Purpose and Scope

Groundwater-quality data, reports, and information are sparse for the counties in northwestern West Virginia that are experiencing intense development of oil, gas, and other hydrocarbons, and were not adequate to assess and characterize the quality of groundwater within the study area.

To address this data gap, 30 rural residential water wells were sampled in 2018 for a broad range of chemical constituents. This study focused on rural residential water supplies in areas with a high density of oil and gas drilling, production, and pipeline transport to provide a baseline dataset of groundwater quality for current assessment and future comparison.

The groundwater samples were analyzed for chemical and physical properties, including nutrients, major ions, trace elements and metals, indicator bacteria, radionuclides, and methane and other dissolved hydrocarbon gases. The groundwater-quality data and summary statistics are presented to assess current groundwater-quality conditions in the region and are compared to drinking-water standards. In addition, the data were analyzed using multivariate statistical methods to assess potential correlations with geology, topographic setting, well depth, and proximity to oil and gas wells. Mineral saturation indices (SIs) were calculated to better understand reduction-oxidation (redox) and other geochemical processes that control groundwater quality. Principal component analysis was used to identify processes affecting groundwater quality and associations among dissolved constituents in groundwater. Concentrations of methane and other dissolved hydrocarbon gases were analyzed to help determine potential sources of, and potential processes affecting, methane and other dissolved hydrocarbon gases in groundwater. Groundwater ages determined from various tracers are presented and a conceptual model for the evolution of groundwater quality is discussed.

Description of Study Area

The study area for this investigation is within the Marcellus Shale oil and gas play in northwestern West Virginia and includes parts of 8 counties (Wetzel, Tyler, Doddridge, Ritchie, Wirt, Calhoun, Gilmer, and Roane) covering approximately 2,498 mi² (fig. 1) and is completely within the Appalachian Plateaus Physiographic Province (Fenneman, 1938; Fenneman and Johnson, 1946). In 2017, the 8-county area produced 913,899 thousand cubic feet (MCF) of natural gas, 3,023,712 barrels (bbl) of oil, and 4,495,150 bbl of natural gas liquids (wet gas; table 1; West Virginia Geological and Economic Survey, 2020a). Oil and gas production is greater in the northern part (Wetzel, Tyler, Doddridge, and Ritchie Counties) than in the southern part (Wirt, Calhoun, Gilmer, and Roane Counties) of the eight-county study area.

The study area consists of rugged, moderately incised hilly terrain with uplifted plateaus capped by resistant layers of relatively flat-lying sandstone and shale. Approximately 57,452 oil and gas wells are present in the 8-county study area as of 2021, some of which are active and others which are no longer producing or have been abandoned (West Virginia Geological and Economic Survey, 2020b).

Average annual precipitation for the study area, based on 1991–2020 station normals (table 2) for the National Weather Service weather station in Middlebourne, West Virginia (town of Middlebourne is shown on fig. 1), is 47.27 inches per year (in/yr). Maximum and minimum monthly temperatures for the station are 84.1 degrees Fahrenheit (°F) in July and 19.4 °F in January (National Oceanic and Atmospheric Administration, 2022). Precipitation is typically distributed unevenly with respect to topography, with areas at higher elevations typically receiving higher average amounts of precipitation than areas at lower elevations.

Public supplies are the principal source of water for 61.7 percent of residents and businesses in the study area. These public water supplies serve 41,478 people and account for 4.86 million gallons per day (Mgal/d) of freshwater withdrawals (2.0 Mgal/d of groundwater and 2.86 Mgal/d of surface water) for residential and commercial use. The remaining water use consists of 3.02 Mgal/d of groundwater withdrawals for residential and commercial supply and serves an estimated 37,774 people who rely on private domestic wells or other unregulated sources such as springs (table 3; U.S. Geological Survey, 2020). Rural residential water systems, primarily from individual wells, supply 38.3 percent of all water used for residential and commercial use in the study area. The majority of water used for public supply in the region (58.8 percent) is derived from surface-water sources, primarily from stream withdrawals, and the remaining part (41.2 percent) is derived from groundwater withdrawals. For rural residential homeowners, however, almost 100 percent of their withdrawals are derived from groundwater (private wells or springs).

Previous Investigations

Recent investigations of groundwater quality within the study area are limited but include Chambers and others (2012, 2015) and Harkness and others (2017). Harkness and others (2017) conducted a comprehensive study of the geochemistry of naturally occurring methane and saline groundwater in an area of unconventional shale gas development and the results are similar to those of this study. In fact, the study areas for the two studies overlapped significantly and both studies collected data in Doddridge, Ritchie, Tyler, and Wetzel Counties, West Virginia. Harkness and others (2017) collected 145 groundwater samples from 105 wells in the aforementioned counties and in Harrison County, West Virginia. In addition to sampling a large number of wells over a 5-county area in northwestern West Virginia where legacy and more recent Marcellus Shale oil and gas production has occurred, that study also analyzed for a broad array of isotopic data that were not collected for our investigation. The Harkness and others (2017) study monitored geochemical variations of drinking-water wells before, during, and after the installation of nearby shale gas wells and concluded that there was no indication of groundwater contamination and subsurface impact from shale-gas drilling and hydraulic fracturing within the study area. The study also found that saline groundwater was common and indicative of an upward flow of Devonian-age brines migrating into shallow aquifers and were modified by water-rock interactions. Occurrence of ethane, propane, and carbon isotope ratios of ethane indicate that thermogenic gas contributes to the overall mixture of natural gas in shallow aquifers in the study area. Dissolved gas in shallow groundwater, however, predominantly contained microbial methane, even though both biogenic and migrated thermogenic gases occur in shallow groundwater in the study area but are unrelated to shale gas development. That study did document surface water contamination that occurred in Tyler County from two injection well sites and a flowback spill related to shale gas development.

