Nutrient and Escherichia coli Attenuation in a Constructed Stormwater Wetland in the North Carolina Coastal Plain

Stormwater best management practices (BMPs) are installed to reduce the delivery of pollutants to surface waters. The objective of this study was to determine the stormwater NO3-N, PO4-P, and Escherichia coli (E. coli) reductions in a constructed wetland in Greenville, North Carolina. Water samples were collected at the inlet and outlet of the wetland before, during, and after 11 storms for NO3-N, PO4-P, and E. coli analysis. Treatment efficiencies for NO3-N (69%) and PO4-P (63%) exceeded the nutrient credit reductions assigned to stormwater wetlands (40% for both) in North Carolina. The E. coli (59%) and PO4-P (63%) concentration reductions in the wetland were similar to the reduction in specific conductivity (62%), possibly because of sedimentation in the wetland that reduced the suspended and dissolved solids with adsorbed E. coli and PO4-P. The relatively large size of the wetland (7% of drainage area), and below average rainfall likely contributed to the exceptional pollutant reduction efficencies.


Urban StormwaterManagement
Stormwater runoff from urban areas can lead to impairment of adjacent receiving waterways via rapid transport of pollutants such as nutrients, sediment, and bacteria that accumulate on impervious surfaces between storms (Davis et al., 2001;Tilley & Brown, 1998).Best management practices (BMPs) such as stormwater wetlands are created to reduce the transport of runoff and pollutants from urban areas to natural waters, thus protecting water quality.Stormwater wetlands are designed and constructed to mimic nutrient and pathogen treatment processes that occur in natural wetlands.Most stormwater wetlands consist of a forebay near the inlet, a shallow-water and shallow-land area, and another deep pool before the outlet (Hunt et al., 2007).The deep pool areas serve to dissipate the energy of influent, allowing sediment to settle and deep pools also provide open water aquatic habitat.Shallow water areas are vegetated channels that connect the deep pools, while shallow land areas are at higher elevations (than shallow water) and consist of different plant species (Hunt et al., 2007).Vegetation in the shallow water and shallow land areas uptake nutrients from runoff, and help slow the flow of influent, hence allowing sediment to settle (Gu & Dreschel, 2008).Constructed stormwater wetlands (CSWs) are often advantageous over other BMPs in that they tend to maintain a continuous flow, involving base flow and storm flow whereas other BMPs (retention ponds, etc.) may only be functioning as a treatment mechanism during and post storm events (Wadzuk et al., 2010).
Stormwater wetlands use physical, chemical and biological processes to treat stormwater.Stormwater wetlands are designed to remove pollutants from runoff via several mechanisms including: microbial breakdown and/or transformation of nutrients, plant uptake, settling, and adsorption (Johengen & LaRock, 1993: Martin & Reddy, 1997;USEPA, 2009).Stormwater wetlands may also reduce the concentration of pathogenic bacteria that can be harmful to public health (American Society of Meteorology, 1999).High concentrations of E.coli may indicate the presence of other disease-causing bacteria, and are a commonly used microbial indicator for non-saline surface waters (USEPA, 2008).High E.coli densities in recreational surface waters have been significantly correlated to human illness and thus threaten public health.Stormwater wetlands can reduce E. coli concentrations via adsorption to suspended sediment and later sedimentation, and/or inactivation from sunlight in open water areas.
The ability of wetlands to attenuate nutrients has been demonstrated for different pollutant sources including urban runoff, agricultural runoff, and wastewater treatment plant discharges (Reed at al., 1995;Raisen & Mitchell, 1995;Kadlec & Knight, 1996;Kovacic et al., 2000;Koskiaho & Puustinen, 2005).Wetland performance in treating stormwater is a function of numerous factors including, but not limited to hydraulic loading rate, detention time, storm intensity, runoff volume, wetland size (Carleton et al., 2001), season (Yousef et al., 1986;Hvited-Jacobsen et al., 1989), maintenance intensity (Hunt & Lord, 2007), and vegetative species and placement (Jenkins & Greenway, 2005).While most studies have shown that stormwater wetlands are effective at reducing nutrient and pathogen influent concentrations, the reported treatment efficiencies are often variable because of the many factors which influence pollutant treatment (Line et al., 2008;Hathaway et al., 2009;Wadzuk et al., 2010 ).

