Examining Chloride in an Agricultural Watershed Using a Chloride Balance Model Open Access

Chloride pollution of surface water and groundwater resources is a concern in many northern regions due to application of road salt during winter months. It is a particular concern in areas with dense roadways and high impervious surface, namely, cities and urbanized settings. Relationships have been established between increasing chloride levels in streams and amount of impervious surface coverage (Kaushal et al., 2005; Cunningham et al., 2009; Daley et al., 2009; and Gardner and Royer, 2010), road density (Heisig, 2000; Rhodes et al., 2001; and Halstead et al., 2014), urbanization (Siver et al., 1996; Trowbridge et al., 2010), and increasing population density (Smith et al., 1987).

Many studies have examined chloride pollution in urban watersheds within northern regions, where road salt use is prevalent and can lead to elevated chloride concentrations (Kaushal et al., 2005; Novotny et al., 2009; and Corsi et al., 2015). Less research has investigated chloride sources in agricultural watersheds, where road salt may be less important than in urban areas. Studies have shown that potash (KCl) fertilizer can be an important source of chloride in agricultural (David et al., 2016) and mixed land use watersheds (Oberhelman and Peterson, 2020). Chloride levels in livestock waste, for example, can be very concentrated (Panno et al., 2006) and may discharge to surface waters and groundwater from land application or storage (Mullaney et al., 2009), potentially impacting groundwater quality (MPCA, 2001). Chloride loads discharged from wastewater treatment plants can also be significant at the watershed scale (Kelly et al., 2010) with chloride contributions from various domestic, commercial, and industrial sources (Overbo et al., 2021). The importance of individual chloride sources can vary across urban, agricultural, and mixed land use watersheds, which has implications for the timing of chloride transport and observed concentrations for surface water bodies with application of road salt in winter months and application of fertilizers throughout spring and summer months.

Modeling approaches provide practical tools to address the timing and transport of chloride from various sources in a watershed. Water flow and solute transport models, however, can require extensive data and input parameters that may not be available for watersheds with limited monitoring data or resources. A range of tools and models have been applied to examine the sources and timing of chloride loading and concentrations in watersheds, such as a spreadsheet-based mass balance model (Kelly et al., 2008), a simple mixing model (Shaw et al., 2012), or the Integrated Catchment (INCA)-Cl model (Jin et al., 2011; Gutchess et al., 2018). Some of these models require frequent time series data with more extensive data and input parameters, and others were developed for watersheds for which road salt was a dominant source of chloride. Development of a model is needed, however, to characterize chloride sources in a mixed land use watershed with limited data and a variety of point and nonpoint chloride sources.

This study examined chloride levels in a mixed land use watershed using a chloride budget and water balance model to identify predominant chloride sources, compare the relative impact and timing of their chloride loading, and assess their impacts on chloride levels.

We started with a chloride budget to estimate monthly chloride loads from nonpoint and point sources within the watershed for the study period of June 2013–May 2019. Then, we conducted a water balance model that estimated monthly stream discharge. The chloride budget and the water balance model outputs were combined to result in a chloride balance model to estimate total monthly chloride load and stream flow-weighted mean monthly chloride concentrations. An overall summary of the inputs, budgets, and models is provided in Figure 1.

Sand Creek watershed is located southwest of the Twin Cities Metropolitan Area in Minnesota, spanning portions of Scott (52%), Rice (28%), and Le Sueur (20%) Counties and draining into the Minnesota River. There are seven subwatersheds; a portion of the subwatershed is unmonitored and downstream of a monitoring station as seen in Figure 2. The monitored portion of the watershed has an area of 237 square miles and contains several reaches that exceed the state standard of 230 mg/L, which is based upon the chronic toxicity water quality criteria of the U.S. Environmental Protection Agency, including East Branch Raven Stream, Lower Raven Creek, Upper Sand Creek, and Lower Sand Creek (MPCA, 2022).

Land use in the monitored watershed is primarily agricultural (Figure 2); approximately 50% of the land is used for agriculture; 7% is developed urban area; and the remaining watershed area is composed of forest, grass, and wetlands (USGS, 2016). Urban areas in the monitored watershed include the cities of New Prague (population 7,635), Montgomery (population 2,910), and Heidelberg (population 124) and portions of the city of Jordan (population 6,081) (U.S. Census Bureau, 2017). New Prague and Montgomery have wastewater treatment plants (WWTPs) within the monitored watershed, and their effluent chloride concentrations frequently exceed 230 mg/L (MPCA, 2014).

