1 Introduction

The combination of increased salinization (Kaushal et al., 2005) and increased alkalinization (Kaushal et al., 2013) in freshwater rivers, lakes, ponds, and wetlands has been termed the “Freshwater Salinization Syndrome” (FSS), with salinization and alkalinization affecting 37% and 90%, respectively, of the drainage area of the contiguous United States over the last half century (Kaushal et al., 2017b). FSS has detrimental effects on biota, human health, and human infrastructure as fundamental changes in water chemistry can have a myriad of direct and indirect ecological and biogeochemical consequences (Kaushal et al., 2017a, 2021). In particular, increased salinity has species-, community-, and ecosystem-level effects as a result of disruptions to niche space and resource utilization at smaller functional levels that can scale up (Hintz & Relyea, 2019). Specific to impacts on humans, FSS has possible health impacts due to increases in chloride in tap water (Leslie & Lyons, 2018).

The recent salinization of rivers has been ascribed to increasing chloride levels due mainly to road salt application (Hintz et al., 2021), treated human waste (Kelly et al., 2010), animal manure (Overbo et al., 2021), and inorganic fertilizers (mainly KCl; Oberhelman & Peterson, 2020). Other sources of chloride can include household water softeners, household cleaning products, atmospheric deposition, dust suppressants, and industrial discharge (Overbo et al., 2021). Recent studies have shown that most of the large increases in salinity in urban rivers are due to road salt application as a deicer and its subsequent transport to the aquatic environment (Corsi et al., 2015). However, even in rural areas, road salt application can affect the salinity of streams (Price & Szymanski, 2014; Soper et al., 2021). Rural areas, especially those dominated by agriculture, can be additionally affected though as other sources of chloride, such as inorganic fertilizers, manure, and septic or wastewater are often present (Falcone et al., 2018).

Several Ohio water bodies have been studied for chloride dynamics in both the Ohio River (Dailey et al., 2014; Leslie & Lyons, 2018) and Lake Erie (Kane et al., 2022) watersheds. These studies have documented increases over time in chloride in rivers and reservoirs in the region, where road salt has been implicated as the main driver of these increases—although other sources listed above, including applications of potash (KCl) concurrent with row-crop agriculture (David et al., 2015) and manure (Overbo et al., 2021), have also contributed. Kane et al. (2022) found that the greatest values and largest increases in chloride watershed yield (kg/ha) in the twenty-first century for the 20 + Ohio watersheds studied were in the urban Cuyahoga River watershed. Interestingly, rural Coldwater Creek (a tributary to Grand Lake St. Marys, the largest inland Ohio reservoir) watershed was second on the Kane et al. (2022) list, warranting additional study. The Coldwater Creek subwatershed is highly agricultural, with areas comprised of almost 70% row crop agriculture and 14% pasture (NRCS Geospatial Data Gateway, 2006). In addition to crop production, associated livestock operations are present and are estimated to produce 66,104 tonnes/year of manure by mostly swine and poultry (GLWWA- Grand Lake/Wabash Watershed Alliance, 2008). Animal manure, particularly that of swine and poultry, can be a significant source of chloride (Müller & Gächter, 2011). Li-Xian et al. (2007) found that chloride accounted for 17.3% of the total soluble salts in poultry manure, more than double the 7% found in pig manure. In addition, potash (KCl) application is relatively common in the subwatershed which can also be a source of chloride in agricultural watersheds (Lax et al., 2017). Spatially, approximately 66% of agricultural operations occur within 300 m of the creek (Filbrun et al., 2013). Finally, human waste can further contribute as a source of chloride in the subwatershed as the St. Henry Wastewater Treatment Plant (WWTP) (secondary treatment level-https://dam.assets.ohio.gov/image/upload/epa.ohio.gov/Portals/35/permits/doc/2PB00027.pdf) discharges into Coldwater Creek through treatment lagoons (technically limited to discharging when streamflow is > 1 cfs; Ohio EPA, 2007).

This variety of potential contributors to the salinization of Coldwater Creek highlights the need for a better understanding of the sources of chloride pollution in this system, particularly when it comes to remediation efforts. Oberhelman and Peterson (2020) found that [Cl−]:[Na +], [Cl−]:[Br−], and [Cl−]:[NO3-N] ratios were the most useful ion ratios for identifying Cl− sources in the Evergreen Lake watershed in rural Illinois. Ratios indicated that road salt was the dominant source of Cl, followed by KCl fertilizer (Oberhelman & Peterson, 2020). A similar approach could be applied to Grand Lake St. Marys – however, as watersheds are unique it is unclear whether a single approach could be applied successfully.

