Stormwater management in nutrient-sensitive watersheds: a case study investigating impervious cover limits and pollutant-load regulations

Watershed management is growing in importance as population continues to increase and land is further developed within source watersheds. A variety of watershed management strategies have been implemented across the globe, each designed to achieve specific hydrologic, water quality, ecological, or societal outcomes (Wang et al. 2016). A common watershed challenge in the southeast United States is non-point nutrient pollution resulting from agricultural and urban land uses (Carpenter et al. 1998). The case study described herein compared two watershed management strategies intended to reduce non-point nutrient pollution from urban areas: impervious cover limits and pollutant-load regulations. Impervious cover limits are often administered using zoning codes that restrict the type of land development permitted within a protected area; for instance, certain areas may only allow low-density residential development, which inherently limits impervious coverage. Pollutant-load regulations, on the other hand, place upper thresholds on pollutant loading rates within a protected watershed. An example pollutant-load regulation in the United States is a total maximum daily load (TMDL), which assigns specific pollutant export loading limits from point and non-point dischargers in a watershed (US Environmental Protection Agency 1991).

The case study site was located in Raleigh, North Carolina, USA, and was also situated within the nutrient-sensitive Falls Lake watershed. Falls Lake is Raleigh's supply reservoir and serves approximately 550,000 residents (NC Division of Water Resources 2016a). The watershed is approximately 2,000 km², comprised of 55% forest cover, 17% agriculture use, 15% urban development, and 13% other land cover. From 2001 to 2011, developed area increased by 11%, whereas forest and agricultural areas decreased by 4% and 2%, respectively (NC Division of Water Resources 2016b, derived from the USGS National Land Cover Database). Rapid urban development is expected to continue within the Falls Lake watershed (NC Office of State Budget and Management 2017); thus, urban stormwater management is a key concern.

Throughout the 1990s and 2000s, local governments in the Falls Lake watershed employed a variety of programmatic practices (public education, pet waste initiatives, etc.) and watershed management strategies (zoning codes and land conservation). Despite these efforts, in 2008 Falls Lake was added to the North Carolina 303(d) list as impaired for turbidity and chlorophyll-a (NC Department of Environment and Natural Resources 2010), indicating excessive algal growth and high nutrient loading (Smith et al. 1999). In response, the Falls Lake watershed TMDL was issued in 2011 which regulated nitrogen and phosphorus export loads from agricultural facilities, wastewater treatment plants, and urban stormwater. The regulation for stormwater runoff was converted to an annual export loading rate, restricting total nitrogen (TN) and total phosphorus (TP) export to 2.47-kg/ha/year and 0.37-kg/ha/year, respectively (NC Administrative Code 2011).

The City of Raleigh currently employs impervious cover limits within the Falls Lake watershed, only allowing Rural Residential zoned development (maximum 12% impervious cover). Once the Falls Lake watershed TMDL was implemented, the development community argued that zoning codes were no longer necessary; provided that new land development met the TMDL regulatory limits, zones other than Rural Residential should be allowable. The development community further suggested that stormwater control measures (SCMs) could be used in protected watersheds to meet the stormwater nutrient load limits set forth by the Falls Lake watershed TMDL (Dietz & Clausen 2008; Wilson et al. 2015; Line & White 2016).

In response to the development community's argument, the case study herein was conducted from 2013 through 2017 on a property within the Falls Lake watershed. The City of Raleigh approved a rezoning request to convert an existing golf practice facility (henceforth pre-development) into a commercial fitness complex (henceforth post-development), despite the property's location within a Rural Residential zone. The objectives of this study were threefold: (1) monitor stormwater quantity and quality before and after development to track changes to annual nutrient export loading rates, (2) determine if SCMs reduced nutrient export loading rates below the regulatory limits, and (3) compare impervious cover limits to pollutant-load regulations as watershed management strategies in nutrient-sensitive watersheds. The paper that follows includes a detailed description of the case study site, the experimental design and methods used during monitoring, data analysis procedures used to investigate hydrologic and water quality changes from pre- to post-development, results of all analyses, and recommendations to the City of Raleigh.

