Marine and Petroleum Geology 37 (2012) 1e6

Contents lists available at SciVerse ScienceDirect

Marine and Petroleum Geology
journal homepage: www.elsevier.com/locate/marpetgeo

Review article

Hydraulic fractures: How far can they go?
Richard J. Davies a, *, Simon A. Mathias a, Jennifer Moss b, Steinar Hustoft c, Leo Newport a

Durham Energy Institute, Department of Earth Sciences, Durham University, Science Labs, Durham DH1 3LE, UK 3DLab, School of Earth and Ocean Sciences, Main Building, Park Place, Cardiff University, Cardiff CF10 3YE, UK c University of Troms?, Department of Geology, Dramsveien 201, N-9037 Troms?, Norway

a r t i c l e i n f o
Article history: Received 14 February 2012 Received in revised form 5 April 2012 Accepted 7 April 2012 Available online 24 April 2012 Keywords: Fracture Pressure Shale Natural Stimulated

a b s t r a c t
The maximum reported height of an upward propagating hydraulic fracture from several thousand fracturing operations in the Marcellus, Barnett, Woodford, Eagle Ford and Niobrara shale (USA) is w588 m. Of the 1170 natural hydraulic fracture pipes imaged with three-dimensional seismic data offshore of West Africa and mid-Norway it is w1106 m. Based on these empirical data, the probability of a stimulated and natural hydraulic fracture extending vertically >350 m is w1% and w33% respectively. Constraining the probability of stimulating unusually tall hydraulic fractures in sedimentary rocks is extremely important as an evidence base for decisions on the safe vertical separation between the depth of stimulation and rock strata not intended for penetration. ? 2012 Published by Elsevier Ltd.

1. Introduction Hydraulic fractures propagate when ?uid pressure exceeds the least principal stress and the tensile strength of the host sediment (Hubbert and Willis, 1957). They continue to propagate until the stress-intensity at the fracture tip is lower than the critical stressintensity of the rock being fractured (e.g. Savalli and Engelder, 2005). These conditions can occur naturally (e.g. Cosgrove, 1995) but they can also be stimulated to recover oil and gas (Simonson et al., 1978), or during injection of water into geothermal boreholes (e.g. Legarth et al., 2005; Julian et al., 2010) and unintentionally as the result of subsurface blowouts (e.g. Tingay et al., 2005). Hydraulic fractures are commonly described in outcrops at centimetre to metre scale (e.g. Cosgrove, 1995 e Fig. 1ab). They can be up to w50 m in height in the Devonian Marcellus shale (e.g. Engelder and Lash, 2008) and sand ?lled fractures (injectites) have been documented to extend hundreds of metres (Hurst et al., 2011). But three-dimensional (3D) seismic data now show that natural hydraulic fractures probably cluster, forming pipe-like features that often extend vertically for even greater distances than this (see L?seth et al., 2001; Zuhlsdorff and Spie?, 2004; Cartwright et al., 2007; Davies and Clarke, 2010). Stimulation of hydraulic fractures as a technique for improved hydrocarbon production from low permeability reservoirs dates
* Corresponding author. E-mail address: richard.davies@durham.ac.uk (R.J. Davies). 0264-8172/$ e see front matter ? 2012 Published by Elsevier Ltd. doi:10.1016/j.marpetgeo.2012.04.001

back to the late 1940s (Montgomery and Smith, 2010). Measurements of the microseismicity they cause (e.g. Maxwell et al., 2002) have shown that they can extend for several hundred metres upwards and downwards from the wellbore (Fisher and Warpinski, 2011). Multiple stages of hydraulic fracture stimulation on multiple wells are routine for the recovery of oil and gas from low permeability sedimentary reservoirs in shale gas provinces in the USA (e.g. King, 2010). Shale gas exploration is starting in many other countries with sediments from Neogene to Cambrian age being potential future targets. Therefore constraining the probability of stimulating unusually tall hydraulic fractures in sedimentary rocks is critically important, as it will help avoid the unintentional penetration of shallower rock strata (Fig. 2) that might be important aquifers or subsurface geological storage sites. Mathematical methods for estimating hydraulic fracturing height are simplistic (Fisher and Warpinski, 2011) and it is generally accepted that we cannot yet accurately predict fracture propagation behaviour in detail, so to date much of what we know of how fractures will behave in situ conditions comes from operational experience (King et al., 2008). Future shale gas targets could be in a variety of different stress regimes and in rocks with varied mechanical properties and ages. Therefore at this stage our approach is to include a wide range of the tallest examples of hydraulic fractures that have different geometries, geological settings and trigger mechanisms. Although hydraulic fractures are 3D, here we compile new and existing data on the extents of only the vertical component of both


