Nitrogen Removal from Urban Stormwater Runoff ... - Semantic Scholar

4 downloads 0 Views 342KB Size Report
parking lots, and rooftops) results in high stormwater runoff flows and volumes .... allowed to remain continuously submerged (Dietz and Clausen,. 2006; Hunt et ...
Nitrogen Removal from Urban Stormwater Runoff Through Layered Bioretention Columns Chi-hsu Hsieh, Allen P. Davis*, Brian A. Needelman

ABSTRACT: Bioretention is a low-impact technology used for the treatment of stormwater runoff in developed areas. The fates of mineral nitrogen compounds in two bioretention columns (RP1 and RP2) with different media-layering characteristics were investigated under multiple loadings of simulated urban runoff. The immediate capture of nitrogen was evaluated, with nitrogen transformation reactions that occurred during the drying periods between rainfall events. A greater proportion of ammonium was removed from runoff in RP2 (68 6 16%), which had a high permeability layer over a lower permeability layer, than in RP1 (12 6 6%), which had the inverse configuration. Both column systems demonstrated nitrate export (9 6 32% and 54 6 22% greater than input for RP1 and RP2, respectively), attributed to washout of nitrate resulting from nitrification processes between runoff loading events. Bioretention media with a less permeable bottom soil layer could form an anoxic/anaerobic zone for promoting nitrification/denitrification processes. Water Environ. Res., 79, 2404 (2007). KEYWORDS: stormwater, ammonia, nitrate, bioretention, runoff, eutrophication. doi:10.2175/106143007X183844

Introduction The creation of urban impervious surfaces (i.e., roadways, parking lots, and rooftops) results in high stormwater runoff flows and volumes and affects water quality. Urban runoff from impervious surfaces is a major component of nonpoint source pollution (U.S. EPA, 1996). Nutrients (including nitrogen and phosphorus) and sediment are typically the leading pollutants of impaired surface waters (including rivers, lakes, reservoirs, ponds, estuaries, and ocean shorelines) and groundwater in the United States (U.S. EPA, 1998). Fertilizers, atmospheric deposition, natural nutrient cycling, and other processes contribute nitrogen to stormwater runoff flows. Connected impervious area and conveyance infrastructure allow rapid transport of nitrogen during wetweather events, leading to regional water quality concerns. Excess inputs of nitrogen into waterways often lead to eutrophication problems (depletion of aqueous oxygen levels) and loss of biodiversity. Nitrogen is present in urban and highway stormwater runoff at approximately 0.6 to 1.4 mg-N/L for total Kjeldahl nitrogen (TKN) and 0.14 to 2.2 mg-N/L nitrate plus nitrite, with ammonium as a minor fraction of the TKN (Davis and McCuen, 2005; Taylor et al., 2005). Because of the critical nature of nutrient pollution, diverse nitrogen transport pathways in the environment, and difficulty in managing many nutrient sources, technologies are * Department of Civil and Environmental Engineering, University of Maryland, College Park, MD 20742; email: [email protected]. 2404

needed to address nutrient inputs to the water environment from a variety of source pathways. Bioretention is a vegetated infiltration practice for managing stormwater runoff from developed areas. In a typical configuration, bioretention includes a layer of approximately 75 to 100 cm of an engineered soil/sand/organic media, supporting a mixed vegetative layer. The bioretention facility is designed at approximately 5 to 7% of the drainage area. Runoff that is directed into the facility is allowed to pool to a maximum depth of approximately 15 cm. A significant fraction of pollutants are removed from the infiltrating runoff through a number of processes, including filtration, sorption, ion exchange, and possibly biological uptake. Infiltration to groundwater is encouraged, although, where local soil drainage is poor, an underdrain is typically included, to prevent long-term saturation of the system. The use of bioretention as a stormwater management practice is becoming prevalent in many areas of the United States, especially the east coast, as municipalities struggle with balancing growth and environmental protection of rivers and estuaries. When exploring the fate of pollutants in bioretention, and specifically with nitrogen compounds, several distinct timeframes of importance must be assessed. The first timeframe is that of the runoff infiltration event. During this time, influent pollutant species will flow through the bioretention media. For capture to occur within the infiltration time of the runoff, transport and reactions must be rapid. For example, at a superficial infiltration rate of 25 cm/h, a period of 1 or 2 hours is required for the water to infiltrate the bioretention depth, but the contact with each individual media particle (i.e., 50 lm) is approximately a few seconds. Such limited contact times allow physical processes and some very rapid chemical reactions, including adsorption and ion exchange, but are too short for slower biogeochemical transformations. Davis et al. (2001, 2006) noted efficient removal (55 to 80%) of ammonia and TKN in pilot-scale box bioretention studies and two selected field investigations. Conversely, in eighteen 6-hour bioretention column studies, only poor-to-fair removal efficiency was noted for ammonium (8 to 24% removal; Hsieh and Davis, 2005a). Ammonium is cationic in aqueous solution and is typically immobilized by negatively charged clays and organic matter in soils through sorption and ion exchange processes (Brady and Weil, 2002; Hook, 1983; Juang et al., 2001). Nitrate, conversely, is minimally held by bioretention media and, as an anion, is very mobile in soils and will not adsorb to soil media to any significant extent. Removal of nitrate was poor, at 1 to 24% in the eighteen 6-hour bioretention column studies (Hsieh and Davis, 2005a). It is expected that physical and chemical processes accounted for most Water Environment Research, Volume 79, Number 12

Hsieh et al.