A USGS statewide assessment of groundwater quality within West Virginia (Chambers and others, 2012) indicated that Pennsylvanian and Permian fractured-bedrock aquifers, which are typical of the current study area, can have elevated concentrations of radon-222, a carcinogenic and radioactive gas known to cause lung cancer. The highest concentrations of radon-222 documented in the statewide assessment were for wells sampled in the Blue Ridge Physiographic Province in Jefferson County in the easternmost part of West Virginia. However, median radon-222 concentrations also were commonly elevated in wells sampled in Permian and Pennsylvanian age bedrock aquifers and typically exceeded the EPA proposed 300-picocurie per liter (pCi/L) maximum contaminant level (MCL) drinking water standard. Other than radon, the statewide study did not find chemical constituents of concern for wells sampled within the general area of study for this investigation.

A water-quality assessment of stream base flow and groundwater in the Marcellus Shale oil and gas play within the Monongahela River basin, an area immediately to the east of the current study area, was conducted in 2011 and 2012 (Chambers and others, 2015). The 2011–12 study was conducted prior to high intensity development of the Marcellus Shale oil and gas play to provide a baseline of groundwater and stream base flow data for future comparison. The principal groundwater findings of the previous study were based on 39 wells and 2 springs sampled June through September of 2011.

Principal components analysis conducted for the 2011–12 study specify that the first principal component indicates gradients of redox conditions and dissolved solids concentrations with strong negative loadings for sodium, bromide, chloride, fluoride, barium, total dissolved solids, specific conductance, pH, and methane, which are inversely proportional to dissolved oxygen, carbon dioxide, argon, nitrogen, and redox potential. These gradients may reflect a continuum from deeper waters with higher concentrations of sodium, bromide, chloride, fluoride, barium and methane (constituents typically found in higher concentrations in deeper brines) to shallow, more dilute, more recent waters with higher concentrations of dissolved atmospheric gases. Whether this continuum reflects a typical gradient from recharge from shallow groundwater to deeper groundwater or indicates other pathways resulting in mixing of shallow groundwater and deeper brines is unknown.

The second principal component had strong positive loadings for calcium, magnesium, potassium, iron, manganese, sulfate, total dissolved solids, specific conductance, argon, nitrogen, carbon dioxide, and hardness, and negative loadings for geologic age, methane, redox potential, bromide, and dissolved oxygen (Chambers and others, 2015). The second component reflects a gradient from conditions typical of shallow groundwater in the Appalachian region, which have high concentrations of total dissolved solids, iron, manganese, calcium, magnesium, and sulfate, to conditions more typical of deeper groundwater. Carbonate dissolution and reduction of pyrite and siderite are dominant processes in the shallow aquifers of the region and can result in elevated concentrations of total dissolved solids, iron, manganese, calcium, and sulfate, whereas higher concentrations of chloride and bromide are more common in deeper waters.

Dissolved gas analyses for the 2011–12 study indicate that the majority of sites (38 of 41) sampled were from a shallow groundwater source. Analysis of groundwater data collected for the study with respect to geologic structure, including wells located near deep through-going faults, did not reveal a clear pattern linking methane concentrations to the presence of geologic structure.

Other prior work focused on hydrogeology, brines and salt in waters, areas affected by coal mining, or basin studies. An assessment of groundwater hydrology of the Little Kanawha River in West Virginia (Hobba, 1980) provided a preliminary assessment of groundwater quality and water-well yields in the basin. Well yields were classified by topographic setting and median well yields ranged from a minimum of less than 1 gallon per minute (gal/min) on hilltops to 14 gal/min in valleys adjacent to the stream side of wide terraces. The report included a map of specific capacity in gallons per minute per foot (gal/min/ft) of drawdown showing areas with groundwater development potential in three categories: less than 2, 2–5, and greater than 5 gal/min/ft. The map also summarized groundwater quality for 795 previously collected water-quality samples and 48 samples collected for the project. Water samples from the 48 sites were analyzed for pH, silica, calcium, magnesium, sodium, potassium, bicarbonate, carbonate, chloride, fluoride, sulfate, nitrate, aluminum, iron, manganese, hardness, and total dissolved solids (TDS), and included maps showing the suitability of groundwater for residential and public supply in the basin.

