Scholarly article on topic 'Geochemical and isotopic variations in shallow groundwater in areas of the Fayetteville Shale development, north-central Arkansas'

Geochemical and isotopic variations in shallow groundwater in areas of the Fayetteville Shale development, north-central Arkansas Academic research paper on "Earth and related environmental sciences"

CC BY
0
0
Share paper
Academic journal
Applied Geochemistry
Keywords
{}

Abstract of research paper on Earth and related environmental sciences, author of scientific article — Nathaniel R. Warner, Timothy M. Kresse, Phillip D. Hays, Adrian Down, Jonathan D. Karr, et al.

Abstract Exploration of unconventional natural gas reservoirs such as impermeable shale basins through the use of horizontal drilling and hydraulic fracturing has changed the energy landscape in the USA providing a vast new energy source. The accelerated production of natural gas has triggered a debate concerning the safety and possible environmental impacts of these operations. This study investigates one of the critical aspects of the environmental effects; the possible degradation of water quality in shallow aquifers overlying producing shale formations. The geochemistry of domestic groundwater wells was investigated in aquifers overlying the Fayetteville Shale in north-central Arkansas, where approximately 4000 wells have been drilled since 2004 to extract unconventional natural gas. Monitoring was performed on 127 drinking water wells and the geochemistry of major ions, trace metals, CH4 gas content and its C isotopes (δ13CCH4), and select isotope tracers (δ11B, 87Sr/86Sr, δ2H, δ18O, δ13CDIC) compared to the composition of flowback-water samples directly from Fayetteville Shale gas wells. Dissolved CH4 was detected in 63% of the drinking-water wells (32 of 51 samples), but only six wells exceeded concentrations of 0.5mg CH4/L. The δ13CCH4 of dissolved CH4 ranged from −42.3‰ to −74.7‰, with the most negative values characteristic of a biogenic source also associated with the highest observed CH4 concentrations, with a possible minor contribution of trace amounts of thermogenic CH4. The majority of these values are distinct from the reported thermogenic composition of the Fayetteville Shale gas (δ13CCH4 =−35.4‰ to −41.9‰). Based on major element chemistry, four shallow groundwater types were identified: (1) low (<100mg/L) total dissolved solids (TDS), (2) TDS>100mg/L and Ca–HCO3 dominated, (3) TDS>100mg/L and Na–HCO3 dominated, and (4) slightly saline groundwater with TDS>100mg/L and Cl>20mg/L with elevated Br/Cl ratios (>0.001). The Sr (87Sr/86Sr=0.7097–0.7166), C (δ13CDIC =−21.3‰ to −4.7‰), and B (δ11B=3.9–32.9‰) isotopes clearly reflect water–rock interactions within the aquifer rocks, while the stable O and H isotopic composition mimics the local meteoric water composition. Overall, there was a geochemical gradient from low-mineralized recharge water to more evolved Ca–HCO3, and higher-mineralized Na–HCO3 composition generated by a combination of carbonate dissolution, silicate weathering, and reverse base-exchange reactions. The chemical and isotopic compositions of the bulk shallow groundwater samples were distinct from the Na–Cl type Fayetteville flowback/produced waters (TDS ∼10,000–20,000mg/L). Yet, the high Br/Cl variations in a small subset of saline shallow groundwater suggest that they were derived from dilution of saline water similar to the brine in the Fayetteville Shale. Nonetheless, no spatial relationship was found between CH4 and salinity occurrences in shallow drinking water wells with proximity to shale-gas drilling sites. The integration of multiple geochemical and isotopic proxies shows no direct evidence of contamination in shallow drinking-water aquifers associated with natural gas extraction from the Fayetteville Shale.

Academic research paper on topic "Geochemical and isotopic variations in shallow groundwater in areas of the Fayetteville Shale development, north-central Arkansas"

ELSEVIER

Contents lists available at SciVerse ScienceDirect

Applied Geochemistry

journal homepage: www.elsevier.com/locate/apgeochem

Applied Geochemistry

Geochemical and isotopic variations in shallow groundwater in areas of the Fayetteville Shale development, north-central Arkansas q

Nathaniel R. Warnera, Timothy M. Kresseb, Phillip D. Haysc, Adrian Downa, Jonathan D. Karrd, Robert B. Jackson a, Avner Vengosh

a Division of Earth and Ocean Sciences, Nicholas School of the Environment, Duke University, Durham, NC 27708, USA b U.S. Geological Survey, Arkansas Water Science Center, Little Rock, AR 72211, USA c U.S. Geological Survey, Arkansas Water Science Center, Fayetteville, AR 72701, USA d Duke Environmental Stable Isotope Laboratory, Duke University, Durham, NC 27708, USA

ARTICLE INFO ABSTRACT

Exploration of unconventional natural gas reservoirs such as impermeable shale basins through the use of horizontal drilling and hydraulic fracturing has changed the energy landscape in the USA providing a vast new energy source. The accelerated production of natural gas has triggered a debate concerning the safety and possible environmental impacts of these operations. This study investigates one of the critical aspects of the environmental effects; the possible degradation of water quality in shallow aquifers overlying producing shale formations. The geochemistry of domestic groundwater wells was investigated in aquifers overlying the Fayetteville Shale in north-central Arkansas, where approximately 4000 wells have been drilled since 2004 to extract unconventional natural gas. Monitoring was performed on 127 drinking water wells and the geochemistry of major ions, trace metals, CH4 gas content and its C isotopes (613CCH4), and select isotope tracers (511B, 87Sr/86Sr, S2H, S18O, 513Cdic) compared to the composition of flowback-water samples directly from Fayetteville Shale gas wells. Dissolved CH4 was detected in 63% of the drinking-water wells (32 of 51 samples), but only six wells exceeded concentrations of 0.5 mg CH4/L. The S13CCH4 of dissolved CH4 ranged from -42.3%c to -74.7%e, with the most negative values characteristic of a biogenic source also associated with the highest observed CH4 concentrations, with a possible minor contribution of trace amounts of thermogenic CH4. The majority of these values are distinct from the reported thermogenic composition of the Fayetteville Shale gas (S13CCH4 = -35.4%c to -41.9%c). Based on major element chemistry, four shallow groundwater types were identified: (1) low (<100 mg/L) total dissolved solids (TDS), (2) TDS > 100 mg/L and Ca-HCO3 dominated, (3) TDS > 100 mg/L and Na-HCO3 dominated, and (4) slightly saline groundwater with TDS >100 mg/L and Cl > 20 mg/L with elevated Br/Cl ratios (>0.001). The Sr (87Sr/86Sr = 0.7097-0.7166), C (513Cdic = -21.3%0 to -4.7%0), and B (511B = 3.9-32.9%c) isotopes clearly reflect water-rock interactions within the aquifer rocks, while the stable O and H isotopic composition mimics the local meteoric water composition. Overall, there was a geochemical gradient from low-mineralized recharge water to more evolved Ca-HCO3, and higher-mineralized Na-HCO3 composition generated by a combination of carbonate dissolution, silicate weathering, and reverse base-exchange reactions. The chemical and isotopic compositions of the bulk shallow groundwater samples were distinct from the Na-Cl type Fayetteville flowback/produced waters (TDS ~10,000-20,000 mg/L). Yet, the high Br/Cl variations in a small subset of saline shallow groundwater suggest that they were derived from dilution of saline water similar to the brine in the Fayetteville Shale. Nonetheless, no spatial relationship was found between CH4 and salinity occurrences in shallow drinking water wells with proximity to shale-gas drilling sites. The integration of multiple geochemical and isoto-pic proxies shows no direct evidence of contamination in shallow drinking-water aquifers associated with natural gas extraction from the Fayetteville Shale.

© 2013 The Authors. Published by Elsevier Ltd. All rights reserved.

CrossMark

Article history: Received 3 December 2012 Accepted 25 April 2013 Available online 14 May 2013 Editorial handling by R. Fuge

q This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. * Corresponding author. Tel.: +1 (919) 681 8050; fax: +1 (919) 684 5833. E-mail address: vengosh@duke.edu (A. Vengosh).

