Water Quality Improvement and Pollutant Removal by Two Regional Detention Facilities with Constructed Wetlands in South Texas

: Stormwater runo ﬀ introduces several pollutants to the receiving water bodies that may cause degradation of the water quality. Stormwater management systems such as detention facilities and wetland can improve the water quality by removing various pollutants associated with the runo ﬀ . The objective of this research project is to determine the performance and e ﬃ ciency of two major regional detention facilities (RDFs) with di ﬀ erent designs and structures in reducing pollutants based on various storm events in McAllen, Texas. The two sites are the McAuli ﬀ e RDF and the Morris RDF; each site was incorporated with a constructed wetland with a di ﬀ erent design and structure to enhance the pollutant removal process. The McAuli ﬀ e RDF reduced the concentration and load of many stormwater constituents in comparison to the Morris RDF. The observed concentrations and pollutant loads of suspended solids were much lower in the runo ﬀ of the inlet compared to the outlet for both sites. The McAuli ﬀ e RDF showed better concentration and load reduction for nutrients, such as nitrogen and phosphorus, of di ﬀ erent species. However, both sites did not show a signiﬁcant improvement of organic material. In addition, the indicator bacteria concentration represented a ﬂuctuation between the inlet and outlet at each site.


Introduction
Urban stormwater runoff contains substantial loads of numerous chemical and physical constituents that may adversely affect the water quality of rivers, channels, and lakes. These constituent loads are caused by rainfall wash-off. They carry various pollutants that cause a decline in aquatic biota and degradation of the water quality, and they are often discharged into untreated surface water [1,2]. Recently, the rapid urbanization in the Lower Rio Grande Valley (LRGV) has increased the stormwater runoff and pollutant loading into the receiving water bodies through the region. One of the affected waterbodies is the Arroyo Colorado Watershed, which flows through the LRGV [1]. The transformation of the watershed from its natural state has contributed to a water-quality problem. The watershed has been exposed to non-point source pollution from the extensive agricultural development that is interspersed with areas of rapid urban development [3]. Henceforth, an overload of nutrients and oxygen-demanding materials was transported across the river stream water; both Similar to detention basins and other GI systems, wetlands are installed to reduce the delivery of pollutants to surface waters, and their performance is commonly reported as being variable due to the site-specific nature of influential factors [29,30].
The removal of nitrogen and phosphorus through wetlands is variable and depends on several factors, such as the load variation, inflow concentration, hydraulic retention time, temperature, hydraulic efficiency, and type of wetland [31]. Nitrogen can be removed through different independent processes, such as denitrification, volatilization, sedimentation, and plant uptake. However, denitrification is considered the major removal mechanism of nitrogen in the wetland. Studies have shown that wetlands typically perform well for nitrate removal because the anaerobic conditions and organic material in wetland sediment create an ideal environment for denitrification. At the same time, due to a lack of oxygen in the wetland filtration bed, the removal of ammonia is limited [32]. Significant nitrate reduction is commonly observed in stormwater wetlands, but total nitrogen reduction depends on the species and concentration of incoming nitrogen [32,33]. Several studies have demonstrated that wetland can effectively remove the total nitrogen (TN) with a median removal efficiency of 37%, and that this reduction significantly correlates with the hydrological loading rate and temperature [31].
Similar to nitrogen removal mechanisms, phosphorus can be removed by sedimentation and plant uptake. In addition, phosphorus can be removed also by sorption or ligand exchange reactions; phosphate replaces hydroxyl or the water group from the surface of iron and aluminum hydrous oxide [32]. A study compared the total phosphorus (TP) removal of 146 studies from wetlands. Results showed that the median removal efficiency of all studies of TP is 49%, and it was significantly correlated with the TP concentration, hydrologic loading rate, and wetland area [31]. Similar to metals, phosphorus can desorb from sediments in a wetland under anaerobic conditions [24]. Humphrey et al., 2014 studied the stormwater water-quality improvement in a constructed wetland in North Carolina. Results showed that high pollutant reduction efficiency is due to the relatively large size of the wetland area and below-average rainfall that likely contributed to improving the water-quality performance [30].
Further investigations are needed to evaluate the sustainable removal performance of constructed wetland and detention basins, which will contribute greater insights into the nutrient treatment process [11,34]. Numerous studies have been conducted on detention basins in a wet climate, but their functionality in removing pollutants in arid/semi-arid regions has not been widely investigated during a rain event for certain land-use types [35]. The best prospects for successful wetland treatment and pollutant removal should be in warmer regions of the world [18], which indicates that the LRGV can be an ideal region for studying wetland performance and its effects on various pollutants. Previous studies have shown the effectiveness of wetland and detention ponds for improving stormwater quality separately under variable rainfall events. However, it is important to evaluate the performance of a regional detention facility (RDF) incorporated with a combination of a constructed wetland and a detention pond to improve the quality of stormwater runoff during different rainfall events. The overarching goal of this study is to quantify the effectiveness of regional detention facilities in improving the quality of stormwater runoff. The main objectives of this study are to compare the performance of two regional detention facilities with a constructed wetland-one with a detention pond and another without a detention pond-to achieve a feasible pollutant load reduction. The second objective is to evaluate the RDFs' performance for pollutant load reduction and to identify the factors influencing their effectiveness at each site.

