Scholarly article on topic 'Environmental signatures and effects of an oil and gas wastewater spill in the Williston Basin, North Dakota'

Environmental signatures and effects of an oil and gas wastewater spill in the Williston Basin, North Dakota Academic research paper on "Environmental engineering"

CC BY-NC-ND
0
0
Share paper
Academic journal
Science of The Total Environment
OECD Field of science
Keywords
{"Unconventional oil and gas production" / Wastewaters / "Brine spills" / "Tight oil production" / "Williston Basin" / "Bakken Formation" / "Endocrine disrupting activity"}

Abstract of research paper on Environmental engineering, author of scientific article — I.M. Cozzarelli, K.J. Skalak, D.B. Kent, M.A. Engle, A. Benthem, et al.

Abstract Wastewaters from oil and gas development pose largely unknown risks to environmental resources. In January 2015, 11.4ML (million liters) of wastewater (300g/L TDS) from oil production in the Williston Basin was reported to have leaked from a pipeline, spilling into Blacktail Creek, North Dakota. Geochemical and biological samples were collected in February and June 2015 to identify geochemical signatures of spilled wastewaters as well as biological responses along a 44-km river reach. February water samples had elevated chloride (1030mg/L) and bromide (7.8mg/L) downstream from the spill, compared to upstream levels (11mg/L and <0.4mg/L, respectively). Lithium (0.25mg/L), boron (1.75mg/L) and strontium (7.1mg/L) were present downstream at 5–10 times upstream concentrations. Light hydrocarbon measurements indicated a persistent thermogenic source of methane in the stream. Semi-volatile hydrocarbons indicative of oil were not detected in filtered samples but low levels, including tetramethylbenzenes and di-methylnaphthalenes, were detected in unfiltered water samples downstream from the spill. Labile sediment-bound barium and strontium concentrations (June 2015) were higher downstream from the Spill Site. Radium activities in sediment downstream from the Spill Site were up to 15 times the upstream activities and, combined with Sr isotope ratios, suggest contributions from the pipeline fluid and support the conclusion that elevated concentrations in Blacktail Creek water are from the leaking pipeline. Results from June 2015 demonstrate the persistence of wastewater effects in Blacktail Creek several months after remediation efforts started. Aquatic health effects were observed in June 2015; fish bioassays showed only 2.5% survival at 7.1km downstream from the spill compared to 89% at the upstream reference site. Additional potential biological impacts were indicated by estrogenic inhibition in downstream waters. Our findings demonstrate that environmental signatures from wastewater spills are persistent and create the potential for long-term environmental health effects.

Academic research paper on topic "Environmental signatures and effects of an oil and gas wastewater spill in the Williston Basin, North Dakota"

ELSEVIER

Contents lists available at ScienceDirect

Science of the Total Environment

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

Environmental signatures and effects of an oil and gas wastewater spill in the Williston Basin, North Dakota

I.M. Cozzarellia,*, K.J. Skalak a, D.B. Kentb, M.A. Englec, A. Benthem a, A.C. Mumford a, K. Haase a, A. Farag d, D. Harper d, S.C. Nagele, L.R. Iwanowiczf, W.H. Oremc, D.M. Akob a, J.B. Jaeschke a, J. Galloway g, M. Kohler b, D.L. Stoliker b, G.D. Jolly a

a US. Geological Survey, National Research Program, Reston, VA 20192, USA b U.S. Geological Survey, National Research Program, Menlo Park, CA 94025, USA c U.S. Geological Survey, Eastern Energy Resources Science Center, Reston, VA 20192, USA

d US. Geological Survey, Columbia Environmental Research Center, Jackson Field Research Station, Jackson, WY 83001, USA e Department of Obstetrics, Gynecology and Women's Health, University of Missouri, Columbia, MO 65211, USA f U.S. Geological Survey, Leetown Science Center, Kearneysville, WV25430, USA g U.S. Geological Survey, North Dakota Water Science Center, Bismarck, ND 58503, USA

CrossMark

HIGHLIGHTS

GRAPHICAL ABSTRACT

1 UOG wastewater (>11 million liters) spilled into Blacktail Creek, ND in January 2015.

' Elevated Na, Cl, Br, Sr, B, Li, NH4, and hydrocarbons were detected in creek waters.

Geochemical baseline deviations persist months after remediation efforts started.

1 B and Sr concentrations, and Ra activities were up to 15 times background in sediment downstream.

1 Biological impacts include reduced fish survival and estrogenic inhibition downstream.

ARTICLE INFO

ABSTRACT

Article history:

Received 16 September 2016

Received in revised form 21 November 2016

Accepted 22 November 2016

Available online 8 December 2016

Editor: Jay Gan

Keywords:

Unconventional oil and gas production

Wastewaters

Brine spills

Wastewaters from oil and gas development pose largely unknown risks to environmental resources. In January 2015,11.4 M L (million liters) of wastewater (300 g/LTDS) from oil production in the Williston Basin was reported to have leaked from a pipeline, spilling into Blacktail Creek, North Dakota. Geochemical and biological samples were collected in February and June 2015 to identify geochemical signatures of spilled wastewaters as well as biological responses along a 44-km river reach. February water samples had elevated chloride (1030 mg/L) and bromide (7.8 mg/L) downstream from the spill, compared to upstream levels (11 mg/L and <0.4 mg/L, respectively). Lithium (0.25 mg/L), boron (1.75 mg/L) and strontium (7.1 mg/L) were present downstream at 5-10 times upstream concentrations. Light hydrocarbon measurements indicated a persistent thermogenic source of methane in the stream. Semi-volatile hydrocarbons indicative of oil were not detected in filtered samples but low levels, including tetramethylbenzenes and di-methylnaphthalenes, were detected in unfiltered water samples downstream from the spill. Labile sediment-bound barium and strontium concentrations (June 2015)

* Corresponding author at: U.S. Geological Survey, National Research Program, 12201 Sunrise Valley Dr., MS 431, Reston, VA 20192, USA. E-mail address: icozzare@usgs.gov (I.M. Cozzarelli).

http://dx.doi.org/mi 016/j.scitotenv.2016.11.157

0048-9697/Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).

Tight oil production were higher downstream from the Spill Site. Radium activities in sediment downstream from the Spill Site were

Williston Basin up to 15 times the upstream activities and, combined with Sr isotope ratios, suggest contributions from the pipe-

Bakken ftimiaticm line fluid and support the conclusion that elevated concentrations in Blacktail Creek water are from the leaking

Endocrine disrupting actiwty pipeline. Results from June 2015 demonstrate the persistence of wastewater effects in Blacktail Creek several

months after remediation efforts started. Aquatic health effects were observed in June 2015; fish bioassays showed only 2.5% survival at 7.1 km downstream from the spill compared to 89% at the upstream reference site. Additional potential biological impacts were indicated by estrogenic inhibition in downstream waters. Our findings demonstrate that environmental signatures from wastewater spills are persistent and create the potential for long-term environmental health effects.

Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Liquid and solid wastes produced during development of oil and gas resources, particularly those from unconventional oil and gas (UOG) resources that are produced using horizontal drilling and hydraulic fracturing technologies, pose potential, but largely unquantified, risks to environmental resources and the health of organisms, including humans. Although several site investigations have been conducted, (e.g., Akob et al., 2016; Lauer et al., 2016; Rivard et al., 2014; Skalak et al., 2014; Warner et al., 2014), the significant variation in the composition of produced water (defined in Engle et al., 2014, referred to as UOG wastewaters herein) from shale and tight oil formations, (e.g., Akob et al., 2016; Blondes et al., 2014; Engle et al., 2016; Harkness et al., 2015; Rowan et al., 2015), as well as differences in disposal methods, climate, and physiography, have limited our understanding of environmental effects of UOG wastes. Recent extensive increases in hydrocarbon production from the tight oil reservoirs in the Williston Basin, including the Bakken and Three Forks Formations, have resulted in increases in water use (Gallegos and Varela, 2015; Scanlon et al., 2016; Scanlon et al., 2014) and new sources and increased volumes of wastewaters (Engle et al., 2014; Gallegos and Varela, 2015; Horner et al., 2016; Kondash and Vengosh, 2015).

Recent studies have raised concerns about the environmental and human health impacts of UOG wastes and have underscored the substantial knowledge gaps in understanding the ecological and human health hazards associated with waste releases (Elliott et al., 2016; Kassotis et al., 2016a; Kassotis et al., 2016b; Vengosh et al., 2014; Yost et al., 2016). Knowledge gaps include identifying pathways to exposure and the chemicals most likely to pose a hazard (Rogers et al., 2015). Tracking UOG wastewaters through their life-cycle of use has required development of new analytical approaches and modeling tools. Recent studies have shown that 87Sr/86Sr ratios, for example, are useful geo-chemical fingerprints of wastewaters because trapped formation waters can develop distinct isotopic signatures, which are conservative and contain high concentrations of Sr (Lauer et al., 2016; Peterman et al., 2012; Stewart et al., 2015; Warner et al., 2013). Further development of analytical methods for trace levels of organics, stable and radioactive isotopes, and compounds of environmental health concern, including endocrine disruptors, are critical to a better understanding of the impact of releases.

Wastewaters from UOG production can be highly saline (often referred to as brine) and contain a complex mixture of chemicals, including toxic and radioactive elements from the formation and chemical additives from production (Barbot et al., 2013; Engle et al., 2014; Gregory et al., 2011 ), which may enter the environment in a variety of ways including direct disposal or well-stimulation practices (DiGiulio and Jackson, 2016; Skalak et al., 2014; Warner et al., 2014), accidental leaks or spills (Akob et al., 2016; Drollette et al., 2015) or beneficial uses such as de-icing on roadways or agricultural irrigation (Adams, 2011; Engle etal., 2014; Lutzetal., 2013; Skalak et al., 2014). Surface releases of wastewaters have been hypothesized to be the source of trace organic compounds in residential wells in the Marcellus Shale region of Pennsylvania (Drollette et al., 2015). Activities at a wastewater disposal

facility in West Virginia caused changes in stream chemistry and microbiology, and measurable endocrine disrupting potential (Akob et al., 2016; Kassotis et al., 2016a). Inorganic contaminants from brine spills in North Dakota have been found to persist at elevated levels in surface water from months to years (Lauer et al., 2016). In addition to the work conducted by Lauer et al. (2016), two previous studies in the Bakken region on the effects of UOG wastewater in the environment found evidence of brine contamination both in surface and groundwater at a total of four different study site locations (Gleason and Tangen, 2014; Mills et al., 2011). These studies highlight the need to determine both the long-term effects and a more comprehensive understanding of the constituents released into the environment from UOG activities. It has become increasingly clear that management of wastewaters from UOG extraction is a key issue in avoiding environmental damage (Lutz et al., 2013; Vidic et al., 2013).

