Hydraulic-Fracture-Height Growth: Real Data

Transcription

1 Hydraulic-Fracture-Height Growth: Real Data Kevin Fisher and Norm Warpinski, SPE, Pinnacle A Halliburton Service Summary Much public discourse has taken place regarding hydraulic- fracture growth and whether fractures could potentially grow up to the surface and create communication pathways for frac fluids or produced hydrocarbons to pollute groundwater supplies. Real fracturegrowth data mapped during thousands of fracturing treatments are presented along with the reported aquifer depths in the vicinity of the fractured wells. These data are supplemented with an in-depth discussion of fracture-growth limiting mechanisms augmented by mineback tests and other studies performed to visually examine hydraulic fractures. These height-growth limiting mechanisms, which are supported by the mapping data, provide insight into why hydraulic fractures are longer laterally and more constrained vertically. This information can be used to improve models, optimize fracturing, and provide definitive data for engineers, regulators, and interest groups. Introduction Hydraulic fracturing is a technology that has been in practice since the late 1940s to improve production from reservoirs (Montgomery and Smith 2010). This technology has been refined through the decades, and fracturing design, execution, quality control, and evaluation have been the source of hundreds of articles in previous literature (Howard and Fast 1970; Gidley et al. 1989; Economides and Nolte 2000). Hydraulic fractures are manufactured flow paths by which hydrocarbons are efficiently extracted from low-permeability rocks. The fracture is constructed by a planned injection of highpressure fluid that penetrates a sufficient volume of reservoir rock to result in economic production. Because fracturing materials (fluids, proppant, and chemicals) and pumping hydraulic horsepower are expensive, the industry has been motivated since the inception of fracturing to understand and control fracture growth. The leastdesirable occurrence is generally excessive height growth, and many studies and journal articles have addressed this topic. The science and engineering of hydraulic fracturing is well understood, and it is clear that there are numerous mechanisms that contribute to fracture containment. These mechanisms deal particularly with the layered Earth structure and are discussed later in some detail. In addition, there is a volumetric argument. For hydraulic fractures to propagate, they must be opened by internal pressure. As a fracture grows larger, the width of the fracture increases in proportion to the fracture height. Because the pressure must remain high to continue feeding the fracture at its deep starting point, the high pressures and a hypothetically large height would require enormous volumes of fluid. It is clear that the fracture can only grow so far before it reaches a physical limit. Furthermore, fluid leakoff into the rocks adjacent to the fracture can waste or thieve much or most of the fracturing fluid for even a moderate-sized fracture. Fracture lengths can sometimes exceed 1,000 ft when contained within a relatively homogeneous layer, but fracture heights, because of the layered geological environment and other physical parameters to be discussed here, are typically much smaller, usually measured in tens or hundreds of feet. Copyright 2012 Society of Petroleum Engineers This paper (SPE ) was accepted for presentation at the SPE Annual Technical Conference and Exhibition, Denver, 30 October 2 November 2011, and revised for publication. Original manuscript received for review 3 November Paper peer approved 8 January Faults have been suggested as mechanisms for enhancing fracture growth, but this ignores the basic understanding of faults in hydrocarbon reservoirs. If there is an open path to the near surface through an existing fault, throughout geologic time, all of the hydrocarbons would have escaped and there would be no reason for exploiting the resource. If the fault is impermeable, then it also must be closed. In that case, the conditions required for extending the fault are essentially identical to those for extending a fracture in a competent rock. The reason is that the strength of the rock is a negligible factor in fracture propagation; almost all of the work is performed in the opening of the fracture against the rock and the in-situ stresses. It also appears that there is some misunderstanding of the permeability conditions that occur in these sedimentary basins. The layered structure of typical sedimentary rocks results in hundreds or thousands of layers that have low vertical permeability. These rocks collectively form the caprock that prevents the hydrocarbons from escaping. This caprock works exactly the same with respect to the fracturing fluids. Once fracturing fluids are emplaced in the reservoir or rocks adjacent to the reservoir, they are there to stay unless they get produced back to the wellbore. If the permeability was sufficiently high that the fracturing fluids could migrate to the surface, then the hydrocarbons would have already escaped and there would be no reservoir to exploit. Fractures grow perpendicular to the direction of least principal stress, or in the direction of maximum stress (in order to open against the smallest stress). At depths deeper than approximately 2,000 ft, the vertical stress or overburden ( ov ) is generally the largest single stress, so the principal fracture orientation is expected to be vertical on deeper wells (Fig. 1). Of the two horizontal-stress orientations, the direction of maximum stress ( hmax ) will dictate the direction the frac grows laterally (length) and the frac width will open against the smallest stress orientation ( hmin ). This paper presents data from microseismic and tiltmeter monitoring technologies that provide definitive evidence of the amount of vertical growth exhibited by industrial hydraulic fractures. The monitoring technologies are briefly discussed, followed by the monitoring results that clearly show that fractures remain in relatively close proximity to the reservoirs in which they are created. Finally, the paper presents the mechanistic arguments that explain why containment of fractures is to be expected in these sedimentary environments. Fracture-Mapping Technologies The two primary methods for full-field monitoring of hydraulic fractures are microseismic and microdeformation. These two technologies monitor different aspects of the fracturing process, so they are complementary and can be used to corroborate information about height growth. Microseismic. Microseismic monitoring is based on detecting and locating the small reservoir movements that take place as a result of the fracturing process (Albright and Pearson 1982; Warpinski 2009). These movements are caused by changes in stress (fracture opening) and fluid pressure (leakoff), and they occur along natural fractures, bedding planes, and other weakness zones in the rocks with which the fracture makes contact. Therefore, it is an excellent technology for monitoring fracture growth by tracking the distribution of the microseismic events. Microseismic monitoring is usually performed by placing long arrays of sensitive receivers in offset wells at a depth that is close to 8 February 2012 SPE Production & Operations