Site Description
The Morris RDF is located behind Morris Middle School at 1400 Trenton Ave, McAllen, Texas, spanning an area of 121,406 m 2 . The RDF is elliptical in shape, and the slope within the RDF is~1%. The RDF was intended to serve as a recreational facility during dry weather. The design includes Sustainability 2020, 12, 2844 4 of 21 a channel on the periphery that serves to drain some runoff from within the basin ( Figure 1A). No microscreen was installed at the upstream of this facility. However, the Morris RDF has one wetland in the middle of the channel connecting the inflow and outflow monitoring points. A constructed wetland was created near the midpoint of this channel ( Figure 1B). This wetland was planted with a mixture of vegetation, including California bulrush (Schoenoplectus californicus) and Olney bulrush (Schoenoplectus americanus). The constructed wetland is elliptical in shape with a width of 22 m and average depth of 0.30 m. The side slope of this wetland is less than 1%. The average width of the incoming and outgoing channel is 4.87 m with a depth of 0.76 m ( Figure 1C,D). The total drainage area of the Morris RDF is over 20.6 km 2 , which is comprised of more than 93% urbanized landscape. According to the National Land Cover Database (NLCD), the dominant percentage of land cover drainage area in the Morris RDF urbanized areas is developed between high to low intensity, such as in parking lots and on pavements; the remaining 7% is either cultivated crops or developed open space ( Figure 2A) in the middle of the channel connecting the inflow and outflow monitoring points. A constructed wetland was created near the midpoint of this channel ( Figure 1B). This wetland was planted with a mixture of vegetation, including California bulrush (Schoenoplectus californicus) and Olney bulrush (Schoenoplectus americanus). The constructed wetland is elliptical in shape with a width of 22 m and average depth of 0.30 m. The side slope of this wetland is less than 1%. The average width of the incoming and outgoing channel is 4.87 m with a depth of 0.76 m ( Figure 1C,D). The total drainage area of the Morris RDF is over 20.6 km 2 , which is comprised of more than 93% urbanized landscape. According to the National Land Cover Database (NLCD), the dominant percentage of land cover drainage area in the Morris RDF urbanized areas is developed between high to low intensity, such as in parking lots and on pavements; the remaining 7% is either cultivated crops or developed open space ( Figure 2A)  Spanning 113,312 m 2 , the McAuliffe Regional Detention Facility (RDF) is located behind McAuliffe Elementary School ( Figure 3A). This RDF serves a drainage area of approximately 5 km 2 . It is a dual-purpose facility, providing recreational opportunities during dry periods and stormwater detention during wet weather. The RDF site boundaries are Nolana Avenue on the north and US 83 Business Expressway. On the south boundary, Ware Road is on the west side, and an eastern boundary extends to N 23rd Street. Runoff generated in the watershed is delivered to the RDF by a man-made drainage channel located upstream of the RDF. The watershed is comprised mainly of urbanized landscapes (83%). The NCLD land cover database showed that the majority of the urbanized area is either medium to low intensity ( Figure 2B). Cultivated crops and developed open space cover about 0.75 km 2 of the total drainage area. The flow to the RDF mainly consists of stormwater runoff along with some groundwater seepage. The dominant hydrologic soil group B (92%) has moderately low runoff potential when thoroughly wet [36]. The soil in the watershed is mainly comprised of type B (92%) with type D (6%) and type C (2%) forming the rest. The detention basin at McAuliffe Elementary School has gradually sloping banks. A larger amount of urban runoff with sediment and nutrients during wet weather discharges to the man-made drainage channel located upstream of the RDF. There are two wet detention ponds that are considered permanent Spanning 113,312 m 2 , the McAuliffe Regional Detention Facility (RDF) is located behind McAuliffe Elementary School ( Figure 3A). This RDF serves a drainage area of approximately 5 km 2 . It is a dual-purpose facility, providing recreational opportunities during dry periods and stormwater detention during wet weather. The RDF site boundaries are Nolana Avenue on the north and US 83 Business Expressway. On the south boundary, Ware Road is on the west side, and an eastern boundary extends to N 23rd Street. Runoff generated in the watershed is delivered to the RDF by a man-made drainage channel located upstream of the RDF. The watershed is comprised mainly of urbanized landscapes (83%). The NCLD land cover database showed that the majority of the urbanized area is either medium to low intensity ( Figure 2B). Cultivated crops and developed open space cover about 0.75 km 2 of the total drainage area. The flow to the RDF mainly consists of stormwater runoff along with some groundwater seepage. The dominant hydrologic soil group B (92%) has moderately low runoff potential when thoroughly wet [36]. The soil in the watershed is mainly comprised of type B Sustainability 2020, 12, 2844 5 of 21 (92%) with type D (6%) and type C (2%) forming the rest. The detention basin at McAuliffe Elementary School has gradually sloping banks. A larger amount of urban runoff with sediment and nutrients during wet weather discharges to the man-made drainage channel located upstream of the RDF. There are two wet detention ponds that are considered permanent pools. They provide more residence time