The case study was conducted on a 10.4-ha property in the piedmont physiographic region of North Carolina. Soils were Cecil series, consisting of a sandy loam surface layer and clay loam subsoil, belonging to hydrologic soil group B (Natural Resources Land Conservation Service 2017). Seasonally high water tables were not detected. Raleigh is situated in a humid subtropical climate region, and the normal annual precipitation depth is approximately 1,100-mm (National Centers for Environmental Information 2018). Prior to redevelopment, the Falls Lake watershed ridgeline split the property; 9.5-ha drained north towards Falls Lake and 0.9-ha drained south via storm sewers towards the Neuse River (Figure 1, left).

Pre-development land cover was approximately 60% lawn/open space, 35% forest, and 5% impervious surface. Post-development land cover was approximately 20% lawn/open space, 35% forest, 30% impervious surface, and 15% SCMs (Figure 1, right). The SCMs were aligned in series and collected runoff from all of the site's impervious surfaces. Sixty percent of the parking lot area drained to biofilters (bioretention cells, sans ponding and internal water storage) (Hunt et al. 2012). The biofilters included at least 0.6 m of media and an underdrain that conveyed through-flow to a storm pipe. Stormwater from the roof, the remaining 40% of the parking area, and the biofilter through-flow was conveyed to a 0.44 ha constructed stormwater wetland. The wetland had an approximate storage capacity of 290 m³ and was designed to store the 25-mm storm's runoff volume. Wetland effluent discharged to a 0.86 ha wet detention pond. The pond had an approximate storage capacity of 16,000 m³ and was designed to capture and store runoff volume from a 380-mm storm (approximately the 1,000-yr, 10-day storm). The wet detention pond was purposefully over-designed to ensure no stormwater would overtop the berm and discharge towards Falls Lake during extreme events. Stormwater was stored in the pond for at least 24 hours following a storm event, then pumped and discharged at a controlled rate to the City's storm drainage system if the water level was above normal pool elevation.

Monitoring results were organized into three distinct groups: pre-development (all pervious), post-development pervious, and post-development impervious-then-treated. This experimental set-up allowed for direct comparisons between pre- and post-development pervious areas and also between pervious and impervious-then-treated areas.

Pre-development monitoring began 8 January, 2013, and ceased 13 January, 2014. Stormwater quantity and quality was measured at the lowest point in the lawn area (the Pervious-Runoff station), with a catchment area of 4.73 ha. Areas south of the Falls Lake watershed ridgeline were not monitored.

Post-development monitoring spanned from 18 November, 2015 to 30 May, 2017. The monitoring station Pervious-Runoff from pre-development remained in place and continued to collect data, but the catchment area was reduced from 4.73 ha to 1.24 ha. Four additional monitoring stations were installed (Parking-Runoff, Wetland-Effluent, WetPond-Effluent, and PumpFlow) to track water quality changes as stormwater progressed through the series of SCMs and to measure the total quantity of stormwater pumped from the pond (Table 1).

Rainfall measurements were collected using a Teledyne ISCO 674 tipping bucket rain gauge underpinned by a manual rain gauge. Flowrates were measured using v-notch weirs in conjunction with Teledyne ISCO 730 bubbler flow modules. Flow-weighted samples were collected using a Teledyne ISCO 6712 sampler, composited, and delivered to the Center for Applied Aquatic Ecology at North Carolina State University for analysis. Event mean concentrations (EMCs) were measured for the following contaminants: total suspended solids (TSS), total nitrogen (TN), nitrate-nitrite (NO_2,3-N), ammonia (NH₃-N), total Kjeldahl nitrogen (TKN), organic nitrogen (ON), total phosphorus (TP), and soluble reactive phosphorus (SRP).

Hydrology was analysed in two ways: (1) event-based rainfall-runoff response and (2) annual volumetric comparisons. The event-based rainfall-runoff analysis was only conducted in the pervious areas (pre- versus post-development), because stormwater conveyed through the SCMs depended more on pump controls than rainfall inputs. Statistically significant differences between pre- and post-development rainfall-runoff responses were tested using linear regression models, an analysis of covariance (ANCOVA), and a Student's t-test.