Soil textures in the watershed range from sandy loam to clay loam (NRCS, 2018). A thick layer of clay-rich glacial till extends across much of the watershed, limiting downward movement of water (Tipping and Runkel, 2008). The Scott County Geologic Atlas characterizes much of Scott County as having moderate infiltration ratings with water taking approximately a month to vertically travel from the surface to a depth of 10 feet, reaching groundwater (Tipping and Runkel, 2008). The hydrologic soil group ranking ranges from A to D, where D soil is typically more clayey (NRCS, 2018). Approximately 60% of the soils in the watershed are poorly drained type C and C/D soils with moderately well-drained type B and B/D soils comprising approximately 20% of soils (NRCS, 2018).

Metropolitan Council Environmental Services (MCES) has a monitoring station for Sand Creek located in Jordan, and it has been in operation since 1989 (METC, 2014). Continuous flow monitoring data is available for the site, and chloride monitoring began in 1999 (METC, 2014). The MCES sampling program collects regular grab samples during baseflow conditions as well as composite samples during storm runoff events (METC, 2014).

Monthly monitoring data were retrieved for New Prague and Montgomery WWTPs through the Minnesota Pollution Control Agency (MPCA) Wastewater Data Browser (MPCA, 2017). Data included 24-hour composite chloride concentrations, total monthly discharge in WWTP effluent, and discharge site locations. Chloride contributions from domestic sources (human excretion and household product use) to WWTPs were estimated based on 2016 Census populations for each community (U.S. Census Bureau, 2017), using data sources described in Table 1 and following methods in Overbo et al. (2021). Chloride from household water softeners discharging to WWTPs was calculated based on estimated prevalence of softening, water use, softener efficiency, and water hardness (Table 1) (Overbo et al., 2021). Seasonal chloride loading from inflow and infiltration of road salt from groundwater into wastewater infrastructure was not estimated.

Commercial organizations, such as laundromats, hotels, and restaurants, may soften water for aesthetic benefits, to reduce detergent use, and to reduce buildup of mineral scaling in pipes, fixtures, and appliances. Commercial water use was estimated using data from the Minnesota Department of Natural Resources (MDNR, 2018) and results from a fitted line for cities lacking data (Overbo et al., 2021). Commercial water softening was calculated based on commercial water use, following the methods described for household water softening. Chloride from use of commercial cleaning agents and other commercial processes were estimated based on a flow-weighted concentration of commercial wastewater (Table 1).

Average annual discharge rate and flow-weighted chloride concentrations calculated for Minnesota industries with available monitoring data, including food processing, waste management, and manufacturing facilities, were used to calculate chloride loading based on the two significant industrial user permits in the watershed’s WWTPs (Overbo et al., 2021). Chloride from chlorination of wastewater was estimated using literature values and data described in Table 1.

Chloride from industries discharging to the environment with National Pollutant Discharge Elimination System (NPDES) permits was estimated using monthly monitoring data from the MPCA Wastewater Browser (MPCA, 2017). Industries with NPDES permits included manufacturing, food processing, iron ore mining, and other facilities. Available discharge and chloride concentration data from the same discharge stations were used to estimate monthly and annual chloride loading.

Chloride contributions from septic systems were estimated for parcels with septic systems adjacent to Sand Creek (shoreland parcels). Chloride loading from other septic systems in the watershed was assumed to infiltrate to shallow aquifers and was not included. Groundwater movement requires many years to transverse a watershed, and even C/D soils of the Sand Creek watershed would allow passage to deeper groundwater during this time period (Setterholm, 2006). The contribution of groundwater to the chloride budget is described later in “Methods.” The number of shoreland septic systems within the watershed was estimated using the septic system data in Table 1 by scaling the area of the county within the watershed and assuming 25% is shoreland (Chen, 2022).

Chloride discharged to septic systems from water-softener discharge, human excreta, and household appliance use were estimated as described previously for domestic wastewater, following methods from Overbo et al. (2021) and data on household water use in Table 1.

