Induced Seismicity Impacts of Unconventional Oil and Gas Development

Alan J. Krupnick and Isabel Echarte

This report was produced as part of The Community Impacts of Shale Gas and Oil Development, an RFF initiative.

JUNE 2017 Induced Seismicity Impacts of Unconventional Oil and Gas Development

Alan J. Krupnick and Isabel Echarte 

Contents Summary ...... 1 Community Risk-Benefit Matrix ...... 1 1. What Is Induced Seismicity? ...... 4 1.1. Why Has Induced Seismicity become a Problem Recently? ...... 5 1.2. What Differentiates Induced Seismicity from Natural Earthquakes? ...... 6 2. Literature Review ...... 7 3. Earthquake Events Analyzed in the Literature ...... 9 4. Causes of Induced Seismicity Related to Oil and Gas Operations ...... 9 4.1. Extraction ...... 9 4.2. Fracking ...... 9 4.3. Enhanced Oil Recovery ...... 11 4.4. Wastewater Disposal ...... 11 5. Hazards Associated with Induced Seismicity and Community Impacts ...... 11 5.1. Valuation Studies for Natural Earthquakes ...... 14 6. Factors that Make an Area Prone to Induced Seismicity ...... 14 6.1. Geology ...... 15 6.2. Injection Depth and Communication with Basement Rocks ...... 15 6.3. Difficulty in Linking Events to Activities ...... 16 7. Seismic Risk: Monitoring and Probabilistic Assessments ...... 18 8. Conclusion ...... 20 9. Appendix ...... 20 9.1. Methods for Mitigating the Probability of an Earthquake ...... 20 9.2. Study Span Chart: Seismicity Impacts ...... 21 References ...... 25

 Krupnick: RFF senior fellow and co-director, RFF Center for Energy and Climate Economics; Echarte: RFF research assistant. This literature review was produced as part of The Community Impacts of Shale Gas and Oil Development, an RFF initiative. The authors would like to acknowledge very helpful reviews by Mark Zoback of Stanford University and Daniel Raimi of Resources for the Future. The authors would also like to thank the Alfred P. Sloan Foundation for funding this effort. Resources for the Future (RFF) is an independent, nonpartisan organization that conducts rigorous economic research and analysis to help leaders make better decisions and craft smarter policies about natural resources and the environment. © 2017 Resources for the Future (RFF). All rights reserved. No portion of this report may be reproduced without permission of the authors. Unless otherwise stated, interpretations and conclusions in RFF publications are those of the authors. RFF does not take institutional positions.

Resources for the Future | Krupnick and Echarte

Summary studies establishing methods to predict This report provides an overview of the the probability of future induced seismic existing state of research on induced events and a third proposing monitoring seismicity related to both unconventional and methods. conventional oil and gas development in the • Three studies analyzing oil- and gas- United States. We provide background on this related induced seismicity assess above- issue, a summary of the most relevant ground impacts, including on the takeaways from each study, and judgments Oklahoma housing market, the about the state of this body of literature. The acceptability of earthquakes (via a topics we discuss include the causes of survey), and shaking intensity. induced seismicity, the hazards associated • Though the literature specifically with induced seismicity, what makes an area addressing above-ground impacts is more likely to experience induced seismicity, sparse, given the robust literature and monitoring and mitigation potential. We connecting seismicity to oil and gas also review several economic studies activities as well as the robust literature assessing the valuation of earthquake risk in on natural earthquakes, we feel the residential housing market, with one study comfortable saying there is an increase in focused on induced seismicity specifically, phsyical damage related to the increase in and one survey on the acceptability of shaking with higher magnitdue seismicity that discusses induced seismicity. earthquakes. The literature connecting oil We additionally critique the current breadth of and gas activities to seismicity is large the literature using a “span chart,” which and of high quality, and as large summarizes the literature along a key magnitude earthquakes have been linked dimension: how many elements of the damage to these activities through the literature, function model (which links oil and gas we are able to infer that there is an activities to burdens, concentrations, increase in damage to buildings based on exposures, impacts, and monetary value) are the relationship between shaking covered by a given study. intensity and magnitude. • In all, our review looks at the findings of • Almost none of the literature, however, 45 studies. The large majority of studies addresses how induced seismicity related are retrospective, concerned with to oil and gas development affects the establishing association with oil and gas stress of residents experiencing this activities for specific events or changes in seismicity. One survey begins to address the rate of seismic events regionally. This this issue by touching on the acceptability literature finds that the change in the rate of induced seismicity, but the survey is of seismic events in the central and not directed at those experiencing this eastern United States is largely seismicity. One study reports findings on attributable to oil and gas operations, property value impacts related to induced namely wastewater injection. A number seismicity. of seismic events in Texas, Oklahoma, Ohio, and other areas that have been Community Risk-Benefit Matrix studied have been attributed to oil and The Community Risk-Benefit Matrix gas operations. identifies specific areas of concern related to • Only recently have studies attempted impacts addressed by the team’s literature more forward-looking analyses, with two review (left column of the matrix), as well as

www.rff.org | 1 Resources for the Future | Krupnick and Echarte impacts for which RFF experts have Lastly, we summarize the findings conducted original research and analysis. (See reported by the literature for each impact— page 3 for the section of the matrix related to whether the studies as a whole report this review, on the induced seismicity impacts increases, decreases, or no relationship of unconventional oil and gas development.) between the impact and an increase in The matrix indicates the quality of the unconventional oil and gas development. The literature for each impact, judged subjectively “heterogeneous” classification indicates that with the color indicating whether we find the the literature reports different outcomes across studies analyzing an impact to be, on average, areas. The “inconsistent” classification of a certain quality. Impacts may be assessed indicates that the literature reports by multiple low-quality studies and a medium- contradictory results (i.e., two studies find an quality study, for example, and we would increase or decrease for a certain impact in the consider this body of literature to be low same context). quality. A high-quality classification indicates View or download the entire matrix, that we trust the results of such studies, including all sections that correspond with including the accuracy, magnitude, and each of the literature reviews by topic direction of the results—meaning, in a produced as part of this initiative: practical sense, that it has no serious or fatal WHIMBY (What’s Happening in My flaws (such as inadequate methodologies) that Backyard?): A Community Risk-Benefit would lead us to question the results. A study Matrix of Unconventional Gas and Oil is considered low quality if we believe we Development cannot trust the results because the study has multiple, serious flaws (e.g., methodology, data, focus, or study design are inadequate to reliably estimate outcomes). A study is considered medium quality if it does not fit in the other two categories. A study is therefore medium quality if it has any such major flaw or if either the methodology, data, focus, or study design lead to questionable results for a number of reasons. Generally, we find the magnitude and direction of these results to be informative, but question the precision.

www.rff.org | 2 Resources for the Future | Krupnick and Echarte

COMMUNITY RISK-BENEFIT MATRIX LITERATURE REVIEW: INDUCED SEISMICITY IMPACTS OF UNCONVENTIONAL OIL AND GAS DEVELOPMENT

KEY

Higher quality: The majority of studies reviewed for an impact are of higher quality. Where there is one study of higher quality, it is marked as such.

Medium quality: The majority of studies reviewed for an impact are of medium quality. Where there is one study of medium quality, it is marked as such.

Lower quality: The majority of studies reviewed for an impact are of lower quality. Where there is one study of lower quality, it is marked as such.

Not reviewed: Research on an impact was not reviewed.

Increase: Studies show a positive, robust ↑ association with an impact (an increase in incidence or magnitude).

Decrease: Studies show a negative, ↓ robust association with an impact (a decrease in incidence or magnitude).

Heterogeneous: Across regions or areas, ↑↓ studies report robust results that differ.

No association: Studies report results that showed no association.

Inconsistent: Studies report differing ~ (contradictory) results.

www.rff.org | 3 Resources for the Future | Krupnick and Echarte

1. What Is Induced Seismicity? existing faults and fractures due to stress Since 2009, the central and eastern United changes from human activities, such as deep States have seen a dramatic increase in the wastewater injection, carbon storage, rate of earthquakes—the region averaged 24 geothermal activities, enhanced oil and gas earthquakes of moment magnitude 3 or larger recovery, surface changes, and sometimes 2 hydraulic fracturing (TNO 2014). Though (M3+)1 per year from 1973 to 2008. The region now averages 318 per year, with 1,010 these types of earthquakes have largely earthquakes occurring in 2015 alone occurred in the past several years, induced (Fitzpatrick and Petersen 2016). Oklahoma seismicity has been known to occur since the surpassed California, historically the most middle of the twentieth century. Some of the seismically active region in the United States, earliest documented cases that have been in the number of M3+ earthquakes per year in studied include the Rocky Mountain Arsenal (RMA; a deep injection well for hazardous 2014 (USGS n.d. a). And in 2015, north- 3 central Oklahoma saw an almost 900-fold chemicals that caused an M4.8 event in 1967 near Denver) and the US Geological Survey’s increase in its rate of M3+ earthquakes from that of before 2009 (Langenbruch and Zoback Rangely, Colorado, fluid injection and 2016). seismicity control experiments that took place from 1969 to 1973 (Raleigh, Healy, and This sudden change in seismic hazard can Bredehoeft 1976; Healy et al. 1968; Hermann, largely be attributed to induced seismicity— Park, and Wang 1981). earthquakes caused by the reactivation of pre- Most induced earthquakes in the central United States are caused by the deep injection

1 of briny wastewater from oil and gas Earthquake magnitudes (denoted by “M” in this production into underground formations for document) refer to moment magnitude, or MW, which measures the size of an earthquake in terms of the disposal. Induced earthquakes related to the energy released. This is a new measure that “gives the oil and gas industry can also be caused by most reliable estimate of earthquake size” (USGS n.d. enhanced oil and gas recovery (in which b). “Moment is a physical quantity proportional to the liquids and/or gases are injected into oil- slip on the fault times the area of the fault surface that slips; it is related to the total energy released in the bearing formations to stimulate production). [earthquake]. The moment can be estimated from And less commonly in the United States, seismograms (and also from geodetic measurements). though more prevalent in Canada, hydraulic The moment is then converted into a number similar to fracturing can also cause induced seismicity other earthquake magnitudes by a standard formula” (British Columbia Oil and Gas Commission (USGS n.d. b). ML, or “local magnitude,” is best known as the Richter scale and is also used in a few studies 2012; Rubinstein and Mahani 2015; Atkinson discussed in this review. This measure is valid for only et al. 2016; Bao and Eaton 2016). certain frequency and distance ranges, so other Underground injection for wastewater measures, such as M , were developed (USGS n.d. b). W disposal and enhanced oil recovery is Both MW and ML are logarithmic in scale. For a comparison of magnitude and shaking intensity above- ground, see USGS.