1. Introduction

The combined technological development of horizontal drilling and hydraulic fracturing has enabled the extraction of hydrocarbons from unconventional sources, such as organic-rich shales, and is reshaping the energy landscape of the USA (Kargbo et al.,

0883-2927/$ - see front matter © 2013 The Authors. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016Zj.apgeochem.2013.04.013

2010; Kerr, 2010). Unconventional natural gas currently supplies ~20% of US domestic gas production and is projected to provide ~50% by 2035 (USEIA, 2010). Therefore, ensuring that unconventional natural gas resource development results in the minimal possible negative environmental impacts is vital, not only for domestic production within the USA, but also for establishing guidance for worldwide development of shale gas resources. Recent work in the Marcellus Shale basin demonstrated a relationship between CH4 concentrations in shallow groundwater and proximity of drinking water wells to shale-gas drilling sites in northeastern Pennsylvania, suggesting contamination of shallow groundwater by stray gas (Osborn et al., 2011a). In addition, a previous study has shown evidence for natural pathways from deep formations to shallow aquifers in northeastern Pennsylvania that may allow leakage of gas or brine, and might pose a potential threat to groundwater in areas of shale gas extraction (Warner et al., 2012). While previous studies have focused on the Pennsylvania and New York portion of the northern Appalachian Basin, many other shale-gas basins currently are being developed that have not been examined for potential effects on water quality. One of the critical aspects of potential contamination of shallow aquifers in areas with shale-gas development is the hydraulic connectivity between shale and other deep formations and overlying shallow drinking water aquifers. Here the quality and geochemistry of shallow groundwater directly overlying the Fayetteville Shale (FS) in north-central Arkansas is investigated. The Fayetteville Shale is an unconventional natural gas reservoir with an estimated total production of 906 billion m3 (USEIA, 2011). Since 2004, approximately 4000 shale-gas wells have been drilled there, including both vertical wells and, more recently, horizontal wells.

In this study, water samples from 127 shallow domestic wells in the Hale, Bloyd and Atoka Formations in north-central Arkansas and six flowback/produced water samples from the underlying FS were analyzed in an attempt to identify possible groundwater contamination. Five of the produced water samples were collected within 21 days of fracturing (i.e., defined as flowback water) and a single sample was collected at about a year following hydraulic fracturing (i.e., defined as produced water). The concentrations of major anions (Cl, SO4, NO3, Br, and dissolved inorganic C [DIC]), cations (Na, Ca, Mg and Sr), trace elements (Li and B), and for a smaller subset of samples dissolved CH4 and selected isotopic tracers (511B, 87Sr/86Sr, S2H, S18O, 513Cdic, and S13CCH4) were determined. Using multiple geochemical and isotopic tracers together with their geo-spatial distribution provides a multidimensional approach to examine potential groundwater contamination in areas of shale gas development. it is hypothesized that shallow groundwater could be contaminated by stray gas migration, possibly associated with poor well integrity, similar to earlier studies (Osborn et al., 2011a). Shallow drinking water could also be contaminated with deeper saline fluids at the same time as the stray gas migration associated with drilling. A third possibility would be natural migration and connectivity between the shallow drinking water aquifers and deeper, higher salinity formation waters through faults or other more permeable pathways (Warner et al., 2012). This study, in conjunction with a United States Geological Survey (USGS) report using the same major element data (Kresse et al., 2012), are to the authors' knowledge the first to report such a comprehensive geochemical evaluation of possible shallow groundwater contamination outside the Marcellus Shale basin (Osborn et al., 2011a; Warner et al., 2012).

2. Geologic setting

The study area is located within the currently active development area for the FS in north-central Arkansas with the majority

of samples collected in Van Buren County and the northern part of Faulkner County (Fig. 1). The area is characterized by a rugged and mountainous landscape to the north and rolling hills to the south, spanning the southern area of the Ozark Mountains, to the northern Arkansas River valley (Imes and Emmett, 1994). The bedrock in the study area comprises the Pennsylvanian-age Hale, Bloyd and Atoka Formations, which are composed of shale with interbedded minor occurrences of relatively permeable sandstone, limestone and coal (Cordova, 1963) (Fig. 2). The shale portion of the Atoka underlies the lowlands because of its lack of resistance to weathering (Cordova, 1963), and thin beds of coal are present throughout but limestone is only present in the north of the study area (Cordova, 1963). The Mississippian-age Fayetteville Shale is the target formation during drilling and lies approximately 5002,100 m below the ground surface (mbgs), with the southern portion of the study area being the deepest. These formations are part of the Western interior Confining System with groundwater flow restricted to the weathered and fractured upper 100 m of bedrock (imes and Emmett, 1994). No one formation within this confining system, even where used for a drinking-water supply, forms a distinct aquifer regionally, and the regional designation as a confining unit indicates that on a regional scale these formations impede the vertical flow of water and confine the underlying aquifers. Domestic wells in the area typically provide limited groundwater yields (imes and Emmett, 1994). The average reported drinking water well depth is 26 m and minimum and maximum of 7.8 m and 120 m, respectively. Wells drilled deeper than 100 m revealed a much more compacted and less permeable section of the formations (imes and Emmett, 1994).

The underlying Fayetteville Shale production zone is ~17-180 m thick and occupies an area of approximately 6500 km2; the area of groundwater samples for this study covered approximately 1/3 of the area of the production zone. The density of shale-gas drilling varied widely across the study area. For the set of drinking-water samples, the total number of unconventional shale-gas wells within 1 km (as measured from the well-head) of a given home ranged from zero to over 14 natural gas wells. This well density represents an area of moderate to intense unconventional shale-gas development similar to other areas of extensive shale gas developments, such as in NE Pennsylvania (Osborn et al., 2011a). importantly, the Fayetteville Shale is the first oil and gas development in this study area. With no records that indicated the presence of historical conventional wells, which may provide possible conduits for vertical migration of stray gas and/ or hydraulic fracturing fluids in other shale-gas plays. Saline water unsuitable for human consumption was identified between 150 and 600 mbgs but generally is at least 300 mbgs in the study area (imes and Emmett, 1994).

The exposed and shallow subsurface geologic formations serving as local aquifers for Van Buren and Faulkner Counties are a series of dominantly sandstone and shale units of the Hale, Bloyd and Atoka Formations (Fig. 2). Subsurface geology, particularly with respect to lateral facies within the Fayetteville Shale, was poorly defined prior to development of gas, and most of the detailed stratigraphic and reservoir analysis were held as proprietary by the companies operating there.

The Fayetteville Shale is a black, fissile, concretionary shale, which contains pyrite and silica replacement fossils in some intervals. The Fayetteville Shale dips from north to south (Fig. 2). The highly organic-rich facies within the Fayetteville Shale is present in the middle and lower part of the formation. Vitrinite reflectance falls within 1.93-5.09%, which corresponds to the dry gas window (imes and Emmett, 1994).

The Hale Formation is made up of two members: the lower Cane Hill Member, which is typically composed of silty shale inter-bedded with siltstone and thin-bedded, fine-grained sandstone,

Fig. 1. Study site location in north-central Arkansas. Unconventional shale-gas wells completed into the Fayetteville Shale are shown in black. Shallow groundwater samples were cataloged based on major element chemistry into four water categories: low-TDS (beige triangles), Ca-HCO3 (blue circles), Na-HCO3 (green squares), and Cl > 20 mg/L (red diamonds).

and the upper Prairie Grove Member composed of thin to massive and Emmett, 1994). The Cane Hill Member of the Hale Formation is limey sandstone. The Hale Formation thickness is up to 90 m (Imes exposed in the extreme northern part of Van Buren County (Fig. 2).

Fig. 2. (a and b) Map of sample locations and bedrock geology in the study area of north-central Arkansas. The majority of samples were collected from the Atoka (southern area) and Hale Formations (northern area). North-to-south geological cross-section in the study area (A-A' line is shown). Geological units gently dip to the south with the Atoka Formation outcropping in the southern portion of the study area. The underlying Fayetteville Shale shoals to the north.

i/I tfl

5 i/> 2 DC

0.001 0.001

CRDS 613C-CH4%o

0.01 0.1 1 10 CRDS Dissolved CH4 ppm

Fig. 3. (a and b) Comparison of results for duplicate samples submitted to a private laboratory (Isotech) by gas chromatography isotope ratio mass spectrometry (GC-IRMS) to those performed using a cavity-ring-down spectroscopy (CRDS) analyzer at Duke University for both dissolved CH4concentrations and S13C-CH4. Dissolved CH4concentrations from the two independent methods showed a good correlation (r2 = 0.90, p < 1 x 10~15) with some variability at higher concentrations. The comparison of the S13C-CH4 values obtained from the two analytical techniques showed a strong correlation (r2 = 0.95, p <1 x 10~15). The CRDS methodology showed some bias at lower S13C-CH4 compared to the private laboratory. Note that this comparison includes samples from other study areas to cover a wide range of concentrations and S13C-CH4 values.

