Performance of Denitrifying Stormwater Biofilters Under Intermittent Conditions

Abstract

Denitrifying stormwater biofiltration systems that include submerged zones containing organic carbon sources can alleviate eutrophication by reducing nitrogen loadings from stormwater runoff. The objective of this research was to evaluate performance of denitrifying biofilters containing wood chip medium under varying dynamic loading rates and antecedent dry conditions (ADCs). Results showed that during ADCs, dissolved organic carbon (DOC) concentrations increased in pore water due to dissolution of the wood chip medium. During a storm event, nitrate (NO₃⁻) removal was enhanced due to high pore water DOC concentrations and mixing of NO₃⁻ from the influent with pore water; however, as DOC was gradually flushed out of the biofilter, higher effluent NO₃⁻ concentrations were observed over time. At lower hydraulic loading rates, effluent NO₃⁻ concentrations were low due to sufficient hydraulic residence time for denitrification. For all storm events tested, an overall NO₃⁻ mass removal efficiency of 85% was achieved. Sulfate reduction was observed at long ADCs (up to 30 days). Phosphate release was observed when NO₃⁻ was absent from the effluent, possibly due to dissolution of iron oxyhydroxides under low redox conditions. Findings from this study are being used to develop a mathematical model of nitrogen removal in denitrifying stormwater biofilters.

Introduction

Urban stormwater runoff is one of the primary sources of impairment to surface waters in the United States (USEPA, 2000). Urban runoff contains nutrients, such as nitrogen and phosphorus, which promote eutrophication. Eutrophication is known to contribute to toxic algal blooms, reduced fish yields and biodiversity, development of hypoxic zones, decreased property values, and reduced recreational use of surface waters (Smith, 2003). Stormwater biofiltration systems (e.g., sand filters and bioretention systems) can be used to remove nutrients from stormwater runoff; however, these systems have also been shown to export nitrate (NO₃⁻) (Collins et al., 2010). To solve this problem, a submerged biofilter containing a medium that serves as an electron donor may be incorporated into stormwater biofiltration systems to enhance denitrification (Kim et al., 2003).

Denitrification is a process that requires the presence of an electron donor, such as organic carbon or elemental sulfur. An overall stoichiometric equation for denitrification using a simple carbohydrate (CH₂O) as an electron donor is shown below: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}5CH_2O + 4NO_3^ - + 4H^ + \rightarrow 2N_2 + 5CO_2 + 7H_2O \tag{1}\end{align*} \end{document}

Denitrification is a stepwise process where NO₃⁻ is converted to nitrite (NO₂⁻), nitric oxide, nitrous oxide, and finally to nitrogen gas. Anoxic conditions are required before denitrification can proceed (Madigan et al., 2009). After oxygen and NO₃⁻ have been depleted, sulfate (SO₄²⁻) reduction, fermentation, and methanogenesis may occur (Madigan et al., 2009). Therefore, bioavailable organic carbon in stormwater biofilters can be utilized through a number of terminal electron accepting processes.

Factors that have been reported to affect NO₃⁻ removal in denitrifying biofilters include: influent NO₃⁻ concentration, oxidation–reduction potential, temperature, detention time, and electron donor bioavailability (Kim et al., 2003; Smith, 2008; Schmidt and Clark, 2013; Peterson et al., 2015). Other factors that may be important include biofilm acclimation, biofilter depth, electron donor hydrolysis rate, transport of influent dissolved organic carbon (DOC), and antecedent dry conditions (ADCs) (Kim et al., 2003; Chun et al., 2009; Chu and Wang, 2013). ADCs are the number of days between storm events. For bioretention systems, decaying vegetation and/or plant exudates may also increase DOC concentrations and affect NO₃⁻ removal in submerged biofilters (Zhang et al., 2011).

Due to ADCs, stormwater biofiltration systems receive influent on an intermittent basis. Prior studies investigating the effect of ADCs on the removal of NO₃⁻ in denitrifying biofilters are limited. Zinger et al. (2007) compared the NO₃⁻ removal performance of biofilters with and without a submerged zone at ADCs between 1 and 8 weeks. After an ADC of 2 weeks, NO₃⁻ removal in biofilters with a submerged zone was greater than in biofilters without a submerged zone. Kim et al. (2003) studied submerged biofilters with ADCs of 7 and 37 days and observed low initial effluent NO₃⁻ concentrations, which increased over time as the influent mixed with the pore water. Cho et al. (2011) observed that NO₃⁻ leaching occurs as a result of increased ADCs; however, their study evaluated unsaturated biofilters.

Several studies have investigated processes that occur in denitrifying biofilters during rainfall events (Dietz and Clausen, 2006; Smith, 2008; and Ergas et al., 2010; Lucas and Greenway, 2011; Lee et al., 2013). However, processes that occur in denitrifying biofilters during ADCs can also affect water quality performance. Barrett et al. (2013) observed phosphate (PO₄³⁻) removal in biofiltration mesocosms that included a submerged zone and hypothesized that PO₄³⁻ precipitation may occur during ADCs. Lynn et al. (2015) observed total Kjeldahl nitrogen (TKN) and DOC export in denitrifying biofilter microcosms, suggesting that the organic carbon from wood chip media will continually be released and then consumed by a variety of microbial terminal electron-accepting processes (aerobic, denitrifying, SO₄²⁻ reducing). Understanding the different transformations that occur in denitrifying biofilters during storm events and ADCs could be useful for improving stormwater biofiltration system designs (Grebel et al., 2013; LeFevre et al., 2014).