for the runoff, as they are aided by sedimentation and infiltration. The first wet pond starting from the upstream has an estimated area of 3400 m 2 , and the second pool has an estimated area of 5139 m 2 . To maintain flow balance in between these two pools, a connection via a concrete pipe has been established. The water from the second pool drains out to the wetland area at the end of the basin through another concrete pipe. pools. They provide more residence time for the runoff, as they are aided by sedimentation and infiltration. The first wet pond starting from the upstream has an estimated area of 3400 m 2 , and the second pool has an estimated area of 5139 m 2 . To maintain flow balance in between these two pools, a connection via a concrete pipe has been established. The water from the second pool drains out to the wetland area at the end of the basin through another concrete pipe. The McAuliffe RDF was designed with a small channel wetland near the outflow monitoring point and the microscreen close to the inflow monitoring point on the upstream. A Coanda screen was installed in a concrete structure built parallel to the McAuliffe inlet ( Figure 3B). There is an increasing need to screen water in surface water collection systems to remove floating debris and small aquatic organisms to protect the receiving water bodies [37]. Previous studies found that a microscreen can remove solids by 3.5 times regardless of mesh size, but it is not effective in lessening the amount of dissolved substances [38]. The main purpose of using a microscreen in this study was to make the water free from debris or other larger particles, such as floatables, which may have clogged the channel by depositing on the channel bed. This screen is a self-cleaning apparatus, which performs without any power requirement. The microscreen used at the McAuliffe inlet RDF site operates based on the Coanda screen principle. The Coanda effect is a hydraulic feedback circuit that produces a fluidic oscillator for which the frequency is linear with the volumetric flow rate of fluid [39]. This screen is a self-cleaning apparatus that does not require any power. The Coanda screen is installed perpendicular to the two concrete chambers for storing and diverting the overflow. The dimensions of this microscreen are 7.6 m long and 0.76 m wide; it can handle a maximum flow of 2 cm/s. If the flow exceeds this maximum capacity, there will be an overflow, resulting in some debris being carried downstream. During normal operation, the incoming water flows over the screen, passes through the openings in the screen, and falls into the outlet underneath, which is approximately 0.46 m deep. A headwall perpendicular to the channel is provided in order to increase the head by 0.37 m. The flow from upstream of the Coanda screen flows through the screen aperture and the debris or other particles larger than the screen openings are trapped upstream. The debris collection chambers assist to hold the debris for a certain time before periodic maintenance is carried out.
A small wetland was constructed just before the McAuliffe RDF outlet; this wetland consists of a channel wetland of primarily Olney bulrush (Schoenoplectus americanus) plants ( Figure 3C,D). The

Sampling, Monitoring, and Analysis
The area velocity flow module (2150 Teledyne ISCO) was used to measure the stream velocity and the stream level. These two parameters can be used, along with the cross-sectional area of the  The McAuliffe RDF was designed with a small channel wetland near the outflow monitoring point and the microscreen close to the inflow monitoring point on the upstream. A Coanda screen was installed in a concrete structure built parallel to the McAuliffe inlet ( Figure 3B). There is an increasing need to screen water in surface water collection systems to remove floating debris and small aquatic organisms to protect the receiving water bodies [37]. Previous studies found that a microscreen can remove solids by 3.5 times regardless of mesh size, but it is not effective in lessening the amount of dissolved substances [38]. The main purpose of using a microscreen in this study was to make the water free from debris or other larger particles, such as floatables, which may have clogged the channel by depositing on the channel bed. This screen is a self-cleaning apparatus, which performs without any power requirement. The microscreen used at the McAuliffe inlet RDF site operates based on the Coanda screen principle. The Coanda effect is a hydraulic feedback circuit that produces a fluidic oscillator for which the frequency is linear with the volumetric flow rate of fluid [39]. This screen is a self-cleaning apparatus that does not require any power. The Coanda screen is installed perpendicular to the two concrete chambers for storing and diverting the overflow. The dimensions of this microscreen are 7.6 m long and 0.76 m wide; it can handle a maximum flow of 2 cm/s. If the flow exceeds this maximum capacity, there will be an overflow, resulting in some debris being carried downstream. During normal operation, the incoming water flows over the screen, passes through the openings in the screen, and falls into the outlet underneath, which is approximately 0.46 m deep. A headwall perpendicular to the channel is provided in order to increase the head by 0.37 m. The flow from upstream of the Coanda screen flows through the screen aperture and the debris or other particles larger than the screen openings are trapped upstream. The debris collection chambers assist to hold the debris for a certain time before periodic maintenance is carried out.
