3 International and UK experience of induced seismic activity arising from hydraulic fracturing
A review of induced seismicity in UOG operations confirms that the probability of felt earthquakes caused by hydraulic fracturing for recovery of hydrocarbons is very small. In the US over 1.8 million hydraulic fracturing operations for shale gas recovery have been carried out and there are only three documented cases of induced earthquakes. The largest of these earthquakes had a magnitude of 3.0. In western Canada, there is evidence of increases in the annual numbers of earthquakes and a number of documented examples of earthquakes with magnitudes larger than 3 in Canada that have been linked to hydraulic fracturing for shale gas recovery. The largest of these was a magnitude of 4.4 earthquake.
Recent increases in earthquake rates and significant earthquakes in many areas of the Central and Eastern United States have been linked to wastewater injection in deep disposal wells rather than hydraulic fracturing, and provide a considerable body of evidence that this activity presents a non-negligible risk of damaging earthquakes. Seismic hazard forecasts for the Central and Eastern United States show increases in earthquake hazard by a factor of 3 or more in some areas of induced earthquake activity. However, although many wastewater injection wells can be associated with earthquakes, the majority are not. Additionally, the nature of the wastewater injected into deep wells varies and while some comes from hydraulic fracturing used in unconventional oil and gas production, much of this is produced water from conventional hydrocarbon production.
The process of hydraulic fracturing in order to increase the permeability of reservoir formations and stimulate the recovery of hydrocarbons is also generally accompanied by microseismicity, commonly defined as earthquakes with magnitudes of less than 2.0 that are too small to be felt. Mapping this microseismicity during hydraulic fracturing operations is widely acknowledged as the best means of characterising stimulated fracture networks in unconventional reservoirs (Maxwell, 2010). Induced seismicity during hydraulic fracture operations has been discussed by a number of authors in the last few years, including Shetema et al. (2012), Warpinski et al. (2012) and Ellsworth (2013).
A report by the National Research Council in the U.S. (NAS, 2012), which examined the scale, scope and consequences of seismicity induced during fluid injection and withdrawal related to energy technologies, concluded that the process of hydraulic fracturing a well, as presently implemented for shale gas recovery, does not pose a high risk for inducing felt seismic events. Similarly, a Royal Society and Royal Academy of Engineering report (2012) also examined the risks associated with hydraulic fracturing during shale gas exploration and production, concluding that the surface impacts of any seismicity induced by hydraulic fracturing would be negligible.
However, there have also been a number of documented examples of rather larger induced earthquakes during hydraulic fracturing operations in both the United States (Holland, 2013) and particularly Canada (BC Oil and Gas, 2012, 2014; Atkinson et al., 2015, 2016; Schultz et al., 2015). The largest of these events had a magnitude of 4.4 Mw, which is, to date, the largest known earthquake induced by hydraulic fracture operations in a hydrocarbon field anywhere in the world.
In this section, we present an extensive review of induced seismicity in UOG operations. Limited data is available globally, and we focus on three geographic areas: central and eastern United States; the West Canada Sedimentary Basin; and Blackpool, UK. Given the limited data available, we also discuss examples of recent seismicity related to waste-water disposal in the eastern United States. We will examine available UOG data from recent examples, such as induced seismicity in Blackpool, UK, as well as data from other analogues to investigate the controlling factors on induced seismicity during fluid injection and examine the relationships between injection volume and pressure and induced seismicity.
3.3 Experience in the United States
In the US, where ~1.8 million hydraulic fracturing operations have been carried out in ~1 million wells (Gallegos and Varela, 2014), felt seismicity induced by hydraulic fracturing appears to be extremely rare and there are only a handful of documented cases of hydraulic fracture induced earthquakes (Holland, 2013; Friberg et al., 2014; Skoumal et al., 2015). The magnitudes of the induced earthquakes in reservoirs such as the Barnett Shale (Maxwell et al., 2006) and the Cotton Valley (Holland, 2011) are typically less than 1 Mw. However, it should be pointed out that most sites of UOG operations lack independent instrumentation for monitoring induced seismicity and that earthquakes with magnitudes of 2.5 or less will fall below the detection thresholds of regional seismic monitoring networks.
Figure 3.1 shows seismicity of Texas and South Oklahoma along with some of the main shale gas plays in the region. The US catalogue contains 7992 events in this region in the period 1850 to present. Fewer than 450 of these are in Texas, of which 149 have M ≥ 3.0.
Figure 3.1. Red circles show the seismicity of Texas and South Oklahoma. Circles are scaled by magnitude. Earthquake data from the U.S. Geological Survey (USGS) ComCat2. The coloured areas show some of the main shale gas plays from U.S. Geological Survey Digital Data Series 69-Z, Map of assessed shale gas in the United States, 2012.