In North Dakota, the boom in tight oil production ramped up in 2008 and has been associated with an increase in the number of spills at production and disposal sites. An analysis of spills data reports from 2008 to 2015 (detailed methodology presented in the SI Methods) available from the North Dakota Department of Health, revealed that > 8000 spills of fluids were recorded (Fig. 1). Despite the fact that volumes and compositions of materials spilled are often not reported, almost 53 million liters of "brine" spills were documented in the reports, representing the vast majority of spills by volume. The documented spills constituted over 75 million liters of total fluids, 17 million liters of which was classified as oil. Most of the spills have occurred in a relatively few number of watersheds in the northwest corner of the state (Fig. 1). This has led to some watersheds (examined at the HUC 10 scale, (Seaber et al., 1987)) cumulatively receiving over 2.5 M L of reported spills and this is likely a conservative estimate. Lauer et al. (2016) demonstrated an association between oil well density and incidents of wastewater (brine) spills in North Dakota.

In January 2015, a UOG wastewater pipeline leak was reported in northwestern North Dakota that allowed an estimated 11.4 million liters of wastewater to flow into nearby Blacktail Creek. Blacktail Creek is a small tributary of the Little Muddy River, which flows into the Missouri River (Fig. 2). We are using the large magnitude spill into Blacktail Creek to conduct in-depth interdisciplinary studies of the environmental fate and potential health effects of contaminants from UOG waste-water releases that enter a surface-water body. This study aimed to identify wastewater constituents that entered the environment, to understand how these components partitioned between water and sediment in the fluvial environment, and to evaluate some of the potential impacts to ecological or human receptors. Samples collected during two time periods, February and June 2015, indicate the presence of wastewater markers and biological impacts in the river; these results demonstrate the persistence of geochemical alterations six months after the spill was discovered. Partitioning of elements from the waste-water spill onto sediment attenuates contaminant concentrations and decreases rates of aqueous transport but could provide a long-term source which could negatively affect the aquatic ecosystems. This work builds on the initial observations of Lauer et al. (2016), who

Fig. 1. Cumulative spills from oil and gas production on watersheds in North Dakota between January 2008 and June 2015. Watershed volumes were calculated by combining reported values of oil, brine, and unidentified liquid waste accidentally released during drilling operations.

Fig. 2. Map of sampling locations along Blacktail Creek and Little Muddy River (A) tributaries of the Missouri River. Panel (B) shows specific sites located along Blacktail Creek and proximity to the area of the spill (red circle). Specific conductance values were those measured in February 2015 (Table S9). Inset in (A) shows location within North Dakota. Map data from: USGS The National Map: National Boundaries Dataset, National Elevation Dataset, Geographic Names Information System, National Hydrography Dataset, National Land Cover Database, National Structures Dataset, and National Transportation Dataset; U.S. Census Bureau - TIGER/Line; HERE Road Data.

collected and analyzed two samples of stream water from the study site in July 2015 and concluded, assuming a contribution of produced water from the Bakken, that the pipeline fluid had entered the creek.

2. Materials and methods

2.1. Site description

The study area is located along Blacktail Creek and the Little Muddy River just north ofWilliston, North Dakota (ND, Fig. 2a). A leak, reported on January 6, 2015, from a shallow underground pipeline that crosses Blacktail Creek resulted in wastewaters from oil production in the Williston Basin to quickly enter into the stream. The leak was approximately 70 m from the stream on the west side of Blacktail Creek; the creek flows approximately 11 km before it joins the Little Muddy River and another 60 km until it flows into the Missouri River. Remediation efforts began in January 2015 and continue at the pipeline leak site, significantly altering the natural hydrology and disturbing much of the surrounding area. Satellite imagery and field reconnaissance personnel indicated that approximately 10,000 m2 of surface soil and sediment has been removed at the site to a depth of 10-30 cm. The stream has been temporarily impounded at several locations by low headwater dams to collect and skim contaminated water. The ground-water between the Spill Site and 4 km downstream was pumped extensively during remediation in the spring and summer of 2015 to the extent that the stream was largely non-flowing in this region during our June 2015 sampling.

Blacktail Creek is a 3rd order tributary and its watershed is approximately 120 km2 in size. Blacktail Creek joins the Little Muddy whose watershed is 1961 km2 and it enters the Missouri River at Lake Sakakawea (formed by the Garrison Dam). Land use in the watershed is 72% agriculture and 4% residential, with the remaining 24% undeveloped (scrub/wetland) (Homer et al., 2015). Analysis of spills reports revealed that within the watershed (at HUC 10 level, Fig. 1), 200,014 L of brine, 35,246 Lofoil, and 10,819 Lof "other" liquid from 50 other reported spills from 76 new wells had been reported in the previous 7 years. A U.S. Geological Survey stream gage (USGS Site ID 06331000) located on Little Muddy, approximately 22.9 km downstream from the Spill Site, has continuous discharge data since 1954. Average annual discharge for Little Muddy River is approximately 0.92 cubic meters per second (cms) (http://waterdata.usgs.gov/nd/nwis/). Flow observations were collected at a temporary stream gage on Blacktail Creek near Marmon, ND (USGS Site ID 06330515) from March to November 2015 located 2.6 km downstream from the Spill Site. Remediation efforts have greatly affected the hydrology of the creek through the use of dams and groundwater pumping. The average daily discharge from March to November 2015 ranged from 0 to 3.39 cms with an average daily discharge of 0.15 cms. During the remediation period, Blacktail Creek had 129 days of zero flow out of the 256 days that the site recorded data. Significant flow variations in the stream occur longitudinally likely due to groundwater flow through observed paleochannels of coarser material along the bank.

2.2. Site sampling and analysis

Samples of sediment and water were collected on February 9-13 and June 15-28, 2015 at 7 sites along Blacktail Creek and the Little Muddy River (Fig. 2). Four of the sampling sites were located on Black-tail Creek: approximately 0.9 km upstream from the spill (representing background conditions for reference, site "BCR"), 0.1 km downstream from the spill (inside the area being remediated, "Spill Site"), 4.7 km downstream from the spill (just outside the area of remediation, site "4.7 km"), and 7.2 km downstream from the spill near the confluence of the Little Muddy River (site "7.2 km") (Fig. 2b). The three sites on the Little Muddy River were located 4.5 km upstream from the confluence with Blacktail Creek (site "LMR", representing the background

(reference) condition on the Little Muddy), at the USGS stream gage (site ID 06331000) located 22.9 km from the spill (site "22.9 km"), and near the confluence of the Little Muddy and the Missouri river 43.8 km from the spill in the town ofWilliston, ND (site "43.8 km"). Site locations were chosen to avoid the active remediation in progress and to focus on examining the downstream impacts. However, the Spill Site did experience active remediation efforts which likely affected the samples we collected there.

Duplicate samples were collected at two sites (7.2 km and at 22.9 km indicated by (1) and (2) in figures and tables). Fish bioassay studies were conducted at sites BCR, LMR, the Spill Site, and 4.7 km, 7.2 km, and 22.9 km downstream from the spill. Water samples for total dissolved ammonium (NH4) analyses were collected after completion of the fish bioassay studies on June 26, 2015. Total dissolved NH4 refers to the sum of all dissolved ammonium species, including NH+ and aqueous NH3.

In February 2015, the surfaces of Blacktail Creek and the Little Muddy River were frozen and water samples were obtained by first augering through the ice layer (approximately 30-50 cm) as described in the Supplemental Information (SI) Methods and shown in Fig. S1. In June 2015, all filtration was performed on site, using in-line filtration. Detailed methods are described in the SI Methods. Field measurements included specific conductance, pH, and dissolved oxygen (DO). Water samples were collected and analyzed for alkalinity, cations, anions, ammonium, strontium (87Sr/86Sr) isotopes, trace inorganics, nonvolatile dissolved organic carbon (NVDOC), semi-volatile hydrocarbons and UOG production organic additives, low-level light hydrocarbons, ex-tractable hydrocarbons, low molecular weight organic acids (LMWOA), and endocrine disrupting activities as described in the SI Methods. Sediment was collected from the upper 1 to 2 cm of the streambed and analyzed for carbon, hydrogen, nitrogen, sulfur (CHNS); barium, strontium, calcium, radium; extractable hydrocarbons; and for invertebrate bioassays. An unfiltered, acidified (HNO3) sample of the pipeline wastewater was supplied by The North Dakota Department of Health and analyzed for NVDOC, ammonium, 87Sr/86Sr isotopes, cations, and anions using the methods described in the SI Methods. Geo-chemical model computations were conducted using the pipeline and surface water samples to assess saturation indices for key solid and gas phases as described in the SI Methods. Ninety-six hour in situ bioassays were conducted during June 17-21, 2015, with Fathead Minnows (Pimephales promelas, FHM) as described in the SI Methods. Survival was recorded every 24 h and field parameters (water temperature, DO, and specific conductance) were recorded hourly. Water samples for cation and anion analyses were collected daily during the in situ bio-assays to assess total major ion concentrations.

3. Results and discussion

Historical discrete specific conductance values at the Little Muddy River stream gage (Fig. S2) retrieved on June 6, 2016 showed a mean specific conductance of 1858 ^S/cm and a median specific conductance of 2060 |jS/cm. Specific conductance in October 2014 and January 2015 exceeded 3000 ^S/cm, representing the two highest recorded values since 1975, presumably reflecting the pipeline break upstream and spillage of wastewater with high total dissolved solids (TDS) into the river. Data collected for the current study, closer to the pipeline break, showed specific conductance values increased by approximately a factor of two 4.7 km downstream from the spill compared with upstream at BCR in February 2015 (Fig. 2b). Typical specific conductance values for streams in western North Dakota range from about 700 to 3300 |jS/cm (Galloway et al., 2012).