Herein, we examine long term trends in chloride concentrations in the Grand Lake St. Mary’s watershed using two tributaries. The goals of our study were to 1) determine mean chloride concentrations apportioned to season in Coldwater Creek and Chickasaw Creek (two of the primary tributaries to GLSM) over the past decade as well as to 2) use chloride and nitrate concentration data to delineate the source of chloride in each of these subwatersheds.

2 Materials and Methods

2.1 Study Site

Grand Lake St. Marys is located in Mercer and Auglaize counties in northwestern Ohio (Fig. 1) and is the largest reservoir in Ohio (mean surface area: 52 km2; mean volume: 8.25 × 107 m3; mean residence time: 236 d; Filbrun et al., 2013). The watershed has been designated by the state of Ohio as a “distressed” watershed due to nonpoint source loading of excess nutrients from the watershed causing cyanobacterial Harmful Algal Blooms (cyanoHABs) in the lake proper (Jacquemin et al., 2018). The eutrophication of Grand Lake St. Marys has negatively affected ecosystem services that the lake provides, including causing “no contact” advisories yearly (Jacquemin et al., 2018) for more than a decade by the Ohio Department of Health. In fact, in most years during spring, summer, and fall, the total microcystin values exceed the World Health Organization (WHO) contact advisory limit of 20 µg/L (Jacquemin et al., 2023), and thus also exceed the USEPA Recommended Human Health Recreational Ambient Water Quality Criterion of 8 µg/L (U.S. EPA, 2019). CyanoHABS in Grand Lake St. Marys are historically among the worst in the United States, with both frequent cyanobacterial blooms (83% of time exceeding the WHO 100,000 cells/mL high risk threshold; Clark et al., 2017) and high microcystin levels (3rd largest total microcystin value (78 µg/L) of 1,161 lakes sampled in the U.S. during May through October 2007; Loftin et al., 2016).

Fig. 1
figure 1

Grand Lake St. Marys (Ohio) and its watershed. Coldwater Creek and Chickasaw Creek subwatersheds are shown, along with their respective primary streams. Star symbols indicate locations of Heidelberg Tributary Loading Program sampling sites. Each site is co-located with a USGS stream gage (Kane et al., 2023). The circle symbol indicates the location of St. Henry Wastewater Treatment Plant. Inset indicates Grand Lake St. Marys location in Ohio

Full size image

Coldwater Creek and Chickasaw Creek are tributaries to Grand Lake St. Marys and enter the lake from the south (Fig. 1). The land use in the Coldwater Creek subwatershed is 69% row-crop agriculture, 14% pasture, 3% forest, 12% urban, and 2% other, while that in Chickasaw Creek subwatershed is 79% row-crop agriculture, 9% pasture, 3% forest, 9% urban, and < 1% other. Land use for each subwatershed was obtained from the NRCS Geospatial Data Gateway (NRCS Geospatial Data Gateway, 2006). A notable difference in land use in the two subwatersheds is that the Coldwater Creek subwatershed contains all of the village of St. Henry (2020 Population- 2,596) and part of the village of Coldwater (2020 Population- 4,774), while Chickasaw Creek contains only the smaller village of Chickasaw (2020 Population- 358) (Ohio Secretary of State, 2021).

2.2 Data Acquisition

2.2.1 Sampling Methods

The Heidelberg Tributary Loading Program (HTLP) currently monitors 22 rivers in Ohio and one in Michigan (River Raisin) that are in the Lake Erie and Ohio River watersheds. As the HTLP has been detailed elsewhere (Baker et al., 2014; Choquette et al., 2019; NCWQR, 2022), as have methods regarding chloride calculations (Kane et al., 2023), only brief synopses of relevant methods follow. The two tributaries to Grand Lake St. Marys sampled for this study have been sampled since the mid-2000s, Chickasaw Creek (2009) and Coldwater Creek (2013), and the data reported here span through 2021.

In the field, the Coldwater Creek and Chickasaw Creek sampling stations contain a submersible pump that continuously pumps water into heated buildings from the river. Inside these buildings, automatic samplers collect discrete samples three times daily (3 samples/d at 8-h intervals (0400 h, 1200 h, 2000 h)) and are refrigerated until retrieved. Each week, the samples are returned to the NCWQR laboratories for analysis. When samples either have high turbidity or collected during high flow conditions, all samples for each day are analyzed. As streamflow and/or turbidity decrease, the analysis frequency shifts to one sample per day (Baker et al., 2014). Each of these stations is paired with a USGS stream gage (Kane et al., 2023) to facilitate loading and other calculations. More details on the sampling program, including a project study plan and quality assurance plan, are available at https://ncwqr.org/monitoring/.