The Bloyd Formation in northwestern Arkansas is formally divided into five members, two of which are limestone members absent in the study area. The lower two thirds of the Bloyd Formation consists dominantly of very thin- to thinly-bedded sandstone with shale interbeds. The upper Bloyd is dominantly a shale with inter-bedded sandstone that is commonly calcareous; the sandstone units can reach a thickness of up to 24 m (Imes and Emmett, 1994). Total thickness for the Bloyd can exceed 120 m in the study area. Exposures of the Bloyd Formation are found in northern Van Buren County (Fig. 2).

The Atoka Formation in the study area consists of a sequence of thick shales that are interbedded with typically thin-bedded, very-fine grained sandstone. The Atoka Formation is unconformable with the underlying Bloyd Formation with a thickness of up to 7500 m in the Ouachita Mountains (Imes and Emmett, 1994). The Atoka Formation is exposed throughout the southern portion of the study area (Fig. 2).

3. Materials and methods

All shallow groundwater samples were collected from private drinking water wells by USGS personnel in July and November 2011. Methods for collection of field parameters (pH, temperature, and specific conductance) and water sampling followed standard USGS protocols (Wilde, 2006). These included sampling prior to any holding tanks or filtration, purging water wells until field parameters stabilized, followed by 0.45 im water filtering on site for water samples collected for trace and major ion analyses. Dissolved gas sample collection followed established protocols (Isotech Laboratories, Inc., 2012). Samples of FS water were collected from production wells (flowback or produced waters) by Arkansas Oil and Gas Commission personnel. Samples were labeled flowback waters if collected within 3 weeks of hydraulic fracturing (5 total samples) and produced water if collected more than 3 weeks after fracturing (1 sample; ~50 weeks following fracturing). All water samples were preserved on ice and shipped to Duke University (Durham, North Carolina, USA), where they were refrigerated until analysis.

Samples for major cations, anions, trace metals, and selected isotopes (O, H, B, Sr and C-DIC) were analyzed at Duke University. Isotech Laboratories performed dissolved gas analysis for concentrations of CH4 and higher-chain hydrocarbons on 20 samples using chromatographic separation followed by combustion and dual-inlet isotope ratio mass spectrometry to measure d13CCH4.

Dissolved CH4 concentrations and d13C-CH4 were determined by cavity ring-down spectroscopy (CRDS) (Busch and Busch, 1997) on an additional 31 samples at the Duke Environmental stable Isotope Laboratory (DEVIL) using a Picarro G2112i. Dissolved CH4 concentrations were calculated using headspace equilibration, extraction and subsequent concentration calculation by a modification of the method of Kampbell and Vandegrift (1998). For each 1-L sample bottle, 100 mL of headspace was generated by displacing water with zero air (CH4-free air) injected with gastight syringes equipped with luer-lock valves. Bottles were shaken at 300 rpm for 30 min to equilibrate headspace with dissolved CH4. The equilibrated headspace was then extracted with gastight syringes while replacing the extracted volume of headspace with deionized water. The extracted headspace was then injected into Tedlar bags (Environmental Supply, Durham, NC) equipped with septum valves and introduced into the Picarro model G2112-i CRDS (Picarro, Inc., Santa Clara, CA). In some cases, dilution into a second Tedlar bag with CH4-free air (zero-air) was required to keep the measured concentration in the optimal range for the instrument. Calculated detection limits of dissolved CH4 were 0.002 mg/L water. Reporting limits for reliable d13C-CH4 were 0.1 mg/L, consistent with Isotech Laboratories (Illinois, USA) reporting values. Concentrations and d13C values were also corrected for instrument calibrations using known CH4 standards from Airgas (Durham, NC) and Isometric Instruments (Victoria, BC).

To confirm the accuracy of the CRDS results, a set of 49 field duplicate groundwater samples was collected and analyzed at Isotech using gas chromatography isotope ratio mass spectrometer (GC-IRMS). These groundwater samples were collected from North Carolina, New York, Pennsylvania and Arkansas in order to span a wider range of both concentrations (<0.002 mg/L through values well above saturation ~100 mg/L) and C isotope values (-30%c through -75%c). The comparison of the field duplicates using these two independent methods showed good correlation for concentration (r2 = 0.90; Fig. 3a) and strong correlation for S13CCH4 (r2 = 0.95; Fig. 3b). Relative standard deviation of dissolved CH4 concentrations determined by CRDS on field duplicates was 9.8%. Reproduc-ibility of 513C measurements determined by CRDS for 8 field duplicate samples ranged from a minimum of 0.07%c to a maximum of 1.0%c. Standard deviation of 513C measurements (n = 6) on a laboratory check standard was 0.55%c over the course of the project.

Major anions were determined by ion chromatography, major cations by direct current plasma optical emission spectrometry (DCP-OES), and trace-metals by VG PlasmaQuad-3 inductively cou-

pled plasma mass-spectrometry (ICP-MS). Four replicate samples showed good reproducibility (<5%) for both major and trace element concentrations. Strontium and B isotopes were determined by thermal ionization mass spectrometry (TIMS) on a ThermoFish-er Triton at the TiMS laboratory in Duke University. The average 87Sr/86Sr of the SRM-987 standard measured during this study was 0.710266 ±0.000005 (SD). The average 11B/10B of NIST SRM-951 during this study was 4.0055 ± 0.0015. The long-term standard deviation of 511B in the standard and seawater replicate measurements was 0.5%c. DIC concentrations were determined in duplicate by titration with HCl to pH 4.5. Values of S18O and S2H of water were determined by thermochemical elemental analysis/continuous flow isotope ratio mass spectrometry (TCEA-CFIRMS), using a ThermoFinnigan TCEA and Delta + XL mass spectrometer at DEVIL. S18O and S2H values were normalized to V-SMOW and V-SLAP. The C isotope ratio of dissolved inorganic C detemined after acid digestion, on a ThermoFinnigan (Bremen, Germany) GasBench II feeding a ThermoFinnigan Delta + XL Isotope Ratio Mass Spectrometer (IRMS) in the DEVIL lab. Several mL (volume depending on DIC concentration) of each sample were injected into 11-mL septum vials that had each been pre-dosed with 150 uL phosphoric acid and pre-flushed for 10 min with He at 50 mL/min to remove air background. Raw 513C of resulting CO2 was normalized vs Vienna Pee Dee Belemnite (VPDB) using NBS19, IAEA CO-8 standards, and an internal CaCO3 standard.

Natural gas well locations (representing locations of the vertical portion of the well) were obtained from the Arkansas Oil and Gas Commission database (Arkansas Oil and Gas Commission, 2012). Arkansas Oil and Gas Commission also provided 228Ra and 226Ra values for five flowback and one produced water sample. Historical water data were gathered from the USGS National Water Information System (NWIS) data base for the six counties that comprise the bulk of permitted and active gas production wells: Cleburne, Conway, Faulkner, Independence, Van Buren and White Counties (Fig. 2). The data set includes 43 groundwater samples collected near the study area prior to shale-gas development during 1948 and 1983 (USGS, 2013).

4. Results and discussion

4.1. Geochemical characterization of the shallow groundwater

The 127 shallow groundwater samples were divided into four major water categories (Fig. 1 and Supplementary data). The first category was low-TDS (<100mg/L) and generally low-pH (pH<6.6; n = 54) water. The second was a Ca-HCO3 water

DIC (mg/L)

A Low TDS □ Na-HC03 O Ca-HC03 O Cl>20 mg/L

Fig. 4. 513C-DIC (%«) and DIC (mg/L) in shallow groundwater samples. The average 513C-DIC (-17%» to -20%») in the bulk groundwater indicates the majority of DIC is derived from weathering of silicate minerals that would approach -22%. Meth-anogens in some of the Na-HCO3 waters would leave DIC with elevated residual d13C-DIC (line).