3.1. NVDOC and major inorganic chemistry

The pipeline wastewater had a TDS concentration of at least 300 g/L, which is near the maximum TDS found in produced waters in the

United States but near the median for those from the Bakken Formation (250 g/L) (Blondes et al., 2014) (Table S7). Ammonium was the third most abundant cation (Table S7, S8), assuming near-neutral pH values in the pipeline fluid. The high NH4 concentration determined in the spilled fluid sample is consistent with high concentrations of NH4 in produced water samples from this region as reported elsewhere (Blondes et al., 2014; Lauer et al., 2016). Thus, the spill contributed a large mass (2 million mol, which equals 28 metric tons as N) of labile nitrogen to Blacktail Creek. As is evident from the data compiled in Table S7, concentrations of most solutes are close to their corresponding median values for produced waters from the Bakken Formation compiled in the USGS Produced Water Database (Blondes et al., 2014). One exception is sulfate (SO4), whose concentration of 2900 mg SO4/L is over 6 times the median sulfate concentration in the database but less than the maximum concentration of 11,000 mg SO4/L. The exact reservoir(s) the pipeline fluid was sourced from is not known. Lauer et al. (2016) showed similarities between the composition of two Blacktail Creek samples from this site and produced waters from the Bakken but did not examine other potential sources. During the interval 2013-2015, ~40% of the new wells in the Bakken-Three Forks petroleum system were tight oil wells drilled into the Three Forks Formation (IHS Markit, 2016). There are no known publicly available produced water compositional or isotopic data for the Three Forks Formation. Thus, these discrepancies may be due to all or some of the water being sourced from non-Bakken reservoirs. Geochemical model computations conducted using the composition given in Table S7, a temperature of 25 °C, and model parameters described in the SI, show that the pipeline fluid composition is slightly supersaturated with halite and supersaturated with respect to celestite (SrSO4,), barite (BaSO4), and both anhydrite and gypsum (at equilibrium with each other at this composition).

Fig. 3 shows the concentrations of major chemical components, NVDOC, chloride (Cl) and sodium (Na) in the pipeline sample and surface water samples. The concentration of NVDOC in the pipeline sample was 34 mg/L carbon (C). The upstream sample from Blacktail Creek (BCR) had NVDOC concentrations of 22 mg C/L and 21 mg C/L in

Fig. 3. Chloride (Cl), sodium (Na), and non-volatile dissolved organic carbon (NVDOC) concentrations (mg/L) in February 2015 (A) and June 2015 (B) in the brine pipeline sample and in water samples from surface water collected along Blacktail Creek and the Little Muddy River, as shown in Fig. 2. Field duplicate samples were collected at sites 7.2 and 22.9 km downstream of the Spill Site and are indicated by (1) and (2). ND = not determined.

February and June, respectively. The LMWOA analyses showed low levels of lactate (0.3-0.5 mg/L) and trace amounts of formate (equal to and <0.1 mg/L) in all samples along Blacktail Creek (Table S9C). Because of the high amount of carbon naturally present in the creek, NVDOC is not a useful indicator of the presence of the wastewater in the surface water.

Chloride and Na are present in the pipeline brine at concentrations that are orders of magnitude higher than naturally present in either Blacktail Creek or Little Muddy River (Fig. 3). The background waters contain Cl concentrations below 15 mg/L and Na concentrations between 200 and 400 mg/L in both February and June 2015. At the Spill Site, concentrations of pipeline indicators were low (Fig. 3), due to the active damming and pumping of water at the site. Consequently, the chemistry at the Spill Site was similar to upstream water chemistry at BCR. However, at 4.7 km and 7.2 km downstream from the spill, Cl and Na were substantially elevated, with Cl present at 67 and 72 times the background concentration at BCR in February, respectively. These observations indicate that despite remediation activities, the downstream waters are impacted by the spilled wastewater. Molar Na/Cl ratios of excess Na and Cl (i.e., background-subtracted) in water samples from the 4.7 km site (0.83) from February matches molar Na/Cl ratios for typical Bakken produced water and the pipeline water sample (0.8-0.9), similar to the observations of Lauer et al. (2016) supporting the argument that produced water from the pipeline was the source of the increased salinity during this event. The pipeline Na/Cl signature will be preserved in water downstream from the spill if no more than a small fraction of the large mass of Na derived from the pipeline spill sorbed to stream sediments. Elevated Cl and Na concentrations were observed at sampling locations as far downstream as 22.9 km on the Little Muddy River. In June 2016 the concentration of Cl at 4.7 km and 7.2 km were 18 and 11 times higher than the concentrations at the upstream BCR site, respectively. However, in June 2015, the Cl signature of the spill was no longer detectable at the gage location 22.9 km downstream, indicating dilution of the wastewater signature. Complete chemistry for all samples is available in Table S9.

Using a simple mass balance mixing model for Cl between the BCR and pipeline data (as described in the SI Methods), we estimate that the pipeline fluid contributed 0.6% of Cl at 4.7 km and 0.44-0.45% of Cl at 7.2 km in February 2015 and 0.14% of Cl at 4.7 km and 0.08% of Cl at 7.2 km in June 2015 (Table S10). Almost no contribution of Cl from the pipeline fluid is predicted from this model at the Spill Site supporting the argument that remediation activities limited movement of pipeline fluid into Blacktail Creek near the Spill Site at the location and time of sampling.

The pipeline sample contained 825 mg/L bromide (Br), whereas Br concentrations were below detection in BCR upstream from the spill and in the LMR reference site (Fig. 4, Table S8 and S9). As was observed for Na and Cl, Br concentrations at the Spill Site reflect the chemistry of upstream water due to active remediation near the Spill Site. Elevated Br concentrations downstream from the spill were detected at 4.7 km and 7.2 km in February 2015, but not at sites farther downstream, where substantial dilution by the higher flows of low-Br water in the Little Muddy River attenuates Br concentrations. Bromide could still be detected at sites along Blacktail Creek downstream from the Spill Site in June 2015 (Fig. 4, Table S9). Other elements that were substantially elevated above background concentrations included Sr, lithium (Li) and boron (B) (Fig. 4). Boron was still detected at concentrations substantially above background at the gage, 22.9 km downstream. This observation is consistent with the findings of Lauer etal. (2016); in their study of a broad sampling of surface waters in ND affected by brine spills, including 2 samples from this study area collected in July 2015, they found that Br, Sr, Li, and B were positively correlated with Cl indicating they may behave conservatively. Many of these same elements were detected in surface waters downstream from a UOG wastewater injection facility in West Virginia (Akob et al.,

Fig. 4. Bromide (Br), strontium (Sr), lithium (Li), and boron (B) concentrations in February 2015 (A) and June 2015 (B) in the brine pipeline sample and in water samples from surface water collected at the sites shown in Fig. 2. Field duplicate samples were collected at sites 7.2 and 22.9 km downstream of the Spill Site and are indicated by (1) and (2). ND = not determined.

2016) as a result of a common source for the formation waters of ancient, evaporated seawater.

Other major anions in this system include SO4, which is naturally high in surface waters of the region (approximately 500 to 1500 mg SO4/L in the upstream background samples) and HCO3 (410635 mg HCO3/L in the upstream samples) (Table S9). Neither of these species were present downstream from the spill at concentrations substantially different from background concentrations. Sulfate accounted for all of the dissolved sulfur detected in samples from all sites within analytical uncertainties. Nitrate and phosphate were below detection in all samples. Cations of calcium (Ca), magnesium (Mg), and potassium (K) were present at very high concentrations in the pipeline sample (Table S8) and were detectable at 1 to 3 times the background concentrations in Blacktail Creek at 4.7 km in February 2015 (Table S9). By June 2015, only K was still detectable above background concentrations at this location. The ammonium concentration measured in the pipeline sample was 2500 mg N/L (Table S8); unfortunately no surface water samples were able to be analyzed for NH4 in February from Blacktail Creek.

Calcium and carbonate concentrations in water samples collected along Blacktail Creek were near saturation with respect to calcite in February (Table S11, Fig. S3). Stream-water samples were highly supersaturated with respect to calcite in June, consistent with findings from Lauer et al. (2016) for stream-water samples collected a few weeks later downstream from the Spill Site. Aqueous compositions along Blacktail Creek show partial pressures of CO2 supersaturated with the atmosphere by a factor of 20 or more in February. In winter the icecover prevents atmospheric exchange, allowing accumulation of CO2 which, in turn, contributes to conditions favorable to dissolution of carbonate minerals. The PCO2 values in June were close to equilibrium with the atmosphere (saturation ratios of 0.6-2), indicating that rates of exchange with the atmosphere were commensurate with rates of CO2 production and consumption in the stream. It is possible that calcite or aragonite (saturation index for aragonite is 0.14 units lower than that for calcite) precipitation in the summer could result in accumulation

of a reservoir of Sr in the sediments, the dissolution of which could provide a source to aquatic communities trapped under ice in the winter.

3.2. Minor and trace elements

The pipeline sample contained elevated concentrations of many other elements that are typically present at low concentrations in surface waters in the region; manganese and iron concentrations, for example, were 5.6 mg/L and 34 mg/L, respectively (Table S8). Manganese concentrations downstream from the Spill Site in February 2015 were approximately double the background upstream concentrations and reached a maximum value of 470 ^g/L (Table S9), which is substantially greater than the Maximum Contaminant Level (MCL) of 50 |ag/L. Manganese and iron are expected to undergo sorption onto stream sediments and oxidation-reduction reactions throughout the water column. The temporal and spatial scales over which these reactions operate to retain or remobilize these elements in Blacktail Creek are unknown. Concentrations of many trace inorganic elements were quantified in stream-water samples (Table S9). The extent to which these elements are derived from the wastewater spill as compared to other anthropogenic (e.g., agriculture) and natural sources was not determined.

Strontium and barium concentrations along Blacktail Creek were elevated downstream from the Spill Site in both February and June (Fig. 4, Table S9). These elements have previously been reported to be useful tracers of UOG wastewater impacts from Marcellus Shale development (e.g. Akob et al., 2016; Brantley et al., 2014). Strontium concentrations 4.7 and 7.2 km downstream from the Spill Site were 7-9 times the BCR concentration in February and twice the BCR concentration in June. Barium concentrations 4.7 and 7.2 km along Blacktail Creek were two to three times the corresponding concentrations at the BCR site in both February and June. Aqueous chemical compositions of stream-water samples collected in February at all sites along Blacktail Creek indicated supersaturation with respect to barite (BaSO4 s), with somewhat higher supersaturations downstream from the Spill Site (Table S11, Fig. S3). By June 2015, only water samples collected downstream from the Spill Site were supersaturated with respect to barite. Lauer et al. (2016) also previously reported that water in Blacktail Creek downstream from the Spill Site was supersaturated with respect to barite. Radium co-precipitates readily with barite (e.g., Brandt et al., 2015; Curti et al., 2010). Therefore, barite precipitation in the wastewater flowing through the pipeline and water downstream from the spill could result in a reservoir of wastewater-derived Ba and Ra in the sediments that could continue to supply these elements to benthic communities and stream water after the pulse of aqueous spill-derived Ba and Ra has been transported downstream.