The Barnett Shale extends over a total area 72,000 km2 and is present across the Fort Worth Basin in north-central Texas (Bruner and Smosna, 2011). The Fort Worth Basin is mapped at surface, though toward the east of the area it is overlain by rocks of Cretaceous age. The Barnett Shale is Mississippian in age (middle Carboniferous), and formed in a foreland basin, created along the margin of convergent tectonic plates (Bruner and Smosna, 2011; Montgomery et al., 2005). The basin is bordered by a NE-SW trending thrust belt (Ouachita Structural Front) and the Muenster Arch, a basement high (Bruner and Smosna, 2011). The shale members are composed of distinct shale units separated by limestone units locally (Montgomery et al., 2005). The unit varies in thickness from 15 to 305 m, but most production is from the northern part of the basin where it is relatively thick (Montgomery et al., 2005).The Barnett Shale is underlain by the Ellenburger Group which is karstified, resulting in high-angle normal faults, karst fault-chimneys and local subsidence features in the Barnett Shale (Bruner and Smosna, 2011). One set of natural fractures with a strike of 100-120 is recognised in the Barnett (Bruner and Smosna, 2011), formed by tectonic movements during the Ouachita Orogeny. A map of the Barnett Shale in Frohlich (2012) shows mapped faults with a dominant NE-SW trend in the south-eastern part of the Barnett Shale, and a lesser population of faults with a NW-SW trend in the north-western part of the Barnett Shale.
Figure 3.2 shows the cumulative number of earthquakes in Texas as a function of time between 1975 and 2015. It is clear that between 1975 and 2005, the number of earthquakes increases at a constant rate, whereas after 2005 it starts to increase. As in other parts of Central and Eastern North America this increase in seismicity rates has been attributed to the injection of brines from oil and gas production into wells that are drilled to dispose of large volumes of waste water over many years (Ellsworth, 2013), and Frohlich et al. (2011) attribute a sequence of earthquakes in the Dallas-Fort Worth region to the disposal of waste-water in deep injection wells
Figure 3.2. Cumulative number of earthquakes in Texas as a function of time between 1975 and 2015. The red squares show all magnitudes, the blue crosses show events of magnitude 3.0 or above. Earthquake data from the U.S. Geological Survey (USGS) ComCat
In the time period from 1975 to 2005 there were 53 earthquakes with a magnitude of 3 or greater in Texas, and, assuming that the catalogue is complete for events of this magnitude and above in this time period, this corresponds to an annual rate of 1.767.
The annual rate for magnitude 3.0 events in Scotland is 0.871, based on the a and b values calculated from the instrumental catalogue of 2.79 and 0.95 respectively. Forty-nine earthquakes with magnitudes of 3.0 or above have been observed in the time period from 1970 to 2014. Scaling the calculated rates by area gives rates of 2.54×10-06/km2 and 7.25×10-06/km2 for Texas and Scotland respectively.
3.4 Earthquakes induced by Hydraulic Fracturing Operations near Blackpool, UK
In Lancashire, UK, 58 earthquakes were linked to fluid injection during hydraulic fracturing at the Preese Hall well in 2011 (de Pater and Baisch, 2011). The largest, on 1 April 2011, had a magnitude of 2.3 and was felt locally. The hydraulic fracturing was carried out during exploration of a shale gas reservoir in the Bowland basin, Lancashire. A further magnitude 1.5 ML earthquake was felt on 27 May, 2011 and also linked to hydraulic fracturing, leading to the suspension of operations at Preese Hall.
The seismicity led to a number of detailed studies of the relationship between the earthquakes and hydraulic fracturing operations (for example, de Pater and Baisch 2011; Eisner et al., 2011). In total, 58 earthquakes were detected in the time period between 31 March and 30 August 2011, nearly all of these occured either during or within a few hours of fracturing operations at Preese Hall. De Pater and Baisch (2011) concluded that the earthquake activity was caused by fluid injection directly into a nearby fault zone, which reduced the effective normal stress on the fault and caused it to fail repeatedly in a series of small earthquakes. A possible causative fault was later identified following a detailed 3-D seismic reflection study (Clarke et al., 2014).
Figure 3.3. Volume of injected fluid (blue line) and earthquakes (red circles, scaled by magnitude) during hydraulic fracturing operations at Preese Hall, Blackpool, between March and June 2011 (after de Pater and Baisch, 2011). There are five distinct hydraulic fracturing stages. Earthquake activity closely correlates with stages 2 and 4. The largest event with 2.3 ML at 02:34 on 1/4/2011 occurred shortly after stage 2. Earthquake data from the British Geological Survey UK Earthquake Catalogue © NERC 2016. Injected fluid volumes provided by Cuadrilla Resources, 2011.
Figure 3.3 shows the injected volume of fluid as a function of time in each of the hydraulic fracture stages carried out at Preese Hall along with the recorded earthquake activity (from de Pater and Baisch, 2011). It is clear that the earthquakes correlate strongly with stages 2, 4 and 5, in which the largest amount of fluid was injected. In two of the hydraulic fracturing stages, 2 and 4, the largest earthquakes occurred approximately ten hours after the start of injection, while the well was shut-in under high pressure. These events were preceded by smaller events, which started immediately after injection, the largest of which was a magnitude 1.4 ML event on 31 March.
No seismicity was observed during stages 1 and 3, and only very weak seismicity occurred during stage 5. The lack of seismicity in stage 3 can be attributed to the smaller pumped volume and the use of flowback. The pumped volume in stage 5 was similar to stages 2 and 4, but there was also flowback, which could explain the lack of larger events. The results show that injected volume and flowback timing are important controlling factors in the level of seismicity, as evidenced from the lack of seismicity during and after stage 3, suggesting that seismicity can be mitigated by modifying job procedure.
Locations for the Blackpool earthquakes were determined by Eisner et al. (2011) and Clarke et al. (2014) among others. Similarity between the recorded events suggested that all the events were from the same location and had the same mechanism. The location is shown in Figure 3.4. It is clear that the location is less than 0.5 km from the well head. In addition, the depths of 3.6 km and 2.9 km, estimated by Eisner et al. (2011) and Clarke et al. (2014) are close to the point of injection (2.3 - 2.7 km) for all 6 stages.