2.2.2 Laboratory Methods

Chloride and nitrate–N and nitrite-N (NO3 + 2) (hereafter referred to as nitrate) concentrations were determined via ion chromatography using a Dionex DX320, Dionex ICS2000, or Dionex ICS2100 (Roerdink et al., 2017). NCWQR samples for chloride were always between the highest and lowest acceptable limits in QA/QC studies using certified reference materials performed during 2011–2022 (Baker, personal communication).

2.3 Data Analysis

2.3.1 Load Calculations

Annual chloride loads were calculated from the concentrations of chloride and mean daily flow from the USGS using the loadflex package in R (Appling et al., 2015) to fill in for days when samples were not collected (< 5% of the time).

To separate patterns into seasons, we calculated daily chloride loads following Richards et al. (2010), and any missing days were interpolated from previous days. Daily mean concentrations were calculated as the daily load divided by daily flow. The seasonal mean daily concentrations were then assessed as the mean of the daily mean concentrations apportioned to season, where fall is September, October, and November; winter is December, January, and February; spring is March, April, and May; and summer is June, July, and August.

2.3.2 Statistical Methods

To assess the influence of land use and season on chloride concentrations across subwatersheds, we used a combination of simple linear regression and analysis of variance (ANOVA). Simple linear regressions were used to test whether mean daily chloride concentrations had changed over time within specific subwatersheds. A three-way ANOVA was used to test the effects of subwatershed, year, and season on chloride concentrations. Data (or the model residuals) were tested for normality and equal variance to satisfy the assumptions of all parametric tests with resulting α-values set at 0.05 to evaluate statistical significance. Minitab 21 (Minitab, 2021) was used to perform linear regressions and three-way ANOVA (subsequent Tukey tests were used to evaluate intergroup differences).

To examine the long-term trends in chloride concentrations and loads, we used Weighted Regressions on Time, Discharge, and Season (WRTDS) (Hirsch et al., 2010) with the EGRET package (Hirsch & De Cicco, 2015) in R (R Core Development Team 2008) to generate daily Flow Normalized Concentration (FNC) values and estimate changes in the relationship between discharge and chloride concentrations over time (Hirsch et al., 2010). WRTDS uses long-term trends, seasonal variation and discharge values to generate estimates of daily concentration over long periods of time (Eq. 1).

$$\text{ln}\left(\text{c}\right)={\beta }_{0}+{\beta }_{1}t+{\beta }_{2}\text{ln}\left(Q\right)+{\beta }_{3}\text{sin}\left(2\pi t\right)+{\beta }_{4}\text{cos}\left(2\pi t\right)+\varepsilon$$
(1)

where c is concentration, Q is discharge, t is time in years, β3 and β4 capture seasonal periodicity, and ε is unexplained variance.

WRTDS has been employed in a wide range of aquatic systems, including Chesapeake Bay, the Mississippi River Basin, and Great Lakes Tributaries (Rowland et al., 2021), and has been employed to discern long-term trends in HTLP nutrient (Choquette et al., 2019) and Lake Erie tributary chloride (Kane et al., 2023) data. Thus, we used WRTDS to visualize long-term chloride trends in Coldwater Creek and Chickasaw Creek.

Daily flow data, which are required for WRTDS, were downloaded from the USGS NWIS data repository (https://waterdata.usgs.gov/nwis, accessed 05/2023) and paired with daily water quality data from the NCWQR. While WRTDS can produce reliable estimates with as few as 60 observations over a decade, it was designed for larger datasets (> 200 concentration samples) (Rowland et al., 2021).

Daily concentration samples from the NCWQR for Coldwater Creek (2013–2021) and Chickasaw Creek (2010–2021) were used to generate the FNC values. Changes in concentration patterns were then compared within each subwatershed across time, as well as between the two watersheds. Diagnostic multiplots from the WRTDS for Coldwater Creek and Chickasaw Creek were also produced (Supplemental Figs. 1–2). These show the (a) discharge/concentration relationship, (b) estimated daily concentrations across the period of record (POR), (c) box plot showing the monthly variation in chloride concentrations, and (d) comparison of the distribution of values sampled for the model versus the entire POR.