2 the zone to be fractured. The microseisms are characterized by the emission of both compressional (P) and shear (S) waves, and these are detected by the receivers in the monitoring well. Robust gridsearch and migration methods are used to determine where the events originated on the basis of wave arrival times and polarizations. It is important to have an accurate velocity model; therefore, significant effort is placed on characterizing and calibrating the model. The distance at which microseisms can be observed depends primarily on the size of the microseism and on noise levels. This distance can almost always be characterized from the observed data, so the limits of the monitoring capabilities are known. In shale reservoirs, microseisms can often be detected at distances up to 5,000 ft, whereas many other reservoirs have much shorter observation distances. In shales (the reservoirs possibly of most concern because of the size of the fracture treatments), the large detection distance allows vertical height to be monitored even if there is significant height growth. If fractures are at relatively shallow depths, then it can also be possible to monitor microseisms using surface arrays. This is typically performed in geothermal reservoirs where the microseisms are much larger and can be observed at greater distances. σ hmin σ ov Fig. 1 Stress orientation. σ hmax Microdeformation. Microdeformation fracture monitoring is the measurement of small displacements either at the surface of the Earth or in boreholes using sensitive tiltmeters. For surface monitoring of deep fractures, these tiltmeters (Wright et al. 1998) are used to measure minute deformation induced by the fracture, but can also be used in offset wellbores to monitor fracture dimensions particularly height. A tiltmeter works on the same principle as a carpenter s level, but the sensitivities are orders of magnitude greater (on the order of nanoradians). When a fracture is created, it deforms the rock surrounding it and this deformation radiates outward, reaching the surface or an observation well. An array of tiltmeters on the surface can be used to measure the deformation pattern and determine some details of the fracture orientation. Dipping and horizontal fractures induce distinct patterns, which are easy to separate from vertical fractures, and it is straightforward to resolve the azimuth and dip of the fractures with a surface array. The amplitude of the tilt signal and the width of the deformed zone can be used to determine the depth to the fracture s center and the fracture volume. If significant height growth were to occur, it would be observable in the surface tiltmeter data. In addition, the mode of fracturing can be closely monitored, such as multiplanar fractures or the case where a vertical fracture turns horizontal, or vice versa. Downhole tiltmeters, when installed in a wellbore near the treatment well with an array sufficiently long to span the fractured interval s thickness, can directly measure the height of a hydraulic fracture. Fracture-Mapping Data Figs. 2 through 5 present plots of data collected on thousands of hydraulic-fracturing treatments in some of the most active shale plays in North America: the Barnett shale of Texas, the Woodford shale of Oklahoma, the Marcellus shale in the northeastern US, and the Eagle Ford shale in south Texas, respectively. More fracture treatments have been mapped in the Barnett than in any other reservoir. The methodology for the following graphs is consistent. Each of the graphs illustrates the fracture top and bottom for all mapped fracture treatments performed in each reservoir from early 0 Barnett Shale Mapped Frac Treatments/TVDs Deepest Water Well Level Frac Top Perf Top Perf Mid Perf Btm Frac Btm 2000 Depth (ft) Archer Clay Denton Harmon Jack Palo Pinto Somervell Wise Bosque Cooke Eastland Hill Johnson Parker Stephens Brown Culberson Erath Hood Montague Reeves Tarrant Frac Stages (Sorted on Perf Midpoints) Fig. 2 Barnett shale measured fracture heights sorted by depth and compared to aquifers. February 2012 SPE Production & Operations 9

3 0 Deepest Water Well Level Frac Top Perf Top Perf Mid Perf Btm Frac Btm 2000 Depth (ft) Andrews Blaine Carter Coal Garfield Johnston Reeves Atoka Canadian Cleveland Culberson Hughes Pittsburg Winkler Frac Stages (Sorted on Perf Midpoints) Fig. 3 Woodford shale measured fracture heights sorted by depth and compared to aquifers through the end of All depths are in true vertical depth. Perforation depths are illustrated by the red band for each stage, with the mapped fracture top and bottom illustrated by colored curves corresponding to the counties where they took place. The deepest reported drinking-water levels in each of the counties where these fractures have been mapped (USGS 2011) are illustrated by the dark blue bars at the top of the individual charts. Note that the depth scale in the vertical axis differs from reservoir to reservoir because of large differences in the depths of the pay zones. As presented, the largest directly measured upward growth of all of these mapped fractures still places the fracture tops several thousand feet below the deepest known water-well level in each of the reservoirs presented. Fig. 2 illustrates that faults do not provide a mechanism whereby hydraulic fractures can propagate to the surface. Most of the larger spikes (both downward and upward) are a result of hydraulic fractures intercepting faults. This is because fault-related microseisms can usually be resolved from fracturing-induced microseisms because of their different characteristics (Wolhart et al. 2006; Warpinski 2009). Hydrau- 0 Deepest Water Well Level Frac Top Perf Top Perf Mid Perf Btm Frac Btm 2000 Depth (ft) 4000 Armstrong Belmont Bradford Butler Cameron Centre Clearfield Clinton Doddridge Elk Forest Greene Harrison Lycoming Marshall Mc Kean Nicholas Potter Putnam Schuyler Susquehanna Taylor Tioga Upshur Washington Westmoreland Wetzel Frac Stages (Sorted on Perf Midpoints) Fig. 4 Marcellus shale measured fracture heights sorted by depth and compared to aquifers. 10 February 2012 SPE Production & Operations

