Surface-Water-Influenced Groundwater Supply (SWIG)

A SWIG, as defined by West Virginia Code §22-30-3 (West Virginia Legislature, 2014), is “a source of water supply for a public water system which is directly drawn from an underground well, underground river or stream, underground reservoir or underground mine, and the quantity and quality of the water in that underground supply source is heavily influenced, directly or indirectly, by the quantity and quality of surface water in the immediate area.” However, this report assumes a previously published, modification of a SWIG to include “a groundwater supply that is heavily influenced by water recharging the well from adjacent rivers, streams, ponds, lakes, irrigation water, or even precipitation that pools at the surface” (Kozar and Paybins, 2016, p. 2). This definition permits the inclusion of additional sources of contamination from shallow surface processes (for example, crop irrigation) that may affect sources prone to rapid infiltration.

Determination of the intrinsic susceptibility and vulnerability is critical when assessing if an aquifer may be a SWIG (Kozar and Paybins, 2016). Intrinsic susceptibility includes factors like hydrogeologic properties that influence fluid transport in the unsaturated zone of an aquifer (Focazio and others, 2002). Vulnerability of an aquifer includes extrinsic factors like proximal contaminant sources and the ease of contaminant transport into and through an aquifer, which may be exacerbated by well construction design, active physical and chemical subsurface processes, and the intrinsic susceptibility of the aquifer (Eberts and others, 2013).

As SWIGs, portions of the Ohio River alluvial aquifer are defined by high intrinsic susceptibility and high vulnerability to nonmicrobial contaminants (for example, volatile organic compounds, nitrate). High transmissivities in the alluvium (4,800 feet squared per day; Kozar and Mathes, 2001) allow rapid infiltration of overland flow, either from precipitation or flooding, and induced infiltration directly from the Ohio River because of pumping (Kozar and Paybins, 2016). Progressive urbanization, industrialization, and agricultural development along the Ohio River has created multiple sources of contamination and removed surficial clays that retard infiltration (Ferrel, 1987). This removal simultaneously increases the intrinsic susceptibility and vulnerability of the alluvial aquifers that are a major groundwater resource in West Virginia, underlining the need to refine current [2019] delineations of SWIGs in the Ohio River alluvium.

In 2019, the U.S. Geological Survey (USGS) began a study in cooperation with the West Virginia Department of Health and Human Resources for the purpose of understanding the influence of surface water on groundwater chemistry in the alluvial aquifers bordering the Ohio River in West Virginia. The study was limited to the alluvial aquifers along the Ohio River in the Appalachian Plateau physiographic province of West Virginia (fig. 1) and does not include alluvial deposits of the Kanawha River or other minor tributaries. This study focused on surface-water influence in the Ohio River alluvial aquifer but not groundwater under direct influence (GWUDI) of surface water as defined by the National Primary Drinking Water Regulation (40 CFR 59570 part 141). As of 2023, no alluvial aquifers in West Virginia are classified as GWUDI, despite high intrinsic susceptibilities, because of a lack of detected pathogens (for example, virus or bacteria) and microscopic particulates in raw process water (Kozar and Paybins, 2016).

Purpose and Scope

This report describes and interprets water-quality data studied by the U.S. Geological Survey in cooperation with the West Virginia Department of Health and Human Resources to help understand the influence of surface water on groundwater chemistry in the alluvial aquifers bordering the Ohio River in West Virginia. This assessment is based on the chemical composition of surface-water samples collected from 4 sites on the Ohio River and groundwater samples collected from 23 wells in alluvial aquifers from June 2019 to January 2020. Chemical analyses of samples included major ions, dissolved organic carbon (DOC), selected man-made organic compounds, dissolved gases, tritium, and carbon-13 isotopes. Specific objectives of this study are 1) characterization of groundwater quality in relation to health-based standards and geochemical processes, 2) determination of recharge sources, 3) estimation of groundwater age, and 4) estimation of the fraction of Ohio River water entering groundwater that supplies wells in the Ohio River alluvial aquifer in the state of West Virginia. Results of multivariate analyses of water-quality samples are discussed to provide insight into geochemical processes controlling groundwater. Binary mixing models and inverse geochemical models were used to estimate the fraction of surface water in groundwater pumped by wells in the alluvial aquifer and compared to fractions estimated from previously published numerical groundwater-flow models. An improved understanding of SWIGs along the Ohio River will aid preventative efforts seeking to protect the alluvial aquifers from potential degradation associated with surface activities.

Study Area and Previous Investigations

The Ohio River alluvial aquifer in West Virginia is in the Appalachian Plateau physiographic province along the West Virginia-Ohio border (fig. 1) with alluvial sediments confined to river terraces and floodplains within the Ohio River Valley (Puente, 1985). It represents one of two major aquifer types and one of five hydrogeologic terrains (lithologies that share similar hydrogeologic properties) in the State. For context, consolidated sedimentary bedrock aquifers represent the other major aquifer type (Schwietering, 1981; Puente, 1985).

The predominantly mountainous terrain and geographic setting have a pronounced effect on climate, which is classified as temperate continental with four well-defined seasons that are locally influenced by topography (Friel and others, 1987; Battelle Memorial Institute, 2003). Average maximum and minimum temperatures from 1900 to 2016 ranged from 40 degrees Fahrenheit (°F) to 65°F, respectively. Mean precipitation, which is affected by orographic lift as atmospheric currents track eastward (Friel and others, 1987), over the same time period was 42.4 inches (Kutta and Hubbart, 2019).