Eastern North Carolina Characteristics and Water Quality Issues
This study was conducted in the city of Greenville, which is in the central coastal plain of eastern North Carolina (NC).The geology of the area may be characterized as a gently southeastward dipping and thickening wedge of sediments resting on an underlying basement complex of Paleozoic age rocks (Lautier, 2001).The sediment is comprised of layers and lenses of sand, clay, silt, limestone, gravel, shell material and combinations of those materials (Lautier, 2001).Ground and surface water interact in a variety of physiographic and climatic landscapes creating the potential for exchange between the two components (Sophocleous, 2002).Therefore, groundwater quality and surface water quality are often linked.The mean monthly precipitation for Greenville ranges from 7.09 cm in November to 14.96 cm in August, and the mean annual precipitation is 106.93 cm (State Climate Office of North Carolina, 2014).Mean air temperatures are warmest in the summer months of July (26.5 °C) and August (25.6 °C) and lowest in the winter months of January (5.3 °C) and February (6.9 °C) (State Climate Office of North Carolina, 2014).
Groundwater and surface water resources in the coastal plain of eastern NC have been influenced by various sources of pollution including onsite wastewater systems, agriculture, and urban runoff (Hathaway et al., 2012;Humphrey et al., 2012).The state of NC enacted regulations (15A NCAC 2B.0258) to reduce the impact of stormwater runoff to nutrient sensitive waters from urbanizing areas such as Greenville, NC (North Carolina Department of Environment and Natural Resources, 2001).Developers must implement stormwater BMPs for construction projects in municipalities affected by stormwater regulations.Developers receive nutrient reduction credits for installing approved stormwater BMPs such as wetlands.However, since prior research has shown that pollutant treatment efficiencies of BMPs are variable, determining the treatment efficiency of common BMPs such as stormwater wetlands is important in ensuring that the regulations meet their intended purposes.The objective of this study was to determine if the NO 3 -N, PO 4 -P, and E. coli treatment efficiency of a constructed stormwater wetland was greater than or equal to the pollutant reduction credit established for the BMP by NC regulations.

Site Selection
The study site was located on the Bellamy student housing development, which was constructed in 2007 to serve students at East Carolina University in Greenville, NC.The Bellamy site consisted of 9.20 hectares (ha) of land.Upon completion of the first phase of development, the housing complex's total impervious area was 5.72 ha or 62% of the total property.Concurrent with the development at this site, NC general statutes (North Carolina Department of Environment and Natural Resources, 2001) required that a stormwater BMP be constructed to detain runoff and reduce nitrogen and phosphorus concentrations from the impervious surfaces before discharging to the existing stormwater conveyance system for the City of Greenville.Drainage from the site eventually discharges to the Tar River, which is classified as nutrient sensitive waters (North Carolina Department of Environment and Natural Resources, 2001).
The constructed stormwater wetland at the Bellamy occupies 0.68 ha or just over 7% of the watershed area (Figure 1).The wetland includes a forebay (influent pond), shallow water and shallow land area, an open water pool, and outlet.There were approximately 9200 plants were installed in the wetland including: Sweet Flag (Acorus calamus), Sawgrass (Cladium jamaicense), Arrow Arum (Peltandra virginica), Duck Potato (Sagittaria latifolia), and Softstem Bullrush (Schoenoplectus tabernaemontani).A wet hydro-seed mix was used to establish

Phosphate Treatment
The mean PO 4 -P inflow concentrations were higher than outflow concentrations for each period of sample collection (Figure 4).The before storm inflow PO 4 -P concentrations (0.19 mg/L) were elevated relative to outflow concentrations (0.07 mg/L), but the differences were not statistically significant (p = 0.08).Samples collected during storms for PO 4 -P analyses had mean inflow concentrations (0.21 mg/L) that were significantly (p = 0.01) higher than outflow concentrations (0.07 mg/L).The after-storm PO 4 -P concentrations were higher for inflow samples (0.18 mg/L) relative to outflow (0.07 mg/L), but the differences were not statistically significant (p = 0.11).When pooling all inflow phosphorus data (mean: 0.19 mg/L) and comparing to all outflow phosphorus data (mean: 0.07 mg/L) statistically significant differences (p = 0.0006) were observed.Mean inflow PO 4 -P concentrations were highest during storms, but outflow PO 4 -P concentrations were similar during each period of sample collection.