Figure 3.4. Epicentre of Preese Hall earthquakes in April and May 2011 (yellow star), as determined by Eisner et al. (2011). The coloured triangles in (a) show permanent monitoring stations operated by the British Geological Survey at epicentral distances of 75 to 99 km (red), 100 to 149 km (orange) and greater than or equal to 150 km (yellow). The red triangles in (b) show temporary stations deployed after the initial earthquakes on 1 April 2011. The blue triangle shows the location of the Preese Hall well. Topography © Crown Copyright 2016 Ordnance Survey 10037272. Earthquake data from the British Geological Survey UK Earthquake Catalogue © NERC 2016.
The Bowland Basin is a thick accumulation of Carboniferous shales, bounded by two NE-SW trending faults (Andrews, 2013). The target for hydraulic fracturing was the Bowland-Hodder unit, a shale-dominated facies up to 1900 m thick, with limestones and turbidite deposits (Andrews, 2013). As in the Midland Valley of Scotland, the NE-SW structural trend is inherited from crustal weaknesses formed in the Lower Palaeozoic (during the Caledonian Orogeny). Fault systems are often normal, with a strong NE-SW influence and formed toward the end of the Carboniferous. There is no surface outcrop of the Carboniferous strata in this area (the strata are overlain at surface by rocks of Triassic and Permian age) and as a result little was known about the density or frequency of faults. The area is assumed to have had no substantial fault activity since the Permian Period (Clarke et al., 2014).
At a regional scale, the Carboniferous geology buried at depth beneath the Blackpool-Preese Hall area could be considered a continuation of the Ribblesdale Fold Belt. Within the Lancashire Coalfield some 30 km to the south-west of Preese Hall, where the Carboniferous rocks are exposed at surface, the density of faulting as mapped by BGS at 1:50,000 scale is similar to that in the Central Coalfield of the Midland Valley of Scotland.
3.5 Earthquakes induced by hydraulic fracturing in the Horn River Basin, British Columbia
Over 200 earthquakes were induced during hydraulic fracturing operations in the Etsho and Tattoo fields in the Horn River Basin, British Columbia, during 2009-2011 (BC Oil and Gas Commission, 2012). The locations of these events are shown in Figure 3.5. The Horn River Basin lies in northeast British Columbia, between Fort Nelson and the Northwest Territories border, with a total area of 11,500 km2, of which ~350 km2 is used for oil and gas exploration and production. The Horn River Group shales occur throughout the basin and are a target for hydrocarbon extraction and exploitation. The shales and associated limestones are Devonian in age and lie at depths of over 2000 m (McPhail et al., 2008), with combined thicknesses of over 400 m in places. The shales are, for the most part, overlain by rocks of Mesozoic (Cretaceous) age.
The basin is bounded to the west by the Bovie Fault, a broadly NNE-SSW trending compressional-fault structure which extends over 200 km, from northeast British Columbia into southern Northwest Territories (Maclean and Morrow, 2004). The fault was established during the late-Carboniferous to early Permian, with motion renewed during the Cretaceous (Maclean and Morrow, 2004). The area is also intersected by the Trout Lake Fault Zone, a long, linear, NE-SW trending basement fault (Williams, 1977). Fault mapping by operators found abundant, sub-parallel N-S trending deep-seated faults within the Etsho and Tattoo areas of the Horn River Basin, with minor secondary NW-SE faulting evident (BC Oil and Gas Commission, 2012), whilst faulting appears to be confined below the lower Fort Simpson shale, extending into the Precambrian basement.
Figure 3.5. Location of the Etsho and Tattoo areas in the Horn River Basin (after BC Oil and Gas Commission, 2012). Red circles show the NRCan reported epicentres (scaled by magnitude). Small black dots show well positions. Black squares show wells with the Tattoo and Etsho areas. The black polygons show producing fields. Field and well data obtained from the B.C. Oil and Gas Commission, available at http://data.bcogc.opendata.arcgis.com. Topography data, GTOPO30, US Geological Survey.
Thirty-eight earthquakes were detected by the regional seismic monitoring network operated by Natural Resources Canada (NRCan) between 8/4/2009 and 13/12/2011. Twenty-one of the earthquakes had magnitudes of 3.0 or greater, and the largest event had a magnitude of 3.8 ML. This event was also felt by workers in the area. The earthquakes occurred in an area where no previous seismicity had been recorded and a report by the BC Oil and Gas Commission (2012) concluded that the earthquakes were caused by fluid injection during hydraulic fracturing in proximity to pre-existing faults.
Hydraulic fracturing operations in the Etsho area took place between February 2007 and July 2011. During this period, over 90 wells were drilled from 14 different locations, with more than 1,600 hydraulic fracturing stages completed. Twenty-seven of the earthquakes detected by NRCan occurred within 10 km of the Etsho area. Seven drilling pads were located within the same area, five of these were conducting hydraulic fracturing operations when events occurred. All seven of the earthquakes detected by NRCan in the Tattoo area occurred within 10 km of wells in which there were hydraulic fracturing operations when the seismicity occurred.