The Ohio River Valley is highly developed and consists of residential, industrial, and agricultural land use. The estimated population of the 10 counties included in this study (fig.1) was 286,954 in 2022, approximately 16 percent of the State’s total population (U.S. Census Bureau, 2023). Development, particularly agriculture, decreases away from the river because of rugged terrain that typifies the Appalachian Plateau physiographic province. Residential land use is a result of the Ohio River facilitating transport across the mountainous topography during settlement, which has spawned a diverse industrial presence including electric-power generation, chemical and metal manufacturing, and petroleum refinement that supports municipal growth alongside agriculture activity (Ferrel, 1987; Bader and others, 1997; Battelle Memorial Institute, 2003). Agricultural land use, which is concentrated in the Ohio River Valley, has declined by 58 percent since 1950 (Kutta and Hubbart, 2019).

Sampling Methods

To prevent environmental contamination, samples were collected and processed inside a mobile field laboratory or a portable processing chamber assembled near the sampling location. Surface-water samples were collected by scientists on a boat at multiple stations across the Ohio River using a 1-liter Nalgene narrow mouth bottle and weighted bottle sampler. Samples were width-integrated from eight vertical stations across the river, and the vertical samples collected were poured into an 8-liter polyethylene churn splitter for sample processing. A multi-parameter water-quality sonde (YSI EXO 2, Yellow Springs, Ohio) was used to measure water temperature, pH, specific conductance, dissolved oxygen, and turbidity according to procedures described in the USGS National Field Manual (U.S. Geological Survey, variously dated).

Groundwater samples were collected at public water system wells in the Ohio River alluvium. Raw-water taps, as identified by the operator at the public water system, were tested with commercially available chlorine test strips to ensure the sample point was located before the system’s disinfection processes. Often located at 3/8-inch hose bibs, raw-water taps came in many configurations including lab faucets, threaded and unthreaded plumbing connections made of various metals, PVC pipes, and other plastic connections. Standard sample tubing was connected to raw-water taps using a combination of nylon connectors, stainless-steel fittings, and hose clamps to ensure an airtight connection. Sample tubing was connected to a flow-through chamber with a YSI multiparameter water-quality sonde, which was calibrated daily and measured temperature, pH, specific conductance, dissolved oxygen, and turbidity. All samples were collected at active high-production wells and did not require purging based on well volume. Field parameters were monitored for a minimum of 25 minutes and readings were recorded every 5 minutes to meet stability criteria according to the National Field Manual (U.S. Geological Survey, variously dated) and to collect enough data to calculate median values for each parameter. After field parameters were recorded, samples were collected in recommended sample containers and preserved according to lab instructions.

All samples (surface water and groundwater) were analyzed for field measurements of water quality (pH, water temperature, specific conductance, dissolved oxygen concentration, and turbidity), alkalinity, major ions, trace elements, nutrients, DOC, and per- and polyfluoroalkyl substances (PFAS; table 2). Subsets of samples at proximal wells were analyzed for microbiological indicators of fecal contamination (table 3), VOCs (table 4), semi-volatile organic compounds (SVOCs; table 5), and pesticides (table 6) to assess potential aquifer degradation from surface-water influence. Field measurements and samples for inorganic analytes, VOCs, SVOCs, DOC, and pesticides were processed using standard USGS protocols described by the USGS National Field Manual for the Collection of Water Quality Data (U.S. Geological Survey, variously dated). These protocols specified that samples collected for determination of dissolved concentrations (major ions, trace elements, nutrients, DOC) were filtered in the field through a 0.45-micron filter and, for some analyses (cations, trace elements), preserved with nitric acid.

Samples collected for determination of major inorganic element, trace element, nutrient, VOC, SVOC, and pesticide concentrations were chilled on ice and shipped overnight to the USGS National Water Quality Laboratory in Lakewood, Colorado, for analysis. Samples collected for determination of microbiological indicators of fecal contamination were chilled on ice and shipped overnight to the USGS Ohio Microbiology Laboratory in Columbus, Ohio. Samples for determination of PFAS concentrations were analyzed by RTI Laboratories in Livonia, Michigan, according to PFAS analysis compliant with U.S. Department of Defense Quality Systems Manual (U.S. Department of Defense, 2019).

Quality Control and Quality Assurance

Quality-assurance samples were collected to provide confidence in analytical results, identify potential sample contamination issues, and describe the magnitude of combined sample and analytical variability. Blank samples were used to determine the extent to which sampling or analytical methods may contaminate samples, which may bias analytical results. Replicate samples are used to determine the variability inherent in collection and analysis of environmental samples. Together, blank and replicate samples were used to characterize the accuracy and precision of water-quality data.