(n = 40), with moderate TDS (100 > TDS < 200 mg/L). The third was a Na-HCO3 water with a wider range of TDS (100 >TDS < 415 mg/L; n = 24). The fourth group was classified as Ca-Na-HCO3 water type with the highest TDS (200 > TDS < 487 mg/L) and slightly elevated Cl (>20 mg/L) and Br/Cl molar ratios >1 x 10-3 (n = 9). The fourth group was identified because the elevated Cl and Br/Cl could potentially indicate contamination from the underlying saline formation water (see description below).

The C isotope ratio of dissolved inorganic C (S13CDIC; n = 81 samples) ranged from -22% to -10% (Supplementary data). The low-TDS and Ca-HCO3 water types had lower DIC concentrations but all water types had similar S13CDIC, while most water samples fell within a narrower and lower range of -20% to -17%. In the Na-HCO3 groundwater a positive correlation was observed between DIC concentrations and S13CDIC values (r2 = 0.49, p <0.05; Fig. 4). The Sr isotope ratios (87Sr/86Sr) varied from 0.7097 to 0.7166 (Fig. 5a). Most of the Ca-HCO3 waters had slightly lower 87Sr/86Sr (mean = 0.71259; n = 12) relative to the Na-HCO3 waters

100 Sr(ng/L)

10,000

• !'■ ■■■

A Low TDS

O Ca-HC03

□ Na-HC03

O CI>20 mg/L

O FS Water

100 B (Mg/L)

10,000

Fig. 5. (a and b) 87Sr/86Sr versus Sr concentration (ig/L) log scale and 5"B% versus B concentration (ig/L) in log scale. The lack of strong Sr and B isotopic relationships exclude possible mixing between the Fayetteville Shale water and the shallow groundwater. Instead the isotopic variations appear to be controlled by weathering and waterrock interaction.

Table 1

Historical water quality results from the USGS National Water Information System database for Cleburne, Conway, Faulkner, Independence, Van Buren, and White Counties, Arkansas.

Latitude Longitude Date Sample pH Specific Temperature Depth of Total Calcium Magnesium Sodium Bicarbonate Chloride Silica Sulfate Nitrate Iron Manganese

sample start conductance (°C) well Dissolved (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L as (lg/L) (lg/L)

collected time (lS/cm) (feet) Solids (mg/L) nitrogen)

351352 914018 9/28/1954 - 6.9 213 - 217 142 21 4.8 17 129 3 14 0.8 0.34 30 -

351352 914018 2/5/1955 - 7.4 223 11 217 121 21 3.8 18 124 3.5 9.4 1.6 1.8 0 -

351352 914018 4/18/1955 - 6.9 213 11 217 120 19 5.7 18 126 3.2 12 0.8 0.16 100 -

350509 924502 1/25/1959 - 7.5 236 - 127 - - - - - - - - - - -

350509 924502 8/25/1959 - 6.9 602 - 127 - - - 63 40 146 - 6.2 1.8 -

350604 924602 1/13/1959 - 8.1 366 - 43.1 - - - - - 62 - - - - -

353123 920028 9/16/1960 - 6.3 311 16.5 74.9 228 33 10 4.7 20 2.8 10 111 0.23 0 -

351026 924438 9/15/1960 - 7.2 371 18.5 33 24 19 8.3 38 110 32 5.7 22 3.4 0 -

352510 924242 9/26/1960 - 6 243 - 50.8 126 8.4 2.8 21 28 24 6.3 22 2.5 0 -

345704 922429 9/26/1960 - 6.8 1210 18 128 790 65 34 145 284 180 5.8 122 0.11 70 -

350510 921150 9/16/1960 - 6.8 1360 17 45.3 1040 51 83 72 88 378 9.1 12 0 0 -

350515 922440 9/16/1960 - 6.3 573 18 22.4 435 21 24 51 33 96 8.6 100 0.16 0 -

350600 922852 1/23/1951 - 8 631 15.5 - - - - - 374 2.5 18 10 -

350432 923214 1/25/1965 - 7.4 422 - 155 263 28 11 48 233 20 22 8.8 0 - 100

350432 923214 10/1/1971 - 6.6 736 21.9 155 411 52 32 53 332 64 12 15 0.11 - 20

350432 923214 3/2/1976 1000 6.9 - - 155 - 41 18 48 251 31 18 13 1400 240

350432 923214 3/2/1976 1005 - - - 155 - 44 21 52 301 33 21 24 120 140

351447 922727 9/26/1960 - 6.7 263 19.5 19.1 192 23 11 15 150 3.2 20 6.4 0.05 80 -

351447 922727 9/30/1953 - 8.1 475 21 19.1 - - - - 270 15 20 0.203 - -

351322 922719 7/24/1997 1000 6.7 378 20.8 416 136 21 5.9 12 - 2.1 29 16 <0.007 6300 600

353323 932618 9/16/1960 - 8 1840 16.5 84.9 1450 95 211 53 980 100 5.6 255 0 0 -

350600 915643 4/15/1955 - 6.8 159 17 42 159 9.7 7.5 9.6 80 6.2 17 5.2 0.14 0 -

351237 914531 6/21/1955 - 8.2 351 16 11 39 8.7 26 210 6 5 0.18 10 -

351947 920453 9/26/1960 - 6.3 154 20 160 7 7.9 4.3 9.4 44 9.5 13 5.6 0.68 0 -

352847 913056 5/19/1964 - 5.9 32 15.5 - 27 107 0.3 2 6 4.5 12 1.2 0.294 10

354055 923020 8/29/1968 - 6.6 202 19 - 121 15 9.1 13 121 1.6 15 4.2 0.11 170

352114 913253 9/1/1983 1415 - 62 19 64 - - - - - 8.8 - - - - -

352114 913352 9/1/1983 1530 - 430 17 - - - - - - 28 - - - - -

352020 913248 9/1/1983 1340 - 895 20.5 318 - - - - - 33 - - - - -

351934 913521 9/1/1983 - - 440 18 65 - - - - - 18 - - - - -

351849 913836 8/30/1983 1245 - 735 22 90 - - - - - 110 - - - - -

351843 913914 8/31/1983 0830 - 930 17 35 - - - - - 140 - - - - -

351822 913834 8/30/1983 1330 - 1260 20 80 - - - - - 260 - - - - -

351725 913358 8/31/1983 1350 - 277 17 60 - - - - - 49 - - - - -

351636 913822 9/20/1983 1515 6.8 390 18 - 217 34 9.5 34 5.4 25 3.4 - 860 70

351606 914232 9/6/1955 - 6.6 158 18.5 141 108 11 6.5 12 86 5 - 3.4 0.43 0 -

352508 913130 9/2/1983 1330 - 610 17 66 - - - - - 38 - - - - -

352414 912949 9/2/1983 - - 297 19 60 - - - - - 7 - - - - -

352351 913107 9/2/1983 0900 - 420 18 100 - - - - - 9.3 - - - - -

352229 913040 9/2/1983 1020 - 280 17 85 - - - - - 70 - - - - -

352256 913124 9/2/1983 0945 - 180 18 154 - - - - - 19 - - - - -

352207 913337 9/2/1983 0745 - 598 22 72 - - - - - 140 - - - - -

352146 913250 9/1/1983 1500 - 540 19.5 176 - - - - - 12 - - - - -

(mean = 0.71543; n = 13). Boron isotope ratios (511B) showed a wide range from 4% to 33%, with a general increase of 511B with B content (Fig. 5b) with no systematic distinction between the water types (p > 0.05). The stable isotope composition of all water types did not show any distinctions (p > 0.05) related to the water composition (Supplementary data) and S18O and S2H variations are consistent with the Local Meteoric Water Line (LMWL) (Kendall and Coplen, 2001) of modern precipitation in the region. This similarity suggests a common meteoric origin, and also indicates that all of the geochemical modifications presented below were induced from water-rock interactions along groundwater flowpaths in the shallow aquifers.