For county highways, Scott County provided data for the amount of road salt used in the watershed from 2013 to 2019. Annual average road salt application rates for Rice and Le Sueur Counties were estimated based on state purchasing records (Overbo et al., 2021) and scaled by the proportion of county highways within the monitored portion of the watershed (MnDOT, 2012).

Annual average road salt applications for city streets were estimated for New Prague, Montgomery, and Heidelberg, the cities within the monitored watershed area. Estimates were based on per capita usage rates from a linear regression of state purchasing data against community population (Overbo et al., 2021). City stormwater documents and Google Earth were used to identify whether cities employed ditches or curb-and-gutter systems to characterize transport of city road salt application through infiltration or drainage, respectively. Packaged and private bulk deicing were estimated based on previously published salt sales statistics (Sander et al., 2007; Overbo et al., 2021).

Monthly road salt application data for state highway routes in the watershed were obtained from MnDOT. Road salt application data were available for MnDOT routes from 2012 to 2021. Two of the four MnDOT routes in the watershed had monthly data, one had annual data for five years (2017–2021), and one did not have available data. Road salt for routes with missing data was estimated using the other routes’ data as follows: Annual road salt application rate normalized by route length was plotted against annual days of snowfall in Jordan (MRCC, 2018) for the known data in the watershed. A linear regression was created between annual days of snowfall and salt application, and the equation of the fitted line was used to estimate salt use for 2013–2016 and 2013–2019, respectively, for the two routes missing annual data (Figure 3). The annual estimates were distributed across months based on the average ratio of monthly road salt contribution to the annual total across the routes with monthly data.

Potash is normally added to the soil as KCl and is ionized to K⁺ and Cl⁻ when dissolved in water, allowing drain tiles to carry chloride to receiving waters. Data were retrieved for each year to determine the watershed area in corn and soybean rotations and Scott County potash fertilizer application rates for corn and soybeans to estimate the annual chloride mass from potash application (Table 1).

The Minnesota Department of Agriculture provided data on the inventoried livestock within the watershed (MDA, 2016). Chloride from livestock manure was assumed to be land-applied within the watershed and was estimated following methods from Overbo et al. (2021) and the data and literature values listed in Table 1.

Statewide sales data from 2020 were used to estimate chloride from potash fertilizer used for turf management and ornamental use (MDA, 2022). Estimates were available for total annual, statewide potash fertilizer use but were not differentiated by turf or agricultural fertilizer type. Estimates were available for total nutrient fertilizer sales for farm and nonfarm use. The ratios of total farm and nonfarm fertilizer use and of potash fertilizer to total fertilizer were used to impute the potash fertilizer use for nonfarm uses of turf management and ornamental use. This estimate was used for all years in the study period.

County dust suppressant use was estimated using county records, application rates, and length of county gravel roads (Marti and Kuehl, 2013; Overbo et al., 2021) and scaled based on the amount of each county’s area in the watershed.

Chloride deposition from precipitation was estimated using 2013 data from the National Atmospheric Deposition Program (NADP, 2016; Overbo et al., 2021). Atmospheric deposition was estimated for the State of Minnesota and scaled based on the watershed area. This estimate was used for all years of the study period.

Monthly chloride contributions from groundwater were estimated using the average chloride concentration of shallow wells in the watershed and a groundwater baseflow contribution estimate from the data sources and estimates in Table 1.

Monthly total discharge from Sand Creek was estimated using the U.S. Geological Society (USGS) Thornthwaite water balance model (McCabe and Markstrom, 2007), a monthly water balance model developed by USGS with a graphical user interface. The model allocates water in the hydrologic system using monthly accounting following methodology developed by Thornthwaite (1948). Data inputs to the model are monthly mean temperature and monthly total precipitation. Output data from the model include monthly evapotranspiration, snow storage, runoff, soil moisture storage, and precipitation.

Monthly mean temperature data for the Jordan monitoring site were retrieved from the National Weather Service (MDNR, 2019). Daily precipitation data for Jordan were retrieved from the Midwestern Regional Climate Center’s cli-MATE data application (MRCC, 2018). Trace precipitation days were estimated as 0.025 inch (METC, 2010), and data from the National Water Service for Jordan were used for dates with missing data (MDNR, 2019).