One of the more comprehensive assessments for a part of the study area was an assessment of groundwater hydrology of the alluvial and bedrock aquifers bordering the Ohio River between Chester and Waverly, West Virginia (Bader and others, 1997), an area immediately to the west of the current study area. Well yields in fractured-bedrock aquifers ranged from 1 to 300 gal/min, with a median of 6 gal/min. Groundwater samples were compared to U.S. Environmental Protection Agency (2019) drinking-water standards. Eight percent of the wells sampled contained chloride concentrations in excess of the 250-milligrams per liter (mg/L) U.S. Environmental Protection Agency (EPA) SMCL drinking water standard for aesthetic effects such as taste, odor, and color. Iron and manganese concentrations ranged from 3 to 52,000 μg/L and from less than 1 to 1,900 μg/L, respectively. Median iron concentrations were 50 micrograms per liter (μg/L) in alluvial aquifers and 16 μg/L in fractured-bedrock aquifers. Median manganese concentrations were 510 μg/L in alluvial aquifers and 56 μg/L in fractured-bedrock aquifers. Arsenic, barium, mercury or phenol were detected at concentrations exceeding EPA drinking-water standards at 7 of the 50 wells sampled. The hardness of the water exceeded 120 mg/L in 91 percent of the samples, which is characteristic of hard to very hard groundwater which can cause scaling of water lines, boilers, hot water heaters, and water distribution systems.

An inventory and assessment of salt water produced in conjunction with oil and gas development in the Little Kanawha River basin of West Virginia, which is immediately southeast of the study area, was conducted (Kirwan and Rauch, 1988) prior to the advent of horizontal well drilling methods. Based on data from 1981 to 1984, their study estimated that 4.2 × 10⁷ gallons (gal) of salt water had been produced as part of oil and gas development in the watershed, and of that amount, 2.6 × 10⁷ gal had been disposed of as surface discharge, either on land or in streams. The 2.6 × 10⁷ gal of saltwater approximated 11.9 percent of the chloride and 1.7 percent of the TDS load within the Kanawha River during the period assessed.

A report on Appalachian connate water (Heck and others, 1964) provides a discussion of connate brines in West Virginia, Pennsylvania, and Ohio. Two additional reports provide a discussion of oil and gas resources in Pleasants, Wood, and Ritchie Counties (Haught, 1955) and Doddridge and Harrison Counties (Haught, 1959) in West Virginia. None of the reports provided information on the quality or quantity of potable groundwater for residential or public groundwater supplies.

The first salt well completed in West Virginia was completed in Kanawha County in 1812. In 1815 at Charleston, West Virginia gas was first encountered while drilling for salt (Eggleston, 2004). In 1859, the Rathbone brothers bored a salt well on burning Springs Run, but rather than encountering salt, they hit petroleum at a depth of 200 feet—which produced 200 barrels a day (Eggleston, 2004). The first purposely drilled well for oil in West Virginia was completed along the Burning Springs Anticline in Wirt County in 1860 by the Rathbone brothers and is known as the Rathbone well (Hennen, 1911). The well was begun in the later part of 1859 and was completed in May of 1860 to a depth of 303 ft and produced oil at a rate of 100 barrels per day (Hennen, 1911). Prior to that, oil was skimmed off the surface of Oil Springs Creek. Discovery of oil and gas produced one of the earliest oil and gas booms in the United States. The natural occurrence of salt water and hydrocarbons at shallow depths in the basin is common, especially along the Burning Springs Anticline, and both oil and gas production and natural flow of brines and hydrocarbons have the potential to locally affect groundwater and surface-water quality within the basin. The report contains several illustrations depicting the conceptual understanding of groundwater flow and interaction of fresh and saline water within the basin.

A study of the hydrology of the U.S. Office of Surface Mining Reclamation and Enforcement (OSMRE) Area 8 Eastern Coal Province in West Virginia and Ohio (Friel and others, 1987), which includes the entire current study area, assessed the quality of groundwater and surface water in an area of limited coal mining activities. About 1 million tons of coal were mined from 26 surface mines and about 6.7 million tons from 6 underground mines in the Area 8 Coal Province in 1980. Coal mining has declined within the study area in recent years (West Virginia Office of Miner’s Health Safety and Training, 2020; West Virginia Geological and Economic Survey, 2020c) and oil and gas production, including prolific development of the Marcellus Shale oil and gas play has increased (West Virginia Geological and Economic Survey, 2020b). Based on samples collected from 158 streams within the Area 8 Eastern Coal Province study area, the median specific conductance of surface water was 260 microsiemens per centimeter (μS/cm) and ranged from 41 to 4,035 μS/cm; the median pH was 7.3 standard units and ranged from 2.9 to 8.1; alkalinity ranged from 4.8 to 2,845 mg/L, with a median of 35.5 mg/L; total iron concentrations ranged from 100 to 565,000 μg/L, with a median of 630 μg/L; and dissolved manganese concentrations ranged from 10 to 5,100 μg/L, with a median of 64 μg/L. Dissolved iron and sulfate concentrations were highest in streams draining mined areas. Water well yields assessed in the study ranged from less than 1 to 350 gal/min. Highest median iron concentrations were from streams draining coal mined areas and were 16 times higher than for streams draining unmined areas.

