Methane and Benzene in Drinking-Water Wells Overlying the EagleFord, Fayetteville, and Haynesville Shale Hydrocarbon ProductionAreasPeter B. McMahon,*,† Jeannie R.B. Barlow,‡ Mark A. Engle,§ Kenneth Belitz,∥ Patricia B. Ging,⊥

Andrew G. Hunt,# Bryant C. Jurgens,∇ Yousif K. Kharaka,○ Roland W. Tollett,◆ and Timothy M. Kresse¶

†U.S. Geological Survey, Colorado Water Science Center, Denver Federal Center, Bldg 53, MS 415, Denver, Colorado 80225, UnitedStates‡U.S. Geological Survey, Lower Mississippi-Gulf Water Science Center, 308 South Airport Road, Jackson, Mississippi 39208, UnitedStates§U.S. Geological Survey, Eastern Energy Resources Science Center, Department of Geological Sciences, University of Texas at ElPaso, El Paso, Texas 79968, United States∥U.S. Geological Survey, New England Water Science Center, 10 Bearfoot Road, Northboro, Massachusetts 01532, United States⊥U.S. Geological Survey, Texas Water Science Center, 1505 Ferguson Lane, Austin, Texas 78754, United States#Crustal Geophysics and 17 Geochemistry Science Center, Denver Federal Center, Bldg 95, MS 963, Denver, Colorado 80225,United States∇U.S. Geological Survey, California Water Science Center, 6000 J Street, Placer Hall, Sacramento, California 95819, United States○U.S. Geological Survey, National Research Program, 345 Middlefield Road, Menlo Park, California 94025, United States◆U.S. Geological Survey, Lower Mississippi-Gulf Water Science Center, 3095 West California, Ruston, Louisiana 71270, UnitedStates¶U.S. Geological Survey, Lower Mississippi-Gulf Water Science Center, 401 Hardin Road, Little Rock, Arkansas 72211, United States

*S Supporting Information

ABSTRACT: Water wells (n = 116) overlying the Eagle Ford,Fayetteville, and Haynesville Shale hydrocarbon production areaswere sampled for chemical, isotopic, and groundwater-age tracersto investigate the occurrence and sources of selected hydrocarbonsin groundwater. Methane isotopes and hydrocarbon gascompositions indicate most of the methane in the wells wasbiogenic and produced by the CO2 reduction pathway, not fromthermogenic shale gas. Two samples contained methane from thefermentation pathway that could be associated with hydrocarbondegradation based on their co-occurrence with hydrocarbons suchas ethylbenzene and butane. Benzene was detected at lowconcentrations (<0.15 μg/L), but relatively high frequencies(2.4−13.3% of samples), in the study areas. Eight of nine samplescontaining benzene had groundwater ages >2500 years, indicatingthe benzene was from subsurface sources such as natural hydrocarbon migration or leaking hydrocarbon wells. One samplecontained benzene that could be from a surface release associated with hydrocarbon production activities based on its age (10 ±2.4 years) and proximity to hydrocarbon wells. Groundwater travel times inferred from the age-data indicate decades or longermay be needed to fully assess the effects of potential subsurface and surface releases of hydrocarbons on the wells.

■ INTRODUCTION

Hydrocarbon production from unconventional oil and gas(UOG) reservoirs in the U.S. using horizontal drilling andhydraulic fracturing technologies has stimulated a substantialamount of research on the potential effects of these activities ongroundwater quality. Despite these efforts, the U.S. Environ-mental Protection Agency concluded that data gaps limitedtheir ability to fully assess the degree of impact in its final report

on the effects of hydraulic fracturing on water resources in theU.S.1 The Marcellus Shale production area in the northernAppalachian Basin is the most extensively studied UOG play

Received: February 9, 2017Revised: May 2, 2017Accepted: May 13, 2017

Article

pubs.acs.org/est

This article not subject to U.S. Copyright.Published XXXX by the American ChemicalSociety

A DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

with respect to understanding the effects of UOG developmenton groundwater quality.2−6 The Barnett (Texas) and Niobrara(Colorado) Shale play areas have also been the focus of severalstudies.6−8 Fewer peer reviewed studies have examinedgroundwater quality in other major U.S. shale gas and tightoil play areas such as the Bakken (Montana and North Dakota),Eagle Ford (Texas), Fayetteville (Arkansas), Haynesville(Louisiana and Texas), and Permian Basin (Texas and NewMexico).9−12

This paper examines the occurrence and sources of methaneand benzene in 116 drinking-water wells overlying the EagleFord (EF), Fayetteville (FV), and Haynesville (HV) Shaleproduction areas using chemical, isotopic, gas, and ground-water-age tracers (Supporting Information (SI) Figures S1−S3). Studies in other areas have suggested possible linksbetween some methane and benzene in groundwater and UOGproduction activities.1,2,6,13 High methane concentrations indrinking water can present a potential explosion hazard,14

whereas high benzene concentrations can result in an increasedcancer risk, among other health concerns.15 And whereasmethane in groundwater is almost entirely from subsurfacesources, benzene in groundwater can be derived fromsubsurface and surface sources.13,16

This work expands the areal coverage of previous studies thatexamined methane in groundwater in parts of the FV and HVplay areas,9,12 provides some of the first data on co-occurrencesof methane, benzene, and other VOCs in drinking-water wellsin the EF−FV−HV play areas, examines possible links betweenbiogenic methane and hydrocarbon degradation, and providesthe first systematic analysis of groundwater age in the playareas, which is used to evaluate the relative importance ofsubsurface and surface sources of benzene. The use ofconsistent sampling and analytical protocols and sets ofanalytes provides an opportunity to compare the occurrenceand sources of methane and benzene among UOG play areasrepresenting a range of hydrogeologic conditions and hydro-carbon development histories.