A dense array consisting of 20 seismometers was deployed by the operator in the Etsho area to study the seismicity in greater detail than was possible with NRCan data. This array operated from 16 June to 15 August 2011 and detected 216 earthquakes ranging from magnitude -0.8 to 3.0 ML, with 19 events greater than 2.0 ML. These earthquakes were interpreted to be related to fault movement and the report by the BC Oil and Gas Commission (2012) concluded that magnitudes from 0.5ML to 1.0ML indicate the transition from fracture driven seismicity to seismicity driven by fault movement. The four earthquakes detected by NRCan in the same time period were relocated by the operator using data from the dense array. The results suggested that the earthquakes were located within 200 m of sections of the borehole where hydraulic fracturing took place.
At both Etsho and Tattoo, all 38 NRCan reported events occurred either during a hydraulic fracturing stage or sometime after one stage ended and another began. No events were recorded before hydraulic fracturing operations began or after the last hydraulic fracturing operations ended.
The average volume of fluid injected into wells for hydraulic fracturing in the Etsho area was approximately 60,000 m3, with a maximum of 138,000 m3 and a minimum of 11,000 m3, with corresponding flow rates of 0.2 m3/s, 0.25 m3/s and 0.13 m3/s, respectively.
In British Columbia, the only previously documented case of induced seismicity, linked to oil and gas activity, occurred in the Eagle Field area, approximately five km north of Fort St. John. Twenty-nine earthquakes with magnitudes from 2.2 to 4.3 ML were recorded from November 1984 to May 1994. Horner (1994) used the Davis and Frohlich (1993) criteria to conclude that the events were induced. High pressure fluid injection for secondary oil recovery was identified as a possible cause. High volume hydraulic fracturing was not employed in the area at that time.
3.6 Earthquakes in the Eola Field, Oklahoma,
In January 2011, a sequence of earthquakes occurred in close proximity to a well, which was being hydraulically fractured in the Eola‐Robberson oil field, south‐central Oklahoma (Holland, 2013). A total of 116 earthquakes were detected by Holland (2013) between 17/01/2011 at 19:06 and 23/01/2011 at 3:13 GMT. Hydraulic fracturing operations in the Picket Unit B Well 4-18 took place between 16/01/2011 at 18:43 and 22/01/2014 at 16:54 GMT. Earthquake magnitudes varied from 0.6 to 2.9 ML, with 16 earthquakes having magnitudes of 2.0 ML or greater. The locations calculated by Holland (2013) suggest that the earthquakes occurred at shallow depths from 2 to 3 km and within ∼2.5 km of the well. The alignment of the hypocentres suggests that the earthquakes occurred on a fault striking N166°E, subparallel to the mapped faults in the area. The first earthquake occurred ∼24 hours after hydraulic fracturing began at the well. This delay is consistent with the diffusion of pore pressure in the subsurface over a distance of ∼2 km. The strong spatial and temporal correlation between hydraulic fracturing and earthquakes suggests that the earthquakes were induced. This correlation is strengthened because hydraulic fracturing operations ceased for ∼2 days due to bad weather, and earthquakes can be observed to cease during this period and resume after hydraulic fracturing had resumed. In addition, no other similar earthquakes were identified at other times before or after hydraulic fracturing.
The Eola-Robberson field lies at the northern edge of the Ardmore Basin, and is part of the Devonian-aged Woodford Shale gas play (Holland, 2013). The Woodford Shale gas play covers virtually the entirety of Oklahoma, and stretches from southern Kansas to west Texas. It is between 15 - 90 m thick, and is found at depths of between 275 and 3960 m (Vulgamore et al., 2007). The field contains a highly folded and faulted thrust system (Holland, 2013), and has undergone numerous phases of tectonic deformation, from initial rifting during the Cambrian (Keller et al., 1983) to transpression (and probable reactivation of Cambrian structures) during the late Carboniferous (Granath, 1989). Numerous major parallel faults trend WNW to ESE, with several NW to SE trending faults intersecting these. The Eola field is block faulted between major faults with a strike-slip component, with fault dips near vertical (Harlton, 1964).
3.7 Earthquakes in Harrison County, Ohio
Friberg et al. (2014) discuss an earthquake sequence detected in Harrison County, Ohio between 7 September and 14 December 2014, and relate it to a hydraulic fracture operation. The sequence consisted of several hundred earthquakes, the largest of which had a magnitude of 2.1 ML. The start and stop of the activity is coincident with the start and stop of hydraulic fracture operations in the nearby Ryser-2, Ryser-3 and Ryser-4 horizontal wells, with some temporal delay. The similarity in the recorded waveforms suggested that the events originated from the same source location and the located earthquakes line up along a steeply dipping east-west trending structure at a depth of 3.2 km immediately beneath the wells. Friberg et al. (2014) conclude that the earthquakes occurred on a fault in the Pre-Cambrian crystalline basement, not in the Palaeozoic formations where the wells were located.
3.8 Earthquakes in Poland Township, Ohio
Skoumal et al. (2015) identified 77 earthquakes in Poland Township, Ohio, between 4-12 March, 2014, that were spatially and temporally correlated with active hydraulic fracture operations. The largest earthquake had a magnitude of 3.0 ML and five other events had magnitude of 2.0 ML or above. The earthquakes occurred during hydraulic fracturing of two horizontal wells that were located 0.8 km away and activity decayed after the Ohio Department of Natural Resources issued a shutdown of the hydraulic fracturing. Precise relative locations calculated for the earthquake hypocentres lined up along a vertical plane. The calculated focal mechanism gave a vertical fault plane in good agreement with both the alignment of the hypocenters and with the regional stress field. Skoumal et al. (2015) conclude that the hydraulic fracturing induced slip on a pre-existing fault that was optimally oriented with respect to the regional stress field.