An assessment of water resources of the Little Kanawha River basin in West Virginia provided a comprehensive evaluation of both groundwater and surface water resources within the basin, both with respect to water quality and water quantity (Bain and Friel, 1972). Average annual streamflow of the river was 3,100 cubic feet per second (ft³/s) or 1.35 cubic feet per second per square mile (ft³/s/mi²) and recharge to groundwater was calculated to be between 5 and 7 in/yr. Water in the Little Kanawha River and its tributaries are a calcium carbonate type of water. Wells inventoried for that project ranged in yield from 1 to 8 gal/min, with median yields in the Dunkard Group, Monongahela Formation, and Conemaugh Group fractured-rock aquifers of 7.0, 6.5, and 4.0 gal/min, respectively. Potable groundwater exists to depths less than 200 ft in the western part of the basin and to depths between 500 and 1,000 ft in the eastern part of the basin. The 1972 report contains maps of chemical constituent distribution including pH, iron, chloride, and hardness.

Selection of Sampling Sites

Sampling site selection was based primarily on two criteria. First, sites were selected to provide data from areas of active or legacy oil and gas production in northwestern West Virginia in areas of current Marcellus Shale gas, oil, and natural gas liquids (NGL) production. Second, sites were selected in areas with sparse groundwater-quality data in northwestern West Virginia. Surficial geology, which commonly is used as a site selection criterion, was not an important factor for site selection in this study. Most of the study area has surficial geology within the Pennsylvanian to Permian age Dunkard Group (50 percent), the Pennsylvanian age Conemaugh Group (22 percent), or the Monongahela Formation (16 percent), all of which have similar lithologies. To a lesser extent, the study area’s surficial geology is within the Allegheny Formation (10 percent) and there are minor exposures of the Kanawha Formation and Quaternary alluvial deposits (1 percent each).

Sample Collection

Groundwater samples were collected from 30 residential water wells in the study area from June 11 to July 18, 2018, by the USGS using standard USGS methods (Wilde and others, 2004; U.S. Geological Survey, 2006). Prior to sampling, wells were purged to remove standing water from the well and ensure that representative water samples were collected. Wells were purged for a sufficient period to allow pH, dissolved oxygen, specific conductance, and water temperature to stabilize, and then water samples were collected and processed. Prolonged 3-volume purging of the wells was not required because the wells sampled were primarily rural residential wells with submersible pumps and are purged daily as part of routine use. Methods applicable for low-yield wells were utilized at a few wells to avoid pulling the water level down to the pump intake, which can cause turbidity issues, aerate the sample water, and potentially damage the well pump. Periodic water-level measurements were made and served as the criteria for length of the well purge in conjunction with close monitoring of dissolved oxygen, pH, water temperature, and specific conductance using a multiparameter water-quality sonde, which was calibrated daily.

For the wells sampled, plumbing and well-casing materials included steel, galvanized steel, polyvinyl chloride (PVC), and other plastics. The purging procedure minimized potential contamination from the well casing and plumbing, as non-contaminating Teflon sample tubing was connected as close to the wellhead as possible, usually at the pressure tank, prior to any treatment such as a water softener or chlorinator, and pumps were kept running as much as possible to prevent sample contamination from the plumbing or backflow from holding tanks. The sample line was then fed to a manifold with sampling ports and a port for connection of a multiparameter water-quality sonde. Field properties were measured in the flowthrough apparatus and monitored until parameters stabilized prior to sampling. To prevent environmental contamination, samples typically were collected and processed inside a mobile field laboratory or a portable processing chamber assembled near the closest spigot on the plumbing system to the well prior to any water treatment equipment such as chlorinators or water softeners, typically at the pressure tank.

Samples for bacterial analysis were collected and processed according to standard USGS methods (Myers and Sylvester, 2014) and processed using the Colilert system (IDEXX, 2019) according to established methods (American Public Health Association, American Water Works Association, and Water Environment Foundation, 2018), a defined substrate liquid-broth medium method for determination of total coliform bacteria and E. coli. Water samples for bacterial analysis were collected by cleansing the spigot on the pressure tank or water valve on the discharge line from the pump with isopropyl alcohol; the spigot then was allowed to thoroughly dry before filling a pre-sterilized 100-milliliter (mL) sample bottle. Bacteria samples were processed within 2 hours of collection and incubated for 24–28 hours, according to USGS standard protocols.

All samples from the 30 wells were analyzed for field properties (pH, water temperature, specific conductance, and dissolved oxygen), turbidity, alkalinity, major ions, metals, trace elements, total coliform and E. coli. bacteria, uranium, radium-226, radium-228, and gross alpha and gross beta radiation (table 5). Radon-222 was analyzed for samples from 28 of the 30 wells (two samples were lost in shipment). A subset of samples from 17 of the 30 wells were collected for analysis of dissolved hydrocarbons, sulfur hexafluoride (SF₆), CFCs, deuterium, oxygen-18, tritium, and helium as a methods development process for a secondary but related pilot study for a new USGS dissolved hydrocarbons and isotope laboratory established concurrent with this study in Reston, Virginia. Samples were processed according to standard USGS protocols (Wilde and others, 2004). Measurements of water temperature, specific conductance, dissolved oxygen concentration, pH, turbidity, and alkalinity were made on site at the time of sampling.