2 The central and eastern US is analyzed in the US 3 An earthquake of this size is generally felt by all with Geological Survey’s seismic hazard forecast that some disturbances, such as broken windows or moved assesses new seismic hazards due to seismicity as well furniture. See the US Geological Survey for more natural earthquakes in a region. information on the relationship between magnitude and intensity.

www.rff.org | 4 Resources for the Future | Krupnick and Echarte regulated under the US Environmental a byproduct of these activities has been so Protection Agency’s (EPA’s) Underground voluminous and so briny that, due in part to Injection Control (UIC) program, which market factors, it has been difficult to put it to covers solution mining wells, industrial and another use or recycle it for oil and gas municipal waste disposal wells, hazardous and operations. The wastewater must be treated or radioactive waste wells, and more. The directly disposed, and underground injection injection wells related to oil and gas provides a method for doing so in a way that operations are known as Class II wells, which is both economical and relatively safe (aside include wells for enhanced oil recovery from induced seismicity). Operators in activities, wastewater disposal, and storage of Oklahoma are currently exploring options to hydrocarbons. transport some wastewater to locations beyond Most Class II injection wells, in fact, are where it is produced. not associated with induced seismicity. In the These new unconventional oil and gas United States, 35,000 wastewater disposal production techniques have allowed resources wells and 80,000 enhanced oil recovery wells in central and north-central Oklahoma are currently active—yet only a handful of specifically (where the largest increase in these wells cause earthquakes felt above earthquakes has occurred) to be redeveloped, ground (Rubinstein and Mahani 2015). and have led to a large increase in produced Induced earthquakes generally have lower water volumes. Murray (2016) presents maximum magnitudes than tectonic estimates of the median water-to-oil and earthquakes; to date, the largest potentially water-to-gas ratios of several plays in induced earthquake was recorded in Pawnee, Oklahoma. The water-to-gas ratios range from Oklahoma on September 3, 2016, of 3.02 in the Missourian to 17.73 in the magnitude 5.8 (USGS 2016a).4 Most induced Mississippian, and the water-to-oil ratios events, however, are small in magnitude and range from 1.63 in the Missourian to 12 in the short in duration (GWPC and IOCC 2015). Mississippian. Ohio, by contrast, has a very low water-to-oil ratio, of 1 (as of 2012). 1.1. Why Has Induced Seismicity There, the injected water is often spent become a Problem Recently? hydraulic fracturing fluid (the liquid and sand The sudden increase in induced seismicity that is injected into a well to stimulate oil or beginning around 2009 is associated with the gas production), as opposed to produced water increased use of unconventional oil and gas (Horton 2012; Kim 2013; Rubinstein and production methods, including the drilling of Mahani 2015; Veil 2015). However, a large horizontal wells, development of shale share of water injected in Ohio is from New formations, and dewatering (which made York, West Virginia, and Pennsylvania (CSIS economical the production of oil and gas from 2017). formations that hold significant amounts of water) (Murray 2015). The water produced as

4 We use the term “potentially” here because we did not find studies assessing the relationship between this event and oil and gas activities. Establishing a causative link is a complicated process discussed further below.

www.rff.org | 5 Resources for the Future | Krupnick and Echarte

FIGURE 1. TECTONIC AND INDUCED EARTHQUAKES (M≥3) IN OKLAHOMA (1979 TO SEPTEMBER 2016)

Source: Langenbruch and Zoback 2016. Notes: The cumulative number of earthquakes is presented in linear (A) and logarithmic scales (B). The map shows the epicenters of earthquakes in Oklahoma.

The situation is particularly extreme in 1.2. What Differentiates Induced Oklahoma and Texas, where, respectively, Seismicity from Natural Earthquakes? 1.087 billion and 1.92 billion barrels per year Although distinguishing induced from were injected in 2012, compared to an average natural seismicity can be complicated, there of 135 million barrels per year in the 29 other are a few tell-tale characteristics of induced states reported by Veil (2015). Notably, events. First, induced earthquakes typically however, in North Dakota (another boom occur at shallower depths than natural state), more gas and oil than water was earthquakes. The former generally occur in the produced in 2012, meaning the state had top 6 kilometers (km) of the Earth’s crust in water-to-oil and water-to-gas ratios of less Oklahoma, for example, while the latter have than 1. New York oil and Alaska gas likewise maximum depths of 25 to 30 km (GWPC and had ratios less than 1 in 2012. Arkansas and IOGCC 2015). Llenos and Michael (2013), Wyoming, also large oil- and gas-producing analyzing rate changes in seismicity in states, both have some of the highest water-to- Arkansas and Oklahoma, found a few key oil ratios, at 26.6 and 26.2, respectively, differences between natural and induced though they injected a fraction of what earthquakes as well. Earthquakes that occur Oklahoma and Texas injected in 2012 (Veil near “active injection wells occur closer 2015). together than those that occur before, after, or far from active injection,” with higher aftershock productivity (Llenos and Michael

www.rff.org | 6 Resources for the Future | Krupnick and Echarte

2013, 2850). The study presents a robust geomechanical modeling and two studies statistical method for distinguishing seismic using probabilistic models to analyze induced rate changes caused by natural forces and seismicity. A handful of studies address changes from induced seismicity. The study above-ground impacts. This section will found that significant changes in the provide an overview of the larger focus and underlying triggering process occurred, topics covered in these studies, with analysis meaning changes in the background rate of of the results discussed throughout later earthquakes as well as aftershock productivity sections of the report. may aid in distinguishing anthropogenic rate Assessing these more general patterns of changes from natural rate changes. seismicity and rate changes proves difficult Additionally, an earthquake swarm (a because historic records of small events are series of earthquakes occurring in a relatively often incomplete, as “most of the sensitive short amount of time) propagating away from systems needed to monitor such events have an injection well over time, for example, been deployed during only the last few provides evidence that the swarm is induced, decades” (GWPC and IOCC 2015, 19). as the pore pressure front from the injection Furthermore, the wide spacing of seismic advances over time (Kim 2013). Such stations leads to uncertainty in locating the hypotheses can be tested using estimates of epicenters and depths of earthquakes, and pressure changes and fluid migration. Van der local networks are often deployed following Elst et al. (2016), however, did find that the larger events (GWPC and IOCC 2015). limits on the maximum magnitudes for Studies that assess these regional or general induced earthquakes are as large as patterns include Keranen et al. (2014), Gono statistically expected, suggesting that induced (2015), Rutqvist et al. (2013), Zhang et al. earthquakes have a tectonic (as opposed to (2013), Frohlich et al. (2016), Hornbach et al. injection-related) control on magnitude. The (2016), Weingarten et al (2015), Llenos and authors state that “induced earthquake Michael (2013), Frohlich (2012), Rubinstein magnitudes should be treated with the same et al. (2014), Walsh and Zoback (2015), maximum magnitude bound that is currently Goebel et al. (2015), Van der Elst et al. used to treat seismic hazard from tectonic (2013), and Van der Elst et al. (2016). earthquakes” (Van der Elst et al. 2016, 4575). Evaluating specific seismic events is more complicated, as it is difficult to “clearly and 2. Literature Review uniquely differentiate between induced and Much of the literature surrounding induced tectonic earthquakes using long-established seismicity focuses primarily on evaluating seismological methods” (GWPC and IOCC patterns and trends of seismicity or attempting 2015, 2). The use of other disciplines, such as to establish causation for specific events— statistics, are necessary (GWPC and IOCC both with a focus on proving that either 2015). Many studies use spatial and temporal changes in regional seismicity or specific correlation to evaluate causality, analyzing earthquake events were due to human activity whether an event occurred near an injection (oil and gas operations in this case). Both well or within a certain amount of time levels of analysis aid in evaluating whether following the start of injection or injection rate seismicity is linked to injection as well as changes, and so on. Such analysis, however, is assisting in the development of mitigation and unable to address factors such as underlying response strategies (GWPC and IOCC 2015). stress, the location of preexisting faults, and Furthermore, several studies use other geomechanical factors (GPWC and

www.rff.org | 7 Resources for the Future | Krupnick and Echarte

IOCC 2015). “Evaluating causation can be a Gono (2015); Rutqvist et al. (2013); Zhang et significant and time-intensive process, al. (2013); Dieterich, Richards-Dinger, and entailing locating the seismic event(s) Kroll (2015); McClure and Horne (2011); and accurately, locating critically stressed faults Hornbach et al. (2015). A better understanding that may have been reactivated, identifying the of these mechanisms is important to inform detailed temporal and spatial evolution of improved regulations to mitigate seismic risk. seismic events where fault slip first occurs and Modeling can also be used for more of any associated aftershocks, characterizing forward-looking studies that assess the the subsurface stress near and on the fault, and probability of seismic risk and perhaps aid in developing a physical geomechanics/reservoir moving towards prediction of this risk. Walsh engineering model that would evaluate and Zoback (2016) model the potential for whether an induced subsurface pore pressure fault slip on mapped faults in two regions in change could initiate an earthquake” (GWPC Oklahoma, while Langenbruch and Zoback and IOCC 2015, 22). Studies that use such (2016) present a statistical model to assess the methods to evaluate the relationship between probability of seismicity in a large area in oil and gas activities and specific seismic Oklahoma. The first study assesses the events include Justinic et al. (2013); Gan and probability of slip on known faults; the second Frohlich (2013); Hornbach et al. (2015); works to assess future probabilities of Shirzaei et al. (2016); McNamara et al. seismicity associated with injection rates (2015); Kim (2013); Keranen et al. (2013); regionally (including on unknown faults). Sumy et al. (2014); Horton (2012); Yeck et al. (2016); Frohlich et al. (2011); Frohlich and Field testing—in which experiments are Brunt (2013); Block et al. (2014); Ake et al. conducted with injection wells to test (2005); Skoumal, Brudzinski, and Currie hypotheses regarding the interaction of (2015); Holland (2011); Holland (2013); and injection parameters and seismicity—is Friberg, Besana-Ostman, and Dricker (2014). another area in which the literature remains Although these studies all address causation in limited, especially as this relates to oil and gas some way, the rigor with which they are able activities. The early experiments in Rangeley, to do so varies. Colorado, tested hypotheses in an area that had experienced seismicity related to Modeling can additionally aid in analyzing enhanced oil recovery (Raleigh, Healy, and geomechanical factors—such as pore pressure Bredehoeft 1976). Ake et al. (2005) described diffusion, “microseismicity” (essentially akin the results of experiments for brine injection at to the energy released by dropping a gallon of 5 Paradox Valley, Colorado, though this milk on the floor), fluid migration, and other injection is not related to oil and gas activities. hydrogeologic conditions—which allows for a more precise identification of the mechanism Few studies, however, address the above- through which injection may cause seismicity. ground impacts of induced seismicity— Studies focusing on both general patterns and including structural damage, human impacts specific events include Keranen et al. (2014); (such as anxiety), or measured shaking. Hough (2014), for example, compared the ground motions caused by induced seismicity 5 The energy released by dropping a gallon of milk is and the shaking caused by tectonic comparable to an M2.0 earthquake, which is similar to earthquakes. the microseismicity associated with hydraulic fracturing. Even fewer studies address the economic impacts and anxiety communities face due to