R.J. Davies et al. / Marine and Petroleum Geology 37 (2012) 1e6

Figure 1. (a) Examples of natural hydraulic fractures in shale (b) close-up of a natural hydraulic fracture ?lled with shale clasts (both examples from onshore Azerbaijan).

natural and stimulated hydraulic fracture systems hosted in sediment from Neogene to Devonian in age from eight different locations (Fig. 3a). We brie?y report on key statistics, compare them and consider which factors control the extent of upward fracture propagation.

Figure 3. (a) Map of the globe showing location of the eight datasets. Red font e datasets for stimulated hydraulic fractures, blue font e datasets for natural hydraulic fractures (pipes). (b) Seismic line from offshore Mauritania showing a representative vertical pipe imaged on 3D seismic re?ection data and its vertical extent (from Davies and Clarke, 2010). (c) Graph of stimulated hydraulic fractures in the Marcellus, Barnett, Woodford and Eagle Ford shales (after Fisher and Warpinski, 2011) and including unpublished data provided by Halliburton for the Niobrara shale. Inset e extract of the graph showing how the vertical extents of fractures were measured. All depths are in true vertical depth (TVD). Coloured spikes e separate hydraulic fractures propagating upwards and downwards from the fracture initiation depth.

1.1. Hydraulic fracturing There are several types of natural hydraulic fracture: injectites (e.g. Hurst et al., 2011), igneous dikes (e.g. Polteau et al., 2008), veins (e.g. Cosgrove, 1995), coal cleats (e.g. Laubach et al., 1998), and joints (e.g. McConaughy and Engelder, 1999). They have been extensively studied. In the case of joints in the Devonian Marcellus Formation, USA, it is even possible to study how they grow on the basis of plume lines that occur at centimetre to metre scale (Savalli and Engelder, 2005). Marcellus shale fractures are thought to form due to gas diffusion and expansion within shale through multiple propagation events. In contrast the tallest examples of hydraulic fractures tend to cluster, are commonly termed chimneys, pipes or blowout pipes (herein we use the term ‘pipe’) and can extend vertically for hundreds of metres (e.g. Cartwright et al., 2007; Huuse et al., 2010). The origin of pipes is not certain, but they probably form due to critical pressurisation of aquifers and oil and gas accumulations (Zuhlsdorff and Spie?, 2004; Cartwright et al., 2007; Davies and Clarke, 2010). Pipe development may be followed by stoping, ?uiddriven erosion and collapse of surrounding strata (Cartwright et al., 2007). Gases that have come out of solution and expand

Figure 2. Schematic diagram showing stimulated hydraulic fractures within a shale gas reservoir, natural hydraulic fractures initiated at a naturally overpressured reservoir, the vertical extent (VE) of hydraulic fractures reported here and the safe separation between shale gas reservoir and shallower aquifer.

