The study of sodium and chloride in the environment has a long history with a particular focus on road salting in urban areas. In many studies, spatial scales are limited (e.g., city) and temporal measurements are coarse (e.g., monthly), with the result that our understanding of the hydrogeochemical dynamics is constrained. Through a unique set of spatial and temporal measurements from the State of Michigan we a) examine the spatial distribution of chloride across a broad geographic area, b) explore the temporal behavior of chloride and sodium over hydrologic events capturing snowmelt and rain through salting seasons, c) evaluate the use of chloride/sodium ratios as a tool for linking sources to concentrations, and d) develop a conceptual framework for processes responsible for their environmental concentrations. Results show 1) the short-term and local impact of urban areas on chloride concentrations is clearly delineated, 2) concentration and ratio variations over the hydrographs differ during salting and post-salting periods, 3) chloride/sodium ratios do not clearly indicate a halite source and can be very high (>5) and this is interpreted to be due to the different environmental behaviors of the two ions, and 4) during salting periods, chloride and sodium are quickly removed from the landscape during first flush and diluted as event water begins to dominate, but in post salting periods, only chloride is diluted. We also find evidence for upwelling of brine in some locations. These two solutes are easily measured indicators of human influences on water quality and it is recommended that they routine be included in water quality assessments. However, we suggest more research is necessary to better understand their cycling on shorter time scales and then how this knowledge can be used to inform our understanding of other chemical cycles in the environment.
Sodium (Na+) and chloride (Cl-) ions occupy a unique role in water quality studies because natural levels are frequently overwhelmed by halite that is intentionally released to the environment through anthropogenic uses (D’Itri, 1992). It is estimated that 50 million metric tons of halite are consumed each year in the U.S., with 30% being used for roadway icing (USGS, 2012). Five states (New York, Ohio, Michigan, Illinois, and Wisconsin) account for 75% of the halite applied, with the major portion of this amount being spread in urban areas (Sleeper, 2013). Chloride levels have been shown to have increased significantly in over 80 percent of urban streams in a U.S. Geological Survey study (USGS, 2014). Single-event application rates typically range from 200 to 400 lb/lane-mi, which can translate to annual applications of 50 tons per mile (Field et al., 1974). Food processing, agricultural, and industrial uses comprise 15, 5 and 4% of the halite used, respectively, with an unknown fraction of this being released directly to the environment in wastes from these operations. Water treatment, primarily home water softeners, constitute about 1% of the total, with nearly all of this being released on-site as regenerant solution. Domestic wastewater, regardless of whether it is treated on-site or collected and treated centrally, represents an indirect source, as most dietary salt is excreted. Considering all of these uses and release pathways, it can be said that with the exception of the small amount of salt that is incorporated into durable products and materials (for example, as chlorinated hydrocarbons), nearly all of the annual consumption of halite is released to the environment in either its crystalline or dissolved forms, and nearly all of the crystalline material dissolves as a result of the very high solubility of halite.
There are, of course, natural sources of Na+ and Cl-. In the mid-continent region of North America, these include atmospheric deposition, and interactions between water and soil, rocks, brines and salt deposits, while additional contributions from salt spray and seawater can be important in coastal regions. In the absence of brines, salt deposits and ocean influences, there are no major sources for Na+ and Cl-. Sodium in natural waters originates from the dissolution of feldspars or cation exchange reactions. The source for Cl- is as an impurity in minerals (e.g., biotite) or rocks (e.g., limestone). Thus, natural concentrations are typically below about 15 mg/L for each ion and are highly variable (Hem, 1970).
There is a large body of literature dating back to the 1970s and continuing to the present, reporting levels of Na+ and Cl- in surface and ground waters and discussing sources and transport processes. From these studies it has been established that road salt is the primary source of significant salinization in many locales (Demers et al., 1990; Gardner and Royer, 2010; Kelly et al., 2010, MacLeod et al., 2011; Dailey et al., 2014). Long-term studies suggest increases over several decades across broad regions including in remote locations with relatively low salt levels (Godwin et al., 2003; Kaushal et al., 2005; Kelly et al., 2008; Medalie, 2012; Perera et al., 2013). Techniques including spatial statistics using geographic information systems, multi-variate analysis, and mass-balance modeling have been used to link these broader trends to land-use in general, and specifically to roadways and urban areas (Mattson and Godfrey, 1994; Fitzpatrick et al., 2007; Novotny et al., 2009; Peters et al., 2009; Halstead et al., 2014).
