RANDALL MARRETT, STEPHEN E. LAUBACH, and JON E. OLSON, The University of Texas at Austin, USA
Recent geologic research on natural
fractures challenges assumptions fre- quently made by geophysicists. Open fractures are not necessarily oriented parallel to the maximum horizontal stress, and fractures do not necessar- ily close when the fluid pressure within them is reduced. Even in the most mechanically favorable envi- ronment, precipitated cements can prop fractures open or seal fractures of any orientation. Fracture sets typi- cally show dispersion in strike, and multiple sets of open fractures can coexist. More importantly, fractures comprise populations that commonly range over orders of magnitude in aperture and length and that occur in nonuniform clusters. Instead of iso- lated, regularly spaced, large, equally Figure 1. Rose diagrams of maximum compressive stress (SHmax, upper row, blue) and open frac- compliant fractures, the Earth presents ture strike (lower row, red) for study areas in Texas and Wyoming. (a) SHmax and (b) fracture complex fractal clustering of fractures strike, East Texas, average ENE fracture strike is similar to SHmax trends, but open fractures have having a wide range of sizes and vari- a spread of 130°; (c) SHmax and (d) fracture strike, four West Texas wells. (e) SHmax and (f) frac- able compliance dictated by natural ture strike, western Green River Basin; (g) SHmax and (h) fracture strike southern Powder River Basin, horizontal well image log data (after Laubach et al., 2004). cements in the fractures and the rock mass. Going beyond anisotropy to document these essential fracture attributes in the interwell discarded. Among seismically important attributes, frac- space is a key challenge for geophysicists. ture openness can determine the extent of mechanical cou- pling across fractures, fracture orientation may control the Motivation for seismic characterization of fractures. direction of velocity anisotropy, and fracture sizes and abun- Fractures are notoriously challenging to study in the sub- dance can control the magnitude of seismic signature surface, where they can have dramatic effects on fluid flow. (Marrett, 1997). Open fracture length may be related to the Direct study of subsurface fractures, using logs or core from magnitude of velocity anisotropy, and could also affect dif- boreholes, is hampered by several sampling problems. fraction patterns. Fractures commonly are nearly vertical, so vertical wellbores Here we outline some recent core, outcrop, and model- are unlikely to intersect many fractures. Sampling proba- based findings on natural fracture populations that suggest bility also is poor because the spacing between conductive that subsurface fracture patterns are highly heterogeneous fractures typically is large in comparison with borehole on a range of scales and in a variety of ways. Thus, there diameters. Additionally, heterogeneity of fractures com- are first-order implications for the expected seismic response monly occurs on length scales that are a fraction of well spac- of fractures that are typically not accounted for in current ing, so important lateral changes in fracture attributes can geophysical approaches. remain undiscovered. Heterogeneity can also be manifest as significant variation of fractures from one layer to the next. Fracture geology. Outcrops commonly contain sets of large, Because of the limitations in wellbore-based observations, more-or-less planar, mostly evenly spaced, barren (no min- seismic detection and characterization of fractures poten- eral fill), opening-mode fractures. Known since the early tially valuable tools for subsurface prospecting. days of geology (Pollard and Aydin, 1988; Cosgrove and The presence of fracture signal in seismic reflection data Engelder, 2004), these joints undoubtedly influence common is detected by studying anisotropic behavior of seismic conceptions of fractures. Yet inherently sparse information velocities and amplitudes (Queen and Rizer, 1990; about subsurface fractures suggests they differ from joints Schoenberg and Sayers, 1995). However, anisotropic attrib- in some important ways. Cores and well logs, particularly utes are challenging to measure, and their interpretation may image logs, provide direct samples of subsurface fractures, rely on assumptions about fracture orientations, shapes, but fracture sampling using wells is notoriously incomplete. openness, sizes, and spatial arrangement that are challeng- For example, prior to the advent of horizontal wells, sub- ing to verify independently. Based on the geologic literature, surface fracture size and spatial distributions were mostly geophysicists frequently assume that fractures are oriented conjectural. Consequently, geologists have long utilized out- parallel to the maximum horizontal stress, that fractures crops containing exhumed fractures as a proxy source of close when the fluid pressure within the fractures is reduced, information on subsurface fractures. Outcrops, however, and that there is a single set of parallel, evenly spaced, open are subject to uplift and weathering-induced fractures that fractures. Many of these assumptions need to be revised or are nonrepresentative of the subsurface.