■ MATERIALS AND METHODS

In 2015 and 2016, samples of groundwater were collected from30 domestic wells in each play area (Supporting Information(SI) Table S1). Twenty-six public-supply wells were alsosampled in EF and HV. One sample of produced water wascollected from a gas well in the Haynesville Shale in Rusk

County, Texas in 2010, and five samples of produced waterwere collected from oil and condensate wells in the Eagle FordShale in Gonzales and Lavaca Counties, Texas in 2015. Waterwells were screened in aquifers composed of semiconsolidatedsands and clays to interbedded shale, sandstone, and limestonethat were generally hundreds to thousands of meters above theUOG-bearing shales (SI Section 1, Figures S1−S4). Waterwells were selected at random from populations of existingwells located ≤1 and >1 km from UOG wells to examinemethane and benzene in groundwater in relation to hydro-carbon-well proximity (SI Section 2). Distances to UOG wellsranged from 0.11 to 25 km, maximum densities of hydrocarbonwells of any kind within 1 km of the water wells ranged from 0to 62 wells, and drill years of the nearest hydrocarbon well ofany kind ranged from 1929 to 2014 (SI Figure S5, Table S2).The 2015 IHS Global data set for oil and gas wells and Statedatabases were used to determine the number and distance ofhydrocarbon wells from sampled water wells.17 UOG wellsrefer to hydrocarbon wells completed in the Eagle Ford,Fayetteville, or Haynesville Shales, and hydrocarbon wells ofany kind refer to UOG wells plus hydrocarbon wells completedin other formations in the study areas.Water-well samples were analyzed for major ions, nutrients,

and trace elements; methane, methane H and C isotopiccompositions, and C1−C5 gas composition; H and O isotopiccomposition of water; noble gas concentrations and isotopiccompositions; tritium, sulfur hexafluoride (SF6), carbon-14 indissolved inorganic carbon (DIC), and δ13C-DIC; and volatileorganic compounds (VOCs) (SI Tables S3−S5). Water fromhydrocarbon wells was analyzed for a subset of theseconstituents (SI Table S6). Details on the methods of samplecollection and analysis are in SI Section 2. Details on assessingVOC detections in groundwater are in SI Section 3.Tritium, SF6, carbon-14, and tritiogenic helium-3 concen-

trations were modeled using the software TracerLPM18 todetermine fractions of post-1950s groundwater in the samplesand mean ages of the pre- and post-1950s fractions. Pre- andpost-1950s groundwater are defined as water recharged beforeor after the early 1950s start of above ground nuclear weaponstesting, respectively. Details on the analysis of groundwater agesare in SI Section 4.Mann−Whitney tests on ranked data were used to test for

significant differences in methane concentrations betweenwater wells ≤1 and >1 km from hydrocarbon wells (SI Table

Figure 1. Br/Cl mass ratios as a function of chloride concentration in groundwater samples from the (A.) Eagle Ford, (B.) Fayetteville, and (C.)Haynesville study areas. Circles in each panel are color-coded according to the methane concentration in the sample. Data for shale water from thisstudy (A. and C.−gray stars) and ref 9 (B.). In B., data for sewage/septic-tank effluent from ref 22, samples with the lowest Br/Cl ratios containedtritium and could be mixed with septic-tank effluent in those rural homeowner wells. In C., data for Wilcox lignite from ref 23. Percentages indicatefractions of shale water in groundwater-shale water mixture.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

B

S7).19 Spearman correlation analysis was used to examinerelations between concentrations of methane and location,density, and age of hydrocarbon wells (SI Table S8). Asignificance level of 0.05 was used for each test.

■ RESULTS AND DISCUSSION

Major-Ion Chemistry of Groundwater. Total dissolvedsolids (TDS) concentrations in groundwater ranged from 99−7750 mg/L in EF (median = 493 mg/L), 34−509 mg/L in FV(median = 159 mg/L), and 76−1401 mg/L in HV (median =511 mg/L) (SI Table S3). The lowest TDS waters werecharacterized as calcium−sodium-chloride (pH < 5) to calcium-bicarbonate and mixed cation-bicarbonate (pH > 6) types (SIFigure S6). The highest TDS waters were mostly characterizedas sodium-bicarbonate to sodium-chloride and sodium-sulfatetypes. These TDS and major-ion characteristics are similar topreviously published data for the areas and have been attributedto varying amounts of calcite, silicate, gypsum, and halitedissolution; ion exchange; iron and sulfate reduction; andpossible mixing with small amounts of saline groundwater.9,12,20

Cross plots of major-ion concentrations from this study aregenerally consistent with those processes (Figure 1 and SIFigure S7).Mixing with even small amounts of brine from shales could

contribute substantial amounts of salt and gas to ground-water.6,9,21 Data for water isotopes, chloride concentrations,and Br/Cl mass ratios in groundwater and shale water indicatethe groundwater contained little (<1%), if any, water from theEagle Ford, Fayetteville, or Haynesville Shales (Figures 1 andS8),.9,22−24 Groundwater samples had water isotopic compo-sitions that plotted on or near global or local meteoric waterlines (SI Figure S8), whereas shale water was isotopicallyenriched compared to groundwater from the same study area.

Several EF samples with relatively high methane and chlorideconcentrations had Br/Cl ratios that appeared to be moreconsistent with dilute groundwater mixing with small amountsof a saline endmember having a Br/Cl ratio resembling that ofseawater than with Eagle Ford Shale water (Figure 1A). Salineformation waters with Br/Cl ratios similar to seawater havebeen reported elsewhere in the Texas coastal plain,24,25 butseawater is used as an analog here because the source of thesaline water is unknown. EF3 was unusual among the mixedsamples because of its high TDS concentration (SI Figure S6),large fraction of saline water (>10%) (Figure 1), and methaneisotopic composition (discussed below).FV17 had the highest chloride, sodium, and boron

concentrations among FV samples. Although the Cl−Br/Cldata indicate FV17 could have contained a small (<1%) fractionof Fayetteville Shale water (Figure 1B), the sodium and boronseem less likely to have been derived from shale water. Thesodium and boron concentrations were highly correlated (rho =0.90, p < 0.001) (SI Figure S7L), and a previous studyconcluded on the basis of boron isotopic data that the sodium−boron relation in groundwater was related to boronmobilization from clays in the aquifer by ion-exchangeprocesses rather than substantial mixing with boron-rich shalewater.9 This apparent lack of shale-water input is consistentwith methane isotopic data for FV17 (discussed below).Several HV samples with higher chloride concentrations also

appeared to contain chloride from a source other than shalewater based on their low Br/Cl ratios (although we only haveone sample to characterize Br/Cl in Haynesville Shale water)(Figure 1C). Halite dissolution might be a relatively importantsource of chloride in HV groundwater. Several of the highestTDS samples were sodium-chloride type waters that plot alongsodium-chloride concentration trend lines with molar ratios