Attempts to identify specific indicators for salt sources have proved useful, but not definitive. Ratios of Cl- to other halogens (iodide, fluoride, and bromide) are the most widely reported, but there are concerns about whether signals are locally specific, that they often overlap, and that there are measurement difficulties at low salinity levels (Howard and Beck, 1993; Davis et al., 1998; Panno et al., 2006; Dailey et al., 2014). Chloride isotope ratios have also been used (Dailey et al., 2014). Other dissolved constituents have been considered, but in most studies, the extent to which these indicate source, as opposed to other environmental processes, is unclear.
Early work suggested that Na+ to Cl- ratios might be used for source identification, but there now seems to be general agreement that using these ratios is complicated by differences in how these two ions interact with environmental solids (Neal and Kirchner, 2000). Chloride is usually considered non-reactive, or conservative, in the environment, as most soils have only limited anion exchange capacity at relevant soil pH values (Foth, 1999). However, cation exchange is significant for most solids, and although highly variable, sodium interactions result in changes in the ratios during transport that must be considered (Shanley, 1994). There is now clear evidence from soil column experiments, environmental data over several seasons, mathematical modeling, and a field “experiment” involving high-level salt contamination, that Na+ is sequestered by soils and sediments by cation exchange as dissolved concentrations increase, for example during spring snowmelt, resulting in initial Na/Cl ratios less than one. With continued inputs of salt as it is released from the landscape, the cation exchange capacity can be approached or exceeded, such that little or no Na+ removal takes place and the ratio decreases, approaching a value of one. As the salt is flushed out of the system, concentrations decrease, Na+ is released from the exchange sites, and ratios increase, often to levels well above one. Finally, as the Na+ sequestered on the solids is depleted, soils approach equilibrium with the non-salted water, and ratios return to one (Werner and diPretoro, 2006; Sun et al., 2012, 2014).
The scenario outlined above can be applied conceptually to an idealized one-dimensional system along a flow path, but it is complicated in most real systems by multiple flow paths that can change occur over seasons, or even over hydrologic events. For example, water reaching a stream at any point in time represents a mixture of water that has moved through these different flow paths and can have different solute concentrations and ratios. Thus, changes in pathways and concentration histories along pathways both affect the observations, and unless it can be argued that a particular pathway dominates at a particular time, it is difficult to separate the two effects. Because of this complexity, the competing roles of mixing of different water sources and Na+ exchange have not been fully investigated for short-term hydrologic events. Cherkauer (1975) compared Na+ and Cl- levels and ratios in an urban and a rural stream in Wisconsin for an autumn rainfall of 2.2 cm. Although the study only discusses the effects of dilution of the streams by the presumably low salinity precipitation, evidence of exchange processes can be found in the differing responses of the two watersheds. Both streams show molar ratios close to one prior to the precipitation event, but the ratio increases for the urban stream in a manner consistent with release of residual sequestered Na+, while the rural stream shows only a decrease attributed to dilution. In addition, two sequential peaks were found for ratio increases in the urban stream suggesting two primary flow paths. Similar findings were reported by Ostendorf (2013) and Cooper et al. (2014) in observing one or more first-flushes of conductivity, accompanied by either more specific measures (but not of Na+ and Cl- across the event) or modeling supporting an interpretation that road salt was the likely source and that exchange mechanisms were operative. Reisch and Toran (2014) took a similar approach for a snowmelt event, using conductivity data to suggest changes in flow paths over the course of the event, and how these relate to temperature and precipitation.
Hysteresis loop graphs are an approach that has been used to explore decoupling of variables that are mechanistically related but do not fully correlate in complex hydrologic systems. Andermann et al. (2012) investigated relationships between precipitation and discharge over annual cycles to quantify the effects of transient storage, while Williams et al. (1989) compared sediment concentration to discharge ratios on the rising and falling limbs of the hydrograph to explore the different patterns of sediment transport during events. Evans and Davies (1998) used the approach for dissolved solutes, linking observed concentration vs. discharge patterns to those predicted by a model based on three water sources whose contributions change over the event, and an assumption that the solutes are conservative. Aubert et al. (2013) employed annual hysteresis loops to investigate how several non-conservative water quality parameters (nitrate, sulfate, dissolved organic carbon, dissolved inorganic carbon) varied as a function of climatic conditions. Sun et al., (2014) used the technique of Aubert et al. (2013) to study annual changes in average monthly Na/Cl ratios. The data spanned a period from 1994 to 2012, but were averaged for periods (e.g., 1944 to 1960, 1961 to 1975). They report a change in hysteresis loop from a counter clock-wise pattern in the early data to a mix of counter clock-wide and clock-wide patterns in the later data. They attributed this change to increases in road salting and in impervious surfaces. To the best of our knowledge, the approach has not been used to investigate the roles of flow path and Na+ exchange for deicing salt in the environment over short (i.e., hydrograph) time scales, but we suggest that it may have potential.