Where sufficient data were available, the ionic charge-balance error (CBE) was calculated to evaluate the electroneutrality of the water sample, identify transcription errors during field activities, and identify laboratory analytical errors. The CBE can be used to assess the accuracy and completeness of field and laboratory results for constituents contributing to sample ionic charge, which typically include the major ions (Freeze and Cherry, 1979). The CBE is calculated by the following formula:

The CBE has a positive value when the sum of cations exceeds the sum of anions but a negative value when the sum of anions is greater than the sum of cations. Calculated CBEs are rarely zero and values as much as 10 percent are typically considered acceptable; however, CBEs may exceed this threshold for waters with low-ionic strength (specific conductance <100 µS/cm) or in acidic waters (Fritz, 1994). Analysis of quality-assurance data showed that no sites had a CBE greater than 10 percent for this study and concentrations of replicate sample pairs differed by small amounts, typically less than 15 percent of the relative percent difference (RPD). Constituents with higher RPD were usually constituents at low concentration at or below the method detection limit. Analytical or sampling variability was considered minimal as a result.

A combination of equipment blanks and field blanks was used to identify and quantify potential sources of contamination. An equipment blank consists of a volume of water of known quality that is processed through the sampling equipment in a laboratory environment. A field blank consists of a volume of water processed through the sampling equipment under the same field conditions in which the samples were processed. Equipment blanks were run through two sets of sampling equipment prior to environmental sampling. One equipment blank had detections above the laboratory reporting level for cobalt, copper, and zinc. That set of sampling equipment was discarded and not used for environmental samples.

Variability for a replicate sample pair was quantified by calculating the RPD of the samples. The RPD was calculated using the following formula:

Statistical Analysis

Both univariate and multivariate statistical methods were used to ascertain significant constituent variations and distributions impacting water quality and water-quality dispersal throughout the study area. Statistical analyses were computed, and graphics created, using the R statistical computing environment, version 3.5.3 (R Core Team, 2022). Nonparametric techniques were used for computing descriptive and multivariate statistics from water-quality data that were censored at multiple levels. Censored values are water-quality results that are reported as less than a laboratory reporting level.

Prior to multivariate and summary statistical analyses, censored data were recoded 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. When 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).

Geochemical Modeling and Interpretation

The surface-water influence on groundwater quality was evaluated through four methods of geochemical modeling and interpretation. Stable isotopes and groundwater age analysis was used to determine recharge sources and assess the aquifer’s susceptibility to contamination from surface-water sources. Binary mixing models and geochemical inverse models were used to delineate the proportion of Ohio River water entering SWIG wells in the alluvial aquifer.

Statistical Summary of Field Parameters and Analytical Results

The sites sampled for this study represent raw-water supplies. Many of these sites have additional treatment after the point sampled, so the statistical summary of results presented here characterizes source water that may not be representative of supplied drinking water. Nevertheless, these data were compared to human-health benchmarks established by EPA (Environmental Protection Agency, 2018) to describe source-water quality relative to drinking-water standards. The EPA’s regulatory primary standards are established to protect human health, are mandatory for public supplies, and define the maximum-contaminant levels (MCL) or highest allowable concentrations in drinking water. Other non-regulatory EPA drinking-water guidelines used to assess this dataset include health advisories (HA) and secondary maximum-contaminant levels (SMCL). Health advisories, which are non-enforceable, provide technical information to state agencies and other public health officials on potential health effects for selected constituents that have no MCL or, in addition to the MCL. Secondary maximum-contaminant levels are listed for selected constituents that pose no known health risk but may have adverse aesthetic effects, such as staining or undesirable taste or odor.

Geochemistry of the Ohio River Alluvial Aquifer

The generalized conceptual model of flow in the Ohio River alluvial aquifer (fig. 2) assumes that recharge to the aquifer is primarily from meteoric sources. Specifically, precipitation falling on the alluvium percolates through the unsaturated zone into saturated sands and gravels, and water infiltrates from the Ohio River in areas of high pumping or during periods of high river stage (Jeffords, 1945; Mathes and others, 1997; Maharjan and Donovan, 2017).

Assessment of major ion chemistry with trilinear diagrams (fig. 3) supports this assumption of two major recharge sources because surface water and groundwater distant from the river have different chemical compositions; samples collected from surface-water sources appear to have a sulfate plus chloride dominated anion abundance, whereas the percentage of anions for wells located greater than 1,000 feet from the river has more carbonate plus bicarbonate. Differentiation of these two groups, with most intermediate wells located between these analyte abundances on trilinear diagrams, suggests that the conceptual model of two end members for recharge sources is representative of the system.

Hydrogeochemical processes controlling solute concentrations in the Ohio River alluvial aquifer were evaluated with multivariate statistical analysis of the available chemical data for the 23 wells sampled using PCA and graphical analysis. The results of the PCA (table 10) show 3 major hydrogeochemical processes—redox, salinity, and carbonate dissolution—predominantly control the geochemical system, with 73 percent of the variance explained by the first 3 components of the PCA. Twelve variables were represented in the PCA by loadings, which correlate individual variables to specific principal components. Positive loadings indicate that as the value of one constituent increases, the value of the correlated constituent also increases; whereas, negative loadings indicate that as the value of one constituent increases, the value of the correlated constituent decreases.

Sources of Salinity

The second principal component (PC2, table 10) has significant (p< 0.01) positive loadings for sodium, chloride, and potassium, which are constituents commonly associated with sources of salinity and may be related to land-use or waste-disposal practices. Sources of salinity and associated constituents (sodium, chloride, potassium, bromide) have been identified using chloride to bromide ratios by several authors (Davis and others, 2005; Katz and others, 2011). Chloride to bromide mass ratios calculated with the data collected for this study ranged from 144 to 1,731 (fig. 4). Chloride to bromide mass ratios with values in this range and measured chloride concentrations from approximately 20 to 120 mg/L in samples from the 23 wells (table 2) are typical of animal waste or sewage (300–1,000 mg/L) and halite dissolution (1,000–10,000 mg/L from natural and anthropogenic salt sources), which are the probable sodium, potassium, chloride, and bromide sources to groundwater in this area (Davis and others, 2005; Mullaney and others, 2009; Katz and others, 2011).