3.3. Isotope signatures

The 87Sr/86Sr isotopic composition of the pipeline wastewater was distinctly more radiogenic than the Sr isotopic composition in water samples collected from background sites BCR and LMR (Fig. 5). The degree to which the composition of the single sample of pipeline fluid we were able to obtain is representative of the composition of the millions of liters spilled is unknown, but likely there was some variability over time in the composition of fluid in the pipeline. To estimate the contribution of the pipeline fluid to the various samples, an 87Sr/86Sr mass balance mixing models between the pipeline fluid and the Blacktail reference site (BCR) was produced (detailed in SI Methods). Strontium compositions in water samples collected along Blacktail Creek 4.7 and 7.2 km downstream of the spill both in February and June 2015 plot closely along mixing lines between the pipeline spill sample and the corresponding BCR sample (Fig. 5), as did the July 2015 samples collected by Lauer et al. (2016) (although the composition of an upstream site was not provided in that study). Deviations between the observed compositions and the actual compositions likely result primarily from

Sr (mg/L)

>100 10

0.7105

0.7100

^ 0.7095 £ сn

0.7090

0.7085

Ill 1 ■

O February 2015

♦ June 2015

4 1% A Pipeline Fluid

□ Lauer etal,, 2016

\\ 4.7 km - Feb 2015 Mixing Line

VC-7.2 km - June 2016 Mixing Line

\\o.i%

43.8 km

22.9 km

4.7 krnV \ / o

v> 0.01%

7.2 km

\\ 22.9 km LMR

BCR Ц-1-1- -I {-Spilh©\ BCR -1-1-

1/Sr (L/mg)

Fig. 5. Ratios of 87Sr/86Sr plotted against the reciprocal of the strontium concentration (mg Sr/L) for aqueous samples from February 2015 (green) and June 2015 (blue) collected at sites shown in Fig. 2. Also shown are data from the site from July 2015 by Lauer et al. (2016). Lines from mixing calculations are shown along Blacktail Creek (squares). External precision of the isotopic measurements (2 s = 0.000013) is smaller than the symbol size.

variations over time in the background composition, as can be seen by comparing the locations of the BCR compositions in February and June 2015 (Fig. 5). Results from the mixing calculations (Table S10) indicate that 0.3 and 0.09% of the Sr in water collected from Blacktail Creek 4.7 km downstream of the Spill Site in February and June 2015, respectively, was derived from the spill. The relative contribution of pipeline fluid at the spill site was negligible, again suggesting that local remediation activities such as groundwater pumping limited movement of pipeline fluid into the stream at this location close to the point of the fluid release. These estimated contributions compare well with mass balance mixing models using Cl data (Table S10). Similarly, water collected from Blacktail Creek 7.2 km downstream from the Spill Site in February and June 2015 had 0.3% and 0.064%, respectively, of its Sr

derived from the spill. The Sr compositions of water samples collected from the 22.9 km downstream site, in February 2015, reflect mixing of water from the reference sites (LMR and BCR) and the spilled pipeline fluid. The isotopic composition of these samples was notably more radiogenic than the background samples from either source, consistent with a contribution from water from the spill.

3.4. Semi-volatile and extractable hydrocarbons in water and sediment

In February 2015, the field team reported visible oil sheen on the water under the ice, downstream from the pipeline break. The semi-volatile hydrocarbon analyses done on unfiltered samples showed the presence of low levels of alkylbenzenes and methyl-, dimethyl-, and trimethyl-naphthalenes in unfiltered samples downstream from the spill (Table 1). Concentrations were in the range 1 to 5 ^g/L. 1-Methyl-naphthalene was detected the farthest downstream at 22.9 km. Extract-able hydrocarbon analyses of filtered samples, however, showed no detectable concentrations of hydrocarbons (Fig. S4) and no measurable concentrations of UOG production additives. This method has been used previously to identify organic compounds in wastewater from UOG production by this USGS laboratory (Engle et al., 2016; Orem et al., 2014) and by others investigating produced water chemistry in shale-gas wells in Colorado, Pennsylvania, Texas and New Mexico (Lester et al., 2015; Maguire-Boyle and Barron, 2014). There was, however, an unresolved complex mixture (UCM) evident in the chromato-grams from all sites along Blacktail Creek in February, that may indicate a mixture of hydrocarbons (straight and branched alkanes and alkenes and aromatics), but at low levels. The UCM was present both upstream and downstream from the spill and is thus not likely to be linked to the wastewater spill and may be due to background organic compounds present naturally in the watershed, or possibly hydrocarbons from general oil and gas production in the region.

These results indicate that the semi-volatile hydrocarbons that were measured at very low levels in the river downstream from the spill at the time of the sampling may have been associated with particulate matter in the creek since they were not detected in filtered samples. Sorption of hydrophobic organic compounds, including naphthalene onto soil and sediments, is well-documented (Eadie et al., 1996; Kile etal., 1995; Moyoetal., 2014) and known to be enhanced at lower temperatures (Piatt etal., 1996). The sediment extracts of soils done in February 2015 contained traces of naphthalene, methylnaphthalenes, and other polycyclic aromatic hydrocarbons but they were below the reporting limits in all samples. Sorption processes might also explain the lack of detection of UOG additives in the river downstream from the spill (the wastewater flowed overland or through sediments for

Semi-volatile organics observed in water samples from Blacktail Creek and Little Muddy River. Compounds which were tentatively identified via automated search against NIST libraries, but were not quantified are indicated by"+"; compounds not detected are indicated by"-". Quantified compounds are in |og/L of the compound and were identified against deuterated internal standards. Field duplicate samples were collected at sites 7.2 and 22.9 km downstream of the Spill Site and are indicated by (1) and (2).

BCR LMR Spill Site 4.7 km 7.2 km (1) 7.2 km (2) 22.9 km (1) 22.9 km (2) 43.8 km Blank

1,3,5-Trimethylbenzene - - - 0.390 1.490 - - - - -

1,2,3,5-Tetramethylbenzene - - - + + + - - - -

1,2,3,4-Tetramethylbenzene - - - 3.64 4.76 4.56 - - - -

n-Ethyl-n,n-dimethylbenzene - - - - + + - - - -

Naphthalene - - - - - - - - - -

2-Methylnaphthalene - - - - - - - - - -

1-Methylnaphthalene - - - 1.71 2.17 2.12 0.13 0.10 - -

Ethylnapthalene - - - - - - - - - -

2,6;2,7-Dimethylnaphthalene - - - + + + + + - -

1,3;1,7-Dimethylnaphthalene - - + + + + - - - -

1,6-Dimethylnaphthalene - - - + + + - - - -

2,3; 1,4; 1,5-Dimethylnaphthalene - - + - - - + + - -

1,2-Dimethylnaphthalene - - - - - - - - - -

1,8-Dimethylnaphthalene - - - - - - - - - -

1,4,6 Trimethylnapthalene - - - + + + + + - -

2,3,6 Trimethylnapthalene - - + + + + + + - -

1,6,7 Trimethylnapthalene - - + + + + + + - -

10s of meters before entering the stream), although the lack of any pipeline sample for organics analyses limits what can be said about the potential fate of organic compounds in the wastewater.

3.5. Light (Cj-C6) hydrocarbons in surface water

Light hydrocarbons were at low concentrations in Blacktail and Little Muddy Creek (Fig. 6, Table S12), but show a distinct thermogenic hydrocarbon signature. Total concentrations in both upstream and downstream locations were higher in February than in June, probably due to the ice cap over the rivers that would have restricted atmospheric venting. Methane is the most prevalent compound in all light hydrocarbon samples, and total C-!-C6 concentrations vary between 39.8 and 91.3 nmol/kg in February and 0.91 and 37.7 in June. In February 2015, the 7.2 km site had the highest ethane and propane concentrations (1 to 2 orders of magnitude higher than BCR). The ratios of the constituents of the dissolved hydrocarbons measurements can be used to infer the origins of the gases in the sample. Biogenic gas is typically dominated by methane, with trace amounts of heavier alkanes (~1 mol% or less); in contrast, thermogenic gases can have significant amounts of higher alkanes in stepwise decreasing abundances in a cracking pattern produced by thermal maturation (Schoell, 1983; Taylor, 2000). The ethane/methane molar ratio increases to 5.74 x 10-1 at 7.2 km from 3.12 x 10-3 at BCR, with concomitant increases in the ratios of methane to higher alkanes, giving a distinct thermogenic signature to the light hydrocarbon ratios that extends to n-hexane, the heaviest alkane reported. The ratios of ethane/methane, propane/ethane, and n-butane/propane were maintained in downstream samples after the confluence of the Blacktail Creek and Little Muddy River. Compared to the BCR and LMR, the concentrations of hydrocarbons, particularly ethane and unsaturated hydrocarbons were generally elevated downstream, including the 43.8 km Site. The ratios

Fig. 6. Low level hydrocarbon concentrations in February 2015 (A) and June 2015 (B) in water samples from surface water collected along Blacktail Creek and the Little Muddy River, as shown in Fig. 2. No samples were analyzed from the Spill Site and 4.7 km collected in February due to observed liquid hydrocarbon sheen on the water surface that precluded trace hydrocarbon analysis. Concentrations are expressed in nanomoles of the specific compound per kilogram water. Field duplicate samples were collected at the 22.9 km site, indicated by (1) and (2). NA = not analyzed.

of propane/ethane and n-pentane/n-butane were also similar at the 22.9 km site. Compared to the BCR and LMR, the concentrations of hydrocarbons, particularly ethane and unsaturated hydrocarbons were generally elevated downstream, including the 43.8 km site. The ratios of propane/ethane and n-pentane/n-butane were also similar at the 22.9 km site.

In June 2015, the Spill Site, 4.7 km, and 7.2 km had the highest methane and ethane concentrations. Although 7.2 km had the highest ethane and propane concentrations in water from the sites sampled in June, the ratios of ethane, propane, and other higher hydrocarbons to methane at the 7.2 km site were lower than in February due to the relatively large amount of methane present (ethane/methane: 1.11 x 10-2, propane/ methane: 5.28 x 10-3). The ratios of propane/ethane and n-butane/ propane were similar to those observed at the site in February and distinct from those observed in the background sites, suggesting that there was a common, lingering source of these compounds at 7.2 km between the two sampling dates. In February, the downstream hydrocarbon concentrations and ratios show dilution and transport of a common source (or set of sources) of thermogenic hydrocarbons. The persistence of these was likely enhanced by the limited ventilation through the iced covered river, higher solubility at low temperatures, and limited microbiological activity in winter. In contrast, the June downstream concentrations of C2-C6 hydrocarbons were low enough that exogenous sources (from microbial activity and water in contact with ambient air with mixing ratios of less than a ppbv) obscure the ratios of compounds that might be transported downstream and diluted from the 7.2 km site during the June sampling event. The ratios and concentrations upstream at 4.2 km and the Spill Site were discordant with those observed at 7.2 km (and downstream in February). If the light hydrocarbons were ultimately sourced from the pipeline spill, this could indicate that the upstream light hydrocarbons were modified or attenuated by the remediation activities, and that they were sorbed to in the riverbed or shallow groundwater around the pipe and are slowly leaving via the discharge into the river.