A small wetland was constructed just before the McAuliffe RDF outlet; this wetland consists of a channel wetland of primarily Olney bulrush (Schoenoplectus americanus) plants ( Figure 3C,D). The Schoenoplectus species complex includes interesting and dynamic wetland species of ecological importance, and the growth of the stem largely depends on the combined nitrogen and phosphorus concentration present in the water [39]. This plant provides some biological treatment by nutrient uptake, infiltration, and reduction in the flow of stormwater from the McAuliffe RDF basin. The wetted area of the wetland is 170.94 m 2 with side slopes of 5:1 horizontal to vertical (H:V) on the left side and 4.5:1 H:V on the right side. The maximum width of the main wetted portion of the wetland is 11.89 m with an average depth of 0.55 m from the water level. The wetland water flows from a concrete outlet with dimensions of 3.3 m × 1.7 m.

Sampling, Monitoring, and Analysis
The area velocity flow module (2150 Teledyne ISCO) was used to measure the stream velocity and the stream level. These two parameters can be used, along with the cross-sectional area of the stream, to calculate the flow rate. The velocity sensors installed in each of the stormwater monitoring sites work based on the Doppler effect. The level of the stream was detected based on the difference in atmospheric and hydrostatic pressures acting on an internal transducer. The velocity and level measurements were recorded by the sensor on a second-by-second basis. However, the data were saved once every 15 s to 24 h, depending on the requirement. The flow modules at the two RDFs were programmed to store data every 5 min. The velocity and level data were retrieved by a computer running the FLOWLINK program to calculate the flow using the velocity and level readings from the velocity sensor. The data recorded by the field instruments were retrieved and viewed on the FLOWLINK software. The retrieved flow data were imported into Excel software to draw the inflow and outflow hydrographs for each rainfall-runoff event and calculate the flow volume. The RDFs' residence time for each event was estimated based on the time between the inflow and outflow peak.
The water-quality sampling protocol adopted in the initial Texas Commission on Environmental Quality (TCEQ) approved the Quality Assurance Project Plan (QAPP) for collecting composite samples for every 41 m 3 of flow [40]. Automated composite samplers (Teledyne ISCO 6712 Portable) were set up at the inlet and outlet of the McAuliffe and Morris RDFs, which collected composite samples based on a user-programmed frequency in a 15 L bottle. The sampling interval was based on flow pacing (sampling process triggered by the flow) and was the same for both sites initially. A peristaltic pump was mounted on the control console that was housed in a protective acrylonitrile butadiene styrene (ABS) plastic casing. The pump was programmed to retry sampling up to a maximum of 3 times. The pump also purged the suction line before and after collecting the sample to ensure that the suction line was not plugged. The autosampler was connected to 2150 via cable, and the 2150 acted as the primary controller for the 6712. All samplers were programmed to enable themselves when certain level-rise conditions were satisfied. Before May 2012, the samplers were initially programmed to draw a fixed aliquot (100 mL) for every 41 m 3 of flow once the event started; this decision was based on a preliminary evaluation of historical rainfall data and drainage areas and estimated runoff coefficients. In our study, composite samples analyzed nutrients, suspended solids, and the bacterial concentration, which were used to calculate the pollutant load reduction on an event-by-event basis. The inflow and outflow volume for each event was obtained from FLOWLINK software.  [40]. The percentage load reduction of pollutants achieved by the RDF for each monitored rainfall event was calculated using the following a mass balance equation: summation of pollutant loads : where, C outlet = concentration at the outlet for event i = 1, 2, 3 . . . n; C inlet = concentration at the inlet for event i = 1, 2, 3 . . . n; and V = event volume In terms of water quality, the dataset of each pollutant was assessed using the Kolmogorov-Smirnov test to ensure the concentration values followed the normal distribution. Based on the characteristics of the data, the significance in the difference between the influent and effluent concentration would be hypothetically justified either by using a paired t-test (parametric) or the Wilcoxon signed-rank test (non-parametric). The correlation between different hydrologic and water-quality parameters was determined by Pearson Multiple Correlation Coefficient (PMCC) analysis. The significance of the correlations was also validated by the regression hypothesis test. In terms of the pollutant load, the significant difference in the amount between the inlet and outlet would be hypothetically tested by using either the t-test (parametric) or Mann-Whitney U test (non-parametric) based on the distribution of the calculated data.

Water-Quality Sampling and Analysis
Figures 4 and 5 represent the comparison of the influent and effluent concentrations at the Morris and McAuliffe RDF sites for different rainfall magnitudes, which were recorded during the monitoring timeframe. Results from the Kolmogorov-Smirnov test suggest that our concentration data do not differ significantly (p > 0.1) from that which is normally distributed. The pollutant concentrations for both RDFs were compared to those reported by the US Environmental Protection Agency (USEPA) National Urban Runoff Program (NURP) for mixed land-use settings.