The general hypothesis driving this study, and much of the body of past work, is that the use of halite is the dominant process adding contaminant Na+ and Cl- to most environments and that urban areas are “hot spots” because of road salting. We have attempted to be more specific and consider that both the spatial and the temporal distribution of these ions are affected by the time and location of sources, the dynamic behavior of transport processes, and Na+ exchange reactions. Therefore, the goals of this paper are to a) examine the spatial distribution of Na+ and Cl- across a broad geographic area, b) explore the temporal behavior of Na+ and Cl- over the primary hydrologic events that release these materials, spring precipitation and snowmelt, c) evaluate the use of Cl/Na molar ratios as a tool for linking sources to observed concentrations, and d) to develop a conceptual framework for of processes responsible for Na+ and Cl- levels found in the environment.
Materials and methods
Spatial patterns data
Two unique data sets were explored. One set resulted from data mining the Michigan Department of Environmental Quality (MDEQ) Waterchem and Wellogic databases (MDEQ, 2014). Concentrations for Na+ and Cl- in drinking water wells were extracted from this database for the whole State of Michigan using the Michigan Ground Water Management Tool developed for MDEQ by Michigan State University (MSU, 2014). This tool allowed not only for the extraction of concentration values from the databases, but also the spatial visualization of the data. The MDEQ-MSU database comprises over 550,000 samples, but does not differentiate the aquifers from which the water is being drawn from. The second database arises from the U.S. Geological Survey’s Regional Aquifer-System Analysis (RASA) program (Mandle, 1986; Dannemiller and Baltusis, 1990) with additions from Long et al. (1988). The RASA-MSU database comprises 700 samples from the three main aquifers in the Michigan Basin. These are the bedrock Marshall and Saginaw Formations (Mississippian and Pennsylvanian ages, respectively) (Fig. 1) and the overlying Glacial Drift (Bauer et al., 1996, Meissner et al., 1996, Wahrer et al, 1996). It should be noted that the dominate deicer used in the State of Michigan is halite (NaCl) (Rustem et al., 1993).
Temporal patterns data
The temporal aspects of the study involved the Red Cedar River, Michigan, U.S.A. (Fig. 2) which has a watershed area of approximately 461 mi2 (1194 km2) and a lengths of 51.1 mi (82.2 km). The watershed is a glaciated landscape characterized predominantly as medium-textured glacial till with abundant eskers. Two-thirds of the watershed is characterized as agriculture, wetlands, and grasslands. At its confluence with the Grand River, the land use is highly urbanized and includes the cities of Lansing and East Lansing, and the campus of Michigan State University (MSU). The population of East Lansing is approximately 100,000 during the time of the sampling, as MSU was in session. Upstream communities of Williamston, Okemos, Webberville and Fowlerville contribute another 30,000 inhabitants. Landuse/land cover is diverse consisting grasslands, wetlands, forests, agricultural, and developed residential, commercial, and industrial land. In portions of the watershed there is high potential for the influence of septic systems on river chemistry. However, much of the land is engineered to drain water from the soil by way of drain tiles, buried conduits that collect and convey groundwater from the soil to the river. Urban areas are largely drained by storm sewers. Spring snowmelt and summer rains are normally the dominant hydrologic events for this river. When precipitation occurs, much of it runs off impervious surfaces and goes directly into storm drains. Waters drained by both drain tiles and storm sewers are not treated and run through pipes to outfalls along the river. A series of maps showing data relevant to the watershed can be found at http://redcedarriver.weebly.com/map-gallery.html.
The data reported here originate from a collection of research and training activities over several decades including independent undergraduate and graduate student projects and student laboratory exercises in an environmental geochemistry class. Data sets from 1994 and 2013 have been selected because they were specifically designed to study chemical behavior during first flush. However, the two projects had different foci and therefore only measurements of river discharge, dissolved Cl-, and meteorological data are similar. Regardless of the task, all samples were collected using the same methodology in which river water was collected in precleaned and river rinsed polyethylene bottles, filtered through 0.45 µm Millipore filter, and stored at 4 °C. The sampling site was on the campus of MSU.
Discharge measurements were obtained from the USGS gaging station (Hydrologic Unit 04050004) located along the river on the MSU campus. This station has been collecting 15 minute discharge data since 1902. By using this station the chemical data reflect the upper 355 mi² of the Red Cedar watershed. The precipitation data were collected from Michigan State University Enviro-weather (Formerly Michigan Automated Weather Network (MAWN), which has a station on the campus of MSU).