Quality Assurance and Quality Control

Three types of QA/QC samples were collected during the project—field blanks, equipment blanks, and replicates. Replicates were run on environmental samples to assess the reproducibility of analytical methods and assess bias that may have resulted in the laboratory analysis. Field blanks were run to assess any contamination that may have resulted from field sampling methods, to evaluate decontamination procedures between sites, and to detect contamination on sampling equipment during transit to or from the sampling site. Equipment blanks were collected in the laboratory at the USGS office in Charleston, West Virginia, and field blanks were run in the field to assess the potential of contamination from blank water used to process field and equipment blanks, the deionized water and detergent solutions used to decontaminate equipment between sampling sites, and to assess whether residual contamination was being carried over from site to site. All quality assurance data for the water samples collected for this study are stored in the quality assurance part of the USGS quality of water data (QWDATA) database for West Virginia (U.S. Geological Survey, written commun., 2022).

Variability for a replicate sample pair was quantified by calculating the relative percent difference (RPD) of the samples. The RPD was calculated using the following formula

where R1 is the concentration of the analyte in the first replicate sample and R2 is the concentration of the analyte in the second replicate sample. Concentrations of replicate sample pairs differed by small amounts, typically less than 15 percent of the RPD. Constituents with higher RPD were typically constituents at low concentrations at or below the method detection limit. Analytical or sampling variability was considered minimal as a result.

CFC, SF₆, Tritium, Dissolved Gas, and Dissolved Hydrocarbon Sampling

Samples for CFCs, SF₆, tritium, and dissolved gas analysis were collected according to procedures described by the USGS Groundwater Dating Laboratory (https://water.usgs.gov/lab). The CFC samples were collected in 125-mL glass bottles with metal foil sealed caps. The SF₆ samples were collected in 1-L borosilicate amber glass bottles with polycone seal caps. Tritium samples were collected in 500-mL high density polyethene bottles with polycone seal caps. Dissolved gas samples were collected in pairs in pre-weighed 125-mL bottle pairs with butyl rubber septa. After sampling, the dissolved gas bottles were stored on ice and refrigerated until analysis to minimize bacterial activity. The CFC, SF₆, and dissolved gas bottles were purged and filled under water to prevent contamination from outside air.

Dissolved hydrocarbon samples were collected by filling a 1-L bottle to overflowing in a bucket then allowing a minimum of three volumes of sample water to flush through the sample bottle. The bottle was then removed from the water and three potassium hydroxide preservative tablets were added to the sample and the sample bottle was capped and taped shut. Samples were kept chilled in a cooler and then transferred to a refrigerator, prior to shipment to the USGS Groundwater Dating Laboratory in Reston, Va.

Analysis of Water Chemistry

Major ions, nutrients, metals, trace elements, and radon-222 were analyzed using standard EPA or USGS methods at the USGS National Water Quality Laboratory in Denver, Co. Trace elements and common ions, such as calcium, magnesium, sodium, potassium, iron, manganese, sulfate, chloride, bromide, and fluoride, were analyzed by inductively coupled plasma (ICP) mass spectrometry; nutrients (nitrate + nitrite, nitrite, total nitrogen, and orthophosphate) were analyzed by digestion or colorimetric methods; and radon-222 was analyzed by liquid scintillation methods. Reporting limits for these analyses are provided in table 5.

CFC, SF₆, and Dissolved Gas Analysis

CFCs (CFC-11, CFC-12, and CFC-113), SF₆, and dissolved gas samples for nitrogen, argon, methane, carbon dioxide, and oxygen analysis were shipped to the USGS Groundwater Dating Laboratory in Reston, Va. for analysis using gas chromatography methods (Busenberg and Plummer, 1992, 2000).

Tritium Analysis

Tritium samples were processed in the USGS Groundwater Dating Laboratory using Helium-3 (³He) ingrowth analysis using methods similar to those described by Schlosser and others (1988) and Papp and others (2012). For ingrowth, 500-gram (g) water samples were transferred in 1-L volume stainless steel vessels fitted with vacuum-tight copper tube stubs. The samples were on a vacuum manifold with a pressure of less than 1×10⁻⁴ torr, created by a stainless-steel liquid nitrogen trap and a rotary vane pump removing air-sourced ³He below the limits of detection. After degassing, the samples were clamped and then stored for at least 6 months to allow for tritium to decay to ³He. After the ingrowth interval, the canisters were attached to an automated vacuum manifold for ³He extraction and analysis on Thermo Scientific Helix SFT Split Flight Tube mass spectrometer. On the manifold, the ³He and any residual gases were extracted from the canisters, the gases were removed with cryogenic traps and reactive getters, and the ³He was quantitatively captured on a Gifford-McMahon cycle cryocooler with charcoal matrix at less than 10 Kelvin (K). The ³He from tritium decay was then expanded into the Helix SFT for analysis. The ³He samples were calibrated against samples of air of known ³He abundance and samples were corrected for any residual air-sourced ³He in the ingrowth canisters. Analysis of replicate samples and blind intercomparison showed method precision better that 1 percent relative standard deviation (RSD) and a lower reporting limit of 0.01 tritium unit (TU).