www.rff.org | 8 Resources for the Future | Krupnick and Echarte induced seismicity. Liu, Ferreira, and Brewer cause of the earthquake and epicenter are the (2016) provide the only study attempting to smallest for hydraulic fracturing (ranging from address these effects, analyzing the housing directly below the wells to 2 km away from market impacts of induced seismicity from the the wells), and the greatest distances (9.65 km 2011 earthquake sequence in Prague, to greater than 10 km) appear to occur when Oklahoma. McComas et al. (2016) multiple injection wells are suspected to be the additionally provide one of the only studies cause of seismicity, based on our review and analyzing perceptions of induced seismicity summarized in Table 1. for energy development—though some studies do analyze perceptions and impacts of induced 4. Causes of Induced Seismicity seismicity caused by other activities, such as Related to Oil and Gas Operations geothermal energy and carbon capture and The majority of induced seismicity events storage. Due to the largely narrow focus of the discussed in this literature review are caused majority of this literature, little is known about by deep underground injection of wastewater. the above-ground impacts of induced There are, however, examples of other oil and seismicity. gas operations that have caused induced seismicity, though they are less common or of 3. Earthquake Events Analyzed in smaller magnitudes than induced seismicity the Literature from wastewater injection in the United Table 1 on the following page lists the States. strongest earthquakes analyzed by the studies covered in this review and their suspected 4.1. Extraction causes. Some studies examined swarms of Some studies showed that extraction earthquakes whereas others only analyzed one (production) wells may also be correlated with event. It is important to note that not all of the seismicity—such as Frohlich and Brunt studies analyze earthquake events or a series (2013), who studied injection and production of earthquakes, such as modeling studies or wells in the Eagle Ford shale play—though those assessing above-ground impacts. This modeling of the subsurface hydrology and table therefore does not include all of the stress would be necessary to address causality. studies discussed throughout this review, and Hornbach et al. (2015) also found that a is additionally not a comprehensive list of all combination of both brine production and earthquakes induced by oil and gas activity. wastewater disposal likely caused the The magnitudes are reported as moment November 2013 earthquakes in Azle, Texas. magnitudes (see p. 4 for a description) unless Though there is evidence for induced otherwise specified. seismicity associated with production, this is The strongest earthquake analyzed in the not the main reason seismicity has increased literature covered in this review is the 2011 in the central United States. Prague, Oklahoma, earthquake, at M5.7, studied by Sumy et al. (2014) and Keranen et 4.2. Fracking al. (2013)—though this quake was actually While hydraulic fracturing is known to triggered by the main shock, which was M5.0. cause microseismicity (see definition on p. 8), The smallest magnitude earthquakes analyzed it sometimes also causes “felt” seismicity in these studies are the result of hydraulic (Rutqvist et al. 2013; Zoback 2016a). Induced fracturing (as opposed to wastewater seismicity felt above ground associated with injection). The distances estimated from the hydraulic fracturing has occurred in the

www.rff.org | 9 Resources for the Future | Krupnick and Echarte

United States, though it is less common than hydraulic fracturing development has been injection-related seismicity depending on the focused, Atkinson et al. (2016) found that only region. Induced seismicity from fracturing is about 0.3 percent of fracking operations are more prevalent in Alberta and British associated with seismicity—though with Columbia in Canada (British Columbia Oil thousands of wells drilled annually, the issue and Gas Commission 2012; Bao and Eaton is still significant in the region. 2016; Atkinson et al. 2016). Even in the western Canadian sedimentary basin, where

TABLE 1. LARGEST MAGNITUDE EARTHQUAKES ANALYZED IN EACH STUDY Study Magnitude Estimated Cause Distance to Date Location (MW)* cause Shirzaei et al. (2016) 4.8 Two wastewater injection wells Within ~10 May 2012 Eastern Texas km McNamara et al. (2015) 4.3 Wastewater injection from 3 wells, Within 9.65 Oct. 2014 Cushing, OK one injecting into the Arbuckle km formation, above the basement§ Kim (2013) 3.9 Deep wastewater injection well ~1 km Dec. 2011 Youngstown, OH Horton (2012) 4.7 Multiple wastewater injection Within 6 km Feb. 2011 Central AR wells Yeck et al. (2016) 5.1 Far-field pressure from clustered, > 12 km Feb. 2016 Fairview, OK high-rate injection wells Keranen et al. (2013) 4.8, 5.0, 5.0 induced by wastewater Within 1.5 Nov. 2011 Prague, OK and 5.7 injection; 4.8 and 5.7 triggered by km stress transfer Sumy et al. (2014) 4.8, 5.0, 5.0 foreshock, induced by fluid Within 1.5 Nov. 2011 Prague, OK and 5.7 injection, triggered cascading km failure of earthquakes Frohlich et al. (2011) 3.3 Wastewater injection well 0.5 km May 2009 Dallas-Fort Worth, TX Rubinstein et al. (2014) 5.3 Wastewater injection from two 2 km Aug. 2011 Raton Basin, CO– nearby wells NM border † Justinic et al. (2013) 2.8 MbLg Two fluid injection wells, one high- Within 3.2 June 2009 Cleburne, TX rate km Hornbach et al. (2015) 3.6 Combination of brine production Within 10 Nov. 2013 Azle, TX and wastewater injection km

Block et al. (2014) 4.4 ML Single, long-term brine injection 8.2 km Jan. 2013 Paradox, CO well Ake et al. (2005) 4.3 Single, long-term brine injection Within 8 km May 2000 Paradox, CO well Friberg, Besana-Ostman, 2.2 Hydraulic fracturing near a fault Directly Oct. 2013 Harrison County, and Dricker (2014) below wells OH

Skoumal, Brudzinski, and 3 ML Hydraulic fracturing ~850 m March Poland Township, Currie (2015) 2014 OH Holland (2011; 2013) 2.9 Hydraulic fracturing 2 km Jan. 2011 South-Central, OK

Frohlich and Brunt (2013) 4.8 Oil/water extraction‡ Within 5 km Oct. 2011 Eagle Ford Shale, TX Notes: Most studies look at series (or swarms) of earthquakes occurring over a short period or even years. The magnitude listed is that of the maximum moment magnitude analyzed by the study. †MbLg refers to the “short-period body-wave magnitude…used in the tectonically ‘stable’ eastern part of North America. … This magnitude is measured from peak motions.” (McCalpin 2009). ‡The study also reviewed other earthquakes occurring in a two-year survey and found those to be associated with fluid injection as well, though this specific earthquake was estimated to be caused mainly by extraction. §Basement rock is the rock layer below which sedimentary rocks are not found.

www.rff.org | 10 Resources for the Future | Krupnick and Echarte

Holland (2011; 2013); Friberg, Besana- 4.4. Wastewater Disposal Ostman, and Dricker (2014); and Skoumal, Wastewater disposal is the source of the Brudzinski, and Currie (2015) found evidence majority of induced seismicity discussed in of earthquakes induced by hydraulic fracturing this review. The water being disposed is in, respectively, south-central Oklahoma; primarily salt water produced with oil and gas; Harrison County, Ohio; and Poland Township, the fluid used in the hydraulic fracturing Ohio. The January 2011 south-central process is generally a small fraction of the Oklahoma sequence caused a range of ML 0.6 produced water, though this fraction varies by to 2.9 earthquakes, with the preferred formation (EPA n.d.). Any fluids returning to hypothesis being that the fluid pressure the surface may be injected into disposal diffused through a more permeable fault or wells. The majority of disposal wells are in fracture system to a critically stressed fault Texas, California, Oklahoma, and Kansas. (Holland 2013; see page 4 for a definition of Wastewater disposal wells make up about 20 ML). The October 2013 Harrison County percent of Class II injection wells (EPA n.d.). earthquakes showed no correlations with maximum injection pressure or fluid volume, 5. Hazards Associated with Induced and were likely caused by hydraulic linkage to Seismicity and Community Impacts a fault (Friberg, Besana-Ostman, and Dricker 2014). The Poland Township earthquakes, The main impact of induced seismicity is ground shaking, which can cause damage to ranging from ML ~1 to 3, may have been caused by elevated pressures from hydraulic buildings, as well as human stress (GWPC and fracture fluids or pressurized formation waters IOCC 2015). Structural damage to modern that entered a fault or fracture system, either buildings generally occurs only if earthquakes from intersecting a well borehole or from the are greater than M5.0 (GWPC and IOCC fractures normally caused during fracking well 2015), and few induced earthquakes have been stimulation (Skoumal, Brudzinski, and Currie above this magnitude. Several induced 2015). earthquakes have caused documented damage—including, for example, the M5.7 4.3. Enhanced Oil Recovery earthquake of 2011 in Prague, Oklahoma; the M5.3 earthquake of 2011 in Trinidad, Enhanced oil recovery is a process by Colorado; and the M4.8 earthquake of 2012 in which water, steam, or carbon dioxide are Timpson, Texas (GWPC and IOCC 2015). injected into an oil- and gas-producing formation through a Class II injection well in Oklahoma also experienced several large order to stimulate production. These earthquakes in 2016, some in the span of a operations generally keep fluid pressure in few weeks. The M5.3 earthquake that reservoirs about equal (as they are used in the occurred on November 7, 2016, in Cushing, extraction reservoir). Enhanced oil recovery Oklahoma caused “substantial” damage, with wells make up about 80 percent of Class II 40 to 50 buildings damaged with cracks or injection wells (EPA n.d.). While earthquakes fallen bricks and facades (Associated Press induced by enhanced oil recovery have 2016), and the M5.8 earthquake of September occurred—as discussed by Nicholson and 3, 2016, in Pawnee also caused damage (Liou Wesson (1992); Gibbs et al. (1973); Raleigh, 2016). The M5.7 earthquake of 2011 in Healy, and Bredehoeft (1976); and Gan and Prague destroyed over a dozen homes and Frohlich (2013)—the discussion of enhanced injured two people (USGS 2016b). Though oil recovery is outside of the scope of this the latter two earthquakes have not yet been exercise. proven to be induced, as they occurred just