Measurements for total dissolved solids were done at the time of collection using Myron L 4PII conductivity/TDS/resistivity meter. Na+ concentrations were measured immediately after collection using flame atomic emission spectroscopy. Because of its concentration stability in the samples, Cl- concentrations were measured in batches following collection.
In this paper we use the ratio Cl/Na rather than Na/Cl, which is frequently used. Our rational is that we found very high Cl/Na ratios that are not typically appreciated using Na/Cl ratios. Considering halite as the dominant contaminate source, these high ratios serve to emphasize the different control on the environmental behaviors of Cl- and Na+.
Results and discussion
Spatial patterns of Cl, Na, and Cl/Na
As mentioned, dissolved concentrations of Cl- in near surface waters can come from various sources, but in most environments, there are no natural mineral/rock sources for this ion. Thus, natural concentrations should be relatively low, a condition that would define natural near-surface waters in most of the State of Michigan. In these aquifers, 10 mg/L has been estimated as the natural concentration of Cl- (e.g., Wharer et al., 1996). Here the dissolution of limestone, which is prevalent in the bedrock and glacial drift might be considered one of the dominant sources. This number is not fixed, of course, and can vary across the state. It has been shown that in a similar setting in a watershed in north eastern Illinois that a clear link to the influence of halite near surface water chemistry could not be established until dissolved Cl- and Na+ concentrations were over 50 mg/L might be consider elevated (Long and Saleem, 1974). We considered for illustration purposes, 100 mg/L as representing Cl- concentrations that are clearly elevated from natural concentrations. The prevalence of Cl- in the near surface waters of the State of Michigan’s is illustrated in Figure 3 (a, b) which shows the distribution for concentrations above 10 mg/L and 100 mg/L, respectively from the MDEQ-MSU data set.
The population density in Michigan is highest in the southern half of the Lower Peninsula so it must be considered that interpretation of this distribution data is biased by the number of wells in a particular area. However, the spatial pattern of Cl- with concentrations above 100mg/L suggests two clear patterns. The first is the cluster of high concentrations in the near-surface waters of the Saginaw and Michigan lowland areas. Population density cannot account for these high concentrations as these areas are highly agricultural and Long et al. (1988) and Hoaglund et al. (2004) have shown that the high concentrations of Cl- in the Saginaw Lowlands are due to the upwelling of brine. Since the geohydrologic situation in the Michigan Lowlands is similar to that of the Saginaw Lowlands (Westjohn and Weaver, 1996; Westjohn et al., 1994), it has been hypothesized that a similar process is causing elevated Cl- concentrations there (Bauer et al., 1996; Fitzpatrick et al., 2007). The second pattern is the higher concentrations of Cl- in near-surface waters in areas such as southeastern Michigan, the Greater Lansing area, Traverse City, Marquette and along highways (e.g., I-94). The like cause of the second pattern is road salt, particularly in urban environments. It is clear from this figure that the impact of anthropogenic activities associated with the use of halite is widespread.
Theoretically, waters highly influenced by halite would have molar Cl/Na ratio near 1, but reported ratios are highly variable, and this has not proven to be a definitive indicator of source.
This is illustrated in the range of Cl/Na ratios in over 12,000 groundwater samples collected from drinking water wells in Allegan County, Michigan, shown in Figure 4, which is a sub set of the MDEQ-MSU database. Thus, these data might reflect the range of values to be expected in most near-surface salted environments. This region is characterized by a mix of agricultural, forested, low-density urban and suburban land use. Sources for Cl- and Na+ included road salting in urban areas and county and state roads, septic systems and other animal wastes, spray irrigation and possible brine upwelling as a portion of the county is in the Michigan Lowlands and overlies the Marshall Formation contains Na-Cl brines at depth (Bauer et al., 1996).
The highest Cl/Na ratios (approaching 15) are constrained to low Cl- and Na+ concentrations (Fig. 4a, b). At Cl- and Na+, concentrations greater than about 600 mg/L, ratios tend to cluster around or slightly above one. Ratios below one for the Cl- plots show a clustering bounded by what might appear to be a mixing curve between solutions containing low Cl/Na ratios and low Cl- concentrations with solutions containing higher ratios (above 1) and high Cl- concentrations. Such a trend is not as clear for Na+, however, there appears to be dominant cluster “peaking” around 1 at concentrations below 200 mg/L In addition, the Na+ plot shows a second dominant clustering bounded by a Cl/Na ratio of 0.5 and Na+ concentrations < 300 mg/L.
