture as well. Secondary fractures and horizontal T fractures are usually considered features of coal fractures and fractures in rocks such as diatomite. These tests indicate that they can occur in any formation. Length Extension Features. Besides the rapid lateral extension seen with the thinner fluids, these tests have also Page 7 of 12

8 8 N.R. WARPINSKI, P.T. BRANAGAN, R.E. PETERSON, S.L. WOLHART, J.E. UHL SPE provided evidence of the way fractures grow and some of the other features which can affect that growth. One example is a relatively staccato growth process, whereby the fracture appears to grow in spurts, usually on one side and then on the other. The fracture image often shows growth out to a certain dimension, a brief stop, and then continued propagation with a flurry of microseisms. When that flurry ends, the next flurry may appear on the opposite wing, resulting in disjointed growth. One reason for this behavior may be the nature of the fluvial reservoirs in which these tests were conducted. Fractures may propagate through separate point bars making up the reservoir, reach the boundary and stop, and then break out somewhere else where the resistance is less. A second feature seen in a few tests is redirection of growth after shut-in periods. The image shown in Figure 8 is an example of the asymmetry that resulted from a shut-in period early in treatment 4-C. After pumping resumed, most of the subsequent growth occurred on the right wing, resulting in a highly asymmetric length pattern. A similar result occurred during injection 4-B, when a leak caused a 6-minute shut down near the end of the treatment and all subsequent growth was again on a single wing. The amount of resultant asymmetry is shown in Figure 12. All microseisms between 280 and 450 ft (85 and 137 m) on the right wing occurred after the leak. No microseisms occurred on the left wing after the leak. Finally, the differences in fracture extension after shut-in were surprising. Injections in the B sand resulted in very little growth after shut in, with the exception of one of the initial water injections. On the other hand, all injections in the C sand resulted in a considerable amount of additional length penetration. Figure 13 shows the final length developed by injection 1-C, in comparison to the length at shut in as shown in Figure 8. The approximately 100 ft (30 m) of additional length growth is typical of most C sand injections. Applications This technology lends itself to multiple applications associated with fracturing, as well as non-fracturing processes. Obvious applications are problem wells and infill programs where fracture length and azimuth are important parameters for designing an infill program. A promising application is the monitoring of multi-zone completions to determine which zones are being stimulated. For new fields, the technology can be applied early to determine optimum placement for subsequent wells. Some of the key observations should also lead to optimized treatment designs. Other applications include monitoring of disposal wells, monitoring of gas storage reservoirs and wells, identification of natural fracture systems in reservoirs, monitoring of subsidence associated with production, and diagnoses of many other possible completion and production activities. Conclusions Microseismic imaging has advanced to the point where it can now be used for application in routine industrial fractures. Using multi-level, accelerometer-based receiver systems, fiberoptic-wireline telemetry, and PC-based processing, fractures will soon be monitored real-time or near-real-time for many applications. Use of this technology has already provided numerous examples of unexpected fracture behavior. Results include better-than-predicted height containment, the observation of secondary fracture strands spawned in standard field environments, rapid lateral length