Magnesium commonly has positive correlation with calcium and significant loading on the principal component that explains carbonate dissolution when magnesium is derived from calcium-carbonate solid-phase sources, but it has a significant (p<0.01) positive loading on PC2 and significant correlation with the constituents responsible for salinity. This positive loading indicates that the main source of magnesium may be from wastewater. Another possible explanation for this significant loading is that additional sodium added to the aquifer through different salinity sources may promote cation exchange of sodium with calcium or magnesium, thus magnesium concentrations may increase more through cation exchange processes rather than dissolution of magnesium-rich carbonates, such as dolomite. Calcium plus magnesium to bicarbonate molar ratios are >1 (fig. 5), indicating that there may be an abundance of magnesium or calcium that cannot be explained by carbonate dissolution alone. Sodium and chloride molar ratios plot close to 1 (fig. 5) indicating that cation exchange may not be an important process controlling magnesium concentrations and the origin of magnesium could be from several different wastewater sources.

Carbonate Dissolution

The third principal component (PC3, table 10) has significant (p<0.01) positive loadings for bicarbonate and calcium, which is indicative of carbonate dissolution. Significant (p<0.01) positive loading for the distance of a well from the Ohio River indicates that calcium and bicarbonate concentrations increase through carbonate dissolution in distal wells. This increase may be indicative of recharge from precipitation in distal wells that is more chemically aggressive than river water and capable of higher rates of carbonate mineral dissolution. The molar ratio of calcium to magnesium shows a much higher mass of calcium (fig. 6), which implies that calcite is the main carbonate mineral involved in carbonate mineral dissolution processes throughout the aquifer. The molar ratio of calcium to bicarbonate confirms this observation, that these two constituents follow a linear one-to-one relationship (fig. 7), and supports the previous assumption that magnesium may be supplied to the aquifer by originating from sources other than calcium, such as weathering magnesium-bearing minerals or through wastewater rather than carbonate dissolution.

Nitrate, Pesticides, Volatile Organic Compounds, ³H, and Dissolved Organic Carbon

Kozar and Paybins (2016) identified bacteria, nitrate, pesticides, volatile organic compounds, and chlorofluorocarbons as indicators of potential surface-water influence on, and vulnerability to contamination from surface or near-surface sources in, groundwater. Chlorofluorocarbons were not collected for this study but ³H was used in its place to indicate infiltration of modern water and assess aquifer vulnerability to surface contamination. Additionally, DOC may indicate strong connections among groundwater and near-surface sources or surface water (Shen and others, 2015). The presence of nitrate and DOC measured in concentrations above reporting levels in all 23 groundwater samples and detections of one or more man-made organic compounds in 22 groundwater samples (tables 8, 9, and 11; McAdoo, Grindle, and Grindle (2022) indicates that the alluvial aquifer is potentially contaminated or recharged by surface or near-surface water. Although no groundwater samples had detections for any of the five microbial constituents analyzed for this study, Kozar and Paybins (2016) stated that detections of bacteria in groundwater from alluvial aquifers in West Virginia may be infrequent because of alluvial sediments acting as a large sand filter that naturally retards the movement of bacteria. However, concentrations of nitrate and DOC and detections of ³H and man-made organic compounds are consistent with potential surface-water or surface-sources’ influence on groundwater. Samples from 19 wells had nitrate concentrations >1 mg/L as N, samples from 9 wells had DOC concentrations >1 mg/L, samples from 18 wells had detectable ³H, samples from 17 wells had detections for VOCs, and samples from 11 wells had detections for pesticides (table 11). Every well sampled for this study in the Ohio River alluvial aquifer had detections for at least one of these indicators of surface-water influence.

Stable Isotope Analysis

Stable isotopes of water (δ²H and δ¹⁸O) were evaluated to determine variation in recharge sources throughout the study area by comparing measured isotopic concentrations at sampling sites (wells and river locations) to the isotopic composition reported in precipitation. Values of δ¹⁸O ranged from −8.33 to −5.63 parts per thousand (permil) and values of δ²H ranged from −54.5 to −40.0 permil (table 12). The LMWL published by Smith and others (2021; δ²H=7.58 x δ¹⁸O+9.16) was used to represent precipitation in the Ohio River Valley. This line deviates from the global meteoric water line (δ²H=8 x δ¹⁸O+10; Craig, 1961) and indicates enrichment of isotopes in precipitation at the local scale (fig. 8). Stable isotope data collected for this study generally follow the LMWL, indicating that recharge to the alluvial aquifer is from a meteoric source of precipitation, with the exception of one site (Mas-0918). River water samples were collected in November 2019 and the stable isotopic composition of these samples (fig. 8) may partly reflect seasonal conditions, with lighter values in cooler months.