Input parameters for the model include runoff factor, direct runoff factor, soil moisture storage capacity, rain temperature threshold, snow temperature threshold, maximum snow-melt rate of snow storage, and the latitude of the location (Figure 4). Soil moisture storage capacity was estimated based on Minnesota values for corn and soybean rooting length and available water capacity of loam (UMN, 2019), the predominant watershed soil type. The direct runoff factor was estimated based on previous modeling of the watershed from 2001 to 2008 (METC, 2010). For the remaining parameters, values suggested by McCabe and Markstrom (2007) were used for initial model inputs.

The model was calibrated using the soil moisture storage capacity, runoff factor, snow temperature threshold, and melting factor as fitting coefficients, for which the model was run using values from the high, low, and midpoints of the estimated ranges and calibrated in increments available in the water balance model, moving from the midpoint toward the high and low values of the range. The calibrations were evaluated using percentage difference from the modeled discharge and discharge estimates provided by the Metropolitan Council (METC, 2021) and visual observations from peaks in the hydrograph. Input parameters selected for the model were similar to those suggested by McCabe and Markstrom (2007) as provided in Table 2.

Chloride mass from monitored point sources were allocated to the chloride mass for the month they were discharged, including WWTP and industrial discharge. Road salt from cities with curb-and-gutter drainage was assumed to enter Sand Creek in the month that it was applied.

Drain tile monitoring data were retrieved from the Minnesota Department of Agriculture Discovery Farm sites to estimate chloride runoff from fields with potash and manure application (MDA, 2023). Data were retrieved from a Blue Earth site in a corn–soybean rotation with manure and urea application for 2012 to 2019 and a Renville site with potash application for 2012 to 2020, which included monthly total runoff and monthly total chloride load. For each of the sites, a linear regression was performed between monthly total runoff and monthly total chloride load in field drain tiles (Figures 5 and 6).

The relationship between monthly total runoff and monthly total chloride load from potash and manure via tile drain established with the Discovery Farm sites data was applied to the monthly runoff generated by the 23 percent of tiled agricultural area in the study watershed (METC, 2014). The potash and manure chloride loads via drain tile were added to the monthly load total from point sources. The drain tile loads were subtracted from the potash and manure chloride loads estimated in the chloride budget, and the remaining load was summed with other nonpoint sources.

Estimated chloride loading from all other, remaining nonpoint sources in the chloride budget were summed over the annual season (June–May) and divided by the total runoff for that period to estimate an annual flow-weighted mean chloride concentration for runoff. This concentration was applied to the monthly runoff to estimate chloride transport and loading from nonpoint sources across the watershed.

Results from the chloride balance model were compared with available monitoring and modeled data for the watershed. Sand Creek chloride sampling results and monitored flow data were retrieved from the Metropolitan Council Environmental Information Management Systems for 2013–2019, the period of study. The Metropolitan Council provided its estimated monthly chloride loading and discharge in the watershed using the FLUX-32 model for the period of study (METC, 2012).

Over the study period, road salt was estimated to contribute on average 1,530 t of chloride per year. Chloride from agricultural sources was estimated to be the second largest source of chloride; manure and potash application contributed on average 656 t and 900 t of chloride annually, respectively. Wastewater was also a major source of chloride with annual averages of 777 t from the two WWTPs in the monitored area, 206 t from shoreland septic systems, and 56 t from permitted industry. In comparison, chloride from groundwater, atmospheric deposition, dust suppressant, and turf fertilizer was relatively minor (Figure 7A).

Public road salt use was estimated as 1,163 t per year, on average, with annual averages of 291 t and 76 t for private bulk use and packaged road salt use, respectively. The greatest amount of road salt was contributed by county roads and highways, which make up a larger proportion of roads in the watershed.

Across the study period, the majority of the chloride loading from manure was estimated to enter the stream through drain tile, ranging from 51 to 94 percent of the manure load across annual seasons (Figure 7B). A smaller proportion of the potash was estimated to be discharged through drain tile, ranging from 23 to 38 percent across the study period seasons (Figure 7B).