These light hydrocarbon measurements are interesting because they show that releases may be traced considerable distance downstream and may linger in the environment for considerable time even though they are highly volatile. Although the controls on the concentrations and ratios of light hydrocarbons at low concentrations in water are not well understood, the differences between upstream, spill, and far downstream sites show that this technique has promise for identifying the extent and compositional origin of UOG materials released into the environment. The concentrations of alkenes may also be indicators of microbial activity, for example the high levels of ethene, a compound released by algae as a byproduct of sulfur reduction (Plettner et al., 2005).

3.6. Sediment chemistry

Labile Ba and Sr concentrations extracted from sediments collected along Blacktail Creek in June 2015 were higher downstream from the Spill Site than upstream (Fig. 7a). Labile sediment-bound Ba and Sr were assayed using short-term, dilute hydrochloric acid extractions (methods detailed in the SI). Sediments collected directly from the Spill Site had labile Ba and Sr concentrations similar to sediments collected upstream. This is likely a result of the remediation operations that had been conducted at the Spill Site, similar to the results seen in the water samples. The isotopic composition of labile Sr extracted from sediments collected downstream in June 2015 exhibited a significant contribution of Sr from the pipeline fluid (Fig. 7b), especially at the 4.7 km site. This observation supports the hypothesis that the small increase in labile Sr on sediments downstream from the spill results from retention of Sr and, likely, Ba from the spill on the sediments.

226Radium activities in sediments from Blacktail Creek and the Little Muddy River ranged from a background of about 10-20 Bq/kg in sediments collected upstream at BCR to 464 Bq/kg in sediments collected from the Spill Site in June 2015 (Fig. 8, Table 2). Ratios of 226Ra/238U

Fig. 7. (A) Potassium-chloride-extractable ammonium and hydrochloric-acid-extractable strontium and barium concentrations from sediments collected along Blacktail Creek in June 2015. (B) Strontium isotopic ratios in HCl extracts of the sediments collected along Blacktail Creek and the pipeline brine aqueous sample. Field duplicate samples were collected at site 7.2 km and are indicated by (1) and (2). External precision of the isotopic measurements (2 s = 0.000013) is too small to be plotted. ND = not determined.

equal to 1.0 (within measurement error) were observed in sediment samples collected upstream from the Spill Site suggesting secular equilibrium in these samples. Given that secular equilibrium in the 226Ra-238U system requires several thousand years, it is likely that the Ra content of these samples represents naturally occurring background concentrations of Ra. Sediments collected in February 2015 had slightly elevated 226Ra/238U ratios over secular equilibrium at the Spill Site. However, it is important to note the February Spill Site sample may not be representative of actual spill concentrations at this location because of remediation efforts, including soil removal, which had already

Fig. 8. Radium-226 activities in river sediment in February 2015 and June 2015 collected along Blacktail Creek and the Little Muddy River, as shown in Fig. 2. "ND" indicates no data were collected.

occurred at the spill location. Additionally, comparison of low activity samples such as these is difficult as the differences of grain size and geometry, interference peaks of other radioactive material, and low counts create greater uncertainty compared to higher activity samples. The 226Ra/238U ratios downstream at 4.7 km in February 2015, had a 9-fold excess of 226Ra over secular equilibrium and suggests that the spill had resulted in contamination of these sediments with 226Ra. These observations are consistent with those of Akob et al. (2016) who found that 226Ra/238U activity ratios in sediments at a wastewater injection facility could be used to detect wastewater spills. Samples collected in February 2015 farther downstream showed no significant elevation of 226Ra over that in secular equilibrium with 238U.

Sediments collected at the Spill Site in June 2015 exhibited higher concentrations of Ra in excess of secular equilibrium, compared with February 2015 samples (Table 2), suggesting additional input of contaminated water or sediments at the site. Furthermore, sediments collected at 4.7 km downstream and at 7.2 km downstream from the Spill Site had a 14-fold and 5.5-fold excess of 226Ra over secular equilibrium, respectively. Lauer et al. (2016) found that sediments collected downstream from the Spill Site in July 2015 had total Ra [226Ra plus 228Ra] activities in the range 550 to 4700 Bq/kg whereas water samples had total Ra activities of 0.3 Bq/L. Sediments from Blacktail Creek with excess 226Ra also had 228Ra/226Ra ratios (Table 2) lower than the value of 1.1 determined on sediments from Blacktail Creek upstream from the Spill Site by Lauer et al. (2016), consistent with contributions from Bakken brines, which have 228Ra/226Ra ratios in the range 0.4-0.5 (Lauer et al., 2016). However, it should be noted that 228Ra has a 5.75 year half-life and our sediment samples were measured up to a little over a year after they were collected. Ratios in our sediment samples may be lower than those determined by Lauer et al. (2016) because our activity ratios have not been corrected for 228Ra decay between when our sediment samples were collected and when they were analyzed.

The chemical form of excess Ra in wastewater-contaminated sediments is unknown. As discussed previously, Ra incorporated into barite is one potential form of sediment-bound Ra. In addition, Ra sorption can occur on sediments exposed to elevated dissolved concentrations of Ra. However, it has been noted by others (Harto et al., 2014; Kraemer and Reid, 1984; Landa and Reid, 1983; Lauer et al., 2016; Nelson et al., 2015) that the degree of Ra sorption to sediment and soil depends on both the quantity and quality of the water present in the sample (such as salinity and chemistry) and the type and character of the sediment or soil (e.g., grain size, organic matter content, cation exchange capacity). The salinity of the brines can inhibit Ra sorption to the soil (which would have a higher potential of occurring close to the spill source where the salinity is greater). Downstream from the spill, the brine becomes diluted with freshwater from surface and groundwater, which could favor Ra sorption onto the sediments. Determining the chemical form of Ra in the sediments is an important topic of further research.

The increases in 226Ra activities in sediment samples collected along Blacktail Creek between February and June 2015, could be a reflection of transport of contaminated material downstream, changing site conditions, and remediation efforts at the spill site. Downstream transport of particle-associated constituents are hypothesized to occur through two potential mechanisms: 1) fluvial transport and storage of fine sediment with precipitates and sorbed constituents, and 2) transport of contaminated groundwater through the channel bed sediments and the hyporheic zone. Field conditions in February were cold and the stream was frozen (Fig. S1), possibly limiting input of contaminated soil to the stream. Remediation efforts were also fully underway with significant soil around the spill having been removed, limiting the availability of contaminated material for transport. In June 2015, sampling was conducted after the spring snowmelt which could have resulted in increased groundwater transport, thereby contaminating soils, and enhancing Ra sorption onto fluvial sediment. These results are consistent with work done on Ra activities conducted along Blacktail Creek by Lauer et al. (2016).

Table 2

Uranium and radium activitiesa in sediments from background, Spill Site, and downstream locations.

Location Date 238U (Bq/kg) 226Ra (Bq/kg) 228Ra (Bq/kg) 226 Ra 238 U 228 Ra 226 Ra

BCR 2/11/15 20 ± 7 12 ± 4 NDb 0.6 ± 0.3 -

LMR No sample

Spill Sited 2/11/15 12 ± 7 30 ± 4 NDb 2.6 ± 1.5 -

4.7 km 2/11/15 18 ± 6 166 ± 3 149 ± 7 9.2 ± 2.9 0.90 ± 0.05

7.2 km 2/10/15 29 ± 5 33 ± 4 ND 1.2 ± 0.3 -

22.9 km 2/10/15 30 ± 7 26 ± 5 ND 0.8 ± 0.2 -

43.8 km 2/10/15 BDc 20 ± 1 BD - -

BCR 6/17/15 9 ± 4 17 ± 2 ND 1.8 ± 0.8 -

LMR 6/17/15 29 ± 3 32 ± 1 21 ± 1 1.1 ± 0.1 0.66 ± 0.11

Spill Site 6/17/15 26 ± 9 464 ± 7 290 ± 1 18 ± 4 0.62 ± 0.01

4.7 km 6/17/15 12 ± 6 159 ± 2 133 ± 2 14 ± 1 0.83 ± 0.02

7.2 km 6/17/15 8 ± 2 41 ± 2 33 ± 1 5.5 ± 1.7 0.74 ± 0.09

22.9 km 6/18/15 11 ± 8 18 ± 5 ND 1.6 ± 1.2 -

43.8 km No sample

a Activity in becquerels (disintegrations per second) per kilogram sediment dry weight. Concentrations can be calculated from the formula: Q = (Ait1/2)i)/{(Na)(ln2)}, where i refers to the specific radionuclide, Cf the concentration (moles/kg sediment), Ai the activity, tV1i the half-life (seconds), Na Avogadro's number (dimensionless), and ln2 is the natural logarithm of 2. b ND = not determined.

c BD = below detection, could not be quantified because of low activity. d Coarse gravelly sample; may effect concentrations.

Labile sediment bound ammonium concentrations were assayed using short-term potassium chloride extractions (as described in the SI Methods) on sediment samples collected in June 2015. Sediment-bound ammonium concentrations 4.7 km downstream from the Spill Site were over 10 times those observed at the background site (BCR) (Fig. 7A, Table S13). Sediment-bound ammonium concentrations were somewhat elevated compared to the background site at the Spill Site and 7.2 km downstream from the Spill Site. This pattern is similar to the pattern in sediment-bound Ba and Sr, consistent with the pipeline spill being the source of all three constituents (Fig. 7A). However, contributions to the observed pattern from variability in natural and anthropogenic sources along Blacktail Creek cannot be ruled out without additional research.

3.7. Potential endocrine disruption activity

No estrogenic receptor activity was noted for the mammalian assays in water downstream from the Spill Site, compared with water collected upstream, in either February or June 2015 (Fig. 9a). However, estrogen receptor inhibition was substantially greater at 7.2 km downstream in February 2015 compared to the upstream waters at BCR. These differences were still measurable but less pronounced in June 2015 (Fig. 9b). However, the water sample from Little Muddy River also showed some estrogen receptor inhibition in the same range as the June 2015 samples, indicating the June observations may be within natural variations in these waters.