Table 1—Media profiles of bioretention columns RP1 and RP2 (Hsieh et al., 2007).

Depth (cm)

Mass (kg)

Bulk density (kg/m3)

Media

0 to 5 5 to 20 20 to 95

0.8 8.2 31

560 1910 1440

Mulch Soil I Sand II

Top

0 to 30

12

1400

Middle Bottom

30 to 85 85 to 95

23 6

1460 2090

Homogeneous media mixture (3 kg mulch, 3 kg soil IV, 6 kg sand I) Sand II Soil IV

Media layer RP1 Top Middle Bottom RP2

of ammonium and nitrate removals, because flowrates were rapid through the columns. A second timeframe of bioretention is the interval between rainfall events. This interval may be as short as a few hours or as long as several weeks. During this time, the bioretention facility may drain and dry significantly. This is sufficient time for more complex chemical and biological transformations to occur in the bioretention media. In most soils, the quantity of nitrogen in organic forms far exceeds mineral nitrogen pools (Stevenson, 1986), with much of the mineral nitrogen present as clay-fixed ammonium (Stevenson, 1994). The carbon-to-nitrogen ratio of surface soils tends to stabilize between 10 and 12 upon humification, regardless of initial conditions (Stevenson, 1994). Therefore, long-term nitrogen retention in bioretention media requires an increase of total soil organic matter. Mineral nitrogen pools are increased by mineralization of organic nitrogen and are likewise decreased by nitrogen immobilization. Through aerobic nitrification processes involving Nitrosomonas and Nitrobacter species, captured ammonium ions in soils may be oxidized, ultimately becoming nitrate, as follows: NH4þ

! ! Nitrosomonas

NO 2

Nitrobacter

NO 3

ð1Þ

This transformed nitrate will accumulate in the bioretention media during the dry interval. With the next rain/runoff event, the accumulated nitrate ions will be leached from the media and into the underdrain or groundwater transport pathway. Leaching of nitrate at concentrations exceeding input levels has been observed in bioretention pilot systems (Davis et al., 2001, 2006). Where soils drain poorly, the held water impedes the diffusion of oxygen, thereby creating anoxic zones in the soil at a suitable oxidation–reduction (redox) potential for denitrification (Brady and Weil, 2002; Meyer et al., 2002). These zones may exist within small micropores of the soil aggregate structure or even individual soil particles in an otherwise drained media (Kremen et al., 2005). Under appropriate conditions, nitrate that has been trapped in these anoxic zones can undergo biological denitrification, resulting in the transformation of nitrate to gaseous nitrogen species, as follows:  NO 3 ! NO2 ! NO " ! N2 O " ! N2 "

ð2Þ

An available organic carbon source is required to sustain this process. For example, Lance et al. (1976) noted that 80% nitrogen November 2007

removal from secondary wastewater was achieved by providing an adequate carbon-to-nitrate-nitrogen ratio favorable for nitrification in soil columns. Even in otherwise aerobic soils, small zones of high microbial activity may develop denitrification potential in zones with high organic matter content (Nielsen and Revsbech, 1998; Peterson et al., 1996; Rice et al., 1988). Attempts have been made to promote microbial denitrification reactions in bioretention by using alternative designs, whereby a portion of the bioretention media is allowed to remain continuously submerged (Dietz and Clausen, 2006; Hunt et al., 2002; Kim et al., 2003). In this study, the fate of mineral nitrogen in runoff and bioretention media was investigated using two three-layered media columns subjected to multiple runoff loadings, allowing sufficient time for drainage and some drying between water loadings. Bioretention columns were used to compare the differences in ammonium and nitrate runoff effluent concentrations during 28 total wetting-drying cycles (12 and 16 each). A sequential combination system of nitrification and denitrification was hypothesized to form in the bioretention media at the interface of high permeability sand and less permeable soil layers. The objective of the work was to understand the moderate-term fate of mineral nitrogen species in bioretention media as affected by media layering configurations. Materials and Methods Two bioretention columns, 19.1 cm (inner diameter) by 110-cm (height), designated as RP1 and RP2, each with a different threelayer media configuration (Table 1), were prepared as described previously (Hsieh and Davis, 2005b; Hsieh et al., 2007). Two types of soil media were obtained from the Prince George’s County (Maryland) Department of Public Works and Transportation. Two types of sand, with different particle size distributions, were obtained from a local home supply store. Before packing into columns, sands were washed using the silica sand washing procedure described by Kunze and Dixon (1989). The mulch was produced from locally collected municipal leaves and grass clippings that had been piled into windrows for composting, obtained from the College Park (Maryland) City Department of Public Works. Media characterization was done by the Soil Testing Laboratory, Department of Natural Resource Sciences and Landscape Architecture, University of Maryland, College Park (Table 2). The RP1 column was designed specifically for high-rate infiltration. The surface layers would provide a mulch/soil zone for vegetation, with sand below to promote rapid infiltration and filtration. With RP2, an inverse configuration put the soil layer at the bottom to allow runoff storage in the sand layer above. A heterogeneous mix was used at the RP2 surface to allow vegetation survival. During each repetition, a synthetic runoff (Table 3) was applied to the bioretention column continually for 6 hours. In total, 12 repetitions were applied, each after a 4- to 7- (typically 6-) day dormant period for RP1, and 16 repetitions after 5 to 14 (typically 6) days drying were completed for RP2. For each repetition, the input synthetic runoff was stored in a 200-L plastic container and was well-mixed. To start the experiment, runoff was pumped into the column from the top, and the first input sample was collected. Because of the high drainage-area-to-bioretention-area ratio (typically 20:1 to 40:1), water pooling quickly occurs within bioretention facilities during rainfall events. Therefore, to better simulate flow conditions in field bioretention, the water head was maintained constant at 15 cm by an overflow hole in the column. Over each 6hour repetition, samples were collected every hour from the bottom 2405