Values for line-conditioned (LC) excess are all generally below the one standard deviation measurement uncertainty (S=1.13) except for Ohio River at Moundsville (site name shortened to OR Moundsville in this report) with a LC-excess of −1.51, and Mas-0918 with a LC excess of −6.48. Negative values of LC-excess indicate evaporative enrichment after precipitation. The reason for elevated LC-excess at OR Moundsville is unknown, but the magnitude of the departure from the LMWL is relatively small in comparison to Mas-0918. This site is within 1,000 feet of wetlands and ponds that may experience evaporation, which subsequently enriches the isotopic concentrations in groundwater recharge in that area.

Carbon Isotope Analysis

Surface-water samples had δ¹³C values between −11.4 and −10.81 permil. Groundwater samples generally had values of δ¹³C that were less than −13.57 permil except for Mas-0918, which had a δ¹³C value of −11.91 permil. Mas-0918 was identified as having an LC-excess indicative of an evaporative source, likely from wetlands in the immediate area, and a bicarbonate/sulfate signature that deviated from other intermediate well samples. This high δ¹³C value may further indicate that the groundwater entering Mas-0918 may be heavily influenced by the proximal wetlands source.

Groundwater samples generally plot to the left of the zero-age line for the open system RFG analytical model (fig. 9; Han and Plummer, 2013). The model was parameterized with values for δ¹³C soil gas and solid carbonates of −25 permil and 2 permil, respectively, which are consistent with typical soil-zone gases and marine carbonate isotopic compositions (Clark and Fritz, 1997). The position of some samples on the figure (close to the zero-age line) indicates that the dominant controls of DIC in the study area are carbonate dissolution and open system exchange with soil CO₂, but several samples do not follow the zero-age line, which may indicate that other processes affect carbon isotope chemistry. The more fractionated δ¹³C signal of these samples is possibly driven by organic carbon oxidation by microbial processes indicated by samples with low dissolved oxygen or high concentrations of iron or manganese. Without appropriate correction for microbial processes and carbon sources, ¹⁴C has limited utility for quantification of groundwater age in samples in which controls on DIC are not accounted for.

Where possible, the final adjusted ¹⁴C (table 12) for these samples was computed, resulting in a range from 100 to 133 percent modern carbon (pmC). The corrected ¹⁴C values are generally consistent with other tracers and indicate that recent recharge along local flow paths is captured at these sites. Carbon-14 has limited utility for quantitative age dating of modern samples because of (1) elevated atmospheric ¹⁴C (for example, >100 pmC) from above-ground nuclear testing since late 1950s and (2) high sensitivity of modern interpreted ages to the geochemical correction. High pmC values of ¹⁴C and the presence of ³H at every site indicates no, or very little, pre-1950s water in any sample collected for this study.

Groundwater Age Tracer Analysis

Measured concentrations of noble gasses were generally within 20 percent of atmospheric solubility equilibrium with Ne, Ar, Kr, and Xe exhibiting average deviations from solubility equilibrium of 20.9, 14.1, 9.2, and 6.1 percent, respectively (table 13). Dissolved noble gas data were well-fit by solubility models with statistically significant model solutions (χ² values less than critical value of 3.84; one-sided, one degree of freedom, 95 percent confidence). One sample (Ohi-0372) was determined to have re-equilibrated with the atmosphere because of sample container malfunction and was not used for further dissolved gas analysis. Noble gas solubility models provided a good estimate of the groundwater recharge temperature (referred to as noble gas recharge temperature; NGT) and the amount of EA for each sample. Noble gas temperatures generally group around the mean annual air temperature modeled for the study area (PRISM Climate Group, 2022). Modeled NGTs ranged from 11.0 to 13.7 °C with two outlier values of 15.1 °C and 16.9 °C at Woo-0121and Mas-0936, respectively.

The He isotopic ratio (R/R_a) indicates a possible contribution from a premodern groundwater source, with a high proportion of terrigenic helium (⁴He_terr) observed in every sample except for Mas-0918. Four samples with R/R_a values between 0.81 and 0.73 were identified as having some ⁴He_terr, but 12 samples with R/R_a values less than 0.70 were identified as having high ⁴He_terr. The only sample in the dataset to have a R/R_a value close to 1 (indicating solubility equilibrium with the atmosphere) was Mas-0918, which was identified by stable isotope analysis as possibly being influenced by a wetland or pond source rather than representing the groundwater conditions in the rest of the aquifer. All samples in the dataset contain tritium (table 12) and are likely modern, which limits the ability to estimate the ⁴He_terr isotopic ratio. Therefore, ³He_trit could not be reliably estimated for most samples in the dataset, which reduced the number of available tracers for age determination. The source of high ⁴He_terr in almost every sample collected for the study is not currently [2019] known but other investigators have also identified high ⁴He_terr accumulation rates in other glacially deposited shallow alluvial aquifers (Solomon and others, 1996).

Measured SF₆ concentrations were corrected for excess air and found to exceed expected concentrations from atmospheric inputs. The contamination source of excess SF₆ observed in the aquifer is not known but high concentrations of SF₆ preclude its use for groundwater age analysis. With area-wide contamination of SF₆, ¹⁴C concentrations that indicated predominantly modern recharge, and high ⁴He_terr observed in most samples that limited the use of ³He_trit, ³H was the only modern tracer left to estimate groundwater age. At six sites, ³H was measured above background levels (approx. 10 TU), which may be indicative of an additional non-atmospheric local source of ³H. Bomb-pulse elevated atmospheric sources of ³H can be distinguished from local ³H sources based on ¹⁴C, which increased in the atmosphere over the same period in response to above-ground nuclear testing. At five of the six samples with elevated ³H concentrations (table 12), it was possible to correct ¹⁴C using the RFG model. At all five sites, elevated ³H concentrations did not correspond to elevated ¹⁴C concentrations, indicating a local source of ³H may be present.