The overall goal of this research was to investigate the processes that occur in denitrifying biofilters under dynamic loading conditions. The specific objectives were to evaluate: (1) the overall water quality performance of denitrifying biofilters; (2) their performance under dormant conditions; and (3) their performance during storm events of varying intensity and ADCs. Although denitrifying biofilters may be incorporated into systems that may contain other elements (e.g., unsaturated sand layer, mulch, vegetation), this study focused solely on processes that occur in denitrifying biofilters with wood chip media. Data from three laboratory-scale columns that were operated in parallel are presented in this article. Although the columns had different depths, they were hydraulically loaded to achieve the same hydraulic residence time. Information on the effect of biofilter depth on performance can be found in Lynn (2014).

Materials and Methods

Reactor description

A schematic of the laboratory setup for the column study is shown in Figure 1. Experiments were carried out in three completely submerged 12.7 cm ID acrylic columns to mimic denitrifying biofilters with depths of 30, 45, and 60 cm. All studies were performed in the laboratory at ∼22°C. The media type used in this study contained a 2:1 (vol/vol) mixture of 0.6–1.3 cm pea gravel (Seffner Rock & Gravel, Tampa, FL) and 1.3–2.5 cm eucalyptus wood chips (Sarasota County, Sarasota, FL). This medium was found to have the best performance among several mixtures tested in a preliminary study in our laboratory (Lynn et al., 2015). The gravel–wood medium (bulk porosity=0.42) was supported by a 7.6 cm under-drain layer of pea gravel to prevent clogging. Due to wall effects, the measured porosity in laboratory columns is lower than the actual porosity observed in field systems using the same medium (Cohen and Metzner, 1981). Stormwater (described below) was pumped from a 0.21-m³ polyethylene reservoir using a separate Cole Parmer Masterflex L/S peristaltic pump (Thermo Fisher Scientific, Waltham, MA) for each column.

FIG. 1.

General laboratory setup for storm event studies. The 30 cm (A), 45 cm (B) and 60 cm (C) columns were operated with equivalent detention times by varying the flow rate.

Reactor operation

Eleven storm events were investigated in this study, as shown in Table 1. Storm Events (SE) Nos. 1, 2, 3, 4, 5, and 9 were defined as normal storm events. These storm events were operated under the same hydraulic loading conditions and only differed in ADCs. For this study, the term ADC is defined as the number of days between the prior and current storm event. The term dormant time is defined as the period of time the columns were not in operation before the storm event. For example, a storm event had an ADC of 8 days (general term), but the dormant time of the columns before the same storm event was 7.6 days (specific term). The normal storm events were also designed to operate for a period of 37.5 h to satisfy the typical requirement of filtering the treatment volume within 24–72 h (UDFCD, 2010; SWFWMD, 2013; NJDEP, 2014). SE Nos. 6 and 7 were operated with a higher influent NO₃⁻ concentration (4 mg/L NO₃⁻-N) and higher hydraulic loading rate, respectively, compared to the normal storm events. SE Nos. 8, 10, and 11 were operated at constant detention times of 1, 2, and 3 h, respectively. The storm events were set up to discharge an equal number of pore volumes from every column and during every storm event with the exception of SE No. 11, where 28.75 pore volumes were discharged from each column.

Table 1.

Characteristics of Each Storm Event

Storm event	Date	ADC (days)	Dormant time (days)	Duration (h)	Pore volumes discharged	Notes
1	11-18-2012	0	1.0	37.5	9	Began one day after acclimation period
2	12-5-2012	16	15.8	37.5	9
3	12-10-2012	4	3.0	37.5	9
4	12-19-2012	8	7.6	37.5	9
5	12-21-2012	0	0.0	37.5	9	Began immediately after SE No. 4
6	12-30-2012	8	7.9	37.5	9	Influent NO₃-N≈4 mg/L
7	1-8-2013	8	7.1	11.25	9	High flow rates
8	1-17-2013	8	8.5	9	9	Constant 1 h detention time
9	2-17-2013	30	30.7	37.5	9
10	3-29-2013	—	—	18	9	Constant 2 h detention time
11	5-6-2013	—	—	86.25	28.75	Constant 3 h detention time

SE Nos. 1, 2, 3, 4, 5, and 9 were defined as normal storm events. A leak occurred in the columns during the ADCs of SE Nos. 10 and 11.

—, not measured.

ADCs, antecedent dry conditions; SE, storm event.

