Characteristics of chloride loading from urban and agricultural watersheds during storm and non-storm periods

The chloride ion (Cl ) can adversely affect an aquatic ecosystem, but it is not clear how Cl moves with runoff and how its transport processes are related to land uses and land cover. This study investigated how the loading characteristics of Cl vary depending on storm events and land cover in a temperate region. We monitored Cl concentrations in three study watersheds that have different compositions of urban and agricultural land uses. In addition, a Mass First Flush ratio (MFFn) was determined to quantify the effect of first flush on Cl loading. Overall, the observed concentrations and loadings in this study were found to be less than those reported in cold northern regions. The monitoring data showed that Cl concentrations and loads observed in an urban watershed were significantly larger than those of a rural watershed. The results suggest water management plans should focus on urbanized areas and their storm water to efficiently reduce chloride loading to downstream waterbodies. However, a further study is recommended to identify the sources and pathways of Cl loaded to waterbodies.


INTRODUCTION
and health implications of Cl À loading to waterbodies, it has not been a major focus of water quality management and planning. As a non-point source (NPS) pollutant, Cl À is difficult to quantify due to its temporal and spatial variations, and its loading characteristics are dependent on land uses (Deletic ).
Chloride ions are naturally present in the form of chlorides such as NaCl, KCl, and CaCl 2 (Rook ), and they are loaded to water bodies from sewage, plant wastewater, seawater intrusion, rainfall, inflow of manure, and excessive disinfectants. Herlihy et al. () concluded that Cl À concentrations could be used as a good surrogate indicator for general human disturbance in a watershed because the Cl À concentrations have various sources, even though studies on agricultural watersheds with large fertilizer inputs are few (Lax et al. ).
Urbanization represented by increased impervious areas reduces the opportunity for direct runoff to infiltrate into the ground before entering rivers and lakes; thus, Cl À concentrations tend to be higher in the streams of urban watersheds. Data collected in Rhode Island, USA, showed that the chloride level sometimes appears as high as 84 mg/L in developed residential and commercial areas while the concentration was around 17 mg/L before development (Hunt et al. ). Also, the widespread adoption of chloride as a deicer has rapidly increased chloride concentrations in water bodies, but the ecological implications of these changes have not yet been fully elucidated and does need long-term monitoring (Daley et al. ).
One important feature of urban runoff is that pollutant discharge in the initial period of a storm event is substantially higher than that in the later period; this is called the First

Monitoring methods
Cl À concentrations were monitored at the outlets of three watersheds: WJ (Woljeong), JS (Jangsu), and PYJ (Pungyeongjeongchen) located in Gwangju city, Korea. WJ and JS are nested in PYJ, which drains an area of 68.9 km 2 .
The study watersheds have distinctive compositions between urban and agricultural land uses (Figure 1). WJ is mostly covered by agricultural areas (62.2%, mostly rice paddy fields) with small urban land uses (6.2%), while JS is relatively highly urbanized with areas covered by a combined sewage system (36.0%) and agricultural areas (28.8%). Overall, PYJ consists of agricultural land uses (46.9%) and urban area (25.7%) ( Table 1). The study sites are located around 35 N and the average annual rainfall at the study sites is 1,391 mm, and the average annual maximum and minimum temperatures are 29.3 C and 1.9 C, respectively. The annual rainfall is concentrated during summertime since the study sites belong to a monsoon area. A combined system is installed in the urban area. the minimum dry period length between two independent events based on the statistical characteristics of historical rainfall events. For instance, the IETD of 17 hours was selected for rainfall records obtained from the weather station (Gwangju) that covers the study watersheds. The period between two consecutive storm events was defined as the non-storm period.
A water pressure gauge (OTT, Germany) was used to measure the depth of flow, and the river cross section was surveyed to calculate flow discharges from measured water depths using a rating curve. The flow discharge data of PYJ were collected from a monitoring station managed by the Korea Ministry of Land, Infrastructure, and Transport.
Water was sampled using an automatic sampler (ISCO 1570, USA) every 1 hour in the first 24 hours of a storm event, and then the sampling interval was increased to 6 hours until 48 hours from the beginning of the event according to the monitoring guideline of the MOE ().
This study was able to capture a total 11 storm events during the monitoring period. Grab sampling (32 times) was conducted weekly or bi-weekly during the non-storm