4 0 Deepest Water Well Level Frac Top Perf Top Perf Mid Perf Btm Frac Btm 2000 Atascosa Burleson 4000 De Witt Fayette Dimmit Frio Gonzales Karnes Depth (ft) La Salle Maverick Webb Live Oak Mc Mullen Frac Stages (Sorted on Perf Midpoints) Fig. 5 Eagle Ford shale measured fracture heights sorted by depth and compared to aquifers. lic-fracture vertical growth in faults is restricted because of limited volumetrics, and the microseismic data clearly present that even the fault intersections do not result in unbounded fracture-height growth. The Woodford shale data, illustrated in Fig. 3, are interesting because of the exceedingly complex geology. As presented, these Woodford completions span a large vertical interval from wells as deep as 14,000 ft to shallow wells at approximately 4,500 ft. The Woodford s geologic structure can include substantial faulting, highly dipping bedding planes, overturned beds where a vertical wellbore could intersect the same series twice, and all manner of geologic complexity. The fracture-height results would be expected to be influenced by this complexity, but the general interpretation is the same as for the Barnett: hydraulic fracture heights are relatively well contained, there is some fault interaction, and the fractures remain separated from the local aquifers by large distances. The Marcellus data in Fig. 4 present a similarly large distance between the top of the tallest fracture and the location of the deepest drinking-water levels. Because it is a newer play (fewer mapped fracture stages at this point) and encompasses several states, the data set is not as comprehensive as that from the Barnett, but it is no less compelling in providing evidence of good physical separation between hydraulic-fracture tops and aquifers. Hundreds of fracture stages are presented in Fig. 4 and are color coded by state. The fractures grow upward much taller than was seen in the Barnett (some fractures grew nearly 1,500 ft), but the shallowest fracture tops are still around 4,800 ft, almost a mile below the surface and thousands of feet below the aquifers in those counties. Only the two shallow tests at the far right in Fig. 4 come within 2,000 ft of the water-well levels, and this occurs only because the wells are very shallow. The Eagle Ford shale data presented in Fig. 5, much like those of the Woodford shale, exhibit very little out-of-zone height growth. Very little growth is observed into the overlying Austin chalk or the underlying Buda limestone. Another important feature of the fractures mapped in all of these shale reservoirs is that the tallest fractures (those with most significant fracture-height growth) occur in the deepest wells in a given reservoir and, in general, the shallowest wells in a given reservoir have the least measured fracture height. Tiltmeter data in Fig. 6 illustrate the vertical and horizontal components from more than 10,000 fractures mapped throughout the past decade with surface tiltmeters in numerous types of reservoirs at all depths across North America. Each point on the map is a separate fracture treatment, and the hydraulic-fracture volume from that particular fracture job is correlated to the horizontal- vs. vertical-fracture volume-percentage scale at the top. The scale is the percentage of fracture fluid in a single treatment distributed in the given fracture dip so that a 0% fracture component would be a fracture that is vertical and 100% would be a fracture that was 100% horizontal. The larger the horizontal component, the less fracture-height growth one would expect (i.e., a 100% horizontal fracture would have a small height equal only to the fracture s width). As illustrated from the blue curve, which is the average of all fracture dips, fractures are largely vertical until the wells get shallower than approximately 4,000 ft, at which point the fracture complexity (ratio of horizontal-to-vertical fracture-volume distribution) begins to increase steadily. Above 2,000 ft, fracture components are largely horizontal, which leads one to expect minimal fracture-height growth in these shallower reservoirs. Although a totally separate methodology (deformation) from microseismic mapping, tiltmeters confirm one of the mechanisms responsible for limited fracture-height growth in shallower wells. As well depth decreases, the overburden (weight of rock above the wellbore) lessens. At some point in shallow wells, the overburden stress will decrease to a point where it is less than the maximum horizontal stress and, at this point, one would expect the fracture growth to be horizontal and not vertical. As wells get shallower, and the overburden stress lessens, mapped fractures are typically observed exhibiting increasingly larger horizontal components. All of the fractures do not necessarily turn horizontal; they might have significant vertical and horizontal components with more of a T-shaped geometry, but the horizontal components can become significant and could thieve away enough fluid to cause a blunting effect, limiting upward fracture-height growth. Even in areas with the largest measured vertical fracture growth (such as the Marcellus), the tops of the hydraulic fractures are still thousands of feet below the deepest aquifers suitable for drinking water. As can be observed from the data in these reservoirs, the huge distance separating the fractures from the nearest aquifers at their closest point of approach demonstrates that hydraulic fractures are not growing into groundwater aquifers and that fracture treatments themselves are unlikely to contaminate them. The results from this extensive fracture-mapping database showed that hydraulic fractures are often better confined vertically (and are also longer and probably narrower) than conventional February 2012 SPE Production & Operations 11