The column experiments were designed to mimic the temporal flows that occur in stormwater biofiltration systems. During a storm event, runoff is conveyed to the stormwater ponding area, where the ponding elevation reaches its maximum height. In this period, the denitrifying biofilter flow rate is high; hence, high hydraulic loading rates were utilized for the initial stage of the experiment. After a storm event, runoff no longer enters the ponding area, while the ponding elevation decreases and conveys runoff through the biofilter. At this stage, water initially retained in the biofilter is discharged and is followed by runoff from the storm event. During this period, the flow rate decreases until the water elevation reaches the top elevation of the denitrifying biofilter. This experimental setup used pumps controlled by timers so there were step decreases in flow rate over each storm event rather than continuous decreases, as shown in Table 2. No pore water samples were collected; however, the water quality characteristics of the initial effluent samples and biofilter pore water were assumed to be similar (Lynn, 2014). The columns were also operated under equivalent detention times by adjusting the hydraulic loading rate of each column. The term detention time is defined as the theoretical time it takes for water to enter and exit the reactor, and duration is defined as the time of each experiment (or storm event). Note that the hydraulic loading rates used in this study can be ensured with an outlet control in the biofilter design (Lucas and Greenway, 2011; LeFevre et al., 2014).

Table 2.

Sample Numbers and Biofilter Flow Rate Patterns for Normal Storm Events (SE Nos. 1, 2, 3, 4, 5, and 9)

			Interval hydraulic loading rate (cm/h) ^a
Sequential sample number	Detention time (h)	Cumulative pore volumes discharged	30 cm	45 cm	60 cm
Initial^b	1^c	1	13.9	21.0	28.1
2	1	1.5	13.9	21.0	28.1
3	2	3.0	6.9	10.5	14.1
4	3	4.5	4.6	7.0	9.4
5	4	6.0	3.5	5.3	7.0
6	6	7.5	2.3	3.5	4.7
7	9	9.0	1.5	2.3	3.1
		Pore volume (L)	1.756±0.054	2.666±0.054	3.560±0.034

Targeted hydraulic loading rates to achieve equal detention times for each column.

Initial refers to the sampled water that was detained in the columns during the dormant period before the storm event and then flushed out during the storm event.

Sample was collected when the columns were operated with a 1 h detention time; however, the sampled water was detained in the columns during the ADCs before the storm event.

Source water

Source water used for this study was local surface water from a pond at the Botanical Gardens at the University of South Florida, Tampa campus. Stormwater was spiked with KNO₃ to achieve a feed NO₃⁻-N concentration above 2 mg/L to mimic nitrified stormwater runoff. This water was chosen to maintain relatively consistent water quality characteristics compared to actual stormwater runoff and to include natural constituents contained in nitrified stormwater compared with dechlorinated tap or synthetic water. Before SE No. 1, the columns were acclimated by the daily addition of one pore volume of source water to each biofilter for 1 month. The biofilters were not inoculated with seed microorganisms based on our prior study (Lynn et al., 2015), which showed that microorganisms present in the source water or wood chips were an effective inoculant. Mean influent characteristics with standard deviations can be found in Table 3. Specific influent characteristics for each storm event can be found in Lynn (2014). The analytical methods for this study, including method detection limits (MDLs), are shown in Table 4.

Table 3.

Overall Storm Event Influent and Effluent Water Quality Characteristics for the 45 cm Column

Influent ^a	NO₃⁻-N 2.14 (0.51)	TKN 0.3 (0.1)	TN 2.5 (0.5)	PO₄³⁻-P 0.12 (0.03)	SO₄²⁻-S 65.4 (5.6)	TSS 6.0 (1.8)	VSS 1.7 (0.3)	DOC 4.8 (0.3)	pH 7.5 (0.2)	DO 5.6 (0.4)
Storm event No.
1^b	0.28 [86]	0.4 [−24]	0.7 [68]	0.03 [70]	50.3 [8]	2.4 [57]	0.4 [74]	5.9 [−37]	6.9	0.1 [98]
2^b	0.05 [97]	0.4 [7]	0.5 [79]	0.03 [73]	51.6 [16]	2.8 [38]	0.5 [68]	22.0 [−368]	6.8	<0.1 [100]
3^b	0.08 [96]	0.5 [−45]	0.6 [75]	0.04 [60]	57.1 [14]	1.8 [55]	0.3 [76]	6.8 [−42]	7.0	0.1 [98]
4^b	0.06 [97]	0.5 [−17]	0.6 [76]	0.04 [73]	62.0 [12]	1.9 [67]	0.2 [87]	11.5 [−130]	6.9	<0.1 [99]
5^b	0.25 [86]	0.5 [−12]	0.8 [65]	0.04 [68]	71.4 [−2]	2.6 [69]	0.6 [61]	5.3 [3]	7.0	0.1 [99]
6	0.40 [89]	0.4 [−49]	1.1 [73]	0.06 [54]	58.1 [13]	2.6 [70]	0.8 [65]	11.2 [−115]	7.1	0.1 [98]
7	0.84 [62]	0.4 [−53]	1.3 [49]	0.04 [74]	56.6 [13]	3.2 [38]	1.0 [41]	9.0 [−94]	7.0	0.3 [95]
8	0.81 [58]	0.5 [−103]	1.4 [38]	0.08 [49]	58.9 [12]	—	—	11.2 [−146]	—	—
9^b	0.06 [97]	0.5 [−55]	0.6 [75]	0.16 [−29]	64.5 [14]	—	—	16.7 [−251]	6.9	—
10	0.39 [80]	0.5 [−234]	1.0 [55]	0.01 [86]	58.2 [5]	—	—	5.6 [−21]	—	<0.1 [100]
11	0.33 [84]	—	—	0.00 [100]	63.4 [−3]	—	—		—	—
Overall	0.32 [85]	0.5 [−42]	0.9 [66]	0.05 [60]	59.3 [9]	2.5 [59]	0.6 [68]	10.5 [−119]	6.9	0.1 [98]

Flow-weighted influent and effluent concentrations (mg/L) are shown with the standard deviation (mg/L) in parenthesis and the total mass removed [%] in brackets. Flow-weighted effluent values are shown for pH.