Hsieh et al.

Table 2—Bioretention column media chemical and mechanical analyses (Hsieh and Davis, 2005a). CECd d10a d60b

Magnesium Phosphorus Potassium Calcium O.M.c

Media

mm

mm d60 / d10 pH

Sand I Sand II Soil I Soil IV Mulch

0.30 0.17 0.09 0.10 0.15

0.84 0.30 0.29 0.32 2.31

2.8 1.8 3.2 3.2 15.4

5.0 7.1 6.7 7.1 7.1

mg/100 g soil 2.5 9.5 28 27 28

4 5 7.5 9.9 56

0.8 3 35 18 35

0.8 2.8 NCe 68 . 44

%

meq/100 g soil

0.10 0.15 4.40 3.50 29.8

0.4 1.1 NCe 15 34

Sand Clay Silt %

%

%

Classification

92 95 71 71

5 3 17 14

3 2 12 15

Sand Sand Sandy loam Sandy loam

a

d10 5 effective size (10th percentile diameter). d60 5 60th percentile diameter. c O.M. 5 organic matter content. d CEC 5 cation exchange capacity. e NC 5 no data collected. b

of the column, to calculate the flowrate and measure pollutant concentrations. Nitrate and ammonium in water samples were analyzed using a Dionex DX-100 ion chromatograph with a Dionex AS4 column and a CS12 column, respectively (Dionex, Sunnyvale, California). The eluant for nitrate analysis was 1.3 mM Na2CO3/1.5 mM NaHCO3; 1.1 mN H2SO4 was used as the eluant for the ammonium measurement. Media nitrate and ammonium levels were analyzed before the experiments and following the cumulative multiple runoff applications. Data and discussion for the other pollutants applied to the columns have been presented elsewhere (Hsieh and Davis, 2005b; Hsieh et al., 2007). Results and Discussion Infiltration rates were controlled by the original media and subsequent captured suspended solids on the media surface. Therefore, flowrates (and hence retention times and mass fluxes) though the columns were different; flow through RP1 was constant, but that through RP2 decreased with time. Ammonium and nitrate concentrations in column effluents varied throughout each 6-hour experiment. Within the columns, strong evidence for nitrogen transformations between runoff applications was apparent. Infiltration rates in the columns have been discussed in detail (Hsieh et al., 2007). Briefly, the infiltration rate throughout all 12 repetitions in RP1 remained constant, at 0.35 cm/min. The upper

soil layer controlled the infiltration rate, and subsequent flow through the lower sand layer was unimpeded. Infiltration through RP2 was more variable, because the rate-controlling soil layer was at the bottom of the column. A buildup of head over the restrictive soil layer, in the media above, was necessary to drive the flow through the soil layer. Overall, a greater flowrate was noted through RP2, initially at 0.51 cm/min, but decreasing to 0.16 cm/min by the 14th repetition, which may be explained by additional resistance resulting from buildup of surface-captured suspended solids (Hsieh et al., 2007). Scraping off the top 5 cm of medium and replacing it with new original material at the end of the 15th repetition brought the infiltration rate back to the initial value (approximately 0.5 cm/min). Ammonium and Nitrate Removals. The overall ammonium removals in both columns are summarized in Figure 1. These data show the mean removal (difference between input and output concentration), taken at 1-hour intervals over each of the 6-hour study periods. Error bars indicate one standard deviation. As shown, RP2 performed considerably better for ammonium removal

Table 3—Makeup of synthetic urban runoff used in this study (Hsieh and Davis, 2005a). Value (mg/L, except pH)

Parameter

Source

pH Total dissolved solids Phosphorus Nitrate Ammonium Lead Suspended solids

7.0

HCl or NaOH

120 3 (as phosphorus) 2 (as nitrogen) 2 (as nitrogen) 0.1 150

Motor oil

20

CaCl2 Na2HPO4 NaNO3 NH4Cl PbCl2 Local soil sieved through a 0.59-mm opening Used oil from local garage

2406

Figure 1—Ammonium removal efficiency over 6-hour studies (mean 6 standard deviation) during repetitive experiments in bioretention columns. Water Environment Research, Volume 79, Number 12

Hsieh et al.