www.rff.org | 11 Resources for the Future | Krupnick and Echarte weeks before the writing of this document, the intensities were generally lower than those Oklahoma Corporation Commission’s Oil and predicted for tectonic earthquakes. Gas Division announced the shut-in of dozens Aside from news reports and anecdotal of wells in both areas as emergency measures evidence, how communities are actually (OCC 2016a; OCC 2016b). affected by induced seismicity is largely In its June 2016 seismic hazard update unaddressed in the academic literature. Liu, (which now includes induced seismic risk), Ferreira, and Brewer (2016), a dissertation, the US Geological Survey updated the chance provide the only attempt (that this review of having a Modified Mercalli Intensity found) to assess the economic impacts of (MMI) VI or greater in north-central induced seismicity in Oklahoma through an Oklahoma to 5 percent to 12 percent, similar analysis of effects on the housing market. This to the chances of California’s damaging approach has a history in the economics earthquakes with some areas with induced literature, whereby housing prices have been earthquakes seeing a hazard that is higher than analyzed with respect to shale gas the 2014 estimate by more than a factor of 3 development, such as Muehlenbachs et al. (Petersen et al. 2016). The MMI index is a (2015). This approach has also been used to measure of above-ground shaking intensity look at natural disasters, such as flooding, and is rated from I to X. An MMI of VI or where it has been found that homes within a greater indicates an earthquake that is strong floodplain experience a drop in market value, and felt by all, with slight damage and heavy as the event acts as an update to resident’s and furniture moved. prospective home buyers’ views about flood Figure 2 on the following page uses these risk (Liu, Ferreira, and Brewer 2016). MMI values to create a map of the potential Analyzing home sales in eight Oklahoma for damage from an earthquake in 2016. The counties between 2000 and 2015, Liu, analysis for the central and eastern United Ferreira, and Brewer (2016) looked at whether States includes both induced and natural the 2011 earthquake in Prague, Oklahoma, earthquakes; on the right side of the map, and also the cumulative intensity of some of the coloring could be due to tectonic earthquakes prior to a housing transaction activity (particularly in eastern Arkansas). The affected home values. The study estimates the analysis for the western states in this map is impact of a home’s proximity to production based on the presumption that the earthquakes and injection wells on home values with a are naturally occurring. hedonic price function—estimating a How or if ground shaking differs between consumer’s willingness to pay for a reduction induced and tectonic earthquakes, however, is in the probability of earthquakes. The study, less clear. Hough (2014), using data from the when analyzing both production and injection US Geological Survey’s “Did You Feel It?” wells, found no effect of injection wells on housing prices. For sales of homes dependent system,6 found that within 10 km of the on groundwater following the Prague epicenter, shaking intensities were similar to expected magnitudes for tectonic earthquakes. earthquake, additional production wells within Outside of this range, however, shaking an area had lower housing prices, with the effect decreasing at greater distances.

6 https://earthquake.usgs.gov/data/dyfi/

www.rff.org | 12 Resources for the Future | Krupnick and Echarte

FIGURE 2. CHANCE OF DAMAGE FROM AN EARTHQUAKE IN 2016

Source: Petersen et al. 2016.

For groundwater-dependent homes within 2 McComas et al. (2016) conducted a web km of a production well, housing prices were survey of attitudes toward induced seismicity 0.95 percent lower following the earthquake; from energy production of 325 US adults (a within 10 km, this effect fell to 0.05 percent. small sample size), who were asked whether This result, however, was not found to be they felt negatively or positively about significant within 1 km of a production well, earthquakes they felt but caused no damage. which may reflect income from royalties Participants had more negative feelings about mitigating the decrease in home prices. The induced earthquakes than natural ones, and study, however, has a number of attributing the benefits to renewable energy or counterintuitive findings. One such finding is climate change mitigation did not make that after the Prague earthquake, the number induced earthquakes more acceptable. Public of production wells actually increased the engagement was the only option that made value of a home between 0.05 percent and induced earthquakes more acceptable: when 0.56 percent for homes not dependent on people believed their peers had a voice in the groundwater (and therefore with lower decision to implement the technology, groundwater contamination risk). Yet other acceptability of induced seismicity was results in the paper indicate that groundwater “significantly higher” (McComas et al. 2016, use increases the value of a home, particularly 27). The participants also generally felt neutral after the earthquake. More work could and about seismicity (an average of 5 on a scale of should be done to better estimate the impact of 0 to 10) if the energy production and its induced seismicity on housing values in revenue benefitted local schools and colleges Oklahoma and elsewhere. and also if it produced benefits shared widely by people across the planet.

www.rff.org | 13 Resources for the Future | Krupnick and Echarte

5.1. Valuation Studies for Natural Though it may sound surprising that a Earthquakes consumer’s willingness to pay for a reduction Though McComas et al. (2016) showed in the probability of property damage by that acceptability varies depending on the choosing a location that has a lower chance of cause of the earthquake (natural versus man- damage from earthquakes (or the amount they made), some of the valuation literature for must be compensated for an increase in natural earthquakes is relevant for illuminating risk)—falls following such an event, the the potential impacts of induced seismicity on authors find that the results are consistent with housing markets. the notion that consumers “initially overestimate the probability of damage from Murdoch, Singh, and Thayer (1993) unlikely events,” or that the hazards are analyzed the 1989 M7.1 Loma Prieta initially communicated poorly. earthquake in northern California and three Because an earthquake releases energy studies analyzed several earthquakes in Japan (Nakagawa, Saito, and Yamaga 2007; Naoi, that was building up, homeowners in a Special Seko, and Sumita 2009; Hidano, Hoshino, and Studies Zone might conclude that after a Sugiura 2015). These studies found evidence tectonic earthquake, a second one is not as that both increases in earthquake risk and likely, whereas homeowners in areas with actual earthquake events have a negative induced seismicity might, in fact, increase impact on property prices. Murdoch, Singh, their perceived future risk because potential and Thayer (1993) found a 2 percent decrease earthquakes there are caused by human in home values, around $5,800, following the activity that will continue in the future. In Loma Prieta earthquake. This study also finds Oklahoma and other areas experiencing that housing values are about $10,770, or 3.7 induced seismicity, therefore, residents may percent, lower for a house included in a potentially increase their willingness to pay Special Studies Zone, comparable to the for a reduction in the probability of property findings in Beron et al. (1997), discussed damage (by moving away from an earthquake- below. prone area). While the results of the valuation studies would most likely differ if conducted Beron et al. (1997), like Murdoch, Singh, in Oklahoma following induced seismicity, and Thayer (1993), assessed homeowners’ they are indicative of changes in risk willingness to pay for a reduction or be perception following earthquakes and the compensated for an increase in risk following communication of earthquake hazard. the Loma Prieta earthquake, which damaged or destroyed 20,000 homes. Homeowners’ 6. Factors that Make an Area Prone to expected loss from inclusion in a higher risk Induced Seismicity area, the Special Studies Zone—an area where Because all the potential options for surface faulting has occurred in the last 11,000 disposal of wastewater are problematic to years—fell from –$11,800 before the event to some extent (and underground injection might –$8,800 after the event. The expected lifetime actually be preferable in some areas), it is cumulative loss from earthquakes, an amount important to understand whether regions may assessed by homeowners based on what they be at risk of induced seismicity, preferably believe to be their assumed risk, fell from prior to the start of injection. The Bakken –$20,000 before the event to –$13,100 after region in North Dakota, for example, has not the event for the six-county Bay area. faced issues with induced seismicity, whereas Oklahoma has seen the majority of impacts in

www.rff.org | 14 Resources for the Future | Krupnick and Echarte the United States. There are a number of in Texas, Frohlich (2012) found that the wells factors that can make a region more prone to nearest earthquakes all reported a maximum induced seismicity. Many of these risk factors, monthly injection rate of more than 150,000 however, are often difficult to assess in the barrels of water per month. Not all wells with absence of earthquakes. In regions that have these injection rates, however, were located low levels of natural seismicity, the near earthquakes. The pressure changes underlying geology is not studied extensively. caused by these injected fluids must reach a In Oklahoma, Ohio, and Arkansas, induced suitably oriented (according to the stress field seismicity has led to the discovery of orientation)7 or critically stressed fault to previously unknown faults. Furthermore, areas cause seismicity. Hornbach et al. (2016) found where more fluids are injected will most likely that seismicity in Texas’s Dallas and Ellis be more prone to induced seismicity. As was Counties, despite not having any injection discuss above, oil is produced alongside large wells, is caused by injection in nearby amounts of brine in Oklahoma, Texas, and counties. Ellis and Dallas are “down-dip” Arkansas, requiring operators to inject more (located at the lower end of a sloping surface water in those regions than in other areas, such or subsurface layer) from the injection sites, as the Bakken region in North Dakota or so denser fluids will flow there. Furthermore, Ohio—where less wastewater is produced. the counties are bounded by a low- Other factors at play include the underlying permeability barrier that blocks this fluid from geology, injection depth, potential for escaping. Gono (2015), also studying injection hydraulic communication into basement rocks and seismicity in the Fort Worth Basin in with large faults (discussed below), and Texas, found spatial and temporal correlations temporal delays as well as spatial uncertainty. between seismicity and pore pressure increases, though not all areas with pore 6.1. Geology pressure increases experience seismic activity, Underlying geology plays a critical role in as the pore pressure increase must encounter a whether an activity will induce an earthquake. favorably oriented fault. The Arbuckle formation in Oklahoma, for example, may be in “hydraulic 6.2. Injection Depth and communication” with the underlying Communication with Basement Rocks basement, meaning pressure changes can Faults that are present in the crystalline propagate to the basement where faults exist basement (a very deep geologic formation (Walsh and Zoback 2015, 1). When many below disposal formations) are, for the most wells inject into the formation, increases in part, those that are of concern. Although faults pore pressure in the Arbuckle may spread away into the basement and eventually trigger seismicity on a critically stressed fault (Walsh 7 “Faults may be more prone to slip under certain stress and Zoback 2015). Injection wells in central conditions and geologic circumstances. In a given stress field, the ratio of shear stress to resisting strength on a Arkansas also faced similar issues when fault depends on the fault orientation. Resisting strength injecting into the Ozark aquifer, as no depends on the stress acting perpendicular to the fault formation exists to prevent fluid migration (i.e., the degree of clamping of the fault)” (GWPC and from the Ozark to the basement rock where IOCC 2015, 54). This means the same shear stress on faults are contained (Horton 2012). two different faults can lead to highly different probabilities of fault slip depending on the direction of In a two-year survey of seismicity and the fault. injection activities in the Barnett shale region

