Monitoring and predicting the fecal indicator bacteria ...

Science of the Total Environment 550 (2016) 1171–1181

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Monitoring and predicting the fecal indicator bacteria concentrations from agricultural, mixed land use and urban stormwater runoff M.A. Paule-Mercado a, J.S. Ventura b, S.A. Memon c, D. Jahng a, J.-H. Kang d, C.-H. Lee a,⁎ a

Department of Environmental Engineering and Energy, Myongji University, 116 Myongji-ro, Cheoin-gu, Yongin-si, Gyeonggi-do 17058, Republic of Korea Department of Engineering Science, College of Engineering and Agro-Industrial Technology, University of the Philippines Los Banos, Los Banos, Laguna 4031, Philippines Institute of Environmental Engineering and Management, Mehran University of Engineering and Technology, Jamshoro, 76062, Sindh, Pakistan d Department of Civil and Environmental Engineering, Dongguk University-Seoul, Seoul 100–715, Republic of Korea b c

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Land use and anthropogenic activities influenced the FIB intra-event variability. • Urban runoff had the highest levels of fecal contamination. • Temperature, TSS and turbidity correlated significantly with FIB concentrations. • MLR identified significant environmental parameter affects on FIB concentrations. • The FIB concentrations were predicted to increase due to urbanization.

a r t i c l e

i n f o

Article history: Received 12 August 2015 Received in revised form 6 January 2016 Accepted 6 January 2016 Available online 16 February 2016 Editor: D. Barcelo Keywords: Fecal indicator bacteria Hydrometeorological Land use and land cover Multiple linear regression Stormwater quality

⁎ Corresponding author.

http://dx.doi.org/10.1016/j.scitotenv.2016.01.026 0048-9697/© 2016 Elsevier B.V. All rights reserved.

a b s t r a c t While the urban runoff are increasingly being studied as a source of fecal indicator bacteria (FIB), less is known about the occurrence of FIB in watershed with mixed land use and ongoing land use and land cover (LULC) change. In this study, Escherichia coli (EC) and fecal streptococcus (FS) were monitored from 2012 to 2013 in agricultural, mixed and urban LULC and analyzed according to the most probable number (MPN). Pearson correlation was used to determine the relationship between FIB and environmental parameters (physicochemical and hydrometeorological). Multiple linear regressions (MLR) were used to identify the significant parameters that affect the FIB concentrations and to predict the response of FIB in LULC change. Overall, the FIB concentrations were higher in urban LULC (EC = 3.33–7.39; FS = 3.30–7.36 log10 MPN/100 mL) possibly because of runoff from commercial market and 100% impervious cover (IC). Also, during early-summer season; this reflects a greater persistence and growth rate of FIB in a warmer environment. During intra-event, however, the FIB concentrations varied according to site condition. Anthropogenic activities and IC influenced the correlation between the FIB concentrations and environmental parameters. Stormwater temperature (TEMP), turbidity, and TSS positively correlated with the FIB concentrations (p N 0.01), since IC increased, implying an accumulation of bacterial sources in urban activities. TEMP, BOD5, turbidity, TSS, and antecedent dry days (ADD) were the most significant explanatory variables for FIB as determined in MLR, possibly because they promoted the FIB growth and survival.

1172

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

The model confirmed the FIB concentrations: EC (R2 = 0.71–0.85; NSE = 0.72–0.86) and FS (R2 = 0.65–0.83; NSE = 0.66–0.84) are predicted to increase due to urbanization. Therefore, these findings will help in stormwater monitoring strategies, designing the best management practice for FIB removal and as input data for stormwater models. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Stormwater runoff picks up biological and chemical pollutants that have accumulated on the impervious or pervious surface during dry days and transports it to the downstream areas. In South Korea, this runoff is the significant cause of water quality impairment and degradation (Memon, et al., 2013). The forms and concentrations of pollutants in stormwater runoff are related to land use and land cover (LULC). For instance, runoff from agricultural, mixed, construction sites and urban land use contained high fecal indicator bacteria (FIB) concentration (Paule, et al., 2015). Most FIB such as Escherichia coli (EC) and fecal streptococci (FS) are use as a surrogate to indicate the presence of fecal matter (Hathaway, et al., 2015). The US Environmental Protection Agency (US EPA) recommends EC and enterococci as the best indicator of health risk for marine and fresh recreational waters (US EPA, 1985 and 1986). The US EPA wanted to exclude streptococci group to be more specific with human fecal contamination. However, the conventional enterococci analysis is sometimes flawed because Streptococcus bovis and Enterococcus faecalis are found in human feces and non-human contamination (Manero and Blanch, 1999), respectively. Moreover, some enterococci such as Enterococcus mundtii and Enterococcus casseliflavus are also found in natural environment (Byappanahalli, et al., 2012), thus, the specificity of enterococci analysis is still subjected to investigation. This study implemented EC and FS measurements to measure FIB in different environments. FS is commonly found in human and other warm-blooded animal digestive system which favors contamination detection of different priority areas included in our study. These bacteria are introduced to the environment in different ways, including, but not limited to, domestic and wildlife animals, untreated sewage discharge, illicit connections between stormwater and sewage systems, agricultural runoff, LULC change, sediments and plants (Cho et al., 2010; Converse et al., 2011; Paule, et al., 2015). FIB survival are affected by complex interactions among environmental parameters — catchment area, hydrometeorological, physicochemical, spatial patterns and LULC management, sedimentation and resuspension, soil types and vegetation (Ge and Frick, 2007; Hathaway et al., 2010; McCarthy, et al., 2012). Numerous studies were conducted to determine the relationship between FIB concentrations and environmental parameters. Nevertheless, these studies were conducted in coastal waters, rivers or streams (Ge and Frick, 2007; Hathaway et al., 2010; Kang et al., 2010; Gonzalez et al., 2012; Gonzalez and Noble, 2014). Also, analyses were typically performed on single grab samples during high flow events. These studies are not directly related to stormwater quality monitoring. However, change of concentrations during a storm event may allow a statistical inference of the intra-event characteristics of FIB contamination. Further, other studies collected a large number of stormwater runoff from agriculture, forest and urban sites in each storm events. Conversely, these studies focused on seasonal and temperature variability of FIB concentrations (Selvakumar and Borst, 2006; Cho, et al., 2010; Converse, et al., 2011). Recent studies on FIB determination focused on the intra-event variability and first flush characteristics. For instance, Krometis et al. (2007) showed declining microbial concentrations in the latter part of the storm event, suggesting this to be an effect of the first flush. Tiefenthaler et al. (2011) proposed the peak flows in the hydrograph of urban stormwater runoff accounts for the highest FIB concentrations. Conversely, other studies focused on the influenced of environmental