Figure 2. δ13C composition of methane as a function of δ2H−CH4 and C1/C2+C3 in groundwater samples from the (A., D.) Eagle Ford, (B., E.)Fayetteville, and (C., F.) Haynesville study areas. Zone boundaries from refs 31 and 32. In A. and D., data for shale gas from this study (star symbols)and ref 30 (dashed gray box), data for fuel spill sites from refs 33, 34, and data for the methanogenic consortium from ref 35. In B. and E., data forshale gas from ref 27. In C. and F., data for shale gas from refs 28 (Texas) and 29 (Louisiana) and data for Wilcox coalbed gas from refs 36 (samplePA-2) and 37 (all others).

Environmental Science & Technology Article

DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

C

near 1 (SI Figures S6 and S7J), suggesting chloridecontributions from halite dissolution. Halite is present in thesubsurface throughout much of the HV region, particularly inthe adjacent East Texas and North Louisiana Salt Basins (SIFigure S3).26 HV16 and HV25 had more enriched Br/Cl ratiosthan most other high methane samples in HV and also arediscussed below.Concentrations of Methane in Groundwater. Methane

was detected at concentrations ≥0.001 mg/L in 81% (EF), 89%(FV), and 100% (HV) of the groundwater samples (SI TableS4; Figures S1−S3). Concentrations exceeded 10 mg/L, aproposed action level for methane in groundwater,14 in 7%(EF), 7% (FV), and 12% (HV) of the samples. The exceedancerate for FV was about 3 times higher than the rate (2%)previously reported for a smaller part of the play area inFaulkner and Van Buren Counties.9 For HV, the exceedancerate was lower than the rate (17%) reported for the Texas partof the play area;12 however, that study focused its sampling on amethane hotspot near the Mount Enterprise fault zone wherewe also detected methane >10 mg/L (well HV16) in this study(SI Figure S3). The highest methane concentrations mostlyoccurred in higher TDS sodium-bicarbonate waters (SI FigureS6), consistent with findings from several other stud-ies.3,9,10,12,20 Some sodium-chloride waters in EF and HV alsocontained high methane concentrations. Methane concen-trations were not significantly correlated with well depth (SITable S8), but in HV, there was a noticeable clustering of highconcentrations in the 40−100 m depth interval, similar toprevious findings in that area (SI Figure S9).12

Methane concentrations were not spatially correlated withhydrocarbon well locations in any of the study areas. Therewere no statistically significant differences in methaneconcentrations between water wells located ≤1 km and >1km from UOG wells or from hydrocarbon wells of any kind (SITable S7); nor were there significant correlations betweenmethane concentrations and distance to either UOG orhydrocarbon wells of any kind (SI Table S8). There werealso no significant correlations between methane concen-trations and number of hydrocarbon wells within 1 km of thewater wells, or between methane concentrations and drillingyear of the nearest hydrocarbon well.Sources of Methane in Groundwater. Most samples that

contained enough methane to measure its isotopic compositionhad values of δ13C−CH4 < −60‰, δ2H−CH4 between −250and −150‰, and C1/C2+C3 molar ratios >1000 (Figure 2);and 87% (FV) to 95% (HV) of the samples that had detectablemethane concentrations did not have detectable concentrationsof C3−C5 hydrocarbons (SI Tables S4 and S5). Thesegeochemical characteristics are typical of biogenic gas producedprimarily by the CO2 reduction pathway.8−10 The predom-inance of biogenic gas in FV and HV is generally consistentwith previous studies in smaller parts of those areas.9,12

Methane isotopic and hydrocarbon-gas compositional datafrom this and previous studies indicate shale gas in the studyareas was thermogenic and compositionally different from mostof the groundwater methane (Figure 2).27−30 The data indicatethe UOG plays were not important sources of methane in thesampled wells. This conclusion is supported by the Br/Cl dataand lack of correlation between groundwater−methaneconcentrations and hydrocarbon well locations, densities, anddrilling years.FV17 was the only FV sample with C1/C2+C3 < 1000, and

Cl−Br/Cl data indicate it could have contained a small (<1%)

fraction of Fayetteville Shale water (Figure 1B). However,FV17 did not contain detectable concentrations of C3−C5hydrocarbons and its methane isotopic data indicate apredominantly biogenic gas (CO2 reduction) component inthe sample (Figure 2B and E).δ13C−CH4 and C1/(C2+C3) values for HV16, HV25, HV29,

and HVP98 plot between the biogenic and thermogenic fieldsin Figure 2F, suggesting a mixed source for methane in thosesamples and/or alterations due to methane oxidation. Methaneoxidation processes clearly affected the δ13C−CH4 and δ2H−CH4 values of HVP98 (Figure 2C). However, HV16, HV25,and HV29 had δ2H values for methane and water close to whatcould be expected for biogenic methane produced by the CO2reduction pathway (Figure 3),31 consistent with the formation

pathway in almost all the other water samples that containedbiogenic methane. The three HV wells were completed in theWilcox Group and had methane isotopic compositions similarto coalbed methane in the Wilcox (Figure 2C). Wilcox coalbedmethane collected east of the study area in north-centralLouisiana was considered primarily biogenic in origin andproduced by the CO2 reduction pathway.37

