Policy Analysis pubs.acs.org/est
Unconventional Oil and Gas Spills: Risks, Mitigation Priorities, and State Reporting Requirements Lauren A. Patterson,*,† Katherine E. Konschnik,‡ Hannah Wiseman,§ Joseph Fargione,∥ Kelly O. Maloney,⊥ Joseph Kiesecker,# Jean-Philippe Nicot,∇ Sharon Baruch-Mordo,# Sally Entrekin,○ Anne Trainor,◆ and James E. Saiers¶ †
Nicholas Institute for Environmental Policy Solutions, Duke University, 2111 Campus Drive, Durham North Carolina 27708, United States, ‡ Environmental Policy Initiative, Harvard Law School, #4123 Wasserstein Hall, Cambridge, Massachusetts 02138, United States § Florida State University College of Law, 424 W. Jefferson Street, Tallahassee, Florida 32306, United States ∥ The Nature Conservancy, 1101 West River Parkway, Suite 200, Minneapolis, Minnesota 55415, United States ⊥ U.S. Geological Survey, Leetown Science Center, Kearnevsville, West Virginia 25430, United States # The Nature Conservancy, Global Lands Team, 117 E. Mountain Avenue, Suite 201, Fort Collins, Colorado 80524, United States ∇ Bureau of Economic Geology, Jackson School of Geosciences, The University of Texas at Austin, 10100 Burnet Road, Building 130, Austin, Texas 78758, United States ○ Department of Biology, University of Central Arkansas, 201 Donaghey Avenue, Conway, Arkansas 72035, United States ◆ The Nature Conservancy, African Program, University of Cincinnati, Department of Biological Sciences, 820G Rieveschl Hall, Cincinnati, Ohio 45221, United States ¶ School of Forestry and Environmental Studies, Yale University, 195 Prospect St., New Haven, Connecticut 06511, United States S Supporting Information *
ABSTRACT: Rapid growth in unconventional oil and gas (UOG) has produced jobs, revenue, and energy, but also concerns over spills and environmental risks. We assessed spill data from 2005 to 2014 at 31 481 UOG wells in Colorado, New Mexico, North Dakota, and Pennsylvania. We found 2−16% of wells reported a spill each year. Median spill volumes ranged from 0.5 m3 in Pennsylvania to 4.9 m3 in New Mexico; the largest spills exceeded 100 m3. Seventy-five to 94% of spills occurred within the first three years of well life when wells were drilled, completed, and had their largest production volumes. Across all four states, 50% of spills were related to storage and moving fluids via flowlines. Reporting rates varied by state, affecting spill rates and requiring extensive time and effort getting data into a usable format. Enhanced and standardized regulatory requirements for reporting spills could improve the accuracy and speed of analyses to identify and prevent spill risks and mitigate potential environmental damage. Transparency for data sharing and analysis will be increasingly important as UOG development expands. We designed an interactive spills data visualization tool (http://snappartnership.net/groups/hydraulic-fracturing/webapp/spills. html) to illustrate the value of having standardized, public data.
■
energy development.3,4 Notwithstanding the incredible potential for energy development, UOG extraction activities have raised concerns about their potential environmental impacts, including dewatering streams as well as surface and groundwater pollution.5−7 Determination of the ecological impacts of UOG development is considered a top science priority to
INTRODUCTION
Global demand for energy, directives to reduce carbon dioxide emissions, and technological advancements in horizontal drilling and hydraulic fracturing, have spurred a rapid increase in alternative and unconventional energy production over the past decade.1 Despite a recent slowdown of activity, energy development will likely continue; with shale gas and tight oil play production estimated to grow 60% by 2040 compared to 2015 production.2 Unconventional oil and gas (UOG) produced from reservoirs with low porosity and permeability, particularly shale reserves, will be a key component of future © XXXX American Chemical Society
Received: November 14, 2016 Revised: February 3, 2017 Accepted: February 13, 2017
A
DOI: 10.1021/acs.est.6b05749 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Policy Analysis
Environmental Science & Technology
Figure 1. Location of UOG wells and reported spills.