Historical groundwater quality data from in or near the study area from the NWIS data base (Fig. 2) included 43 samples collected prior to shale-gas development between 1948 and 1983 (Table 1). Although collected from the same formations, the majority of historical samples were collected to the east, and only three sampling sites overlapped with the intensely sampled part of the study area (Fig. 2); therefore, a complete statistical comparison to historical data was not possible. However, the reported chemical composition of the water samples collected prior to shale gas development in the area was consistent with the Ca-HCO3 and Na-HCO3 water types, with a predominance of Na-HCO3 water type in the Atoka Formation (Fig. 2) as reported in previous studies (Cordova, 1963). Likewise, the range of concentrations in this study fell within the minimum and maximum reported values in the NWIS (Table 1).

4.2. Methane sources in shallow groundwater

Dissolved CH4 concentrations were determined in 51 of the 127 water samples from wells collected for this study (Supplementary data). Methane was detected (>0.002 mg/L) in 63% of wells (32 of the 51), but only six wells had concentrations >0.5 mg CH4/L, with a single sample point (28.5 mg/L) above the potential recommended action level in the USA [10 mg/L] (Eltschlager et al., 2001) (Fig. 6). Dissolved CH4 concentrations were not higher closer to shale gas wells (Fig. 6 and Supplementary data), nor was any statistical difference (student t-test) apparent between concentrations in groundwater of 32 wells collected within 1 km of shale-gas production and 19 wells >1 km away from gas wells (p > 0.1; Supplementary data).

The C isotope ratios of CH4 (S13CCH4) was measurable in 14 of 51 samples (dissolved CH4 > 0.1 mg/L) and ranged from -42.3% to -74.7% (Fig. 7), but the range in S13CCH4 in the six samples with concentrations greater than 0.5 mg/L was systematically (p <0.01) lower (-57.6% to -74.7%). This result provides evidence for a predominantly biogenic origin of the dissolved gas (i.e., <-55%) (Coleman et al., 1981; Whiticar et al., 1986; Grossman et al., 1989; Whiticar, 1999). Additionally, S13CCH4 of 13 out of 14 samples with measurable S13CCH4 did not overlap the reported values (Zumberge et al., 2012) for Fayetteville Shale production gas (Fig. 7). The only one sample with a S13CCH4 value (-42.3%) that approaches the values reported for shale gas had low CH4 concentration (0.15 mg/L).

Samples with trace (<0.5 mg/L) CH4 concentrations and S13CCH4 values between -42 and -60% could reflect either flux of deep-source thermogenic gas (Schoell, 1980) or a mixture of biogenic and thermogenic gas. The sample with the highest S13CCH4 value (-42.3%) also had a low Cl- concentration (2 mg/L). The combined low Cl- and CH4 concentrations rule out likely contamination from underlying fluids (gas and water) (see discussion below). Further evidence for a biogenic origin of CH4 in the shallow groundwater was provided by the lack of detectable higher chain hydrocarbons (C2 < 0.0005 mol%) in the 20 samples analyzed at the commercial laboratory (C2+ was detected in only 1 of 20 samples analyzed).

a 25 %

u û 20

•a 0) 10

l/l l/l

A Low TDS 0Ca-HC03 ■ Na-HC03 O Cl>20 mg/L

Potential Action Level

■ ■

0 5 10 15

Distance to nearest natural gas well (km)

Fig. 6. Dissolved CH4 concentrations (mg/L) in domestic wells plotted versus distance from the domestic wells to nearest natural gas well. Only one of 51 wells analyzed contained CH4 at concentrations above the potential action level set by the Department of Interior (10 mg/L). There is no statistically significant difference in dissolved CH4concentrations from wells collected within 1 km of a gas well and those collected >1 km from a well. The highest dissolved CH4 concentrations were detected in Na-HCO3 water.

Fig. 7. Histogram of S13C-CH4 (%) values of dissolved CH4 plotted in comparison to published values for Fayetteville Shale produced gas S13C-CH4 (%) (Zumberge et al., 2012). Concentrations of the dissolved CH4 in the studiedshallow groundwater samples are indicated by color. The majority of samples, including all of those at higher CH4 concentrations plot at more negative S13C-CH4 values, indicating that a shallow biogenic origin likely contributes to the formation of CH4. The lone sample that overlaps with Fayetteville Shale values may represent migration of stray production gas, but at very low concentrations.

The single detection of a higher chain hydrocarbon (C2 = 0.0277 mol%) was in a sample with a relatively high C1/C2 -+ ratio = (C1/C2 = 730), consistent with a biogenic source (~1000) (Schoell, 1980; Coleman et al., 1981) (Supplementary data). The distribution of dissolved CH4 concentrations and S13CCH4 values (Fig. 8a) suggest a local, shallow origin of dissolved CH4 unrelated to shale-gas extraction in the vast majority of samples.

If the CH4 was sourced from biogenic processes within the shallow aquifers, the ground water chemistry should provide further

613C-CH4%0

Na-HCOj r2 =0.46

50 100 150 Na (mg/L)

Na-HCOj r2 =0.79

200 400 600 DIC (mg/L)

-10 -12 -14 -16 -18 -20

-22 -80

613C-CH4%o

A LowTDS O Ca-HC03 □ Na-HC03 O CI>20 mg/L

200 400

DIC (mg/L)

Fig. 8. Dissolved CH4(mg/L) versus 513C-CH4 (%o)(8a), Na (mg/L)(8b), and 513C-DIC (%«)(8c) in shallow groundwater samples. The correlations observed between CH4 and Na (r2 = 0.46) and DIC (r2 = 0.79) indicate that the highest CH4 is found in Na-HCO3 groundwater. At the higher DIC and CH4 concentrations the depleted S13C-CH4 indicates that methanogens likely contribute to the formation ofCH4. 513C-DIC versus 513C-CH4(%»)(8d) and DIC (mg/L)(8e) in shallow groundwater. The average 513C-DIC (-17%» to -20%») in the bulk groundwater indicates the majority of DIC is derived from weathering of silicate minerals that would approach -22%. Methanogens in some of the Na-HCO3 waters would leave DIC with elevated residual 513C-DIC (green arrow).

support for its biogenic origin (Aravena et al., 1995). Median dissolved CH4 concentrations were highest in the Na-HCO3 water type, with positive correlations to Na and DIC (r2 = 0.46 and 0.79, respectively; Fig. 8b and c). In addition, the positive correlation between DIC concentrations and 513CDiC values (r2 = 0.49, p <0.05; Fig. 8e) could suggest that methanogenesis is occurring within the formations, perhaps within the minor coal beds (Imes and Em-mett, 1994) under reducing conditions. If the minor concentrations of observed CH4 were sourced from microbial CO2 reduction, generation would be expected of 513CCH4 of -70% to -80% (Coleman et al., 1981) parallel to elevated residual S13CDIC (e.g., >+10%) during CH4 production (Aravena and Wassenaar, 1993). However, in the study the majority of the S13CDIC values are significantly lower (-20% to -17%), demonstrating that methanogens are not the main control of S13CDIC in the aquifer. In the low TDS water only trace levels (CH4 < 0.8 mg/L) of dissolved CH4 were recorded (n = 9) and only two low-TDS samples had detectable higher S13CCH4 (-42.3 and -59.6%). These values could indicate either a minor presence of thermogenic gas in the shallow aquifers or bacterial oxidation (Coleman et al., 1981).