Lignite in relatively close vertical proximity to water wellscould increase the likelihood of coalbed methane mixing withgroundwater. Lignite sample PA-2 and well HV16 were locatedin southern Panola County, Texas and had similar methaneisotopic compositions and C1/C2+C3 ratios, although PA-2contained detectable C3−C5 hydrocarbons and HV16 did not(Figures 2C, 2F).36 PA-2 was collected from a depth of 112m,36 and HV16 from 64 m (SI Table S1). Previous studiessuggested the Mount Enterprise fault in Panola County couldbe a migration pathway for coalbed methane to enter overlyingaquifers.12 Thirty-four to 44% of the domestic wells in Caddoand Bossier Parishes, Louisiana, where HV25 and HV29 werelocated, had at least one lignite layer described in their boringlogs.38 HV25 and HV29 also did not contain C3−C5hydrocarbons. Groundwater interaction with lignite couldaccount for the Br/Cl enrichment in HV16, given that lignitesamples from the Wilcox Group (n = 44) had Br/Cl ratiosranging from 0.006 to 0.076 (median 0.027),23 well above theBr/Cl ratio in HV16. Boron enrichment in several highermethane HV samples, including HV16 and HV25, relative tolower methane samples, could be further evidence forgroundwater interaction with lignite (SI Figure S7M). Co-occurrence of high boron and biogenic methane concentrations

Figure 3. δ2H isotopic compositions of water and methane ingroundwater samples from the study areas. Isotopic relations for theacetate fermentation and CO2 reduction pathways for methanogenesisfrom refs31 and32, respectively.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

D

in groundwater were also reported in the lignite-rich FortUnion Formation in the northern Great Plains.10

13C enrichment of Wilcox coalbed methane in north-centralLouisiana relative to typical biogenic gas (Figure 2C) has beenattributed to substrate depletion during methanogenesis.37

δ13C−CH4 and δ13C−DIC data for HV16, HV25, and HV29appear consistent with that process (SI Figures S10A, S10B).HV25 was the only high-methane sample that appeared to besubstantially stripped of light noble gases (SI Figure S10C),suggesting direct contact with a local gas accumulation.6,39

Given that methane in HV25 was compositionally differentfrom Haynesville Shale gas (Figures 2C, 2F) and other lines ofevidence presented here, it seems unlikely that the inferredlocal gas accumulation was from the Haynesville Shale.EFP72 was the only sample with a C3−C5 hydrocarbon

detection (0.0004 mol % propane) that also had a methaneconcentration >1 mg/L (6.7 mg/L), δ13C−CH4 > −50‰(−43.7‰), and C1/(C2+C3) < 1000 (551), indicating itcontained thermogenic gas. EFP72 was also unusual in that itwas the deepest (1299 m) and hottest (water temperature of60.8 °C) water well sampled. The depth and temperature ofthat well are consistent with geothermal gradients previouslyreported for that area (2.7−3.6 °C/100 m)40 (SI Figure S11A),suggesting this was not an anomalous temperature associatedwith leakage of deeper, hotter fluids. The water isotopiccomposition and Br/Cl ratio of EFP72 indicate it containedlittle (<1%), if any, Eagle Ford Shale water, and its TDSconcentration was comparable to some of the low-methane EFsamples (SI Figure S6). Cumulatively, the data indicatethermogenic methane in EFP72 was not due to substantialmixing between groundwater and gas-rich formation waterfrom the Eagle Ford Shale such as might occur from large scalenatural brine migration21 or leakage from a nearby UOG well(closest UOG well 1.8 km away). The methane in EFP72 alsodoes not appear to be from substantial mixing with a migratedfree-gas phase based on its 20Ne/36Ar ratio (0.16) relative to air-saturated water (∼0.15) (SI Figure S11B). Elevated 20Ne/36Arratios (0.4 to >1) were observed in some groundwatercontaining thermogenic methane in the Marcellus play areaand attributed to solubility partitioning of light noble gases intoa migrated free-gas phase.6 Given the depth and temperature ofEFP72 and that the top of the Eagle Ford Shale in that area was∼2000 m deeper than EFP72,41 the simplest explanation of thedata is that there was some connection between the deepaquifer in that location and a relatively shallow accumulation ofthermogenic gas, although not through extensive direct contactgiven the limited 20Ne and 36Ar gas stripping in the sample (SIFigure S10C). Similar conditions were observed in groundwaterin parts of the Barnett UOG play area in north Texas.39

Evidence for Methane from Hydrocarbon Degrada-tion. EF3 and FV7 were notably different from the othersamples that contained biogenic methane in that they appearedto contain relatively large fractions of methane produced by theacetate fermentation pathway based on their methane andwater δ2H values (Figure 3). Although the methane isotopiccomposition of EF3 was somewhat unusual, a similar result wasobtained when the well was resampled (Figure 2A). EF3 had anisotopic composition similar to some methane at fuel-spill sitesin Minnesota and California that was produced by thefermentation pathway and subjected to varying degrees ofoxidation (Figure 2A).33,34 Methane can also be produced bythe fermentation pathway during crude oil degradation (Figure2A).35 The EF3 data could be consistent with fermentation

methane produced during crude oil degradation, withsubsequent isotopic alteration by oxidation processes. EF3contained trace amounts of butane, high chloride and helium-4concentrations relative to the shallow depth of the well (SIFigures S11C, S11D), and high sodium and boron concen-trations, indicating mixing with a deeper saline fluid (Figure 1Aand SI Figure 7K). EF3 was essentially tritium (<0.1 TU) andcarbon-14 (<1% modern carbon, pmc) dead (SI Table S4), soany hydrocarbons potentially mixed in EF3 were likely from asubsurface source such as natural hydrocarbon migration or aleaking hydrocarbon well. Three UOG wells were locatedwithin 1 km of EF3 (the closest well at 500 m) (SI Table S2),but Br/Cl data suggest groundwater in EF3 mixed with a salineendmember other than Eagle Ford Shale water (Figure 1A).Importantly, FV7, the other sample besides EF3 containing

fermentation methane, clearly contained fuel hydrocarbonsethylbenzene, xylenes, butane, and the manufactured fueladditive methyl tert-butyl ether (MTBE) (SI Table S5)possibly indicating a link between fermentation methane andhydrocarbon degradation in that sample.33−35 The presence ofMTBE and tritium (0.96 TU) in FV7 suggest a surface sourcefor the hydrocarbons in that sample. If fermentation methane inEF3 and FV7 was associated with degradation of oil or fuel, itwould have implications for interpreting the presence ofbiogenic methane in shallow groundwater in hydrocarbonproduction areas. Biogenic methane in shallow groundwater inthose settings is not typically thought to be related tohydrocarbon reservoirs or production,8−10 but data for EF3and FV7 may indicate methane produced by the fermentationpathway was associated with degradation of hydrocarbons.