Targeted influent NO₃⁻-N concentration for all storm events was 2.0 mg/L, except for SE No. 6 (4 mg/L).

SE Nos. 1, 2, 3, 4, 5, and 9 were defined as normal storm events for this study.

DOC, dissolved organic carbon; TKN, total Kjeldahl nitrogen; TN, total nitrogen; TSS, total suspended solids; VSS, volatile suspended solids.

Table 4.

Analytical Methods and Method Detection Limits

Parameter	MDL (mg/L)	Method	Reference	Instrument	Company
DOC	0.11	5310 B	Rice et al. (2012)	Shimadzu TOC-V CSH, TOC/TN Analyzer	Shimadzu Scientific Instruments (Columbia, MD) SAB
TN	0.03	4500-N	SAB
DO	—	4500-O G	SAB	Orion 5 Star	Thermo Scientific, Inc. (Beverly, MA)
pH	—	4500-H B	SAB	SAB
NH₄⁺-N	0.07	4500NH3C/D	SAB	850 Professional Ion Chromatograph	Metrohm AG (Herisau, Switzerland)
NO₃⁻-N	0.01	300.1	USEPA (1997)
NO₂⁻-N	0.04	SAB	SAB	SAB	SAB
PO₄³⁻ -P	0.02	SAB	SAB	SAB	SAB
SO₄²⁻-S	0.01	SAB	SAB	SAB	SAB
TSS	—	2540 D	Rice et al. (2012)	Analytical Balance	Mettler Toledo (Greifensee, Switzerland) SAB
VSS	—	2540 D	SAB	SAB
TKN	—	—	—	—	—
Org-N	—	—	—	—	—

DOC concentrations were estimated from nonpurgeable organic carbon concentrations.

TKN concentrations were calculated by difference TN−NO₂⁻-N−NO₃⁻-N.

Organic nitrogen (Org-N) concentrations were calculated by difference TKN−NH₄⁺-N.

DO, dissolved oxygen; MDL, method detection limit; SAB, same as above.

Data analysis

Flow-weighted influent and effluent concentrations were calculated for each storm event and were used to calculate mass removal efficiencies (Table 3). Initial sample data were assumed to represent the water quality characteristics of the pore water during the dormant period. Dormant period values from SE Nos. 10 to 11 were not included in dormant period means due to a leak that occurred during the dormant period of these storm events. The following analyses were determined through linear regression to calculate the coefficient of determination (r²): first-order kinetics for SO₄²⁻; zero-order kinetics for PO₄³⁻-P; and increases in TKN during ADCs.

Results

Flow-weighted influent and effluent concentrations and mass removal efficiencies for all measured water quality parameters for the 45 cm column are shown in Table 3 [See Lynn (2014) for results from the 30 and 60 cm columns]. Average mass removal efficiencies for NO₃⁻, TKN, and total nitrogen (TN) were 85%, 43%, and 66%, respectively. Influent and effluent NH₄⁺-N concentrations were consistently below the detection limit (BDL) of 0.07 mg/L. Removal of PO₄³⁻, SO₄²⁻, total suspended solids (TSS), volatile suspended solids (VSS) and dissolved oxygen (DO), decrease in pH, and export of DOC were observed. During normal storm events, the lowest NO₃⁻, TN, and SO₄²⁻ mass removal efficiencies and the lowest mean effluent DOC concentrations were observed during the 0-day ADC storm events (SE Nos. 1 and 5). The lowest PO₄³⁻ mass removal efficiency was observed during the 30-day ADC storm event (SE No. 9). Similar phenomena were observed in the 30 and 60 cm columns.

Mean water quality data for the water that discharged from the columns just after the dormant period (assumed to be similar to the pore water) are shown in Figure 2. These data represent the initial samples that were taken after ∼0.25 pore volumes were discharged from the columns. Pore water DOC concentrations increased to a maximum of 114±25 mg/L for the 16-day ADC storm event and then decreased to 76±14 mg/L for the 30-day ADC storm event (Fig. 2a). Sulfate concentrations rapidly decreased between the 0 and 8-day ADC storm events and then remained low with increasing ADCs (Fig. 2b). A first-order kinetic model was fit to the SO₄²⁻ data (0.42 h⁻¹; r²=0.99). PO₄³⁻ concentrations were observed to increase approximately linearly with increasing dormant period (0.040 mg/L-d; r²=0.83) (Fig. 2c). Although TKN export (mean=0.2 mg/L) was observed to consistently occur during dormant conditions (Fig. 2d), it was not correlated with dormant time (r²=0.13). Similarly, pH was observed to decrease during dormant conditions (data not shown), but was not correlated with dormant time (r²=0.41).

FIG. 2.