Dissolved Hydrocarbon Analysis

Dissolved hydrocarbon concentrations were analyzed by the USGS Dissolved Gas Laboratory. The dissolved C₁ to C₆ hydrocarbons (methane, ethane, ethene, ethyne, propane, propene, propyne, n-butane, isobutane, 1-butene, n-pentane, isopentane, 2- and 3-methylpentane, hexane, and benzene) were analyzed using a custom purge and trap gas chromatography-flame ionization detector and atomic emission detector (GC-FID/AED) system developed for quantifying trace volatile gases in water. This analytical method is sensitive to trace constituents in the picomole range, while maintaining enough dynamic range to enable quantification of samples with considerable amounts of dissolved gas and has been employed in numerous studies where trace hydrocarbon concentrations are of interest (Haase and others, 2014; Cozzarelli and others, 2017; Orem and others, 2017; Kozar and others, 2020). Duplicate samples for dissolved hydrocarbons were collected in 1-L borosilicate glass bottles, as discussed above, and refrigerated until analysis to inhibit bacterial activity and degassing. Dissolved hydrocarbon data for this study are published in a USGS data release (Haase and others, 2022).

The dissolved hydrocarbon analytical system was calibrated using a suite of certified gas standards with blend accuracies greater than 10 percent. Additionally, quality control samples consisting of homogenized tap water, nitrogen-purged water, and water purged with air containing parts-per-million to parts-per-billion by volume mixing ratios of hydrocarbon gases were collected in the laboratory in a manner identical to the field samples and analyzed to verify analytical performance (Haase and others, 2022). Methane was quantified exclusively on the flame ionization detector (FID) because concentrations are typically much higher than other hydrocarbons (Prinzhofer and others, 2000). The FID detection limit is 4.97 nanograms per kilogram (ng/kg) and the calibration precision is 0.9 percent RSD. The C₂ to C₆ hydrocarbons were simultaneously measured on the atomic emission detector (AED) and the FID, with the reported value coming from the highest sensitivity detector that was still in the range of linearity, which was typically the AED for most measurements of these compounds. The detection limits of the AED are typically in the range of 6.75 to 25.1 × 10⁻³ ng/kg, with calibration precisions between 1 and 14 percent, replicate quality control samples had concentration RSDs between 4 and 10 percent, and median percent differences between duplicate field samples were 2 to 56 percent.

Geospatial Analysis

Geospatial analysis for this project was conducted by incorporating requisite project data into ArcGIS version 10.6 (Esri, 2020). Data layers compiled or created for the ArcGIS Arc-Map project included but were not limited to (1) the locations of water wells sampled during the project, (2) locations of current and legacy vertical and more recent horizontal oil and gas wells, (3) a geologic map of the study area, and (4) significant geographic features such as locations of public water lines, roads, streams, cities, towns, county boundaries, and land use. The primary tasks of the spatial analysis were to determine (1) the extent of oil and gas well production near the sampled residential water wells (within either a 500-m or 1,000-m radius) in the study area, and (2) the geologic formation in which the sampled water wells were completed. Thus, GIS was used to determine the extent of oil and gas production and geology surrounding each residential water well sampled, which were the primary factors considered to potentially affect groundwater quality in the study area, although other factors, such as well depth and topographic setting, also were evaluated as part of the study.

Statistical and Graphical Analysis

Statistical and graphical techniques were used to summarize and compare water chemistry and field parameters among different sites according to spatial location, geology, proximity to oil and gas wells, and topographic setting. Statistical analyses were conducted and graphics were created with the R statistical computing environment version 3.4.0 (R Core Team, 2017). Non-parametric techniques were used for computing descriptive and multivariate statistics from water-quality data that were, in some instances, censored at multiple levels. Censored data are low-level concentrations of chemicals with values between zero and the laboratory’s reporting limit. The robust regression on order statistics (ROS) survival analysis method was used to calculate summary statistics for censored data because of the relatively small sample sizes and to avoid transformation bias of non-normal water-quality data (Helsel, 2012). Scatter plots were used to understand the relation between ion concentrations, mineral saturation indices, and pH. Tukey-type boxplots, censored at the highest reporting limit (Lorenz, 2018) were created to understand the relation between water quality and site characteristics, whereas trilinear Piper diagrams (Piper, 1944; Back, 1966) were used to show a graphical representation of major ion chemistry. Boxplots and Piper diagrams were created with the USGS smwrGraphs R package and, when necessary, the robust ROS method was employed to impute values for censored water-quality data (Lorenz and Diekoff, 2017). For statistical analysis, data for the wells sampled in the Conemaugh Group and Monongahela Formation were combined for multivariate statistical analysis because of the low sample size (only 7 and 4 wells sampled in the Monongahela Formation and Conemaugh Group, respectively), but were graphed as separate units in the boxplots and trilinear diagrams. Hillside and hilltop wells were also combined and labeled as upland wells for multivariate statistical and graphical analysis.

Prior to multivariate statistical analysis, and because of the presence of multiple censoring levels, censored data were re-coded to u-Scores with the codeU function in the USGS smwrQW package (Lorenz, 2018). The u-Score is the sum of the sign of the differences between each value and all other values and is equivalent to the rank but scaled so the median is equal to zero. Using u-Scores allows for the computation of multivariate relations without requiring censoring at the highest reporting limit and retains information at multiple reporting limits. In cases where a column of data has only one censoring level, the u-Scores are the same as ordinal methods of ranking for one reporting limit (Helsel, 2012).