Concentrations of Benzene in Groundwater. Thehighest benzene concentration detected in any water samplewas 0.127 μg/L (HV29, 2015 sample; SI Table S5), nearly 40times lower than the 5 μg/L federal drinking-water standard.15

Even though benzene concentrations were low, benzenedetection frequencies (DFs) were relatively high. Benzenewas detected at concentrations ≥0.013 μg/L (long-termmethod detection level; SI Section 3) in 9.3% (EF), 13.3%(FV), and 2.4% (HV) of the water samples, which are about1.5−8 times higher than benzene DFs reported in two nationalgroundwater data sets with similar detection levels.42,43 Aprevious study in the EF play area did not detect benzene ingroundwater,11 which could be due in part to differences inbenzene detection levels that were used (ref 11 did not report abenzene detection level). Benzene and methane were weaklycorrelated in EF but not correlated in FV or HV (SI Table S8).Benzene detections occurred in more varied water types andTDS ranges than methane (SI Figure S6), probably reflecting,in part, contributions of benzene to both shallow groundwaterfrom surface sources and to deeper groundwater fromsubsurface sources.

Surface versus Subsurface Sources of Benzene inGroundwater. Groundwater recharged after the onset ofUOG production activities in ∼2004 (FV) to ∼2008 (EF, HV)could potentially contain benzene and other hydrocarbons fromsurface releases associated with those activities such asaccidental spills or pipeline leaks. Tritium, tritiogenic helium-3, SF6, and carbon-14 were used to evaluate the timing ofrecharge. Age-dating has been used in previous studies toanalyze sources of hydrocarbons in groundwater overlying oiland gas fields,10,13,16 but this is the first systematic analysis ofgroundwater age in these play areas. Tritium was detected atconcentrations >0.1 to 0.2 TU in 87% (FV), 20% (EF), and

Environmental Science & Technology Article

DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

E

14% (HV) of the samples (SI Figure S12), indicating theycontained at least some groundwater recharged after the early1950s (post-1950s) (SI Section 4).18,44 The large percentage ofFV samples containing post-1950s recharge likely reflects theshallow water-well depths there (SI Table S1), and higherrecharge rates driven by the relatively low mean annual airtemperature and high mean annual precipitation in FV.45,46

Conversely, 86% (HV), 80% (EF), and 13% (FV) of thesamples were tritium dead (<0.1 to 0.2 TU) and appeared toonly contain pre-1950s recharge. For the pre-1950s samples inEF and HV, median carbon-14 and helium-4 concentrationswere 10 and 5 pmc, respectively, and 4.4 × 10−7 and 1.6 × 10−6

cm3STP/gH2O (∼10−100 times air saturation levels), respec-tively, suggesting many of those samples had ages older thanseveral thousand years.6,10,39

The presence of post-1950s recharge in some samples doesnot necessarily mean the water was recharged after the onset ofUOG development, so the lumped-parameter modelingsoftware TracerLPM18 was used to estimate fractions of post-1950s water in the samples and the mean ages of the pre- andpost-1950s fractions (SI Table S9) (SI Section 4). Despite thelarge percentage of FV samples that contained post-1950srecharge, only three of them had mean ages that overlapped theperiod of UOG development (Figure 4). The data indicategroundwater in the depth zones of the drinking-water wells wasnot particularly vulnerable to surface releases of contaminationpotentially associated with UOG activities, at least not in the∼7−11 year time frame of UOG development relative to whenthe wells were sampled (blue bands in Figure 4). The wellscould become more vulnerable with time as potentiallycontaminated groundwater overlying the wells moves deeperinto the aquifers. This is particularly relevant in FV where meanages of post-1950s recharge suggest groundwater travel timesfrom recharge areas to many of the wells were on the order of15−40 years.Of the nine samples with benzene detections, only FV12

contained benzene that could be from a surface releaseassociated with UOG production activities based on itsgroundwater age (Figure 4). FV12 contained about 80%post-1950s water, had a mean age for that fraction (10 ± 2.4years) that overlapped UOG development, and was 0.36 kmfrom an UOG well. FV-12 was a low-TDS, mixed-cation-bicarbonate type water, consistent with the groundwater beingrelatively young and less chemically evolved (Figure S6). Theremaining eight samples only contained pre-1950s water with

mean ages ranging from 2,600 to >30,000 years, indicating thebenzene in those samples was from subsurface sources such asnatural hydrocarbon migration or leaking hydrocarbon wells.Those samples were sodium-bicarbonate type waters withrelatively high TDS concentrations (Figure S6). Theoccurrence of benzene in old groundwater is consistent withprevious work near California oil fields.16 The results differfrom findings in the Marcellus play where a predominantlysurface source for diesel-range organic compounds and otherhydrocarbons in groundwater was observed,13 which couldreflect differences in recharge conditions and histories ofhydrocarbon production between the areas.Previous studies have shown that old groundwater ages can

be indicators of slow groundwater velocities.10,47−49 A study inthe EF area reported groundwater velocities on the order of 1−3 m/year in parts of an aquifer containing groundwater >5000years old,47 suggesting any benzene releases into old ground-water from leaking UOG wells might have only traveled shortdistances in the eight years since UOG production commenced.The closest UOG well to an EF water well with a benzenedetection was 1.26 km away (SI Table S2), suggesting thebenzene detections probably were not from UOG wells.An UOG source can also be ruled out for the highest

benzene concentration detected in pre-1950s groundwater(HV29). Benzene was detected in HV29 by the USGS in 2004(SI Table S5), about 4 years before to the start of UOGproduction in the HV play area but well after the drilling year(1955) of the nearest hydrocarbon well (0.13 km away) (SITable S2). HV29 also had 62 hydrocarbon wells within 1 km,92% of which predated UOG activity. This informationindicates benzene in HV29 was not from UOG productionactivities, but an association with older hydrocarbon productionactivities is possible given the proximity and age of thosehydrocarbon wells. HV29 also contained trace amounts ofchloroform and bromodichloromethane that could be asso-ciated with well chlorination, a recommended well-maintenancepractice in parts of Louisiana.50