parameters to FIB concentrations in stormwater runoff (Selvakumar and Borst, 2006; Hathaway et al., 2010). Therefore, the accurate models that predict the FIB concentrations in stormwater runoff should consider both influence of intra-event and inter-summer period and environmental parameters. Several studies used the multiple linear regressions (MLR) to identify the significant environmental parameters that influence FIB concentrations (Ge and Frick, 2007; Hathaway et al., 2010; Kang et al., 2010; Gonzalez et al., 2012; Gonzalez and Noble, 2014). MLR models are cost-effective tools to predict the behavior of FIB concentrations during a storm event under different LULC. However, as was the case with the previously cited studies, these studies conducted in coastal waters, rivers, or streams and samplings performed using single grab samples. The goal of this study was to examine the impacts of environmental parameters on FIB concentrations collected in agricultural, mixed and urban catchments. The specific objectives were to (1) contrast the FIB concentrations with respect to intra-event and inter-summer season; (2) determine the relationship between FIB concentrations and environmental parameters; and (3) develop regression models to predict the influence of the environmental parameters and LULC change on FIB concentrations. The following environmental parameters were used in this study: catchment area; hydrometeorological — antecedent dry days (ADD), runoff flow (FLOW), and runoff volume (RUNVOL); and physicochemical — pH, 5-day biological oxygen demand (BOD5), chemical oxygen demand (COD), chloride (Cl−), total nitrogen (TN), total phosphorus (TP), turbidity (TURB), total suspended solids (TSS) and stormwater temperature (TEMP). As parameters for FIB determination, E. coli (EC) and fecal streptococcus (FS) were used. These FIB parameters were selected because they are used mostly as indicators for the bacteriological safety of water and; their detection and quantification in the laboratory is easier and economical than those of pathogens (Cho et al., 2010; Rowny and Stewart, 2012). However, numerous limitations are associated with their application including short survival in water body, ability to multiply after releasing into water column and resistance to disinfection (Savichtcheva and Okabe, 2006).

2. Materials and methods 2.1. Study area Stormwater was collected at the outlets of three monitored catchment within Geum-Hak stream, in Yongin City, Gyeonggi Province, South Korea (Fig. 1). This stream is one of the tributaries of the Paldang reservoir, the major source of drinking water for the Seoul metropolitan area and nearby provinces (Paule, et al., 2014). Table 1 presents the general characteristics of the stormwater monitoring sites. Site 1 was categorized as “agriculture” because over 50% of the total area use for cultivation, and no livestock farms exist there. Although Site 2 dominated by a forest, it is categorized as “mixed land use”. The upstream region of this site has an ongoing LULC development into residential and commercial complexes whereas the downstream region is mainly an urban area. Site 3 was categorized as “urban” because 100% ground land use. These monitoring sites were selected based on the catchment area, imperviousness, and LULC. The study area has a separate sewer for more than 20 years, which is sufficient infiltration and exfiltration in the stormwater drainage pipes.

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

1173

Fig. 1. Location of study sites and the LULC as of 2013.

strategy was determined by the hydrodynamic characteristics during stormwater runoff and the site-specific characteristics (Table 2).

2.2. Sampling strategy Sixteen storm events from June 2012 to December 2013 were monitored, and a total of 254 grab samples (n = 3–7 samples at each site) were collected. Sampling was initiated when storms were forecasted to produce at least 4 mm of rainfall and with 72 h ADD. The shortest acceptable sampling duration for each event was based on the principle of sampling from the beginning of the runoff until water samples became visually clear (≤30 NTU). These sampling criteria were established to: ensure that there is an adequate flow would discharge; allow some build-up of pollutants during the dry weather period and; and ensure that the storm events would be a “representative” (e.g. typical for the area in terms of intensity, depth and duration). The stormwater runoff was measured and collected at the end of drainage channel of stormwater of each monitoring site (Fig. 1). Flow meter was installed in monitoring sites prior to the start of precipitation or stormwater runoff. In Site 1, current velocity meter (AEM1-D, Japan) was used. In Sites 2 and 3, however, automatic flow meter (PCM F NIVUS, Germany) was installed. Continuous flow measurements were performed and recorded using 1 min (automatic) and 10 min (manual) interval at the same location as stormwater samples were taken. To characterize the intra-event variability of the FIB concentrations, hydrograph stages (initial, peak and final) were collected. Sampling

2.3. Hydrometeorological and physicochemical data Rainfall data were measured through an automated rainfall gauge (HB-3207-09, Casella, UK) installed about 100 m from Site 2 (Fig. 1). The runoff volumes were calculated for each storm event using Eq. (1). The rainfall amount and the associated runoff volume for each event were used to calculate the volumetric runoff coefficients (Eq. (2)). Other meteorological variables (e.g. ADD, total rainfall and average rainfall intensity) were obtained from the Korean Meteorological Administration (http://web.kma.go.kr/eng/index.jsp), which have the nearest rain gauge station about 2,000 m from Site 2. RUNVOL ¼

F t Dt 60

ð1Þ

where RUNVOL is the runoff volume (m3); Ft is the sample flow during the sample collection (m3/h) and Dt is time duration during the sample collection. RC ¼

V PA

ð2Þ

Table 1 General characteristics of stormwater monitoring sites. Site ID

Area (m2)

Average slope (%)

Imperviousness (%)

LULC compositiona

Possible source of fecal contamination

Agriculture

Site 1

577,100

17.01

8

Domestic and wild animal feces, sewer and septic cross connection leakages and overflow

Mixed land use

Site 2

1,451,500

20.98

20

Urban

Site 3

41,200

1.41

100

Agriculture (56%) bare land (5%) forest (18%) grassland (13%) ground (8%) Agriculture 6%) bare land (30%) forest (35%) grassland (9%) ground (20%) Ground (100%)

a

Extracted from MOE and MAF 2010.

Domestic animal feces, sewer and septic cross connection leakages and overflow

1174

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

Table 2 Interval and number of samples collected in each monitoring sites. Hydrograph stage

Initial Peak Final

Site 1

Site 2

Site 3

Interval (min)

Number of samples

Interval (min)

Number of samples

Interval (min)

Number of samples

0–120 180–360 N360

1 2–3 1

0–30 120–240 or 300–360 N360

1 2–5 1

0–30 60–180 N360

1 2–4 1

where RC is the runoff coefficient, V is the total runoff volume (m3), P is the total rainfall depth (m), and A is the catchment area (m2). Measurements of pH, stormwater temperature (TEMP) and turbidity were conducted in situ with a calibrated multiprobe (Horiba U-50 Multi-Probe, Japan). TSS was analyzed using a vacuum filtration. COD was analyzed by the dichromate reduction method, and a Wagtech DRB200 apparatus was used for digestion. Analytical analyses for other water quality parameters, including BOD5, TN, TP, and Cl− were performed based on Standard Methods for Examination of Water and Wastewater (APHA 2005 and 2009).