3.9 Earthquakes in the Montney Trend, British Columbia
The Montney Trend is a 29,850 km2 siltstone formation that stretches from the British Columbia-Alberta border, near Dawson Creek, 200 km northwest to the Rocky Mountain foothills. It lies at the western edge of the West Canada Sedimentary basin. Unconventional gas development in the Montney began in the mid-2000s, and by 2014 the region had become British Columbia's most important natural gas producing area. In 2014, the Montney had over 1,700 active natural gas wells.
Natural Resources Canada recorded 231 earthquakes in the Montney Trend between May 2013 and October 2014, with magnitudes from 1.0 to 4.4 Mw. A study by the BC Oil and Gas Commission (2014) found that 193 were correlated in both space and time with hydraulic fracturing operations in the Doe-Dawson, Septimus, Altares, Beg-Town and Caribou gas producing areas. Another 38 earthquakes were found to be correlated with wastewater disposal wells in the Graham and Pintail areas. The maximum injected volume in any stage was approximately 2,200 m3.
The largest event had a magnitude of 4.4 Mw and was located in the Caribou area, approximately 200 km northwest of Fort St. John. A magnitude 4.2 Mw earthquake was recorded on 28 May, 2013, in the Septimus area, 10 km south of Fort St. John.
Figure 3.6. Coloured circles show earthquakes recorded by Natural Resource Canada between May 2013 and end October 2014. The circles are coloured by date. The yellow stars show the locations of the three largest earthquakes in the sequence, with magnitudes of 4.4, 4.2 and 3.8 Mw. The blue shaded areas show the gas producing areas linked to the seismicity, while the small black dots show well positions. The grey squares show towns. Earthquake data from the National Earthquake DataBase (NEDB), compiled by Natural Resources Canada. Well data obtained from the Alberta Energy Regulator, available at https://www.aer.ca/data-and-publications.
No injuries or property damage were linked to the seismicity and the recorded ground motions were below damage thresholds. However, the event triggered an automatic shutdown of a nearby gas plant and precautionary flaring of gas. Several hundred people were without power for a prolonged period. Several instances of casing deformation occurred within the horizontal portion of shale gas wellbores, but no loss of integrity within the wells and no impact on the vertical portions of wellbores. Atkinson et al. (2015) found that high-frequency ground motions were lower than those predicted by ground motion prediction equations commonly used in seismic hazard assessments, possibly as a result of a low stress drop. However, Atkinson et al. (2015) also suggest that moderate-induced earthquakes in the magnitude range 4-5 may be damaging as a result of the expected shallow focal depths.
3.10 Crooked Lake, Alberta
A sequence of more than 160 earthquakes occurred between November 2013 and December 2014 near Crooked Lake, Alberta, Canada. Schultz et al. (2015) find that the seismicity in the Crooked Lake Sequence is correlated both in space and time with hydraulic fracturing operations in the McKinley and Waskahigan fields, approximately 30 km west of Fox Creek. The largest event in the sequence had a magnitude of 3.8 Mw. Earthquake activity has continued in this region and a magnitude 4.4 earthquake on 12 January 2016, 15 km west-northwest of Fox Creek is also suspected to be due to hydraulic fracturing.
Hydraulic fracturing in the area is used to exploit the Duvernay Formation, an organic-rich shale with an average depth of approximately 3400m. Hydraulic fracturing operations typically consist of multi-staged pressure treatments with average pressures, pump rates and volumes of 60 MPa, 9 m3/min and 2700 m3, respectively.
Figure 3.7. Map showing triggered seismicity west of Fox Creek, Alberta, Canada. The red circles show seismicity recorded by Natural Resources Canada, in the period November 2013 to May 2016. The grey circles show locations calculated for the Crooked Lake sequence between November 2013 and December 2014 by Schultz et al. (2015). Earthquake symbols are scaled by magnitude. Earthquake data from the National Earthquake DataBase (NEDB), compiled by Natural Resources Canada. Topography data, GTOPO30, US Geological Survey.
3.11 Raton Basin, Colorado-New Mexico
Coal-bed methane is extracted by direct drilling into a coal seam, which allows both gas and produced water to flow to the surface. Hydraulic fracturing is sometimes used to release gas from a coal seam. Australia, Canada and the United States all have commercial coal bed methane production. In the United States, most CBM production came from the Rocky Mountain States of Colorado, Wyoming, and New Mexico. A number of notable earthquakes have been observed in the Raton Basin, situated along the Colorado-New Mexico border, since coal-bed methane production began there in the 1990's (Figure 3.8).
The Raton Basin is a coal-bearing sedimentary basin, between the Great Plains to the east and the Rio Grande rift to the west. It is approximately 150 km long and 75 km wide at its maximum. Geological mapping within the Raton Basin reveals little evidence for faulting. The basin is bounded by thrust faults at its western edge and by a west-dipping northwest-striking normal fault along its eastern side. There are few mapped faults within the basin, although Robson and Banta (1987) identified two normal faults in the basin that are buried in the Precambrian basement. These do not outcrop at the surface and there is no evidence that they have been active in the Quaternary period. Similarly, the USGS Quaternary Fault and Fold Database (2016) does not contain any active faults within the basin.