Cl- and Na+ are not correlated to total dissolved solids (TDS) at concentrations below 500 mg/L (Fig. 5a, b). Fewer samples are shown here because only a portion of the Allegan County data reported TDS values. At concentrations above 500 mg/L, there appears to be a rough correlation, but the extent of the scatter suggests that other ions contribute to TDS and that these are variable in a manner unrelated to Cl- and Na+. Ratios of Cl/Na show no trend with respect to TDS at any concentration increment below TDS values of 1500 mg/L (Fig 5c). The lack of correlation with TDS is typical for groundwater chemistry in the Lower Peninsula of Michigan (Fig. 5d).
These data show that there is no simple relationship between Cl-, Na+, Cl/Na ratios, and TDS. Cl/Na ratios below 1 might be expected in uncontaminated near-surface waters as there are no natural rock sources for Cl- in this area, while dissolved Na+ can be derived from weathering of rock and exchange reactions. Weathering processes are most likely limited to low Na+ concentrations, whereas exchange reactions involving release of halite-derived Na+ could produce much higher levels. Hypotheses for ratios above 1 include application of alternative deicers and selective retardation of Na+ during transport. Applications of CaCl2 and KCl for deicing could account for some high Cl/Na ratios but these deicers are not used widely so their effect on water chemistry would likely be minimal compared to the large amount of halite used on landscapes.
Upwelling of brine or its application for deicing might also explain the high Na+ and Cl- concentrations observed in some samples. Brine in the lower formations of the Michigan Basin was created from the evaporation of ancient seawater (e.g., Wilson and Long, 1993a, 1993b). A common technique to examine for the influence of brine with near-surface water is to plot an indicator chemical that reflects the degree of evaporation against solute concentrations (Long et al., 2009). The concentration changes of the solutes in seawater during evaporation are well known. A comparison is then made between the groundwater date and expected trends. Typically, Br- is used as the indicator of the degree of evaporation as its behavior is conservative through much of the evaporation sequence. When Br- data are not available, Cl- is often used. When sea water evaporates, Cl- and Na+ concentrations increase conservatively along the trajectory shown on Figure 6a (McCaffrey et al., 1987). When the brines mix with the near-surface freshwater, they are diluted down this trajectory. The groundwater from the three main aquifers in the Michigan Basin (Fig 6a) characterize the expected trajectories when near surface waters interact with brine from the evaporation of seawater (e.g., Wilson and Long, 1993a). At concentrations below 1000 mg/L, there is much scatter in the data with Cl/Na ratios varying from high to low. A larger proportion of the data plot above (Na+ rich) the potential mixing line than below (Cl- rich). The Na+ concentration appear to be converging to a concentration of about 300 mg/L at low Cl- concentrations, while there is no clear trend for Cl-. Above Cl- concentrations of 1,000 mg/L the data tightly cluster along the evaporation trajectory line. The data at higher Cl- concentrations on Figure 6a that plot off of the evaporation trajectory have been concentrated to a point that Cl- and Na+ are no longer conservative as halite begins to precipitate out (McCaffrey et al., 1987). These solutions are not influencing Cl- and Na+ concentrations in near-surface waters.
The trends in Cl- and Na+ concentrations for the Allegan County data are similar (Fig. 6b). The MDEQ-MSU data set at low concentrations involved only one significant figure causing the data patterns at low concentrations. Similar to Michigan Basin groundwater, the majority of the data plot above the potential mixing with Na+ converging near 300 mg/L. A difference from the Michigan Basin groundwater trends is that there are significantly more data that plot below the potential mixing line (Cl- rich) with a possible trend to 10 mg/L at low Na+ concentrations. The preponderance of data plot below a Cl- concentration of 1000 mg/L, but because of the scatter in Cl/Na ratios, a specific source is not apparent. However, it is clear that there is an influence of solutions with high concentrations of Na+ and Cl- and the trend towards seawater indicates the potential for near-surface water chemistry to be influenced by brines.
Temporal patterns of Cl-, Na+, and Cl/Na: 1994 data set
The 1994 data (Fig 7) were collected during the first 100 days of the year. The collection period was characterized as a strong El Niño and included four storm events. Only Cl- concentrations were measured. The low-flow or base-flow discharge appears to be about 100 ft3/sec (cfs) for the period. For the first 28 days, the discharge is relatively constant even though there are some small precipitation events, as these were snowfalls that did not immediately affect stream discharge. The first major high discharge event occurred around the 28th day. The delay from precipitation to the discharge peak is approximately 5 days, which is typical for the Red Cedar River at this location. There are four major peak discharge events during this time period. It can be noted that discharge does not return to the base flow value of 100 cfs, rather it increases over the course of the four events.
