Experimental Design

Highway runoff, surface water, and groundwater were investigated in three areas with different experimental deicing treatments along sections of Route 86 and Route 73 in the Adirondack Park (fig. 1). The watersheds were divided into upgradient and downgradient subwatersheds, relative to the intersection of Route 86 or Route 73 and the selected stream. Upgradient subwatersheds were monitored to understand background conditions and were presumed to be minimally affected by deicing practices because those areas have little to no impervious surface cover. The downgradient subwatersheds were monitored to quantify potential impacts from ongoing winter snow and ice management practices on water quality. One watershed was identified as a control, and winter deicing was done in the typical way without altering the application rate or the deicing mix composition. Two watersheds were subjected to different treatments than the control and identified as “Treatment A” and “Treatment B,” explained as follows.

The control treatment by the NYSDOT on Route 73 consisted of a typical application of deicing salt (George T. Laundrie, professional engineer, NYSDOT, written commun., 2021). The deicing salt was sodium chloride that also contained a prewetting agent, coloring agent, and a corrosion inhibitor; this was used on all three treatment areas. Deicing salt application rates in the control section ranged from 90 to 360 pounds (lb) per lane mile. Treatment A reduced the salt concentration of the deicing mixture. Sand was added to the deicing salt, creating a variable abrasive-deicing mixture. Treatment B reduced the application rate of the typical deicing salt.

Individual application rates were determined relative to temperature, weather, and pavement conditions in accordance with the NYSDOT snow and ice guidelines (NYSDOT, 2006). Application rate and frequency on highways were adjusted as needed to maintain safe driving conditions (George T. Laundrie, professional engineer, NYSDOT, written commun., 2021). In the Treatment A highway reach, an approximate 50:50 sand-salt mixture was applied at 250 lb per lane mile during the 2019 to 2020 winter period, and a 90:10 sand-salt mixture was applied at 750 lb per lane mile during the 2020 to 2021 and 2021 to 2022 winter periods. Application rates on the Treatment B highway reach were about 10 percent lower than those on the control highway, although deicing salt was applied more frequently to stretches of the Treatment B highway reach than in the control or Treatment A reaches during winter snow and ice events. A 23- to 24-percent brine solution was occasionally applied to the pavement in the Treatment A and B reaches at a rate of 40 to 50 gallons per lane mile (George T. Laundrie, professional engineer, NYSDOT, written commun., 2021); however, application amounts and rates for brine solutions were not supplied by the NYSDOT for this analysis.

The NYSDOT provided estimates of individual material application rates, materials used, and lane miles treated for each highway reach (Fisher and others, 2025). The Town of North Elba provided annual application rates for the town road within the control watershed. Annual average daily traffic volume estimates were about 4,064 and 4,032 vehicles per day in 2019 for Route 73 and Route 86, respectively (NYSDOT, 2020).

Annual chloride mass balance (sum of total chloride inputs and total chloride export) was determined for the Treatment B watershed. Surface electromagnetic surveys were collected upgradient and downgradient from the Treatment B highway reach along the general inferred direction of subsurface flow at the groundwater monitoring sites. Electromagnetic surveys map the apparent electrical conductivity of the subsurface and can be used to estimate changes in the specific conductance of groundwater. As many as three nested groundwater monitoring wells were upgradient and downgradient from the highway reaches but outside of the upgradient and downgradient surface water subwatersheds. Groundwater monitoring wells were installed at different depths (23 to 38.8 ft below land surface at wells adjacent to the Control highway reach, 15 to 19.5 ft below land surface at wells adjacent to the Treatment A highway reach, and 8.3 to 28.8 ft below land surface at wells adjacent to the Treatment B highway reach). Discrete samples were collected from upgradient and downgradient monitoring wells for water quality analysis. In combination, electromagnetic survey and discrete groundwater monitoring well data show the transport of chloride from deicing salt runoff through the Treatment B watershed (fig. 2).