below the NURP average (25,000 MPN/100 mL), approximated from the typical range of 400-50,000 MPN/100 mL [41]. Our paired t-test results indicate that the observed influent and effluent concentrations from the Morris RDF is not significantly (p > 0.05) different for the parameters analyzed. The NOx intake concentration was around 0.99 or 0.197 mg/L for most of the events; thus, those values were approximated to 1.0 & 0.2 mg/L. The maximum influent concentration of NOx (1.0 mg/L), TKN (4.20 mg/L), TN (5.20 mg/L), TP (1.4 mg/L), E.coli (61,310 MPN/100mL), and TSS (2,270 mg/L) was reported on 18 October 2012, induced from 35 mm of rainfall depth, which was depleted to 5.20, 6.20, 1.1, 853, and 12 mg/L prior to reaching the outlet, respectively. Surprisingly, the E. coli concentration was also observed higher at the outlet (57,940 MPN/100 mL) for the same event. The maximum BOD5 intake (57 mg/L) was observed from 4 mm rainfall depth (occurred on 1 July 2012), which was depleted to 29 mg/L at the outlet. An increase in the effluent concentration was observed for TP, TSS, and E. coli for about 25% of the total sampling events. Although there was no significant change observed in NOx concentration, less than 60% of the total rainfall events were reported with lower TKN, TN, and BOD5 concentrations at the outlet.  The concentration of all pollutants was relatively higher from May 2012 to October 2012. The number of BOD 5 and E. coli samples was less than for the other parameters due to the holding-time limitation after the sample collection. The holding for BOD 5 and E. coli was 48 and 24 h, respectively. Some of the samples passed the test standards holding time and were not analyzed for either BOD 5 or E. coli. Therefore, reliable BOD 5 and bacterial data were cumbersome to collect due to their shorter holding time. The median influent concentration of NO x , TKN, TP, TSS, and BOD 5 was observed much closer to the NURP standard, which was approximately 1.1 to 1.6 times higher than the standard values. However, the mean influent E. coli concentration (15,079 MPN/100 mL) was below the NURP average (25,000 MPN/100 mL), approximated from the typical range of 400-50,000 MPN/100 mL [41]. Our to reaching the outlet, respectively. Surprisingly, the E. coli concentration was also observed higher at the outlet (57,940 MPN/100 mL) for the same event. The maximum BOD 5 intake (57 mg/L) was observed from 4 mm rainfall depth (occurred on 1 July 2012), which was depleted to 29 mg/L at the outlet. An increase in the effluent concentration was observed for TP, TSS, and E. coli for about 25% of the total sampling events. Although there was no significant change observed in NO x concentration, less than 60% of the total rainfall events were reported with lower TKN, TN, and BOD 5 concentrations at the outlet.  Table 1 demonstrates the water-quality results summary for the Morris RDF for all parameters of analysis. Almost all the pollutants showed an elevated median concentration at the outlet of the Morris RDF. Although there was some depletion observed for TKN (11%), TN (7%), TP (9%), and BOD5 (22%) on average, the median TKN, TSS, and BOD5 concentration at the outlet slightly exceeded NURP typical values by 1.2, 1.1, and 1.6 times, respectively. Despite the relatively higher median than the NURP standard, the median TSS depletion efficiency was achieved by 49%, which was the maximum among all the pollutants analyzed. However, the expected TSS removal efficiency from a conventional detention facility is 75% [41,42].   Table 1 demonstrates the water-quality results summary for the Morris RDF for all parameters of analysis. Almost all the pollutants showed an elevated median concentration at the outlet of the Morris RDF. Although there was some depletion observed for TKN (11%), TN (7%), TP (9%), and BOD 5 (22%) on average, the median TKN, TSS, and BOD 5 concentration at the outlet slightly exceeded NURP typical values by 1.2, 1.1, and 1.6 times, respectively. Despite the relatively higher median than the NURP standard, the median TSS depletion efficiency was achieved by 49%, which was the maximum among all the pollutants analyzed. However, the expected TSS removal efficiency from a conventional detention facility is 75% [41,42]. Figures 6 and 7 demonstrate the results of the water-quality analysis of composite samples collected at the inlet and outlet points of the McAuliffe RDF. A total of eight rainfall events were considered for the analysis. The last two samples (for events that fell on 9 January 2013, and 30 April 2013) were taken after the installation of the Coanda microscreen. The depletion of nutrient constituents was found inconsistent for the samples analyzed from unprotected inlet conditions. The results from the paired t-test suggested that there was no significant (p > 0.1) difference between the inlet and outlet concentration of pollutants analyzed from the McAuliffe RDF. It appears that the effluent concentration of nutrients and organic solids (TKN, TN, TP, and BOD 5 ) was reported higher in less than 50% of the total sampling events before installing the microscreen. The maximum NO x inlet concentration 1 Inflow Volume, 2 Outflow Volume, 3 Concentration for E. coli = MPN (Most Probable Number)/100mL, 4 Standard Deviation, 5 25 Percentile, 6 75 Percentile, 7 Removal Efficiency, 8 Number of Samples, 9 N/A= Not available, 10 Mean concentration. Figures 6 and 7 demonstrate the results of the water-quality analysis of composite samples collected at the inlet and outlet points of the McAuliffe RDF. A total of eight rainfall events were considered for the analysis. The last two samples (for events that fell on 9 January 2013, and 30 April 2013) were taken after the installation of the Coanda microscreen. The depletion of nutrient constituents was found inconsistent for the samples analyzed from unprotected inlet conditions. The results from the paired t-test suggested that there was no significant (p > 0.1) difference between the inlet and outlet concentration of pollutants analyzed from the McAuliffe RDF. It appears that the effluent concentration of nutrients and organic solids (TKN, TN, TP, and BOD5) was reported higher in less than 50% of the total sampling events before installing the microscreen. The maximum NOx inlet concentration (2. Table 2 demonstrates the results summary of the influent and effluent concentrations for all the water-quality parameters of analysis for the McAuliffe RDF. Nutrient and BOD 5 showed somewhat contradictory concentration results at the outlet. Among all nutrient constitutes, NO x depletion was