Figure 3.8. Circles show earthquake activity in the Raton Basin (USGS ComCat, 2016). The circles are scaled by magnitude. The two red circles show the magnitude 5 earthquake on 10 August 2005 and the magnitude 5.3 earthquake on 23 August 2011. The grey shaded area shows the extent of the basin (from Coleman and Cahan, 2012).
Coal-bed methane production in the Raton Basin began in 1994 and has continued to present. Production is from the Raton, Vermejo and Trinidad formations, at depths from 200 to 800 m depth. Considerable formation water is produced with the methane some of this is disposed of in deep wells. Wastewater disposal began in Colorado in 1994 and in New Mexico in 1999, with injection primarily in to the Dakota formation (Johnson, 1969), at depths between 1250 and 2100 m. Figure 3.9 shows how earthquake activity has increased following the start of coal-bed methane production.
Rubinstein et al. (2014) investigated seismicity in the Raton basin from 1972 to 2013 and conclude that the disposal of wastewater from the coal-bed methane field in deep injection wells is responsible for inducing the majority of the seismicity since 2001. Evidence for this includes a major increase in seismicity shortly after major wastewater injection began in 1999 and the fact that most of the seismicity lies within 5 km for active wells and is at a shallow depth. Also, both the volume of injected water and the injection rates are high.
There have been three notable sequences of seismicity since 2001: August-September 2001; August-September 2005; and August-September 2011. The August-September 2005 sequence included a magnitude 5 earthquake, while the August-September 2011 sequence included a magnitude 5.3 event, the largest recorded earthquake in the area. Earthquakes within the 2001 and 2011 sequences lie within 2 km of high volume injection wells. Two wells adjacent to the magnitude 5.3 earthquake in August 2011 injected 4.9 million cubic meters of wastewater prior to the earthquake.
Figure 3.9. Cumulative number of earthquakes in the Raton Basin with magnitudes greater than 3 as a function of time. Activity increases dramatically shortly after the start of coal-bed methane production (after Rubinstein et al., 2014). Earthquake data from the U.S. Geological Survey (USGS) ComCat
3.12 Induced Seismicity and WasteWater Disposal
There is mounting evidence that the disposal of the wastewater related to oil and gas production in Class II Underground Injection Control (UIC) wells can lead to increases in observed seismicity rates and damaging earthquakes. Ellsworth (2013) shows that that numbers of earthquakes in central and eastern U.S. have increased dramatically in the last few years. This is demonstrated in Figure 3.10, which shows that more than 300 earthquakes with M ≥ 3 occurred in the 3 years from 2010 to 2012, whereas the average number/year from 1967 to 2000 is 21. In addition, several of the largest earthquakes in the U.S. midcontinent in the last few years may have been triggered by nearby disposal wells, and of the seven earthquakes of magnitude 4.0 or greater that occurred east of the Rocky mountains, six are thought to have been induced. The magnitude 4.0 Mw earthquake on 31 December 2011 in Youngstown, Ohio, appears to have been induced by the disposal of wastewater in a UIC well at depths of up to 3.0 km (Kim, 2013). The magnitude 4.7 earthquake in central Arkansas in 2011 has also been linked to disposal of wastewater in a UIC well at depths of 2-3 km (Horton, 2012). The magnitude 4.4 Mw earthquake on 11 September 2011, near Snyder, Texas, occurred in an oil field where injection for secondary recovery has been inducing earthquakes for years (Davis and Pennington, 1989). The largest of these was a magnitude 5.7 earthquake in Prague, central Oklahoma that was located close to active UIC wells, which destroyed 14 homes and injured two people (Kerenan et al., 2013).
Figure 3.10. Cumulative count of earthquakes with M ≥ 3 in the central and eastern United States, 1967-2012 (after Ellsworth, 2013). Earthquake data from the U.S. Geological Survey (USGS) ComCat
These results provide a significant body of evidence that wastewater disposal by injection in to UIC wells poses a significant seismic risk. The report by the National Research Council in the U.S.A. (NAS, 2012), which examined the scale, scope and consequences of seismicity induced during fluid injection and withdrawal related to energy technologies, concluded that injection for disposal of wastewater derived from energy technologies into the subsurface does pose some risk for induced seismicity, but very few events have been documented over the past several decades relative to the large number of disposal wells in operation.
Figure 3.11 shows a USGS map showing the wastewater injection wells in Central and Eastern US that have (red) and have not (grey) been associated with earthquakes. The fluid injection wells associated with earthquakes are defined as being within a 15 km radius and active at the time of an earthquake. Named areas that have been impacted by these induced earthquakes are delineated by the polygons. This shows that although many wells can be associated with earthquakes, the majority are not. Rubinstein et al. (2015) state that while there are approximately 35,000 active wastewater disposal wells and 80,000 active enhanced oil-recovery wells in the United States, only a very small number of these are known to have induced felt earthquakes. Factors controlling the occurrence of earthquakes are likely to include the size and state of stress of nearby faults as well as fluid pressure changes that are large enough to induce earthquakes.