2.3.3 Geospatial Methods

GeoTiff files of road salt application for the United States for the years 1992–2015 for Coldwater Creek and Chickasaw Creek were downloaded from the USGS science base catalog (sciencebase.gov) (Bock et al., 2018). GeoTiffs were converted into a raster stack in R using the “raster” library. The raster stack was then clipped to shape files of the Coldwater Creek and Chickasaw Creek subwatersheds. Individual cell values from the clipped raster stack were then extracted, summed by year and converted to metric tons per year by subwatershed. Linear regressions were then performed to 1) determine temporal trends in yearly road salt application in each subwatershed and 2) determine if there was a significant relationship between yearly road salt application and each of the seasonal mean concentrations for chloride for each of the two subwatersheds. Due to a shorter temporal frame for river sampling, road salt vs. chloride seasonal mean regressions were performed only for 2013–2015 for Coldwater Creek and 2010–2015 for Chickasaw Creek. To compare vehicular traffic on SR 219 (a road that crosses both creeks) the ODOT (Ohio Department of Transportation) TMMS (Traffic Monitoring Management System) (https://odot.ms2soft.com/TDMS.UI_Core/portal) was used to determine mean ± standard error AADT (Annual Average Daily Traffic) (2013–2015) at Main St. SR219 W of T101 Fleetwood Rd., E of Coldwater (Location ID 7454) (Coldwater Creek) and SR219 W of Auglaize County Line (Location ID 7854) (Chickasaw Creek).

2.3.4 Chloride Source Delineation

To delineate the source of chloride in samples taken from Coldwater Creek and Chickasaw Creek, we followed Oberhelman and Peterson (2021) and plotted chloride concentrations vs. nitrate concentrations taken from the same water sample. As Oberhelman and Peterson’s (2021) study occurred in the Eastern Cornbelt Plains ecoregion in Illinois and Grand Lake St. Marys and its watershed lie in the same ecoregion, we used the same cutoff values for source delineation. Following the breakpoints established by Oberhelman and Peterson (2021) study, samples with < 2.5 mg/L nitrate and < 18 mg/L chloride or ≥ 2.5 mg/L nitrate and < 18 mg/L chloride were classified as background, those with < 2.5 mg/L nitrate and > 18 mg/L chloride were classified as road salt, and those with ≥ 2.5 mg/L nitrate and ≥ 18 mg/L chloride were classified as animal, septic, or fertilizer, with samples < 10 mg/L nitrate classified as animal or septic and samples ≥ 10 mg/L nitrate classified as inorganic fertilizer.

2.3.5 Wastewater Plant Sewage Lagoon Discharge Analysis

St. Henry Wastewater Treatment Plant sewage lagoon number of discharge days and volume of discharge were determined for each year for 2017–2021 and a mean ± standard deviation for this range of time was determined using data retrieved from the US EPA ECHO database (https://echo.epa.gov/effluent-charts#OH0020028).

3 Results

3.1 Yearly and Seasonal Trends in Chloride Concentration

Neither Coldwater Creek (Fig. 2a) nor Chickasaw Creek (Fig. 2b) had significant changes in mean daily flow-weighted mean concentration for chloride as apportioned to each season over the period of record for each river (p > 0.05). Coldwater Creek and Chickasaw Creek had similar patterns in flow-weighted mean chloride concentrations with respect to season (Table 1), with the greatest mean concentrations in fall, the lowest mean concentrations in spring and summer/winter mean concentrations in between.

Fig. 2
figure 2

Seasonal mean concentrations for fall (September–November), winter (December-February), spring (March–May), and summer (June–August) from (a) Coldwater Creek (2013–2021) and (b) Chickasaw Creek (2010–2021). The USEPA chronic water-quality criterion of 230 mg/L is denoted by a dashed line on each graph

Full size image
Table 1 - Mean chloride concentrations (mg/L) (± Standard Error) by season for Coldwater Creek (2013–2021) and Chickasaw Creek (2010–2021) with the number of years of sampling (n)
Full size table

Chickasaw Creek and Coldwater Creek flow-weighted mean chloride concentrations were significantly different from each other (3-way ANOVA; F(1,70) = 121.47, p < 0.001), across years (F(12,70) = 2.36, p < 0.022), and among seasons (F(3,70) = 36.99, p < 0.001), and the interaction between rivers and seasons was significant (F(3,70) = 11,14, p < 0.001). Coldwater Creek had a greater yearly mean chloride concentration than Chickasaw Creek (243.59 mg/L vs. 97.37 mg/L) (Tukey’s test; p < 0.05). Across both streams, mean annual chloride concentrations were higher in 2017 and 2021 compared to 2019 (Tukey’s test; p < 0.05). Also across both streams, fall had greater yearly mean chloride concentrations compared to winter or summer, which were both greater than spring (p < 0.05). The mean chloride concentration was greatest in Coldwater Creek during the fall (p > 0.05) compared to all other river/season combinations (Table 2). This was followed by winter and summer in Coldwater Creek, which did not significantly differ from one another (p > 0.05) (Table 2). Furthermore, the Coldwater Creek spring mean was less than the Coldwater Creek summer mean (p < 0.05) but did not differ from fall, winter, or summer in Chickasaw Creek (p > 0.05) (Table 2). Finally, the spring mean concentrations for Chickasaw Creek was less than the fall in Chickasaw Creek (p < 0.05) but did not differ from spring in Coldwater Creek or winter and summer in Chickasaw Creek (p > 0.05) (Table 2).