Data Collection at Highway Monitoring Stations

Design of Highway Monitoring Stations

Highway monitoring stations and paired noncontact and embedded pavement sensors were installed at a single location along each highway reach and operated by the USGS (fig. 1; table 2). Highway monitoring stations included automatic water samplers and sensors for the measurement of highway runoff stage and specific conductance, water and air temperature, snow depth, and precipitation (fig. 3). Noncontact and embedded pavement sensors measured surface pavement temperature and pavement condition (whether the pavement was dry, damp, or wet, or covered in snow, ice, or black ice). Highway condition data are available from the National Water Information System (NWIS; USGS, 2023).

Table 2.    

Highway condition and runoff water-quality monitoring stations along New York State Routes 73 and 86, Essex County, New York.

[The number of lanes and speed limits were obtained from direct field observations. Site information from U.S. Geological Survey (2023). ft², square foot; mph, miles per hour; NY, New York]

Table 2.    Highway condition and runoff water-quality monitoring stations along New York State Routes 73 and 86, Essex County, New York.

USGS site identification number	Station Name	Treatment area	Latitude (decimal degrees)	Longitude (decimal degrees)	Drainage Area (ft2)	Number of lanes	Speed limit (mph)
441411073544501	Control, south side Route 73 catch basin outlet, near North Elba NY	Control	44.236500	−73.912639	9,840	2	55
441729073571801	Abrasives, east side Route 86 catch basin outlet, near Lake Placid NY	Treatment A	44.291519	−73.955261	4,800	2	55
442004073540301	Low salt, west side Route 86 catch basin outlet, near Lake Placid NY	Treatment B	44.334531	−73.901081	8,740	2	35

Paved asphalt waterways (about 200 ft long on one side of the highway) and drop inlets with 2-ft sumps were installed at each highway monitoring station beginning in November 2019 and completed by December 2020. These waterways diverted highway runoff to the drop inlets where continuous measurements and automatic sample collection occurred. Discharge was measured at a two-stage volumetric weir (rated for about 0.95 cubic feet per second [ft³/s]) installed in the outlet pipe of each drop-inlet sump. Continuous water-level measurements were converted to continuous discharge values by programming the data loggers with the stage-discharge relation for the weirs. Highway runoff discharges were recorded every 10 minutes for discharges less than 0.004 ft³/s. Specific conductance and discharge were recorded at 1-minute intervals for discharges greater than 0.004 ft³/s, which corresponded to the minimum flow needed to submerse the specific conductance sensors. Composite samples of highway runoff were retrieved concurrently with discrete water quality sample collection.

Meteorological data were measured at each monitoring station at a height of about 7 ft above the ground surface. Precipitation data were measured with a 7.9-in diameter electronic weighing-bucket sensor (fig. 3). Meteorological data were measured at 5-minute intervals. Data were retrieved automatically through cellular telemetry.

Noncontact pavement sensors were mounted at an angle of about 40 degrees on masts about 20 ft above ground level. These sensors observed an elliptical field of measurement of about 13 in by 70 in, which primarily captured the lane adjacent to the highway stations. The noncontact sensors used electro-optical technology to monitor the near infrared spectral differences of the roadway surface and could detect the difference between the state of water on the pavement: water, snow, or ice (High Sierra Electronics, 2022). The sensor measured pavement temperature with an infrared thermometer, and air temperature and relative humidity were measured within the device housing. The noncontact pavement sensors were calibrated to dry pavement conditions about every 2 months. Pavement sensor data were measured at 5-minute intervals. Data were retrieved automatically through cellular telemetry.

The embedded pavement sensors (Lufft IRS31Pro–UMB) were installed in August 2020 at about 4 ft from the fog line, or just on the inside of the typical position of the right tire for vehicles traveling on the road (fig. 3). The embedded pavement sensors measured temperature just beneath the pavement surface and at a depth of about 1 ft below the pavement surface. The instrument accuracy was plus or minus 0.5 degrees Celsius (°C; High Sierra Electronics, 2022). High-resolution fixed-image cameras were mounted at highway monitoring stations at about the same height as the noncontact pavement sensors and focused on the general location monitored by the noncontact and embedded pavement sensors. Cameras were operated in accordance with guidelines established by the manufacturer (High Sierra Electronics, 2022).