Figure 2—Ammonium removal (mean 6 standard deviation) for each hour sample during the 1st to 12th repetitive experiments (RP1) and the 2nd to 16th repetitive experiments (RP2).

efficiency (overall average for all repetitions 5 68 6 16%) than did RP1 (12 6 6%). During each 6-hour repetition, low variability in ammonium removal efficiency in RP1 was demonstrated (standard deviation ranged from 0.7 to 2%) whereas greater variability occurred in RP2 (standard deviation ranged from 7 to 31%). The ammonium removal by RP2 for the first few runoff applications was approximately 60%, but later increased to approximately 80%. Scraping off and replacing the media in repetition 15 did not affect the ammonium removal.

Figure 3—Nitrate removal efficiency over 6-hour studies (mean 6 standard deviation) during repetitive experiments in bioretention columns. November 2007

Figure 4—Nitrate removal efficiency (mean 6 standard deviation) for each hour sample during repetitive experiments in bioretention columns.

A more telling picture of the dynamic nitrogen removal/ transformation process results by combining percent removals for each hour over the entire 12 or 16 repetitions for each column (Figure 2). Similar low ammonium removal occurred in RP1 throughout all 6 hours, demonstrating a steady-state, relatively inefficient treatment. In contrast, for RP2, consistently efficient removal of ammonium was noted during the first 2 hours of runoff input (90 6 2% for the 1st hour and 92 6 2% for the 2nd hour; Figure 2). The removal efficiency gradually decreased to 51 6 16% after 6 hours, although the variability was high throughout these latter data points. The greater ammonium removal in RP2 can be attributed to the higher organic content from the larger amount of mulch, with a corresponding greater cation exchange capacity to promote ammonium adsorption. The increase in ammonium removal with time may be related to the increasing retention time through the RP2 column, although the high removal was maintained even at the final repetition after the media replacement, when the infiltration rate was increased to its original value. Figure 3 shows the summary of repetition results of nitrate removal throughout the entire experimental program, while the average nitrate removals on an hourly basis are presented in Figure 4. Based on the data of Figure 3, the media of both RP1 and RP2 overall demonstrated poor removal, and, in fact, the overall net result for both was the production of nitrate (‘‘removal’’ of 29 6 32% for RP1 and 254 6 22% for RP2). For RP1, significant nitrate washout from the column was noted during the 2nd to 5th repetitions (range from 221 to 264% removal), most likely resulting from washout of nitrate present in the initial media (to be discussed below). After this period, the average removal efficiency for nitratenitrogen was consistent and steady over each 6-hour experiment, ranging from 8 6 1% to 19 6 1%. Similar to RP1, abundant nitrate originally in the media apparently washed out from RP2 during the initial runs (initial ‘‘removal’’ efficiency from approximately 280 to 2102%), but the average became somewhat steadier, at 243 6 4% removal during the final 7 repetitions. The media scrape/replacement, again, did not affect nitrate removal. 2407

Hsieh et al.