R.J. Davies et al. / Marine and Petroleum Geology 37 (2012) 1e6


during ?uid advection may also contribute to their development (Brown, 1990; Cartwright et al., 2007). They are recognised on seismic re?ection data on the basis of vertically aligned discontinuities in otherwise continuous re?ections (Fig. 3b and Cartwright et al., 2007; L?seth et al., 2011). Stimulated hydraulic fractures are created to signi?cantly increase the rate of production of oil and gas from ?ne-grained, low permeability sedimentary rocks such as shale. Commonly a vertical well is deviated so that it is drilled strata-parallel through the shale reservoir (Fig. 2). The production casing is perforated and hydraulic fractures stimulated by injecting saline water with chemical additives. ‘Proppant’ (for example sand) is used to keep the fractures open (see King, 2010). Hydraulic fracture stimulation from a horizontal borehole is usually carried out in multiple stages with known volumes and compositions of ?uid (e.g. Bell and Brannon, 2011). Rather than pipes forming, clustering of fractures commonly occurs along planes, which are theoretically orthogonal to the least principle stress direction. So there are fundamental differences in the geometry of these fracture systems compared to those that cluster to form pipes, the reasons for which are not yet understood. Hydraulic fractures can be also be stimulated unintentionally for example as an underground blowout (e.g. Tingay, 2003) and they can unintentionally be caused by the injection of waste water at high enough rates to generate pore pressures which exceed the pressure required for hydraulic fracturing (e.g. L?seth et al., 2011). 2. Datasets Hundreds of pipes have recently been identi?ed on 3D seismic re?ection surveys in continental margin settings (Davies and Clarke, 2010; Hustoft et al., 2010; Moss and Cartwright, 2010; L?seth et al., 2011). We compile new data based upon these surveys on the vertical extents of 1170 pipes (e.g. Fig. 3ab). Pipe heights were measured by recording their bases and tops in two-way-travel time and converting these to heights using estimated seismic velocities for the host successions (Davies and Clarke, 2010; Moss and Cartwright, 2010; Hustoft et al., 2010). Errors are related to the seismic resolution and the estimation of the velocity of the sediment and are probably <20%. Because of the limits of vertical seismic resolution the numbers of hydraulic fractures that have vertical extents of less than 100 m are probably underestimated. We have not included in this study the heights of vertically extensive injectites or igneous intrusions because their likely modes of formation are quite different to stimulated hydraulic fractures and pipe structures. The vertical and horizontal extents of stimulated hydraulic fractures used to recover hydrocarbons can be estimated using the energy released by the hydraulic fracturing which is recorded as microseismicity in a nearby borehole (e.g. Maxwell et al., 2002). Both

tension and shear fractures are detected. We used a compilation of microseismic events (Fisher and Warpinski, 2011). They presented measurements of the fracture tops and bottoms for ‘thousands’ of mapped fracture treatments performed in the Barnett, Woodford, Marcellus and Eagle Ford shale gas formations recorded from early 2001 to the end of 2010 (their Figs. 2e6 respectively). A ?fth unpublished dataset from the Niobrara shale (Colorado, USA) was compiled in the same way and released by Halliburton for publication here. These operations varied in terms of the depth of fracturing operations, the execution of the fracturing process and the geological setting. It represents the majority of the data released into the public domain at the time of writing. Because we did not have access to the primary database our measurements of fracture height were made by digitising their published and unpublished graphs (Fisher and Warpinski, 2011 e Fig. 3c). This will have introduced errors but again this will be mainly with the shorter hydraulic fractures (with vertical extents < 100 m) therefore the numbers of these are also underestimated. Because it is the taller fractures we focus on, this bias does not change our main conclusions. There are also errors associated with the microseismic method, mainly due to estimating the velocity of the rock between the hydraulic fracture and the monitoring well (e.g. Maxwell et al., 2002). There is signi?cant uncertainty in the depth of the initiation of fracture systems caused by underground well blowouts, hence we do not draw on this source of data in this paper. But the depth of the fracture initiation is well constrained where waste water is being injected, so here we use a recently reported example from the Tordis Field, offshore Norway where the fractures eventually intersected the seabed (L?seth et al., 2011). This example provides some additional context for the natural and stimulated datasets described. 3. Results 3.1. Natural hydraulic fractures Offshore of Mauritania 368 vertical pipes were identi?ed over an area of 1880 km2 (Davies and Clarke, 2010 e Fig. 3ab) and it was possible to measure the vertical extent of 360 of these. They are hosted on a passive continental margin, probably within ?ne-grained turbidites and foram-nannofossil hemipelagites of Neogene age (Henrich et al., 2010). A graph of frequency against vertical extent shows the most common vertical extent is between w200 m and w300 m. The tallest conduit is w507 m (Fig. 4a). The average vertical extent is 247 m. Offshore of Namibia, we measured 366 vertical chimneys in a succession composed of ?ne-grained claystones of Miocene to Recent age (Moss and Cartwright, 2010). The average vertical extent is 360 m. The maximum vertical extent is w1106 m (Fig. 4a). Vertical chimneys also form offshore mid-Norway within the Oligocene to

Figure 4. (a) Graph of frequency against vertical extent for natural hydraulic fractures (pipes) identi?ed on 3D seismic data from offshore Mauritania, Namibe Basin and midNorway. (b) Probability of non-exceedance against fracture height for offshore Mauritania, offshore mid-Norway and Namibe Basin.


R.J. Davies et al. / Marine and Petroleum Geology 37 (2012) 1e6

Figure 5. Graphs of frequency against hydraulic fracture height for (a) upward and (b) downward propagating fractures in the Marcellus, Barnett, Woodford, Eagle Ford and Niobrara shales. Graphs of probability of exceedance against height of (c) upward propagating fractures and (d) downward propagating fractures.