In the wastewater budget, household water softening was found to be a major source of chloride for Montgomery and New Prague WWTPs (averaging 40 and 53 percent of the chloride discharge, respectively) as shown in Figure 8. The proportion of chloride from commercial water softening was also estimated to be relatively high for New Prague WWTP (27 percent), whereas industrial discharge was a greater contributor to Montgomery WWTP (35 percent). For Montgomery WWTP, on average across the study years, 19 percent of the total chloride load was not accounted for though this unaccounted chloride was largely driven by 5 months across the study period with significantly higher chloride concentrations in the monitoring data and estimated chloride loads compared with other months. Chloride loading from WWTPs was fairly consistent across seasons and study years and was a major source during low-flow winter months along with curb-and-gutter road salt (Figure 7B).

Across the study period, water softening was estimated to contribute 94 percent of the chloride load discharged through septic systems in the watershed.

One industry had an NPDES permit in the watershed with chloride monitoring, a food processing facility. It had three monitored discharge sites during the period of study. Its two surface water discharges were used until 2016 and contributed, on average, 4 t of chloride per year. The spray irrigation site had greater discharge than the surface water sites and averaged 53 t per year throughout the study period.

Shallow groundwater chloride concentrations were estimated as 1 mg/L based on average results from six wells in the upper quaternary aquifers, which were sampled between 1993 and 2015.

The runoff estimated by the model was generally shown to be in good agreement with Metropolitan Council estimates for the watershed discharge throughout the study period (Figure 9A). Both models generally saw higher discharge in spring months, indicative of melting periods, and some higher discharge during winter months.

Compared to Metropolitan Council volume estimates, there was an average 56 percent and median 39 percent difference between the values with a Nash–Sutcliffe Efficiency Index of 0.54.

The chloride balance model showed higher chloride loading across summer and early fall (May–November) with an average of 416 t across these seasons compared with an average of 252 t across the winter months (December–April). The loading from the chloride balance model generally followed the same trends as the discharge with higher chloride loads typically happening during spring snowmelt months and occasional peaks during the late fall or early winter months (Figure 9B).

The estimated chloride loading calculated from the chloride balance model generally showed similar trends and magnitudes compared with Metropolitan Council estimates. Months in which the two estimates had greater discrepancies generally corresponded to months with greater discrepancy in estimated discharge (late summer 2013 and 2014, winter 2015–2016, and early 2019) (Figure 7A). Compared with Metropolitan Council load estimates, there was an average 48 percent difference between the values (median 39 percent), with a Nash–Sutcliffe Efficiency Index of 0.53. The chloride loading estimates from the chloride balance model were typically lower than those provided by the Metropolitan Council.

Over the study period, the average flow-weighted mean monthly chloride concentration was estimated as 28 mg/L from May through November and 88 mg/L from December through April. Flow-weighted mean monthly concentrations were generally within the same magnitude as grab sample results from 2015 to 2017 (Figure 9C). Flow-weighted monthly average chloride concentrations from the chloride balance model were generally highest in February, a low-flow month. Elevated chloride concentrations in the flow-weighted mean estimates lag the elevated concentration indicated by grab samples in late 2013, 2014, and 2017 or do not appear as in summer 2018.

Based on livestock permitting records, there was generally a higher concentration of livestock animal units within the impaired subwatersheds compared with Porter Creek, SC Ditch 10, and West Branch Raven Stream (Figure 10). Both of the municipal watersheds in the monitored area of the watershed are within impaired subwatersheds as well as Seneca Foods, a food processing facility.

The percentage of area in cultivated crops was similar within the impaired and nonimpaired subwatersheds; nonimpaired SC Ditch 10 and West Branch Raven Stream had the highest percentage of cultivated crop land use compared with the other subwatersheds (Table 3). The impaired subwatersheds generally had higher proportions of highway miles in the watershed relative to their area although West Branch Raven Stream is not impaired and also had a relatively high proportion of highway miles for its area (Table 3).

Road salt was found to be the largest source of chloride even in this agricultural watershed. Agricultural sources, however, contributed significant amounts of chloride in the watershed, including upper reaches with chloride impairments with more predominant row crop agriculture and a high concentration of feedlots, in which greater application of potash and manure are expected. This is consistent with findings by Hubbart et al. (2017) in analysis of a mixed land use watershed in Missouri, in which chloride concentrations were highest midwatershed, in part due to the increased loading and low dilution compared with upper and lower reaches. Fertilizer and manure have been identified as significant sources of chloride in other watershed studies; Thunqvist (2004) found them to be the second largest chloride source after road salt in a Swedish watershed, and Olson et al. (2003) found that repeated application of manure resulted in mobilization of chloride to groundwater, significantly affecting soil and groundwater quality when manure application rates were high. The impaired subwatersheds had a number of large feedlots, generally a greater concentration of animal units, and the WWTPs and permitted industry in the watershed, suggesting that agricultural and wastewater chloride loading can add substantial chloride to streams and result in an impaired designation for chloride.