2.4. Fecal indicator analyses Stormwater samples were collected in 4-L sterile polyethylene bottles at each site, stored at 4 °C, and brought to the laboratory not exceeding 6 h for immediate analysis. EC (APHA 9221F) and FS (ES 05706.1a) were determined using the American (APHA, 2005 and 2009) and Korean (Korea Ministry of Environment, 2011) standards for the examination of water. Samples for FIB concentration evaluation were analyzed using the 5-tube most probable number (MPN) method. In the presumptive tests, a series of sample dilutions were dispensed in sets of tubes containing lactose broth (BD Difco™, USA) and azide dextrose broth (BD Difco™, USA) to permit the growth of EC and FS, respectively. Production of gas, acid formation, or abundant growth in the test tubes after 24 to 48 h of incubation at 35 ± 0.5 °C constituted a positive presumptive reaction. For EC, the positive tubes were transferred using a loop into a tube containing EC-MUG medium (BD Difco™, USA) and then incubated 44.5 ± 0.2 °C for 24 ± 2 h. For the final confirmation of FS presence, streaking of subsamples from all presumptive positive tubes on Enterococcosel agar (BD Difco™, USA) was performed. To finally confirm the presence of EC, tubes were viewed under long-wavelength UV (365 nm) light for fluorescence. For FS final confirmation, brownish black colonies with brown halos could be observed after 24 h ± 2 h of incubation at 35 °C ± 0.5 °C.

2.6. Statistical analysis All parameters were log10-transformed to reduce skewness prior to any analysis. Box plots and pollutographs (in a selected month) were used to examine the intra-event variability of FIB concentrations. The intra-event FIB concentrations in each monitoring site were analyzed to determine the differences between early-summer (May to August) and late-summer (September to December) seasons with temperature range from 20.1–25.3 °C and 6.5–18.6 °C, respectively. A one-way ANOVA was performed to determine the significant differences of the FIB concentrations in each monitoring site. Pearson's correlation coefficient matrix was used to identify the relationship between measurement concentrations of FIB and the relationship between FIB and environmental parameters. Furthermore, an MLR was used to identify the significant environmental parameters that explained FIB concentration variation and to predict the impact of LULC and LULC change on FIB concentrations. The general equation of an MLR model is: h
i  n  E Ln y j ¼ α0  ∑i¼1 βij xij where E[Ln(yj)] represents the expected values of the log10-transformed FIB concentrations (MPN/100 mL) in an event j, α0 is the regression constant, βij is the regression coefficient of dependent variables – environmental parameters and % area of LULC – xi, and xij is the dependent variables i in the event j. Each model was limited to five dependent variables in order to prevent over-parameterization. Variance inflation factors (VIFs) were calculated for determining the likeliness of collinearity issues in the MLR. VIF values greater than 10 indicate that the collinearity among the predictor variables in the MLR model is strong enough to warrant corrective action (Gonzalez and Noble, 2014). Performance of the MLR model was evaluated using: coefficient determination (R2), standard error of estimate (SEE), and Nash– Sutcliffe efficiency coefficient (NSE). All statistical analyses were performed in SigmaPlot 12.3 (Systat Software, Inc., San Jose, CA) with p ≤ 0.05.

2.5. Spatial analysis

3. Results and discussion

ArcGIS 10 (Redlands, CA, USA) was used to delineate the boundaries of the monitored catchments and necessary geoprocessing such as LULC fractionation and average slope of the respective catchments. The 2010 LULC maps acquired from the Korean Ministry of Environment (MOE) and the Ministry of Agriculture and Forestry (MAF), Republic of Korea, in a raster format with 4 m resolution were used to measure the LULC composition of the study area. This LULC data set is the most recent available cover for the full extent of the study area. Monthly field visits were made to update and validate the LULC change in the study area. To reduce the number of independent variables, the original 23 LULC classifications were condensed into five broad land use types: agriculture, (paddy and dry fields), bare land (gravels, bare ground, bare rocks, soil digging and soil filling), forest (deciduous, mixed coniferous, and broadleaf forests), grassland, and ground (parking lot, residential, road, and commercial).

3.1. Rainfall and runoff characteristics Table 3 summarizes the hydrological characteristics during the monitoring period. All monitored events had minimum: 3 days of ADD and 4 mm of total rainfall, which means there was enough time to build-up of pollutants during dry weather periods and to produce an adequate flow discharge, respectively. The average total rainfall was 34.3 mm which represents at least 80% of the total number of storm events occurred in Korea per year. Site 2 produced higher total runoff volume than Sites 1 and 3. However, the higher average runoff coefficient that was found in Site 3 was calculated to be 0.85, which is consistent with the areas specially used for parking lot (0.84), commercial areas (0.83) and streets (0.85) (Kim, et al., 2007; Flint, et al., 2007; Caltrans, 2011). Higher runoff coefficients in this site are expected because this monitoring site is 100% impervious cover (IC) with limited infiltration.

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

1175

Table 3 Statistical summary of monitored storm events (2012 to 2013). Parameters

Unit

Site

ADD

Day

Total rainfall

mm

Average rainfall intensity

mm/h

Runoff duration

h

Total runoff volume

m3

Runoff depth

m

S1 S2 S3 S1 S2 S3 S1 S2 S3 S1 S2 S3 S1 S2 S3 S1 S2 S3 S1 S2 S3

Runoff coefficient

Minimum

Maximum

3

31

4

0.2

9 6.16 6.5 178.27 2,206.28 128.54 0.0003 0.0015 0.0031 0.08 0.38 0.78

Average

Standard deviation

7.9

24.6

97.5

34.3

7.9

15.8

3.4

2.23

15 13 12 18,467.87 123,123.49 3,615.30 0.0332 0.0848 0.0878 0.34 0.87 0.93

12.25 9.75 9.25 3,439.54 40,327.02 1,201.19 0.0062 0.0278 0.0292 0.18 0.81 0.85

3.20 2.22 1.99 18,547 50,927 1,957 0.0132 0.0451 0.0642 0.08 0.14 0.2

Site 1 (S1); Site 2 (S2); Site 3 (S3).