Table 4—Nitrate-nitrogen in bioretention column media. RP1

RP2

Nitrate in media (mg NO2 3 -N/kg-media) Layer

Media

Initial

Final

Difference

Top Middle Bottom Total

Mulch Soil I Sand II

1355 12 4

123 23 0.2 to 2

1230 211 3.8 to 2

Nitrate mass lost (mg-N) 984 290.2 899 1790

Examining the hourly sampling (Figure 4), slight nitrate export was consistently noted for RP1, under fairly steady-state conditions. For RP2, however, an interesting pattern was observed; the 1-hour initial data point indicates overall moderate nitrate removal (29%), with high variability. This variability resulted from a temporal change in the nitrate results. The 1-hour result showed nitrate export for the first 5 repetitions, indicative of media washout. After the 6th repetition, however, high nitrate removal efficiency (75 6 22%) consistently was noted in the 1st-hour effluent of RP2 during each 6-hour experiment. In the 2nd-hour sample, however, a large amount of nitrate washed out from the column, and overall nitrate removal efficiency decreased to 2204 6 37%. Subsequently, the export of nitrate decreased gradually, with removal increasing from 2125 6 31% in the 3-hour to 218 6 6% in the 6-hour sample. Export of nitrate has been noted with pilot-scale box bioretention systems (Davis et al., 2006). In these box systems, nitrate concentrations up to twice that of the influent were noted at various depths using a sandy loam homogeneous soil media. More nitrate was exported at shallow depths, apparently resulting from mineralization of captured organic nitrogen. Controlled field studies with synthetic runoff (Davis and McCuen, 2005; Hsieh and Davis, 2005a) have shown slight nitrate removals of 5 to 20%; 10 to 30% nitrate removal from low nitrate inputs (0.09 to 0.13 mg/L NO3-N) have been noted with grab samples from rain events from two new bioretention facilities (Hsieh and Davis, 2005a). Considering both the ammonia and nitrate results, the configuration of the bioretention media significantly affected the fate of these nitrogen species from the runoff. In RP1, the less permeable mulch and soil I comprised the upper media layer (0 to 25 cm deep). Most of the runoff efficiently drained out from the RP1 column after each repetition, because bottom sand I (75 cm) had only 5% silt1clay content. Consequently, little water stayed in the column during the dormant time between each 4- to 7-day repetition period. As a result, the RP1 effluent throughout each 6-hour repetition was composed primarily from the runoff input. Because of the low available water, microbial mineralization of organic nitrogen does not appear to have been significant. Conversely, the RP2 media contained the less permeable soil in the bottom layer (at the 85 to 95 cm depth), with sand and sandy homogeneous layers above. Without sufficient head for complete drainage, a fraction of the runoff water was held for some time in the RP2 column after each experimental repetition. As such, based on the runoff infiltration rate data (Hsieh et al., 2007), it can be estimated that residual water that remained in the lower media (2 to 25 cm away from the bottom of the column) from the previous experiment was the primary constituent of the first experimental sample, instead of the newly applied input runoff. This water 2408

Nitrate in media (mg NO2 3 -N/kg-media) Media

Initial

Final

Difference

Media mixture Sand I Soil IV

345 4.3 30

46 to 51 4.5 14

299 to 294 20.2 16

Nitrate mass lost (mg-N) 3560 24.6 96 3650

holdup affected the fate of both nitrate and ammonium in the bioretention column. The very high ammonium removal in the initial runoff samples of RP2 apparently resulted from drainage of water held during the 4- to 7-day drying timeframe within the bioretention column. Relatively slow chemical and/or biological removal of ammonium would occur in this held water during dormant periods. As this residual water washed out, ammonium removal consistently decreased over the subsequent 4 hours of runoff flow, but still remained at good-tomoderate levels, as a result of the cation-adsorbing characteristics of the media. These hypotheses are supported similarly by the nitrate results of RP2. For the first 5 repetitions, the nitrate level in the column effluent was high, indicative of nitrate that had mineralized between events being washed from the media. Afterward, however, a fairly consistent pattern emerged. The first effluent sample contained water that was held in the lower, soil portion of the column and demonstrated a moderate degree of nitrate removal—approximately 30%. This water was likely held in the soil layer and, being saturated, developed anoxic conditions during the intervals between runoff applications, allowing denitrification to occur within the media. The second sample was mostly composed of the water that was first flushed through the upper sandy media (11 to 60 cm away from the bottom). Ammonium that was previously held within this media layer would be readily transformed to nitrate via microbial nitrification processes during the nonloading time increment. With the subsequent runoff event, this mobilized nitrate moved unabated through the rest of the column. Thus, a large slug of nitrate was consistently found in the second output sample. Subsequent effluent resulted from the newly input runoff, and nitrate levels were approximately equal to the input, unaffected by the column media. The importance of biological processes during the between-event timeframe is evident. In the case of RP1, the water drained rapidly from the media, not allowing significant retention and nitrification/ denitrification within the column. In the case of the RP2 configuration with the lower permeability media at the bioretention column bottom, greater amounts of water were held during the draining periods. This resulted first in nitrification throughout the column and then, in the low permeability/high fines saturated lower zone, denitrification, resulting in the net loss of both nitrate and ammonia. In a similar investigation, Lance et al. (1976) concluded that nitrate formed through oxidization of captured ammonium from infiltrating wastewater during the drying period subsequently washed out of the column, either in a small concentrated volume or in a diffuse manner, depending on the water infiltration rate. Mineral Nitrogen Mass Balances. The overall nitrate changes in the bioretention columns are estimated in Table 4. The initial Water Environment Research, Volume 79, Number 12

Hsieh et al.

Table 5—Mass balance analysis of ammonia-nitrogen and nitrate-nitrogen from sequential events in bioretention columns. RP1 Parameter

NH+4

NO2 3

Input from runoff (g nitrogen) Output from effluent (g nitrogen) Mass removal from infiltrating flow (%) Decrease in media (g nitrogen) Net decrease (g nitrogen) (A2B1D)

1.26 1.10 13% 0* 0.16

1.60 1.85 (216%) 1.79 1.54

RP2 NH+4

+

NO2 3

2.86 2.95 (23%) 1.79 1.70

NH+4

NO2 3

NH+4 + NO2 3

1.46 0.6 59% 0* 0.86

1.42 2.22 (256%) 3.65 2.85

2.88 2.82 (2%) 3.65 3.71

* Assumed value; see text for details.