relatively better with a 37% mean and 13% median reduction. However, a noticeable depletion of TSS concentration was observed in 83% of the first six samples collected during the monitoring period. Before installing the microscreen, the overall depletion efficiency of TSS concentration was moderate (50%), and the median concentration (41 mg/L) in the effluent remained below NURP guidelines (67 mg/L). A substantial improvement in the TSS mean depletion (up to 98%) was noticed just after the installment of the Coanda microscreen (last two sampling events). The maximum TSS concentration (836 mg/L) was observed on 9 January 2013, which was substantially depleted to 13 mg/L. Also, the nutrient constituents (TKN, TN, TP) and BOD 5 were moderately depleted (30-80%) by the Coanda microscreen and well below the NURP standard. However, a further assessment would be important to ensure the long-term performance of the microscreen. The mean E. coli concentration (18,118 MPN/100 mL) at the McAuliffe outlet was within the NURP typical average (400-500 MPN/100 mL). Alike, at the Morris outlet, the maximum E. coli concentration was observed to be exceedingly higher (86,640 MPN/100 mL) than at the McAuliffe outlet; it was spiked with one very large accumulation on 5 April 2012. Overall, E. coli reduction was poor for about 40% of the total time monitoring events with an intense accumulation near the outlet.  Table 2 demonstrates the results summary of the influent and effluent concentrations for all the water-quality parameters of analysis for the McAuliffe RDF. Nutrient and BOD5 showed somewhat contradictory concentration results at the outlet. Among all nutrient constitutes, NOx depletion was relatively better with a 37% mean and 13% median reduction. However, a noticeable depletion of TSS concentration was observed in 83% of the first six samples collected during the monitoring period. Before installing the microscreen, the overall depletion efficiency of TSS concentration was moderate (50%), and the median concentration (41 mg/L) in the effluent remained below NURP guidelines (67 mg/L). A substantial improvement in the TSS mean depletion (up to 98%) was noticed just after the installment of the Coanda microscreen (last two sampling events). The maximum TSS concentration (836 mg/L) was observed on 9 January 2013, which was substantially depleted to 13 mg/L. Also, the nutrient constituents (TKN, TN, TP) and BOD5 were moderately depleted (30-80%) by the Coanda microscreen and well below the NURP standard. However, a further assessment would be important to ensure the long-term performance of the microscreen. The mean E. coli concentration (18,118 MPN/100 mL) at the McAuliffe outlet was within the NURP typical average (400-500 MPN/100 mL). Alike, at the Morris outlet, the maximum E. coli concentration was observed to be exceedingly higher (86,640 MPN/100mL) than at the McAuliffe outlet; it was spiked with one very large accumulation on 5 April 2012. Overall, E. coli reduction was poor for about 40% of the total time monitoring events with an intense accumulation near the outlet.   Overall, both of the RDFs showed a relatively better removal of TSS than the other pollutants analyzed. In our study, the mean depletion efficiency of TSS concentration (81%) by the McAuliffe RDF was achieved 21% higher than that from the Morris RDF. This result was closer to the desired removal efficiency (80%) achieved by GI controls [42]. The depletion of TSS concentration was more consistent in the McAuliffe RDF for most of the sampling events. However, the median TSS influent concentration in the Morris RDF (150 mg/L) was substantially higher than the McAuliffe RDF (93 mg/L) as it serves a relatively bigger basin (4 times higher than the McAuliffe RDF) with a higher percentage of high-density impervious cover (17.1%). For the rainfall that occurred on 9 January 2013, the lowest TSS effluent concentration achieved from the Morris and McAuliffe RDFs was 15 and 13 mg/L, respectively, which were considerably below the normally expected discharge concentration from a conventional facility (30 mg/L) [10,43]. It is important to note that the inlet modification with the installation significantly enhanced the depletion of TSS concentration. In conventional detention facilities, the primary mechanism of pollutant removal is sedimentation resulting from gravitational settling [43]. Several factors can influence the variability in the quality of effluents achieved in RDFs through the sedimentation process. Among them, three important hydrologic variables can have a possible impact on a RDF's settling mechanism: the residence time, the volume of feed, and the temperature [11]. Both RDFs are hydrologically connected to a vegetated wetland to improvise the overall retention and settling process. The water holding depth and aerial footprint of the McAuliffe RDF are much higher than that of the Morris RDF, including the drainage channels, two sequential wet ponds, and a wetland. These act like wet ponds and they offer more residence time for the runoff, thus aiding sedimentation. For the same rainfall event that occurred on 9 January 2013, the residence time of the McAuliffe RDF (4 h) was estimated at almost 8 times higher than that for the Morris RDF (0.5 h). This event could account for the McAuliffe RDF being 98% in terms of TSS concentration removal while the Morris RDF was 70%. In addition, the greater depth of the McAuliffe RDF's downstream wetland might have improved the storage volume for larger storm events and increased the overall residence time in the McAuliffe RDF. Temperature also affects the viscosity of the sediment particles in contact with water and, thereby, improves the settling process [11]. During our monitoring period, the local temperature varied in a range between 24 and 31 • C, which might have enhanced dynamic viscosity and the particle settling velocity. Apart from the hydrologic benefits, the inlet modification by the installation of the Coanda microscreen resulted in an enhanced TSS concentration (81% on average), which slightly exceeded the acceptable removal efficiency (80%) from other GI controls [42]. It is possible that a polluted stream entering the McAuliffe RDF, generated from the watershed, passed through a rigorous prescreening of large suspended solid particles and later received direct rainfall volume (cleaner than the stormwater runoff), which contributed to the significant depletion of concentrations in some cases. Thus, it can be said that the Coanda screen could be used in the RDF when comparable performance is required.