3.2. Land use and land cover In terms of LULC change, only in Site 2 had a rapid LULC change over the monitoring period. Fig. 2 shows the LULC composition before ground-work (2010) and during ground-work (between 2012 and 2013). In 2010, agriculture, bare land, forest, grassland and ground accounted for 17.16, 4.79, 41.86, 12.45 and 23.74% of the total area, respectively. Between 2012 and 2013, however, these LULC changed rapidly due to the ongoing LULC development into residential and commercial complexes. Results revealed the bare land increased exponentially (578.38%) whereas agriculture (78.82%), forest (12.24%), grassland (16.31%) and ground (30.91%) reduced in spatial extent. 3.3. FIB measurement Overall, the FIB concentrations were significantly higher in Site 3 than in Sites 1 and 2 (Fig. 3). These biological contaminations caused by the FIB often result from several mechanisms such as runoff from IC, commercial market, sewer and septic overflows and feces of domestic animals. In addition, 15–30% of the significant bacteria were attached to sediment during storm events, which could lead to resuspension of bacteria (Sidhu et al., 2012). Further, in Site 2, flow is always present owing to forest runoff, did not significantly differ from Site 3 which showed the second highest FIB concentrations (Fig. 3). Illicit connection or leakages in wastewater system was observed during the monitoring period. The other possible sources of FIB contamination on this site was agricultural runoff, feces from domestic and wild animals, forest, IC, and the ongoing LULC development. Moreover, in Site 1 the flow is always present, similar in Site 2. This monitoring site, however, obtained the lowest FIB concentrations and with significantly different from other sites (ANOVA, p b 0.001) (Fig. 3). Intensive application of manure to cropland and forest runoff was observed in this monitoring site. The scope of this study did not include identification of the FIB sources; hence, microbial tracking was not performed. However, examining the anthropogenic activities in the study area can provide clues to these direct sources. 3.4. Intra-event variability Intra-event variability of the FIB concentrations differed across the monitoring sites (Fig. 3). In Site 1, FIB concentrations increased during

peak concentration then decreased as the storm progressed. The peak FIB concentrations had a significant correlation with stormwater runoff (r = 0.813, p = 0.024) than initial (r = 0.257, p = 0.031) and final (r = 0.385, p = 0.018). In contrast to other studies, the peak FIB concentrations (e.g. EC) may appear before the peak flow (Davies-Colley et al., 2008, Stott et al., 2011). In this study, farming practices and agricultural land saturation (56% of the total area) affects the correlation between stormwater runoff and peak FIB concentrations. In Korea, the agricultural land, specifically rice paddies receive an intensive application of fertilizers between May and November (Paule, et al., 2014), in which monitoring period occurred. Farmers keep rice paddies flooded after application of fertilizer to facilitate an increase in nutrient uptake by the rice plants. However, large amounts of precipitation and saturated agricultural land can cause the overland flow of fertilized paddies (Lee et al., 2010). Further, in Site 2, the FIB concentrations remained constant during the collection of the initial and final samples and increased during peak flow (Fig. 3). Peak FIB concentrations had a significant positive correlation (r = 0.891, p = 0.041). These results support previous studies that the initial FIB concentrations were due to runoff from surrounding urban land use impact, then the peak concentration was caused by overflowing sewers, illicit connections or leaking of sanitary sewers (McKergow and Davies-Colley, 2010; McCarthy, et al., 2012; Chow, et al., 2013; Liang, et al., 2013). Moreover, Site 3 showed the highest FIB concentrations among sites (Fig. 3). The FIB concentrations decreased as the storm progressed, which is consistent with the first flush phenomenon (higher concentration of FIB at the beginning of runoff) (Stumpf, et al., 2010). Considering that Site 3 is a small piece of urban land, the bacterial buildup on the IC would be minimal. The results of this study are consistent with other studies that, greater concentrations of FIB in developed and more impervious watersheds (Schoonover and Lockaby, 2006; Mallin, et al., 2000). FIB concentrations varied with time as a function of flow over the course of storm events (Fig. 4). Tiefenthaler et al. (2011) reported the FIB concentrations in urban or nonurban watershed increased during peak flow and then decreased to base levels within 2 h. However, in this study, for instance, the event concentration was collected on July 22, 2013. The peaks FIB concentrations in Site 1 occurred during peak flow subsequently decreased gradually after 4 h. In Site 2, the initial peak FIB concentrations occurred within 2 h start of stormwater runoff and then decreased gradually after 4 h. The initial peak concentration was high due to impact from surrounding urban land use, whereas the concentration from the construction site resulted in later peak

1176

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

Fig. 2. LULC patterns in mixed land use.

concentration. It was observed that the pervious areas discharged impact occurred after a few hours of runoff. Therefore, sampling was continued until the runoff becomes clear (≤30 NTU) with a decrease in flow

rate. In Site 3, the first flush was observed, where the FIB concentrations were high during the initial 3 h of the stormwater runoff then decreased gradually. Therefore, initial stormwater runoff sampling is important to

Fig. 3. Summary of Escherichia coli (A) and Fecal streptococcus (B) (log10 MPN/100 mL) collected from stormwater monitoring sites. The dashed line represents the Korean recreational water quality standard for fecal coliform: 2.3 log10 MPN/100 mL (or 200 MPN/100 mL).

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

Fig. 4. FIB intra-event concentrations (log10 MPN/100 mL) during July 22, 2013 sampling collected from stormwater monitoring sites. The dashed line represents the Korean recreational water quality standard for fecal coliform.

capture and hence time-weighted sampling is considered. The results indicate that the FIB concentrations varied depending on the characteristics of the stormwater runoff and site-specific conditions. 3.5. Inter-summer variability The FIB concentrations varied depending on summer seasonal trend (Fig. 5). The FIB concentrations significantly correlated with stormwater temperature (TEMP) (r ≤ 0.85, p b 0.05). Overall, the FIB concentrations, ADD and precipitation data were higher in the early-summer season. Therefore, enhanced accumulation of FIB contamination in stormwater

1177

Fig. 5. FIB intra-summer season concentrations (log10 MPN/100 mL) of stormwater samples collected from monitoring sites.

during early-summer indicates its association with increased temperature, higher ADD and greater amounts of precipitation. Similar to the results of Selvakumar and Borst (2006), the results of the present study suggest that any design for FIB removal (e.g., structural best management practices) must be based on data collected during the warmer months since the FIB concentrations are higher under these conditions. 3.6. Correlation analysis Table 4 presents the results of correlation analyses between FIB concentrations and environmental parameters in all monitoring sites.