For the annual volumetric comparisons, gaps in rainfall data were supplemented with observations from the Raleigh-Durham airport 15 km west (NC Climate Office 2018). Although the timing, depths, and intensities of specific storms varied across 15 km, the supplemental rainfall only needed to be representative of expected rainfall patterns at the site for the purposes of the annual volumetric comparison. A continuous 365-day subset from each monitoring period was selected to compute the estimated annual loading rates to adjust for different monitoring durations during pre- and post-development. These subsets were selected based on two criteria: (1) lowest occurrence of equipment malfunction and (2) similarity of annual rainfall depths and rainfall patterns between the selected time periods.

The 365-day continuous monitoring periods for pre- and post-development were specified as 12 January 2013 through 11 January 2014, and 22 April 2016 through 21 April 2017, respectively (Table 2). The pre-development monitoring period experienced 80 days of equipment malfunction, leading to missing rainfall and runoff data. Missing storm events were backfilled using data from the Raleigh-Durham airport solely for the annual comparisons. The rain gauge did not malfunction during post-development monitoring, so no missing storms needed replacement. The most noticeable difference between pre- and post-development rainfall was one extreme event (Hurricane Matthew) during post-development, a 175-mm storm, whereas the largest storm during pre-development was 67 mm. Total annual rainfall during pre-development was 1,093 mm and during post-development was 1,078 mm, both within 5% of the normal annual rainfall depth. Thus, comparisons between pre- and post-development annual monitoring periods were deemed viable despite one extreme event during post-development.

Pre- and post-development rainfall-runoff response in the pervious areas was monitored at the Pervious-Runoff station. Simple linear regression models were developed using rainfall depth, 5-minute peak rainfall intensity, and 5-day antecedent rainfall as independent variables and runoff depth as the dependent variable to determine which predictors had a significant effect. All variables were log-transformed. Variable selection for ANCOVA followed a forward selection procedure (only variables with significant effects in their respective simple linear regression models were included in the ANCOVA). Rainfall depth was the only significant predictor (p = 0.002, Table 3). A one-way ANCOVA was conducted to test for significant changes to runoff depth using monitoring period as the independent variable and rainfall depth as the covariate. The monitoring period did not have a statistically significant effect (p = 0.396, Table 3). Many storms did not generate surface runoff from the pervious areas (33 pre-development, 27 post-development), and these storms were omitted from the regression and ANCOVA analyses. Instead, these events were used to test for changes to the runoff threshold. A Student's t-test (with data log-transformed) was conducted to determine if the runoff threshold was significantly different between monitoring periods. The test did not find a significant difference in the runoff thresholds (p = 0.18, Table 3).

The only statistically significant finding was event rainfall depth had an effect on runoff depth. The lack of runoff-producing storms (only 13 pre- and 24 post-development) reduced statistical power and the ability to discern effects in the regression models. However, the findings still demonstrated trends worth noting. For instance, although the effect of monitoring period in the ANCOVA reduced model was insignificant, the runoff depth was expected to increase approximately 250% from pre- to post-development. Regarding runoff threshold, all three summary statistics listed (median, mean, and maximum rainfall depth) decreased from pre- to post-development, suggesting that the ability to retain rainfall decreased following development.

The pervious areas underwent multiple landscape changes from pre- to post-development. The contributing area decreased from 4.73 ha to 1.24 ha, vegetation was removed, soils were compacted, and slopes were increased. Landscape changes are known to impact rainfall-runoff response, as previously demonstrated by Gregory et al. (2006) and Line & White (2016). The analyses, although statistically insignificant, suggest that earthwork during construction increased the runoff response from pervious areas. Further research may be needed to confirm this assumption, but the findings were consistent with previously conducted studies.