Pavement-condition data are a semiquantitative measurement, where a “damp” pavement condition means that there is water on the road but an insufficient amount to cause tire spray, a “wet” pavement condition means that the road has sufficient water to cause tire spray, and a “standing-water” condition means that a wake is observed behind the tires of moving vehicles. “Snow,” “ice,” and “black ice” conditions mean that snow, ice, and black ice, respectively, were observed on the road surface. Images were collected with the fixed-image cameras and used in combination with sensor data to validate the accuracy of semiquantitative pavement conditions. Pavement conditions were checked, embedded temperature sensors were inspected, and the pavement surface temperature was tested against a reference meter during routine site visits for water quality sample collection (about monthly).

Surface Water Data Collection

Six surface water sites were monitored from November 2019 to August 2022 for specific conductance (table 1) and sampled routinely for chloride. Discrete samples were manually collected about every 2 months and analyzed by the USGS NWQL. Specific conductance and water temperature were measured at 15-minute intervals using Eureka Manta2 water-quality multiprobe sondes. The Eureka Manta2 water-quality multiprobe sondes were installed in the stream channel and deployed from November 2019 through November 2022. Routine sensor maintenance included calibration checks and cleaning to reduce fouling and prevent sensor drift. Water-quality sampling and continuous data collection for surface water were performed by USGS personnel following USGS protocols (Wagner and others, 2006; USGS, variously dated).

Stream stages were recorded at 15-minute intervals using pressure transducers and data loggers at three sites from November 2019 to November 2022. Manual discharge measurements were made at regular intervals and during high-flow conditions. Gage height-discharge relations were developed for each station to determine the discharge at 15-minute intervals from stage measurements (Sauer, 2002). Stream discharge was computed in accordance with standard USGS protocols (Rantz and others, 1982). Water-quality and discharge data are available from NWIS (USGS, 2023).

Groundwater Data Collection

Near Surface Geophysical Methods

Surface electromagnetic surveys were used to map spatiotemporal changes in groundwater salinity adjacent to a section of the Treatment B reach (Terry and others, 2024). Results support chloride mass balance for surface-water sites in this watershed. Surface electromagnetic surveys were collected using a Geophex GEM–2 in an approximate 500 square foot grid pattern in areas upgradient and downgradient from the Treatment B highway reach between July 2019 and October 2022. Hand-carried surface electromagnetic surveys can be used to map relative subsurface bulk electrical conductivity variations (as from variable groundwater salinity) down to several ft below land surface (bls). Electrical geophysical methods have been used for similar purposes in other studies; for example, mapping the extent of a high chloride groundwater plume downgradient from Interstate 93 in New Hampshire (Toran and others, 2010) and delineating recharge patterns downgradient from a deicing salt storage facility in Maine (Slater and Sandberg, 2000).

The GEM–2 is a multiband electromagnetic instrument consisting of a Global Positioning System (GPS) receiver and battery powered electronics mounted to a rigid ski in which a transmitter and receiver coil are embedded at a fixed offset (5.45 ft). The transmitter emits an electromagnetic field at multiple frequencies that induces eddy currents in subsurface conductors (such as groundwater with elevated salinity), which are measured by the receiver. The instrument was carried flat (horizontal coplanar) at waist height (about 3.3 ft) at a comfortable walking speed (about 3 miles per hour). GEM–2 quadrature and in-phase data were measured at six frequencies (3930, 8670, 19110, 28950, 42210, and 93090 hertz) in real-time using a Bluetooth connection to a Trimble Nomad or similar Windows mobile device using WinGEM2 software (Geophex Ltd., 2021). Raw binary data were exported to comma separated value format using the EMExport software and converted to apparent electrical conductivity using the Invertor software (Geophex Ltd., 2021). Apparent electrical conductivity represents a volume-averaged electrical conductivity and is not associated with a specific depth. Apparent conductivity data were linearly interpolated to 10 ft by 10 ft grids using the R “interp” function available in the akima package (Akima and Gebhardt, 2022). The spatial distribution of mapped electrical conductivities throughout the survey area were examined and compared across seasonal data collection events at the site. Data associated with GEM–2 surface electromagnetic surveys are available in Terry and others (2024).