Wastewater was also an important chloride source with WWTPs being the third largest individual chloride source, following road salt and potash. Water softening was estimated to be the largest source of chloride to the WWTPs, contributing approximately half of the total chloride loading to Montgomery and New Prague WWTPs. Chloride loading from other commercial processes, domestic sources of chloride, chlorination of drinking water and wastewater, and background drinking water levels were minor in comparison. The chloride budget underestimated chloride loading for the Montgomery WWTP, which had a few months of significantly elevated chloride levels in its effluent monitoring data. The underestimated loading could potentially be attributed to above-average industrial or commercial discharge or chloride from hauled septage, which is accepted by the WWTP (MPCA, 2018) and was not estimated in the analysis.

Results from the water balance model generally showed good agreement with modeled results from the Metropolitan Council with similar trends and magnitude. Discrepancies in chloride loading estimates generally occurred during periods in which there were also discrepancies in discharge between the water balance model and the Metropolitan Council hydrologic model as runoff was a used as a surrogate for discharge and also drove the timing of nonpoint chloride loading in the chloride balance model. The watershed chloride loading was generally lower than Metropolitan Council estimates, which could indicate underestimating chloride loads or potentially a longer retention time in the watershed.

The estimated flow-weighted mean chloride concentrations were generally higher in low-flow conditions, consistent with previous research, which attributed the trend to groundwater contributions or spring snowmelt with dissolved road salt (METC, 2014). Results from the chloride balance model indicate that, during low-flow periods, discharge from WWTPs also helped drive the elevated chloride concentrations and contribute to chronic water quality standard exceedances, supporting previous suggestions that WWTPs are important factors in the watershed chloride impairments (MPCA, 2016), particularly as they discharge to chloride-impaired reaches of Sand Creek.

Although flow-weighted mean monthly chloride concentrations are not directly comparable to grab samples, there were similarities between the magnitude, trends, and seasonality of the grab sample and flow-weighted mean monthly chloride concentrations. Estimated flow-weighted mean chloride concentrations and grab sample data generally showed higher chloride concentrations in late winter months, which can be attributed to road salt use and runoff (Hubbart et al., 2017). Sand Creek monitoring data showed elevated chloride concentrations in late summer and early fall months in some years that were not observed in the flow-weighted estimates, but the greater variability in summer months could be attributed to mobilization of chloride retained in subsoils (Kelly et al., 2008; Gardner and Royer 2010; and Beom et al., 2021). The highest chloride concentrations occurred in 2013, 2014, and 2017 although elevated chloride concentrations were seen in grab sample data ahead of the modeled concentrations. This could be indicative of chloride runoff from short-term application or melting events that were not captured in the model, which had a monthly time step. Additionally, seasonal fluctuations in stream chloride concentrations can be buffered by chloride storage in soil or groundwater within floodplain areas hydrologically connected to the stream (Ledford et al., 2016), which were not investigated in the chloride balance model.

Results from the chloride balance model (and Metropolitan Council model) showed that, on average, chloride loading was greater during nonwinter months (May–November), suggesting that agricultural practices had a large influence upon chloride loading. This timing could also indicate retention and delayed transport of chloride from road salt applied in winter months. Past research in Minnesota has shown significant chloride loading in snowmelt periods along with delayed chloride transport (Herb et al., 2017; Klimbal 2020). In a study following chloride transport from road salt application through stormwater conveyance, Klimbal (2020) found that 70 percent of expected chloride was observed in surface runoff during late winter and early spring melting events, but approximately 30 percent was not sensed during the monitoring period. This is supported by findings by Herb et al. (2017) who found significant retention of chloride from road salt application with between 44 and 94 percent of chloride not reaching receiving waters across study watersheds, potentially due to retention in unsaturated soil, shallow groundwater, ponds, and wetlands. Significant chloride retention and release over time has been observed in soil column experiments (Erickson et al., 2019). Several mass balances investigating chloride in urban watersheds have found significant chloride retention in groundwater and subsoils (Novotny et al., 2009; Eyles and Meriano, 2010; and Perera et al., 2013). Delayed transport of chloride has also been observed in research in settings with more agricultural and mixed land use compared with urban land use (Frey et al., 2012; Oswald et al., 2019; and Beom et al, 2021). Whereas the chloride balance model assumed that chloride from nonpoint sources was transported to the stream within a year, the approach and findings do not preclude the possibility of chloride loading from past years and seasons. The chloride loads estimated by the Metropolitan Council model were based on sampling data and exceeded the loading estimates from the chloride balance model, which could indicate loading from legacy chloride within the watershed that was applied prior to the study period.