1106 THE LEADING EDGE SEPTEMBER 2007
environment, reactive fracture sur- faces are susceptible to accumulating cement deposits. Core studies show (Figure 2) and modeling studies pre- dict that in otherwise open fractures, isolated deposits of cement are com- mon. Laboratory tests show that par- tial mineral fill can make fracture aperture insensitive to static changes in effective stress. The prevalence of strong, spatially isolated mineral bridges that resist fracture closure is not widely appreciated. Cement precipitation in the host rock during or immediately after frac- ture formation is another mechanism that can increase the resistance of nat- ural fractures to closing. This process, which is probably widespread, essen- Figure 2. Subsurface and outcrop data on open and sealed fractures. (a) Rocks may have many tially freezes fractures open. Lander et sets of open fractures or none (modified from Laubach and Ward, 2006). (b) Even in the most al. (2002) showed that, without seal- mechanically favorable environment, precipitated cements can seal any orientation fracture (mod- ified from Laubach, 2003). ing fractures, as much as 20% whole rock volume of quartz cement can pre- cipitate in a rock’s pore space after fractures form. Thus, fractures do not necessarily close when the fluid pres- sure within them is reduced, even if they lack cement bridges (Figure 2). A simple calculation demonstrates how host rock stiffening can affect fracture aperture compressibility (Olson et al., 2007). Assuming linear elasticity and plane strain, the compliance of frac- ture aperture (opening per unit dri- ving stress) for a fracture of a given total length, L, can be written as
where v is Poisson’s ratio, Δσ is the dri-
ving stress, and E is Young’s modulus. If after initial fracture opening (Figure 3, A to B), diagenetic cementation in the host rock increased Young’s mod- ulus by a factor of 5, for instance, frac- Figure 3. Plot of fracture opening (aperture) versus driving stress. When driving stress is zero ture compliance would be reduced by (A), the fracture is closed. Increasing driving stress by increased pore pressure or reducing crack- that same factor and it would take five normal compression causes the crack to open (A to B). If diagenesis occurs at time B, crack clos- times the stress change to close the ing takes more stress because of host rock stiffening effect (after Olson et al., 2007). fracture (Figure 3, B to C) as it took to open it. We do not believe that a consensus exists in the geologic Of course, fracture aperture can be closed without any community about the general attributes of subsurface frac- kinematic aperture change by the diagenetic process of tures. Nevertheless, geologic observations do test the valid- cement precipitation, which is insensitive to fracture com- ity of assumptions about fractures that geophysicists accept pliance or fracture-stress orientation relationships. Empirical as generally true. For example, Figure 1 shows that open evidence shows that heterogeneous patterns of infilling of fractures are not necessarily oriented parallel to current-day large fractures by cements is a common occurrence. Core maximum horizontal stress. In these examples, stress ori- demonstrates that sealed and open fractures having identi- entation data from reliable measurements are consistent cal strike can be interspersed over vertical distances that with regional stress maps. Yet observations from extensive range from a few meters or less to decimeters and over lat- core collections show that open fractures can have arbitrary eral distances of m to km. Production and core data demon- strike relative to SHmax (maximum horizontal compression). strate that it is the degree of cement fill in fractures rather Production data show that these fractures also govern fluid than fracture orientation that limits fluid flow. Flow occurs flow. only where fractures are not sealed with cement. In the Chemical processes can account for resistance of frac- absence of reliable measurements of both open fracture tures to closure. Fractures at depth in sedimentary basins strike and SHmax, these features should not be presumed to are exposed to hot (>80°C), mineral-laden water. In this be parallel. Even in the most mechanically favorable envi-
SEPTEMBER 2007 THE LEADING EDGE 1107