5 Depth, ft Percent Horizontal Component Fig. 6 Microdeformation (tiltmeter) measurements of horizontal vs. vertical fracture components with depth. wisdom or models predict. Examination of data from other studies might help explain why fracture-height confinement is generally better than that predicted by fracture models. Fracture-Containment Mechanisms On the basis of mineback work performed in the 1970s and 1980s at the Nevada test site (Warpinski et al. 1981a, 1981b, 1982a, and 1982b), it is clear that hydraulic fractures are much more complex than envisioned by conventional models of the process. Figs. 7 through 9 illustrate three examples of how complexity is induced by geologic factors and the wellbore itself. A multistranded vertical fracture is presented in Fig. 7 in which the number of fracture strands increases where layering becomes more frequent, as represented by the horizontal banding near the bottom of the photograph, but also where a horizontal component has propagated to the left. This propped fracture was created with a crosslinked fluid carrying three stages of colored sand (red, black, blue). The apparent curvature of the fracture is actually the curvature of the mineback face; the main fracture is vertical at this location. Fig. 8 presents a fracture initiating from a wellbore (the white circular area near the bottom; the wellbore was filled with cement) and having its path offset at a series of natural fractures. The viewpoint in this photograph is looking up at the fracture, which starts out as a single fracture, but becomes two fractures after intersecting and filling one of the natural fractures that is oblique to the hydraulic-fracture trajectory. Fig. 7 Example mineback fracture showing multiple fracture strands and horizontal components. Fig. 8 Example mineback fracture near borehole showing offsets at natural fractures and two parallel fractures. 12 February 2012 SPE Production & Operations

6 Simple Fracture Complex Fracture Complex Fracture With Fissure Opening Complex Fracture Network Fig. 9 Example mineback fractures initiating from perforations in a horizontal wellbore; multiple fractures have initiated from multiple perforations below the wellbore. Fig. 9 presents a dyed-water fracture that has been initiated from a horizontal, cased, and cemented wellbore through six perforations on both the top and bottom. Those fractures that initiated from the bottom perforations are visible, and there are at least five separate strands that can be identified in the photograph. It appears that each perforation was the initiation point for a separate fracture and these all propagated independently for some distance (for scale, the wellbore diameter is approximately 5 in.). The mineback did not go farther down, so it is not known if the fractures ever coalesced into fewer strands. It is clear from the mineback examples in Figs. 7 through 9 that fractures are much more complex than envisioned by early modelers and practitioners. Fig. 10 is a commonly used schematic showing a hierarchy of complexity (Fisher et al. 2002; Warpinski et al. 2009). Instead of the simple planar fracture presented in the upper left, fractures in common geologic environments show varying degrees of complexity, from the simple complex fracture that is relatively planar in the upper right, to the complex fracture network in the lower right. As a result of this complexity, fractures tend to grow shorter than they would if they were simple planar features. The multiple strands provide additional walls that increase friction and thus raise the fluid pressure, causing wider cracks, and the additional walls also provide a large amount of added surface area for leakoff of the injected fluid. As a result, complex fractures are shorter and wider than simple fractures and both their height growth and length growth are reduced. Fig. 11 Geometry for fracture-height calculations. Fig. 10 Schematics of levels of complexity observed in hydraulic fractures. Given this complexity in a relatively simple fracturing environment, it becomes clear that the height-growth mechanisms in a complex sedimentary basin that is perturbed by structure and natural fractures will likely be dominated by the layering and heterogeneities of both properties and stress. A brief discussion of each of the important mechanisms follows. In-Situ Stress. The in-situ stress contrasts clearly have the most significant effect on fracture-height growth. The importance of stress was recognized early on (Perkins and Kern 1961) and has been extensively studied in modeling (Simonson et al. 1978; Voegele et al. 1983; Palmer and Luiskutty 1985), mineback tests (Warpinski et al. 1982a, b), and numerous laboratory experiments. Fracture-height growth can be easily restricted if the layers above and below have higher stress than the reservoir rock, and this is a common occurrence in sedimentary basins. An equilibrium (static) analysis of the linear elastic fracturemechanics behavior of a fracture surrounded by rocks with higher stress, as illustrated in Fig. 11, was given by Simonson et al. (1978) for a symmetric case (stresses above and below are equal). They obtained the following equation: 2 1 h KIc 2 P = [ 2 1] sin, (1) H H / 2 where P is the net pressure in the fracture, 1 is the stress in the pay zone, 2 is the stress in the bounding layers, h is the thickness of the pay zone, H is the total fracture height, and K Ic is the fracture toughness of the bounding layers. In this equation, the first term on the right side is a result of the stress contrasts, while the second term is a result of fracture toughness. For standard laboratory values of fracture toughness, the second term is generally small. In general, this equation is conservative because there are other dynamic factors that affect the amount of height growth that will occur. For example, if P > 2, the equation would predict unlimited growth, but the flow resistance through the narrower width of the high-stress region would restrict growth vertically compared to laterally. Similar equations can be developed for nonsymmetric stress contrasts, but more complete dynamic analyses are usually performed in fracture models. This mechanism is only effective if there are sufficiently high stresses in the sedimentary layers. The most complete record of stress in a basin is probably that from the multiwell and multisite experiments in the Piceance basin of Colorado (Warpinski et al. 1985; Warpinski and Teufel 1989; Branagan et al. 1997). The stress data are illustrated in Fig. 12 and represent the stresses measured by smallvolume hydraulic fractures in both reservoir (blue symbols) and nonreservoir (red symbols) rocks. As would be expected, the general trend is one of increasing stress with depth, but there are large variations across layers that act to trap fractures in low-stress zones because it requires less energy to grow against a low stress. Because the layered February 2012 SPE Production & Operations 13