Chloride Estimation in Highway Runoff

Continuous specific conductance was used to estimate chloride concentrations in highway runoff using regression equations (Hem, 1985; Spaetzel and Smith, 2022). Regression equations were developed using concurrent measurements of continuous specific conductance and chloride concentration measured in discrete and composite samples of highway runoff. Regression equations were applied to continuous specific conductance data to estimate chloride concentrations at 1-minute intervals for each highway monitoring station. These regression equations were developed with the MOVE.1 technique on the basis of concurrent measurements of log-transformed values of specific conductance and log-transformed concentrations of discrete and composite chloride in highway runoff samples (Helsel and others, 2020). The MOVE.1 technique minimizes errors in the x and y directions and has demonstrated utility when applied to water quality of highway runoff in previous studies (Stonewall and others, 2019; Smith and others, 2023). MOVE.1 generates a unique equation, which can be used to estimate either variable from the other; the estimations reflect the variance of the measured concentrations (Helsel and others, 2020). The MOVE.1 equation (eq. 2) is given as

The ionic ratio of highway runoff can vary across ranges of specific conductance spanning several orders of magnitude (Smith and others, 2023). Owing to this variability, two-part regressions were selected to represent the range in specific-conductance values and chloride concentrations of highway runoff (table 4). In the discrete and composite runoff sample data (n = 83), continuous specific conductance ranged from 24 to 73,200 microsiemens per centimeter (μS/cm) at 25 °C, and chloride concentrations ranged from 0.75 to 29,200 milligrams per liter (mg/L). Part 1 equations were applied to estimate chloride concentrations at low specific conductance (less than 150 μS/cm for the control highway reach and less than 290 μS/cm at the Treatment A and B highway reaches). Part 2 equations were applied to estimate chloride concentrations at high specific conductance (greater than or equal to 150 µS/cm at the control highway reach and greater than 290 µS/cm at the Treatment A and B highway reaches).

These multisegment regression equations were developed using pooled sample data from all stations (table 4). The relative proportion of individual ion contributions to specific conductance can vary as a function of site characteristics or magnitude of specific conductance (Smith and others, 2023). Thus, two-part regressions were selected to better represent the large range in specific-conductance values and variable ion contributions at these stations. Although regression equations were applied to nonoverlapping ranges of specific conductance (table 4) to produce estimates of chloride concentrations, the ranges of observed data used to develop each segment overlap to produce a continuous function (Granato, 2006). The root mean square error was 15–19 percent for part 1 equations and 1.7 percent for part 2 equations.

Flow-weighted chloride concentrations were calculated to account for the potential effects that differences in highway runoff rates may have on chloride concentrations. Flow-weighted chloride concentrations for highway runoff were calculated for a given date using the following equation:

Estimation of Annual Chloride Yields and Mass Balance in Surface Water

An annual chloride mass balance for the Treatment B watershed for water years 2021 and 2022 was estimated. A water year is defined as the period from October 1 to September 30 and designated by the year in which it ends. Total annual inputs for the Treatment B watershed included the sum of chloride from deicing, atmospheric deposition (estimated for the downgradient watershed only), and chloride export from the upstream surface water station, the last of which constitutes background chloride contributions. Contributions of chloride from septic effluent were assumed to be negligible because residential development was low to absent in the watershed. These data were compared annually for the Treatment B watershed and expressed as pounds of chloride. The difference between the sum of inputs and downstream chloride loading represents the estimated mass of chloride stored annually in the Treatment B watershed.