Some of the wastewater injected into UIC wells comes from hydraulic fracturing used in unconventional oil and gas production, while some is produced water from conventional hydrocarbon production. Produced water is the salty brine that is held in the same pore space as oil and gas and is extracted at the same time at nearly all oil wells. The nature of the wastewater varies from place to place. For example, the fluids disposed of near earthquake sequences in Youngstown, Ohio (Horton, 2012), and Guy, Arkansas (Kim, 2013), are believed to consist largely of spent hydraulic fracturing fluid, whereas the vast majority of the fluid that is injected in disposal wells in Oklahoma is produced water (Murray, 2013).
Figure 3.11. USGS map displaying 21 areas impacted by induced earthquakes as well as the location of fluid injection wells that have and have not been associated with earthquakes. Credit: U.S. Geological Survey, Open-File Report OFR-2016-1035
As a result of this, the USGS has recently produced a 1-year seismic hazard forecast for the Central and Eastern United States (CEUS) that includes contributions from both induced and natural earthquakes. Conversion of ground shaking to seismic intensity (Figure 3.12) indicates that some places in Oklahoma, Kansas, Colorado, New Mexico, Texas, and Arkansas may experience damage if the induced seismicity continues unabated. The chance of having Modified Mercalli Intensity (MMI) VI or greater (damaging earthquake shaking) is 5-12 percent per year in north-central Oklahoma and southern Kansas, similar to the chance of damage caused by natural earthquakes at sites in parts of California.
Figure 3.12. USGS map displaying the intensity of potential ground shaking with a 1% probability of being exceeded in one year. Both natural and human induced earthquakes are included. Credit: U.S. Geological Survey, U.S.G.S. Open-File Report OFR-2016-1035
The process of hydraulic fracturing a well, as presently implemented for shale gas recovery, is generally considered to pose a low risk of inducing either felt, damaging or destructive earthquakes (e.g. NAS, 2012; Royal Society, 2012). Although hydraulic fracturing is generally accompanied by microseismicity, the magnitudes of these events are usually less than 2.0 making them too small to be felt by people. Given the large number of stimulations that have been carried out in in the US and Canada and the small number of felt earthquakes, the probability of felt earthquakes appears to be very small.
There are at least seven documented examples of earthquakes with magnitudes greater than two that have been conclusively linked to hydraulic fracturing for shale gas exploration/recovery. Seismicity is most notable in Canada, where a magnitude of 4.4 Mw near Fort St John in August 2014 is largest known earthquake suspected to have been induced by hydraulic fracture operations in a hydrocarbon field anywhere in the world. This event resulted in an automatic shutdown of a nearby gas plant and precautionary flaring of gas. Felt seismicity associated with hydraulic fracturing of the Preese Hall well near Blackpool in 2011 led to the suspension of all such operations in the UK for several years. In this case, the magnitudes of the observed earthquakes were relatively small, with only one event in the sequence of over 50 exceeding a magnitude of 2.0 ML.
The mechanism for these larger earthquakes is generally well-understood and is constrained by observations and modelling. Fluids injected during the hydraulic fracturing process can change the stress conditions on pre-existing faults making it easier for them to slip. This can happen because the injected fluid increases the pore fluid pressure on the fault, which reduces the frictional resistance and makes it easier for the fault to slip. The reactivation potential of a fault depends both on the existing stress field and magnitude of the stress change caused by the injected fluid. If a fault is close to failure it may only require a small stress perturbation to cause it to slip. The energy released in larger induced events is a result of the long term accumulation of strain rather than the energy from the injected fluid. As a result maximum magnitudes for such induced earthquakes will be similar to those for tectonic earthquakes in the region.
Figure 3.13. Earthquakes in the Western Canada Sedimentary basin (WCSB) with a magnitude of 3.0 or above (after Atkinson et al. (2016). The boxes delineate an area parallel to the foothills of the Rockies where induced seismicity has been observed. Earthquake data from the National Earthquake DataBase (NEDB), compiled by Natural Resources Canada. Topography data, GTOPO30, US Geological Survey.
There is a significant contrast between experience in central United States and Canada. In the central United States, most induced seismicity is linked to disposal of co-produced wastewater from oil and gas extraction (Ellsworth, 2013). Atkinson et al. (2016) suggest that most recent cases of induced seismicity in western Canada are highly correlated in space and time with hydraulic fracturing, and state that the observed maximum magnitude of events associated with hydraulic fracturing appears to exceed the predictions of the McGarr (2014) relationship between the volume of injected fluid and the maximum expected magnitude.
Figure 3.13 shows earthquakes in the Western Canada Sedimentary basin (WCSB) with a magnitude of 3.0 or above (after Atkinson et al., 2016) for the time period 1985-2015. The boxes delineate an area parallel to the foothills of the Rockies where induced seismicity has been observed. Figure 3.14 shows the cumulative number of events as a function of time for the same data as shown in Figure 3.13. It is clear that between 1985 and 2005, the number of earthquakes increases at a constant rate, whereas after 2005 it starts to increase at a greater rate and this corresponds to the increase in the number of hydraulically fractured wells.
The difference in response to hydraulic fracturing in the US and Canada is not well understood. It may be related to higher background seismicity rates in British Columbia and western Alberta, however, this remains speculative. In the time period from 1985 to 2005 there were 99 earthquakes with a magnitude of 3 or greater in the area of interest shown in Figure 3.13, and, assuming that the catalogue is complete for events of this magnitude and above in this time period, this corresponds to an annual rate of 4.95. Taking into account the area of the zone of interest (465,963 km2) gives a scaled seismicity rate of 7.93×10-06/km2, which is almost three times the value calculated for Texas (2.54×10-06/km2) and twice the value calculated for Scotland (4.94×10-06/km2).