affect which spills are reported and what information are provided. Spill data collected by states provide a rich opportunity for gaining insight into when spills are most likely to occur, where they are most likely to occur, and the underlying causes. Better insight into these three factors would provide regulatory agencies and industry decision-makers with important information on where to target efforts for locating and preventing future spills. However, due to different reporting requirements, the types of spills reported and details about the type, volume, location, timing and cause of the spills reported vary across states. Moreover, some states collect or store data in hard copy or image pdfs, inhibiting analysis unless personnel later digitize the data. As UOG expands, efforts to reduce spill risk would benefit from making data more uniform, accessible, and informative to well operators, regulators, and the public. Objectives. In this paper we cleaned and analyzed state spill databases related to unconventional oil and gas (UOG) wells in four states: Colorado,14 New Mexico,15 North Dakota,16 and Pennsylvania17 (Figure 1). We defined UOG wells as those having been hydraulically fractured, as explicitly identified in the databases we acquired, or, where this information was unavailable, as indicated by horizontal drilling or water use. These states were selected because of their significance as oil or natural gas producing states and because extensive spill data could be obtained. We assessed data availability for 11 other states; however the data were either difficult to obtain or we could not determine which wells were linked to UOG activity, leaving four states in this study. We focused on UOG wells as a manageable subset of oil and gas related spills and one of growing importance as UOG activity increases. Although some of the spills that occur at UOG wells are unique to UOG
inform energy policy and conservation and management of natural systems.8 Hydraulic fracturing enables resource extraction from both conventional and UOG wells by injecting high-pressure fluids mixed with chemical additives to fracture nonporous, lowpermeability rock and release the trapped gas or oil.9,10 One of the public’s greatest concerns is the impact of surface spills− particularly of the chemicals used in hydraulic fracturing−to the environment.11 Federal law establishes common minimum reporting requirements for some but not all spills (Supporting Information (SI), Section A). The Clean Water Act and Oil Pollution Act require reporting of discharges that create a sheen on navigable waters. The Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) and the Resource Conservation Recovery Act (RCRA) require reporting of spills of hazardous substances above a threshold volume, but exempt fracturing fluids and flowback from this requirement. CERCLA also excludes oil spills from reporting requirements. Beyond these minimum federal standards, in most cases states determine when and how oil and gas related spills are reported (exceptions are for wells on federal or Tribal lands, or on the Outer Continental Shelf). States often use “spill” and “release” interchangeably; hereafter, we generically use “spill”.12,13 State law specifies the type of spill event that triggers a reporting requirement usually based on whether the spill exceeded a threshold volume or concentration and whether it was contained or migrated beyond the well site. These rules also indicate the reporting method (i.e., verbal or written), timing, and content. These factors, as well as the frequency of inspection by regulators for independent detection of spills, B
DOI: 10.1021/acs.est.6b05749 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Policy Analysis
Environmental Science & Technology
given that only a subset of all spills are required to be reported to each state. Colorado. In 2000, an interagency memorandum required companies to verbally report any spill entering or threatening to impact surface waters as soon as practicable.24 In addition, companies were required to report spills exceeding 5 barrels (210 gal) on state Form 19 within 10 days of discovery (SI Figure S2). In 2014, the reporting thresholds were lowered and time periods for reporting were shortened;25 requiring any spill or release greater than 1 barrel (42 gal) that escapes secondary containment to be reported. Also in 2014, Colorado amended Form 19 to capture information about the material spilled and the cause for the spill.25 Spills data were obtained through the Oil & Gas Conservation Commission. New Mexico. New Mexico’s Oil and Conservation Division’s 2001 rules required companies to verbally report within 24 h any “major release” of oil, gas, produced water, condensate, oil field waste including regulated radioactive materials or other oil field related chemicals, contaminants, or mixtures.26 A major release was defined as greater than 25 barrels (1050 gal), or a release that resulted in a fire or in substantial damage to property or the environment, would reach a water course, or might “with reasonable probability endanger public health” or water quality.27 In addition to the verbal reports for major releases, written reports were required within 15 days on Form C-141 to document any “major release” or “minor release”, that is, releases of at least 5 barrels (210 gal). The location and cause of the spill must be reported, along with the material and the source of the release, as well as whether the release reached a waterway. The Division reorganized the rules in 2008, but the substantive reporting provisions remained the same.27 Spills data were