mulch material contained a high concentration of nitrate (1355 mgN/kg-mulch), which was present as the surface layer in RP1 and as part of the homogeneous mixture that made up the top 30 cm in RP2. The lower nitrate levels found in the upper layers at the end of the column experiments demonstrate that most of this nitrate was either leached from the mulch and mixed media or immobilized to organic nitrogen forms. Evaluating the other media layers, the nitrate concentration of soil I in RP1 increased by 11 mg NO2 3 -N/kg-media, whereas that of the soil IV layer in the bottom layer of RP2 decreased (216 mg NO2 3 N/kg-media). As expected, the nitrate levels in the sand I layers in both columns were low initially, and nitrate accumulation was not found; the sand layers have minimal potential for nitrate accumulation (0.2 to 4.5 mg NO2 3 -N/kg-media). Input and output mass (M) of both ammonium and nitrate for each column throughout the repetitive periods was calculated as follows: M¼

td n X X 1

QC t

ð3Þ

i¼1

Where 5 runoff flowrate (L/min), 5 input or output nitrogen species concentration (mg/L), 5 measurement time increment (minutes), 5 number of samples collected during the runoff duration, and n 5 number of repetitions (12 for RP1 and 16 for RP2).

Q C t td

The total input and output ammonium-nitrogen values were 1.26 g and 1.10 g in the water phase, respectively, for RP1, and 1.46 g in and 0.60 g out for RP2 (Table 5). Therefore, the net percent ammonium removal was 13% for RP1 and 59% for RP2. Similarly, the input and output masses for nitrate-nitrogen were 1.60 g and 1.85 g, respectively, for RP1, and 1.42 g and 2.22 g for RP2. Accordingly, 0.25 g nitrate-nitrogen (16% of input) was exported from RP1, whereas 0.8 g (56% of input) was exported from RP2. The total mineral nitrogen input (NH4-N 1 NO3-N) for RP1 was 2.86 g, and output was 2.95 g, for a small net export of mineral nitrogen. With RP2, the input was 2.88 g, with an output of 2.82, for a 2% reduction. Therefore, essentially all mineral nitrogen that was applied to the column was found in the effluent, with 3% export found in RP1 and 2% removal in RP2. The mass of nitrate change in each media layer (Mr) through leaching, denitrification, and immobilization was calculated as follows: November 2007

Mr ¼ S 3 mp

ð4Þ

Where S 5 media mass used in each column layer (kg), and mp 5 gain in nitrate content per unit mass of each media after the repetitive column runs (mg/kg). Therefore, the overall system mass balance for nitrate was calculated as follows: Mlost ¼ Min  Mout þ

allX media

Mr

ð5Þ

i¼1

Where Mlost 5 nitrate mass totally lost from the system through denitrification or immobilization into organic forms; Min and Mout 5 nitrogen mass added to the column in the runoff and leaving the column in the underflow, respectively, calculated via eq 3; and Mr 5 nitrate changes in the various media layers, determined via eq 4. The results are shown in Table 5. Totally, RP1 showed a decrease of 1.54 g nitrate- nitrogen, while RP2 showed a decrease of 2.85 g, nearly all from the media. These results support the hypothesis of denitrification in both columns, especially in RP2, which held more water during drying cycles. Ammonia accumulation in the media was not monitored. However, the lack of ammonium removal from the RP1 water flow and the dynamic nature of the ammonium during the 6-hour experiments of RP2 suggest that the media layers are not accumulating ammonium. With the assumption that no ammonium is accumulating, the mass balance details of Table 5 can be completed. In both cases, the results indicate an overall net mineral nitrogen decrease in the system, with the bulk of the decrease coming from decreases in nitrate in the media. Overall, it is evident that nitrification processes proceeded in the upper media of RP2 because of the efficient ammonium removal from the influent, especially in the first two hourly samples and the slug of nitrate in the second sample. Additionally, because nitrate was regularly reduced in the first hourly sample, which mostly came from the water held in the bottom soil layer of RP2, active denitrification was apparently occurring in this zone, which is also supported by the significant decrease of nitrate-nitrogen in the media of RP2, according to the mass balance analysis. It is not clear, 2409

Hsieh et al.