Estrogenicity using yeast reporter strains was noted in samples collected at a number of sites during February and June (Fig. 9c). Estrogenicity was detected in samples from all sites evaluated during the February collection. The process control was always below detection. Net estrogenicity was greatest at the site 4.7 km downstream from the spill (2.13 ± 0.11 ng/L). Modest estrogenic activity was observed both up- and downstream from the Spill Site. Calculated estrogen equivalents (EEQ) ranged from 1.30-2.13 ng/L across the downstream site gradient from BCR to 4.7 km. In general, estrogenicity was higher in February compared to June. The maximum observed EEQ in June was 0.54 ±0.11 ng/L at BCR, upstream from the Spill Site. It is not clear if this estrogenic signal is associated with UOG activity as a modest signal was identified upstream from the spill location on Black-tail Creek. Endocrine disrupting activity has been associated with UOG spills in Colorado and West Virginia (Kassotis et al., 2014; Kassotis et al., 2016b); to our knowledge, this is the first report of EDC activity associated with UOG spills in ND. Although these results indicate that a full understanding of the endocrine disrupting chemical (EDC) activity

Fig. 9. Estrogen receptor (A) activity and (B) inhibition of surface water samples collected at sites indicated via mammalian reporter gene assay. (C) Estrogen equivalents (EEQ, relative to 17b-estradiol) of OASIS HLB extracted water samples using the bioluminescent yeast estrogen screen (BLYES). Values are mean of triplicate reads; error bars indicate standard error of the mean. BD = below detection. ND = not determined (sample not analyzed). Field duplicate samples were collected at sites 7.2 and 22.9 km downstream of the Spill Site and are indicated by (1) and (2). The duplicate sample from 4.7 km was not evaluated due to sample loss during shipping.

of water downstream from these types of spills requires further study, the elevated EDC activity, specifically inhibition of estrogen receptor activity, downstream from the spill site in February is consistent with previous reports of an association between EDC activity in water and UOG activities (Kassotis et al., 2014; Kassotis et al., 2016b). Previous work by others has shown antiestrogens can negatively impact aquatic organisms, and in the current study we found moderate antiestrogenic activity within a potential bioactive range (Madureira et al., 2015; Roepke et al., 2005).

3.8. Aquatic health studies

Ninety-six hour survival of early life stage Fathead Minnows (FHM) downstream from the wastewater pipeline rupture was significantly reduced at 1 of 4 experimental sites approximately 6-months after the spill was detected. Survival of <48 h aged FHM after 96 h was 89.2% at BCR, 94.7% at LMR, 89.7% at the Spill Site, 89.2% at 4.7 km, 2.5% at 7.2 km, and 74.2% at 22.9 km (Table S14). The reduction in survival observed at 7.2 km is statistically significant at p < 0.05. Mortality of two resident Madtom Catfish (Noturus sp.) was observed at 7.2 km at 72 h and an oily rainbow-colored sheen was observed on the water surface at 96 h (Fig. S5), although the relationship between this observation and fish survival is unknown. Although no systematic native fish sampling was conducted, live resident fish were observed at all sites, with the exception of the 7.2 km site where no live resident fish were observed (Fig. S5b and c). There was no statistical difference between survival at the LMR and the 22.9 km downstream site.

Large diel fluctuations in dissolved oxygen and temperature were recorded at all sites in June 2105; oxygen saturation ranged from < 10% to over 200% (Fig. S6), and water temperatures ranged from 15 to 28 °C, except at site 7.2 km. At the 7.2 km site declining temperatures were observed at 72 and 96 h, dropping to 7.95 °C. Total dissolved salt (TDS) concentrations, hardness, and SO4 concentrations were generally consistent among sites. Concentrations of Cl and bicarbonate were stable at all sites except at 7.2 km, where they increased from 194 to 568 mg/L and from 41.6 to 727 mg CaCO3/L, respectively. To a lesser degree, Cl but not bicarbonate increased at the 4.7 km site, where 316 and 499 mg/L were measured at 0 and 96 h respectively with a slight decrease to 308 mg/L at 24 h (Table S15).

Reduced temperatures and increased Cl and HCO3 concentrations suggests a pulsed upwelling of groundwater into the stream at the time that mortalities were observed. Variations in stream discharge observed during our field investigations and are consistent with the observations of paleochannels by environmental remediation consultants working at the site (H. Rhodes, personal communication, June 2015); these paleochannels are highly conductive and could act as conduits of spill contaminated groundwater into the surface water. The changes observed in water chemistry were accompanied by a decrease in surface water flow, and highlight the need for further study to assess the hydro-logic conditions and potential contaminant pathways at the site.

Changes in water chemistry and visual observations concurrent with observed mortalities were increased Cl and bicarbonate concentrations, decreased water temperature, and a rainbow-colored oily sheen on the water's surface. The median lethal concentration (LC 50) of Cl and HCO3 to FHM is approximately 4000 mg Cl/L and 1250 mg HCO3/L respectively in laboratory studies (Harper et al., 2014; Mount et al., 1997). But in waters with high hardness, calcium may have an ameliorating effect on ion toxicity (Soucek et al., 2011). Under the hardness conditions found in Blacktail Creek, the increased concentrations of Cl and HCO3 alone did not appear sufficiently high to induce acute toxicity at the 7.2 km site.

The 2500 mg N/L NH4 concentration (as defined above as total NH4) in the pipeline spill fluid had the potential for causing chronic or even acute toxicity to fish and aquatic life. The toxic form of NH4 is the aqueous NH3 species, sometimes referred to as "un-ionized ammonia". The U.S. EPA acute NH4 criterion for the protection of aquatic life is 17 mg N/L and the chronic NH4 criterion is 1.9 mg N/L (U.S. EPA,

2013). Note that the U.S. EPA (2013) term "ammonia" refers to what we call "total dissolved ammonium" (NH4). Ammonium concentrations of 10 mg N/L were measured at the 7.2 km site in February (Table S9) but aqueous NH3 concentrations were <0.1 mg N/L owing to the low pH values (Table S11). Lauer et al. (2016) reported NH4 concentrations in water samples collected in July 2015 from Blacktail Creek near the Spill Site of 17-21 mg N/L, equal to or exceeding the U.S. EPA criterion for acute toxicity. Aqueous NH3 concentrations could not be computed because Lauer et al. (2016) did not report pH values or temperatures for individual samples. After completion of the experiment with FHM described above, NH4 concentrations were determined in water samples collected at BCR, LMR, the Spill Site, and at 4.7 km, 7.2 km, and 22.9 km downstream. Concentrations at all sites were 0.1 mg N/L or less except the 7.2 km site, which had an NH4 concentration of 3.4 mg N/L (Table S16). This concentration exceeds the U.S. EPA criterion for chronic toxicity to aquatic life. Thurston et al. (1983) reported acute aqueous NH3 toxicity to FHM ranging from 0.75 to 3.4 mg N/L; the toxicity was reduced as temperature increased to 23 °C, the upper temperature limit of the study. At the pH and temperature at site 7.2 km (8.83, 27 °C), the NH4 concentration of 3.4 mg N/L corresponds to a concentration of aqueous NH3 of 0.85 mg N/L, within the range of concentrations Thurston et al. (1983) found to cause acute toxicity to FHM. The U.S. EPA criterion for acute toxicity at a temperature of 27 °C and pH of 8.8 is 0.44 mg N/L of total NH4 (U.S. EPA, 2013). Thus, these measurements, under the conditions observed at our site, show that chronic or even acute toxic levels of NH4 could have been reached in Blacktail Creek.

In situ experiments take advantage of diel fluctuations in physical and chemical conditions that cannot be recreated in the laboratory but are typical in prairie stream conditions (Farag et al., 2014). As a result, these experiments provide relevant information about the resiliency of fish in Blacktail Creek and the Little Muddy River six months after the UOG wastewater pipeline rupture, and during continued remediation efforts. Early lifestage FHM did not survive 96 h at the 7.2 km site. Almost all of the mortalities from this site were observed 24 h following the discovery of dead resident fish at the site. Diel fluctuations in DO were large at all sites measured and neither the BCR nor the 7.2 km sites went anoxic (Fig. S6), yet significant mortality of FHM was observed only at the 7.2 km site, and mortalities of resident madtoms occurred at the same site within 24 h of the experiments with the FHM. These observations suggest that changes in DO were not the cause of death in the madtoms or FHM.

Laboratory invertebrate health studies showed an inconclusive impact on amphipods. Differences in survival, growth (average mg per individual), and total biomass (total mg per replicate) of amphipods, midges, and mussels among sites were evaluated by one-way analysis of variance of rank-transformed data, with differences among means evaluated using Tukey's test. Survival of all three species was unaffected. Growth of midges and mussels and biomass of all three species differed significantly among sites, but sites downstream from the spill location generally did not show significant reductions in these endpoints relative to reference sites (Table S17). Overall, there was a non-significant trend of lower growth and biomass of amphipods at two sites downstream from the Spill Site (e.g., growth amphipod BCR 0.76 mg ±0.02 vs. Spill Site 0.67 mg ± 0.04 and 4.7 km 0.59 mg ± 0.07) and to a lesser extent the 7.2 km site (0.73 mg ± 0.07). Although these results do not demonstrate strong or consistent toxic effects of sediments from reaches of Blacktail Creek or Little Muddy Creek affected by the spill, they do indicate that longer-term investigations of sediment toxicity after spill or release events are warranted.

4. Implications

Although there are critical knowledge gaps regarding the effects of contaminants released to the environment during UOG waste management activities, it is apparent from this study that the type of spill

(pipeline rupture), the constituents from the spill (brine with some hydrocarbons mixture), physiography, and location of the spill (discharge adjacent to a stream) will influence the environmental pathways and effects. This research advances our understanding and quantifies potential impacts through the analyses of Ra and Sr concentrations and isoto-pic compositions, trace inorganic and organic compounds, as well as endocrine disrupting effects and bioassays with model organisms. This set of analytical tools provides insights into potentials for human exposures. Concentrations of many wastewater-derived contaminants in stream water were several times background concentrations, but still relatively low compared to U. S. EPA drinking water standards. Nevertheless, wastewater-derived elements and radioisotopes partitioned onto sediments, potentially providing a long-term source of Ba, Ra, and other contaminants to aquatic life. Results from this study and Lauer et al. (2016) show that radium activities were significantly above the U.S. EPA action level for 226Ra in surface soils, which should not exceed 5 pCi/g (185 Bq/kg). Episodic increases in NH4 were at levels high enough to be toxic to aquatic life, particularly in the spring and summer when pH values increase during peak photosynthetic activity, shifting speciation in favor of the more toxic aqueous NH3 form.

