MRCH 2014 The etter usiness Publication Serving the Exploration / Drilling / Production Industry New 3-D Seismic Vector ttribute Explains Hydraulic Fracture ehavior y Martin Haege Shawn Maxwell Lars Sonneland nd Mark Norton CLGRY nalysis of about 100 production logs from horizontal wells in multiple shale gas reservoirs indicates that nearly 30 percent of all perforation clusters do not contribute to production. To optimize well placement and completion strategies and improve unconventional shale play economics, more operators are trying to unravel how petrophysical and geomechanical properties influence hydraulic fracture propagation. Induced fracture geometries along horizontal laterals in shale reservoirs vary from simple linear planes to complex branching networks. Primary reasons include the textural anisotropy, mineralogy, and varying diagenetic histories of various shales; differences in in-situ horizontal stress ratios; and the density and distribution of natural fractures as well as the rock s weak zones. Pre-existing faults can compromise reservoir stimulation by, for example, diverting hydraulic fractures into waterbearing zones. Therefore, a detailed understanding of the natural fault network is essential. FIGURE 1 Figure 1 depicts the location of three Montney horizontal laterals and a monitor well, showing the select microseismic events that are sized by magnitude and color-coded to indicate from which well they come. Grid size is 100 meters. Rock Fabric To characterize unconventional reservoirs more accurately, operators are finding it increasingly critical to integrate independent data regarding surface reflection seismic and microseismic. During and after multistage hydraulic fracturing treatments, passive microseismic surveys reveal the approximate location, direction and length of induced fractures. efore drilling or completing a well, certain attributes extracted from surface 3- D reflection seismic data can locate reservoir rock discontinuities such as faults and pre-existing fractures that may impact stimulation efficiency. In striving to image these larger features, however, traditional seismic fault extraction workflows tend to smooth or average away most of the finer-scale discontinuities and pre-existing zones of weakness known as rock fabric. Until recently, these subtler features, which often test the limits of surface 3-D reflection seismic resolution, typically were treated as noise. Nevertheless, research indicates that the prevalence of rock fabric can significantly impact hydraulic fracture behavior, completion quality and subsequent production. The following research study demonstrates how integrating microseismic data with a new 3-D seismic vector attribute can more precisely characterize the rock fabric in a target shale reservoir. Combined with other, more familiar analyses of geomechanical and geophysical attributes such as Poisson s Ratio mapping the rock fabric should enable geoscientists and engineers to make better informed decisions about well placement and completion. Unexpected Fracture ehaviors The Upper Montney formation of northeastern ritish Columbia consists of fine to very fine grained siltstone with low matrix permeabilities in the micro- Darcy range, and porosities from 3 to 6 percent. To achieve economic gas pro- Reproduced for Schlumberger with permission from The merican Oil & Gas Reporter www.aogr.com
FIGURE 2 Figure 2 illustrates vectors overlaid on a vertical seismic section. To the left is a normal vector field and to the right are rotation vectors, weighted by the magnitude of the rotation of the 3-D normal vector field. duction from a target interval at about 1,750 meters true vertical depth, three wells were drilled with 1,500-meter long horizontal sections and completed with multistage fracture stimulation. During treatment, microseismic data were recorded in a near-vertical monitoring well in the center of the study area. To ensure a uniform dataset unbiased by the distance-dependent detection threshold, the study used only 10 of the 21 stages and 992 of the 3,695 events closest to the monitor well (Figure 1). Initial analysis of the location and magnitude of microseismic events revealed one well in which several hydraulic fractures behaved unexpectedly. First, a larger number of high magnitude events occurred along well C in the study area s southeast portion than they did around wells or to the northwest. Second, those microseismic events were not distributed evenly around well C. Instead, most of them propagated on the lateral s northeast rather than southwest side. Third, the shape of the induced fracture network around well C appeared more scattered and heterogeneous than the more linear patterns associated with the other two wells. To unravel the meaning of these observations, Progress Energy teamed with Schlumberger for an in-depth research study of the Montney reservoir s smallscale rock fabric, comparing the passive seismic data with a new 3-D seismic vector attribute. Mapping natural faults and fractures manually can be time-consuming and subjective. This is why geoscientists commonly run the 3-D seismic reflection volume through an enhanced edge detection algorithm to highlight and extract significant spatial discontinuities. However, as already noted, enhancing the signal-tonoise ratio to identify larger-scale features in typical fault extraction workflows may suppress or smooth away most of the subtle features. The new seismic vector attribute, developed by the Schlumberger Research Center in Stavanger, is a polynomial reconstruction of the seismic traces that enables detailed rock fabric analysis even at subseismic resolution. The magnitude of rotation of the 3-D normal vector field captures subtle, small-scale changes in seismic reflections (Figure 2). These changes effectively characterize subtle FIGURE 3 pre-existing discontinuities or faults in the imaged reservoir rock that normal fault extraction techniques generally fail to depict. s such, the two methods can be complementary. To investigate how rock fabric influences hydraulic fracture behavior in the Montney study area, it was necessary to properly smooth the 3-D seismic vector attribute volume. Comparing the XYZ location of a particular microseismic event which is actually somewhere within an uncertainty ellipsoid, rather than a precise point in space with the corresponding location within a surface 3-D reflection seismic volume, involves a number of uncertainties. First, the velocity models used to process the surface seismic and microseismic datasets may be different. Second, microseismic receiver coverage and the distance between source and receiver may cause uncertainties in the exact event location. Third, microseismic events that are recorded during stimulation typically have a source radius shorter than 10 meters, which may be close to the limit of 3-D seismic resolution. Finally, the surface seismic acquisition design can bias subsurface illumination. Therefore, the new 3-D seismic vector attribute volume was smoothed with a Gaussian filter, which effectively honored all the spatial uncertainties while it preserved as much geological detail as possible. Unlike the typical enhancement process for large-scale faults, smoothing done throughout this 3-D volume varies according to the amount of uncertainty associated with each XYZ location. s a result, the research team meaningfully integrated information from both The image on the left depicts the degree of rock fabric (in which darkness signifies greater amounts) mapped directly onto individual microseismic events, sized by magnitude. The figure on the right depicts a rock fabric density map.