Implications. UOG operations did not contribute sub-stantial amounts of methane or benzene to the sampleddrinking-water wells. However, groundwater travel timesinferred from the age-data indicate decades or longer may beneeded to fully assess the effects of potential subsurface andsurface releases of hydrocarbons on the wells.10 In EF and HV,where there was a long history of non-UOG hydrocarbonproduction (Figure S5), the few water wells with detections of

Figure 4. Mean age of the pre-1950s (lower half of y-axes) and post-1950s (upper half of y-axes) fractions of water as a function of the fraction ofpost-1950s water in groundwater samples from the (A.) Eagle Ford, (B.) Fayetteville, and (C.) Haynesville study areas. Blue bands indicate theperiod of UOG development relative to the sample year in each study area. Error bars represent 1-sigma errors. Orange symbols indicate sampleswith hydrocarbon detections. Hydrocarbon abbreviations: B, benzene; E, ethylbenzene; T, toluene; X, xylenes; MTBE, methyl tert-butyl ether; C3,propane; C4, butanes; C5, pentanes. Samples with benzene detections highlighted in red.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

F

benzene or C3−C5 generally appeared to be closer to non-UOG hydrocarbon wells than to more recently drilled UOGwells (Figure 5), suggesting a connection between hydrocarbon

detections and older wells (e.g., HV29). We cannot, however,rule out a possible natural subsurface source for somehydrocarbon detections (e.g., EFP72).16

Our results may not be representative of the entire studyareas. Only 130 hydrocarbon wells of any kind, out of acombined population in the study areas of >100 000 wells (SIFigure S5),17 were within 1 km of the sampled water wells.Sampling near more hydrocarbon wells could be warranted.Generally, samples were not collected at the water table wherethe aquifers are most vulnerable to surface releases ofcontamination. We also did not collect samples from multipledepths at any one location. It could be helpful to understandvertical variability in groundwater chemistry in close proximityto hydrocarbon wells.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.est.7b00746.

Further information on study design, sample collectionand analysis, assessing VOC detections, and groundwaterage-dating approach; additional figures and tables. Thedata are also available for download from https://doi.org/10.5066/F77D2SC4 (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: pmcmahon@usgs.gov.ORCIDPeter B. McMahon: 0000-0001-7452-2379NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSWe thank well owners who allowed us to sample their wells,and Brian Varela for doing the GIS analysis of the IHS dataset.This article was improved by the constructive reviews of

Stephen Opsahl and anonymous reviewers for the journal. Thiswork was funded by the USGS National Water-QualityAssessment (NAWQA) Project and Energy ResourcesProgram. Any use of trade, firm, or product names is fordescription purposes only and does not imply endorsement bythe U.S. Government.

■ REFERENCES(1) U.S. Environmental Protection Agency, Hydraulic Fracturing forOil and Gas: Impacts from the Hydraulic Fracturing Water Cycle onDrinking Water Resources in the United States (Final Report). EPA-600-R-16-236Fa; Washington, D.C.: 2016. https://www.epa.gov/hfstudy.(accessed December 13, 2016).(2) Osborn, S. G.; Vengosh, A.; Warner, N. R.; Jackson, R. B.Methane contamination of drinking water accompanying gas-welldrilling and hydraulic fracturing. Proc. Natl. Acad. Sci. U. S. A. 2011,108 (20), 8172−8176.(3) Molofsky, L. J.; Connor, J. A.; Wylie, A. S.; Wagner, T.; Farhat, S.K. Evaluation of methane sources in groundwater in northeasternPennsylvania. Groundwater 2013, 51 (3), 333−349.(4) Baldassare, F. J.; McCaffrey, M. A.; Harper, J. A. A geochemicalcontext for stray gas investigations in the northern Appalachian Basin:Implications of analyses of natural gases from Neogene-throughDevonian-age strata. AAPG Bull. 2014, 98 (2), 341−372.(5) Siegel, D. I.; Smith, B.; Perry, E.; Bothun, R.; Hollingsworth, M.Pre-drilling water-quality data of groundwater prior to shale gasdrilling in the Appalachian Basin: Analysis of the Chesapeake EnergyCorporation dataset. Appl. Geochem. 2015, 63 (1), 37−57.(6) Darrah, T. H.; Vengosh, A.; Jackson, R. A.; Warner, N. R.;Poreda, R. J. Noble gases identify the mechanisms of fugitive gascontamination in drinking-water wells overlying the Marcellus andBarnett Shales. Proc. Natl. Acad. Sci. U. S. A. 2014, 111 (39), 14076−14081.(7) Fontenot, B. E.; Hunt, L. R.; Hildenbrand, Z. L.; Carlton, D. D.;Oka, H.; Walton, J. L.; Hopkins, D.; Osorio, A.; Bjorndal, B.; Hu, Q.H.; Schug, K. A. An evaluation of water quality in private drinkingwater wells near natural gas extraction sites in the Barnett Shaleformation. Environ. Sci. Technol. 2013, 47 (17), 10032−10040.(8) Li, H.; Carlson, K. H. Distribution and origin of groundwatermethane in the Wattenberg oil and gas field of northern Colorado.Environ. Sci. Technol. 2014, 48 (3), 1484−1491.(9) Warner, N. R.; Kresse, T. M.; Hays, P. D.; Down, A.; Karr, J. D.;Jackson, R. B.; Vengosh, A. Geochemical and isotopic variations inshallow groundwater in areas of the Fayetteville Shale development,north-central Arkansas. Appl. Geochem. 2013, 35 (1), 207−220.(10) McMahon, P. B.; Caldwell, R. R.; Galloway, J. M.; Valder, J. F.;Hunt, A. G. Quality and age of shallow groundwater in the BakkenFormation production area, Williston Basin, Montana and NorthDakota. Groundwater 2015, 53 (S1), 81−94.(11) Hildenbrand, Z. L.; Carlton, D. D.; Meik, J. M.; Taylor, J. T.;Fontenot, B. E.; Walton, J. L.; Henderson, D.; Thacker, J. B.; Korlie, S.;Whyte, C. J.; Hudak, P. F.; Schug, K. A. A reconnaissance analysis ofgroundwater quality in the Eagle Ford shale region reveals two distinctbromide/chloride populations. Sci. Total Environ. 2016, 575 (1), 672−680.(12) Nicot, J.-P.; Larson, T.; Darvari, R.; Mickler, P.; Slotten, M.;Aldridge, J.; Uhlman, K.; Costely, R. Controls on methane occurrencein shallow aquifers overlying the Haynesville Shale Gas Field, EastTexas. Groundwater 2017, DOI: 10.1111/gwat.12500.(13) Drollette, B. D.; Hoelzer, K.; Warner, N. R.; Darrah, T. H.;Karatum, O.; O’Connor, M. P.; Nelson, R. K.; Fernandez, L. A.;Reddy, C. M.; Vengosh, A.; Jackson, R. B.; Elsner, M.; Plata, D. L.Elevated levels of diesel range organic compounds in groundwater nearMarcellus gas operations are derived from surface activities. Proc. Natl.Acad. Sci. U. S. A. 2015, 112 (43), 13184−13189.(14) Eltschlager, K. K., Hawkins, J. W., Ehler, W. C., Baldassare, F.Technical measures for the investigation and mitigation of fugitivemethane hazards in areas of coal mining. U.S. Department Interior Office