E. coli Analysis
The geometric means of E.coli concentrations for inflow samples were elevated relative to outflow samples for each period of sample collection, and for all of the pooled inflow and outflow data (Figure 5).Geometric mean inflow E. coli concentrations before (log 10 2.95 or 887 MPN/100 mL), during (log 10 3.02 or 1036 MPN/100 mL), and after (log 10 2.89 or 784 MPN100 mL) storms were higher compared to outflow geometric mean concentrations (before: log 10 2.57 or 374 MPN/100 mL; during: log 10 2.62 or 413 MPN/100 mL; after: log 10 2.52 or 320 MPN/100 mL).These differences were not statistically significant, as the p-values exceeded 0.05 for each comparison.However, when all the data was pooled, the geometric mean inflow E. coli concentration (log 10 2.95 or 896 MPN/100 mL) was significantly (p = 0.03) higher than the geometric mean outflow concentration (log 10 2.56 or 367 MPN/100 mL).The geometric mean E. coli inflow and outflow concentrations were highest during storm events.

Nutrient and Bacteria Treatment Efficiency
The constructed stormwater wetland at the Bellamy reduced mean inflow NO 3 -N concentrations by 69%, mean PO 4 -P concentrations by 63%, and geometric mean E. coli concentrations by 59% before discharge.The treatment efficiency established by the state of NC was a 40% reduction for nitrogen and phosphorus, thus the wetland was performing well relative to the regulatory expectations.Also the mean inflow specific conductivity of stormwater was reduced by 62% before reaching the outlet.

Specific Conductivity and Temperature
Similar to the nutrient and E. coli concentration trends, inflow specific conductivity was elevated relative to outflow specific conductivity for each month, and for the entire study (Figure 6).The mean monthly inflow specific conductivity was similar between July 2009 and March 2010 (range: 572 to 710 µS/cm), but lower at the start of the project in June 2009 (340 µS/cm).The mean monthly outflow specific conductivity was similar between June 2009 and January 2010 (range: 187 to 274 µS/cm), but increased during the winter months of February (329 µS/cm) and March (350 µS/cm).
Figure 7. Mean monthly inflow and outflow water temperature (°C), and pooled data (All).During summer months, outflow was warmer than inflow, and during winter months outflow was cooler than inflow

Rainfall Data
Precipitation during the study totaled 64.75 cm or 39% less than the long-term average (106.93cm) for the City of Greenville, NC (State Climate Office, 2014).November and December of 2010 were the only two months when the observed precipitation was greater than or equal to the long-term average (Figure 8).
Figure 8. Observed precipitation at the Airport in Greenville, NC during the study, as compared to the long-term monthly average (normal).The study period was relatively dry compared to the long-term average