seismic datasets and began to make sense of well C s unexpected fracture behaviors. Integrated Investigation The first stage of the integrated investigation was to explore the relationship between the degree of rock fabric present and the magnitude of microseismic events that occurred in the same area. The strength, or seismic moment, of a microseismic event reflects the amount of fracture movement. The greater the movement, the higher the seismic moment and the magnitude. related measure of seismicity, the b-value, reflects the relative number of small to large magnitude microseismic events. The higher the b-value, the greater the number of small events. Global earthquake seismology demonstrates that the earth s tectonic b-value is approximately 1.0. Thus, b-values significantly greater than 1.0 imply the presence of many more low magnitude events than one would expect to occur naturally. nalyzing the seismic moment and b-value of microseismic events, therefore, can provide independent validation of information about rock fabric that is derived from the 3-D reflection seismic data. For example, in relatively homogeneous reservoir rock where there are no pre-existing small-scale discontinuities (hence a low degree of rock fabric) one would expect hydraulic fracturing to generate a greater number of low magnitude microseismic events. In other words, the rock would exhibit lower seismic moment FIGURE 4 FIGURE 5 Figure 5 shows microseismic events sized by magnitude and overlaid on maps of minimum Poisson s Ratio (left), and mean rock fabric (right). and b-values greater than 1.0. In more heterogeneous reservoir rock, where there are considerable pre-existing discontinuities (hence a high degree of rock fabric), one would expect hydraulic fracturing to reactivate many of these weak zones, producing a smaller number of low magnitude microseismic events. In other words, it would exhibit higher seismic moment and lower b-values closer to the typical tectonic value of 1.0. This, in fact, corresponded exactly with observations in the study area after rock fabric density was compared with maps of seismic moment and b-values. The 3-D seismic vector attribute was mapped directly onto individual microseismic events (Figure 3). This was used to compute a rock fabric density map (Figure 3). Figure 4 shows a contour map of the logarithm of seismic moment density and Figure 4 displays a contour map of the b-value with the microseismic events overlaid on both maps. distinct correlation was observed between the cluster of events with a high seismic moment and a low b-value of about 1.0 that surrounded well C, as illustrated in Figure 4 and Figure 4, and the higher degree of rock fabric in the southeast portion of the study area (Figure 3). This suggested that hydraulic fracturing was reactivating small-scale pre-existing faults and fractures in the southeastern region. Regions with a low degree of rock fabric around wells and correlated with low seismic moment and high b- values greater than 1.5, which suggested these were more homogeneous rocks lacking natural weak zones. In addition, the scattered shape of the induced fracture network to the northeast of well C corresponded with the higher density of rock fabric on that side of the well. s a result, the degree of rock fabric around well C nicely explained the unexpected hydraulic fracture behaviors that the research team had set out to understand. Figure 4 illustrates microseismic events that are colored according to the degree of rock fabric and sized by magnitude, overlaid on contour maps of seismic moment (left), and b-values (right). Stage Two t this point in the research study, it appeared the new 3-D seismic vector attribute was indeed capable of accurately identifying the presence and distribution
of subtle pre-existing discontinuities in the Montney Shale reservoir. The integrated investigation s second stage, therefore, was to consider potential relationships between the degree of rock fabric and other 3-D seismic attributes meant to detect additional influences on hydraulic fracture behavior in unconventional plays. For example, variations in horizontal stress ratios within a particular interval have been correlated with different fracture growth patterns. reas of high stress anisotropy usually tend to create more planar fractures, while areas of low stress anisotropy create more complex, branching fracture networks. Complex fractures usually generate greater reservoir contact, which is essential to enhancing production. For this reason, most operators prefer to drill wells in and stimulate areas of lower horizontal stress. Mapping Poison s Ratio (PR) from 3-D seismic data differentiates areas of higher and lower stress. For this study, minimum PR values within the interval of interest were obtained by performing amplitude-versus-offset analysis on the 3-D reflection seismic data. For the same interval, computing the mean seismic attribute value for each seismic trace creates a rock fabric map. Overlaying microseismic events on maps of minimum PR (Figure 5) and mean rock fabric (Figure 5) yielded several valuable observations. For one, as expected, most microseismic events around wells and and a few small magnitude events southwest of well C occur in regions of low PR, or low horizontal stress. However, the majority of microseismic events around well C occur northeast of the lateral in an area of relatively high PR, which runs counter to conventional wisdom. Why do fractures behave differently