obtained from the Oil Conservation Division. North Dakota. Prior to 2010, companies in North Dakota had to report “any fire, leak, spill, blowout, or release of fluid” except for spills less than 1 barrel (42 gal) that stayed on site.28 Reports, which were strictly verbal, had to include the location and cause of the incident, and the amount and type of fluid involved. Effective April 1, 2010, a time limit was placed on this verbal report (within 24 h of discovery); in addition, a written report was required to be submitted within 10 days of cleanup “unless deemed unnecessary by the director” of oil and gas of the Industrial Commission.29 Effective April 1, 2014, North Dakota added an online reporting requirement within 24 h of discovery.30 Spills data were obtained from the Oilfield Environmental Incident reports provided by the Department of Health-Environmental Health. These data do not include RCRA spills. Pennsylvania. Pennsylvania’s 2001 rules required companies to report by telephone to the Department of Environmental Protection any “reportable release of brine” or the discharge of any substance which would endanger downstream users of water, result in or create a danger of pollution of Pennsylvania waters, or damage property.12,31,32 The report had to include the location and cause of the incident. “Reportable release of brine” was defined as “spilling, leaking, emitting, discharging, escaping or disposing” of at least 5 gallons in 24 h of brine containing more than 10,000 mg/L total dissolved solids (TDS), or of at least 15 gallons of brine with a lower TDS concentration. In October 2016, Pennsylvania’s new rules went into effect; these will require written spill reports. 33 Pennsylvania does not have a separate spill data set, therefore spill data for our analysis were pulled from the Department of Environmental Protection’s notice of violations (NOV) data-
development, such as spills of fracturing chemicals and flowback, other spills, such as spills of produced water, can occur at both conventional and unconventional wells. Our objective was two-fold. First, we assessed the regulatory framework for how spills were reported in each state and how that shaped the data collected. Second, we quantified (1) the probability of a spill by well age (with spud−initial drilling− year representing age 0), (2) the risk of a spill occurring at different locations on a well pad, (3) the volume released, and (4) the underlying cause of the spill. The regulatory assessment and the spill data analysis were used to ascertain how improved data collection and standardization procedures might enable higher resolution data analysis that could decrease the likelihood for spills. We built an interactive database to showcase this opportunity (http://snappartnership.net/ groups/hydraulic-fracturing/webapp/spills.html).
■
MATERIALS AND METHODS We obtained state oil and gas spills data for spills occurring between 2005 and 2014 at UOG wells that were spudded as early as 1995. We constrained UOG well spud dates to limit the range of technology and techniques to unconventional shale well activity18 and capture the transition to mainstream UOG activity. Analysis of spills data was limited by difficulties obtaining earlier data and little UOG production related to shale gas or tight oil plays occurred prior to 2005.2 Spill data were matched by American Petroleum Institute (API) numbers to UOG wells. Data for Colorado UOG wells were obtained from the IHS Enerdeq database,19 and from state databases in New Mexico,20 North Dakota,21 and Pennsylvania.22 When the distinctions between conventional and UOG wells were not clear, we did not include the well to avoid including non-UOG wells in our data set. Pennsylvania was the only state to provide a separate well database for unconventional wells. The IHS data for Colorado provided the volume used to complete wells so we included all horizontal wells and those wells using more than 1 Mgal for completion. In North Dakota and New Mexico, only horizontal wells were included. Further detail on how we identified UOG wells in each state can be found in Section B of the Supporting Information. Spill Data. Spills are events in which an unauthorized discharge released material, regardless of whether the spill stayed within the boundaries of the pad or migrated into groundwater or surface waterways. For a spill to occur there must be (1) a means by which materials were released from a particular point at or near the well site (hereafter referred to as a pathway) and (2) a causal mechanism (such as human error or equipment failure) that resulted in a release from the particular pathway. For example, a tank (pathway) may release produced water to the surface due to equipment failure (causal mechanism) from a valve malfunction. We focused on identifying the pathway and the causal mechanism for each spill to determine the most likely routes for spills and thereby suggest where monitoring, development of new practices, and focused interventions may have the most benefit. A detailed analysis of the materials spilled and potential environmental impacts of spills is described in a separate paper.23 The best source of information regarding UOG spills are state reporting records. Each state’s reporting thresholds and requirements differ, such that the subset of spills that are reported vary by state (SI, Section B provides details on state spills data). The number of spills reported is likely conservative C
DOI: 10.1021/acs.est.6b05749 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Policy Analysis
Environmental Science & Technology
Figure 2. Common pathways for spills.