7 Fig. 13 Mineback photograph of dyed-water fracture propagating across interface. Fig. 12 Stress data from the Piceance basin. depositional environment in sedimentary basins has such variability in stress, it is highly unlikely that fractures could propagate very far vertically. Also illustrated for reference is the lithostatic stress in these tight rocks. Many of the nonreservoir lithologies that are clay rich or organics rich have stresses that are near or at the lithostatic stress. Material-Property Contrasts. There are three issues of concern with material-property contrasts between layers. The first one deals with the effect of a fracture approaching an interface with modulus contrasts, as proposed by Simonson et al. (1978). The second one deals with the effect of modulus on the width of the fracture and the increased or diminished flow resistance caused by a width change. The third one is caused by differences in the fracture toughness in the various layers. While Simonson et al. (1978) showed that a material-property interface in an ideal situation could blunt fracture growth, years of fracturing experience (Nolte and Smith 1981), fracture-diagnostic monitoring (Warpinski et al. 1998; Wright et al. 1999), mineback testing (Warpinski et al. 1982a, b), and other research (Smith et al. 1982; Teufel and Clark 1984; Palmer and Sparks 1991) have demonstrated that this is not the case. Fig. 13 illustrates an example of a dyed-water fracture that has propagated through an interface from a low-modulus material into a high-modulus material (Warpinski et al. 1982a, b). A more complete discussion of the role of the interface has been given by Cleary (1978), where the complexities of the interface, the micromechanics of the fracturing process, the potential for blunting and twisting (no longer only Mode I fracture growth), and various other factors make any systematic conclusions from the interface mechanics quite problematic. What is clear is that crossing interfaces requires additional energy and can hinder vertical growth. Modulus contrasts clearly have an effect on the width of the fracture and can be expected to enhance or restrict fluid flow appropriately. Cleary (1980) provided a time-constant analysis of the effect of modulus, while Van Eekelen (1982) developed a relationship based on relative height changes in the layers. As discussed by Van Eekelen (1982) and Smith et al. (2001), these effects are generally small and cannot be expected to provide significant containment of fractures. Gu and Siebrits (2008) also showed that low-modulus layers surrounding a higher-modulus pay zone can be restrictive because of a lowered stress-intensity factor, but this also depends on the relative fracture toughness of the different materials. Fracture toughness can have a significant impact on fracture growth if the stress contrasts are small, as can be seen by setting 2 = 1 in Eq. 1. A large value of K Ic will either induce a high pressure or restrict the height or both. For a homogeneous formation, the stress-intensity factor at the top of the fracture can be computed (Rice 1968) if the net stress distribution is known by K I 1 H / 2 = p( y) H / 2 H / 2 H / 2 + y H / 2 y dy, (2) where p(y) is the net stress distribution vertically. If the stress-intensity factor exceeds the fracture toughness of the material, the fracture will propagate. Obviously, the situation becomes more complex (and not analytic) for layered materials with different elastic properties, but Eq. 2 provides a rough estimate of the fracture stability. Laboratory experiments have generally proven that fracture toughness varies over only a limited range (Hsiao and El Rabaa 1987), which suggests that fracture-toughness effects will be rather limited. However, the scale dependence of fracture toughness (or potentially other types of tip effects) is not well understood for large-scale fractures, so there may be potential for fracture containment because of this mechanism (Shlyapobersky et al. 1998). Weak Interfaces. It is well known that weak interfaces can blunt fracture growth, and such a mechanism is often cited for the use of the Khristianovich, Geertsma, and De Klerk models (Nierode 1985). Examples of blunting have been noted in mineback experiments (Warpinski et al 1982a, b; Warpinski and Teufel 1987; Jeffrey et al. 1992; Zhang et al. 2007) and laboratory experiments (Anderson 1981; Teufel and Clark 1984). Fig. 14 illustrates an example of a mineback fracture (Warpinski et al. 1981a) terminating at a weak interface. The fracture was created with a crosslinked gel having three stages of colored proppant (sand) in order of black, red, and blue. Dense sand is observed right up to the interface, which suggests that it was wide, as would be expected with a blunted fracture. The fracture was approximately 30 ft tall and February 2012 SPE Production & Operations