Recent ?ne-grained mudstone and siliceous mudstones of the Brygge, Kai and Naust Formations (Hustoft et al., 2010). 66% of these terminate at the seabed. Of the 446 pipes the average vertical extent is 338 m. The maximum vertical extent is w882 m (Fig. 4a). Graphs of hydraulic fracture height against the probability of non-exceedance of this height for each dataset show the probability of pipes exceeding a range of vertical extents (Fig. 4b). Based upon these data the probability of a pipe extending vertically >350 m is w33% (Fig. 6ab). 3.2. Stimulated hydraulic fractures Our compilation of data from the USA shales (Fisher and Warpinski, 2011 e Fig. 3c) shows that generally the Marcellus is the shallowest reservoir, then the Niobrara, Barnett, Woodford and Eagle Ford. The maximum upward propagation of fractures initiated in these reservoirs is w588 in the Barnett shale but in each case the vast majority of hydraulic fractures propagate much shorter distances (Fig. 5ab). The maximum upward propagation recorded to date in the Marcellus shale is w536 m. The graphs show that the probability of an upward propagating fracture exceeding a height of 200 m, for example, is highest for those initiated in the Marcellus then the Barnett, Woodford, Niobrara and Eagle Ford shale reservoirs. Based upon these data the probability that an upward

propagating hydraulic fracture extends vertically >350 m is w1% (Fig. 6ab), but the probability is probably lower than this because we cannot measure all of the shorter fractures. We cannot accurately estimate the average vertical extent for the same reason. 3.3. Unintentionally stimulated hydraulic fractures At the Tordis Field, offshore Norway, waste water produced due to oil production was injected at w900 m below the surface. This caused hydraulic fractures to propagate approximately 900 m to the seabed. Pressure pro?les from the injection well show a stepped fracturing of the overburden (L?seth et al., 2011). The injection lasted for approximately 5.5 months, while the leakage to sea?oor may have occurred for between 16 and 77 days (L?seth et al., 2011). 4. Interpretation and discussion 4.1. Vertical extent Offshore mid-Norway there are controls on the locations of the bases of the pipes as many emanate from overpressured strata and 66% terminate at the present-day seabed (Hustoft et al., 2010) and these controls cause the peak in the frequency versus depth plot between 300 and 350 m (Fig. 4a). Both of these factors have an

Figure 6. (a) Graph of frequency against fracture height for all stimulated and natural hydraulic fractures. (b) Graph of probability of non-exceedance against fracture height for stimulated and natural hydraulic fractures.

R.J. Davies et al. / Marine and Petroleum Geology 37 (2012) 1e6


in?uence on the shape of the probability of exceedance versus height curves (Fig. 4b). In contrast only 12 of the 360 pipes from offshore Mauritania terminate at a palaeo-seabed and a small number of pipes in the Namibe Basin do this. Despite the limitations of the datasets it is clear most of the natural hydraulic fractures reported here are 200e400 m in height and that very few natural fracture systems reported to date propagate beyond a height of 700 m. The tallest is 1106 m, which is comparable to the tallest injectites documented (Hurst et al., 2011). Lastly hydraulic fractures that cluster to form pipe structures generally propagate upwards further than stimulated hydraulic fractures (Fig. 6ab). The vast majority of stimulated hydraulic fractures have a very limited vertical extent of <100 m (Figs. 3c and 5a) and the tallest is w588 m. These taller hydraulic fractures probably form by intersecting existing faults which has been recognised on the basis of changes in the clustering and the magnitude of microseismic events (Warpinski and Mayerhofer, 2008). 4.2. Controls on vertical extent Stimulated hydraulic fractures probably form by a number of small fracture propagation events rather than a single one. The stimulation of the hydraulic fractures in shale gas reservoirs normally takes place over time periods of only 1e2 h for a single fracturing stage. When this period has been increased, for example in the Barnett Shale, where pumping for an 11.7 h period took place with a total volume of c. 5565 m3, the maximum height of hydraulic fractures was only w266 m (Maxwell et al., 2002). During pumping periods of up to 11.3 h into the Barnett Shale the relationship between height and pumping time and volume was shown not to be strong (Shelley et al., 2011) and fractures stopped propagating vertically after only 1e3 h. Therefore there is little relationship between pumping time and fracture height when measured over these timescales. In the shale gas provinces local geology such as variations in lithology, provide natural barriers to propagation because of higher con?ning stress or high permeability which allows the ?uid to bleed off (Fisher and Warpinski, 2011). For example in the Barnett shale the Viola and Ellenberger limestones located below the Barnett shale can limit the downward propagation of hydraulic fractures (King et al., 2008). We propose that natural hydraulic fractures that cluster to form pipe structures have greater vertical extents for a number of reasons. There is much more ?uid and much longer timescales available for multiple stages of fracture propagation. A volume of w6 ? 109 m3 is reported for an aquifer in the North Sea (UK) (Heward et al., 2003). Although the ?ux of only some of the ?uid from an aquifer would cause pressure to drop to hydrostatic levels and therefore only some of this ?uid would have a role in pipe development, there are orders of magnitude higher volumes of ?uid available than used in fracture stages in shale gas wells. Gas that comes out of solution during ascent of ?uid in natural hydraulic fractures could also have a contributory role in propagation (Cartwright et al., 2007). The pipes recorded here are hosted within ?ne-grained, relatively homogenous successions on continental margins where there are fewer mechanical boundaries and generally low permeability strata that do not allow ?uid to bleed off and therefore pressure to drop. Lastly there are signi?cant geometrical differences between natural and stimulated hydraulic fractures and we do not know what in?uence this has on height. Despite these differences there are similar trends in the datasets. The vast majority of both natural and stimulated hydraulic fractures included in this study are <500 m in vertical extent (Fig. 6a). This is again because of variations in-situ stress, weak interfaces, material property contrasts and high permeability layers