It is likely that an earthquake similar in magnitude to the largest events linked to hydraulic fracturing in the West Canada Sedimentary Basin (4.4 MW) would be strongly felt across much of the Midland Valley of Scotland and could even cause some superficial damage. However, earthquakes with magnitudes similar to those observed in Blackpool and Garvin County would be unlikely to cause any damage, although they could be felt by people close to the epicentre and may cause some concern to the local population.
Figure 3.14. Cumulative number of earthquakes with a magnitude of 3.0 or above (blue squares) within the parts of the WCSB in the area delineated by the rectangles in Figure 3.13 from 1985 to 2015 (after Atkinson at al, 2016). The blue line shows the increase in number of hydraulically fractured wells. Earthquake data from the National Earthquake DataBase (NEDB), compiled by Natural Resources Canada. Well data obtained from the Alberta Energy Regulator and the B.C. Oil and Gas Commission, available at https://www.aer.ca/data-and-publications and http://data.bcogc.opendata.arcgis.com.
Studies of earthquake activity in the Raton Basin (Rubinstein et al, 2014) an area that has produced coal-bed methane since 1994, provides strong evidence that a this activity is related to the subsequent disposal of wastewater from the coal-bed methane extraction process by injection into deep wells, rather than from the extraction process itself. Literature was not located concerning induced seismicity and coal-bed methane extraction in Canada, Australia or other parts of the USA, suggesting that this is not a major issue in those areas.
The observed increases in earthquake rates and significant earthquakes in many areas of the Central and Eastern United States, which have been linked to the disposal of wastewater by injection into UIC wells (Ellsworth, 2013; Keranen et al., 2014; Weingarten et al., 2015), provide a considerable body of evidence that this activity has a non-negligible contribution to the seismic hazard. Previous National Seismic Hazard Models (NSHM) for the United States published by the USGS (e.g. Petersen et al., 2014) did not consider non-tectonic events. However, Petersen et al. (2016) have now published 1-year seismic hazard forecast for 2016 for the Central and Eastern United States (CEUS) that includes contributions from both induced and natural earthquakes. This forecast shows increases in earthquake hazard by a factor of 3 or more in some areas of induced earthquake activity. In areas where induced activity appears to have stopped seismic hazard returns to the 2014 result. Petersen et al. (2016) suggest that some places in Oklahoma, Kansas, Colorado, New Mexico, Texas, and Arkansas may experience damage if the induced seismicity continues unabated. The chance of having Modified Mercalli Intensity (MMI) VI or greater (damaging earthquake shaking) is 5-12 percent per year in north-central Oklahoma and southern Kansas, similar to the chance of damage caused by natural earthquakes at sites in parts of California.
Of the considered case studies, the Bowland Basin (Blackpool) is the most geologically similar to the Midland Valley of Scotland and is also on a similar physical scale to the Midland Valley of Scotland, unlike the US and Canadian examples which are at least twice the size. For example, the Barnett Shale in the Fort Worth Basin (Texas) extends for 72,000 km2, whilst the Midland Valley covers around 7,000 km2.
The Midland Valley and the Bowland Basin are around 1,500 km from the nearest plate boundary (the Mid-Atlantic Ridge). The Horn River Basin is less than 1,000 km from the subducting margin of the Pacific Plate. While Oklahoma and Texas are also far from any plate boundaries, the Eola-Robberson field (Oklahoma) has a complex tectonic history with a faulted thrust system, and the Fort Worth (Texas) Basin is bordered on its south-east side by a major thrust front.
Both the Midland Valley and the Bowland Basin have unconventional targets in Carboniferous rocks, and both have a similar NE-SW or ENE-WSW dominant structural grain. However, the prospective shale resource in the Bowland Basin consists of thick deposits of shale deposited in a deep marine setting (Andrews, 2013), whilst the shales in the Midland Valley were deposited in a lacustrine, fluvio-deltaic and shallow marine depositional environment and are intercalated in a stacked sequence with numerous relatively thin shales (Monaghan, 2014). Many of the US gas shales were also deposited in a deep marine setting, e.g. the shales in the Fort Worth Basin were deposited in a deep foreland basin (Bruner and Smosna, 2011). In general, the UK shale targets are thicker than their American counterparts.
The faults that cut the Carboniferous strata in the Midland Valley are mapped at surface, whereas the faults that cut the shales in the Horn River Basin do not penetrate the surface and are hidden by an overlying clay-rich mudstone above the prospective organic-shale units. The faults that cut the Midland Valley and Bowland Basin are mostly normal, with a component of strike-slip and reverse, and were last active during the late-Carboniferous to early-Permian. Conversely, the Eola field in Oklahoma is dominated by transpressional strike-slip faulting and thrusts (Kilic and Tapp, 2014).
No major faults intersect the Midland Valley of Scotland, unlike the Horn River Basin, which is cut by the 200 km long Bovie Fault, which was last active during the Cretaceous. The faults which cut the Barnett Shale are mostly associated with the Oachita Thrust front or basement structures (Walper, 1982; Montgomery at al., 2005), but may also be associated with dissolution of the underlying karstic Ellenburger Group (Gale et al., 2007): there is no such thrust control to the structure of the Midland Valley nor a comparable underlying karstic unit.