velocity is decreased in all azimuths and the net effect can be less velocity anisotropy than from either of the frac- ture sets alone. The case of nonorthog- onal fracture sets might be diagnosed if fractures have different normal and shear compliances. That rock may con- tain fracture sets having differing ori- entations has long been appreciated, but recent studies show that rocks may Figure 4. Fracture patterns generated using randomly oriented starter cracks and a subcritical have many sets of fractures that are propagation model for increasingly anisotropic prefracture strain states (from a to c) followed by open concurrently. And even given isotropic biaxial extension. only one set of fractures (formed by one deformation event), the assump- tion that all individuals within that set are parallel may not always be rea- sonable. Some core data sets show substantial dispersion in strike for nominally coeval fractures that cannot be ascribed to core orientation errors. Locally, fracture sets in outcrop also show wide strike dispersion. This vari- able fracture orientation in some cases can be attributed to perturbations of stress fields caused by the presence of large faults (Rawnsley et al., 1992) or to nearly isotropic loading conditions that may promote random or orthog- onal fracture patterns to develop dur- ing a single loading event. Geo- mechanical modeling results (Figure 4) show how differences in strain anisotropy for a given deformation event can significantly affect the ori- entation of natural, opening-mode fracture sets. Beyond orientation and fracture cementation, another crucial fracture attribute is size. Subsurface fractures in a single set commonly have aper- tures and lengths that range over orders of magnitude in size, with small fractures far more abundant than large fractures (Marrett et al., 1999). In such cases, fractures of different size share orientations, kinematics, and timing Figure 5. Abundance of fractures as a function of fracture size. Fractures commonly have aper- relative to diagenetic events, so they tures and lengths that range over orders of magnitude in size and follow power-law frequency are most simply interpreted as differ- distributions. Fractures from the Marble Falls Limestone, Texas, follow a common power-law ent size fractions of a single fracture distribution of apertures, even though one subset of the data was collected at outcrop scale using set with a common genesis. One con- a hand lens whereas other data were collected petrographically in a microscope (Marrett et al., sequence of the broad spectrum of 1999). The power-law regression to data from microfractures is extrapolated for comparison with fracture sizes is that fracture intensity, data collected along a 60-m line of observation at outcrop scale. the abundance of fractures in space, is inherently scale-dependent, and varies ronment, precipitated cements can seal any orientation frac- as a function of minimum observed fracture size. As the ture (Laubach, 2003). Recognition of, and distinction among, threshold for counting fractures is decreased, fracture inten- uncemented fractures, partially cemented fractures con- sity increases rapidly. Another consequence is that average taining cement bridges, and completely cemented fractures fracture size is poorly defined and depends sensitively to is essential in effective reservoir characterization and is a detection threshold. challenge for seismic methods. This might be feasible if min- These aspects of fracture size distributions are prob- eral fill within a fracture (e.g., mineral bridges) affects shear lematic, because models of seismic velocity anisotropy vari- compliance across the fracture differently than normal com- ation with fracture characteristics (e.g., Thomsen, 1993) are pliance (Sayers and Dean, 2001). defined in terms of average fracture size and fracture inten- Given that open fractures exist in the subsurface, their sity. To the extent that fracture size distributions follow sys- orientation is often surmised from velocity anisotropy. tematic patterns, theory can be modified to account for However, multiple sets of fractures are common, and if a realistic fracture parameters (e.g., Marrett, 1997). For exam- second set forms at a high angle with the first set, then ple, fracture apertures commonly follow power-law distri-
1108 THE LEADING EDGE SEPTEMBER 2007
butions (Figure 5), which can be char- acterized by two parameters that replace fracture intensity and average fracture size in formulations of the magnitude of velocity anisotropy. Models of seismic velocity anisotropy typically presume a statis- tically uniform arrangement of frac- tures in space (e.g., Thomsen, 1995), but subsurface fractures commonly are clustered (Figure 6a). Moreover, the largest fractures tend to occur in clusters. For example, Figure 6b shows the autocorrelation function (Davis, 2002) using logarithmically graduated lags for the data shown in Figure 6a. Positive autocorrelation characterizes the fractures for almost all lags less than a few meters, indicating the frac- tures occur in meter-scale clusters. Negative autocorrelation dominates lags of 5–15 m, and a spike of positive autocorrelation occurs for lags of 15–25 m. This pattern of autocorrela- tion suggests that ~15 m wide domains of unusually low fracture intensity lie between fracture clusters, the centers of which are spaced about 20 m apart. Fracture clustering may affect seis- mic response in at least two ways. First, because the largest fractures tend to be clustered, the probability for them to be connected is much higher that it would otherwise be. Mechanical connectivity among fractures should magnify