Table 2 - Results of Tukey test comparisons for Season*River. Means that do not share a letter are significantly different (α = 0.05)
Full size table

3.2 WRTDS

The chloride concentrations in Coldwater Creek showed a substantial increase from 2013 to 2021 at the 50th and 90th percentiles of discharge but decreased slightly at the 10th percentile of discharge (Fig. 3a). Chickasaw Creek chloride concentrations stayed relatively similar from 2010 to 2021 (Fig. 3b).

Fig. 3
figure 3

Chloride concentration estimates over time for a) Coldwater Creek (2013–2021) and (b) Chickasaw Creek (2010–2021). The lines are the 10th, 50th, and 90th flow rates (top to bottom) over the period of record. Data are centered on July 1st of each year. Note that the concentration scales differ by plot

Full size image

The contour plots show the seasonal periodicity of chloride concentrations for both rivers (Fig. 4 a-b). Yearly chloride concentrations have declined in Coldwater Creek, at least at the 0.02–0.03 cms flow rates in Coldwater Creek (Fig. 4a), while in Chickasaw Creek, decreases are evident at low discharges during 2013–2017 before returning to pre-2013 levels after 2017 (Fig. 4b).

Fig. 4
figure 4

Contour plots of chloride concentrations by discharge for (a) Coldwater Creek (2013–2021) and (b) Chickasaw Creek (2010–2021). The color legends indicate chloride concentrations in mg/L. Black lines are 5th and 95th flow percentiles. Note that both the discharge and chloride concentration scales differ between the plots

Full size image

The seasonal patterns visualized in the concentration difference plots show that Coldwater Creek concentrations have increased over the period of record during the summer and early fall at almost all discharge levels and decreased during the winter at lower discharge levels (Fig. 5a), while in Chickasaw Creek the concentrations have stayed relatively the same or even decreased, except at low flows during the winter, where increases in chloride concentrations are evident (Fig. 5b).

Fig. 5
figure 5

Contour plots of the difference in chloride concentrations by discharge for (a) Coldwater Creek (2012–2020) and (b) Chickasaw Creek (2008–2020). The color legends indicate change in chloride concentrations in mg/L over the period of record

Full size image

3.3 Road Salt Application

For Coldwater Creek yearly (1992–2015) road salt average application amounts were 1170 ± 72 metric tons (mean ± standard error) and also increased over time (p = 0.02, r2 = 0.22), but were not significantly correlated with mean chloride concentration for any season (p > 0.05). Similarly for Chickasaw Creek yearly (1992–2015) road salt application average application amounts were 1005 ± 61 metric tons (mean ± standard error) and increased over time (p = 0.03, r2 = 0.20), but yearly road salt application amounts were not significantly correlated with chloride mean concentration for any season (p > 0.05). For the vehicular traffic comparison among the creeks, mean + standard error AADT (Annual Average Daily Traffic) (2013–2015) for SR219 was 5219 ± 505 for Coldwater Creek and 1349 ± 66 for Chickasaw Creek.

3.4 Chloride Source Delineation

Plots of chloride vs. nitrate for Coldwater Creek (Fig. 6a) and Chickasaw Creek (Fig. 6b) were similar in shape, with Coldwater Creek having higher chloride values at low nitrate levels (Fig. 6a) and Chickasaw Creek having higher nitrate levels at low chloride levels (Fig. 6b). With respect to chloride source delineation, both Coldwater Creek and Chickasaw Creek had < 4% of samples classified as background (Table 3). With respect to road salt, Coldwater Creek had 37% of samples classified as road salt, while Chickasaw Creek had 25% of samples in this category (Table 3). 62% of Coldwater Creek samples were classified as animal, septic, or fertilizer, with 36% of samples classified as animal or septic and 25% of samples classified as fertilizer (Table 3). Chickasaw Creek had a greater number of samples classified as animal, septic, or fertilizer (72%), with fewer samples classified as animal or septic (30%) and more (42%) classified as fertilizer (Table 3).