Potential health effects are indicated by fish bioassays, in which fish experienced mortality, and endocrine disrupting activity was observed downstream from the spill. There was a clear increase in antagonism below the spill site. This increased antagonism previously has been associated with UOG impacted waters (Kassotis et al., 2016a; Kassotis et al., 2014). The total estrogenicity measured above and below the spill site was modest compared to other locations in the U.S. where strain BLYES has been utilized to measure estrogen agonism. Measures of estrogenicity > 1 ng/L are typically more characteristic of waters impacted by wastewater reclamations discharge or animal feeding operations (Ciparis et al., 2012; Iwanowicz et al., 2016). Published accounts of a reasonable aquatic organism adverse effects threshold for estrogenicity range from 0.73 ng/L to 2 ng/L (Caldwell et al., 2012; Wu et al., 2014; Young et al., 2002). Conley et al. (2016) have suggested 1 ng/L as an adverse effects trigger value. A number of the February measures proximate to the Spill Site exceed this value suggesting the potential for adverse effects in resident aquatic organisms. The identity of these endocrine-active chemicals is not known based on the current analyses, but the bioassay endpoints clearly indicate their presence at biologically meaningful concentrations.

The results of this work suggest that whereas wastewater spills introduce both organic and inorganic constituents into the environment, elements from brines (such as NH4, Ra, Ba, and Sr) occur at higher concentrations and may persist longer in the environment due to partitioning onto sediment. This is an especially significant finding considering that 3 times as many brine spills were reported (2007-2015) in North Dakota as compared to oil spills during the same period.

Our observations show that initial remediation of the spill effectively removed some types of contamination while allowing others to persist. Temporal sampling indicates that contaminated groundwater and soil can reintroduce different contaminants at variable rates leading to potentially different site management concerns between initial contamination and long-term impact. Partitioning of chemicals onto the sediment limits movement of wastewater components downstream but could provide a long-term source of contaminants to aquatic organisms. Sediment-bound forms of elements like Ba, Sr, and Ra could include ion exchangeable species as well as elements incorporated into solid phases like calcite or aragonite, and barite. Carbonate minerals that form during the summer may dissolve in response to rising PCO2 values during periods when ice-cover prevents atmospheric exchange, releasing co-precipitated contaminants. Increasing Ca concentrations in response to dissolution of carbonate minerals could perturb ion exchange equilibria, potentially mobilizing elements like Ba, Sr, and Ra. Understanding chemical forms of wastewater-spill derived contaminants retained in sediments is an important area of future research.

Acknowledgements

This project was supported by the USGS Toxic Substances Hydrology Program, USGS Energy Resources Program, and the USGS Fisheries Program. The authors are grateful to Joanna Thamke and two anonymous reviewers for their helpful comments. The authors would like to thank Matthew Olson, Bill Damschen, and Robert Lundgren for assistance in the field, Katherine Akstin, Kalla Fleger, Meagan Mnich, Michael Doughten, and Tracey Spencer for laboratory assistance, and John Swiecichowski for assistance with tables. We would like to thank John M. Besser and Chris D. Ivey for conducting invertebrate sediment bioas-says and statistical interpretations of the data. Sincere thanks to Jenn Cornelius Green and Chris Kassotis for their work to process water samples and test for EDC activity. In addition, we would like to thank the local landowners who allowed access to field sites.

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2016.11.157.

References

Adams, M.B., 2011. Land application ofhydrofracturing fluids damages a deciduous forest stand in West Virginia. J. Environ. Qual. 40 (4), 1340-1344.

Akob, D.M., Mumford, A.C., Orem, W.H., Engle, M.A., Klinges, J.G., Kent, D.B., et al., 2016. Wastewater disposal from unconventional oil and gas development degrades stream quality at a West Virginia injection facility. Environ. Sci. Technol. 50 (11), 5517-5525.

Barbot, E., Vidic, N.S., Gregory, K.B., Vidic, R.D., 2013. Spatial and temporal correlation of water quality parameters of produced waters from Devonian-Age shale following hydraulic fracturing. Environ. Sci. Technol. 47 (6), 2562-2569.

Blondes, M.S., Gans, K.D., Thordsen, J.J., Reidy, M.E., Thomas, B., Engle, M.A., et al., 2014. U.S. Geological Survey National Produced Waters Geochemical Database v2.1 (PROVISIONAL). United States Geological Survey.

Brandt, F., Curti, E., Klinkenberg, M., Rozov, K., Bosbach, D., 2015. Replacement of barite by a (Ba,Ra)SO4 solid solution at close-to-equilibrium conditions: a combined experimental and theoretical study. Geochim. Cosmochim. Acta 155,1-15.

Brantley, S.L., Yoxtheimer, D., Arjmand, S., Grieve, P., Vidic, R., Pollak J., et al., 2014. Water resource impacts during unconventional shale gas development: the Pennsylvania experience. Int. J. Coal Geol. 126,140-156.

Caldwell, D.J., Mastrocco, F., Anderson, P.D., Lange, R., Sumpter, J.P., 2012. Predicted-no-effect concentrations for the steroid estrogens estrone, 17beta-estradiol, estriol, and 17alpha-ethinylestradiol. Environ. Toxicol. Chem. 31 (6), 1396-1406.

Ciparis, S., Iwanowicz, L., Voshell, J., 2012. Effects of watershed densities of animal feeding operations on nutrient concentrations and estrogenic activity in agricultural streams. Sci. Total Environ. 414, 268-276.

Conley, J.M., Evans, N., Mash, H., Rosenblum, L., Schenck, K., Glassmeyer, S., et al., 2016. Comparison of in vitro estrogenic activity and estrogen concentrations in source and treated waters from 25 U.S. drinking water treatment plants. Sci. Total Environ. http://dx.doi.org/10.10167j.scitotenv.2016.02.093.

Curti, E., Fujiwara, K., Iijima, K., Tits, J., Cuesta, C., Kitamura, A., et al., 2010. Radium uptake during barite recrystallization at 23 ± 2 °C as a function of solution composition: an experimental 133Ba and 226Ra tracer study. Geochim. Cosmochim. Acta 74 (12), 3553-3570.

DiGiulio, D.C., Jackson, R.B., 2016. Impact to underground sources of drinking water and domestic wells from production well stimulation and completion practices in the Pavillion, Wyoming, Field. Environ. Sci. Technol. 50 (8), 4524-4536.

Drollette, B.D., Hoelzer, K., Warner, N.R., Darrah, T.H., Karatum, O., O'Connor, M.P., et al., 2015. Elevated levels of diesel range organic compounds in groundwater near Marcellus gas operations are derived from surface activities. Proc. Natl. Acad. Sci. 112 (43), 13184-13189.

Eadie, B.J., Morehead, N.R., Klump, J.V., Landrum, P.F., 1996. Distribution of hydrophobic organic compounds between dissolved and particulate organic matter in Green Bay waters. J. Great Lakes Res. 18 (1), 91-97.

Elliott, E.G., Ettinger, A.S., Leaderer, B.P., Bracken, M.B., Deziel, N.C., 2016. A systematic evaluation of chemicals in hydraulic-fracturing fluids and wastewater for reproductive and developmental toxicity. J. Expo. Sci. Environ. Epidemiol. http://dx.doi.org/ 10.1038/jes.2015.81.

Engle, M.A., Cozzarelli, I.M., Smith, B.D., 2014. USGS Investigations of Water Produced During Hydrocarbon Reservoir Development. US Geological Survey, Fact Sheet. 2014-3104 pp. 1-4.

Engle, M.A., Reyes, F.R., Varonka, M.S., Orem, W.H., Ma, L., Ianno, A.J., et al., 2016. Geochemistry of formation waters from the Wolfcamp and "Cline" shales: insights into brine origin, reservoir connectivity, and fluid flow in the Permian Basin, USA. Chem. Geol. 425, 76-92.

Farag, A.M., Harper, D.D., Skaar, D., 2014. In situ and laboratory toxicity of coalbed natural gas produced waters with elevated sodium bicarbonate. Environ. Toxicol. Chem. 33 (9), 2086-2093.

Gallegos, T.J., Varela, B.A., 2015. Trends in hydraulic fracturing distributions and treatment fluids, additives, proppants, and water volumes applied to wells drilled in the United States from 1947 through 2010—data analysis and comparison to the literature. U.S. Geological Survey Scientific Investigations Report 2014-5131, p. 15.

Galloway, J.M., Vecchia, A.V., Vining, K.C., Densmore, B.K., Lundgren, R.F., 2012. Evaluation of water-quality characteristics and sampling design for streams in North Dakota, 1970-2008. US Geological Survey Scientific Investigations Report. No. 2012-5216.

Gleason, RA, Tangen, B.A., 2014. Brine contamination to aquatic resources from oil and gas development in the Williston Basin, United States. U.S. Geological Survey Scientific Investigations Report 2014-5017, p. 127.

Gregory, K.B., Vidic, R.D., Dzombak, D.A., 2011. Water management challenges associated with the production of shale gas by hydraulic fracturing. Elements 7 (3), 181-186.

Harkness, J.S., Dwyer, G.S., Warner, N.R., Parker, K.M., Mitch, WA, Vengosh, A., 2015. Iodide, bromide, and ammonium in hydraulic fracturing and oil and gas wastewaters: environmental implications. Environ. Sci. Technol. 49 (3), 1955-1963.

Harper, D.D., Farag, A.M., Skaar, D., 2014. Acute toxicity of sodium bicarbonate, a major component of coal bed natural gas produced waters, to 13 aquatic species as defined in the laboratory. Environ. Toxicol. Chem. 33 (3), 525-531.

Harto, C.B., Smith, K.P., Kamboj, S., Quinn, J.J., 2014. Radiological dose and risk assessment of landfill disposal of technologically enhanced naturally occurring radioactive materials (TENORM) in North Dakota. Report No. ANL/EVS-14/13. Argonne National Laboratory, Chicago, IL.

Homer, C.G., Dewitz, JA., Yang, L., Jin, S., Danielson, P., Xian, G., et al., 2015. Completion of the 2011 National Land Cover Database for the conterminous United States-representing a decade of land cover change information. Photogramm. Eng. Remote. Sens. 81 (5), 345-354.

Horner, R., Harto, C.B., Jackson, R.B., Lowry, E.R., Brandt, A.R., Yeskoo, T., et al., 2016. Water use and management in the Bakken shale oil play in North Dakota. Environ. Sci. Technol. 50 (6), 3275-3282.

IHS Markit, 2016. U.S. Petroleum Information/Dwights LLC Database. PIDM 2.5.3: Data Management System, Englewood, Colorado.