Principal components analysis (PCA) was used to identify relations between the major chemical and hydrological processes that could explain dissolved element concentrations in the water-quality dataset. Principal component analysis is a multivariate statistical analysis method that allows rapid analysis of large datasets and extracts the eigenvectors and eigenvalues from a covariance matrix. It can be used to understand the intercorrelations of many variables and provide insight into underlying hydrogeochemical processes (Davis, 1986). PCA was computed with the principal function in the R psych package (Revelle, 2022), which first computes correlation coefficients (Spearman’s rho) for the raw u-Scores and then performs a PCA on the resulting correlation matrix (app. 1). Varimax rotation was applied to simplify the structure of the PCA model, which maximized the differences in components and aided in the interpretation of results. Water-quality variables that had missing values or were censored in more than 40 percent of the values were excluded from the PCA. The variable loadings from the varimax-rotated PCA were used to determine the master variables for each rotated component. Resulting loadings from the PCA and statistically significant correlations (p less than 0.01) from the correlation matrix were retained and used for further interpretation of the dataset.

The Wilcoxon Signed Rank Test (Helsel, 2012) was run to determine whether there were statistically significant differences between (1) the geologic formations, (2) the topographic setting (uplands or valleys) in which the water wells are present, (3) the depth of water wells sampled (less than 95 ft in depth or greater than or equal to 95 ft in depth), and (4) with respect to the number of oil and gas wells in proximity to the water wells sampled for the study (pre-1930 oil and gas wells, Marcellus Shale oil and gas play wells only, and oil and gas wells within 500 or 1,000 m of residential water wells sampled for the study). For the Wilcoxon Signed Rank Test, p less than 0.05 was used to determine if differences between groups of data are statistically significant. When p is less than 0.05, it indicates that groups are statistically different from one another, whereas if p is greater than 0.05, it indicates that differences between the two compared groups are not statistically significant.

Additional analyses were conducted by preparing trilinear diagrams showing the overall composition of the groundwater for the wells sampled with respect to the (1) geologic formations, (2) topographic setting (uplands or valleys) in which the water wells are present, and (3) depth of water wells sampled (less than 95 ft in depth or greater than or equal to 95 ft in depth). These factors commonly are found to be related to variability of groundwater quality in West Virginia. A similar approach for data analysis was used for a companion study of groundwater quality in areas of active and legacy coal mining in southern West Virginia (Kozar and others, 2020).

Finally, statistical tables of the mean, median, maximum, and minimum concentrations and the standard deviation of the data were developed for the data collected from the residential water wells sampled and assessed with respect to (1) overall groundwater quality as compared to EPA, USGS, OSMRE, and other drinking-water standards, and with respect to the (2) various geologic formations, (3) topographic setting (uplands or valleys) in which the residential wells are present, and (4) depth of wells sampled (less than 95 ft in depth or greater than or equal to 95 ft in depth).

Geochemical Modeling

Aqueous speciation and mineral saturation indices were computed with the geochemical modeling software PHREEQC (Parkhurst and Appelo, 2013). The mineral saturation index (SI) is a measure of whether a mineral has the potential to dissolve or precipitate depending on the conditions of the solution. The mineral SI is determined by dividing the ion activity product (IAP) by the thermodynamic solubility product (Ksp) and then taking the logarithm of the quotient [log(IAP/ Ksp)]. When a solution is at equilibrium, the SI is zero. In solutions where the SI is greater than zero (IAP is greater than Ksp), the solution is said to be supersaturated and the specified mineral, if present, is not likely to dissolve and may precipitate. In solutions where the SI is less than zero (IAP is less than Ksp), the solution is said to be undersaturated and the mineral, if present, could dissolve and would not precipitate (Benjamin, 2002).

Quality Assurance Results

The QA samples consisted of one replicate environmental sample pair for various analyses, one field blank, and one equipment blank. Analyte concentrations of a replicate sample pair typically differed by small amounts, less than 5 percent RPD. The only constituents that exceeded the 5-percent RPD threshold were radionuclides including gross alpha activity (one replicate pair with an RPD of 32 percent), radium-224 (one replicate pair with an RPD of 34 percent), radium-226 (one replicate pair with an RPD of 14 percent), uranium-234 (one replicate pair with an RPD of 12 percent), uranium-235 (one replicate pair with an RPD of 6 percent), and uranium-238 (one replicate pair with an RPD of 31 percent). However, the relatively large RPD for the replicate samples for radionuclides (excluding radon-222) reflects relatively greater uncertainty at low values near method reporting levels, as the field and replicate analyses for radium-224, radium-226, radium-228, and uranium were all at very low concentrations or less than the method detection limits. Of the remaining 39 constituents, the RPD could not be quantified accurately for nine constituents, as analytical results for both replicate samples were less than the method detection limits. The constituents where RPD could not be quantified accurately included beryllium, cadmium, chromium, cobalt, copper, mercury, nickel, silver, zinc, antimony, selenium, and uranium, principally due to a large number of non-detect concentrations and possibly to limitations in analytical methods.