Initial sample (collected after 0.25 pore volumes discharged) mean effluent dissolved organic carbon (DOC) (a); SO₄²⁻-S (b); PO₄³⁻-P (c) concentrations; and TKN increase (d) data from all three columns plotted with the dormant time before each storm event for SE Nos. 1 through 9. TKN increase was calculated by subtracting influent from effluent TKN concentrations. The line represents a zero-order PO₄³⁻-P release model. Error bars represent standard deviation. SE, storm event; TKN, total Kjeldahl nitrogen.

The mean values for water quality data from the 45 cm column for all the normal storm events (SE Nos. 1 through 5 and 9) are shown in Figures 3 and 4. Similar trends were observed in the 30 and 60 cm columns. Initial pore water NO₃⁻ removal efficiencies of at least 98% were observed for all of the normal storm events. After the initial pore water volume was discharged, mean effluent NO₃⁻ concentrations increased (second sample taken) and then decreased as the hydraulic loading rate decreased (Fig. 3a). Mean effluent DOC concentrations decreased from the average pore water value of 49.9 mg/L to a minimum of 5.2 mg/L and then increased slightly as the hydraulic loading rate continued to decrease (Fig. 3b). Initial mean effluent PO₄³⁻-P concentrations decreased from 0.37 mg/L to BDL and then increased after the hydraulic loading rate decreased to 5.3 cm/h (Fig. 3c). Mean effluent SO₄²⁻ concentrations rapidly increased up to the mean influent SO₄²⁻ concentration (Fig. 3d). Higher mean TKN concentrations, lower mean pH values, and lower mean TSS and VSS concentrations were observed in the effluent compared to the influent for all sampling events (Fig. 4a–d).

FIG. 3.

Mean influent and effluent NO₃⁻-N (a); DOC (b); PO₄³⁻-P (c); and SO₄²⁻-S (d) concentrations from the 45 cm column during the normal storm events. Mean initial sample effluent DOC (49.9 mg/L), PO₄³⁻-P (0.37) and SO₄²⁻-S (19.4) concentration and second sample effluent DOC (21.3 mg/L) and SO₄²⁻-S (47.7) concentration data are not shown. Error bars represent standard deviation.

FIG. 4.

Mean effluent TKN (a); pH (b); total suspended solids (TSS) (c); and volatile suspended solids (VSS) (d) data from the 45 cm column during the normal storm events. Error bars represent standard deviation.

Effluent NO₃⁻ and PO₄³⁻ concentration data from the 45 cm column constant flow storm events (SE Nos. 8, 10 and 11) are shown in Figure 5a and b. During these storm events, effluent NO₃⁻ concentrations increased over time throughout the experiment. Effluent PO₄³⁻ concentrations for SE Nos. 8, 10, and 11 decreased to BDL after three, three and four pore volumes were discharged, respectively, and remained BDL for the remainder of the storm events. Similar trends were observed in the 30 and 60 cm columns. A plot of effluent NO₃⁻ versus DOC concentration for all columns and constant flow storm events (excluding SE No. 11) is shown in Figure 6. The effluent NO₃⁻ concentrations were significantly lower (paired t-test; p-value <0.001) in the samples that contained an effluent DOC concentration greater than 10 mg/L (NO₃⁻-N=0.10±0.10 mg/L) compared to lower than 10 mg/L (NO₃⁻-N=0.53±0.38 mg/L).

FIG. 5.

Effluent NO₃⁻ (a) PO₄³⁻ concentration (b) data from the 45 cm column during the constant flow storm events of SE No. 8 (21.0 cm/h), SE No. 10 (10.5 cm/h) and SE No. 11 (7.0 cm/h). SE No. 11 sample data after 15 pore volumes were discharged are not shown.

FIG. 6.

Effluent NO₃⁻ and DOC concentration data from the 30, 45 and 60 cm columns during the constant flow 21.0 cm/h (SE No. 8) and 10.5 cm/h (SE No. 10) storm events.

Discussion

Overall results (Table 3) show that consistent NO₃⁻ and TN removals can be achieved in properly designed wood chips biofilters. However, some export of TKN was observed, possibly due to leaching of organic nitrogen from the wood chips (Schmidt and Clark, 2012; Peterson et al., 2015). The overall PO₄³⁻ results provide evidence that denitrifying biofilters are capable of retaining phosphorus. The low SO₄²⁻ removal efficiencies observed were similar to the findings by Lynn et al., (2015) and Elgood et al., (2010) who studied denitrifying biofilter microcosms using stormwater and mesocosms using wastewater, respectively. The overall TSS and VSS results (Table 3) indicate that denitrifying biofilters can be utilized for other stormwater treatment applications such as polishing filters to improve TSS and VSS removal in wet detention systems. However, the high TSS and VSS removals indicate that problems with clogging may be a concern. Hatt et al., (2007) evaluated the performance of a gravel filter and observed clogging throughout their study. However, their filters incorporated an underlain layer of fine sand, which may have induced system clogging. The overall pH results (Table 3) indicate that volatile fatty acids (VFAs) are produced through hydrolysis of the wood chips, which dissolve into the pore water and decrease the pH.