in this area? gain, rock fabric offers a possible explanation. In particular, the region immediately southwest of well C exhibits a low degree of rock fabric, while the region northeast of the well exhibits pervasive rock fabric. pparently, at least in this area, the unusual hydraulic fracture behavior either is driven more by rock fabric than by the in-situ stress regime or, more likely, it involves a complex interaction between these two, and perhaps other, important reservoir properties. Indeed, the interaction between stress and rock fabric may explain both well C s relatively high production and its relatively steep initial pressure decline. In any case, PR by itself would not have explained the number and distribution of FIGURE 6 Potential Out-of-Zone Stimulation MRTIN HEGE is a senior geophysicist specialist for Schlumberger based in Calgary. He specializes in microseismic monitoring and integrating microseismic data with other geophysical methods, such as 3-D reflection seismic. Haege also worked as a consultant for the United Nations and participated in several field exercises that used microseismic techniques to detect aftershocks associated with potential nuclear explosions. Haege earned a Ph.D. in engineering from the University of Stuttgart, Germany, and a master s in geology from the University of Karlsruhe, Germany. SHWN MXWELL is a microseismic adviser and chief geophysicist for Schlumberger based in Calgary. Through previous positions with various service companies, he has helped pioneer commercial microseismic hydraulic fracture imaging services to the oil and gas industry. Maxwell earned a Ph.D. in microseismology from Queen s University in Canada. He serves on various microseismic focused committees and workshops around the globe, and is passive seismic associate editor for Geophysics. high magnitude microseismic events in this particular area. This illustration of the automated extraction of patches (signified by the colored planes) indicates areas with a high degree of rock fabric. Using this information, engineers can adjust the stimulation treatment in real time to prevent potential out-of-zone fracturing. LRS SONNELND is a research director at Schlumberger. He has managed several of the company s research and development programs from research centers in Ridgefield, Ct.; Cambridge, UK; and Stavanger, Norway. Sonneland received the ssociation of Chartered Engineers Technical ward, the Geophysical ward and the Schlumberger est R&D Project ward in recognition for his major role in developing 3-D and 4-D seismic technologies, Schlumberger s interpretation software systems, and geophysical reservoir characterization and monitoring methodology. MRK NORTON is the geophysics team lead at Progress Energy. Since leaving ExxonMobil Corp. in 2005, Norton has worked for Husky Energy and Real Resources. He joined Exxon- Mobil in 2000 as a staff geophysicist in the company s Calgary office and later worked in the deepwater Gulf of Mexico out of the company s Houston office. He earned a.sc. in geophysics from the University of Manitoba, where he specialized in electromagnetics and completed his undergraduate thesis on magnetotellurics.
Implications nd pplications For operators to optimize production in unconventional shales, it is becoming increasingly critical to unravel differences in the behavior of hydraulic fractures. Insitu stresses, rock properties and pre-existing zones of weakness both large-scale faults and small-scale discontinuities are among the primary influences on fracture propagation. This research study suggests that understanding rock fabric may become an essential factor in explaining otherwise puzzling fracture behaviors. Detailed knowledge of spatial rock fabric distribution may assist in well placement and fracture engineering design, contributing to more intelligent completions. For example, it may enable engineers to adjust the stimulation treatment in real time to prevent potential out-of-zone fracturing (Figure 6). For this application, each microseismic event recorded during passive seismic monitoring may be used as a seed point within the 3-D seismic volume for automatically extracting small pre-existing discontinuities or areas of notable rock fabric. The search volume would be guided by uncertainties in the location of microseismic events, and the search radius by seismic moment, a function of the source radius. Of course, the prevalence of rock fabric is only one of many reservoir properties for operators to consider when attempting to predict the best possible well location or completion strategy. In other research studies, high rock fabric has been associated with greater production in some places and lower output in others. Depending on local conditions, operators may decide either to avoid areas of high rock fabric or to target them intentionally. Combinations of various attributes almost certainly will prove more important than depending too heavily on any single one. Editor s Note: The preceding article was adapted from a technical paper the co-authors presented at the 2012 Society of Exploration Geophysicists nnual Meeting, held Nov. 4-9 in Las Vegas (http://dx.doi.org/10.1190/segam2012-0301.1). dditional information on using hydraulic fracturing microseismicity for enhanced reservoir characterization is available in SPE 140449, a technical paper the co-authors presented at the Society of Petroleum Engineers 2011 Hydraulic Fracturing Technology Conference, held Jan. 24-26 in The Woodlands, Tx.