Whereas high chloride retention in soils and groundwater contribute to increasing chloride concentrations in streams, high chloride residence times indicate that increases in stream concentrations could continue over long time periods. Research in a central New York watershed found a chloride residence time of 20–30 years, indicating long-term retention and suggesting that stream chloride concentrations would continue rising over several decades (Gutchess et al., 2016). Legacy chloride can also contribute to the long-term increases in groundwater chloride concentrations (McDaris et al., 2022). Previous modeling in the Sand Creek watershed showed that flow-adjusted chloride concentrations increased by about 37% in the 20-year period of 1999 through 2019 (METC, 2021), which could reflect high retention and gradual transport of legacy chloride to the stream.

The chloride balance model was developed as a tool for watershed study with limited monitoring data and without data-intensive modeling. The model inputs were simplified, using the same input parameters for each year, whereas some parameters, such as the direct runoff factor, could be expected to change annually. The monthly time step, however, yields limited data resolution for the watershed, not providing insight into the impacts of individual chloride application, runoff, or melting events. The input data of monthly precipitation does not take into account event factors, such as rainfall depth, rainfall intensity, or antecedent dry days, which can have a first flush effect on contaminants (Gupta and Saul 1996; Kim et al., 2004; and Jeung et al., 2019).

Future research in the watershed could further investigate major chloride sources to identify implications from the timing of their application, transport, and release for watershed chloride loading and in-stream concentrations. Additional monitoring along reaches of the stream using chloride-bromide ratios would help identify and characterize significant chloride sources along the impaired streams and subwatersheds (Gutchess et al., 2016). Modeling to identify relative contributions of surface water, groundwater, and urban groundwater can also inform watershed management decisions (Montgomery, 2023).

Road salt was found to be the highest source of chloride in this primarily agricultural watershed, followed by potash fertilizer applications, WWTP contributions, and feedlot manure applications. Wastewater contributions were especially important in low-flow periods along with road salt from curb-and-gutter drainage. Results from the chloride balance model suggest potential chloride retention within the watershed across winter and nonwinter seasons with major melting and precipitation events influencing chloride loading and levels within the stream. Using runoff as the basis for chloride transport from nonpoint sources is a useful indicator for local application as surface runoff can be readily estimated. Overall, the results suggest that using a chloride budget and water balance model approach can yield useful results for watershed management in a predominantly agricultural watershed and provide an applicable approach for mixed land use watersheds.

Funding for this paper was provided by the Minnesota Environment and Natural Resources Trust Fund as recommended by the Legislative–Citizen Commission on Minnesota Resources under the project “Impacts of Adding Salt to Our Minnesota Lakes, Rivers, and Groundwater” as funded under Legal Citation: M.L. 2016, Chp. 186, Sec. 2, Subd. 04n. The Trust Fund is a permanent fund constitutionally established by the citizens of Minnesota to assist in the protection, conservation, preservation, and enhancement of the state’s air, water, land, fish, wildlife, and other natural resources. We would also like to thank the following individuals for providing data to support this research: Melissa Bokman, Joe Wiita, Pete Schmitt, Greg Wagner, and Jon Utecht from Scott County; Wendy Surprise from the Minnesota Department of Transportation; and Katie Brosch Rassmussen from the Minnesota Department of Agriculture. The first author also thanks Bob Tipping, Bruce Wilson, and Jeff Peterson at the University of Minnesota for their review of her thesis with a draft of this paper as one chapter.