Overall results (Table 3) also show how ADCs affect the performance of denitrifying biofilters. During periods with frequent rainfall events (SE Nos. 1 and 5), lower NO₃⁻ and higher PO₄³⁻ removal efficiencies may occur compared to dryer periods with longer ADCs (SE Nos. 2, 3, 4, and 9). Moreover, longer ADC storm events will likely discharge higher DOC loadings. Nitrate removal results observed in this study were similar to the results reported by Kim et al., (2003) who observed high NO₃⁻ mass removal efficiencies in saturated denitrifying biofilter columns that were subjected to long ADC storm events. However, the PO₄³⁻ removal results are in contrast with Barrett et al., (2013) who hypothesized that PO₄³⁻ removal could increase with an increase in ADCs due to PO₄³⁻ precipitation.

Dynamic NO₃⁻ removal performance of the columns showed that during normal storm events the initial mean effluent NO₃⁻ concentration was near or equal to MDL, increased to the highest effluent concentration at the highest hydraulic loading rate, and then decreased as the hydraulic loading rate decreased (Fig. 3a). Based on these data alone, an increase in detention time will increase NO₃⁻ removal as was previously observed by others (Kim et al., 2003; Smith, 2008; Lucas and Greenway 2011; Peterson et al., 2015). However, mixing of the influent and pore water may also affect effluent NO₃⁻ concentrations (Kim et al., 2003) (Fig. 5a). In addition, DOC flushing will decrease the rate of heterotrophic denitrification and will result in higher effluent NO₃⁻ concentrations over time (Figs. 5a and 6). Similar results were reported by Yang et al. (2010), who observed that higher carbon to nitrogen ratios in rain gardens yield greater NO₃⁻ removal efficiencies, and Warneke et al. (2011) who observed that bioavailable carbon in the pore water limited NO₃⁻ removal rates in microcosms that contained high initial NO₃⁻ concentrations. The results of this study indicate that many factors control NO₃⁻ transformations in denitrifying biofilters. The findings from this study are currently being incorporated into a mathematical model to gain further insight into the performance of stormwater denitrifying biofilters (Lynn, 2014).

Understanding pore water DOC concentrations during ADCs (Fig. 2a) and storm events (Fig. 3b) can be useful in identifying the underlying processes that occur in denitrifying stormwater biofilters. Higher pore water and mean effluent DOC concentrations compared with influent DOC concentrations indicate that a DOC dissolution process, such as hydrolysis, was occurring during ADCs and during operation. Aerobic and anaerobic bacteria and fungi can hydrolyze solid organic substrates such as wood (Malherbe and Cloete, 2002) to VFAs which dissolves into and increases the pore water DOC concentration. In this study, pore water DOC concentrations (Fig. 2a) increased with increasing ADCs up to 16 days, indicating that the wood chip DOC dissolution rate usually exceeded the DOC utilization rate during ADCs. The decrease in pore water pH may also be an indicator of VFA production. The decrease in pore water DOC concentrations at very long dormant periods (30 days) may indicate that DOC dissolution rates decreased over time (Malherbe and Cloete, 2002; Cameron and Schipper, 2010; Schmidt and Clark, 2012) or that other microbial processes, such as methanogenesis consumed DOC. Methane emissions have been observed in denitrifying bioreactors used for wastewater treatment (Warneke et al., 2011), but methanogenic processes have not been studied in denitrifying stormwater biofilters. Nonetheless, higher mean effluent compared to influent DOC concentrations were observed during storm events (Fig. 3b). Future research on this topic is warranted because higher DOC loadings could cause an increase in biological oxygen demand (BOD), which may inhibit DO replenishment in downstream ecosystems. Problems with BOD export may be a particular concern during start up (Lynn et al., 2015) and may be reduced by flushing the media before installation.

To further illustrate the effect of ADCs on the dynamic relationship between NO₃⁻ and DOC, mean concentration data from the first three samples (Table 2) collected during the 0-day ADC (SE No. 5) and 8-day ADC (SE No. 4) normal storm events are shown in Figure 7a and b. The highest NO₃⁻ removal efficiencies (all at least 94%) and effluent DOC concentrations were observed in the initial sample taken from all columns and both ADC storm events. After the initial pore volume was flushed out, an increase in NO₃⁻ (Fig. 7a) and decrease in DOC (Fig. 7b) concentrations were observed from the second samples during both storm events. When the second and third samples were taken, significantly lower NO₃⁻ (second sample p-value=0.042 and third sample p-value=0.023) and significantly higher DOC (0.014 and 0.002) concentrations were observed in the effluent samples taken during the 8-day ADC compared with the 0-day ADC storm event. Similar high NO₃⁻ removal efficiencies (Table 3) from the 4+ADC normal storm events were observed.

FIG. 7.

Mean effluent NO₃⁻ (a) and DOC (b) concentration data from the first three samples taken during the 0- (SE No. 5) and 8 (SE No. 4)-day antecedent dry conditions (ADC) normal storm events from the 30, 45 and 60 cm columns. Error bars represent standard deviation.