Data analysis
The Event Mean Concentration (EMC) represents a flowweighted average concentration, which is computed as the total pollutant mass divided by the total runoff volume for the duration of an event (Ballo et al. ). In this study the EMC was used as an index to characterize the water quality of runoff of a storm event (Novotny & Olem ). MFFn which quantifies the FFe by calculating the size of first flush intensity and the accumulated load against the cumulative runoff during a rainfall event was calculated by Equation (1). The 'n' value represents the percentage amount of runoff volume that carries the pollutant of interest. For instance, the MMF 10 of Cl À means the amount of Cl À load transported by the first 10% of runoff in a rainfall event.
MFFn ¼ where C(t) and Q(t) are the pollutant concentration and the runoff volume as a function of time, respectively, and n is the percentage of the runoff volume (ranging from 0% to 100%).
M and V are the total mass of the emitted pollutant and total runoff volume, respectively.
The annual Cl À load by the direct runoff of storm events was estimated by multiplying the average direct runoff load per unit rainfall depth and the 10-year average annual rainfall depth as Equation (2), (MOE ).
where N is the number of storm events. Annual export of Cl À was estimated by adding the load of base flow and direct runoff load. The load of base flow was determined by multiplying flow rate of the base flow and average concentration observed during the non-storm period.
The Eckhardt filter (Eckhardt ) was applied to separate direct runoff and baseflow from the streamflow of individual storm events by Equation (3).
where b is baseflow (m 3 /s), k is the time step (min), y is total streamflow (m 3 /s), BFI max is the baseflow index, and a is a recession constant.
The differences between the Cl À concentrations of water samples taken in the different study watersheds and periods (storm and non-storm periods) was tested at the significance level of 0.05. When the concentration measurements do not follow the normality assumption, the null hypothesis, there is no difference between the Cl À concentrations, as evaluated using the non-parametric Kruskal-Wallis test.

RESULTS AND DISCUSSION
Land use and Cl À concentration () showed that Cl À might not be an appropriate proxy for human influences within agricultural settings.  On the other hand, WJ did not show such a trend presumably because paddy fields have the capacity of storing storm water runoff that carries chloride. In some storm events, for the same reason, higher EMCs for some events were observed in JS (urbanized) than WJ (agricultural or rice paddy fields) (Table 3). In addition, PYJ produced lower chloride EMCs than did JS, as water drained from agricultural areas (WJ) might dilute urban-originated (JS) high Cl À concentrations in PYJ.        and rainfall events. The monitoring data showed that urban land uses produced a greater amount of Cl À load than did agricultural ones. The FFe was stronger in the urban watersheds than the agricultural one, which might be attributed to the fact that the urban land has more impervious areas and combined sewer systems. In addition, the storage or ponding capacity of rice paddy fields distributed over the agricultural watershed was believed to lessen the FFe and make the streamwater slowly respond to storm events. Such results suggest water management plans and implementation should focus on urbanized areas to effectively reduce Cl À loads to downstream waterbodies.

Seasonal effects on concentration of Cl
There must be many different factors that affect Cl À loading to waterbodies, and this study tried to find the statistical association between Cl À loads and the environmental characteristics (rainfall and land use). The observed Cl À concentrations and loadings from the study watersheds located in a temperate climate region were found to be much lower than those reported in cold northern areas. Such a finding implies that salt application as deicer could contribute to Cl À loadings. The hydrological connectivity of the locations to which Cl À is applied to downstream waterways and waterbodies would influence the timing and magnitude of Cl À retention and release.
Urbanization may increase the connectivity between the Cl À sources and downstream waterbodies through sewer drainage systems and thus accelerate Cl À loading. For improved watershed-scale Cl À management, it will be necessary to identify the source areas and major transport mechanisms of Cl À and handle Cl À loads at the sources or along the pathways. The use of environment-friendly deicing agents and advanced road washing machines can also help reduce chloride loadings. Furthermore, climate change is projected to increase the frequency and intensity of rainfall events, which will probably impact Cl À loading, as the FFe demonstrated. Long-term monitoring of streamflow and groundwater will help elucidate the urbanization and climate change impacts on Cl À loading and maintain the ecological health of watersheds.

ACKNOWLEDGEMENT
This study was financially supported by Chonnam National University, Republic of Korea (Grant number: 2019-3866).

DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.