4.3. Water-rock interactions and mixing with external fluids

The geochemical variation from low TDS, Ca-HCO3 and Na-HCO3 water types infer different modes of water-rock interaction. The low-TDS waters could reflect an early stage of groundwater re-

charge without much mineralization induced from water-rock interaction, while the Ca-HCO3 waters suggest dissolution of carbonate minerals in the aquifers. A Na-HCO3 water type typically (e.g., Cheung et al., 2010) indicates silicate weathering and ion exchange processes (e.g., reverse base-exchange reaction). In the majority of the shallow groundwater samples, regardless of the water type, DIC nearly balanced the sum of Na, Ca and Mg concentrations (in equivalent units; Fig. 9). DIC could be generated in the shallow aquifers by weathering of silicate minerals in the shale, dissolution of marine carbonate by H2CO3 produced through oxidation of organic matter, or bacterial SO4 reduction. Silicate weathering would mobilize Na, Ca, Mg and Sr with a radiogenic 87Sr/86Sr signature and B with a wide 511B range from 0%, which would characterize structural B in silicate minerals (Lemarchand and Gaillardet, 2006), to 15-20% in "desorbable" B on marine clay surfaces (Spivack and Edmond, 1987). The S13CDIC value would reflect the isotopic fractionation between DIC species and would be expected to be similar to the composition of the H2CO3 that triggered the silicate weathering (~-22%c). If, instead, dissolution of marine carbonate minerals was occurring, one would expect contributions of Ca, Mg and Sr with a low 87Sr/86Sr (~0.7082) for the Pennsylva-nian-age marine formation (Burke et al., 1982), and B with 511B of a marine carbonate signature (~20%) (Vengosh et al., 1991). Dissolution of marine carbonate would generate HCO- with 513Cdic ~-11%c, assuming a closed system with equal proportions of marine calcite dissolution (S13CDIC ~0%) and H2CO3 (S13CDIC —22%),

A Low TDS

O Ca-HC03

□ Na-HC03

O Cl>20 mg/L

DIC (meq/L)

Fig. 9. The sum of Na, Ca and Mg (meq/L) versus dissolved inorganic C(DIC; meq/L) in shallow groundwater samples. Note that DIC balances the majority of the total cations in shallow groundwater samples across all water types.

and that all DIC-bearing species would be in isotopic equilibrium (McCaffrey et al., 1987). Carbonate dissolution could contribute Ca that would be exchanged with Na from exchange sites on clay minerals, resulting in Na-HCO3 water. In such a scenario, the Ca concentrations would be inversely correlated with Na.

Examining all of these geochemical and isotopic constraints, it is clearly shown that neither of these two mechanisms (i.e., silicate weathering versus marine carbonate dissolution combined with base-exchange reaction) were explicitly consistent with the geo-chemical variations measured in the shallow groundwater in this study. For example, in most of the groundwater samples, including those defined as the Na-HCO3 type, Na was positively correlated with Ca, indicating contributions from both elements that would reflect silicate weathering. In contrast, the most DIC-rich (Fig. 10a) waters showed an inverse relationship between Na and Ca (Fig. 10b) that typically mimics reverse base-exchange reactions. Likewise, all of the water types showed a positive correlation (r2 = 0.79) between Na and B concentrations (Fig. 10c), a combination that could reflect mobilization from exchangeable sites on clay minerals. The most DIC-rich waters have a lower Ca/Na ratio and lower Na relative to B concentration (Fig. 10c), inferring a different source. The 511B of the Na-HCO3 waters (16.5-33%; Fig. 5b) was also consistent with B sourced from exchangeable sites on marine clay minerals.

In contrast, relatively low S13CDIC (-20% to -17%) (Supplementary data) and radiogenic 87Sr/86Sr ratios (0.7097-0.7166) (Fig. 5a) in the majority of the studied groundwater rule out the possibility that marine carbonate dissolution was the major process that controlled the generation of Ca-HCO3 water. Nonetheless, given that the shale in the study area is carbonate-rich (Imes and Emmett, 1994), carbonate dissolution likely contributed Ca and HCO3, with Ca exchanged with Na to generate Na-HCO3 water. Reverse base-exchange reaction would remove Ca and Sr, and the uptake of Sr is not expected to modify its original isotopic ratio (i.e., 87Sr/86Sr ratio of the Pennsylvanian-age marine carbonate). One possible explanation for the high 87Sr/86Sr ratio is that the carbonate in the shale was diagenetically-modified from bacterial SO4 reduction with modified fluids containing radiogenic 87Sr/86Sr and depleted S13CDIC relative to the original composition of the marine carbonates. Given that the groundwater has a radiogenic 87Sr/86Sr ratio (0.7097-0.7166) that is similar to the composition of the local shale formations (Kresse and Hays, 2009), it was concluded that the water chemistry was controlled by both silicate

80 120 Na (mg/L)

Fig. 10. The DIC, Ca and B concentrations versus Na in shallow groundwater samples.

mineral weathering and dissolution by diagenetically modified carbonate cement followed by ion-exchange reactions. Further study is needed to characterize the composition of the carbonate cement and delineate the specific mechanism that has caused evolution of the groundwater into a Na-HCO3 composition.

The fourth shallow groundwater type, the higher-Cl- waters, shows a strong correlation between Cl and Br (r2 = 0.89; Fig. 11a) with a high Br/Cl ratio (>1 x 10-3) that is similar to the elevated Br/Cl in the FS brine (see below). This geochemical composition could be interpreted as mixing of shallow groundwater with underlying formation water, similar to the salinization phenomena observed in NE Pennsylvania (Warner et al., 2012). However, the variations of other dissolved constituents such as B and Sr were not correlated with Cl- (Fig. 11f-h), and their isotopic ratios, including 87Sr/86Sr (Fig. 5a), S13CDIC (Fig. 4), and the majority of 511B (Fig. 5b) were distinctly different from expected mixing relations with the FS brines (Supplementary data). This infers that the composition of the groundwater with (Cl > 20 mg/L) was modified by weathering and water-rock interaction. The ability to delineate the exact saline end-member that generated the saline groundwa-ter is limited.

Finally, neither the Na-HCO3 water type, nor the fourth water type with Cl > 20 mg/L were located closer to shale-gas wells (Supplementary data), which rules out the likelihood of salinization induced from shale gas exploitation and migration of fluids associated with natural gas wells. Instead, a geographical distribution of the water types was observed; the majority of Na-HCO3

—. 10

A Low TDS

O Ca-HC03

■ Na-HC03

O Cl>20 mg/L

O FS Water

1 100 10,000 CI (mg/L)

1 100 10,000 CI (mg/L)

100 10,000 CI (mg/L)

100000

100 10,000 CI (mg/L)

100 10,000 CI (mg/L)

100 10,000 CI (mg/L)

100 10,000 CI (mg/L)

A Low TDS

O Ca-HC03

■ Na-HC03

O Cl>20 mg/L

O FS Water

100 10,000 CI (mg/L)

Fig. 11. The variations of major elements as normalized to Cl- in shallow groundwater and the FS saline water. The composition of the FS water infers modified seawater through evaporation and halite precipitation (high Br/Cl ratio), water-rock interactions (enrichment of Na, Sr, Mg and Ca relative to the expected evaporated seawater curve), followed by dilution with meteoric water. Note that there is no apparent relationship between concentrations of constituents in shallow groundwater and the deeper FS waters. The negative correlation between Cl and DIC indicates that dilution is the main factor for the high DIC in the formation water. The positive correlation of Na/Cl and Ca/ Cl with DIC concentration indicates that Na, Ca and DIC within the FS are likely sourced from carbonate dissolution combined with base-exchange reactions that have modified the original composition of the FS water.

samples were identified in the southern portion of the study area (Fig. 1) and at lower average elevations (Supplementary data), which could indicate increased Na and DIC in the southern portion of the study area, corresponding to a regional groundwater flow and increased water-rock interaction along regional flow paths (Imes and Emmett, 1994; Kresse et al., 2012) and/or greater predominance of shale lithology in the low lying regions (Cordova, 1963).

4.4. The Fayetteville shale flowback and produced waters

The FS flowback and produced water samples (Supplementary data) were saline (TDS ~20,000 mg/L), yet the present data show that the salinity is substantially lower than produced waters from other shale gas basins (e.g., Marcellus brine with TDS ~200,000 mg/L; Table 2). The FS saline water was composed of Na-Cl-HCO3, with a linear correlation (r2 = 0.39) between Cl~ and Br~ and a high Br/Cl ratio (~4 x 10~3 to 7 x 10~3; Fig. 11a). This composition infers modified evaporated seawater (seawater evaporation, salt precipitation, followed by dilution with meteoric water) with Na, Sr, Mg and Ca enrichments relative to the expected evaporated seawater curve (McCaffrey et al., 1987) (Fig. 11a-h). The S18O (-2.1%c to -0.5%c) and S2H (-19.8%c to -15.2%c) of the formation water samples plot to the right of the 52H/518O LMWL (Kendall and Coplen, 2001) (Fig. 12). DIC content was elevated

Table 2

Typical produced water TDS (mg/L) concentrations. Source: Kimball, 2012 citation of USGS produced water database - available at http://energy.cr.usgs.gov/prov/prodwat/ data.htm.