Interannual Variations in Winter Conditions, Deicing Salt Applications, and Highway Runoff Quality

Winter 2022 weather and pavement conditions were more severe than 2021 conditions, with more total snowfall (fig. 4C), more observations as snow and ice (fig. 5J–L), and lower air (fig. 4A) and pavement temperatures (fig. 5G–I). However, differences in winter pavement conditions across treatments were less pronounced (fig. 6). Total precipitation was higher during winter 2022 than winter 2021 and was characterized by more frequent, higher intensity snowfall events (fig. 4B–C), larger ranges in daily air temperature (fig. 4A), and more frequent runoff events (fig. 5A). The number of days with surface pavement temperatures at or below freezing was higher during winter 2022 than winter 2021 at all highway monitoring stations (fig. 6). More pavement conditions were observed as snow or ice during winter 2021 than 2022 at the control station. Daily ranges in pavement temperature were higher during winter 2022 and high-volume runoff events also occurred more often during 2022 (fig. 5). Pavement condition observations were not recorded consistently at the Treatment A and B stations for winter 2021. Snowfall totals in 2021 and 2022 (131.9 and 153.2 inches, respectively) were notably higher than the average annual snowfall total (105.4 inches; National Oceanic and Atmospheric Administration, 2024).

Chloride applied for deicing mirrored winter severity except when affected by management practices, such as increased applications to improve conditions in particularly icy areas of the roadway or during adverse conditions. Fewer observations as snow or ice were recorded at the Control station during the winter of 2022 than in 2021 (fig. 5). More deicing salt was applied to the control and Treatment B reaches in the 2022 winter season (fig. 6). Less chloride was applied for deicing in the Treatment A watershed during winter 2022, because the ratio of salt to sand was reduced during winter 2021 (a 50:50 sand-salt mixture at a rate of about 250 lb per lane mile was used through December 2020, and a 90:10 sand-salt mixture at a rate of 750 lb per lane mile was used through 2022). Higher flow-weighted chloride concentrations were observed in highway runoff during winter 2022 than winter 2021 (fig. 6B; table 5) when more salt was applied to Treatment B and the control highway reaches (fig. 6D). Highway runoff rates and water quality were not collected with consistent frequency in the Treatment A reach during 2021. Chloride concentrations in highway runoff peaked during winter months when runoff events occurred after deicing salt application and declined in subsequent months from meteorological spring (March, April, and May) through early fall (September and October; fig. 5). Although winter 2022 for the control reach was characterized by higher total snowfall rates (fig. 4C), higher total deicing salt application, and lower pavement temperatures (fig. 6), sporadic snowmelt and subsequent runoff events also occurred more often (fig. 5).

Chloride Concentrations in Highway Runoff and Road Surface Conditions Under Different Deicing Salt Application Scenarios

Comparisons of pavement surface conditions for different deicing salt application scenarios were limited to data collected during winter 2022, because gaps in coverage during winter 2021 at the Treatment A and B stations hindered comparisons across treatments (fig. 5). The total percentage of pavement conditions as snow or ice during the winter 2022 season was lower at the Treatment B station (10.4 percent) than at the control or Treatment A stations (14.2 and 14.8 percent, respectively). Although nearly three times less chloride was estimated as applied to the Treatment A highway reach in 2022 (fig. 6; table 6), the percentage of pavement conditions as snow or ice at this station was only about 4 percent greater than that of the Treatment B station and within 1 percent of the control station during winter 2022.

Total chloride applied as deicing salt was also different across treatments. Total chloride application by weight was greatest in the Treatment B highway reach, despite an approximately 10 percent lower application rate than the control reach. This may have occurred because deicing salt was applied more frequently during winter snow and ice events on highways in the Treatment B section than other sections (Fisher and others, 2025) and the total lane mileage was greater in this highway reach than in other reaches (table 1). Furthermore, the application rates may have been increased on steep grades, winding roadways, or in problem areas where snow or shading affected highway conditions (Joseph B. Thompson, NYSDOT, written comm. 2023) because pavement surface temperatures tended to be lower in the Treatment B section than in other treatment sites (fig. 6). Chloride applied for deicing was lowest in the Treatment A watershed, consistent with the application of a mixture of abrasives and deicing salts in this watershed (fig. 6).