Assessment of non-atmospheric sources of ³H yielded possible contribution from a nuclear power plant on the Ohio River, upstream of every site in the study area. This facility regularly releases ³H into the Ohio River at concentrations that are higher than atmospheric background levels (Paciello and others, 2019). Helium-3 was not sampled in the Ohio River for this study but 35 ³H samples were collected in the Ohio River at a surface-water site near Newell, West Virginia (USGS 03109670) between April 17,  2002, and August 2,  2010 (U.S. Geological Survey, 2023; https://waterdata.usgs.gov/nwis/inventory?agency_code=USGS&site_no=03109670). This surface-water site is adjacent to Hnc-0043, and samples collected there had an average value for ³H of 19 TU. This ³H value supports the possibility of elevated ³H in the Ohio River but without the ability to corroborate age dates across multiple tracers and the possibility of a non-atmospheric source of ³H affecting concentrations in the aquifer, obtaining accurate estimates of groundwater age in the aquifer is not probable with these data. What can be determined from the age analysis for this study is that all water captured by wells in the Ohio River alluvial aquifer is likely from a relatively modern source.

Statistical Relations Between Ohio River Water and Groundwater Chemistry

The two major sources of recharge to the Ohio River alluvial aquifer in West Virginia are local precipitation and infiltration from the Ohio River (Jeffords, 1945). Hierarchical agglomerative cluster analysis (HACA) was used to confirm and further characterize these sources. The primary output from HACA is a dendrogram depicting the clustering structure of the input data (fig. 10). Individual samples are represented by vertical lines at the bottom of the figure, and the merging of similar clusters are represented by horizontal lines connecting clusters. Clusters that merge toward the bottom of the dendrogram have similar water chemistry. Clusters that do not merge until the top of the dendrogram represent groups of samples that have less similar water chemistries. Specific analytes and parameters used in the HACA for this study included δ²H, δ¹⁸O, δ ¹³C, pH, specific conductance, dissolved oxygen, DOC, calcium, magnesium, potassium, sodium, chloride, fluoride, bicarbonate, silica, sulfate, iron, and manganese. This resulted in a HACA with 4 surface-water sites and 23 groundwater wells (table 1) that included field parameters, major ions, trace metals, and isotopes important in end-member characterization. Constituent values were converted to u-scores, and the distance matrix was computed as described in the “Methods of Study” section.

The HACA resulted in 3 clusters represented by capital letters (A, B, C). The agglomerative coefficient, a numerical value used to assess the structure of the clustering, ranges from 0 to 1, with values closer to 1 representing better cluster structure (Kaufman and Rousseeuw, 1990). The agglomerative coefficient produced was 0.77, which provides confidence in the HACA structure. The SIMPROF test was applied to the cluster analysis and identified 11 significant clusters (p<0.01), of which 3 clusters (A, B, C) were used for further analysis.

Cluster B includes all four surface-water samples and represents the surface-water end member of the system. Clusters C and B merge before coming together with cluster A at the final merge point in the HACA. This merge point indicates that sites in cluster C are more closely associated with surface-water sites in cluster B than other groundwater sites in cluster A. Cluster C includes 10 wells that have an average distance of 217 ft from the river and are representative of sites that may be highly influenced by Ohio River water chemistry. Cluster A includes 13 wells that have an average distance of 906 ft from the river. This cluster includes wells greater than 1,000 ft from the river and wells that may have some influence from the Ohio River water chemistry but are probably more influenced by water from the recharge area of the aquifer, as depicted in the conceptual model (fig. 2). Results indicate that two hydrochemical facies are contributing water to the SWIG wells in the alluvium but the Ohio River is the most likely source contributing an increasing proportion of water to wells as distance to the river decreases.

Mixing models to estimate Ohio River Water Influence on Groundwater Chemistry

Jeffords (1945) states that analyses of major anions are adequate for use in determining the source of water recharging any well field in the Ohio River alluvium and that the greatest difference in Ohio River water and groundwater not influenced by the Ohio River is in the respective concentrations of sulfate and bicarbonate. Data collected for this study show a similar result, with anions that represent noncarbonate hardness (sulfate and chloride) comprising a higher proportion of total hardness in surface-water samples and anions that represent carbonate hardness (carbonate and bicarbonate) comprising a higher percentage of total hardness in groundwater samples greater than 1,000 ft from the river (fig. 3). Likewise, comparison of the equivalents of bicarbonate as a percentage of anions with the equivalents of sulfate as a percentage of anions (fig. 11) shows differentiation of three groups. The first group consists of surface-water samples collected in the Ohio River and is characterized by sulfate comprising over 35 percent of anion equivalents. The second group is comprised of samples greater than 1,000 ft from the Ohio River and is characterized by bicarbonate comprising over 70 percent of anion equivalents. The third group consists of samples with mixtures of Ohio River water and groundwater because they plot between the distal wells and surface water sites. Samples from three wells do not follow this general association between bicarbonate and sulfate, with the sample from Mas-0918 (444 ft from river) containing a relatively high percentage of bicarbonate compared to other wells at intermediate distances (400–650 ft) from the Ohio River, the sample from Ple-0071(1,473 ft from river) containing a low percentage of bicarbonate relative to other wells located greater than 1,000 ft from the Ohio River, and the sample Mas-0936 (650 ft from river) containing relatively more sulfate but less bicarbonate than samples from other wells at intermediate distances (400–650 ft) from the Ohio River, plotting in an area that is not indicative of surface water from the Ohio River or groundwater from more distal wells (fig. 11). These three sites may be indicative of the heterogenous nature of the Ohio River alluvial aquifer or may be supplied by an unknown source with an unidentified bicarbonate/sulfate signature.