8 ~ Fig. 14 Mineback photograph of fracture terminating at a weak interface. ft long, and Fig. 14 shows only a small segment that was left after mining to show the interfacial effects. While it is generally expected that weak interfaces will be most important in stopping fracture-height growth at shallow depths where friction caused by the overburden stress is at a minimum, other factors such as overpressuring or embedded particulates (equivalent to a fault gouge) can clearly minimize frictional effects even at great depths. Weak interfaces have the potential of totally stopping vertical fracture growth, initiating interface fractures, or causing offsets in the fracture. Fig. 15 presents an example of a fracture that is crossing unhealed natural fractures (Warpinski et al. 1981b), which is equivalent to the case of a weak interface with some permeability along the interface. This example illustrates offsets at the fractures at a location close to the wellbore. Cement was used as the fracturing fluid for this test to preserve the width of the fracture. Such offsets would clearly restrict fracture growth because of the narrower width of the fracture in the offset and the possibility of proppant bridging. In addition to restricted-growth effects, weak interfaces above and below the reservoir can decouple the fracture walls (Barree and Winterfeld 1998; Gu et al. 2008), which results in poor coupling of the fracture pressure in the reservoir to the fracture outside of the weak interfaces. This reduced coupling creates narrower fractures in the layers across the interface and much wider fractures within the reservoir rock. Layered Interfaces. All of the mechanisms previously discussed can be bundled together to describe fracturing across a succession of interfaces. The possibility that such layered media could contain hydraulic fractures has been derived from fracture diagnostic information (Warpinski et al. 1998; Wright et al. 1999; Griffin et al. 1999). It is easy to conceive of multiple mechanisms serving to blunt, kink, offset, bifurcate, and restrict growth in various layers, much as a composite material hinders fracture growth across it. Various methods are now being used to model such behavior (Wright et al. 1999; Miskimmins and Barree 2003; Weijers et al. 2005). Several of the mechanisms are presented in Fig. 16, which is a mineback photograph of a fracture propagating upward across several interfaces. The left side is the unaltered photograph, while the right side has the fracture accentuated with a line drawn over it. Kinking, offsetting, and bending occurred as the fracture made its way through the layers. In other cases, additional fractures are initiated or some fractures are terminated. February 2012 SPE Production & Operations Fig. 15 Mineback photograph of fracture terminating at a weak interface. Fig. 17 presents a schematic of several types of behavior that have been observed in minebacks or laboratory tests. The result of these behaviors could be any combination of complexity, restriction, or termination of the fracture as it propagates across the layered medium, often collectively termed the composite layer effect. Restrictions should be common if kinking or offsets occur because the width in the kink or offset will necessarily be less than in the vertical part of the fracture from both geometric and stress considerations. These kinks, offsets, and restrictions lead to less fracture-height growth than one would expect from a single, simple fracture. Fluid-Pressure Gradient. Simonson et al. (1978) described the relative effects of fluid-pressure gradient compared with the rockstress gradient. This is one mechanism that is conducive to height growth, but for normal fracture heights of a few hundred feet, it is a small effect. However, it is possible to evaluate what would happen if there were tall fractures in a medium without all of the stress variations that typically are found in sedimentary basins. Fig. 18 illustrates that the lower fluid-pressure gradient would cause a large overpressure at the top of a tall fracture. This case has a linearly varying stress profile (no stress contrasts in layers), a normal fluid-pressure gradient, and a tall fracture that is growing upward so that the entry point is at the bottom. England and Green (1963) provide a solution for a 2D crack that can be used to determine the fracture width for a similar distribution. The net pressure distribution is derived as the difference between the fluid-pressure gradient and stress-pressure gradient, assuming that the net pressure at the bottom is essentially zero. The fracture height is H, and the vertical coordinate is y. The average pressure in the fracture (at the centerline) is p and the gradient is given by k, so that the net pressure anywhere in the fracture is given by p + ky. The equations for fracture width and area, respectively, in this case are 15

9 ~ 2 ft Fig. 16 Mineback photograph (and line drawing) of fracture kinking, offsetting, and turning as interface is crossed. and ( ) 41 wy ( )= E A = ( ) 2 2 ky H p y ph, (3) 2E where v is Poisson s ratio and E is Young s modulus. The minimum p needed to keep the fracture open is kh/4. The profile on the right in Fig. 18 is calculated for a total fracture height of 2,000 ft in a material with E = psi, v = 0.2, k = 0.26 psi/ft (the difference between a 0.7-psi/ft stress gradient and a 0.44-psi/ft fluid gradient), and p = 130 psi. The width in the upper portion of the fracture is approximately 0.15 ft. The important point here is that the fracture volume will become enormous as the fracture grows upward, making it impossible for typical fracture-fluid volumes to grow to these extents. For this case, if a typical total fracture length of 1,000 ft is assumed, the fracture volume would be 35,000 bbl, assuming zero leakoff. If the fracture height grows to 4,000 ft, the volume requirement is 280,000 bbl. However, it is obvious that leakoff would also be enormous because of such large fracture areas. In short, the stress conditions Fig. 17 Schematic of pathologies for fracture behavior in a layered sequence. 2 that would encourage fracture growth upward would also result in large fracture widths that require enormous volumes of fluid to continue propagating. These volumes are an order of magnitude more than what is typically pumped in a fracture treatment, so the volumetrics will not support unlimited height growth. Faults. Faults offer a potential conductive surface that can be through-going in the layered media, minimizing many of the restrictive mechanisms discussed previously, with the exception of remaining sensitive to stress variations. While faults can offer somewhat better conductive paths, it is not likely that they are conductive over sizeable fractions of the depth because any oil or gas in the reservoir would have escaped through such conduits and there would not be any hydrocarbon exploitation success in that area. Microseismicity has proven that faults can be activated by fluid injection and that much-larger-magnitude events can be generated, presumably because of the large fault surface area available to move. Such faults are commonly and easily seen in the microseismic data, and there often is some limited additional height