wellhead, and unknown (Figure 2; SI Table S3). Across all states, between 93% (PA) and 98% (ND) of spills fell into these main pathways. Of the four states assessed in this study, only Colorado and New Mexico categorize the underlying cause of a spill in their aggregated data sets. We folded their standardized designations into four general categories: equipment failure, environmental conditions, human error, and unknown. We used narratives and other reported data from North Dakota and Pennsylvania to assign a causal category to each spill in those states as well. Equipment failures ranged from leaks due to corrosion to specific mechanical problems within equipment such as valves, flanges, gaskets, sight glasses, fire tubes, and polish rods. Spills from equipment that failed due to extreme conditions, such as cracked valves due to freezing, were listed as spills caused by environmental conditions. Other environmental conditions included flooding, heavy rain, snowmelt, high winds, rock slides, well kicks (incidents in which pressure causes brines and other fluids to enter the wellbore), valves opened by cattle coming into contact with equipment, and gophers chewing through flowlines. Human errors included incorrect valve positions, miscommunication between well operators and transportation staff, flowlines being punctured by construction equipment, vandalism, and illegal dumping. Analysis. We normalized the data four different ways to compare spill rates between states: the average rate from 2005 to 2014, annual rate, life-year rate, and the combination of annual and life-year rates. We use the term “rate” to refer to the occurrence rate (frequency) of spill events. Examples for all calculations can be found in the SI Calculations spreadsheet. We calculated the average spill rate by dividing the total number of spills by the number of well-years observed within the state. A well-year is observed each year since the well was drilled and spill data were obtained (2005−2014). For example, a well drilled in 2010 would have five well-years observed (from 2010 to 2014) for a spill in 2014, but if that well had been drilled prior to 2005, it would have 10 well-years observed (from 2005 to 2014).
base for UOG (SI, Section B). This necessarily limited the spill data to those where an inspector issued an NOV, possibly leading to an underestimation of the number spills in our analysis. Using the violation code and comments, we included in our analysis only those NOV’s categorized as having a spill or the potential to result in a spill (n = 1293). Data Cleaning. To ensure consistency among states, the same analyst cleaned each state database. The data provided by each state were reconciled with a consistent classification system for the volume units, material, pathway, and cause of spill. Some states provided the data in separate fields, while others only provided the information through narrative descriptions. When there were conflicting data for a spill (e.g., the cause or volume of the spill differed from the narrative), the narrative was taken to be correct. Spill volumes were converted from reported units (typically barrels) into cubic meters and gallons. The material spilled was categorized as drilling waste, chemicals, hydraulic fracture solution, saltwater, freshwater, oil products, diesel, equipment oil, and unknown. We employed similar methods to those reported in an EPA analysis of eight states34 in that we used one analyst and treated narrative descriptions as the governing factor to identify common pathways and causal mechanisms. However, we focused on spills at all stages of well development while the EPA analysis focused on spills connected to the hydraulic fracturing process. Pathway categories were developed based on descriptive narratives of incidents. Pathways identify the specific point on or near the well site at which a spill occurred, except the blowout pathway, which describes an event in which pressure builds up in the wellbore and can lead to underground or surface releases. Pathways that were relatively small and occurred infrequently were lumped into a category of “other” (SI, Section C). More commonly reported pathways include: blowouts, drilling equipment (active mud system, drill rig, and shakers), completion equipment (blender, chemical totes, and storage containers), tanks, pits, flowlines, heater treaters, stuffing boxes, pumps, transportation, leaks around the D
DOI: 10.1021/acs.est.6b05749 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Policy Analysis
Environmental Science & Technology We calculated the annual spill rate by dividing the number of spills in a given year by the cumulative number of wells drilled, including that year (see SI Calculations spreadsheet). Given the rapid and recent expansion of unconventional wells, most of our observations are for young wells (median well age in 2014 was 3 years). It may not be appropriate to extrapolate this spill rate over the entire lifetime of a well. Therefore, we also calculated the life-year spill rate for each lifeyear of a well (i.e., all new wells, all one year old wells, etc.). We based the life-year of a well on its spud date. We rounded spills to the closest life-year, such that life-year 1 represents spills that occurred between six and 18 months after a spud date. Lifeyears ranged from 0 (wells spudded in 2014) to 19 (wells spudded in 1995) (SI Table S7). While spills may have occurred at some of these wells prior to 2005, we only have spill data from 2005 to 2014. The volume of material spilled was assessed using quantile regressions (quantreg package in R) at 10% intervals. For each state, a chi-square test was performed between the observed and expected number of wells with multiple spills. The expected number of wells is based on the binomial probability of more than one spill occurring at a well.