Fig. 6
figure 6

Chloride (mg/L) vs. nitrate (mg/L) for (a) Coldwater Creek (2013–2021) and (b) Chickasaw Creek (2010–2021). Lines are used to delineate the source of chloride (Oberhelman & Peterson, 2021), and cutoff values are listed in the text and Table 3

Full size image
Table 3 - Coldwater Creek (2013–2021) and Chickasaw Creek (2010–2021) chloride source delineation using water sample nitrate and chloride concentrations. Cutoff points are from Oberhelman and Peterson (2021)
Full size table

3.5 Wastewater Plant Sewage Lagoon Discharge

The St. Henry Wastewater Treatment Plant sewage lagoon discharged an average of 135 ± 0.09 days/year (mean ± standard deviation (SD)) between 2017–2021. Days with flow per year ranged from 103 days in 2021 to 168 days in 2017. Mean daily volume ± SD (including zero flow days) was 224 ± 33.2 million gallons of effluent between 2017 and 2021 and ranged from 178.4 (2021) to 263.5 (2019) million gallons (see ECHO database for additional records).

4 Discussion

Other studies have found that increased road salt application over time has led to increases in river chloride concentrations (Daley et al., 2009; Kane et al., 2022), and river chloride concentration is often (but not always) correlated with the amount of urban land use in the watershed (Bird et al., 2018, Gardner & Royer, 2010, Halstead et al., 2014, Winter et al., 2011). Since we have been monitoring the two rivers for less than 15 years and the road salt dataset ends in 2015, we do not have the ability to determine whether in the long term there is a relationship between road salt application and river chloride in these two subwatersheds. However, the finding that the application of road salt has increased in both subwatersheds over the time period of 1992–2015 is consistent with the linear increase in road salt applied in the U.S. during the 1975–2015 time period (Sparacino et al., 2022) suggesting future analysis pending additional data is needed.

Seasonally, for each subwatershed, fall had the largest FWMC for chloride (Table 1). This is consistent with other studies conducted in agricultural watersheds that have also found the greatest values of chloride during low flow fall seasonal conditions (Kane et al., 2022; Oberhelman & Peterson, 2021). Furthermore, Leslie and Lyons (2018) found that reservoirs in nearby central Ohio in the Scioto River watershed can also have high chloride concentrations in the fall. This contrasts with urban watersheds (e.g. Cuyahoga River in Ohio) where chloride concentrations were greatest in winter (Kane et al., 2022), as they have been found to be in other midwestern rivers concurrent with road salt for winter roadways (Corsi et al., 2015). Finally, several studies have shown that soils may retain and release chloride throughout the year, thus obscuring seasonality (Mackie et al., 2022; Molloseau & Steinman, 2023; Robinson et al., 2017).

Historically, there have been few other studies of nearby watersheds with respect to chloride, but two recent studies provide comparable data. In the upstream reaches of the Scioto River to the east of Grand Lake St. Marys at a site in Prospect (Ohio), Wichteric (2022) measured mean chloride concentrations of 51 mg/L, which drains land that is 79% agriculture. In another agricultural watershed to the west of Grand Lake St. Mary’s, Frisbee et al. (2022) found that baseflow chloride concentrations in three headwater sites in the Wabash River watershed in Ohio and Indiana ranged from 56 to 131 mg/L, which was 1.9 to 4.3 times higher than that measured elsewhere in the Wabash River. Two of these sites are in Ohio and lie just west of Grand Lake St. Marys, which also drains into the Wabash River watershed. Frisbee et al. (2022) used several ionic ratios to delineate the source of the elevated chloride and posited that the high chloride levels they found are due to a combination of KCl from fertilizer and NaCl from road salt use. For Chickasaw Creek, mean (± standard error) chloride levels were 91.99 ± 0.80 mg/L (n = 6,310), whereas Coldwater Creek mean chloride levels were more than double Chickasaw Creek with 214.37 ± 3.03 mg/L (n = 4,144). While the Chickasaw Creek mean is below the greatest chloride values measured in the headwater streams feeding the Wabash River, the Coldwater Creek mean exceeds the highest value found for the Wabash headwater streams (131 mg/L). Furthermore, the greatest value of chloride found in Coldwater Creek of 1652 mg/L (9/20/2020) is 12.6 times this value.

Panno et al. (2006) used plots of halide ratios vs. chloride to delineate chloride sources into the categories of precipitation, pristine aquifer, landfill leachate, field tiles, seawater, basin brines, and road salt and septic effluent and total nitrogen vs. chloride to successfully delineate sources of chloride into two categories: landfill leachate, septic effluent, animal waste, and affected waters as one category and road salt and affected waters as another. Our monitoring program does not analyze halides other than chloride and fluoride, so we used the approach of Oberhelman and Peterson (2020) and used chloride vs. nitrate plots to delineate the likely sources of chloride in the Coldwater and Chickasaw Creek subwatersheds. One of our source delineation findings was that the less urban Chickasaw Creek had a lower percentage of chloride attributed to road salt (25% vs. 37% for Coldwater Creek) (Table 3). However, this is a secondary source compared to the main source of animal, septic, or fertilizer for both (> 60%). Other studies in the Midwest that have tried to delineate the source of chloride have made similar findings. David et al. (2015) found that the major source of chloride to the Embarras and Kasksia rivers was from potash (KCl) application to crops, with road salt playing a secondary factor (especially for the Embarras River, which drains urban runoff in the cities of Champaign and Urbana, Illinois).