www.rff.org | 15 Resources for the Future | Krupnick and Echarte are present at many different depths, those that before felt seismicity occurred in Cleburne, reach into or are present in the basement are Texas, and Shirzaei et al. (2016) also found a large enough to cause earthquakes that can be five-year delay in Timpson, Texas. Frohlich felt and potentially damaging (Zoback 2016b). (2012) noted that, for the most part, injection Hydrogeologic communication between the wells in the Barnett shale region operated for a wastewater injection zone and the basement, year or more before the onset of earthquakes. where larger and favorably oriented faults Hornbach et al. (2016) found a delay of six may exist, is perhaps the main factor in years from the start of injection in neighboring whether injection will lead to induced counties to seismicity in Dallas and Ellis seismicity (Zhang et al. 2013, Shirzaei et al. Counties in Texas. As is evident from Figure 3 2016). In some areas, such as the Bakken below, the levels of wastewater injection shale region in North Dakota, there are increased more slowly in Oklahoma than did barriers that prevent this from occurring. In the number of earthquakes. Higher levels of Oklahoma, however, the disposal reservoir wastewater injection began years before the associated with the earthquakes—the highly rapid increase in the rate of earthquakes permeable Arbuckle formation—is in occurred—with injection rates doubling over communication with the underlying crystalline the past 17 years and seismic rates increasing basement, in which there are optimally beginning in 2009 (Walsh and Zoback 2015). oriented faults. Even when injection stops or slows, several studies found that the buildup of 6.3. Difficulty in Linking Events to pressure caused by fluid injection requires Activities time to return to pre-injection levels (Shirzaei There are a number of reasons why these et al. 2016; Langenbruch and Zoback 2016; risk factors may not become evident Horton 2012; National Research Council immediately as injection starts—including 2013). Horton (2012) found that following the temporal displacement, inducement at a shutdown of disposal wells associated with a distance, and triggering by other earthquakes. continuous swarm of M≤ 4.7 earthquakes in A lag in induced seismic events following the central Arkansas during 2010 and 2011, the start of injection is apparent in many of the swarm did not stop, though the rate and size of studies we reviewed. Some seismicity might earthquakes decreased in the months begin within months or may take decades to following. In modeling seismic rates in occur following the start of injection. Seismic Oklahoma, Langenbruch and Zoback (2016) risk therefore does not become immediately found that despite decreases in injection apparent. In a study of the 2011 Prague occurring due to both regulation and economic earthquake, for example, Keranen et al. (2013) forces, heightened levels of seismicity will reported that felt seismicity did not begin in persist for several years before returning to the area until 17 years following the start of normal (historic) rates. The study found that injection. The authors hypothesized that the probability of an earthquake exceeding a injection over that period of time could have certain magnitude annually returns to 2009 to refilled a compartment until a critical volume 2013 levels by 2025. A National Research injected was reached. Council (2013) study noted that despite the Justinic et al. (2013) likewise reported that fact that injection operations ceased at the earthquakes began years after the start of the Rocky Mountain Arsenal in 1966, earthquakes first injection well and six to eight months continued for years, until the late 1980s. following a doubling of injection volume

www.rff.org | 16 Resources for the Future | Krupnick and Echarte

FIGURE 3. CHART COMPARING CUMULATIVE M2.5+ EARTHQUAKES AND MONTHLY FLUID INJECTION IN OKLAHOMA

Source: Walsh and Zoback (2015).

Another reason for the difficulty in injection-related seismicity to occur up to 35 assessing risk is that although induced km away from disposal wells. seismicity often occurs in close proximity to Induced earthquakes—those where the injection wells or other sources of pressure majority of stress changes are caused by changes, some studies find that earthquakes human activities—can also trigger other can be induced up to 20 or even 35 km away earthquakes. Earthquakes are considered from the source. This means that one triggered when other factors (such as community conducting fluid injection underlying stress conditions) are responsible activities may impact another nearby for the majority of a stress change that caused community—the issue is therefore more an earthquake. Hornbach et al. (2016) regional than local. Yeck et al. (2016), observed seismicity both near injection wells studying the M5.1 earthquake in Fairview, in the Fort Worth basin and more than 10 km Oklahoma, for example, found that far-field away from injection wells in Dallas and Ellis pressurization from a group of closely located, Counties. For the 2011 Prague earthquakes, high-rate injection wells in long-term Keranen et al. (2013) found that the first operation located 12 to 20 km from the event, M5.0, was induced by fluid pressure, earthquake sequence caused a partially but the larger M5.7 earthquake (while it can unmapped 14 km fault to rupture. Using a 3D be considered a consequence of injection model of pore pressure diffusion for injection activities) was triggered by stress transfer rates from 89 wells within 50 km of the from the first event. With faults that are close earthquake swarm that began in 2008 in Jones, to failure, induced earthquake events can Oklahoma, Keranen et al. (2014) found therefore cause additional seismic activity.

www.rff.org | 17 Resources for the Future | Krupnick and Echarte

Sumy et al. (2014), conducting a slightly following seismicity, found that the M5.0 different study that focused on how the M5.0 foreshock triggered a “cascading failure of earthquake in Prague interacted with the earthquakes” along a fault system (1,904) M4.8 earthquake in Timpson, Texas, was 7. Seismic Risk: Monitoring and actually caused by deeper injection wells Probabilistic Assessments outside of this area of uplift. Surface uplift, Although most of the literature focuses on therefore, may not be indicative of potential establishing the causes of regional rate hazard. The authors found that these changes or specific seismic events, there is a techniques are useful for understanding small subset of recent seismic literature that changes in stress and pore pressure at depth. has implications for both regional and Although such analysis can aid in assessments localized risk assessments. Walsh and Zoback of seismic hazard, it is not clear from the (2016) used a probabilistic model to focus on results of this study that monitoring surface the potential for a known and mapped fault to deformation alone can assess risk. slip given changes in pore pressure, focusing on Oklahoma. Such a model is limited in its Two studies (Walsh and Zoback 2016; ability to predict actual risk, as many induced Langenbruch and Zoback 2016) assess the earthquakes occur on unmapped faults. probability of a future earthquake, though Langenbruch and Zoback (2016) were able to through different methods. Walsh and Zoback advance probabilistic assessments by (2016) built a model to focus on injection analyzing the probability of earthquakes in rates as one of the main determinants of Oklahoma given regional injection rates. This seismicity in Oklahoma and performed a model cannot predict the potential for quantitative risk assessment to evaluate the localized earthquakes (such as the potential conditional probability of fault slip for for a specific well to cause an earthquake), but mapped faults in north-central Oklahoma. The it can assess regional policies within results—a “cumulative distribution function of Oklahoma (such as the Oklahoma Corporation the pore pressure required to cause slip on Commission’s recent volume reduction each fault segment” (991)—can be used to plans8) to examine the effect of a policy assess the probability of slip on a known fault change on the state or larger regions of the from a given pore pressure increase caused by state. injection in order to avoid injection near a potentially active fault. The authors then One study, Shirzaei et al. (2016), analyzed mapped known faults with color coding the potential for remote sensing of surface where, in response to a 2 MPa pressure changes to indicate subsurface pressure perturbation, faults with <1 percent changes, focusing on a series of earthquakes conditional probability of slip in response are in Texas. The authors found a line-of-sight green, those with >33 percent chance are red, uplift of above 3 millimeters per year near one and those in between are yellow (see Figure 4 pair of injection sites, but found that the 2012 below). The results are said to be very conservative considering the assumptions made regarding geomechanical uncertainties, 8 See the Oklahoma Corporation Commission’s though they did find that the majority of “Earthquake Response Summary (as of November 19, 2016)” for more information. mapped faults are not likely to be activated by (http://www.occeweb.com/News/2016/11-23- modest pore pressure changes. 16EARTHQUAKE%20ACTION%20SUMMARY.pdf)

www.rff.org | 18 Resources for the Future | Krupnick and Echarte

FIGURE 4. FAULT MAP COLORED WITH CONDITIONAL PROBABILITY OF FAULT SLIP FROM PORE PRESSURE CHANGE

Source: Langenbruch and Zoback (2016). Notes: Green represents <1% of a probability of slip in response to 2 MPa pore pressure perturbation; red indicates a >33% change in response to the same pressure change.

The issue with this model is that a large earthquake decreased as injection rates number of induced earthquakes are not decrease, though the probability remained associated with mapped faults (Walsh and elevated for several years before returning to Zoback 2016). The study used Oklahoma for historic levels. During peak injection rates in its analysis, but the results produced by the 2015, the probably of an earthquake greater technique would likely be more accurate in than M5.0 was as high as 80 percent; other areas where pressure does not spread as following the decrease of injection rates due to quickly or as far as it does in Oklahoma. regulatory intervention and economic changes, Langenbruch and Zoback (2016) took this the model predicts this probability will fall to probabilistic analysis a step further—toward 37 percent in 2017. Although the model was establishing a model to find the probability of successful in predicting many of 2014’s large induced seismic hazard more generally, not earthquakes, it was not able to predict the just for specific known faults. The model M5.7 earthquake in Prague—meaning that, predicts the probability of exceeding an M5.0 according to the model, the earthquake could earthquake based on injection rates and have been caused by factors other than several other parameters in two large areas injection rates. The model also is not able to recently regulated by the Oklahoma predict localized seismicity, as the results are Corporation Commission. The authors found averaged over two large areas in Oklahoma, that their model was able to predict many of and does not rule out the potential for the large earthquakes of 2014 in Oklahoma, fluctuations in seismic hazards on the local and that the probability of exceeding a M5.0 level.