Figure 3. Annual spill rate. Dashed vertical lines represent changes in reporting requirements.
analysis, and we did not discern a noticeable change in spill rates from those few months of data. The requirement that the spill escape secondary containment may also drive down spill reports. Colorado also provided unique identifying information for pits, tanks, flowlines, etc., but it did not always link these spills to an individual well. This reporting method may have contributed to lower Colorado spill rates since we only included spills that could be linked to an unconventional well API. Pennsylvania’s reporting requirements remained the same throughout this time period; however, a spike in UOG well development in 2009 corresponded with a doubling of the number of inspectors from 35 to7635 and a spike in annual spill rates (Figure 3). The rates declined as the number of wells increased (1208 wells in 2009 to 4478 wells in 2011), due to a combination of factors including the centralization of regulators into a new Office of Oil and Gas Management within the Pennsylvania Department of Environmental Protection, operators gaining more experience with best management practices, and a shift in inspection to focus on drilling which increased the number of wells that could be inspected while simultaneously decreased the likelihood of finding a violation.36 Frequency of inspection is a critical factor for reporting spills in Pennsylvania since spills data were obtained from notices of violations. Meanwhile, pit spills decreased from 132 in 2010 to 28 in 2011, which may correspond with an increased effort by the Pennsylvania Department of Environmental Protection to enforce pit regulations (S. Perry, Pennsylvania Department of Environmental Protection, Personal Communication, May 25, 2016). Spill rates increased in New Mexico between 2012 and 2013, corresponding with increased production. Over this period, storage was inadequate to contain the increased quantities of produced water and oil, and some truckers determined it was cheaper to illegally dump produced water than to find adequate storage (D. Sanches, New Mexico Oil Conservation Division, Personal Communication, June 1, 2016), which may have accounted for the increased spill rate. Life-Year Spill Rates. The spill rate by life-year of the well was greatest in the first few years (SI Table S8) when drilling, completion, and the highest amount of production occurred.37,38 The average spill rate ranged between 2% (Colorado) to 15% (North Dakota) during the first three years of well life; with rates decreasing as the well matured. We quantified the spill rate by life-year when at least 100 well years were present to increase the robustness of the analysis (Figure 4). Our data set only included new wells spudded from 1995; further research is needed on spill rates for older wells exceeding 20 years of age.