in sedimentary successions, particularly heterogeneous ones, which provide natural barriers to fracture propagation. At the Tordis ?eld the average rate of injection of water was 7000 m3 a day for 5.5 months (total volume w 1,115,000 m3). Fractures grew from w900 m depth to the surface through Cenozoic (Tertiary) rather than Palaeozoic strata. But this volume of ?uid is over w123 times greater than typically used for fracture stages in the shale gas reservoirs and over a time period hundreds of times longer. L?seth et al. (2011) reported that pressure pro?les from the injection well show a stepwise fracturing of the overburden and that fractures actually propagated for 900 m reaching the surface (the seabed). The complexity of fracture propagation, role of bed boundaries and pre-existing natural fractures was also demonstrated at centimetre to metre scale through the mine-back experiments carried out in the USA where stimulated fractures were exposed by excavating them (e.g. Cipolla et al., 2008). Propagation continues once the stress at the new boundary exceeds the least principal stress and the tensile strength of the host sediment and the stress-intensity at the fracture tip is lower than the critical stress-intensity of the rock being fractured. Therefore to develop the vertically most extensive fracture systems there needs to be long, probably multiple episodes of injection of high pressure ?uid (probably >> 1 day). There would be several steps in the propagation of the fracture system, breaking through permeable beds and mechanical boundaries. Mechanically homogeneous successions with low permeability will result in vertically more extensive fractures than heterogeneous formations with variable permeability and con?ning stress. 4.3. Implications and further work Further research should include additional datasets, particularly from new settings that have not undergone fracturing treatments to further increase con?dence that exceptional propagation heights have been captured. Additional data may allow for a better understanding of several potential relationships between the height of fractures and variables such as the type of stress regime (i.e. conducive for shear failure or tensile failure), rock type, volume of pumped ?uid and pumping time. There are some geological scenarios where there could be connectivity of permeable reservoirs through a signi?cant thickness of overburden. For example sand injectites can cut through w1000 m of shale (e.g. Hurst et al., 2011) and this could, given long enough pumping time cause critical pressurisation of shallower strata and therefore shallower fractures. These and other geological scenarios should be considered and modelled. Lastly, stimulated hydraulic fractures have been proposed as a mechanism for methane contamination of aquifers located 1e2 km above the level of the fracture initiation in the Marcellus shale (Osborn et al., 2011). Because the maximum upward propagation recorded to date in the Marcellus shale is w536 m this link is extremely unlikely (Davies, 2011; Saba and Orzechowski, 2011; Schon, 2011). Other mechanisms for contamination such as the leakage of biogenic or thermogenic gas from porous and permeable strata behind well casing and natural migration of methane are more likely. 5. Conclusions Natural hydraulic fracture pipes have the potential to propagate upwards further than stimulated ones. The maximum upward propagation recorded for a stimulated hydraulic fracture to date is w588 m in the Barnett shale in the USA. Based upon the data