Figure 5. Distance of sampled water wells to nearest unconventionalhydrocarbon (UOG) well (circles) or hydrocarbon well of any kind(triangles). Samples are color-coded according to whether they had adetection of selected hydrocarbons. In general, only UOG wells werepresent in the Fayetteville study area. Location data for hydrocarbonwells are listed in SI Table S2.

Environmental Science & Technology Article

DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

G

of Surface Mining. 2001; Pittsburgh, PA; http://www.osmre.gov/resources/blasting/docs/MineGasesDust/Methane.pdf (accessed No-vember 29, 2016.(15) U.S. Environmental Protection Agency, Table of regulatedd r i n k i n g wa t e r c o n t am i nan t s . h t t p s : / /www . ep a . g o v /dwstandardsregulations (accessed September 19, 2016).(16) Landon, M. K.; Belitz, K. Geogenic sources of benzene inaquifers used for public supply, California. Environ. Sci. Technol. 2012,46 (16), 8689−8697.(17) IHS Global, U.S. Well History and Production Database−2015:IHS Global, Englewood, Colo.. http://www.ihsenergy.com. (accessedDecember 13, 2016).(18) Jurgens, B. C., Bohlke, J. K., Eberts, S. M., 2012, TracerLPM(Version 1): An Excel® workbook for interpreting groundwater agedistributions f rom environmental tracer data. Techniques and MethodsReport 4-F3; U.S. Geological Survey: VA, 2012.(19) Helsel, D. R., Hirsch, R. M. Statistical Methods in WaterResources; Elsevier: New York, 1992.(20) Grossman, E. L.; Hahn, R. W.; Fritz, S. J. Origin of gaseoushydrocarbons in the Sparta Aquifer in Brazos and Burleson Counties,Texas. Trans. Gulf Coast Assoc. Geol. Soc. 1986, 36 (1), 457−470.(21) Warner, N. R.; Jackson, R. B.; Darrah, T. H.; Osborn, S. G.;Down, A.; Zhao, K.; White, A.; Vengosh, A. Geochemical evidence forpossible natural migration of Marcellus Formation brine to shallowaquifers in Pennsylvania. Proc. Natl. Acad. Sci. U. S. A. 2012, 109 (30),11961−11966.(22) Katz, B. G.; Eberts, S. M.; Kauffman, L. J. Using Cl/Br ratios andother indicators to assess potential impacts on groundwater qualityfrom septic systems: A review and examples from principal aquifers inthe United States. J. Hydrol. 2011, 397 (3−4), 151−166.(23) Palmer, C. A., Oman, C. L., Park, A. J., Luppens, J. A. The U.S.Geological Survey coal quality (COALQUAL) database version 3.0.Data Series 975; U.S. Geological Survey: VA, 2015. https://pubs.er.usgs.gov/publication/ds975 (accessed January 6, 2017).(24) Blondes, M. S., Gans, K. D., Rowan, E. L., Thordsen, J. J., Reidy,M. E., Engle, M. A., Kharaka, Y. F., Thomas, B. U.S. Geological SurveyNational Produced Waters Geochemical Database v2.2 (PROVI-SIONAL). http://energy.usgs.gov/EnvironmentalAspects/EnvironmentalAspectsofEnergyProductionandUse/ProducedWaters.aspx#3822349-data (accessed December 13, 2016).(25) Land, L. S.; McPherson, G. L. Origin of saline formation waters,Cenozoic section, Gulf of Mexico sedimentary basin. AAPG Bull. 1992,76 (9), 1344−1362.(26) Adams, R. L. Basement tectonics and origin of the Sabine Uplift.Gulf Coast Assoc. Geol. Soc. Trans. 2009, 59 (1), 3−19.(27) Zumberge, J.; Ferworn, K.; Brown, S. Isotopic reversal(‘rollover’) in shale gases produced from the Mississippian Barnettand Fayetteville formations. Mar. Pet. Geol. 2012, 31 (1), 43−52.(28) Stolper, D. A.; Lawson, M.; Davis, C. L.; Ferreira, A. A.; SantosNeto, E. V.; Ellis, G. S.; Lewan, M. D.; Martini, A. M.; Tang, Y.;Schoell, M.; Sessions, A. L.; Eiler, J. M. Formation temperatures ofthermogenic and biogenic methane. Science 2014, 344 (6191), 1500−1503.(29) Darling, B. K. Stable isotope forensics as a means of identifyingthe source of methane in a domestic water well, Caddo ParishLouisiana. http://loga.la/wp-content/uploads/2014/09/LOGA-Conference-Paper-BK-Darling-rev.pdf (accessed September 22, 2016).(30) Harrington, J., Whyte, C., Muehlenbachs, K., Darrah, T. Usingnoble gas and hydrocarbon gas geochemistry to source the origin offluids in the Eagle Ford Shale of Texas, USA. AAPG Datapages Searchand Discovery Article #41710. http://www.searchanddiscovery.com/pdfz/documents/2015/41710harrington/ndx_harrington.pdf.html(accessed December 13, 2016).(31) Whiticar, M. J.; Faber, E.; Schoell, M. Biogenic methaneformation in marine and freshwater environments: CO2 reduction vs.acetate fermentation−Isotopic evidence. Geochim. Cosmochim. Acta1986, 50 (5), 693−709.