www.rff.org | 19 Resources for the Future | Krupnick and Echarte

A lack of seismic data impedes the policy change (such as rate reduction, volume possibility of more studies like Langenbruch reduction, or injection into specific subsurface and Zoback (2016), and, due to formations) will definitively prevent any undersampling, a lack of information about further earthquakes from occurring. However, existing geologic structures diminishes the certain regulations (such as traffic light usefulness of studies such as Walsh and systems, which we discuss in the Appendix) Zoback (2016). Almost all studies in this can mitigate the risk of induced seismicity, review echo the importance of quantifying particularly when combined with high-quality stress and pore pressure changes in the crust in monitoring. order to predict the occurrence of or an uptick in seismicity. Such data, however, remain 9. Appendix elusive for a number of reasons, such as a lack of observations or integration with models 9.1. Methods for Mitigating the (Shirzaei et al. 2016). Probability of an Earthquake There is no magic-bullet technique for 8. Conclusion preventing any induced events from occurring Overall, the literature focuses largely on with high-volume, deep-disposal wells in establishing that earthquakes or regional regions that have been shown to be prone to seismic rate changes are attributable to oil and induced seismicity, though several methods gas activities. Very few studies address above- have been implemented by jurisdictions to ground impacts (Hough 2014; Liu, Ferreira, mitigate risk (discussed further below). Here, and Brewer 2016; McComas et al. 2016), with we discuss what the findings in this literature only the latter two addressing how imply for the ability to mitigate induced communities and people might be affected by seismic risk. induced seismicity from energy development. Several studies have shown the potential Only two studies, Walsh and Zoback (2016) to manage the number of earthquakes that and Langenbruch and Zoback (2016), present occur, but the magnitude depends more on the probabilistic models that assess existing or underlying geology and tectonics, of which future seismic hazard as well as analyze the little is often known. Injection controls, such effectiveness of policy changes meant to as volume of fluid injected, injection rate, mitigate such hazard. Shirzaei et al. (2016) reservoir size, and other factors, do not appear also presented a model that can be used to to have an effect on the upper bounds of inform and update hazard assessments using induced earthquake magnitudes (Van der Elst surface deformation data. et al. 2016; Yeck et al. 2016). Although More studies, however, are needed injection controls can affect nucleation—the regarding the community impacts so that fact that an earthquake occurs—they cannot industry and regulatory responses can be determine magnitude. Van der Elst et al. better informed and more efficient. Further (2016) argued that “the probability of research in probabilistic or predictive analysis inducing an earthquake of a given size is of induced seismic risk would additionally directly related to the probability of prove beneficial in both informing regulators encountering—and triggering—an of risk and aiding in mitigating risk. However, appropriately stressed and tectonically as we discuss in the Appendix, it is difficult connected nucleation site” (4584). Yeck et al. for regulators to guarantee that an injection (2016) likewise emphasized “the role well will not cause seismicity or that a certain preexisting geologic structures play in induced earthquake sequences” (10,205). For hazard

www.rff.org | 20 Resources for the Future | Krupnick and Echarte assessments, fault length and orientation, context of areas of higher seismic risk. More among other factors, are important, though study regarding this topic is necessary before they may not be known until a fault is concluding that underground injection in such reactivated and presents seismic activity areas is less risky than injection in areas that (McNamara et al. 2015). have lower risk from natural earthquakes— Several studies suggest that lowering particularly as one study found that some early injection volumes could reduce the probability twentieth-century earthquakes in the Los of an earthquake (McNamara et al. 2015; Angeles area were probably induced by oil Weingarten et al. 2015; Langenbruch and production (Hough and Page 2016). Zoback 2016). Studying seismicity near Policy response to this uncertainty has Dallas-Fort Worth, Azle, and Cleburne, Texas, focused on what are known as traffic light Hornbach et al. (2016) found correlations with systems to mitigate risk for site-specific injection volumes but not injection rates Ake operations. Traffic light systems, such as those et al. (2005) found that reducing the rate of in Alberta, Ohio, and the United Kingdom, as injection by one-third significantly reduced well as the “advisory” system used in the monthly seismic rate. One method that is Oklahoma use monitoring and enable different largely considered not successful in mitigating regulations to take effect depending on risk is injecting fluid without pressure at the different levels of seismic activity. In Alberta, wellhead (Rubinstein et al. 2014; Weingarten for example, if an M ≥2 earthquake occurs et al. 2015). The weight of the water column within 5 km of activities, an induced can cause pressure and stress changes at seismicity plan must be implemented and seismic depth (Rubinstein et al. 2014). operations can resume with permission from Another method for mitigating seismicity the regulator. For earthquakes above M4.0, would be to re-inject wastewater into the however, activities must be halted and can reservoir from which it came, where pressure only resume with permission from the is being released by the production of oil and regulator. Walters et al. (2015) discussed how gas (Zoback 2016a). Additionally, having to incorporate these anthropogenic factors into geologic formations that act as barriers seismic risk assessments in more detail. between injection formations and deeper Although such measures do not guarantee that formations with faulting may also help seismicity will not occur, they do provide (Zoback 2016a, Shirzaei et al. 2016). Using an regulators with a way to manage factors that analytical-numerical model, Zhang et al. could increase the risk of nucleation. (2013) found that confining units and barriers 9.2. Study Span Chart: Seismicity (effectively a “bottom seal” on injection) had Impacts the single largest effect in preventing induced seismicity in the underlying crystalline We found 39 studies on induced seismicity basement. Avoiding injection in areas with from oil and gas development during our high seismicity seems like an obvious literature review—these are presented in the mitigation technique, but Goebel et al. (2015) chart on page 24 and grouped in the left presented preliminary evidence that areas column according to their methodologies, with already prone to natural seismicity, such as six categories in all. The top row of the chart Kern County, California, may face lower risks contains a representation of the impact from induced seismicity from fluid injection. pathway elements (defined below). The Goebel et al. (2015) is the only such study we elements covered by each study are shown at a found that analyzes induced seismicity in the

www.rff.org | 21 Resources for the Future | Krupnick and Echarte glance by the color-coded boxes that uses simulation modeling or uses correspond to the to the impact pathways. geologic/seismologic analysis of ex post We found one survey, McComas et al. earthquake data). (2016). We found nine statistical studies, Another notable takeaway from this chart which use data analysis to assess above- is that only three studies (Walsh and Zoback ground impacts or the associations among 2016; Langenbruch and Zoback 2016; regional seismic trends and certain oil and gas Shirzaei et al. 2016) are probabilistic, or development activities. Frohlich et al. (2016) working toward predictive models. Predictive is labeled as historical. It used a five-question and probabilistic models are important in that test to analyze historic earthquakes and assess they can measure baselines of future activity how likely it is that a given earthquake is against which policies can be compared. Such induced. studies enable more effective policy implementation. Only one study of the three We found eight studies using “geomechanic modeling, simulations,” that mentioned above (Langenbruch and Zoback model regional seismic trends, conduct fault 2016) predicts future changes in the rate of simulations and reactivations, and (in one seismicity as a regional probability. Further instance) a specific seismic event. Three research is clearly needed to expand studies are labeled probabilistic and predictive understanding in this area. Further work is or aid in hazard assessments in that they use also needed to address public concerns about methodologies that have implications for induced seismicity. assessing future changes in seismic rates Impact Pathway Elements related to oil and gas operations using a Activities variety of methods described in the literature review. Finally, the seventeen Activities can include a number of oil and geologic/seismologic studies most often gas operations, though for seismicity, analyze seismic data to characterize specific activities would refer to wastewater disposal, seismic events—including “relocating” an enhanced oil recovery, hydraulic fracturing, earthquake, assessing the focal mechanisms, and so on. This element can also indicate the conducting stress analysis, and more. These general presence of oil and gas development studies also assess the relationship between in a community rather than specific activities these seismic events and oil and gas activities or processes. in a number of ways. These six categories are Burdens not exclusive—for example, some geologic/seismologic studies may also use Burdens are the initial consequences of the some statistical analysis. above activities, such as emissions caused by diesel pumps in a fracking operation, or, here, It is important to note that this span chart an earthquake event. does not portray elements that would determine study quality (which we present in Exposure the Community Risk-Benefit Matrix on page Exposure measures the amount of a 3.) Also, as we can claim no expertise in this substance or other type of burden that enters academic area, this chart as well as the the body or the amount and intensity of literature review itself do not address earthquakes a region might be exposed to over distinctions in geologic or seismological a given time period. Exposure can also include analysis across the papers in these categories the element of dose or cover a measure of the (aside from differentiating when such analysis

www.rff.org | 22 Resources for the Future | Krupnick and Echarte number of people or animals exposed to a indirectly addressed exposure to shaking by burden. analyzing data on shaking intensity for Impacts induced earthquakes from the US Geological Survey’s “Did You Feel It?” system. These Impacts include both physical and mental three studies are far from conclusive regarding outcomes that affect human or animal the community impacts of earthquakes— populations, such as increased preterm birth or policymakers must therefore rely on aboveground damage to dwellings from qualitative reports and anecdotal evidence to induced seismicity. assess whether a mitigating response is Scope of the Studies sufficient. Further, there are no studies that address trade-offs that residents might be Looking down the columns, the vast willing to make. Perhaps local residents want majority of these studies (36 of 39) address no risk of induced seismicity, regardless of the only activities and burdens—these are, for the cost, or perhaps they are willing to trade off most part, studies assessing causation between the potential for some induced seismicity for oil and gas activities and specific seismic an increase in energy production in the region. events or regional changes. McComas et al. More such studies would be needed to address (2016) and Liu, Ferreira, and Brewer (2016) these issues. are the only studies to address the impacts communities face from induced seismicity. McComas et al. (2016) conducted a survey to assess public attitudes toward seismic events and the acceptability of induced seismic events, whereas Liu, Ferreira, and Brewer (2016) studied housing price changes related to oil and gas development activities and seismicity in Oklahoma. Hough (2014)