compliance and enhance velocity anisotropy compared with isolated fractures of the same size and abundance. Second, velocity ani- sotropy may be heterogeneous on length scales that are long compared with individual fractures but short Figure 6. Arrangement of fractures in space. Instead of regularly spaced fractures of comparable compared with seismic wavelengths. size, complex clustering of fractures having a wide range of sizes is common. (a) Fractures in the Marble Falls Limestone, Texas, are arranged in clusters, where the largest-aperture fractures tend The extent to which such heterogene- to lie. (b) Autocorrelation function (Davis, 2002) of indicator series (value of 1 within fractures, ity can be teased from the seismic sig- value of 0 between fractures) documents concentration of fractures inside of clusters that are ~5 nal remains to be addressed. m wide and spaced ~20 m apart. This is the same data set as used in Figure 5 to represent out- crop-scale observations. Discussion. The well-known chal- lenges of obtaining meaningful geo-
1110 THE LEADING EDGE SEPTEMBER 2007
logic samples of fractures using boreholes (see, for exam- Structural Geology, 2004). “Diagenesis in porosity evolution of ple, Narr, 1996; Mauldon and Mauldon, 2005) underline the opening-mode fractures, Middle Triassic to Lower Jurassic la need for seismic information on hard-to-measure attributes. Boca Formation, NE Mexico” by Laubach and Ward Unfortunately, most seismic data analysis techniques cur- (Tectonophysics, 2006). “Permeability, porosity, and shear-wave rently practiced are based on equivalent or effective media the- anisotropy from scaling of open-fracture populations” by Marrett ories which assume unrealistic fracture geometries and (RMAG Fractured Reservoirs: Characterization and Modeling distributions. Nonetheless, seismic methods offer the hope of Guidebook, 1997). “Extent of power-law scaling for natural frac- measuring key fracture attributes between boreholes and on tures in rock” by Marrett et al. (Geology, 1999). “Fracture sampling length scales that are most meaningful to fluid-flow simula- on a cylinder: From scanlines to boreholes to tunnels” by Mauldon tion. Appreciating the complexity of fracture systems as illus- and Mauldon (Rock Mechanics and Rock Engineering, 1997). trated by outcrop and subsurface geologic investigations will “Estimating average fracture spacing in subsurface rock” by Narr be essential for creating new seismic analysis tools and (AAPG Bulletin, 1996). Combining Diagenesis and Mechanics in improved processing and analysis of seismic data. Natural Fracture Network Characterization by Olson et al. (Geological Society of London Special Publication 270, 2007). “Progress in Suggested reading. The Initiation, Propagation, and Arrest of Joints understanding jointing over the past century” by Pollard and and other Fractures by Cosgrove and Engelder (Geological Society Aydin (GSA Bulletin, 1988). “An integrated study of seismic of London Special Publication 231, 2004). Statistics and Data anisotropy and the natural fracture system at the Conoco bore- Analysis in Geology by Davis (Wiley, 2002). Predicting and hole test facility, Kay County, Oklahoma” by Queen and Rizer Characterizing Fractures in Dolostone Reservoirs: Using the Link (Journal of Geophysical Research, 1990). “Joint development in per- between Diagenesis and Fracturing” by Gale et al. (Geological Society turbed stress fields near faults” by Rawnsley et al. (Journal of of London Special Publication 235, 2004). “Feasibility of seismic Structural Geology, 1992). “Azimuth-dependent AVO in reservoirs characterization of multiple fracture sets” by Grechka and containing nonorthogonal fracture sets” by Sayers and Dean Tsvankin (GEOPHYSICS, 2002). “Lithologic and structural controls (Geophysical Prospecting, 2001). “Seismic anisotropy of fractured on natural fracture distribution and behavior within the Lisburne rock” by Schoenberg and Sayers (GEOPHYSICS, 1995). “Weak elas- Group, northeastern Brooks Range and North Slope subsurface, tic anisotropy” by Thomsen (GEOPHYSICS, 1993). “Elastic anisotropy Alaska” by Hanks et al. (AAPG Bulletin, 1997). “Interaction due to aligned cracks in porous rock” by Thomsen (Geophysical between quartz cementation and fracturing in sandstone” by Prospecting, 1995). TLE Lander et al. (AAPG Annual Convention, 2002). “Practical approaches to identifying sealed and open fractures” by Laubach Acknowledgments: We gratefully acknowledge comments from S. Fomel and (AAPG Bulletin, 2003). “Are open fractures necessarily aligned support by the U.S. Department of Energy Office of Basic Energy Sciences with maximum horizontal stress?” by Laubach et al. (Earth and and the Fracture Research and Application Consortium. Planetary Science Letters, 2004). “Coevolution of crack-seal texture and fracture porosity in sedimentary rocks: Cathodoluminescence Corresponding author: jolson@mail.utexas.edu observations of regional fractures” by Laubach et al. (Journal of