TDS (mg/L) DIC (mg/L)

Fayetteville Shale 25,000 1300a

Barnett Shale 60,000 610b

Woodford Shale 110,000

Haynesville Shale 120,000

Permian basin 140,000

Marcellus Shale 180,000 140b

a This study.

b EPA workshop on hydraulic fracturing - http://www.epa.gov/hfstudy/ 12_Hayes_-_Marcellus_Flowback_Reuse_508.pdf.

6180 (VSMOW %o)

Fig. 12. S2H versus S18O values in shallow groundwater and the Fayetteville Shale

brines. The relationship between S18O and S2H in shallow groundwater is consistent

with the local meteoric water line (LMWL) while the Fayetteville Shale brines plot

to the right of the LMWL and could reflect mixing between depleted S18O and S2H

low-saline water and S18O and S2H-enriched brines.

(800-1800 mg/L) compared to other produced waters in other shale basins in the USA (Table 2), and had a distinctive elevated 513CDiC (-12.7%c to +3.7%c), which may reflect the composition of the injected hydraulic fracturing fluid or methanogenesis. Boron (5nB = 26-30%c; Supplementary data and Fig. 5b) and Sr (87Sr/86Sr = 0.7090-0.7111; Supplementary data and Fig. 5a) isoto-pic fingerprints were different than would be expected for unaltered Mississippian-age evaporated seawater, which would generate 5"B>39%0 and a less radiogenic 87Sr/86Sr ratio of ~0.7082 (Burke et al., 1982). The 226Ra and 228Ra activities were relatively low (14-260 pCi/L) (Supplementary data) compared to Appalachian brines (Rowan et al., 2011) with a 228Ra/226Ra range of 0.1-0.5, which is similar to Appalachian brines. This relatively low Ra level could have important implications for management strategies and evaluation of possible environmental effects, following disposal of the flowback and produced waters.

The chemical composition of the five flowback samples reflects mixing between the original formation water (represented by the produced water) and lower-saline water that was injected as fracturing fluids. Given the higher salinity of the formation water (relative to the injected water) its chemistry overwhelmingly controlled the composition of the flowback waters. Similar results were observed in the composition of flowback water from the Marcellus Formation (Haluszczak et al., 2013). Overall, the combined geochemical data from five flowback and one produced water samples indicate that the FS water is likely the remnant of seawater that evaporated beyond the halite saturation stage (McCaffrey et al., 1987). Similar to the Appalachian brines (Dresel and Rose, 2010; Warner et al., 2012) the evaporated seawater was modified by water-rock interaction that resulted in Na, Sr, Mg and Ca enrichments and alterations of the original marine 511B and 87Sr/86Sr isotopic fingerprints. The brine was subsequently diluted by meteoric water with lower S18O and S2H values that reduced the original salinity to levels lower than seawater (TDS < 32,000 mg/L).

Another unique characteristic of the FS is the substantial DIC enrichment that is inversely correlated (r2 = 0.55) with Cl" content (Fig. 11c) with high 513Cdic values (-12.7%c to 3 .7%c) relative to shallow groundwater (Supplementary data). This suggests that the FS water is diluted with DiC-rich water. The elevated positive S13CDIC could infer methanogenesis in the low-saline water that diluted the original FS brine. Alternatively, dissolution of the limestone matrix with a 513C of ~1.0%c (Handford, 1986) coupled with reverse base-exchange reaction within the FS would generate Ca, Na (from base-exchange) and DIC with a positive S13CDIC signature. This is confirmed by the correlation of Na/Cl and Ca/Cl ratios and inverse correlation of Cl with DIC (Fig. 11g, d and c). Combined, the chemistry and isotopic results indicate a major modification and dilution of the original FS brine composition.

5. Conclusions and Implications

This study examined water quality and hydrogeochemistry in groundwater from shallow aquifers in an attempt to delineate possible groundwater contamination. Three types of potential contamination were considered (1) stray gas contamination; (2) migration of saline fluids from depth that were directly associated with drilling and exploration of the underlying Fayetteville Shale; and (3) natural migration of saline fluids from depth through permeable geological formations. The results of this study clearly show lack of saline fluid contamination (scenario #2) in drinking water wells located near shale gas sites, which is consistent with previous studies in shallow groundwater in the Marcellus in northeastern Pennsylvania (Osborn et al., 2011a; Warner et al., 2012). However, the lack of apparent CH4 contamination with thermo-

/ a Low TDS

° Ca-HC03

Na-HC03

Cl>20 mg/L

J®* ° FS Water

6S>,° -- -LMWL

genic C isotope composition in shallow groundwater near shale gas sites in the Fayetteville Shale differs from results reported for shallow groundwater aquifers overlying the Marcellus Formation (Os-born et al., 2011a). It has been proposed that the stray gas contamination likely resulted from poor well integrity that allowed leakage and migration of CH4 to the shallow aquifers (Jackson et al., 2011; Osborn et al., 2011a,b). In this study no direct evidence was found for stray gas contamination in groundwater wells located near shale gas sites and most of the CH4 identified (mostly low concentrations) had a S13CCH4 composition that is different from the fingerprint of the Fayetteville Shale gas.

Likewise, this study did not find geochemical evidence for natural hydraulic connectivity between deeper formations and shallow aquifers (Warner et al., 2012) that might provide conduits for flow of saline fluids from depth to the shallow groundwater. The spatial distribution of the slightly saline groundwater (Cl > 20 mg/L) that could be derived from dilution of the FS brine or another saline source was not associated with the location of the shale gas wells. Shallow groundwater samples for this study were collected from formations that are part of the Western Interior Confining System (Imes and Emmett, 1994). A previous investigation has shown that these formations impede the vertical flow of groundwater and restrict groundwater movement for domestic supply wells to only local near-surface flow systems (Imes and Em-mett, 1994). The natural impermeability and apparent lack of deformation of these formations seems to prevent hydraulic connectivity that might allow the flow of saline fluids between deep saline formations and shallow drinking water aquifers in north-central Arkansas.

The lack of fracture systems that would enable hydraulic connectivity is very different from the geological formations overlying the Marcellus Shale in the Appalachian basin (Warner et al., 2012 and references therein). These differences could be explained by two structural deformation scenarios: (1) recent glaciation and isostatic rebound of shallow bedrock that was reported in the Appalachian and Michigan basins (Weaver et al., 1995); and (2) tectonic deformation that shaped particularly the Appalachian Basin (Lash and Engelder, 2009). These natural deformation events could explain the increased hydraulic connectivity and pathways that provide conduits for fluids and gas between the deeper production zones and shallow ground-water in the shallow geological formations overlying the Marcellus Shale in the Northern Appalachian Basin but apparently not in the study area in Arkansas.

Previous studies in the Marcellus Basin have suggested that the CH4 leakage to shallow drinking water wells is most likely attributable to poor well integrity (Osborn et al., 2011a). Such human factors could also explain the lack of CH4 contamination in Arkansas, possibly due to: (1) better wellbore integrity; and/or (2) a lack of conventional oil and gas development in north-central Arkansas prior to the shale gas extraction from the Fayetteville Formation (Kresse et al., 2012).

In conclusion, this study demonstrates the importance of basin-and site-specific investigations in an attempt to determine the possible effects of shale gas drilling and hydraulic fracturing on the quality of water resources. The study shows that possible ground-water impacts from shale-gas development differ between basins and variations in both local and regional geology could play major roles on hydraulic connectivity and subsurface contamination processes. Based on the results of this and previous studies (Osborn et al., 2011a; Warner et al., 2012), it is concluded that systematic monitoring of multiple geochemical and isotopic tracers is necessary for assessing possible groundwater contamination in areas associated with shale gas exploration as well as the possible hydraulic connectivity between shallow aquifers and deeper production zones.

Acknowledgments

Funding for this study was provided by the Nicholas School of the Environment and the Center on Global Change at Duke University and by funding to the Nicholas School from Fred and Alice Stanback. Field sampling activities were funded by Shirley Community Development Corporation and Faulkner County, Arkansas. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/jj.apgeochem. 2013.04.013.

References

Aravena, R., Wassenaar, L.I., 1993. Dissolved organic carbon and methane in a regional confined aquifer, southern Ontario, Canada: carbon isotope evidence for associated subsurface sources. Appl. Geochem. 8, 483-493.

Aravena, R., Wassenaar, L.I., Barker, J.F., 1995. Distribution and isotopic characterization of methane in a confined aquifer in southern Ontario, Canada. J. Hydrol. 173, 51-70.