Chloride concentrations at low-flow discharge are about 34 mg/L. While the discharge is constant for the first 28 days, there are two peaks in the Cl- concentration that indicate additional inputs. The peaks around days 10 and 24 are from minor snowmelts, although not enough melting occurred in either case to measurably change discharge. The precipitation event on Day 27 produced a first-flush of Cl-, peaking at 240 mg/L on Day 28, well before peak discharge on Day 32. It can be noted that the Cl- peak is extremely sharp, suggesting that the precipitation immediately liberated a pool of Cl- from the landscape as a pulse that reached the stream quickly. The return in Cl- to levels at or below those observed at base-flow suggests that this mass was likely carried by surface runoff.
A similar temporal pattern is seen for a snowmelt event around day 45, with Cl- concentrations preceding the discharge peak by several days. It can also be seen, however, that the concentration peaks at a much lower level, about 100 mg/L, and the peak is much broader. This seems to suggest that the additional Cl- being liberated is either less accessible to the water that transports it, or the transport pathway is less direct. This second event produces a much larger change in discharge, which may explain why the Cl- levels drop to less than half the base-flow levels, and then recover linearly on the falling limb of the hydrograph. This observation is consistent with the effects of dilution by the precipitation and snowmelt that follows the first flush of Cl-. The first flushes of Cl- associated with events 3 and 4 again show a similar temporal pattern but the magnitude of the concentration peak is diminished. This suggests that after the first two events, there is little remaining road salt on the landscape. Decreases in Cl- concentrations at peak discharges still occur however. It also appears that the amount of the decrease in Cl- concentrations is related to the magnitude of the discharge peak and the trajectory of recovery related to shape of the falling limb curve. For example for event 2, the falling limb is rapid and relatively linear. The recovery of the Cl concentration increases linearly. As the shapes of the falling limb become nonlinear (e.g., events 3 and 4) the recovery trajectory of Cl is also nonlinear but still inversely reflects the trajectory of the falling limb.
The relationship between discharge and Cl- for the Red Cedar River in the 1994 data might be considered typical for rivers (e.g., Cooper et al., 2014). These patterns would include a delay in peak discharge after the start of an event (snowmelt, rain), and evidence of first flush of chemicals. However, not typically reported in the literature is the actual change in Cl- concentrations over a hydrograph and the concentration decrease following peak discharge. The patterns observed in the 1994 data are similar to data sets for the Red Cedar River collected for the same time period every year from 1999 to 2014.
These patterns can be related to changes in the pathways of water to the river and the relative contributions of water from these pathways to the stream hydrograph (Fig. 8a and b). We will ignore the effect of evaporation because it would not play a role during the course of a hydrologic event. Pathway I, direct precipitation input is not considered to be a major contributor to the mass of water or of chemical input into the river. Pathway II, overland flow or surface runoff caused by the precipitation event, is directly responsible for the concentration increase (i.e., first flush) of Cl- in the river. This water can profoundly change the chemistry of the river, but not necessarily the mass of water and discharge. Increasing discharge is slightly delayed and is the result of Pathway II and the increasing importance of pathway III (interflow). At the discharge peak, Cl- concentrations are low due to dilution and then increase as pathways IV (base or bankflow) and eventually V (deep base flow) become more dominant during the falling limb.
The hysteresis loop plots also shown in Figure 7 (b, c, d, e) provide a different way of looking at the changes across the four events. The plot for event I represents simple first-flush behavior, where we see a clockwise progression with concentration, increasing and peaking with the first increases in flow, decreasing at flow continues to increase until base-flow levels are reached, followed by a horizontal return toward the starting point as discharge returns to pre-event levels. The plot for event II also clearly shows first-flush behavior, but in addition, the effect of dilution is present with the concentrations at high flow reaching levels below those found pre-event, and concentration increasing as discharge subsides. In the plots for events III and IV, we see little evidence of first flush (note the different scale), but dilution behavior remains clear.
Temporal patterns of Cl, Na, and Cl/Na: 2013 data sets
Two series of data are available for 2013. Both data series are characterized by a period of a very weak La Niña. The first chemical data series (Fig. 9a) covers Julian days 7 to 89 (January 7 to March 29) with four larger events and three relatively smaller events observed. Values for Cl- and TDS are available for this data set. The second data series (Fig. 9b) covers a later time period of Julian days 97 to 163. (April 7 to June 11). There are two large events and three small events during this time period. The largest event exceeded the bankfull discharge level. Cl- and Na+ values and Cl/Na ratios are available for this data series. Low-flow discharge values are similar to those of 1994. Low-flow Cl- concentrations (∼70 mg/L) are higher than in 1994. To better explore the patterns Cl-, Na+, and Cl/Na, values for discharge, Cl-, Na+, TDS were normalized to their highest value during the measurement period. Normalization allows for better characterization of the relationships and rates of change in the relationships over time.