First, the dewatering pump malfunctioned or was deactivated for approximately 180 total days during post-development monitoring (notice the long gaps in pump operation in Figure 2). The pump ran for approximately 15% of the monitoring period with a dewatering rate of approximately 200 lpm. The pump was only activated following very large storm events (Hurricane Matthew, October 2016) or following a series of storms in short succession (June and July 2016). A lack of pump inspection and maintenance caused operational issues, a common problem for new SCM technologies installed on private property (Gee & Hunt 2016). Second, the pond's high storage capacity kept water levels below the normal pool elevation for long periods of time, allowing for high evaporation losses. Lastly, the wet pond infiltrated significant volumes of water. Water level recession rates were analysed during periods without inflow and outflow, and levels receded approximately 13 mm/day quicker than evaporation alone could have accounted (NC Climate Office 2018). Although this infiltration rate may seem negligible, the conditions within the wet pond allowed for infiltration to occur continuously over the entire post-development monitoring period. It was estimated that roughly one-half of the wet pond's inflow volume was lost due to infiltration. Although wet detention ponds are not designed to infiltrate water, similar results have been observed in other monitoring studies in North Carolina, USA (Line et al. 2012; Baird 2015).

Pollutant concentrations from pervious areas were compared between pre- and post-development monitoring periods at the Pervious-Runoff station. Eight runoff events were sampled during pre-development and 13 were sampled post-development (Figure 3 and Table 5).

None of the tests detected significant differences between pre- and post-development pollutant concentrations, likely due to the small sample size (n = 8) from pre-development monitoring. That being said, the box plots show an increase in TSS and a decrease in TN, TKN, ON, NH₃-N, and SRP. Increased concentrations of TSS may have been due to unconsolidated soils and exposed ground following construction, whereas decreases in nutrients may have been attributed to the elimination of fertilizer application following redevelopment. The median TN and TP concentrations observed from pervious areas were both less than those reported by Skipper (2008) at a similar study site in North Carolina, suggesting that the stormwater runoff from the pervious surfaces was relatively clean.

Pollutant concentrations were measured throughout the suite of SCMs in the post-development impervious-then-treat area (stations Parking-Runoff, Wetland-Effluent, and WetPond-Effluent). Sixteen paired events were sampled at the Parking-Runoff and Wetland-Effluent stations, but only nine unpaired samples were collected at the WetPond-Effluent station due to the infrequency of pumping (Figure 4 and Table 6). One-sided location-shift tests were conducted to identify significant reductions.

The median TN and TP concentrations observed at the Parking-Runoff station were both less than those reported by Passeport & Hunt (2009) at parking lots across North Carolina, indicating that the stormwater runoff entering the wetland was relatively clean. This was likely due to the biofilters treating 60% of runoff from the parking area. The wetland significantly reduced TN, NO_2,3-N, and NH₃-N concentrations and the wet pond further significantly reduced TSS, TN, NO_2,3-N, and TP concentrations. The predominant pollutant removal mechanism in the wetland and wet pond was sedimentation. The oversized wet pond could store runoff for weeks (sometimes months) at a time, which further increased sedimentation and reduced pollutant concentrations below values observed in other wet pond monitoring studies in North Carolina (Wu et al. 1996; Mallin et al. 2002; Winston et al. 2013). Other nutrient pollutant removal mechanisms in wetlands, such as gross filtration and plant uptake, may not have been fully realized because the wetland was still in the maturation period and vegetative cover was not fully established (Merriman & Hunt 2014).

Estimated annual export loading rates were normalized by unit area for comparison to the Falls Lake watershed TMDL regulations of 2.47 kg/ha/yr for TN and 0.37 kg/ha/yr for TP (Figure 5). The estimated annual export loads from the pervious areas doubled from pre- to post-development for nearly all pollutants, as shown in Figure 5 by the orange and green bars, respectively (color graphics are available online). Although no statistically significant differences in pollutant concentrations were detected (Table 5), the annual depth of surface runoff from the pervious areas more than doubled (Table 4). The annual TN and TP loads from the pervious areas during post-development monitoring would have exceeded the Falls Lake watershed TMDL regulations; however, only stormwater from impervious surfaces was subject to the regulation.

The post-development annual export loads from the impervious-then-treated areas (Figure 5, blue) were lower than the annual loads from the pervious areas for nearly all pollutants. The TN and TP annual loads were both compliant with the Falls Lake watershed TMDL regulations (Figure 5, dotted lines in the TN and TP plots). Thus, nutrient loads from impervious areas were reduced from pre- to post-development using a combination of pollutant removal mechanisms and volume reduction, which has also been demonstrated in other monitoring studies such as Hunt et al. (2006). Although the results from this study were promising, the findings are not meant to suggest that a wetland and wet pond in series is the best design for minimizing nutrient export loads; rather, any combination of SCMs that reduce stormwater volume and treat pollutants is appropriate within a protected watershed.