More deicing materials were applied to all highway reaches in winter 2022, and higher annual flow-weighted chloride concentrations were observed in highway runoff in the control and Treatment B watersheds. Highway runoff water quality data were not available for the Treatment A station for part of winter 2021, because the paved asphalt waterway to the drop inlet was not sealed at this location until December 20, 2020, and highway runoff was minimal from December 20, 2020, until March 26, 2021. Thus, comparisons of highway runoff quality at the Treatment A highway monitoring station with highway runoff quality at other stations were not possible for winter 2021 data. Despite lower total chloride application at the Treatment A highway reach in 2022, median winter flow-weighted chloride concentrations in highway runoff were not substantially lower than concentrations for other treatments in 2022 (p > 0.05; fig. 5B). This can be attributed to high chloride concentrations that coincided with a greater number of runoff events at the Treatment A station compared with the Treatment B and control stations because no other sources of chloride were likely to contribute to the watersheds.

Flow-weighted chloride concentrations in highway runoff spanned several orders of magnitude across all treatment reaches (table 6), routinely exceeding the criterion established for the protection of aquatic habitats. The U.S. Environmental Protection Agency recommends a chronic criterion for aquatic life of a 4-day average chloride concentration of 230 mg/L, not to be exceeded more than once every 3 years (U.S. Environmental Protection Agency, 1988). The recommended acute criterion concentration for chloride is 860 mg/L, corresponding to a 1-hour average concentration with a recurrence interval of less than once every 3 years (U.S. Environmental Protection Agency, 1988). Although the highway runoff collected in this study may not directly constitute aquatic habitat for sensitive organisms for which the criteria were developed, the frequency and magnitude of exceedances of the criterion in this controlled experiment hold relevance to planning, management, and highway design considerations aimed at minimizing the deleterious effects of deicing to aquatic organisms.

Median flow-weighted chloride concentrations in highway runoff in this study (0.991 to 6.585 mg/L; table 5) were lower than median chloride concentrations of highway runoff collected at 11 highway-monitoring stations in Massachusetts from 2005 to 2007 (10 to 70 mg/L; Smith and Granato, 2010). Traffic volumes for highways in Massachusetts were generally higher than those in the Adirondack Park, and chloride concentrations in highway runoff increased with increasing traffic volume (Smith and Granato, 2010). Chloride concentrations in highway runoff were higher and more variable during periods of frequent deicing salt application, declined over time after seasonal applications ended, and did not return to baseline levels for several months in highway runoff in the control and Treatment B reaches (fig. 5). Higher and more variable specific conductance in urban streams receiving direct road runoff has been associated with temporal proximity to salt applications (Lakoba and others, 2020).

Discerning the effects of salt application on pavement conditions and highway runoff across treatment stations can be challenging because of the temporal and spatial variability of meteorological factors, highway orientation and slope, traffic conditions, and uncertain interactions among these and other variables.

Chloride Storage and Transport at the Watershed Scale

Chloride concentrations in surface water varied by season: higher concentrations of chloride were observed during summer months than in winter months, coincident with low flow conditions. Chloride concentrations were higher in surface water in subwatersheds downgradient from Route 73 and Route 86 than in upgradient subwatersheds. Average annual chloride concentrations in downgradient surface water were higher at the Treatment B site than at downgradient sites in the control or Treatment A watersheds (table 7). Differences in hydrogeology across watersheds may affect the residence time of chloride in these systems. Furthermore, the downgradient surface water station in the Treatment B watershed was a greater distance from the road than either the control or Treatment A downgradient stations (table 2), and monitoring at this station may have captured seasonal groundwater discharge of deicing-salt affected groundwater, whereas baseflow chemistry at downgradient surface water sites within several hundred feet of State highways is likely to be dominated by groundwater sourced from upgradient areas with lower chloride concentrations. Maximum chloride concentrations at downgradient surface water stations (table 7) were lower than those in surface waters downgradient from Interstate 95 in Connecticut (160 mg/L; Brown and others, 2011), because land use, traffic volumes, and deicing salt application rates were higher in the Connecticut study.