In the early data series (Fig. 9a), Cl- and TDS concentration changes do not always correlate. For the first hydrologic event, Cl- shows evidence of first flush while TDS does not. If fact, TDS decreases slightly as Cl- concentrations rise. Preceding the second event, which is the highest discharge for the period, both Cl- and TDS show first-flush behavior, as well as reaching their highest values for the time period. Both Cl- and TDS concentrations decrease similarly in magnitude to low values at peak discharge and then increase, however the trajectories of their recoveries differ. Total dissolved solids concentrations recover more quickly and the Cl- trajectory is similar to that observed in the 1994 data. For both Cl- and TDS, first flush clearly precedes events 3 and 6 and again their trajectories for recovery are not the same. The magnitude of the concentration increase for event 6 is second highest for these parameters with that for Cl- being greater than that for TDS. Although concentration changes for event 6 are relatively high for the period, the discharge peak for 6 is also one of the lowest. Concentration decreases at peak discharge are not observed. Event 7, the second largest event during this period, starts at Julian day 69 with a precipitation event. Cl- concentrations increase slightly before the precipitation event as result of snow melt. There is no change in TDS concentration. Both concentrations greatly decrease in a similar magnitude at peak discharge, but the slopes for recovery differ.
For the second data series (Fig. 9b), Cl- and Na+ concentration changes do not always correlate. Both Cl- and Na+ exhibit first-flush behaviors during hydrologic event 1.. Their slopes of increase, decrease, and recovery are similar. The pattern in their recovery trajectories are similar to those observed for Cl- in 1994 and the earlier measurement period in 2013. Prior to event 2, the highest discharge event in this period, both Cl- and Na+ show first flush curves with concentration peaks that are significantly less than in event 1. Both of their concentrations decrease at peak discharge but the decrease for Na+ is much less than that for Cl-. The shapes of decrease and recovery trajectories below the peak discharge differ from those observed in the 1994 and early 2013 data. After this event, changes in Cl- and Na+ concentrations generally parallel one another until event 5, where Cl- concentrations substantially decrease and recover while those of Na+ slightly increase and recover. Event 5 produced only a modest increase in discharge, and appeared in late May, long after the deicing season.
As might be expected, the Cl/Na ratios vary because of the differential changes in Cl- and Na+ concentrations. During the first flush for event 1, the ratio increases and then declines before the peak concentrations. Low ratios (near 1) occur before the concentration lows for Cl- and Na+ and then increase before a rapid rise during first flush for event 2. Ratios fall below 1 at peak discharge and then increase to relatively steady ratios around 1.6. Ratios decrease at peak discharge for event 5. It appears that event 1 releases much of the Cl- and Na+ on the landscape through pathway II, while we see significant dilution of Cl- with the very high discharge associated with event II. Na+ shows very little dilution behavior, perhaps because pathway III has become more important, and the release of previously sequestered Na+ maintains dissolved concentration levels. This is seen even more dramatically in event 4, perhaps because pathway IV is of greater importance and the pool of Na+ available for release has increased. It is interesting to note the Na+ levels appear to start decreasing at about Day 150, possibly suggesting that the supply of sequestered Na+ is starting to by depleted.
Despite the visual complexity of the integrated plots, the interpretation of the Cl- concentration hysteresis loops (Fig. 10) is essentially the same as offered for the 1994 data. We can clearly see the first flush events, the dilution that occurs at peak discharge, and the recovery to pre-event conditions. A hysteresis plot of the Cl/Na ratios (Fig. 11) for the second 2103 data set is also informative. The highest ratios occur for the first couple of first-flush events, and we only see ratios less than 1 under the highest discharge conditions. This indicates that over most of this period (the first half of the year), Na+ is being sequestered. We can speculate from the release that occurs under diluting conditions and the literature on annual cycles that events in the second half of the year are responsible for most of the Na+ release that eventually occurs.