extension when fracturing with thin fluids, and other results that are not normally considered in fracture design or evaluation procedures. Acknowledgments The authors would like to thank the Gas Research Institute for sponsoring this research and the US DOE for co-funding the filed experiments at M-Site. References 1. Warpinski, N.R., Lorenz, J.C., Branagan, P.T., Myal, F.R. and Gall, B.L., Examination of a Cored Hydraulic Fracture in a Deep Gas Well, SPE Prod. & Fac., 45, , August Fast, R.E., Murer, A.S. and Timmer, R.S., Description and Analysis of Cored Hydraulic Fractures, Lost Hills Field, Kern County, California, SPE 24853, SPE ATCE, Washington, DC, Oct. 4-7, Branagan, P.T., Peterson, R.E., Warpinski, N.R., and Wright, T.B.: Characterization of a Remotely Intersected Set of Hydraulic Fractures: Results of Intersection Well No. 1-B, GRI/DOE Multi-Site Project, SPE 36452, 1996 ATCE, Denver, Oct Branagan, P.T., Warpinski, N.R., Peterson, R.E., Hill, R.E., and Wolhart, S.L.: Propagation of a Hydraulic Fracture into a Remote Observation Wellbore: Results of C-Sand Experimentation at the GRI/DOE M-Site Project, SPE 38574, 1997 SPE ATCE, San Antonio, TX, Oct Warpinski, N.R. and Teufel, L.W., Influence of Geologic Discontinuities on Hydraulic Fracture Propagation, JPT, 39 pp , February Jeffrey, R.G., Settari, A. and Smith, N.P., A Comparison of Hydraulic Fracture Field Experiments, Including Mineback and Geometry Data, with Numerical Fracture Model Simulations, SPE 30508, 1995 SPE ATCE, Dallas, TX, October Warpinski, N.R., "Interpretation of Hydraulic Fracture Mapping Experiments," paper SPE 27985, Tulsa Centennial Petroleum Engineering Symposium., Tulsa, OK, , Aug. 1994,. 8. Albright, J.N. and Pearson, C.F., "Acoustic Emissions as a Tool for Hydraulic Fracture Location: Experience at the Fenton Hill Hot Dry Rock Site," SPEJ, 22, pp , August Thorne, B. J., "An Assessment of Borehole Seismic Fracture Diagnostics," SPE 18193, 63rd SPE ATCE, Houston, TX, pp , October 2-5, Hart, C. M., Engi, D., Fleming, R. P. and Morris, H. E., "Fracture Diagnostics Results for the Multiwell Experiment's Paludal Zone Stimulation," SPE 12852, SPE/DOE Unconventional Gas Recovery Symposium, Pittsburgh, PA, Page 8 of 12

9 SPE MAPPING HYDRAULIC FRACTURE GROWTH AND GEOMETRY USING MICROSEISMIC EVENTS DETECTED 9 pp , May 13-15, Vinegar, H.J., Wills, P.B., De Martini, D.C., Shlyapobersky, J., Deeg, W.F.J., Adair, R.G., Woerpel, J.C., Fix, J.E. and Sorrells, G.G., "Active and Passive Seismic Imaging of a Hydraulic Fracture in Diatomite," JPT, 44, pp , 88-90, January Vandamme, L., Talebi, S. and Young, R.P., Monitoring of a Hydraulic Fracture in a South Saskatchewan Oil Field, JCPT, 33, 27-33, January Brady, J.L., Withers, R.J., Fairbanks, T.D. and Dressen, D., Microseismic Monitoring of Hydraulic Fractures in Prudhoe Bay, SPE28553, 69 th SPE ATCE, New Orleans, LA, Sept , Warpinski, N.R., Wright, T.B., Uhl, J.E., Engler, B.P., Drozda, P.M. and Peterson, R.E., Microseismic Monitoring of the B- Sand Hydraulic Fracture Experiment at the DOE/GRI Multi- Site Project, SPE 36450, 1996 ATCE, Denver, CO, October Truby, L.S., Keck, R.G. and Withers, R.J., Data Gathering for a Comprehensive Hydraulic Fracture Diagnostic Project: A Case Study, paper SPE 27506, 1994 IADC/SPE Drilling Conf., Dallas, TX, Feb Warpinski, N.R., Branagan, P.T., Peterson, R.E., Fix, J.E., Uhl, J.E., Engler, B.P., and Wilmer, R.: Microseismic and Deformation Imaging of Hydraulic Fracture Growth and Geometry in the C-Sand Interval, GRI/DOE M-Site Project, SPE 38573, 1997 SPE ATCE, San Antonio, TX, Oct Walker, R.N., Cotton Valley Hydraulic Fracture Imaging Project, SPE 38577, 1997 SPE ATCE, San Antonio, TX, October Peterson, R.E., Wolhart, S.L., Frohne, K.