Iwanowicz, L.R., Blazer, V.S., Pinkney, A.E., Guy, C.P., Major, A.M., Munney, K., et al., 2016. Evidence of estrogenic endocrine disruption in smallmouth and largemouth bass inhabiting northeast U.S. national wildlife refuge waters: a reconnaissance study. Ecotoxicol. Environ. Saf. 124,50-59.

Kassotis, C.D., Tillitt, D.E., Davis, J.W., Hormann, A.M., Nagel, S.C., 2014. Estrogen and androgen receptor activities of hydraulic fracturing chemicals and surface and ground water in a drilling-dense region. Endocrinology 155 (3), 897-907.

Kassotis, C.D., Iwanowicz, L.R., Akob, D.M., Cozzarelli, I.M., Mumford, A.C., Orem, W.H., et al., 2016a. Endocrine disrupting activities of surface water associated with a West Virginia oil and gas industry wastewater disposal site. Sci. Total Environ. 557-558, 901-910.

Kassotis, C.D., Tillitt, D.E., Lin, C.-H., McElroy, J.A., Nagel, S.C., 2016b. Endocrine-disrupting chemicals and oil and natural gas operations: potential environmental contamination and recommendations to assess complex environmental mixtures. Environ. Health Perspect. 124 (3), 256-264.

Kile, D.E., Chiou, C.T., Zhou, H., Li, H., Xu, O., 1995. Partition of nonpolar organic pollutants from water to soil and sediment organic matters. Environ. Sci. Technol. 29 (5), 1401-1406.

Kondash, A., Vengosh, A., 2015. Water footprint of hydraulic fracturing. Environ. Sci. Technol. Lett. 2 (10), 276-280.

Kraemer, T.F., Reid, D.F., 1984. The occurrence and behavior of radium in saline formation water of the U.S. Gulf Coast region. Chem. Geol. 46 (2), 153-174.

Landa, E.R., Reid, D.F., 1983. Sorption of radium-226 from oil-production brine by sediments and soils. Environ. Geol. 5 (1), 1-8.

Lauer, N.E., Harkness, J.S., Vengosh, A., 2016. Brine spills associated with unconventional oil development in North Dakota. Environ. Sci. Technol. 50 (10), 5389-5397.

Lester, Y., Ferrer, I., Thurman, E.M., Sitterley, KA, Korak, J.A., Aiken, G., et al., 2015. Characterization of hydraulic fracturing flowback water in Colorado: implications for water treatment. Sci. Total Environ. 512-513, 637-644.

Lutz, B.D., Lewis, A.N., Doyle, M.W., 2013. Generation, transport, and disposal of wastewater associated with Marcellus shale gas development. Water Resour. Res. 49 (2), 647-656.

Madureira, T.V., Malhäo, F., Pinheiro, I., Lopes, C., Ferreira, N., Urbatzka, R., et al., 2015. Estrogenic and anti-estrogenic influences in cultured brown trout hepatocytes: focus on the expression ofsome estrogen and peroxisomal related genes and linked pheno-typic anchors. Aquat Toxicol. 169,133-142.

Maguire-Boyle, S.J., Barron, A.R., 2014. Organic compounds in produced waters from shale gas wells. Environ. Sci. Process. Impacts 16 (10), 2237-2248.

Mills, C.T., Goldhaber, M.B., Stricker, C.A., Holloway, J.M., Morrison, J.M., Ellefsen, K.J., et al., 2011. Using stable isotopes to understand hydrochemical processes in and around a Prairie Pothole wetland in the Northern Great Plains, USA. Appl. Geochem. 26 (S97-S100) (Supplement).

Mount, D.R., Gulley, D.D., Hockett, J.R., Garrison, T.D., Evans, J.M., 1997. Statistical models to predict the toxicity of major ions to Ceriodaphnia dubia, Daphnia magna and Pimephales promelas (fathead minnows). Environ. Toxicol. Chem. 16 (10), 2009-2019.

Moyo, F., Tandlich, R., Wilhelmi, B.S., Balaz, S., 2014. Sorption of hydrophobic organic compounds on natural sorbents and organoclays from aqueous and non-aqueous solutions: a mini-review. Int. J. Environ. Res. Public Health 11 (5), 5020-5048.

Nelson, A.W., Eitrheim, E.S., Knight, A.W., May, D., Mehrhoff, M.A., Shannon, R., et al., 2015. Understanding the radioactive ingrowth and decay of naturally occurring radioactive materials in the environment: an analysis of produced fluids from the Marcellus shale. Environ. Health Perspect. 123 (7), 689-696.

Orem, W., Tatu, C., Varonka, M., Lerch, H., Bates, A., Engle, M., et al., 2014. Organic substances in produced and formation water from unconventional natural gas extraction in coal and shale. Int. J. Coal Geol. 126, 20-31.

Peterman, Z.E., Thamke, J., Futa, K., Preston, T., 2012. Strontium isotope systematics of mixing groundwater and oil-field brine at Goose Lake in northeastern Montana, USA. Appl. Geochem. 27 (12), 2403-2408.

Piatt, J., Backhus, D., Capel, P., Eisenreich, S., 1996. Temperature-dependent sorption of naphthalene, phenanthrene, and pyrene to low organic carbon aquifer sediments. Environ. Sci. Technol. 30 (3), 751-760.

Plettner, I.N.A., Steinke, M., Malin, G., 2005. Ethene (ethylene) production in the marine macroalga Ulva (Enteromorpha) intestinalis L. (Chlorophyta, Ulvophyceae): effect of light-stress and co-production with dimethyl sulphide. Plant Cell Environ. 28 (9), 1136-1145.

Rivard, C., Lavoie, D., Lefebvre, R., Se'journe', S., Lamontagne, C., Duchesne, M., 2014. An overview of Canadian shale gas production and environmental concerns. Int. J. Coal Geol. 126, 64-76.

Roepke, T.A., Snyder, M.J., Cherr, G.N., 2005. Estradiol and endocrine disrupting compounds adversely affect development of sea urchin embryos at environmentally relevant concentrations. Aquat. Toxicol. 71 (2), 155-173.

Rogers, J.D., Burke, T.L., Osborn, S.G., Ryan, J.N., 2015. A framework for identifying organic compounds of concern in hydraulic fracturing fluids based on their mobility and persistence in groundwater. Environmental Science & Technology Letters 2 (6), 158-164.

Rowan, E.L., Engle, M.A., Kraemer, T.F., Schroeder, K.T., Hammack, R.W., Doughten, M.W., 2015. Geochemical and isotopic evolution of water produced from middle Devonian Marcellus shale gas wells, Appalachian basin, Pennsylvania. AAPG Bulletin 99 (2), 181-206.

Scanlon, B.R., Reedy, R.C., Nicot, J.P., 2014. Comparison of water use for hydraulic fracturing for unconventional oil and gas versus conventional oil. Environ. Sci. Technol. 48 (20), 12386-12393.

Scanlon, B.R., Reedy, R.C., Male, F., Hove, M., 2016. Managing the increasing water footprint of hydraulic fracturing in the Bakken Play, United States. Environ. Sci. Technol. 50 (18), 10273-10281.

Schoell, M., 1983. Genetic characterization of natural gases. AAPG Bull. 67 (12), 2225-2238.

Seaber PR, Kapinos FP, Knapp GL. Hydrologic Unit Maps. U.S. Geological Survey Water-Supply Paper 2294,1987, pp. 1-63.

Skalak K.J., Engle, MA, Rowan, E.L., Jolly, G.D., Conko, K.M., Benthem, A.J., et al., 2014. Surface disposal of produced waters in western and southwestern Pennsylvania: potential for accumulation of alkali-earth elements in sediments. Int. J. Coal Geol. 126, 162-170.

Soucek, D.J., Linton, T.K., Tarr, C.D., Dickinson, A., Wickramanayake, N., Delos, C.G., et al., 2011. Influence of water hardness and sulfate on the acute toxicity of chloride to sensitive freshwater invertebrates. Environ. Toxicol. Chem. 30 (4), 930-938.

Stewart, B.W., Chapman, E.C., Capo, R.C., Johnson, J.D., Graney, J.R., Kirby, C.S., et al., 2015. Origin of brines, salts and carbonate from shales of the Marcellus formation: evidence from geochemical and Sr isotope study of sequentially extracted fluids. Appl. Geochem. 60, 78-88.

Taylor, S.W., Sherwood Lollar, B., Wassenaar, I., 2000. Bacteriogenic ethane in near-surfaceaquifers: implications for leaking hydrocarbon well bores. 34. Environ. Sci. Technol. 34 (22), 4727-4732.

Thurston, R.V., Russo, R.C., Phillips, G.R., 1983. Acute toxicity of ammonia to fathead minnows. Trans. Am. Fish. Soc. 112 (5), 705-711.

U.S. EPA, 2013. Aquatic life ambient water quality criteria for ammonia - freshwater 2013. U.S. Environmental Protection Agency 822-R-13-001.

Vengosh, A., Jackson, R.B., Warner, N., Darrah, T., Kondash, A., 2014. Acritical review of the risks to water resources from unconventional shale gas development and hydraulic fracturing in the United States. Environ. Sci. Technol. 48 (15), 8334-8348.

Vidic, R.D., Brantley, S.L., Vandenbossche, J.M., Yoxtheimer, D., Abad, J.D., 2013. Impact of shale gas development on regional water quality. Science 340 (6134), 1235009.

Warner, N.R., Christie, CA, Jackson, R.B., Vengosh, A., 2013. Impacts of shale gas wastewater disposal on water quality in Western Pennsylvania. Environ. Sci. Technol. 47 (20), 11849-11857.

Warner, N.R., Darrah, T.H., Jackson, R.B., Millot, R., Kloppmann, W., Vengosh, A., 2014. New tracers identify hydraulic fracturing fluids and accidental releases from oil and gas operations. Environ. Sci. Technol. 48 (21), 12552-12560.

Wu, F., Fang, Y., Li, Y., Cui, X., Zhang, R., Guo, G., et al., 2014. Predicted no-effect concentration and risk assessment for 17-[beta]-estradiol in waters of China. Rev. Environ. Contam. Toxicol. 228,31-56.

Yost, E.E., Stanek J., DeWoskin, R.S., Burgoon, L.D., 2016. Overview of chronic oral toxicity values for chemicals present in hydraulic fracturing fluids, flowback, and produced waters. Environ. Sci. Technol. 50 (9), 4788-4797.

Young, W.F., Whitehouse, P., Johnson, I., Sorokin, N., 2002. Proposed Predicted-no-effect-concentrations (PNECs) for Natural and Synthetic Steroid Oestrogens in Surface Waters. (Research and Development, Technical Report P2-T04/1). Environment Agency, Bristol.