1178

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

Table 4 Correlation analysis between FIB concentrations and environmental parameters in monitoring sites. Indicator

EC

FS

Parameter

Area TEMP FLOW RUNVOL ADD pH TURB TSS BOD5 COD TN TP Cl− Area TEMP FLOW RUNVOL ADD pH TURB TSS BOD5 COD TN TP Cl−

Agricultural

Mixed

r

p-Value

r

p-Value

r

Urban p-Value

0.381 0.645 0.273 0.283 0.525 −0.301 0.724 0.698 0.328 0.318 −0.220 0.285 0.254 0.341 0.686 0.281 0.318 0.529 −0.330 0.643 0.758 0.248 0.310 −0.381 −0.561 0.208

0.049 0.025 0.036 0.015 0.029 0.013 0.040 0.034 0.014 0.016 0.028 0.025 0.016 0.050 0.019 0.018 0.021 0.024 0.017 0.041 0.024 0.023 0.020 0.028 0.018 0.024

0.434 0.723 0.243 0.316 0.585 −0.467 0.828 0.794 0.793 0.821 0.413 0.351 −0.325 0.498 0.768 0.371 0.421 0.551 −0.143 0.783 0.831 0.761 0.874 0.538 −0.544 −0.218

0.045 0.017 0.038 0.035 0.028 0.049 0.018 0.031 0.031 0.012 0.013 0.005 0.031 0.041 0.002 0.041 0.025 0.0174 0.033 0.001 0.046 0.040 0.018 0.018 0.030 0.014

0.838 0.847 0.301 0.328 0.548 −0.296 0.847 0.831 0.874 0.851 0.404 0.420 −0.281 0.870 0.852 0.418 0.349 0.517 −0.087 0.805 0.836 0.785 0.821 0.385 −0.521 −0.388

0.021 0.040 0.023 0.021 0.0125 0.041 0.019 0.026 0.021 0.024 0.028 0.020 0.025 0.014 0.027 0.017 0.034 0.027 0.045 0.011 0.035 0.021 0.019 0.013 0.031 0.015

Note: Catchment area (AREA, m2); stormwater temperature (TEMP, °C); flow (FLOW, m3/s); runoff volume (RUNVOL, m3); antecedent dry days (ADD, days); pH; turbidity (TURB, NTU); FIB (MPN/100 mL); stormwater quality parameters (mg/L);|r| ≥ 0.6 are values are in bold type; all values presented are significant at p b 0.05.

A significant correlation between FIB concentrations and catchment area is expected because of the greater opportunity for fecal contamination and hence heightened concentrations of FIB and markers in the outfall (Converse et al., 2011). However, in this study, Site 3 is a small urban area and has a more significant positive correlation with the FIB concentrations (r ≤ 0.87, p ≤ 0.05) than Sites 1 and 2 (r b 0.50, p b 0.05). Sources of FIB, the LULC composition, and soil conditions may intensify the correlation between the FIB concentrations and the catchment area. Leakages and overflows from sewers, septic systems, and wastewater treatment plants are common sources of FIB in urban runoff. Furthermore, Site 3 is a 100% IC and may be prone to FIB from runoff. Mallin et al. (2000) found the amount of watershed imperviousness would be the cause for the 95% variability in the average FIB abundance. However, coniferous and broadleaf forests (sandy loam) and rice paddies (silty-clay loam) are the main landscapes in either Site 1 or 2. According to the slope gradient, sandy-loam and silty-clay loam has moderate-to-rapid permeability and moderately slow permeability, respectively (NAAS, 2013). These conditions might result in an increased storm magnitude (intensity, duration) necessary to cause significant differences in FIB concentrations. In all monitoring sites, TEMP significantly correlated with FIB concentrations (r ≤ 0.847, p b 0.05). The high correlations could be due to the persistence of FIB growth in warmer TEMP, which provided an opportunity for higher growth and survival rate during storm events and summer season (McCarthy et al., 2012). Most FIBs are mesophiles (optimum growth between 20 °C and 45 °C). The warmer TEMP results to faster metabolism and earlier inactivation of bacteria. However, colder TEMP preserves the viability of bacteria by slowing down the metabolism (UWRRC, 2014). FLOW and RUNVOL have weak positive correlations with the FIB concentrations in all sites (r ≤ 0.421, p ≤ 0.041). This result hypothesized that an additional parameters (e.g. sediment or IC) needs to consider in FIB variability. For instance, Nagels et al. (2002) and Muirhead et al. (2004) separately monitored the FIB concentrations during storm

events. Their studies revealed that an increase of FIB concentrations was associated with fine particles. Thus, the rate and transport of FIB in water columns and sediments are a complex interaction of physical, chemical, and biological processes (Cho et al., 2010). In addition, the FIB concentration in stormwater runoff dramatically increased during rainfall events compared to baseline conditions. This increase was due to FIB contamination washed away from IC and terrestrially associated fecal contamination that scoured from the land and transported via stormwater to receiving waters (Stumpf et al., 2010). In general, FIB concentrations expected to increase with an increase of ADD. Because this condition may allow the accumulation of bacteria in the watershed before they wash out during storm event (Gonzalez et al., 2012). In this study, however, these parameters had weak positive correlation (r ≤ 0.585, p ≤ 0.029). Results suggest that bacterial concentrations are affected by other factors (e.g. TEMP and amount of precipitation). Further, equilibrium rapidly develops between the new fecal materials added to the landscape and the decay of older fecal material that reaches the landscape within a few days following a storm (Ackerman and Weisberg, 2003). The FIB concentrations positively correlated with turbidity (r b 0.850, p b 0.42) and TSS (r b 0.840, p b 0.05). Turbidity associated with TSS, since these pollutants are usually correlated. Results suggest that FIB are adsorbed onto resuspended particles, bound to particulate matter and transported to the receiving waters. Applications of organic fertilizer and soil tilling were the possible source of FIB and TSS in Site 1, respectively. Tiefenthaler et al. (2011) identified high FIB concentrations in agriculture were due to regular applications of animal-based fertilizers. Further, in Sites 2 and 3, the source of FIB could be due to natural (e.g., animals, vegetation, soils and sediments, storm drain biofilms, etc.), illicit sewer connections or leaks in sanitary sewer (Stumpf et al., 2010); and for TSS could be due to road construction, lot clearing, IC, and ongoing extensive LULC development (only in Site 2). In this study, however, no assessment of particle size distribution conducted over the entire monitoring period. This assessment may help to clarify the dynamics of bacteria-particle source associations. FIB concentrations and organic matter had significantly correlated (BOD5 and COD) (r b 0.874, p b 0. 05) in Sites 2 and 3, which suggest the influence of inflow of sewage and wastewater effluent. However, weak positive is correlated in Site 1 (r b 0.330, p b 0.05), which suggest that the organic matter in stormwater is mainly due to application of fertilizers. Then, the FIB concentrations depend on physiochemical conditions and the nature of the organic matters (Kagalou et al., 2002). TN has a positive correlation with the FIB concentrations in Sites 2 and 3 (r b 0.540 p b 0.05) and weak negative correlations in Site 1 (r b 0.220, p b 0.030). Further, TP has a positive EC (r b 0.430, p b 0.026) and negative FS (r b − 0.520, p b 0.032) correlation in all monitoring sites, respectively. The positive correlation between TN and FIB concentrations in Sites 2 and 3 attributes to problems associated with the sewer. The correlation between TP and FIB concentrations possibly result of fertilizers' application to agricultural area and deposition of animal wastes. Chloride had a negative correlation with the FIB concentrations in Sites 2 and 3 (r ≤ − 0.281, p b 0.05) and a positive correlation in Site 1 (r b 0.255, p b 0.05). Chloride concentrations in stormwater associated with the percentage of IC and the quality of rock salt formation (Erickson et al., 2013). Further, the chloride concentrations are higher in wastewater than in raw water, and these high concentrations may be harmful to microorganisms (APHA, 2005 and 2009). However, comparing information in the literature, it suggested that chloride concentrations were insufficient to inhibit FIB growth such as EC (Bromley, 2003). 3.7. MLR analysis Based on MLR analysis, TEMP, BOD5, TURB, TSS, and ADD are the most influential independent variables for the FIB concentrations across the monitoring sites (Table 5). TEMP is an important factor related to