Groundwater-flow simulations in the Ohio River alluvium by Kozar and McCoy (2004) estimated fractions of Ohio River water contribution to the groundwater system at Parkersburg, Lubeck, Glendale, and Point Pleasant, to be 0.75, 0.39, 0.72, and 0.04, respectively (table 14). The sites Parkersburg, Lubeck, Glendale, and Point Pleasant specified in Kozar and McCoy (2004) were represented in this study by Woo-0121, Woo-0215, Mal-0410, and Mas-0920, respectively. Binary-mixing models were computed with analytical data collected from this study and used as a first-order approximation for fractions of the two major recharge sources. Several analytes were considered for binary mixing, including δ²H, δ¹⁸O, and chloride, but the binary mixing model that had the most significant correlation (r=0.97, p<0.03) with groundwater modeling results was the ratio of bicarbonate equivalents as a percentage of anions to sulfate equivalents as a percentage of anions, which was shown to have a linear relation in figure 11. Silica (SiO₂) concentrations were found to have significant negative correlation (r=−0.91, p<0.08) with groundwater modeling results, showed similar behavior as bicarbonate (lower concentrations in Ohio River samples and higher concentrations in distal wells), and were also considered for binary mixing models (table 14). The Ohio River end member (table 14) was represented by the average of results of the four surface-water samples (table 1). The alluvial well end member was represented by the average of results for four wells greater than 1,000 ft from the Ohio River (Mas-0852, Jac-0128, Jac-0057, Woo-0196), with results from two wells excluded for the following reasons: Mas-0920 (1,475 ft from river) results were excluded because an estimate of the end member fractions for this site was needed to compare to the value reported in the groundwater-flow model, and Ple-0071 (1,473 ft from river) as this site was identified as possibly having a contribution from an unknown recharge source with a different bicarbonate/sulfate signature (fig. 11).

The estimated fractions of Ohio River water contribution to the groundwater system from the groundwater-flow model (Kozar and McCoy, 2004) were used to validate the results of the silica and equivalent-ratio binary mixing models. Fractions of river water in the aquifer computed using the silica binary-mixing model were significantly correlated with those determined from the groundwater-model for 4 sites (Woo-0121, Woo-0215, Mal-0410, and Mas-0920; r=0.91, p less than 0.09), with three sites showing 6 percent difference or less. The percentage of Ohio River water calculated by the silica binary mixing model for Woo-121 was 29 percent less than the percentage of Ohio River water calculated in the groundwater model results for Ohio River @ Mile 183.0 (site name shortened to OR Parkersburg in this report), which suggests that the silica binary-mixing model may not be accurate at all sites or that Woo-121 may not be representative of the Parkersburg area as simulated by the groundwater model. The percentage of Ohio River water calculated with the equivalent ratio (bicarbonate equivalent as percentage of anion equivalents to sulfate equivalent as percentage of anions equivalents) binary mixing model was significantly correlated with the percentage of Ohio River water estimated by the groundwater model at the four sites (r=0.97, p less than 0.03). Of the 4 sites, the percentage of Ohio River water for Woo-0121 was most different between these 2 models, with the equivalent ratio model results being 15 percent greater than the percentage of Ohio River water calculated in the groundwater model results for OR Parkersburg. The disparity among models in calculated results of the same site indicates that non-conservative mixing may be important at some sites where geochemical processes are not accounted for in binary mixing models.

Inverse geochemical models that met the specified uncertainty (10 percent or less) were computed for Mas-0920 and 13 of the groundwater sites located less than 1,000 ft from the river (table 14). The percentage of Ohio River water computed by inverse models was significantly correlated with the percentage of Ohio River water estimated by the groundwater model (r=0.98, p<0.02). The percentage of Ohio River water for Woo-0121, Woo-0215, and Mal-0410 calculated with the inverse model was within 8 percent of the results reported from the groundwater model. The percentage of Ohio River water for Mas-0920 was 16 percent greater than the results in the groundwater model for Point Pleasant, which indicates that the inverse model may be more accurate for sites less than 1,000 ft from the river. The average of all three geochemical models (silica binary mixing model, equivalent ratio binary mixing model, inverse model) yields the most significant correlation (r=0.99, p<0.006) with the results from the groundwater model. By using the average of all three models, all 4 sites had average results that are within 7 percent of the groundwater model results (table 14, fig. 12).