www.rff.org | 23 Resources for the Future | Krupnick and Echarte

STUDY SPAN CHART: SEISMICITY STUDIES

www.rff.org | 24 Resources for the Future | Krupnick and Echarte

References British Columbia Oil and Gas Commission. Ake, Jon, Kenneth Mahrer, Daniel O’Connell, 2012. “Investigation of observed and Lisa Block. 2005. “Deep-Injection and seismicity in the Horn River basin.” Closely Monitored Induced Seismicity at https://www.bcogc.ca/node/8046/downloa Paradox Valley, Colorado.” Bulletin of the d. Seismological Society of America 95(2): CSIS (Center for Strategic and International 664–683. DOI: 10.1785/0120040072. Studies). 2017. “Managing Societal and Associated Press. 2016. “Substantial damage Environmental Risks/Impacts of Onshore after earthquake rattles major Oklahoma Oil and Gas Development.” Workshop. oil hub.” November 07. April 20. http://www.foxnews.com/us/2016/11/07/d Dieterich, James H., Keith B. Richards- amage-reported-no-injuries-as-5-0- Dinger, and Kayla A. Kroll. 2015. earthquake-rattles-central-oklahoma.html. “Modeling Injection-Induced Seismicity Atkinson, Gail M., David W. Eaton, Hadi with the Physics-Based Earthquake Ghofrani, Dan Walker, Burns Cheadle, Simulator RSQSim.” Seismological Ryan Schultz, Robert Shcherbakov, Kristy Research Letters 86(4): 1-8. DOI: Tiampo, Jeff Gu, Rebecca M. Harrington, 10.1785/0220150057. Yajing Liu, Mirko van der Baan, and EPA (US Environmental Protection Agency). Honn Kao. 2016. "Hydraulic fracturing n.d. “Class II Oil and Gas Related and seismicity in the Western Canada Injection Wells.” Sedimentary Basin." Seismological https://www.epa.gov/uic/class-ii-oil-and- Research Letters 87(3): 631–647. DOI: gas-related-injection-wells. (Accessed 10.1785/0220150263. January 13, 2017). Bao, Xuewei, and David W. Eaton. 2016. Fitzpatrick, Jessica, and Mark Petersen. 2016. “Fault activation by hydraulic fracturing in “Induced Earthquakes Raise Chances of western Canada.” Science. DOI: Damaging Shaking in 2016.” Science 10.1126/science.aag2583. Features, United States Geological Survey Beron, Kurt J., James C. Murdoch, Mark A. (USGS). March 28. Thayer, and Wim P.M. Vijverberg. 1997. https://www2.usgs.gov/blogs/features/usgs "An analysis of the housing market before _top_story/induced-earthquakes-raise- and after the 1989 Loma Prieta chances-of-damaging-shaking-in-2016/. earthquake." Land Economics 73(1): 101– (Accessed January 12, 2017). 113. Friberg, Paul A., Glenda M. Besana-Ostman, Block, Lisa V., Christopher K. Wood, William and Ilya Dricker. 2014. “Characterization L. Yeck, and Vanessa M. King. 2014. of an Earthquake Sequence Triggered by “The 24 January 2013 ML 4.4 earthquake Hydraulic Fracturing in Harrison County, near Paradox, Colorado, and its relation to Ohio.” Seismological Research Letters deep well injection.” Seismological 85(6): 1295–1307. DOI: Research Letters 85(3): 609–624. DOI: 10.1785/0220140127. 10.1785/0220130188. Frohlich, Cliff. 2012. “Two-year survey comparing earthquake activity and injection-well locations in the Barnett Shale, Texas.” PNAS 109(35): 13934– 13938. DOI: 10.1073/pnas.1207728109.

www.rff.org | 25 Resources for the Future | Krupnick and Echarte

Frohlich, Cliff, and Michael Brunt. 2013. GWPC (Groundwater Protection Council) and “Two-year survey of earthquakes and IOGCC (Interstate Oil and Gas Compact injection/production wells in the Eagle Commission). 2015. Potential Injection- Ford Shale, Texas, prior to the Mw4.8 20 Induced Seismicity Associated with Oil & October 2011 earthquake.” Earth and Gas Development: A Primer on Technical Planetary Science Letters. 379: 56–63. and Regulatory Considerations Informing DOI: 10.1016/j.epsl.2013.07.025. Risk Management and Mitigation. Frohlich, Cliff, Heather DeShon, Brian Stump, StatesFirst Induced Seismicity Injection Chris Hayward, Matt Hornbach, and Jacob Work Group (ISWG). I. Walter. 2016. “A Historical Review of Healy, J.H., W.W. Rubey, D.T. Griggs, and Induced Earthquakes in Texas.” C.B. Raleigh. 1968. “The Denver Seismological Research Letters. 87(4): 1– Earthquakes.” Science 161(3848): 1301– 17. DOI: 10.1785/0220160016. 1310. Frohlich, Cliff, Chris Hayward, Brian Stump, Hermann, R.B., S. Park, and C. Wang. 1981. and Eric Potter. 2011. “The Dallas–Fort “The Denver earthquakes of 1967-1968.” Worth earthquake sequence: October 2008 Bulletin of the Seismological Society of through May 2009.” Bulletin of the America 71(3): 731–745. Seismological Society of America 101(1): DOI: 10.1785/0120120109. 327–340. DOI: 10.1785/0120100131. Hidano, Noboru, Tadao Hoshino, and Ayako Gan, Wei, and Cliff Frohlich. 2013. “Gas Sugiura. 2015. "The effect of seismic injection may have triggered earthquakes hazard risk information on property prices: in the Cogdell oil fied, Texas.” PNAS Evidence from a spatial regression 110(47): 18786–18791. discontinuity design." Regional Science DOI: 10.1073/pnas.1311316110. and Urban Economics 53: 113–122. DOI: Gibbs, J., J. Healy, C. Raleigh, and J. Coakley. 10.1016/j.regsciurbeco.2015.05.005. 1973. “Seismicity in the Rangely, Holland, Austin A. 2011. “Examination of Colorado, area: 1962–1970.” Bulletin of Possibly Induced Seismicity from the Seismological Society of America Hydraulic Fracturing in the Eola Field, 63(5): 1557–1570. Garvin County, Oklahoma.” Oklahoma Goebel, T.H.W., E. Hauksson, F. Aminzadeh, Geological Survey Open-File Report OF1- J.-P. Ampuero. 2015. “An objective 2011. method for the assessment of fluid ———. 2013. “Earthquakes Triggered by injection induced seismicity and Hydraulic Fracturing in South-Central application to tectonically active regions in Oklahoma.” Bulletin of the Seismological central California.” Journal of Society of America 103(3): 1784–1792. Geophysical Research: Solid Earth DOI: 10.1785/0120120109. 120(10): 7013–7032. DOI: Hornbach, Matthew J., Heather R. DeShon, 10.1002/2015JB011895. William L. Ellsworth, Brian W. Stump, Gono, Valerie. 2015. “On Shaky Ground: Chris Hayward, Cliff Frohlich, Harrison Understanding the Correlation between R. Oldham, Jon E. Olson, M. Beatrice Induced Seismicity and Wastewater Magnani, Casey Brokaw, and James H. Injection in the Fort Worth Basin.” Thesis. Luetgert. 2015. “Causal factors for University of Texas Austin. seismicity near Azle, Texas.” Nature

www.rff.org | 26 Resources for the Future | Krupnick and Echarte

Communications 6: 6728. Keranen, K.M., M. Weingarten, G.A. Abers, DOI: 10.1038/ncomms7728. B.A. Bekins, S. Ge. 2014. “Sharp increase Hornbach, Matthew J., Madeline Jones, in central Oklahoma Seismicity since 2008 Monique Scales, Heather R. DeShon, M. induced by massive wastewater injection.” Beatrice Magnani, Cliff Frohlich, Brian Science 345(6195): 448–451. Stump, Chris Hayward, Mary Layton. DOI: 10.1126/science.1255802. 2016. “Ellenburger wastewater injection Kim, Won-Young (2013). “Induced and seismicity in North Texas.” Physics of Seismicity associated with fluid injection the Earth and Planetary Interiors 261(A): into a deep well in Youngstown, Ohio.” 54–68. DOI: 10.1016/j.pepi.2016.06.012. Journal of Geophysical Research: Solid Horton, S. 2012. “Disposal of Hydrofracking Earth 118: 3506–3518. Waste Fluid by Injection into Subsurface Langenbruch, Cornelius, and Mark D. Zoback. Aquifers Triggers Earthquake Swarm in 2016. “How will induced seismicity in Central Arkansas with Potential for Oklahoma respond to decreased saltwater Damaging Earthquake.” Seismological injection rates?” Science Advances 2(11). Research Letters 83(2): 250–260. DOI: 10.1126/sciadv.1601542. DOI: 10.1785/gssrl.83.2.250. Liou, Tiffany. 2016. “Oklahoma Insurance Hough, Susan E. 2014. “Shaking from Commissioner Tours Pawnee Earthquake injection-induced earthquakes in the Damage.” News 9. September 8. central and eastern United States.” Bulletin http://www.news9.com/story/33052903/ok of the Seismological Society of America lahoma-insurance-commissioner-tours- 104(5): 2619–2626. pawnee-earthquake-damage. DOI: 10.1785/0120140099. Liu, Haiyan, Susana Ferreira, and Brady Hough, Susan E., and Morgan Page. 2016. Brewer. 2016. “Housing market impacts of “Potentially Induced Earthquakes during wastewater injection induced seismicity the Early Twentieth Century in the Los risk.” Agricultural and Applied Economics Angeles Basin.” Bulletin of the Association, 2016 Annual Meeting, July Seismological Society of America 106(6). 31–August 2. Boston, Massachusetts. DOI: 10.1785/0120160157. Llenos, Andrea L., and Andrew J. Michael. Justinic, Ashley Howe, Brian Stump, Chris 2013. “Modeling Earthquake Rate Hayward, and Cliff Frohich. 2013. Changes in Oklahoma and Arkansas: “Analysis of the Cleburne, Texas, Possible signatures of Induced Earthquake Sequence from June 2009 to Seismicity.” Bulletin of the Seismological June 2010.” Bulletin of the Seismological Society of America 103(5): 2850–2861. Society America 103(6): 3083–3093. DOI: DOI: 10.1785/0120130017. 10.1785/0120120336. McCalpin, James P. 2009. “Earthquake Keranen, Katie M., Heather M. Savage, Magnitude Scales,” Appendix 1 in Geoffrey A. Abers, and Elizabeth S. Paleoseismology, 2nd ed., edited by James Cochran. 2013. “Potentially induced P. McCalpin. Academic Press. earthquakes in Oklahoma, USA: Links McClure, Mark W., and Ronald N. Horne. between wastewater injection and the 2011 2011. “Investigation of injection-induced Mw 5.7 earthquake sequence.” Geology seismicity using a coupled fluid flow and 41(6): 699–702. DOI: 10.1130/G34045.1. rate/state friction model.” Geophysics