■
RESULTS AND DISCUSSION Average Spill Rates. Between 2005 and 2014 there were 6648 spills reported across the four states based on each state’s reporting requirements and our definition of UOG wells (SI Table S4). Our results exceed the number of spills found by EPA (n = 457) for eight states between 2006 and 2012 because we included spills that occurred during all stages of unconventional production (from drilling through production) while EPA focused on those spills explicitly related to hydraulic fracturing.34 In our study, North Dakota reported the most spills (n = 4453) and the highest overall spill rate at 12.2% (SI Table S6). Pennsylvania reported 1293 spills (4.3%), whereas New Mexico (426 spills; 3.1%) and Colorado (476 spills; 1.1%) each reported fewer than 500 spills. Some wells within each state experienced multiple spills (from 0.12% of wells in Colorado to 7.32% of wells in North Dakota); and these wells with multiple spills contributed to a large proportion of spills. The proportion of spills coming from a well with more than one spill accounted for 26% of all spills in Colorado, 37% in New Mexico, 53% in North Dakota, and 47% in Pennsylvania. The number of wells with multiple spills was 1.3 (North Dakota) to 4.6 (Colorado) times higher than expected due to chance alone, indicating that wells with one spill have a higher probability of spilling in the future (SI Table S5). Annual Spill Rates. Within New Mexico and Colorado the annual spill rate fluctuated within a percent or two; whereas in North Dakota and Pennsylvania, the annual spill rate fluctuated by more than 13% (SI Table S6). Fluctuations were likely influenced by changes to state spill reporting requirements, demonstrating how state policies directly impact efforts to identify and accurately assess UOG risk, their causes and potential mitigating remedies. For instance, North Dakota’s transition from a verbal notification to a written reporting requirement coincides with an increase in spill rates of 3−4% after 2010 (Figure 3). Similarly, the lowering of a reporting threshold for a spill could increase the number of reported spills. From 2000, Colorado only required spills larger than 210 gallons to be reported; after 2014, the state lowered the threshold to 42 gallons when a spill escaped secondary containment. This change occurred during our last year of E
DOI: 10.1021/acs.est.6b05749 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Policy Analysis
Environmental Science & Technology
Figure 4. Spill rates by life-year of well (left) and number of UOG wells that experienced each life-year (right). New wells included those with a life year of 0−2 years (3 total years).
Figure 5. Spill rate by life-year of the well. Bubble size is scaled to the maximum spill rate that occurred in each state. The maximum spill rate is given (4% in CO, 9% in NM, 23% in ND, and 12% in PA).
Annual Spill Rate by Life-Year. Given the observed change in spill rate between younger and older wells, it was necessary to control for life-year to understand changes in spill frequency during the decade we analyzed. We examined spill rate by year and life-year to differentiate whether spill rates were influenced by improvements in technology or management practices over time or by the age of the well. For Colorado, North Dakota, and Pennsylvania the maximum spill rate was consistently within the first three life-years of the well over time (Figure 5). New Mexico experienced an increase in spill rate across all life-years in 2013 when production began to rapidly increase.
Spill Volume. Reported volumes of spills ranged from as small as 0.004 m3 (1 gal) to as large as 3756 m3 (991 200 gal). Spill volume was not always reported. The frequency of a spill report including the volume spilled varied between states, ranging from 27% in Pennsylvania to 98% in North Dakota (SI Table S9). Pennsylvania may have “missing” volumes data because reporting of spills has only been required by telephone; agency guidance discouraged written notification.39 The 2016 regulations will require a written report for spills exceeding 42 gallons or when a spill threatens to pollute Pennsylvania waters.33,40 Another reason for missing volumes data is that the spills data for Pennsylvania comes from its enforcement F
DOI: 10.1021/acs.est.6b05749 Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Policy Analysis
Environmental Science & Technology records. Inspectors may not have known the volume of spills encountered after the fact, or may have observed a potential spill that had not yet manifested into an actual spill. When volumes were reported, the median volume of reported spills in Colorado (3 m3 or 798 gal) and New Mexico (4.9 m3 or 1302 gal) were larger than North Dakota (0.8 m3 or 210 gal) and Pennsylvania (0.5 m3 or 120 gal) (Figure 6). The larger median sizes in Colorado and New Mexico may reflect the higher volume reporting threshold relative to Pennsylvania and North Dakota (SI Figure S2).
Figure 7. Number of spills reported for each state by pathway category. Spills reported as pits or tanks in Pennsylvania were divided equally between the two categories.
(N = 519 spills), with the majority of those spills related to issues of loading and unloading material (N = 452; 87%). North Dakota experienced a large number of spills related to heater treaters and stuffing boxes (N = 682), while Pennsylvania had a high number of pit-related spills (N = 345), the majority of which occurred prior to 2011. The data indicate that for every 1000 wells drilled, 12 wells experienced a tank related spill and 11 wells experienced a spill related to flowlines (Table 1). Approximately one well per thousand had a blowout, which is comparable with reported unconventional blowout frequencies of 1−3 per 1000 wells in Texas42 and 1 blowout per 1000 onshore wells drilled between 1975 and 1990.43 In another study, the number of blowouts reported at California oil and gas wells decreased by 80% to 0.2 blowouts per 1000 wells due to improvements in production practice.44 Issues related to well casing and well communication are rare (