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

1179

Table 5 Multiple linear regression analysis based on FIB concentrations and various environmental parameters in monitoring sites.a FIB

Site

Explanatory variable

αo

b

p

VIF

R2

SEE

NSE

EC

Agricultural

TEMP ADD COD TN TURB TURB TEMP BOD5 ADD COD BOD5 COD TURB ADD TEMP TEMP TURB TP BOD5 ADD TEMP ADD TSS TN BOD5 pH TSS ADD BOD5 TEMP

5.12

0.031 0.038 0.031 0.026 0.405 0.297 0.563 0.171 0.123 0.109 0.321 0.309 0.112 0.472 0.142 0.309 0.051 −0.115 0.132 0.157 0.411 0.054 0.028 0.205 0.079 −0.083 0.088 0.049 0.351 0.035

0.023 0.012 0.045 0.024 0.039 0.036 0.024 0.015 0.019 0.029 0.037 0.032 0.035 0.015 0.042 0.022 0.018 0.032 0.016 0.042 0.008 0.006 0.012 0.035 0.045 0.047 0.016 0.009 0.027 0.034

1.324 1.077 1.193 1.137 1.205 1.207 1.138 1.572 1.285 1.373 1.366 1.259 1.294 1.284 1.137 1.261 1.170 1.514 1.513 1.100 1.16 1.161 1.233 1.322 1.546 1.088 1.245 1.212 1.242 1.088

0.83

0.79

0.82

0.84

0.79

0.83

0.83

0.73

0.82

0.87

0.79

0.86

0.91

0.89

0.90

0.85

0.82

0.83

Mixed

Urban

FS

Agricultural

Mixed

Urban

6.266

4.357

3.59

4.9

4.793

Note: Antecedent dry days (ADD, days); stormwater temperature (TEMP, °C); turbidity (TURB, NTU); FIB (MPN/100 mL) stormwater quality parameters (mg/L); αo (coefficient value for the intercept); b (coefficient value for the explanatory variables); and p (p-value for the explanatory variables). All values presented are log10 and significant at p b 0.05. Square of coefficient determination (R2); standard error of estimate (SEE); Nash–Sutcliffe model efficiency coefficient (NSE). a All equations are expressed as: log10EC = 5.12 + 0.031TEMP + 0.038ADD + 0.031COD + 0.026TN + 0.405TURB.

the FIB concentrations. On the other hand, BOD5 had a positive regression coefficient (b ≤ 0.351, p b 0.05). BOD5 had directly affected the dissolved oxygen (DO) in the stormwater runoff. A low DO levels may be an indication of a recent contamination event (such as untreated wastewater), which could elevate bacterial levels. Low DO levels in stormwater runoff may also be due to the consumption of oxygen with the bacterial decomposition of organic matter (Gonzalez et al., 2012). The effect of high BOD5 on the FIB concentrations was almost similar to that of a low DO levels: aquatic organisms become stressed and they subsequently suffocate and die (USEPA, 2010). FIB had positive correlations with TURB (b ≤ 0.405, p N 0.05) and TSS (b ≤ 0.088, p b 0.05). High TURB and TSS is a suitable environment for FIB survival because it reduces the exposure to sunlight and can carry nutrients to support microbial growth in the stormwater. Finally, ADD had a weak positive correlation (b ≤ 0.472, p b 0.05) to FIB concentrations. Results suggest the ADD had less impact on FIB concentrations. The most likely explanation of this difference that FIBs more liable than physicochemical constituents, as typically decay within a few days. The MLR models had a reasonably high degree of “goodness of fit” because the FIB concentrations had R2, and SEE values greater than 0.83 and 0.79, respectively (Table 5). Furthermore, the NSE value calculated for assessing the prediction performance of the MLR models. The calculated NSE values ranged from 0.82 to 0.86, which is the closer to 1 (Table 5 and Fig. 6). Therefore, the MLR models have good prediction performance. Three theoretical LULC scenarios developed based on current historical LULC (Table 6). Scenario 0 is the present LULC of each monitoring site. The ground cover ranged from 8 to 100% in the original LULC; catchments with proportions between 30 and 80% were not presented; Scenarios 1 and 2 filled the gap in IC data. Scenario 3 predicts the effects of further urban development (Site 1 and 2) and agriculture (Site 3).

FIB concentrations predicted to increase as urban development increases (Table 6). The agricultural area (Site 1) correlated with FIB thus the ground cover decreases. Lowest FIB concentrations found in agricultural area (except in scenario 2) and expected to remain so into the future. FIB concentrations in this area, however, will not decrease into the Korean recreational water quality standard for fecal coliform: 2.3 log10 MPN/100 mL. Moreover, the EC levels are higher in scenario 0 (Site 3), scenario 1 to 3 (Site 2); whereas the FS levels are higher in scenario 0 (Site 3), scenario 1 (Site 1), scenario 2 (Site 2) and scenario 3 (Site 3). These attributed to dominant ground cover (N30% of the total area). The regression analyses highlight the importance of including both agricultural and urban LULC in FIB models. The model will help the city stormwater planners to determine the impacts of LULC alteration on FIB concentrations. However, the limitation of this method lies in the presence of point-source-pollutants (PS) or combined sewer flows (CSF) wherein the risk for erroneous estimates of FIB concentrations becomes highly probable. 4. Conclusions The variability of FIB concentrations in agricultural, mixed land use, and urban catchment was examined. These monitoring sites varied in terms of catchment area, LULC and possible source of contaminants in order to analyze the intra-event and inter-summer variability, their correlations, as well as application of MLR to the evaluation of effects of environmental parameters and LULC on FIB concentrations. The following conclusions were drawn from the obtained findings: • The highest FIB concentrations were observed in the following site order: urban N mixed catchment N agricultural. The EC ranged from

1180

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

• The intra-event variability and strength of the first flush of FIB concentrations differ across the monitoring sites. Those influenced by catchment size, farming practices, agricultural runoff saturation, overflowing sewers, illicit connections or leaking of sanitary sewers, ongoing LULC development and bacterial buildup on the IC; • FIB concentrations had a significant variation between early- and latesummer seasons. Higher FIB concentrations in the early-summer due to increase of anthropogenic activities and TEMP. This condition provide a more persistent growth environment for FIB, therefore higher concentrations during subsequent storm events and; • Pearson correlation showed numerous significant relationships of the FIB concentrations with stormwater quality and hydrometeorological factors. While MLR model provide the relationships in detail. TEMP, BOD5, TURB, TSS, and ADD appeared to be significant variables in explaining the FIB concentration variability. The MLR models provided greater insights into potentially influential factors affecting the FIB concentrations. The models, therefore, have the potential to facilitate the establishment of water quality criteria that will help to prevent or minimize the adverse effects of contaminants on human and aquatic lives and likewise, identify the long-term changes in the water quality of a target watershed.