Sulfate, NO₃⁻, and DOC concentration data are useful for understanding other terminal electron-accepting processes that occur during ADCs (Fig. 2a, b) and storm events (Fig. 3a, d). Over the first 16 days of dormancy, DOC (Fig. 2a) and SO₄²⁻ (Fig. 2b) concentration data were inversely correlated (DOC vs. 1/SO₄²⁻-S; r²=0.78). Moreover, when effluent NO₃⁻ concentrations (Fig. 3a) were present during the storm events, low or no removal of SO₄²⁻ (Fig. 3d) was observed (excluding the effects of mixing). These results can be explained by the combined effects of hydrolysis and biological SO₄²⁻ reduction. Sulfate-reducing bacteria can utilize SO₄²⁻ and DOC to produce gaseous hydrogen sulfide when NO₃⁻ is absent (Yamashita et al., 2011; Tang et al., 2013). Previous studies of denitrifying bioreactors for wastewater treatment had conflicting findings as to whether or not biological SO₄²⁻ reduction occurs in these systems (Elgood et al., 2010; Warneke et al., 2011; Yamashita et al., 2011). One concern with SO₄²⁻ reduction in denitrifying biofilters with long ADCs is the potential for odor production; however, one benefit may be metal precipitation (Rittmann and McCarty, 2001).

TKN (primarily as organic nitrogen) and DOC concentration data are useful in understanding how different organic constituents are released at varying redox conditions. During storm events (i.e., aerobic/anoxic), effluent TKN concentrations (Fig. 4a) decreased with decreasing hydraulic loading rate possibly because biofilm scouring was reduced. During ADCs (i.e., anoxic/anaerobic), pore water DOC concentrations (Fig. 2a) generally increase with increasing ADCs, but the increase in pore water TKN concentrations (Fig. 2d) is not correlated with ADCs. This indicates that different processes control TKN and DOC release from the wood chips or biofilm. Similar results were observed by Schmidt and Clark (2012) who observed no correlation between TKN and DOC in a denitrification wall for wastewater treatment; however, Peterson et al., (2015) observed similar trends between TKN and DOC in a denitrifying biofilter that was evaluated under alternating saturated and unsaturated conditions.

Phosphate concentration data show that, in general, PO₄³⁻ was retained during operational periods (Fig. 3c) and released during dormant periods (Fig. 2c) or when hydraulic loading rates were low (<3.5 cm/h) and NO₃⁻ was absent from the effluent. These results indicate that PO₄³⁻ transformations in stormwater biofilters are influenced by changes in microbial factors and/or redox conditions. The results are consistent with prior studies, which have documented increases in dissolved phosphorus after the development of reducing conditions in submerged soils and sediments, possibly due to dissolution of iron oxyhydroxides (Shober and Sims, 2009).

Conclusions

Processes that occur in submerged denitrifying stormwater biofilters under intermittent conditions were investigated using a gravel–wood media mixture. Based on the results from this study, the following conclusions were drawn:

• During dormant conditions, NO₃⁻ was almost completely reduced, DOC and PO₄³⁻ concentrations increased and SO₄²⁻ concentrations decreased in the pore water.

• During operational conditions, effluent DOC and TKN concentrations were higher and effluent PO₄³⁻ concentrations were lower than the influent.

• Higher NO₃⁻ removal efficiencies were observed with increasing ADCs, decreasing hydraulic loading rates or decreasing operation time.

• When NO₃⁻ was present in the effluent, up to 100% PO₄³⁻ removal was observed.

• Data from this study are currently being used to develop a nitrogen removal model for denitrifying stormwater biofilters.

Footnotes

Acknowledgments

The authors would like to thank the staff from Sarasota County, FL for providing the eucalyptus wood chips. This article is based upon the work supported by the Environmental Protection Agency Center for Reinventing Aging Infrastructure for Nutrient Management (RD835569) and the National Science Foundation (Grant Number 0965743). Any opinions, findings, and conclusions or recommendations expressed in this article are those of the author(s) and do not necessarily reflect the views of the funding agencies.

Author Disclosure Statement

No competing financial interests exist.

References

Barrett

M.E.

, Limouzin

, and Lawler

D.F.

(2013). Effects of media and plant selection on biofiltration performance. J. Environ. Eng. ASCE, 139, 462.

Cameron

S.G.

, and Schipper

L.A.

(2010). Nitrate removal and hydraulic performance of organic carbon for use in denitrification beds. Ecol. Eng., 36, 1588.

Cho

, Yoon

, Song

, and Ahn

(2011). The effects of antecedent dry days on the nitrogen removal in layered soil infiltration systems for storm run-off control. Environ. Technol., 32, 747.

Chu

, and Wang

(2013). Denitrification performance and biofilm characteristics using biodegradable polymers PCL as carriers and carbon source. Chemosphere, 91, 1310.

Chun

J.A.

, Cook

R.A.

, Eheart

J.W.

, and Kang

M.S.

(2009). Estimation of flow and transport parameters for woodchip-based bioreactors: I. laboratory-scale bioreactor. Biosyst. Eng., 104, 384.

Cohen

, and Metzner

A.B.

(1981). Wall effects in laminar flow of fluids through packed beds. Am. Inst. Chem. Eng. J., 27, 705.

Collins

K.A.

, Lawrence

T.J.

, Stander

E.K.

, Jontos

, Kaushal

S.S.

, Newcomer

T.A.

, Grimm

N.B.

, and Eckberg

M.L.C.

(2010). Opportunities and challenges for managing nitrogen in urban stormwater: A review and synthesis. Ecol. Eng., 36, 1507.

Dietz

M.E.

, and Clausen

J.C.