R.J. Davies et al. / Marine and Petroleum Geology 37 (2012) 1e6 Hubbert, M.K., Willis, D.G.W., 1957. Mechanics of hydraulic fracturing. Transaction of the American Institute of Mining, Metallurgical, and Petroleum Engineers Incorporated 210, 153e168. Hurst, A., Scott, A., Vigorito, M., 2011. Physical characteristics of sand injectites. Earth-Science Reviews 106, 215e246. Hustoft, S., Bünz, S., Mienert, M., 2010. Three-dimensional seismic analysis of the morphology and spatial distribution of chimneys beneath the Nyegga pockmark ?eld, offshore mid-Norway. Basin Research 22, 465e480. Huuse, M., Jackson, C.A.-J., Van Rensbergen, P., Davies, R.J., Flemings, P.B., Dixon, R.J., 2010. Subsurface sediment remobilization and ?uid ?ow in sedimentary basins: an overview. Basin Research 22, 342e360. Julian, B.R., Foulger, G.R., Monastero, F.C., Bjornstad, S., 2010. Imaging hydraulic fractures in a geothermal reservoir. Geophysical Research Letters 37, L07305. King, G.E., Haile, L., Shuss, J., Dobkins, T.A., 2008. Increasing Fracture Path Complexity and Controlling Downward Fracture Growth in the Barnett Shale. SPE. 119896. King, G.E., 2010. Thirty Years of Shale Gas Fracturing: What Have We Learned? SPE. 133456. Laubach, S.E., Marrett, R.A., Olson, J.E., Scott, A.R., 1998. Characteristics and origins of coal cleat: a review. International Journal of Coal Geology 35, 175e207. Legarth, B., Huenges, E., Zimmermann, G., 2005. Hydraulic fracturing in a sedimentary geothermal reservoir: results and implications. International Journal of Rock Mechanics and Mining Sciences 42, 1028e1041. L?seth, H., Wensaas, L., Arntsen, B., Hanken, N., Basire, C., Graue, K., 2001. 1000 m Long Gas Blow Out Chimneys. In: 63rd EAGE Conference and Exhibition, Extended Abstracts, 524 p. L?seth, H., Wensaas, L., Arntsen, B., Hanken, N.-M., Basire, C., Graue, K., 2011. 1000 m long gas blow-out chimneys. Marine and Petroleum Geology 28, 1047e1060. Maxwell, S.C., Urbancic, T.J., Steinsberger, N., Zinno, R., 2002. Microseismic Imaging of Hydraulic Fracture Complexity in the Barnett Shale. SPE. 77440. McConaughy, D.T., Engelder, T., 1999. Joint interaction with embedded concretions: joint loading con?gurations inferred from propagation paths. Journal of Structural Geology 21, 1637e1652. Montgomery, C.T., Smith, M.B., 2010. Hydraulic fracturing: history of an enduring technology. Journal of Petroleum Technology 62, 26e32. Moss, J.L., Cartwright, J., 2010. 3D seismic expression of km-scale ?uid escape chimneys from offshore Namibia. Basin Research 22, 481e501. Osborn, S.G., Vengosh, A., Warner, N.R., Jackson, R.B., 2011. Methane contamination of drinking water accompanying gas-well drilling and hydraulic fracturing. Proceedings of the National Academy of Sciences 108, 8172e8176. Polteau, S., Mazzini, A., Galland, O., Planke, S., Malthe-S?renssen, A., 2008. Saucershaped intrusions: occurrences, emplacement and implications. Earth and Planetary Science Letters 266, 195e204. Saba, T., Orzechowski, M., 2011. Lack of data to support a relationship between methane contamination of drinking water wells and hydraulic fracturing. Proceedings of the National Academy of Sciences 108, E663. Savalli, L., Engelder, T., 2005. Mechanisms controlling rupture shape during subcritical growth of joints in layered rock. Geological Society of America Bulletin 117, 436e449. Schon, S.C., 2011. Hydraulic fracturing not responsible for methane migration. Proceedings of the National Academy of Sciences 108 (37). Shelley, R.F., Lyle, P.E., Lehman, V., Shah, K., 2011. Survey of More than 1,000 Fracture Stage Database with Net Pressure in the Barnett Shell, Part 2. SPE. 143330. Simonson, E.R., Abou-Sayed, A.S., Clifton, J.J., 1978. Containment of Massive Hydraulic Fractures, SPE, J 18(1), 27e32 pp. Tingay, M.R.P., Hillis, R.R., Morley, C.K., Swarbrick, R.E., Drake, S.J., 2005. Present day stress orientation in Brunei: a snapshot of ‘prograding tectonics’ in a Tertiary delta. Journal of the Geological Society, London 162, 39e49. Tingay, M.R.P. 2003. In situ stress and overpressures of Brunei Darussalam. PhD thesis, University of Adelaide. Warpinski, N.R., Mayerhofer, M.J., 2008. Stimulating Unconventional Reservoirs: Maximizing Network Growth while Optimizing Fracture Conductivity. SPE.114173. Zuhlsdorff, L., Spie?, V., 2004. Three-dimensional seismic characterization of a venting site reveals compelling indications of natural hydraulic fracturing. Geology 32, 101e104.