(32) Schoell, M. The hydrogen and carbon isotopic composition ofmethane from natural gases of various origins. Geochim. Cosmochim.Acta 1980, 44 (5), 649−661.(33) Conrad, M. E.; Templeton, A. S.; Daley, P. F.; Alvarez-Cohen,L. Isotopic evidence for biological controls on migration of petroleumhydrocarbons. Org. Geochem. 1999, 30 (8), 843−859.(34) Revesz, K.; Coplen, T. B.; Baedecker, M. J.; Glynn, P. D.Methane production and consumption monitored by stable H and Cisotope ratios at a crude oil spill site, Bemidji, Minnesota. Appl.Geochem. 1995, 10 (5), 505−516.(35) Richnow, H. H.; Zengler, K.; Gehre, M.; Michaelis, W.; Widdel,F. Methane formation during the degradation of n-alkanes and totalcrude oils. J. Conf. Abstr. Goldschmidt 2000, 5 (2), 844.(36) Warwick, P. D., SanFilipo, J. R., Karlsen, A. W., Barker, C. E.et al. Results of Coalbed Methane Drilling in Panola County, TX, Open-File Rep 2005−1046; U.S. Geological Survey: VA, 2005.(37) McIntosh, J. C.; Warwick, P. D.; Martini, A. M.; Osborn, S. G.Coupled hydrology and biogeochemistry of Paleocene-Eocene coalbeds, northern Gulf of Mexico. Geol. Soc. Am. Bull. 2010, 122 (7/8),1248−1264.(38) Carlson, D. Lignite’s occurrence within aquifers of northernLouisiana. Proc Baton Rouge Geol Soc, First Annual GroundwaterSymposium 2007, 1 (1), 43−53.(39) Wen, T.; Castro, M. C.; Nicot, J. P.; Hall, C. M.; Larson, T.;Mickler, P.; Darvari, R. Methane sources and migration mechanisms inshallow groundwaters in Parker and Hood Counties, Texas−A heavynoble gas analysis. Environ. Sci. Technol. 2016, 50 (21), 12012−12021.(40) Pfeiffer, D. S.; Sharp, J. M. Subsurface temperature distributionsin south Texas. Gulf Coast Assoc. Geol. Soc. Trans. 1989, 39 (1), 231−245.(41) U.S. Energy Information Administration, Update to the EIAEagle Ford Play Maps. http://www.eia .gov/maps/pdf/EIA%20Eagle%20Ford%20Play%20update%2012-29-14.pdf. (accessedJanuary 30, 2017).(42) Zogorski, J. S., Carter, J. M., Ivahnenko, T., Lapham, W. W.,Moran, M. J., Rowe, B. L., Squillace, P. J., Toccalino, P. L. The Qualityof Our Nation’S Waters−volatile Organic Compounds in the Nation’SGroundwater and Drinking Water Supplies, Circular 1292; U.S.Geological Survey: VA, 2006.(43) Arnold, T. L., DeSimone, L. A., Bexfield, L. M., Lindsey, B. D.,Barlow, J. R., Kulongoski, J. T., Musgrove, M., Kingsbury, J. A., Belitz,K. Groundwater Quality Data from the National Water-QualityAssessment Project, May 2012 through December 2013. Data Series997; U.S. Geological Survey:VA, 2016.(44) Michel, R. M. Tritium Deposition in the Continental United States,1953−1983, Water-Resources Investigations Report 89-4072; U.S.Geological Survey:VA, 1989.(45) McMahon, P. B.; Plummer, L. N.; Bohlke, J. K.; Shapiro, S. D.;Hinkle, S. R. A comparison of recharge rates in aquifers of the UnitedStates based on groundwater-age data. Hydrogeol. J. 2011, 19 (4),779−800.(46) National Oceanic and Atmospheric Administration, 1981−2010climate normals. https://www.ncdc.noaa.gov/data-access/quick-links#normals (accessed September 25, 2016).(47) Pearson, F. J.; White, D. E. Carbon 14 ages and flow rates ofwater in Carrizo Sand, Atascosa County, Texas. Water Resour. Res.1967, 3 (1), 251−261.(48) Plummer, L. N.; Eggleston, J. R.; Andreasen, D. C.;Raffensperger, J. P.; Hunt, A. G.; Casile, G. C. Old groundwater inparts of the upper Patapsco aquifer, Atlantic Coastal Plain, Maryland,USA: evidence from radiocarbon, chlorine-36, and helium-4. Hydro-geol. J. 2012, 20 (7), 1269−1294.(49) McMahon, P. B.; Thomas, J. C.; Hunt, A. G. Groundwater agesand mixing in the Piceance Basin natural gas province, Colorado.Environ. Sci. Technol. 2013, 47 (23), 13250−13257.(50) Schellinger, D., Iqgal, J., Carney, B. Testing your water well forsafety. Louisiana State University Research and Extension; http://www.lsuagcenter.com/nr/rdonlyres/20ab91b7-7efd-4d0b-a2eb-11e1bddaedf3/18042/watertest.pdf (accessed December 13, 2016).

Environmental Science & Technology Article

DOI: 10.1021/acs.est.7b00746Environ. Sci. Technol. XXXX, XXX, XXX−XXX

H