In addition to potash application, the other main agricultural source of chloride would be through manure. There are 16,478 animal units in the Coldwater Creek subwatershed, as compared to 43,961 animal units in the Chickasaw Creek subwatershed (GLWWA- Grand Lake/Wabash Watershed Alliance, 2008). Compared with one another, Coldwater Creek subwatershed had 10.4% of the poultry, 2.1% of the dairy, 1.1% of the hog, and 2.0% of the beef for a total of 15.7% of all the animal units in the GLSM watershed vs. Chickasaw Creek which had 23.8% of the poultry, 3.6% of the dairy, 10.7% of the hog, and 3.7% of the beef for 41.8% of all the animal units in the GLSM watershed (GLWWA- Grand Lake/Wabash Watershed Alliance, 2008). Although these numbers predate our sampling, they are close enough to demonstrate that the Chickasaw Creek subwatershed has more animal units but a lower number of samples classified as animal or septic (30%) than Coldwater Creek (37%). This may be due to the distressed watershed rules that mandate export of poultry litter from the watershed as a market commodity, as opposed to more typical land application seen in watersheds outside of GLSM (Jacquemin et al., 2018). In addition to the fact that the absolute or relative amounts of different animals may have changed since the GLWWA report, Coldwater Creek subwatershed has villages with larger populations and industries (and thus carries a larger wastewater influence), coupled with the fact that our source delineation technique used does not allow us to separate what is human-based from (other) animal-based chloride.

Overall, 28.69% of the Coldwater Creek samples are above the U.S. EPA. secondary standard for drinking water of 250 mg/L of chloride (U.S. EPA, 2013), while only 2.65% of samples in Chickasaw Creek surpass this value. Based on the U.S. EPA. threshold for chronic chloride toxicity of 230 mg/L and acute toxicity threshold of 860 mg/L (U.S. EPA, 1988), Coldwater Creek had 31.40% samples meet or exceed the chronic value and 0.29% meet or exceed the acute level, while Chickasaw Creek had only 4.11% of samples at or above the chronic level and no samples above 860 mg/L. Nine out of the 12 Coldwater Creek samples that exceeded 860 mg/L were taken between September and November (3 samples taken in December) and thus occurred during the time of typically low flows in Coldwater Creek (WRTDS analyses from this study). Although we have no direct measurements of the water in the St. Henry WWTP sewage lagoon, in late summer/fall 2022, we conducted sampling upstream and downstream of the water treatment outflow into Coldwater Creek. We found that immediate downstream samples (N = 5) had a greater median chloride value of 659.9 mg/L compared to the median value of 270.5 mg/L of the immediate upstream samples (N = 5), and the upstream/downstream sites were significantly different from one another (Mann‒Whitney test, W = 17.0, p < 0.037) (Kane et al., 2023). Furthermore, multiseason and multiyear sampling (not only chemical but also biological) would help elucidate the role that the sewage lagoon plays in chloride dynamics in Coldwater Creek and the differences between a subwatershed with a sewage lagoon (Coldwater Creek) and subwatershed without one (Chickasaw Creek).

More extensive biological surveys could help serve as an indicator of potential impacts manifesting in the ecosystem beyond high chemical values. Importantly, chloride is known to have effects on stream macroinvertebrate survival (Hong et al., 2023; Miltner, 2021) and community composition (Shenton et al., 2022), as is also true for fish (Hintz & Relyea, 2019). Monitoring both macroinvertebrates and fish in Coldwater Creek was part of a biological and water quality survey undertaken by the Ohio Environmental Protection Agency (https://epa.ohio.gov/static/Portals/35/tmdl/Study%20Plan/UGMR-Wabash-QAPP-FS.pdf, and the results should be forthcoming. Assessment tools such as the Chloride Contamination Index for invertebrates (Williams et al., 2000) and fish community surveys (Happel & Gallagher, 2022) should be used and conducted in the Coldwater and Chickasaw Creek subwatersheds to determine whether high chloride levels negatively affect the biota. However, it would likely be difficult to separate chloride toxicity from habitat degradation and other pollutants (e.g. ammonium) in these systems. In addition, sites beyond Grand Lake watershed should also employ chloride monitoring in conjunction with investigating potential biological impacts.