Summary and conclusions
The results of the spatial analysis of Cl- and Na+ across the state of Michigan show that elevated Cl- and Na+ are widespread. The higher Cl- concentrations in urban areas and along interstates are consistent with halite as the dominant source. In addition, there is evidence for the influence of upwelling brine in some locations. Ratios of Cl/Na do not approach a diagnostic value (e.g. ∼1) until Cl- concentrations reach 1000 mg/L. Below 1000 mg/L the ratios are highly variable above and below 1. Very high Cl/Na ratios (> 5) are not observed at Cl- and Na+ concentrations above about 100 mg/L. Exchange reactions of Na+ are more dominant at low Na+ concentrations. Because of the dynamic nature of exchange processes, and the likelihood of mixed water sources/pathways, Cl/Na ratios are not all that useful for spatially distributed data with no temporal control. One might speculate however, that in this setting, any ratio above 1 at concentrations below 1000 mg/L can be attributed to the use of halite. Chloride and sodium concentrations are not related to TDS at TDS values < 600 mg/L, patterns in relationships differ for Cl- and Na+ at TDS concentrations > 600 mg/L and there is no relationship of Cl/Na to TDS at TDS values < 1500 mg/L.
The results of the temporal analysis show that most of the release of Na+ and Cl- from the application of road salt occurs in a first-flush when overland flow is dominant. Chloride, sodium, and total dissolved solids decrease to low values at peak discharge as interflow and bank flow becomes more dominant. Chloride concentrations recover from dilution along a trajectory that is inversely related to the trajectory of falling limb as base flow and bank flow become more dominant as “event” water is flushed out. Sodium concentration changes track that of chloride during first flush, decrease, and recovery phases. However, the magnitude of concentration changes is dampened compared to Cl- as a result of exchange reactions. As the spring season progresses (in the absence of road salting), the concentration changes of Na+ and Cl- can be decoupled as the pathways for Na+ and Cl- transport shift from being direct (e.g., overland flow) to increasingly less direct. Sodium concentrations become more controlled by water-rock interactions, with sequestration occurring during most of the spring season. Chloride behavior remains conservative: its concentration is influenced by the legacy of road salt in the system and it is more sensitive to changes in discharge than is sodium. Therefore, during storm events Na+ is sourced from water-rock reactions (e.g., desorption) while Cl-, with no direct contaminant source or rock source, is diluted. The differential behavior of Cl- and Na+ result in changes in Cl/Na ratios over a hydrograph. The pattern of the change in Cl/Na ratios is primarily driven by changes in Cl- concentrations. Similar to what was learned in the spatial analysis, Cl- concentrations does not always track TDS values, suggesting that one must use caution in using TDS or specific conductance values to infer Cl- cycling.
In terms of our goals, these results show that 1) the impact of urban areas on Cl- concentrations in the environment can be clearly delineated, 2) concentration and ratio changes over the hydrographs differ during salting and post-salting periods with Cl- concentrations the dominant factor causing ratio changes, 3) Cl/Na ratios do not clearly indicate a halite source (e.g., ratio of 1) and can be very high because the different environmental behaviors of the two solutes whose concentration trends are often decoupled, and 4) during salting periods, Cl- and Na+ are quickly flushed from the landscape during first flush and diluted as event water begins to dominate, while in post salting periods, only Cl- is diluted. Additional findings provide evidence for 1) the influence of upwelling brine on Cl- and Na+ concentrations and 2) possible concentrations limits on Cl- and Na+ in the environment. These limits might be on the order of 300 mg/L for Na+ at low Cl- concentrations and 10 mg/L for Cl- at low Na+ concentrations.
Dissolved Cl- and Na+ are two of the most easily measured chemical indicators of human influence on the environment. As demonstrated here, Cl- concentrations in near-surface waters definitely show the fingerprint of human activities on the landscape. We conclude, however, from our results and those of others, that the study of the general hypothesis that halite is the dominant process adding contaminant Na+ and Cl- to most environments and that urban areas are “hot spots” because of road salting, has been fully tested. However, we suggest that more measurements are necessary to better understand the Cl- and Na+ patterns observed here including the unique Cl/Na ratios in terms of their different environmental behaviors, dynamic nature of pathways and the combination of sources supplying each pathway. Finally, we suggest that this knowledge might be used to inform our understanding of the cycling of other chemicals in the environment.
Data accessibility statement
The data for the project can be obtained from:
U.S.G.S. RASA data: Dannemiller and Baltusis, 1990.
Red Cedar River event data: These data are part of ongoing student thesis research and are not currently available.
© 2015 Long et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
We would like to thank Tina Beals, Eunsang Lee, Wu Huiyun, Amira Oun, and the 2013 MSU Environmental Geochemistry class for assistance in sample collection and analysis. We very much appreciate the insightful discussions with Dr. Irene Xagoraraki about this research.
Portions of the work were funded by the U.S.G.S. Michigan RASA project, Michigan State University, MSU Water Initiative.
The authors have no competing interests.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.