www.rff.org | 27 Resources for the Future | Krupnick and Echarte

76(6): WC181–WC198. metropolitan area.” Regional Science and DOI: 10.1190/GEO2011-0064.1 Urban Economics 37(1):87–99. DOI: McComas, Katherine A., Hang Lu, Katie M. 10.1016/j.regsciurbeco.2006.06.009. Keranen, Maria A. Furtney, Hwansuck Naoi, Michio, Miki Seko, and Kazuto Sumita. Song. 2016. “Public perceptions and 2009. "Earthquake risk and housing prices acceptance of induced earthquakes related in Japan: Evidence before and after to energy development.” Energy Policy massive earthquakes." Regional Science 99: 27–32. DOI: and Urban Economics 39(6): 658–669. 10.1016/j.enpol.2016.09.026. DOI: 10.1016/j.regsciurbeco.2009.08.002. McNamara, D.E., G.P. Hayes, H.M. Benz, National Research Council. 2013. Induced R.A. Williams, N.D. McMahon, and R.C. Seismicity Potential in Energy Aster. 2015. “Reactivated faulting near Technologies. National Academies Press: Cushing, Oklahoma: Increased potential Washington, DC. DOI: 10.17226/13355. for a triggered earthquake in an area of https://www.nap.edu/catalog/13355/induce United States strategic infrastructure.” d-seismicity-potential-in-energy- Geophysical Research Letters 42(20): technologies 8328–8332. DOI: Nicholson, C., and R. Wesson. 1992. 10.1002/2015GL064669. “Triggered earthquakes and deep well Muehlenbachs, Lucija, Elisheba Spiller, and activities.” Pure and Applied Geophysics Christopher Timmins. 2015. "The housing 139(3): 561–578. market impacts of shale gas development." Oklahoma Corporation Commission. 2016a. American Economic Review 105(12): “Earthquake Response Summary (As of 3633–3659. DOI: 10.1257/aer.20140079. September 12, 2016).” Murdoch, James C., Harinder Singh, and https://www.occeweb.com/News/2016/09- Mark Thayer. 1993. “The Impact of 12- Natural Hazards on Housing Values: The 16EARTHQUAKE%20ACTION%20SU Loma Prieta Earthquake.” Journal of the MMARY.pdf. (Accessed January 17, American Real Estate and Urban 2017). Economics Association V21(2): 167–184. ———. 2016b. “Advisory 4 – Cushing Area.” Murray, Kyle E. 2015. Class II Saltwater November 8. Disposal for 2009-2014 at the Annual-, https://www.occeweb.com/News/2016/09- State-, and County- Scales by Geologic 12- Zones of Completion, Oklahoma. Open- 16EARTHQUAKE%20ACTION%20SU File Report OF5-2015. Oklahoma MMARY.pdf. (Accessed January 17, Geological Survey. 2016). ———. 2016. “Modeling Approach for Petersen, Mark D., Charles S. Mueller, Estimating Co-Produced Water Volumes Morgan P. Moshcetti, Susan M. Hoover, and Saltwater Disposal Volumes in Andrea L. Llenos, William L. Ellsworth, Oklahoma.” In American Geophysical Andrew J. Michael, Justin L. Rubinstein, Union (AGU) Annual Fall Meeting. San Arthur F. McGarr, and Kenneth S. Francisco, CA. Rukstales. 2016. 2016 One-Year Seismic Nakagawa Masayuki, Makoto Saito, Hisaki Hazard Forecast for the Central and Yamaga. 2007. “Earthquake risk and Eastern United States from Induced and housing rents: evidence from the Tokyo Natural Earthquakes. US Geological

www.rff.org | 28 Resources for the Future | Krupnick and Echarte

Survey (USGS). Open-File Report 2016- Sumy, Danielle F., Elizabeth S. Cochran, 1035 Version 1.1, June 2016. Katie M. Keranen, Maya Wei, and https://pubs.usgs.gov/of/2016/1035/ofr201 Geoffrey A. Abers. 2014. “Observations 61035ver1_1.pdf. of static Coulomb stress triggering of the Raleigh, C.B., J.H. Healy, and J.D. November 2011 M5.7 Oklahoma Bredehoeft. 1976. “An experiment in earthquake sequence.” Journal of earthquake control at Rangely, Colorado. Geophysical Research: Solid Earth 119: Science 191(4233): 1230–1237. 1904–1923. DOI:10.1002/2013JB010612. DOI: 10.1126/science.191.4233.1230. TNO. 2014. “Literature review on Injection- Rubinstein, Justin L., William L. Ellsworth, Related Induced Seismicity and its Arthur McGarr, and Harley M. Benz. relevance to Nitrogren Injection.” TNO 2014. “The 2001-Present Induced 2014 R11761. Earthquake Sequence in the Raton Basin USGS (US Geological Survey). n.d. a. of Northern New Mexico and Southern “Earthquake Facts & Earthquake Fantasy.” Colorado.” Bulletin of Seismological https://earthquake.usgs.gov/learn/topics/m Society of America 104(5): 2162–2181. egaqk_facts_fantasy.php. (Accessed DOI: 10.1785/0120140009. January 12, 2017). Rubinstein, Justin L. and Alireza Babaie ———. n.d. b. “Moment magnitude, Richter Mahani. 2015. “Myths and facts on scale – what are the different magnitude wastewater injection, hydraulic fracturing, scales, and why are there so many?” enhanced oil recovery, and induced https://www2.usgs.gov/faq/categories/982 seismicity.” Seismological Research 8/3357. (Accessed January 18, 2017). Letters 86(4): 1–8. DOI: ———. 2016a. “Magnitudes for Oklahoma 10.1785/0220150067. Earthquakes Shift Upward.” September 7. Rutqvist, Jonny, Antonio P. Rinaldi, Frédéric https://www.usgs.gov/news/magnitudes- Cappa, and George J. Moridis. 2013. oklahoma-earthquakes-shift-upward. “Modeling of fault reactivation and (Accessed January 12, 2017). induced seismicity during hydraulic ———. 2016b. “M5.8 North Central fracturing of shale-gas reservoirs.” Journal Oklahoma Earthquake of 03 September of Petroleum Science and Engineering 2016.” October 13. Earthquake 107: 31–44. DOI: Information Center. 10.1016/j.petrol.2013.04.023. https://earthquake.usgs.gov/ Shirzaei, Manoocher, William Ellsworth, earthquakes/eqarchives/poster/2016/20160 Kristy Tiampo, Pablo González, and 903.pdf. (Accessed January 17, 2017). Michael Manga. 2016. “Surface uplift and Van der Elst, Nicholas J., Morgan T. Page, time-dependent seismic hazard due to fluid Deborah A. Weiser, Thomas, H.W. injection in eastern Texas.” Science Goebel, and S. Mehran Hosseini. 2016. 353(6306): 1416–1419. “Induced earthquake magnitudes are as Skoumal, Robert J., Michael R. Brudzinski, large as (statistically) expected.” Journal and Brian S. Currie. 2015. “Earthquakes of Geophysical Research: Solid Earth Induced by Hydraulic Fracturing in Poland 121(6): 4575–4590. DOI: Township, Ohio.” Bulletin of the 10.1002/2016JB012818. Seismological Society of America 105(1): Van der Elst, Nicholas J., Heather M Savage, 189–197. DOI: 10.1785/0120140168. Katie M. Keranen, and Geoffrey A. Abers.

www.rff.org | 29 Resources for the Future | Krupnick and Echarte

2013. “Enhanced Remote Earthquake 10,198–10,207. DOI: Tiggering at Fluid-Injection Sites in the 10.1002/2016GL070861. Midwestern United States.” Science Zhang, Yipeng, Mark Person, John Rupp, 341(6142): 164–167. DOI: Kevin Ellett, Michael A. Celia, Carl W. 10.1126/science.1238948. Gable, Brend Bowen, James Evans, Karl Veil, John. 2015. “US Produced Water Bandilla, Peter Mozley, Thomas Dewers, Volumes and Management Practices in and Thomas Elliott. 2013. “Hydrogeolgic 2012.” Prepared for the Ground Water controls on induced seismicity in Protection Council. Veil Environmental, crystalline basement rocks due to fluid LLC. injection into basal reservoirs.” Ground http://www.gwpc.org/sites/default/files/Pr Water 51: 525–538. DOI: oduced%20Water%20Report%202014- 10.1111/gwat.12071. GWPC_0.pdf. Zoback, Mark D. 2016a. “Managing the Risk Walsh III, F. Rall, and Mark D. Zoback. 2015. of Earthquakes Triggered by Waste Water “Oklahoma’s recent earthquakes and Disposal and Hydraulic Fracturing.” saltwater disposal.” Science Advances Presentation prepared for “The Shale Gas 1(5): 1–9. DOI: 10.1126/sciadv.1500195. and Oil Wastewater Disposal Cycle: ———. 2016. “Probabilistic assessment of Managing Earthquake and Other potential fault slip related to injection- Environmental Risks. Resources for the induced earthquakes: Application to north Future, Stanford Woods Institute for the central Oklahoma, USA.” Geology 44(12): Environment, and Stanford Natural Gas 991–994. DOI: 10.1130/G38275.1. Initiative. October 18. http://www.rff.org/events/event/2016- Walters, Randi Jean, Mark D. Zoback, Jack 10/shale-gas-and-oil-wastewater-disposal- W. Baker, and Gregory C. Beroza. 2015. cycle-managing-earthquake-and-other. “Characterizing and Responding to Seismic Risk Associated with Earthquakes ———. 2016b. “Managing the Risks of Potentially Triggered by Fluid Disposal Injection-Induced Seismicity.” Keynote and Hydraulic Fracturing.” Seismological presentation, Panel 4, Onshore Research Letters 86(4): 1110–118. DOI: Unconventional Hydrocarbon 10.1785/0220150048. Development: Legacy Issues, induced seismicity, and innovations in managing Weingarten, M., S. Ge, J.W. Godt, B.A. risk. National Academy of Sciences. Bekins, and J.L. Rubinstein. 2015. “High- December 2. http://nas- rate injection is associated with the sites.org/uhroundtable/past- increase in US mid-continent seismicity.” events/onshore-workshop/onshore- Science Magazine 348(6241): 1336–1340. workshop-panel-4/. DOI: 10.1126/science.aab1345. Yeck, W. L., M. Weingarten, H. M. Benx, D. E. McNamara, E. A. Bergman, R. B. Hermann, J. L. Rubinstein, and P. S. Earle. 2016. “Far-field pressurization likely caused one of the largest injection induced earthquakes by reactivating a large preexisting basement fault structure.” Geophysical Research Letters 43(19):

www.rff.org | 30