presented here the probability that stimulated hydraulic fractures extend vertically beyond 350 m is w1%. Microseismic measurement of fracture propagation is an essential monitoring tool which allows us to provide an evidence base for the setting of the minimum vertical separation between the shale gas reservoir and shallower aquifers. This is a comprehensive compilation of data, but of course should be added to with new fracture height data from other regions, as the different geological conditions may result in unusually short or tall fractures. Building upon this dataset and deriving probabilities from it will help inform industry and academic geoscientists and engineers, regulators, non government organisations and publics on safe separation distances and help ensure environmentally safe shale gas operations. Acknowledgements We are very grateful to Steve Wolhart (Halliburton) for providing the unpublished data from the Niobrara shale (Fig. 3). Katie Roberts is thanked for providing ?eld photographs (Fig. 1ab), Jonny Imber, Roger Crouch, Jon Trevelyan and Joe Cartwright are thanked for discussion, although the opinions are those of the authors. Both Mads Huuse (Manchester University, UK) and Terry Engelder (The Pennsylvania State University, USA), provided very thorough, constructive and informative reviews that allowed important revisions to the manuscript. References
Bell, C.E., Brannon, H.D., 2011. Redesigning Fracturing Fluids for Improving Reliability and Well Performance in Horizontal Tight Shale Applications. SPE. 140107. Brown, K.M., 1990. The nature and hydrogeological signi?cance of mud diapirs and diatremes for accretionary systems. Journal of Geophysical Research 95, 8969e8982. Cartwright, J., Huuse, M., Aplin, A., 2007. Seal bypass systems. American Association of Petroleum Geologists 91, 1141e1166. Cipolla, C.L., Warpinski, N.R., Mayerhofer, M.J., 2008. Hydraulic Fracture Complexity: Diagnosis, Remediation and Exploitation. SPE. 115771. Cosgrove, J.W., 1995. The expression of hydraulic fracturing in rocks and sediments. In: Fractography: Fracture Topography as a Tool in Fracture Mechanics and Stress Analysis. Geological Society Special Publication No 92, pp. 187e196. Davies, R.J., Clarke, A.L., 2010. Storage rather than venting after gas hydrate dissociation. Geology 38, 963e966. Davies, R.J., 2011. Methane contamination of drinking water caused by hydraulic fracturing remains unproven. Proceedings of the National Academy of Sciences 108 (43), E871. Engelder, T., Lash, G.G., 2008. Marcellus Shale Play’s Vast Resource Potential Causing Big Stir in Appalachia. The American Oil and Gas Reporter. Fisher, K., Warpinski, N., 2011. Hydraulic Fracture-Height Growth: Real Data. SPE. 145949. Henrich, R., Cherubini, Y., Meggers, H., 2010. Climate and sea level induced turbidite activity in a canyon system offshore the hyperarid Western Sahara (Mauritania): the Timiris Canyon. Marine Geology 275, 178e198. Heward, A.P., Scho?eld, P., Gluyas, J.G., 2003. The Rotliegend reservoir in Block 30/ 24, UK Central North Sea: including the Argyll (renamed Ardmore) and Innes ?elds. Petroleum Geoscience 9, 295e307.


How far can they go
Only those who risk going too far can ever know how far they can go
How Far can Structural Grammar Go
Computer simulation of hydraulic fractures
Multiple Hydraulic Fractures in Horizontal Wells
Sports and Study--Can they go together
IPTC-12303-MS-P Application of Horizontal Wells with Multiple Hydraulic Fractures
Block-level Virtualization How far can we go