The Ohio River Alluvial Aquifer is Susceptible to Contamination from Surface Water

Meteoric recharge from the Ohio River and precipitation on the alluvium and subsequent recharge through the vadose zone are the main sources of water entering the aquifer, but isotope ratios and concentrations of major ions indicate other surface-water sources may affect groundwater chemistry at individual sites (Mas-0918, Ple-0071, Mas-0936) on a local scale (fig. 8, 11). Generally, groundwater from wells located over 1,000 ft from the river, which represented the chemistry of the precipitation dominated portion of the aquifer, had a chemical signature that was different than Ohio River water (fig. 8, 10, 11), with a higher percentage of bicarbonate relative to sulfate observed in wells located closer to the Ohio River. These results are consistent with Maharjan and Donovan (2017), who observed low total dissolved solids specific conductance, alkalinity, and enriched δ ¹³C values of dissolved inorganic carbon, indicative of induced infiltration in proximal (less than 1,000 ft from Ohio River) groundwater wells. Also consistent with Maharjan and Donovan (2017) was the observation that distal (more than 1,000 ft from Ohio River) alluvial well water chemistry exhibited higher total dissolved solids, specific conductance, alkalinity, and relatively depleted δ¹³C values and was not affected by exfiltration from the Ohio River.

Every well sampled for this study (distal and proximal) is intrinsically susceptible and potentially vulnerable to surface contamination. Kozar and Paybins (2016) identify bacteria, nitrate, pesticides, VOCs, and chlorofluorocarbons as indicators of surface-water influence in groundwater wells. No microbial fecal indicators were found in any groundwater sample from wells throughout the study area, indicating these sites are not likely GWUDI, but all sites sampled for this study had detections for at least one water-quality indicator of surface-water influence (table 11). These observations suggest that one tracer or parameter is not sufficient to assess surface-water-influence in the Ohio River alluvium, but when multiple constituents are included in the analysis, it is apparent that every site in the Ohio River alluvial aquifer is influenced by surface water to some extent and vulnerable to surface contamination.

Ohio River Water Influences Groundwater Chemistry in the Alluvial Aquifer

Jeffords (1945) stated that the approximate amount of river water recharging groundwater wells in the Ohio River alluvium can be determined by obtaining the percentage of bicarbonate and sulfate and comparing these with similar data on water from the Ohio River and from wells that obtain little or no recharge from the river. Likewise, results from binary mixing models and inverse models created for this study suggest that sulfate, silica, or bicarbonate concentrations adequately predict the fraction of Ohio River water entering alluvial wells for preliminary investigations of surface-water influence. These three constituents are commonly measured in water-quality analyses for regulatory purposes and are more readily available for preliminary assessment of surface-water influence than many other analytes, such as isotopes. Binary mixing models do not consider time variability of end-member signatures or the reactive nature of sulfate, bicarbonate, and silica. Nevertheless, results from binary mixing models showed significant correlation with the inverse models computed for this study and groundwater models published by Kozar and McCoy (2004). Using the average of the fraction of Ohio River water computed from the three geochemically based models (silica binary mixing model, equivalent ratio binary mixing model, and inverses models) yielded the highest correlation with fractions estimated using the groundwater-flow model (table 14).

In the absence of extensive analytes and geochemical or groundwater-flow modeling capabilities, preliminary assessment of the fraction of Ohio River water entering groundwater wells in the Ohio River alluvium may be estimated for most sites (including sites not specified in this study) using the linear relation between the equivalent ratio of bicarbonate to sulfate and the fraction of water computed by the average of the three geochemical models (fig. 13A, table 14). Additionally, the linear relationship of sulfate concentration with the proportion of Ohio River water computed from the average of three geochemical models (fig. 13B) or the linear relationship of silica concentration with the proportion of Ohio River water computed from the average of three geochemical models (fig. 13C) may be used to provide a preliminary assessment of the potential influence from surface water entering wells in the Ohio River alluvium, although this approach might not be sufficient for samples with high concentrations of sulfate or silica without equivalent-weight normalization. For samples with high concentrations, the equivalent ratio of bicarbonate to sulfate may yield better results for outlier sites because it normalizes the concentrations by the molecular weight and anion percentage. Although first-order calculations of the fraction of Ohio River water presented in this study may be sufficient for a broad understanding of the Ohio River alluvial aquifer or a preliminary assessment of an individual groundwater site, focused reaction-transport models or other types of biological or geochemical data could enhance understanding in localized areas.

This study indicates that the chemistry of groundwater wells in the Ohio River alluvium is influenced by recharge from the Ohio River but distance from the Ohio River was not always a good indication of the amount a SWIG site will be influenced by Ohio River water chemistry. The data and related analyses indicate that Ohio River water chemistry has more influence at some sites over 400 ft from the Ohio River (for example Mal-0410) than at some sites located less than 100 ft from the Ohio River (for example Mas-0934) and that distance from the Ohio River was not significantly correlated with the computed average fraction of Ohio River water, although models indicated the smallest fraction of river water at the well most distant from the river Mas-0920 (table 14). This general finding of inconsistent relations between distance from the river and estimated fraction of river water in groundwater indicates that influence from the Ohio River on alluvial well water chemistry may be affected by local-scale heterogeneity in alluvial sediments (permeability and mineralogy), well construction, or transient pumping by the local public water system in addition to river proximity. Pumping can induce infiltration from the river, and the analysis presented in this study could be further refined if pumping were included as a factor.