The annual mass of chloride imported and exported from the Treatment B watershed was estimated to show potential differences in chloride inputs and storage across water years. Chloride inputs included deicing salt application, atmospheric deposition, and background chloride contributions from upstream surface water sources.

Deicing salt application was the primary source of chloride annually, regardless of treatment. Deicing salt application is frequently identified as the primary source of salinity in watersheds in New York (Heisig, 2000; Kelly and others, 2007; Shaw and others, 2012; Gutchess and others, 2016). An annual chloride mass balance for the Treatment B watershed estimated that chloride from upstream surface waters constituted less than 2 percent of the annual chloride export from the watershed and that deicing salt application constituted 97 percent of the total chloride inputs to the downstream watershed (fig. 7). Kelly and others (2007) estimated that about 91 percent of the sodium chloride inputs to a rural stream in southeastern New York were from deicing application. Deicing salt application accounted for an estimated 67 to 99 percent of total chloride inputs to two Connecticut watersheds (Brown and others, 2011). More chloride was applied as deicing salt in winter 2022 (fig. 6), and surface water exported a higher mass of chloride in water year 2022 (fig. 7). More chloride was exported by surface water in the Treatment B watershed than annual estimates of chloride inputs during water years 2021 and 2022. The higher export of chloride in the Treatment B watershed indicates that additional sources of chloride are influencing surface water other than those accounted for as inputs in this mass balance exercise, such as the discharge of groundwater that has been affected by legacy deicing salt applications in previous years (Kelly and others, 2007; Rossi and others, 2022). Chloride mass balance for watersheds impacted by deicing practices in New York and Connecticut have resulted in net storage of chloride annually, but these watersheds received higher annual chloride inputs from deicing and were larger than the Treatment B watershed (Brown and others, 2011; Gutchess and others, 2016). Higher application rates in previous years may have resulted in chloride storage in the Treatment B watershed, which would account for the additional chloride exported during study years (fig. 7). Residence time estimates for chloride in other New York watersheds range from 20 to 70 years, although these watersheds received higher annual applications of chloride from deicing and sewage inputs than the Treatment B watershed (Shaw and others, 2012; Gutchess and others, 2016).

Chloride and Seasonal Saltwater Plumes in Groundwater

Chloride concentrations in groundwater varied seasonally and were consistently higher in monitoring wells downgradient from Routes 73 and 86 than those in upgradient wells (p < 0.005). Median chloride concentrations in downgradient monitoring wells were higher at the control site than at either treatment site (table 8). Higher chloride concentrations have been observed in wells downgradient of highways in New York (Kelly and others, 2018) and in proximity to roads in neighboring States (Harte and Trowbridge, 2010; Cassanelli and Robbins, 2013; Rhodes and Horton, 2015; Levitt and Larsen, 2020) and have been attributed to runoff from deicing material applications. The lack of alternate sources of salinity and low chloride concentrations in upgradient groundwater indicate that downgradient groundwater quality in each of the Treatment areas is affected by deicing practices (table 8).

Surface electromagnetic surveys conducted at the Treatment B groundwater site adjacent to Route 86 in June and October 2022 indicated higher apparent electrical conductivity downgradient from Route 86 than upgradient from the highway (fig. 8). In June 2022, electrical conductivity was highest across a narrow band oriented parallel to Route 86. In October 2022, electrical conductivity was highest primarily downgradient from Route 86, with dilution along plume margins. These patterns suggest that in June, high-salinity groundwater was constrained to the near-pavement region of the surveyed area, whereas in October, groundwater flow driven by spring and summer recharge led to the lateral transport and dispersal of a seasonal high-conductivity plume downgradient from Route 86. Temporal variability in downgradient groundwater chloride concentrations in the Treatment B area mirrored that of electrical conductivity downgradient from Route 86, with lower chloride occurring during meteorological fall 2022 compared to summer.