In addition to invertebrate and vertebrate sampling, warm conditions combined with high chloride values and high overall specific conductance also set up the right environmental conditions for altered and imbalanced algal communities (Hartman et al., 2021). For example, Prymnesium (also referred to as golden algae) production has been linked to high salinity conditions and can produce hemolytic toxins (prymnesins) (Sobieraj & Metelski, 2023) that can eliminate animals. In 2022, a Prymnesium bloom occurred in Europe that was referred to as an “ecological disaster” (Sługocki & Czerniawski, 2023). This bloom in Central Europe’s Oder River killed 249 tons of fish along almost the whole course of the river (Sobieraj & Metelski, 2023). The researchers found conductivity levels from 1000 to 5000 µS/cm (Sobieraj & Metelski, 2023) associated with the bloom. In the case of both the Oder and the nearby Vistula River, the high conductivities were due to salt pollution from coal mines (Woźnica et al., 2023). Recent toxic blooms of Prymnesium have also occurred in Lake Koronia in Greece in 2019, killing fish and birds (Demertzioglou et al., 2022), and in West Virginia/Pennsylvania in a small stream (Dunkard Creek) in the Ohio River drainage in 2009, killing fish and unionid mussels (Hartman et al., 2021). The cause of the bloom was once again due to discharge from a coal mine (Hartman et al., 2021). Hartman et al. (2021) found that they could assess the risk of a site for Prymnesium blooms using specific conductivity, chloride levels, and pH. They found that a site that met or exceeded 1500 µS/cm was considered “moderate risk” for P. parvum (36.05% of Coldwater Creek and 6.45% of our Chickasaw Creek water samples met or exceeded this value, data not shown). Sites that exceeded the specific conductivity threshold and either exceeded the chloride level threshold (228 mg/L) or had pH > 7.5 were classified as “high risk” (Hartman et al., 2021). Because of the high levels of chloride in Coldwater Creek, algal samples should be taken and analyzed when high conductivities are found (as this can be measured in the field) to determine if Prymnesium is in the stream water, especially in the late summer and fall when low flows are expected in the system.

Long term climate change predictions for the nearby Great Miami River watershed (as well as many others in the Midwest) suggest that there will be greater low and average streamflows in the near future (Shrestha et al., 2019). Since we saw the greatest increases in chloride concentrations at intermediate and higher flows, more precipitation and more flashy precipitation events would likely lead to increased flows in Coldwater Creek, which could also lead to greater concentrations of chloride. It is hard to predict whether more precipitation would affect sewage lagoon discharges. On one hand, it could make them more frequent as more water would be falling on the lagoon. On the other hand, when the discharges reach the creek there may be a greater volume of water to dilute the chloride levels in the lagoon-derived water. More study of the concentrations of chloride in the sewage lagoon, including modelling the effects of climate change on both discharge from the lagoon and streamflow in Coldwater Creek are needed. At a larger region scale, changing climates combined with ever present runoff issues provide even more opportunity for chloride imbalance in streams. While thus further emphasizing that while this study may focus specifically on a single watershed in Ohio, much more work needs to be done around the Midwest to better understand these patterns.

5 Conclusions

Both Coldwater and Chickasaw creeks have elevated chloride levels, with Coldwater Creek having greater concentrations and more frequent exceedances of chronic and acute toxicity levels. Most significantly, based on the WRTDS analyses, chloride concentrations are increasing at the 50th and 90th percentile flows in Coldwater Creek. Further, the sources of the increased chloride concentrations in the Coldwater Creek subwatershed are likely a mix of road salt, inorganic and organic fertilizer, and sewage-lagoon discharge, as evinced by the chloride source delineation analyses and sewage lagoon upstream/downstream analysis. Therefore, several different types of remedial measures need to be enacted to address the high chloride concentrations in Coldwater Creek.

On a larger scale, climate change may be influencing these increases and/or may lead to future increases in chloride in Coldwater Creek and beyond. Predicted warmer temperatures, paired with greater chloride concentrations, make conditions more likely for harmful algal blooms of Prymnesium parvum. Furthermore, since source delineation points to several sources of chloride in the subwatershed (road salt, animal or septic, and fertilizer) managing chloride levels in this particular subwatershed are going to be a challenge, especially since most of these are nonpoint, diffuse sources of pollution. Thus, we recommend a thorough assessment of the chloride dynamics in the subwatershed, including the St. Henry WWTP sewage lagoon system as one of the few identifiable point sources likely contributing to high chloride levels in Coldwater Creek. It is important to note however, that this subwatershed is not the only place in the Midwest or larger US that exhibits high chloride concentrations and loading as a result of a combination of nonpoint and point source inputs. This case study can hopefully help to serve as a starting point for further investigation and provide baseline data for future conservation work.