however, that the nitrification/denitrification cycle induced in the RP2 column was beneficial from a stormwater management perspective. In both columns, considering the runoff flows only, the total mineral nitrogen that was put into the column in the aqueous phase was recovered in the effluent (23% difference in RP1 and 2% in RP2), and the overall effect on stormwater runoff was not significant. However, it does appear that denitrification occurred in the media. Given a more extended study, the denitrification may shift to produce a greater effect on runoff nitrate levels following decreases in the nitrate content of the media. This appears to be evident considering the 5 final repetitions of RP2. Ammonium removal is approximately 85% (Figure 1, 0.32 mg-N/L effluent), while nitrate excess over input is approximately 44% (Figure 3, 3.1 mg-N/L effluent). Combined, the average net mineral nitrogen is 3.4 mg/L from a 4.2 mg/L input, corresponding to a modest 20% nitrogen removal. The approximate locations of nitrification and denitrification processes in the RP2 column is diagrammed in Figure 5. These data support the importance of media configuration on nitrogen removal, as was similarly shown for phosphorus removal in bioretention (Hsieh et al., 2007). In bioretention box studies with a uniform sandy loam soil layer, nitrate removals from runoff were always less than 20%, with several instances of output nitrate concentrations higher than input (Davis et al., 2006). This occurs as accumulated nitrogen is nitrified between runoff applications, but no anoxic provision exists for promoting denitrification. With field installations, minor nitrate removal has been noted, likely for the same reasons. In column studies examining different media and media mixes, the greatest nitrate removal (49%) was found in a column that was comprised of nearly all mulch, demonstrating the importance of organic matter and slow flow (Hsieh and Davis, 2005a). In this study also, media layerings that had the least permeable layer at the surface produced negligible nitrate removal, while those with a more restrictive layer below higher permeability media demonstrated measurable, albeit small (13 to 27%), nitrate removal. In the field study of Dietz and Clauson (2006), engineering adjustment of the system drain promoted saturation in the facility and demonstrated redox potentials that could be indicative of denitrification conditions. The result was a statistically significant reduction of 18% in total mineral nitrogen. As demonstrated by the present column studies, bioretention cells with similar media, but different media configurations, can exhibit different nitrogen fate behaviors because of the contrasting environmental conditions that result, creating unique nitrification/ denitrification behavior in the columns. Media layering and resulting residual saturation play important roles in bioretention pollutant capture and fate, may explain some of the variability noted in field and laboratory nitrogen studies, and can possibly be exploited for nitrogen removal from runoff. Summary and Conclusions In summary, a total of twelve and sixteen 6-hour repetitions, respectively, were conducted in columns RP1 and RP2, to evaluate the removal efficiency of ammonium and nitrate from simulated runoff. The removal efficiency of ammonium was low in RP1 (13% on a mass basis), but was improved in RP2 (59% mass basis). However, nitrate was exported from both columns (16% of input for RP1, mass basis, and 56% of input for RP2). Considering both mineral nitrogen species, RP1 demonstrated a slight net export of mineral nitrogen, while RP2 showed a minimal 2% total mineral nitrogen reduction. Recent data on ammonium concentrations in 2410

Figure 5—Relative appearance of nitrification/denitrification processes in bioretention column with restricting soil layer below more permeable layer. urban runoff suggest that the values used in this study are somewhat high. Lower ammonium concentrations may show different levels of capture in bioretention media and may also result in less export of nitrate. The mulch used in this study had a high nitrate-leaching potential, which contributed to the poor removal performance noted in both bioretention columns. The mulch lost a significant amount of nitrate. Considering this nitrate source, mass balance calculations suggest that nitrification/denitrification and/or immobilization processes occurred in the columns. The two different media configurations demonstrated different performance efficiencies for ammonium and nitrate removals. Nitrification/denitrification processes were promoted in RP2 because of the specific layering media configuration in this column. The upper media layers generally stayed aerobic, favoring microbial nitrification processes. Through nitrification processes Water Environment Research, Volume 79, Number 12

Hsieh et al.

proceeding during wetting-drying cycles, ammonium in water that was held in the column between application events was transformed to nitrate, and high ammonium removal efficiency was noted when this water was drained through the columns. The lower layers in the RP2 column apparently developed anoxic conditions, and water held in this region saw a small fraction of nitrate transformed via denitrification. Both single and dual layers of bioretention media, with organicrich mulch throughout and with the bottom media layer designed to be less permeable than the upper layers, have the potential to form anoxic zones to encourage the nitrification/denitrification processes. A low available nitrate content is desired and recommended for the mulch/organic matter used in bioretention. Without washout of nitrate from the surface mulch layer, nitrate removal efficiency may be improved through the denitrification processes that will occur during the drying periods that occur between rainfall events. Longterm retention of total nitrogen in bioretention media would require a net increase in organic nitrogen. Establishing nitrification/denitrification zones in bioretention may prove beneficial in promoting total nitrogen removal from runoff. This can be partially achieved by configuring a less permeable media layer below a more permeable one. This configuration will allow denitrification to proceed in the upper aerobic zone, with the possibility of some denitrification in the less permeable layer. Supplemental organic matter from mulch should supply adequate carbon necessary for the denitrification process. During the time between rainfall events when water flows are reduced, slower microbial processes can become dominant and control the local biogeochemistry. This agrees with the nitrogen performance noted by Dietz and Clausen (2006) in saturated field bioretention facilities. Credits This work was supported by the Cooperative Institute for Coastal and Estuarine Environmental Technology, University of New Hampshire, Durham. Submitted for publication May 16, 2006; revised manuscript submitted March 19, 2007; accepted for publication March 20, 2007. The deadline to submit Discussions of this paper is February 15, 2008. References Brady, N. C.; Weil, R. R. (2002) The Nature and Properties of Soils, 13th ed.; Pearson Education Inc.: Upper Saddle River, New Jersey. Davis, A. P.; McCuen, R. H. (2005) Stormwater Management for Smart Growth; Springer: New York. Davis, A. P.; Shokouhian, M.; Sharma, H.; Minami, C. (2001) Laboratory Study of Biological Retention for Urban Stormwater Management. Water Environ. Res., 73 (1), 5–14. Davis, A. P.; Shokouhian, M.; Sharma, H.; Minami, C. (2006) Water Quality Improvement Through Bioretention Media: Nitrogen and Phosphorus Removal. Water Environ. Res., 78 (3), 284–293. Dietz, M. E.; Clausen, J. C. (2006) Saturation to Improve Pollutant Retention in a Rain Garden. Environ. Sci. Technol., 40, 1335–1340.