In terms of nutrients, the mean effluent concentration of TKN, TN, and TP was similar or slightly lower than the influent concentration for both RDFs. A smaller portion of nutrients was removed from both RDFs. For the Morris RDF, a slight depletion of nutrient concentration was observed for most of the rainfall events. The side-slope drainage area surrounding the RDF is the mostly vegetative or agricultural land cover and may have been a potential source of the organic and nutrient load generation through the erosion of the topsoil. During heavy rainfall events, these excess nutrients could be carried out by the agricultural runoff (i.e., fertilizers). This condition might induce a potential nutrient recharge near the RDF. Comparatively, the McAuliffe site covers a higher percentage of cultivated croplands (3.2%), which could lead to a significant amount of nutrient generation in the facility. Besides the negative value of the percentage, the BOD 5 mean and median reduction in the McAuliffe RDF indicates that there is a chance of organic solid accumulation in the wetland bed.
Maintaining low organic levels and adequate plant support are important to maintain the aerobic condition in the water for the effective oxidization of NH 4 + or other organic nitrogen to produce more mineral NO 3 form, which is later consumed by the roots of most plants [44]. The mean depletion of TKN (11%), TN (7%), TP (9%), and BOD 5 (22%) was relatively better in the Morris RDF. For most nutrient constituents, the percentage average depletion was negative in the McAuliffe RDF, which indicated an elevated concentration before reaching the RDF outlet. Several studies have found that higher BOD 5 and TN (predominantly present as NH 4 + ) removal is possible when a wetland bed is vegetated with plants of the bulrush genus (Schoenoplectus) because of their deeper root penetration (30-60 cm), which results in aeration and microbial nitrification [45].  Table 3. The significance of the correlations was validated by the regression hypothesis test. The positive correlations between the pollutants and inflow volume in the Morris RDF suggest that much of the nutrient and TSS generation was triggered beyond the RDF inlet, perhaps because the basin area serves a highly intense impervious land cover (17%). Conversely, the negative correlations between the McAuliffe inflow volume and outlet pollutant concentration suggested that much of the pollutant generation was not triggered by the total feed volume; rather their concentration underwent potential changes through the subsequent retention and sedimentation process throughout the RDF length, improvised by two sequential wet ponds. The same explanation could be appropriate for the negative correlation (R = −0.8) between the inflow volume and outlet BOD 5 concentration in the Morris RDF.
Possibly, the progression of nutrients throughout the RDFs was associated with sediment transportation, perhaps because of the ability of sediment particles to adhere to nitrate and phosphate on its surface. In the McAuliffe RDF, this hypothesis is strongly supported by the positive correlation (R ≥ 0.80, p < 0.05) between nutrient constituents (TKN, TN, and TP) and TSS concentration. For the McAuliffe RDF, strong positive correlations (R > 0.85, p < 0.05) were observed between BOD 5 and the rest of the probable organic sources (TKN, TN, TP, and TSS), perhaps because a large number of organics were carried by the solids. Adversely, high sediment concentration can interrupt NO x generation because it raises the issues of dissolved oxygen depletion and nitrification [46]; the hypothesis can be supported by the strong negative correlation (R = −0.80) between the TSS and NO x concentrations in the McAuliffe RDF. In general, high organic compounds are highly associated with an escalated bacterial population. Bacteria consume organic compounds for their growth and reproduction and deplete oxygen from the water [47]. This hypothesis can be supported by the strong correlation (R > 0.8) between E. coli and all probable organic sources (TKN, TN, TP, TSS, and BOD 5 ). However, the bacteria-nutrient interaction was more dominant (R > 0.95) in the McAuliffe RDF compared to the Morris RDF (R > 0.8). There was a close association (R = 0.91) between TKN and TP in the McAuliffe RDF.