The results of this study provide a useful tool for predicting stormwater FIB concentrations of a given LULC and contribution of environmental parameters to FIB concentrations. Also, it can be use as an input data for stormwater models and in calculations of a total pollution load management system, similar to the calculation of the total maximum daily load, in a cost-effective manner. However, assessment of particle size distribution must be considered to clarify the dynamics of bacteria-particle source associations. Also, the microbial tracking should be performed in order to identify the FIB sources.

Acknowledgments Fig. 6. Measured and predicted values of FIB concentrations in stormwater samples collected from monitoring sites using MLR.

2.51 (Site 1) to 7.39 (Site 3) log10 MPN/100 mL, while the FS ranged from 2.48 (Site 1) to 7.21 (Site 3) log10 MPN/100 mL. The obtained MPN values exceeded the Korean recreational threshold limits (2.3 log10 MPN/100 mL), which means the stormwater can be a substantial source of NPS in downstream;

This research was supported by the Korea Environmental Technology and Industrial Institute, Next Generation Eco-Innovation Project (No. 413-111-003). The authors also would like to thank the following students for assisting with sample collection and analyses during storm events: Myongji University — Sunhae Kang, Bum-Yeon Lee, Shin-Jeong Park, Chinzorig Sukhbaatar and Raja Umer; Dongkuk University — Lady Shanee Alfonso-Tarun, Dong Hoon Lee, Min Yong Lee and Krish Madarang.

Table 6 Predictive log10 FIB concentrations of three theoretical LULC scenarios based on current historical LULC trends.a Scenario

0

1

2

3

Site

1 2 3 1 2 3 1 2 8 1 2 3

Percentage land use and land cover

EC

FS

AGRI

BARE

FOR

GRASS

GRND

MPN

R2

NSE

MPN

R2

NSE

56 6 0 35 9 20 16 10 10 10 15 85

5 30 0 8 20 10 5 5 5 2 12 5

18 35 0 14 35 15 16 5 10 5 18 3

13 9 0 13 6 25 13 10 10 13 5 2

8 20 100 30 30 30 50 70 65 70 50 5

3.90 6.16 7.89 5.14 5.23 4.30 4.90 6.65 6.05 5.99 6.33 6.30

0.73 0.75 0.73 0.76 0.83 0.82 0.83 0.85 0.71 0.85 0.81 0.76

0.74 0.76 0.74 0.77 0.83 0.83 0.84 0.85 0.72 0.86 0.82 0.77

3.94 4.55 7.4 4.8 4.06 3.73 4.45 6.12 5.51 5.59 5.54 6.66

0.68 0.65 0.76 0.79 0.82 0.83 0.83 0.83 0.74 0.82 0.82 0.77

0.67 0.66 0.77 0.81 0.83 0.84 0.84 0.84 0.75 0.83 0.83 0.78

a All equations are expressed as:〖logEC =3.9+0.0209AGRI+0.158BARE-0.081FOR-0.063GRASS+0.0339GRND and logFS =3.78+0.0296AGRI+0.114BARE-0.085FOR0.0636GRASS+0.0362GRND; agriculture (AGRI); bare land (BARE); forest (FOR); grassland (GRAS); and ground (GRND); FIB are expressed in MPN/100 mL. All values presented are log10 and significant at p b 0.05.

M.A. Paule-Mercado et al. / Science of the Total Environment 550 (2016) 1171–1181

References Ackerman, D., Weisberg, S.B., 2003. Relationship between rainfall and beach bacterial concentrations on Santa Monica bay beaches. Water Health 1 (2), 85–89. APHA, 2005. Standard Methods for the Examination of Water and Wastewater. 21st ed. American Public Health Association, Washington DC. APHA, 2009. Standard Methods for the Examination of Water and Wastewater. Revised online ed. American Public Health Association, Washington, DC. Bromley, J., 2003. Evidence for the resuscitation of chloride-induced viable but nonculturable Escherichia coli in estuarine environments. Nat. Resour. 90. Byappanahalli, M.N., Nevers, M.B., Korajkic, A., Staley, Z.R., Hardwood, V.J., 2012. Enterococci in the environment. Microbiol. Mol. Biol. Rev. 76 (4), 685–706. Caltrans, 2011. Stormwater Quality Handbooks. California: CTSW-RT-11-255.08.01. Cho, K., Kang, J.-H., Lee, S., Park, Y., Kim, J.-W., Kim, J., 2010. Meteorological effects on the levels of fecal indicator bacteria in an urban stream: a modeling approach. Water Res. 44 (7), 2189–2202. Chow, M.F., Yusop, Z., Toriman, M.E., 2013. Level and transport pattern of faecal coliform bacteria from tropical urban catchments. Water Sci. Technol. 67 (8), 1822–1831. Converse, R.R., Piehler, M.F., Noble, R.T., 2011. Contrasts in concentrations and loads of conventional and alternative indicators of fecal contamination in coastal stormwater. Water Res. 45 (15), 5229–5240. Davies-Colley, R., Nagels, J., Lydiard, E., 2008. Stormflow-dominated loads of faecal pollution from an intensively dairy-farmed catchment. Water Sci. Technol. 57 (10), 1519–1523. Erickson, A.J., Weiss, P.T., Gulliver, J.S., 2013. Optimizing Stormwater Treatment Practices: A Handbook of Assessment and Maintenance. Springer Publishing, NY, USA. Flint, K.R., Davis, A.P., ASCE, F., 2007. Pollutant mass flushing characterization of highway stormwater runoff from an ultra-urban area. Environ. Eng. 626 (6), 616. Ge, Z., Frick, W.E., 2007. Some statistical issues related to multiple linear regression modeling of beach bacteria concentrations. Environ. Res. 103 (3), 358–364. Gonzalez, R.A., Noble, R.T., 2014. Comparisons of statistical models to predict fecal indicator bacteria concentrations enumerated by qPCR- and culture-based methods. Water Res. 48, 296–305. Gonzalez, R.A., Conn, K.E., Crosswell, J.R., Noble, R.T., 2012. Application of empirical predictive modeling using conventional and alternative fecal indicator bacteria in Eastern North Carolina waters. Water Res. 46 (18), 5871–5882. Hathaway, J.M., Hunt, W.F., ASCE, M., Simmons III, O.D., 2010. Statistical evaluation of factors affecting indicator bacteria in urban storm-water runoff. Environ. Eng. 136 (12), 1360–1368. Hathaway, J.M., Hunt, W.F., McCarthy, D.T., 2015. Variability of intra-event statistics for multiple fecal indicator bacteria in urban stormwater. Water Resour. Manag. 29 (10), 3635–3649. Kagalou, I., Tsimarakis, G., Bezirtzoglou, E., 2002. Inter-relationships between bacteriological and chemical variations in Lake Pamvotis—Greece. Microb. Ecol. Health Dis. 14 (1), 37–41. Kang, J.-H., Lee, S., Cho, K., Ki, S., 2010. Linking land-use type and stream water quality using spatial data of fecal indicator bacteria and heavy metals in the Yeongsan river basin. Water Res. 44 (14), 4143–4157. Kim, L.-H., Ko, S.-O., Jeong, S., Yoon, J., 2007. Characteristics of washed-off pollutants and dynamic EMCs in parking lots and bridges during a storm. Sci. Total Environ. 376 (1–3), 178–184. Korean Ministry of Environment, 2011. Korean standard methods for the examination of drinking water. Notice 2011–21 Ministry of Environment (in korean). Krometis, L.H., Characklis, G.W., Simmons III, O.D., Dilts, M.D., Likirdopulos, C.A., Sobsey, M.D., 2007. Intra-storm variability in microbial partitioning and microbial loading rates. Water Res. 41 (2), 506–516. Lee, J.Y., Yang, J.S., Kim, D.K., Han, M.Y., 2010. Relationship between land use and water quality in a small watershed in South Korea. Water Sci. Technol. 62 (11), 2607–2615. Liang, Z., He, Z., Zhou, X., Powel, C.A., Yang, Y., He, L.M., Stoffella, P.J., 2013. Impact of mixed land-use practices on the microbial water quality in a subtropical coastal watershed. Sci. Total Environ. 449, 426–433.