Discussion
The mean NO 3 -N and PO 4 -P treatment efficiency for the wetland exceeded the nutrient treatment efficiency credit (40% for N and P) for each nutrient.The exceptional performance of the BMP may be due to the relatively large stormwater wetland area in relation to total drainage area.The Bellamy wetland encompassed 7% of the watershed, and thus was 2-4 percentage points larger than most constructed wetlands with similar sized drainage areas.Research has shown that wetlands with larger surface area to drainage area ratios can be more effective at reducing some nutrient concentrations than wetlands with smaller ratios (Line et al., 2008), possibly because the increased hydraulic residence time in larger wetlands allows more opportunity for pollutant removal via sedimentation, plant uptake and other processes (Carlton et al., 2001;Gu & Dreschel, 2008).The relatively low amount of precipitation may have also influenced the treatment efficiency, by providing more internal storage for stormwater and increased hydraulic retention time.If water levels in the wetland are low because of below average rainfall, when a storm event does occur, there may be little outflow because of the internal storage in the wetland, and evapotranspiration.Another potential contributing factor to the performance of the wetland was the maintenance intensity.A management company was contracted to maintain the stormwater wetland.Part of the maintenance agreement was the harvesting cattails to prevent the eventual loss of biodiversity.The harvesting of cattails was witnessed during data collection.Prior research has indicated that harvesting wetland plants may be beneficial to the treatment efficiency of the BMP (Lenhart et al., 2012).When plants are harvested and removed from the wetland, so are the nutrients within the plants and any sediment attached to the plants (Lenhart et al., 2012).The harvesting of wetland plants represented a mechanism for nutrient, sediment, and E. coli (sorbed to sediment or on harvested plants) reduction in the BMP.Also, by harvesting the wetland plants, there is a reduction in the potential mineralization and release of nutrients upon death of the plants (Lenhart et al., 2012).
Other maintenance practices such as trash removal, preventing blockage of the inlet and outlet, and monitoring and removing excess sediment in the forebay, can improve the treatment efficiency and aesthetics of the wetland (Hunt & Lord, 2007).The Bellamy wetland was routinely maintained during the period of this study, and because the wetland was relatively new, the accumulated sediment did not need to be removed.These factors including a relatively large wetland area, below average rainfall, routine harvesting of wetland plants and other maintenance, may have contributed to the high nutrient and E. coli treatment efficiency of the wetland.As the watershed continues to develop, and the wetland ages and accumulates sediment, the nutrient and E. coli treatment efficiencies may decline.
The mean stormwater specific conductivity reduction from the wetland inlet to outlet (62%) was similar to the mean PO 4 -P (63%) and geometric mean E. coli (59%) reductions, but lower than NO 3 -N reduction (69%).Specific conductivity is influenced by the dissolved salt and solid content of waters (Allhajar et al., 1990), therefore as suspended and dissolved sediment moves from the inlet through the wetland towards the outlet, processes which remove sediment from the water such as sedimentation can also reduce the specific conductivity of water.Sedimentation may be a dominant removal mechanism for solids in the wetland, and a dominant mechanism for the reduction in specific conductivity of inflow stormwater.Both PO 4 -P and E. coli can bind to solids, and thus be removed via sedimentation (Hunt, 1999;Davis et al., 2000;Characklis et al., 2005;Jamieson et al., 2005).The NO 3 -N reduction efficiencies were higher (than PO 4 -P, E. coli, and specific conductivity reductions) in the wetland possibly because of processes in addition to plant uptake and sedimentation acting to remove NO 3 -N, such as denitrification (Bourgues & Hart, 2007).Denitrification is the microbial conversion of NO 3 to N 2 gas.Denitrification occurs in anaerobic environments such as natural and constructed wetlands, where labile carbon and NO 3 are abundant (Mitsch & Gosselink, 2000).This additional removal pathway for nitrogen in wetland environments, may explain why the NO 3 -N reduction efficiencies were greater than PO 4 -P and E. coli reduction efficiencies.
The mean inflow water temperatures were cooler during the summer months and warmer during the winter months than outflow water temperatures, possibly because of frontal systems that generated rainfall (Brooks et al., 2003).During the winter, the mean inflow water temperatures were warmer than outflow water temperatures possibly because of warm fronts that contribute relatively warm runoff to the wetland, while during the summer, cold fronts delivered relatively cool runoff to the wetland.

Conclusions
The stormwater wetland at the Bellamy housing complex was more efficient at reducing NO 3 -N and PO 4 -P than the reduction credits assigned by the State of NC.The high treatment efficiency of the wetland was most likely influenced by the large wetland to drainage area percentage (7%), the relatively low precipitation, the maintenance intensity, and age of the wetland.These factors most likely resulted in the wetland's high storage capacity and treatment efficiency for urban runoff.The PO 4 -P and specific conductivity reduction percentages were similar (62 and 63%) possibly because sedimentation was a dominant reduction process for solids and PO 4 -P.The NO 3 -N treatment efficiency of the wetland exceeded the PO 4 -P and E. coli treatment efficiency possibly because of the denitrification removal pathway that is specific to NO 3 -N.A follow up study to monitor the wetland's nutrient and E.coli concentrations may be beneficial for determining the temporal variability of treatment.

Figure 3 .
Figure 3. Nitrate concentrations before (B), during (D) and after (A) rain events for inlet (I) and outlet (O) sampling locations.Circles indicate the mean, stars indicate statistical outliers.Nitrate concentrations were always lower near the outlet relative to the inlet

Figure 4 .
Figure 4. Phosphate concentrations before (B), during (D) and after (A) rain events for inlet (I) and outlet (O) sampling locations.Circles indicate the mean, stars indicate statistical outliers.Phosphate concentrations were always lower near the outlet relative to the

Figure 5 .
Figure 5. Log 10 E. coli concentrations before (B), during (D) and after (A) rain events for inlet (I) and outlet (O) sampling locations.Circles indicate the mean, stars indicate statistical outliers.E. coli concentrations were always lower near the outlet relative to the inlet