The total post-development annual export loads were the sum of loads from the pervious and impervious-then-treated areas, normalized by the total area, for TN and TP (Figure 5, purple). These values were also less than the Falls Lake regulatory limits. Furthermore, the TSS, TP, and SRP total export loads were lower than the loads from pre-development pervious areas, despite the increased contributing area, imperviousness, and stormwater volume. TSS and TP load reductions can be attributed to the high amount of sedimentation within the wet pond. However, the changes to site-wide export loads of nitrogen species were mixed, due to high loads from pervious areas and less reduction potential from the on-site SCMs. Constructed wetlands and wet detention ponds have a wide range of nitrogen reduction potentials (Hammer & Knight 1994; Wu et al. 1996; Mallin et al. 2002), so this result was not altogether unexpected. SCMs, particularly bioretention, which include additional pollutant removal mechanisms such as filtration, sorption, and microbial transformations, would likely have reduced nitrogen loads more than this particular wetland or wet pond (Hunt et al. 2012).

This case study had site-specific design parameters that limited application beyond this location. First, the wet pond was highly oversized. Few developers have the financial or spatial availability to construct a wet pond as large as the one monitored during this case study. Second, the wet pond infiltrated high volumes of influent stormwater, which does not represent the normal hydrologic performance of a wet pond. Lastly, the pump to dewater the wet pond malfunctioned for much of the post-development monitoring period. Despite these limitations, the benefits of stormwater volume reduction and the hydrologic changes that resulted from land disturbance were still demonstrated, and management recommendations can be provided to municipalities located within nutrient-sensitive watersheds.

A case study was conducted in the City of Raleigh, North Carolina, USA, to investigate nutrient export loads from a commercial development in a protected watershed. The study monitored stormwater quantity and quality before and after land development to determine if pollutant-load regulations could be met if impervious cover limits were overridden. A suite of SCMs were installed at the site including an oversized wet pond with a dewatering scheme that pumped the stormwater back into the City's storm drainage system. Results from monitoring suggested that annual export loads from pervious areas increased following disturbance and runoff from impervious areas that was treated by a series of SCMs had very low pollutant concentrations and consequent annual loads. Annual export loading rates for TN and TP met the Falls Lake watershed TMDL pollutant-load regulations.

Based on the results of this case study, two recommendations can be provided regarding stormwater management in nutrient-sensitive watersheds.

• (1)

  Pollutant-load regulations appear to outperform impervious cover limits, especially if all areas disturbed during development are subject to regulation.

The intent of impervious cover limits is to reduce hydrologic change and protect water quality following land development. However, this case study demonstrated that pervious areas subjected to earthwork and compaction may also have an altered hydrologic response. Since impervious cover limits ignore stormwater from pervious surfaces, total pollutant export loads may be under-predicted. Furthermore, impervious cover limits may not require stormwater to be treated, instead relying on the imperviousness threshold to protect the watershed. Pollutant-load regulations, on the other hand, directly address the pollutants of concern and require stormwater treatment when necessary.

• (2)

  SCMs in protected watersheds should reduce effluent volumes and feature multiple pollutant removal mechanisms.

In protected watersheds, redundancy is the best way to ensure that pollutant export loads are minimized. Pollutant loads in stormwater can be addressed using two strategies: volume reduction and pollutant treatment. As demonstrated in this case study, both strategies were needed to meet the Falls Lake watershed TMDL regulations for TN and TP. The most assured way to meet pollutant-load regulations is to select SCMs that reduce discharge volumes and treat contaminants using multiple pollutant removal mechanisms.

The authors thank Amos Clark (and McAdams Company) for providing engineering plans and design details, Jack Miles (Life Time Fitness^®, Inc.) for providing field support, Aaron Koehler and Justin Schmidt (Life Time Fitness^®, Inc.) for their guidance through monitoring, the Center for Applied Aquatic Ecology at North Carolina State University for laboratory analysis of water quality samples, and Life Time Fitness^®, Inc. for funding this research.