growth associated with fault reactivation. The additional height growth tends to range from a few hundred to as much as 1,000 ft in isolated cases. Nevertheless, it is clear from abundant microseismic and surface-tiltmeter data that extensive fracture growth does not occur in faults. High-Permeability Layers. High-permeability layers have a significant effect on fracture growth because they can either act as thief zones that accept fluid and reduce the fracture-driving forces (typical if gas saturated) or can induce a large poroelastic backstress that clamps down on the fracture (typical if liquid saturated). This type of containment mechanism is difficult to measure in field tests but can be proven by means of modeling or laboratory tests (de Pater and Dong 2009). Implications The implications of these monitoring data sets and the review of mechanisms are straightforward and clear. Under normal circumstances, where hydraulic fractures are conducted at depth, there is no method by which a fracture is going to propagate through the various rock layers and reach the surface. This fact is observed in all of the mapping data and is expected on the basis of the application of basic rock-mechanics principles deduced from mineback, core, laboratory, and modeling studies. The vertical growth of fractures in sedimentary basins is hindered by the layering of the materials. The varying material properties of 16 February 2012 SPE Production & Operations

10 Fig. 18 Example calculation of fracture size for tall fracture. the layers and the variable interface properties, together with the large stress contrasts that are largely a result of these properties and interface variations, create an environment where vertical fracture growth is hindered and lateral fracture growth is favored. Outside factors, such as faults, are regularly observed in the monitoring data. Although intersection of a hydraulic fracture with these faults can result in some additional height growth, the amount of growth has been found to be limited everywhere it has been observed. It is clear that a fault can provide a path for fractures to propagate through a limited region in the vicinity of the reservoir, but the amount of growth is restricted by the same factors that resist height growth in nonfaulted regions, and also by the volumetrics of the process. As noted previously, any area where there might be an open fault that extends to the surface would also be devoid of hydrocarbons because they would have leaked out over geologic time. Near-surface hydraulic fracturing is a situation that is primarily controlled by the stress conditions in the reservoir. At depths of less than 1,000 to 2,000 ft, the vertical stress is the minimum stress in all sedimentary basins where measurements have been made and hydraulic fractures will primarily be horizontal. While a number of shallow individual layers may have a horizontal in-situ stress that is the minimum stress so that fractures within that layer are vertical, the bulk of a near-surface sedimentary rock mass would have horizontal fractures that do not propagate vertically. Any mixed growth having both horizontal and vertical fractures would significantly limit vertical growth, more so than in deeper layers. This behavior is clearly observed in the tiltmeter data taken from many basins and depths. Conclusions Real data collected using microseismic and microdeformation fracture-mapping technologies on many thousands of hydraulicfracturing jobs indicate that hydraulic-fracture heights are relatively well contained. The directly measured height growth is often less than that predicted by conventional hydraulic-fracture propagation models because of a number of containment mechanisms. Many of those mechanisms can be best explained by careful review of several mineback experiments, where real hydraulic fractures in the subsurface were able to be physically viewed and studied. Some of those mechanisms include complex geologic layering, changing material properties, the presence of higherpermeability layers, the presence of natural fractures, formation of hydraulic-fracture networks, and the effects of high fluid leakoff. The effects of pre-existing faults are noted and observed in the mapping results, and the relatively negligible effect of faulting on hydraulic-fracture-height growth is discussed. Fracture physics, formation mechanical properties, the layered depositional environment, and other factors all conspire to limit hydraulic-fracture-height growth, causing the fracture to remain in the nearby vicinity of the targeted reservoirs. This certainly is a positive feature of hydraulic fracturing and allows many otherwise noncommercial-quality reservoirs to produce hydrocarbons commercially and safely. Public discourse continues around hydraulic fracturing. The authors have shown real fracture-growth data from thousands of treatments in several of the most active shale plays where hydraulic fracturing is a must-have, where without it, there would be little to no production or reserves growth. It might be instructive to step back from the public debate and recognize, from real data collected starting more than a decade ago and from governmentsponsored mineback studies performed as far back as the 1970s, that fracture physics, height growth, and containment mechanisms have already been extensively studied and documented in an effort to make hydraulic fracturing more effective. These same early studies, performed outside of today s highly charged climate of debate, reveal significant and relevant data to promote informed discussions about hydraulic-fracture growth and its environmental impact. Nomenclature A = fracture area, L 2, ft 2 [m 2 ] E = Young s modulus, M/LS 2, psi [MPa] h = thickness of reservoir, L, ft [m] H = fracture height, L, ft [m] K = vertical net pressure gradient in fracture, M/S 2, psi/ft [MPa/m] K I = stress-intensity factor, M/S 2 L, psi in. [kpa m] K I c = fracture toughness, M/S 2 L, psi in. [kpa m] p = average pressure in fracture, M/LS 2, psi [MPa] P = pressure in fracture, M/S 2 L, psi [MPa] February 2012 SPE Production & Operations 17