1181

Mallin, M.A., Williams, K.E., Esham, E.C., Lowe, R.P., 2000. Effect of human development on bacteriological water quality in coastal watersheds. Ecol. Appl. 10 (4), 1047–1056. Manero, A., Blanch, A.R., 1999. Identification of Enterococcus spp. with a biochemical key. Appl. Environ. Microbiol. 65 (1), 4425–4430. McCarthy, D.T., Hathaway, J.M., Hunt, W.F., Deletic, A., 2012. Intra-event variability of Escherichia coli and total suspended solids in urban stormwater runoff. Water Res. 46 (20), 6661–6670. McKergow, L.A., Davies-Colley, R.J., 2010. Stormflow dynamics and loads of Escherichia coli in a large mixed land use catchment. Hydrol. Process. 24 (3), 276–289. Memon, S.A., Paule, M.A., Park, S.-J., Lee, B.-Y., Kang, S., Umer, R., Lee, C.-H., 2013. Monitoring of land use change impact on stormwater runoff and pollutant loading estimation in Yongin watershed Korea. Desalin. Water Treat. 51 (19–21), 4088–4096. Muirhead, R.W., Davies-Colley, R.J., Donnison, A.M., Nagels, J.W., 2004. Faecal bacteria yields in artificial flood events: quantifying in-stream stores. Water Res. 38 (5), 1215–1224. Nagels, J.W., Davies-Colley, R.J., Donnison, A.M., Muirhead, R.W., 2002. Faecal contamination over flood events in a pastoral agricultural stream in New Zealand. Water Sci. Technol. 45 (12), 45–52. National Academy of Agriculture Science (NAAS), 2013. htt://soil.rda.go.kr. Paule, M.A., Memon, S.A., Lee, B.-Y., Umer, S.R., Lee, C.-H., 2014. Stormwater runoff quality in correlation to land use and land cover development in Yongin, South Korea. Water Sci. Technol. 70 (20), 218–225. Paule, M.A., Ventura, J.-R.S., Memon, S., Lee, B.-Y., Jahng, D., Kang, M.-J., Lee, C.-H., 2015. Fecal contamination in Yongin watershed: association to land use and land cover and stormwater quality. Desalin. Water Treat. 53 (11), 3026–3038. Rowny, J., Stewart, J., 2012. Characterization of nonpoint source microbial contamination in an urbanizing watershed serving as a municipal water supply. Water Res. 46 (18), 6143–6153. Savichtcheva, O., Okabe, S., 2006. Alternative indicators of fecal pollution: relations with pathogens and conventional indicators, current methodologies for direct pathogen monitoring and future application perspectives. Water Res. 40 (13), 2463–2476. Schoonover, J.E., Lockaby, B.G., 2006. Land cover impacts on stream nutrients and fecal coliform in the lower Piedmont of West Georgia. J. Hydrol. 331 (3–4), 371–382. Selvakumar, A., Borst, M., 2006. Variation of microorganism concentrations in urban stormwater runoff with land use and seasons. Water Health 4 (1), 109–124. Sidhu, J.P., Hodgers, L., Ahmed, W., Chong, M.N., Toze, S., 2012. Prevalence of human pathogens and indicators in stormwater runoff in Brisbane, Australia. Water Res. 46 (20), 6652–6660. Stott, R., Davies-Colley, R., Nagels, J., Donnison, A., Ross, C., Muirhead, R., 2011. Differential behaviour of Escherichia coli and Campylobacter spp. In a stream draining dairy pasture. J. Water Health 9 (1), 59–69. Stumpf, C.H., Piehler, M.F., Thompson, S., Noble, R.T., 2010. Loading of fecal indicator bacteria in North Carolina tidal creek headwaters: hydrographic patterns and terrestrial runoff relationships. Water Res. 44 (16), 4704–4715. Tiefenthaler, L., Stein, E.D., Schiff, K.C., 2011. Levels and patterns of fecal indicator bacteria in stormwater runoff from homogenous land use sites and urban watersheds. Water Health 9 (2), 279–290. USEPA, 1985. Test methods for Escherichia coli and enterococci in water by the membrane filter procedure (Method #1103.1). EPA 600/485076. U.S. Environmental Protection Agency, Environmental Monitoring and Support Laboratory, Cincinnati, OH. USEPA, 1986. Bacteriological ambient water quality criteria for marine and fresh recreational waters. EPA 440/584002. Environmental Protection Agency, Office of Research and Development, Cincinnati, OH, U.S. USEPA, 2010. Predictive Tools for Beach Notification. Review and Technical Protocol vol. I (http://water.epa.gov/scitech/). UWRRC, 2014. Pathogens in Urban Stormwater Systems. Urban Drainage and Flood Control District, Denver, CO.