However, the analysis of the variation of event-specific concentrations does not explain the differences observed in the pollutant load reduction. The following section of the paper discusses the results in terms of load reduction estimates for all pollutants analyzed from both RDFs.

Pollutant Mass Load Reduction
Water-quality data were collected for concentration along with the flow data at the inlet and outlet from each site to calculate the pollutant load reduction. Overall, during the monitoring period, the McAuliffe RDF exhibited better load reduction in comparison to the Morris RDF. A summary of load reduction at the inlet and outlet for each site for the different pollutants is presented. Figure 8 shows the box and whisker plots of the total load for all the collected samples of each pollutant at the outlet monitoring station on each site.

Summary and Conclusions
The analyses used in evaluating the performance and efficiency of the two RDFs in McAllen, Texas were able to give some insight into pollutant reduction in the constructed facilities. Both the McAuliffe and the Morris RDFs were incorporated with constructed wetlands to enhance the urban For the Morris RDF site, only TSS loads were significantly (p < 0.05) different between the inlet and outlet. While the loads for TN, TSS, TP, BOD 5 , TKN, and E. coli were not significantly less at the outlet than the inlet. On the other hand, for the McAuliffe site, NO x compounds the load, which includes nitrate and nitrites. They were significantly (p < 0.05) different between the inlet and outlet. However, the load for the TN, TSS, and TP were significantly less at the outlet than the inlet. However, there was a significant difference (p < 0.05) between the loads of BOD 5 , TKN, and E. coli between the two monitoring stations at the McAuliffe RDF.
Statistically significant reductions in loads of TSS were observed from both of the RDFs. For the Morris RDF, the median value of the inflow and outflow for the TSS load was 4309 ± 97,388 and 2316 ± 38,130 kg, respectively, for eight collected samples. The average and median load reductions were 53% and 60%, respectively. Only one stormwater event showed a negative reduction with a value of −64%, while in the McAuliffe RDF, the median value of the inflow and outflow for the TSS load was 1173 ± 3581 and 637 ± 681 kg, respectively, for eight collected samples. The average and median load reductions were 48% and 75%, respectively. The median average indicated a higher removal as only one event occurred on 4 May 2012; the TSS load at the outlet was higher than the inlet. For this event, TSS load reduction was negative due to both the TSS concentrations, and the outflow flow volume was higher at the outlet. Results from both sites indicate that TSS can be removed from the detention basin in the semi-arid coastal region under certain conditions, such as with certain concentrations, rainfall depth, and flow reduction. TSS is removed in detention facilities through sedimentation of the solid particles at the bottom of the basin. The removal mechanisms are achieved by reducing the water velocity and, hence, allowing the solids to settle before reaching the outlet.
For nutrients such as nitrogen and phosphorus, the Morris RDF did not show any significant difference for the NO x species. However, in the McAuliffe RDF, the NO x total load at the outlet was significantly (p > 0.05) less than at the inlet. Generally, there are two ways for NO x depletion to occur; it is either absorbed by plants as a form of nitrate, or it is biologically converted to nitrogen gas through the denitrification process. However, the detention time between the inlet and outlet was short and may not be enough to trigger the denitrification process. It is possible that NO x reduction was achieved through plant assimilation in the wetland section. This process can be supported by the difference in the wetland area between both sites. The wetland area in the McAuliffe RDF was 258 m 2 , and the average load reduction was 47%; the wetland in the Morris RDF was 9.3 m 2 with an average reduction of 6%. The McAuliffe RDF's large wetland area could enhance the NO x uptake from the urban stormwater runoff. Due to the relatively small area of the Morris RDF wetland, minimum NO x removal was observed. Both sites did not show any significant difference in TKN load reduction between the inlet and outlet at each monitoring station. However, the McAuliffe RDF showed better performance due to the runoff reduction volume. From eight storm events, the TKN load at the outlet was lower than at the inlet in seven events; the average pollutant load was 24 and 25 kg, respectively. In the Morris RDF, the TKN load was higher at the outlet in four storm events. The average load at the outlet and inlet was 156 and 137 kg, respectively. TN and TP loads showed significant reductions in the McAuliffe RDF only. As the McAuliffe RDF showed better performance in NO x and TKN removal than the Morris RDF, the TN total load at the McAuliffe outlet was expected to be lower than the at the inlet and the Morris RDF because the TN comprises all the nitrogen species, including NO x and TKN. Several studies have shown that a detention basin can reduce the TP in stormwater runoff [11,12]. Detention basins improve runoff water quality through the settling out of suspended particles that may carry contaminants such as phosphorus [35]. Both the McAuliffe and Morris RDFs showed reductions in TP load in comparison to the inflow load at each site. However, the McAuliffe RDF has an extended retention time compared to the Morris RDF due to the larger wetland volume and the presence of two wet basins. Therefore, the McAuliffe RDF showed an enhanced removal of solids and phosphorus.
Similar to the TKN load reduction, both sites did not show any significant difference between the BOD 5 and E. coli outlet and inlet loads for each site. The average BOD 5 load reduction for the McAuliffe and Morris RDFs was 22 and 19, respectively. Some events showed a negative reduction for BOD 5 .