11 W = fracture width, L, ft [m] Y = vertical distance from center of fracture, L, ft [m] ν = Poisson s ratio 1 = stress in reservoir, M/LS 2, psi [MPa] = stress in layers outside reservoir, M/LS 2, psi [MPa] 2 Acknowledgments The authors thank the management of Halliburton for permission to publish this paper. References 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. SPE J. 22 (4): SPE-9509-PA. org/ /9509-pa. Anderson, G.D Effects of Friction on Hydraulic Fracture Growth Near Unbonded Interfaces in Rocks. SPE J. 21 (1): SPE PA. Barree, R.D. and Winterfeld, P.H Effects of Shear Planes and Interfacial Slippage on Fracture Growth and Treating Pressures. Paper SPE presented at the SPE Annual Technical Conference and Exhibition, New Orleans, September. Branagan, P., Peterson, R., Warpinski, N., and Wright, T Results of Multi-Site Project Experimentation in the B-Sand Interval: Fracture Diagnostics and Hydraulic Fracture Intersection. Topical report GRI-96/0225, DOE/MC/ , Gas Research Institute, Chicago, Illinois (December 1997), techrpts/dec1997.pdf. Cleary, M.P Primary Factors Governing Hydraulic Fractures in Hetrogeneous Stratified Porous Formations. Paper 78-Pet-47 presented at the ASME Energy Technology Conference, Houston, 5 9 November. Cleary, M.P Analysis of Mechanisms and Procedures for Producing Favourable Shapes of Hydraulic Fractures. Paper SPE 9260 presented at the SPE Annual Technical Conference and Exhibition, Dallas, September. de Pater, C.J. and Dong, Y Fracture Containment in Soft Sands by Permeability or Strength Contrasts. Paper SPE presented at the SPE Hydraulic Fracturing Technology Conference, The Woodlands, Texas, USA, January. Economides, M.J. and Nolte, K.G Reservoir Stimulation, third edition. New York: John Wiley & Sons. England, A.H. and Green, A.E Some Two-Dimensional Punch and Crack Problems in Classical Elasticity. Math. Proc. Cambridge Philos. Soc. 59 (2): Fisher, M.K., Wright, C.A., Davidson, B.M., et al Integrating Fracture Mapping Technologies to Optimize Stimulations in the Barnett Shale. Paper SPE presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, 29 September 2 October. Gidley, J.L., Holditch, S.A., Nierode, D.E., and Veatch, R.W. Jr Recent Advances in Hydraulic Fracturing, No. 12. Richardson, Texas: Monograph Series, SPE. Griffin, L.G., Wright, C.A., Davis, E.J., Weijers, L., and Moschovidis, Z.A Tiltmeter Mapping to Monitor Drill Cuttings Disposal. Paper ARMA presented at the 37th Annual US Symposium on Rock Mechanics (USRMS)(Vail Rocks 99), Vail, Colorado, USA, 6 9 June. Gu, H. and Siebrits, E Effect of Formation Modulus Contrast on Hydraulic Fracture Height Containment. SPE Prod & Oper 23 (2): SPE PA. Gu, H., Siebrits, E., and Sabourov, A Hydraulic-Fracture Modeling With Bedding Plane Interfacial Slip. Paper SPE presented at the SPE Eastern Regional/AAPG Eastern Section Joint Meeting, Pittsburgh, Pennsylvania, USA, October MS. Howard, G.C. and Fast, C.R Hydraulic Fracturing, Vol. 2, Richardson, Texas: Monograph Series, SPE. Hsiao, C. and El Rabaa, A.W Fracture Toughness Testing Of Rock Cores. Paper ARMA presented at the The 28th U.S. Symposium on Rock Mechanics (USRMS), Tucsan, Arizona, USA, 29 June 1 July. Jeffrey Jr., R.G., Brynes, R.P., Lynch, P.J., and Ling, D.J An Analysis of Hydraulic Fracture and Mineback Data for a Treatment in the German Creek Coal Seam. Paper SPE presented at the SPE Rocky Mountain Regional Meeting, Casper, Wyoming, USA, May. Miskimins, J.L. and Barree, R.D Modeling of Hydraulic Fracture Height Containment in Laminated Sand and Shale Sequences. Paper SPE presented at the SPE Production and Operations Symposium. Oklahoma City, Oklahoma March. Montgomery, C.T. and Smith, M.B Hydraulic Fracturing: History of an Enduring Technology. J Pet Technol 62 (12): (Special Section Hydraulic Fracturing: The Past, Present, and Future). Nierode, D.E Comparison of Hydraulic Fracture Design Methods to Observed Field Results. J Pet Technol 37 (10): SPE PA. Nolte, K.G. and Smith, M.G Interpretation of Fracturing Pressures. J Pet Technol 33 (9): SPE-8297-PA. org/ /8297-pa. Palmer, I.D. and Luiskutty, C.T A Model of the Hydraulic Fracturing Process for Elongated Vertical Fractures and Comparisons of Results With Other Models. Paper SPE presented at the SPE/DOE Low Permeability Gas Reservoirs Symposium, Denver, May. dx.doi.org/ /13864-ms. Palmer, I.D. and Sparks, D.P Measurement of Induced Fractures by Downhole TV Camera in Black Warrior Basin Coalbeds. J Pet Technol 43 (3): , SPE PA. org/ /20660-pa. Perkins, T.K. and Kern, L.R Widths of Hydraulic Fractures. J Pet Technol 13 (9): SPE-89-PA. Rice, J.R Mathematical Analysis in the Mechanics of Fracture. In Fracture: An Advanced Treatise, ed. H. Liebowitz, Vol II, Chap. 3, New York: Academic Press. Shlyapobersky, J., Issa, M.A., Issa, M.A. et al Scale Effects on Fracture Growth Resistance in Poroelastic Materials. Paper SPE presented at the SPE Annual Technical Conference and Exhibition, New Orleans, September. Simonson, E.R., Abou-Sayed, A.S., and Clifton, R.J Containment of Massive Hydraulic Fractures. SPE J. 18 (1): SPE-6089-PA. Smith, M.B., Bale, A.B., Britt, L.K., Klein, H.H., Siebrits, E., and Dang, X Layered Modulus Effects on Fracture Propagation, Proppant Placement, and Fracture Modeling. Paper SPE presented at the SPE Annual Technical Conference and Exhibition, New Orleans, 30 September 3 October. Smith, M.B., Rosenburg, R.J., and Bowen, J.F Fracture Width Design vs. Measurement. Paper SPE presented at the SPE Annual Technical Conference and Exhibition, New Orleans, September. Teufel, L.W. and Clark, J.A Hydraulic Fracture Propagation in Layered Rock: Experimental Studies of Fracture Containment. SPE J. 24 (1): SPE-9878-PA. USGS National Water Information System: Web Interface. USGS Water Data for the Nation, (accessed June 2011). Van Eekelen, H.A.M Hydraulic Fracture Geometry: Fracture Containment in Layered Formations. SPE J. 22 (3): SPE PA. Voegele, M.D., Abou-Sayed, A.S., and Jones, A.H Optimization of Stimulation Design Through the Use of In-Situ Stress Determination. J Pet Technol 35 (6): SPE PA. org/ /10308-pa. Warpinski, N Microseismic Monitoring: Inside and Out. Distinguished Author Series, J Pet Technol 61 (11): SPE MS. Warpinski, N.R. and Teufel, L.W Influence of Geologic Discontinuities on Hydraulic Fracture Propagation. J Pet Technol 39 (2): SPE PA. Warpinski, N.R. and Teufel, L.W In-Situ Stresses in Low Permeability, Nonmarine Rocks. J Pet Technol 41 (4): SPE PA. Warpinski, N.R., Branagan, P., and Wilmer, R In-Situ Stress Measurements at U.S. DOE s Multiwell Experiment Site, Mesaverde Group, Rifle, Colorado. J Pet Technol 37 (3): SPE PA February 2012 SPE Production & Operations