-H., Branagan, P.T., Warpinski, N.R., Wright, T.B.: Fracture Diagnostics Research at the GRI/DOE Multi-Site Project: Overview of the Concept and Results, SPE 36449, 1996 SPE ATCE, Denver, CO, October Malott, C., Theoretical Limitations of Microseismic Transducer Systems, Third Conference on Acoustic Emission/Microseismic Activity in Geologic Structures and Materials, Penn State Univ., , October Sleefe, G.E., Warpinski, N.R. and Engler, B.P., The Use of Broadband Microseisms for Hydraulic Fracture Mapping, SPE26485, 68 th SPE ATCE, Houston, TX, Oct. 3-6, McEvilly, T.V. and Majer, E.L.A., A real-time field processor for microearthquake networks, Annual Report, Earth Sciences Division, Lawrence Berkeley Laboratory, Nelson, G.D. and Vidale, J.E., Earthquake Locations by 3D Finite Difference Travel Times, Bull. of the Seismological Society of America, 80, p. 395, April Batchelor, A.S., Baria, R. and Hearn, K., Monitoring the Effects of Hydraulic Stimulation by Microseismic Event Location: A Case Study, SPE 12109, 58 th ATCE, San Francisco, CA, October 5-8, Phillips, S., Microseismic Fracture Mapping in the Giddings Field, TX, personal communication, Withers, R.J. and Rieven, S., Fracture Development during Cuttings Injection Determined by Passive Seismic Monitoring, Expanded Abstract, 66 th Annual International Mtg., SEG, Denver, CO, Nov 10-15, Walker, R.N., Cotton Valley Hydraulic Fracture Imaging Project, SPE 38577, 1977 SPE ATCE, San Antonio, TX, October 5-8. Appendix - Glossary of Terms A/D system. The analogue-to-digital conversion system. Aliasing. The distortion of digitized signals caused by an insufficiently fast sampling rate. The general sampling rule is that the sampling frequency must be at least twice that of the highest frequency component of the signal. Anti-aliasing filter. A low pass filter inserted prior to the A/D system to prevent aliasing of the signal. Bandwidth. The frequency range of interest, as for example the frequency range over which a microseism has significant energy content or over which a transducer is capable of measuring. Bit resolution. The smallest interval into which the recording range is divided. For example, 16 bits = 2 16 = 65,536 discrete intervals which can be sampled. If the maximum unipolar voltage of a 16-bit A/D system is 1 volt, the bit resolution of this system is 1/65,536 or volts. Common-mode rejection. The ability of an electronic system to cancel any electronic noise pick-up that is common to both the positive and negative polarities. Dynamic range. The spread between the lowest and highest amplitudes to be detected and recorded by a system. This is usually expressed in db (decibels), which is a log scale. For amplitudes, 20dB = 1 order of magnitude. (e.g., 100 db = 10 5, so signals ranging from 0.01 mvolt to 1 volt could be recorded. Elastic wave. Waves or vibrations in a solid body having sufficiently small amplitudes that Hooke s law is applicable. Electrical noise floor. The highest noise level due to electronics and power supply only. (e.g., the noise level if the system were operating in a perfectly vibration-free room). Low-resonance geophones. Geophones are characterized by the resonant frequency of their spring mass systems. Typically, geophones can perform adequately at frequencies up to about 25 times the resonant frequency. Thus, a low-resonance 10-Hz geophone could adequately detect vibrations up to about 250 Hz. Modal analysis. Analysis of the modes of vibration of a body or system. Particle motion. The vibratory motion of a particular element or particle as a wave impinges on it (Lagrangian sense). Radiation pattern. The wave amplitude distribution as a function of location. For example, the radiation pattern of the s wave from a fault slip has maximum amplitude at 45 angles to the slip plane. Sampling rate. The rate at which data are sampled (e.g., a 1/8 msec sampling rate yields 8000 samples per second). Simultaneous sample and hold. A system which holds the voltage levels of all channels for a sufficiently long time Page 9 of 12