Constituent concentrations for one field blank collected at the first site sampled generally were less than the method detection limit, indicating that field collection and processing procedures for samples were adequate to prevent cross contamination of environmental samples collected for the project. Trace metal concentrations for barium, cobalt, lead, manganese, and molybdenum indicate low level contamination or analytical bias with blank concentrations of 1.2, 0.04, 0.03, 0.717, and 0.064 µg/L, respectively. Due to the low concentrations and large number of non-detects for cobalt and lead, some bias in those data is possible, but environmental concentrations for barium, manganese, and molybdenum were orders of magnitude higher than the concentrations in the blank samples. For the replicate sample submitted for radionuclide analysis, concentrations of gross alpha activity, gross beta activity, radium-224, radium-228, uranium-235, and uranium-238 were all below the method detection limits, and radium-226 and uranium-234 were detected at concentrations of 0.022 and 0.043 pCi/L, respectively. Given that the replicate samples were sampled at the same time as the field blank and had similar concentrations, and concentrations of the radionuclide constituents were typically below the method detection limit, no significant variability with respect to analytical methods is evident.

Equipment blank data are QA samples collected in a controlled laboratory environment to assess the cleanliness of the equipment used, to test the quality of the water used for field and laboratory blanks, and to assess laboratory analytical contamination. One equipment blank was collected for the study and the data for common ion and trace metal analysis all were less than the method detection limits, indicating no bias with respect to the equipment, blank water, or laboratory analysis methods for these constituents. Equipment blank concentrations of gross alpha activity, gross beta activity, radium-224, radium-226, radium-228, uranium-235, and uranium-238 were all below the method detection limits, and uranium-234 was detected at a concentration of 0.032 pCi/L.

Overall, the QA samples indicate that data collected for the project were acceptable for the goals of the project, with minor potential field collection contamination with respect to cobalt and lead, with field blank concentrations of 0.040 and 0.030, µg/L, respectively.

Relation to Drinking-Water Standards

A primary impetus for this project was concern that trace metals, radionuclides, or other contaminants present in water from unregulated residential wells may pose a health threat to individuals that rely on these sources for their water supply. In addition, there is concern that legacy as well as current oil and gas development and production activities may adversely affect the quality of groundwater upon which residents in the study area rely for their primary source of water. As a result, groundwater-quality data collected for this study are compared to public drinking-water standards, even though those regulations are not enforceable for rural residential water supplies.

Analyte concentrations were compared to several drinking-water standards (table 7) including EPA’s maximum contaminant levels (MCLs), maximum contaminant level goals (MCLGs), health-based values (HBVs), surface water treatment rule (SWTR) treatment techniques (TT), proposed MCLs, alternate proposed maximum contaminant levels (AMCLs) for radon-222, and drinking water equivalent levels (DWEL). Full descriptions of EPA primary and secondary drinking-water standards and health advisories may be accessed at EPA web sites (U.S. Environmental Protection Agency, 2022a; 2022b). USGS health-based screening levels (HBSLs) and the OSMRE level of concern (LOC) and immediate action level (IAL) for methane in groundwater also are provided to assess the overall quality of groundwater in the study area. The EPA HBV of 20 mg/L for sodium for those on a sodium-restricted diet also is included. Full descriptions of the USGS HBSLs given by Norman and others (2018) may be accessed online at the USGS Health-Based Screening Levels website (https://water.usgs.gov/water-resources/hbsl/). Secondary maximum contaminant levels are standards that relate to aesthetic issues such as color, odor, taste, and staining of plumbing fixtures, rather than health-based criteria.

Overall, 21of 30 constituents analyzed exceeded drinking-water standards in one or more samples, including arsenic, turbidity, and bacteria that exceeded MCLs; pH, hardness, chloride, fluoride, TDS, iron and manganese that exceeded SMCLs; radon-222 that exceeded a proposed MCL; manganese that exceeded an HBSL; and sodium that exceeded the EPA HBV (table 7). Boxplots showing the distribution of some of the constituents that exceeded one or more of the drinking-water standards in 10 percent or more of the wells sampled are shown in figure 7.

Relation to Geologic Formation

The composition of the bedrock in which the water wells are completed can affect the quality of groundwater derived from the rural residential wells sampled for this study. There is substantial variability with respect to lithologic composition between the various geologic formations (fig. 8A, table 10) and even within individual strata that comprise the geologic formations (figs. 9, 10). There were statistically significant differences with respect to calcium, magnesium, potassium, silica, and turbidity (table 11). Calcium, magnesium, potassium, silica, and turbidity concentrations were statistically significantly higher in the Dunkard Group than in the Conemaugh Group or Allegheny Formation (tables 10, 11). Although not statistically significant, median sodium concentrations are highest in wells sampled in the Monongahela Formation (fig. 11). The surface geology does not allow for assessing from which specific strata (sandstone, limestone, siltstone, shale, or coal) a particular well may be deriving water. In fact, wells may obtain water from multiple intervals and therefore the following discussion is based on the major geologic groups and formations within the study area (Conemaugh Group, Monongahela Formation, or Dunkard Group).