Arkansas Oil and Gas Commission, 2012. Arkansas Online Data System. <http:// www.aogc.state.ar.us/JDesignerPro/JDPArkansas/AR_Welcome.html>.

Burke, W.H., Denison, R.E., Hetherington, E.A., Koepnick, R.B., Nelson, H.F., Otto, J.B., 1982. Variation of seawater 87Sr/86Sr throughout Phanerozoic time. Geology 10, 516-519.

Busch, K., Busch, M., 1997. Cavity ring-down spectroscopy: An ultratrace absorption measurement technique. Am. Chem. Soc. Symp. Series 720, Oxford.

Cheung, K., Klassen, P., Mayer, B., Goodarzi, F., Aravena, R., 2010. Major ion and isotope geochemistry of fluids and gases from coalbed methane and shallow groundwater wells in Alberta, Canada. Appl. Geochem. 25,1307-1329.

Coleman, D.D., Risatti, J.B., Schoell, M., 1981. Fractionation of carbon and hydrogen isotopes by methane-oxidizing bacteria. Geochim. Cosmochim. Acta 45, 10331037.

Cordova, R., 1963. Water Resources of the Arkansas Valley Region, Arkansas. US Geological Survey, Washington, DC.

Dresel, P., Rose, A., 2010. Chemistry and origin of oil and gas well brines in western Pennsylvania. Pennsylvania Geol. Surv., 4th series Open-File Report OFOG 1001.0. Pennsylvania Department of Conservation and Natural Resources.

Eltschlager, K., Hawkins, J., Ehler, W., Baldassare, F., 2001. Technical measures for the investigation and mitigation of fugitive methane hazards in areas of coal mining. U.S. Department of the Interior, Office of Surface Mining Reclamation and Enforcement.

Grossman, E.L., Coffman, B.K., Fritz, S.J., Wada, H., 1989. Bacterial production of methane and its influence on ground-water chemistry in east-central Texas aquifers. Geology 17, 495-499.

Haluszczak, L.O., Rose, A.W., Kump, L.R., 2013. Geochemical evaluation of flowback brine from Marcellus gas wells in Pennsylvania, USA. Appl. Geochem. 28,55-61.

Handford, C., 1986. Facies and bedding sequences in shelf-storm deposited carbonates—Fayetteville Shale and Pitkin Limestone (Mississippian), Arkansas. J. Sed. Petrol. 56, 123-137.

Imes, J., Emmett, L., 1994. Geohydrology of the Ozark Plateaus aquifer system in parts of Missouri, Arkansas, Oklahoma, and Kansas. U.S. Geol. Surv. Prof. Paper 1414D.

Jackson, R.B., Osborn, S.G., Vengosh, A., Warner, N.R., 2011. Reply to Davies: hydraulic fracturing remains a possible mechanism for observed methane contamination of drinking water. Proc. Nat. Acad. Sci. 108, E872.

Kampbell, D.H., Vandegrift, S.A., 1998. Analysis of dissolved methane, ethane, and ethylene in ground water by a standard gas chromatographic technique. J. Chromatog. Sci. 36, 253-256.

Kargbo, D.M., Wilhelm, R.G., Campbell, D.J., 2010. Natural gas plays in the marcellus shale: challenges and potential opportunities. Environ. Sci. Technol. 44, 56795684.

Kendall, C., Coplen, T., 2001. Distribution of oxygen-18 and deuterium in river waters across the United States. Hydrol. Process. 15, 1363-1393.

Kerr, R.A., 2010. Natural gas from shale bursts onto the scene. Science 328,16241626.

Kimball, R., 2012. Key Considerations for Frac Flowback/Produced Water Reuse and Treatment. NJ Water Environment Association Ann. Conf. Atlantic City, NJ, May 2012. <http://www.aaees.org/downloadcenter/2012NJWEAPresentation-RobertKimball.pdf>.

Kresse, T., Hays, P., 2009. Geochemistry, comparative analysis, and physical and chemical characteristics of the thermal waters east of Hot Springs National Park, Arkansas, 2006-09. U.S. Geol. Surv. Scient. Invest. Rep. 2009-5263.

Kresse, T., Warner, N., Hays, P., Down, A., Vengosh, A., Jackson, R., 2012. Shallow groundwater quality and geochemistry in the Fayetteville Shale gas-production

area, north-central Arkansas, 2011. U.S. Geol. Surv. Scient. Invest. Rep. 20125273.

Lash, G.G., Engelder, T., 2009. Tracking the burial and tectonic history of Devonian shale of the Appalachian Basin by analysis of joint intersection style. Geol. Soc. Am. Bull. 121, 265-277.

Lemarchand, D., Gaillardet, J., 2006. Transient features of the erosion of shales in the Mackenzie basin (Canada), evidences from boron isotopes. Earth Planet. Sci. Lett. 245, 174-189.

McCaffrey, M., Lazar, B., Holland, H., 1987. The evaporation path of seawater and the coprecipitation of Br and K with halite. J. Sed. Petrol. 57, 928-937.

Osborn, S.G., Vengosh, A., Warner, N.R., Jackson, R.B., 2011a. Methane contamination of drinking water accompanying gas-well drilling and hydraulic fracturing. Proc. Nat. Acad. Sci. USA 108, 8172-8176.

Osborn, S.G., Vengosh, A., Warner, N.R., Jackson, R.B., 2011b. Reply to Saba and Orzechowski and Schon: Methane contamination of drinking water accompanying gas-well drilling and hydraulic fracturing. Proc. Nat. Acad. Sci. USA 108, E665-E666.

Rowan, E., Engle, M., Kirby, C., Kraemer, T., 2011. Radium content of oil- and gas-field produced waters in the northern Appalachian Basin (USA)— Summary and discussion of data. U.S. Geol. Surv. Scient. Invest. Rep. 2011-5135.

Schoell, M., 1980. The hydrogen and carbon isotopic composition of methane from natural gases of various origins. Geochim. Cosmochim. Acta 44, 649-661.

Spivack, A.J., Edmond, J.M., 1987. Boron isotope exchange between seawater and the oceanic crust. Geochim. Cosmochim. Acta 51, 1033-1043.

US-EIA, 2010. Annual Energy Outlook 2010 with Projections to 2035, Washington, DC. <http://www.eia.doe.gov/oiaf/aeo/>.

US-EIA, 20l1. Review of Emerging Resources: U.S. Shale Gas and Shale Oil Plays. US Dept. of Energy, <http://www.eia.gov/analysis/studies/usshalegas/>.

USGS, 2013. National Water Information System Database, <http:// waterdata.usgs.gov/nwis> (accessed 01.04.13).

Vengosh, A., Kolodny, Y., Starinsky, A., Chivas, A.R., Mcculloch, M.T., 1991. Coprecipitation and isotopic fractionation of boron in modern biogenic carbonates. Geochim. Cosmochim. Acta 55, 2901-2910.

Warner, N.R., Jackson, R.B., Darrah, T.H., Osborn, S.G., Down, A., Zhao, K., White, A., Vengosh, A., 2012. Geochemical evidence for possible natural migration of Marcellus Formation brine to shallow aquifers in Pennsylvania. Proc. Nat. Acad. Sci. USA. http://dx.doi.org/10.1073/ pnas.1121181109.

Weaver, T.R., Frape, S.K., Cherry, J.A., 1995. Recent cross-formational fluid flow and mixing in the shallow Michigan Basin. Geol. Soc. Am. Bull. 107, 697-707.

Whiticar, M.J., 1999. Carbon and hydrogen isotope systematics of bacterial formation and oxidation of methane. Chem. Geol. 161, 291-314.

Whiticar, M.J., Faber, E., Schoell, M., 1986. Biogenic methane formation in marine and freshwater environments: CO2 reduction vs. acetate fermentation—Isotope evidence. Geochim. Cosmochim. Acta 50, 693-709.

Wilde, F., 2006 Collection of water samples (ver. 2.0): U.S. Geological Survey Techniques of Water-Resources Investigations, Book 9 (Chapter A4).

Zumberge, J., Ferworn, K., Brown, S., 2012. Isotopic reversal ('rollover') in shale gases produced from the Mississippian Barnett and Fayetteville formations. Mar. Petrol. Geol. 31, 43-52.