November 2007

Hook, J. E. (1983) Movement of Phosphorus and Nitrogen in Soil Following Application of Municipal Wastewater. In Chemical Mobility and Reactivity in Soil Systems, Nelson, D. W. (Ed.); Soil Science Society of America: Madison, Wisconsin, 241–255. Hsieh, C.-h.; Davis, A. P. (2005a) Evaluation and Optimization of Bioretention Media for Treatment of Urban Storm Water Runoff. J. Environ. Eng., 131 (11), 1521–1531. Hsieh, C.-h.; Davis, A. P. (2005b) Multiple-Event Study of Bioretention for Treatment of Urban Storm Water Runoff. Water Sci. Technol., 51 (3–4), 177–181. Hsieh, C.-h.; Davis, A. P.; Needelman, B. A. (2007) Bioretention Column Studies of Phosphorus Removal from Urban Stormwater Runoff. Water Environ. Res., 79, 177–184. Hunt, W. F.; Jarrett, A. R.; Smith, J. T. (2002) Optimizing Bio-Retention Design to Improve Denitification in Commercial Site Runoff, ASAE meeting paper No. 022233; American Society of Agricultural and Biological Engineers: St. Joseph, Michigan. Juang, T. C.; Wang, M. K.; Chen, H. J.; Tan, C. C. (2001) Ammonium Fixation by Surface Soils and Clays. Soil Sci., 166, 345–356. Kim, H.; Seagren, E. A.; Davis, A. P. (2003) Engineered Bioretention for Removal of Nitrate from Stormwater Runoff. Water Environ. Res., 75 (4), 355–367. Kremen, A.; Bear, J.; Shavit, U.; Shaviv, A. (2005) Model Demonstrating the Potential for Coupled Nitrification Denitrification in Soil Aggregates. Environ. Sci. Technol., 39 (11), 4180–4188. Kunze, G. W.; Dixon, J. B. (1989) Pretreatment for Mineralogical Analysis, Methods of Soil Analysis, Part 1—Physical and Mineralogical Methods, 2nd ed.; Agronomy Society of America and Soil Science of America: Madison, Wisconsin, 91. Lance, J. C.; Whisler, F. D.; Rice, R. C. (1976) Maximizing Denitrification During Soil Filtration of Sewage Water. J. Environ. Qual., 5, 102–107. Meyer, R. L.; Kjæ´r, T.; Revsbech, N. P. (2002) Nitrification and Denitrification Near a Soil-Manure Interface Studied with a NitrateNitrite Biosensor. Soil Sci. Soc. Am. J., 66, 498–506. Nielsen, T. H.; Revsbech, N. P. (1998) Nitrification, Denitrification, and NLiberation Associated with Two Types of Organic Hotspots in Soil. Soil Biol. Biochem., 30, 611–619. Peterson, S. O.; Nielsen, T. H.; Frostegard, A.; Olesen, T. (1996) O2 Uptake, C Metabolism and Denitrification Associated with Manure Hotspots. Soil Biol. Biochem., 28, 341–349. Rice, C. W.; Sierzega, P. E.; Tiedje, J. M.; Jacobs, L. W. (1988) Stimulated Denitrification in the Microenvironment of a Biodegradable Organic Waste Injected Into Soil. Soil Sci. Soc. Am. J., 52, 102–108. Stevenson, F. J. (1986) Cycles of Soil: Carbon, Nitrogen, Phosphorus, Sulfur, Micronutrients; John Wiley & Sons: New York. Stevenson, F. J. (1994) Humus Chemistry: Genesis, Composition, Reactions; John Wiley & Sons: New York. Taylor, G. D.; Fletcher, T. D.; Wong, T. H. F.; Breen, P. F.; Duncan, H. P. (2005) Nitrogen Composition in Urban Runoff—Implications for Stormwater Management. Water Res., 39 (10), 1982–1989. U.S. Environmental Protection Agency (1996) Managing Urban Runoff, EPA-841/F-96-004G; U.S. Environmental Protection Agency: Washington, D.C. U.S. Environmental Protection Agency (1998) The Quality of Our Nation’s Waters—A Summary of the National Water Quality Inventory: 1998 Report to Congress, EPA-841/S-00-001; U.S. Environmental Protection Agency: Washington, D.C., http://www.epa.gov/305b/98report/ 98brochure.pdf.

2411