(2006). Saturation retention in a rain garden. Environ. Sci. Technol., 40, 1335.

Elgood

, Robertson

W.D.

, Schiff

S.L.

, and Elgood

(2010). Nitrate removal and greenhouse gas production in a stream-bed denitrifying bioreactor. Ecol. Eng., 36, 1575.

10.

Ergas

S.J.

, Sengupta

, Siegel

, Pandit

, Yao

, and Yuan

(2010). Performance of nitrogen-removing bioretention systems for control of agricultural runoff. J. Environ. Eng. ASCE, 136, 1105.

11.

Grebel

J.E.

, Mohanty

S.K.

, Torkelson

A.A.

, Boehm

A.B.

, Higgins

C.P.

, Maxwell

R.M.

, Nelson

K.L.

, and Sedlak

D.L.

(2013). Engineered infiltration systems for urban stormwater reclamation. Environ. Eng. Sci., 30, 437.

12.

Hatt

B.E.

, Fletcher

T.D.

, and Deletic

(2007). Treatment performance of gravel filter media: Implications for design and application of stormwater infiltration systems. Water Res., 41, 2513.

13.

Kim

, Seagren

E.A.

, and Davis

A.P.

(2003). Engineered bioretention for removal of nitrate from stormwater runoff. Water Environ. Res., 75, 355.

14.

Lee

L.Y.

, Tan

, Wu

, Yeo

S.K.Q.

, and Ong

S.L.

(2013). Nitrogen removal in saturated zone with vermicompost as organic carbon source. Sustain. Environ. Res., 23, 85.

15.

LeFevre

G.H.

, Paus

K.H.

, Natarajan

, Gulliver

J.S.

, Novak

P.J.

, and Hozalski

R.M.

(2014). Review of dissolved pollutants in urban storm water and their removal and fate in bioretention cells. J. Environ. Eng. ASCE, 141, 04014050.

16.

Lucas

W.C.

, and Greenway

(2011). Hydraulic response and nitrogen retention in bioretention mesocosms with regulated outlets: Part II-Nitrogen retention. Water Environ. Res., 83, 703.

17.

Lynn

T.J.

(2014). Evaluation and modeling of internal water storage zone performance in denitrifying bioretention systems. Graduate Theses and Dissertations. Available at: http://scholarcommons.usf.edu/etd/5260 (accessed April 3, 2015).

18.

Lynn

T.J.

, Yeh

D.H.

, and Ergas

S.J.

(2015). Performance and longevity of denitrifying wood-chip biofilters for stormwater treatment: A microcosm study. Environ. Eng. Sci., 32, 321.

19.

Madigan

M.T.

, Martinko

J.M.

, Dunlap

P.V.

, and Clark

D.P.

(2009). Brock Biology of Microorganisms. 12th edition. San Francisco, CA: Pearson Benjamin Cummings.

20.

Malherbe

, and Cloete

T.E.

(2002). Lignocellulose biodegradation: Fundamentals and applications. Rev. Environ. Sci. Biotechnol., 1, 105.

21.

New Jersey Department of Environmental Protection (NJDEP). (2014). New Jersey Stormwater Best Management Practices Manual. Chapter 9.9. Division of Watershed Management. September. Trenton, NJ: State of New Jersey.

22.

Peterson

I.J.

, Igielski

, and Davis

A.P.

(2015). Enhanced denitrification in bioretention using woodchips as an organic carbon source. J. Sustain. Water Built Environ. 04015004.

23.

Rice

, et al. (2012). Standard Methods for the Examination of Water and Wastewater. 22nd edition. Washington: American Public Health Association, American Water Works Association, and Water Environment Federation.

24.

Rittmann

B.E.

, and McCarty

P.L.

(2001). Environmental Biotechnology: Principles and Applications. New York, NY: McGraw-Hill.

25.

Schmidt

C.A.

, and Clark

M.W.

(2012). Efficacy of a denitrification wall to treat continuously high nitrate loads. Ecol. Eng., 42, 203, 211.

26.

Schmidt

C.A.

, and Clark

M.W.

(2013). Deciphering and modeling the physicochemical drivers of denitrification rates in bioreactors. Ecol. Eng., 60, 276, 288.

27.

Shober

A.L.

, and Sims

J.T.

(2009). Evaluating phosphorus release from biosolids and manure-amended soils under anoxic conditions. J. Environ. Qual., 38, 309.

28.

Smith

D.P.

(2008). Sorptive media biofiltration for inorganic nitrogen removal from storm water. J. Irrig. Drain. Eng. ASCE, 134, 624.

29.

Smith

(2003). Eutrophication of freshwater and coastal marine ecosystems. Environ. Sci. Pollut. Res., 10, 126.

30.

Southwest Florida Water Management District (SWFWMD). (2013). Environmental Resource Permit Applicant's Handbook Volume II. Part IV 4.1(b). October 1. Brooksville, FL: SWFWMD.

31.

Tang

, Ontiveros-Valencia

, Feng

, Zhou

, Krajmalnik-Brown

, and Rittmann

B.E.

(2013). A biofilm model to understand the onset of sulfate reduction in denitrifying membrane biofilm reactors. Biotechnol. Bioeng., 110, 763.

32.

United States Environmental Protection Agency (USEPA). (1997). Method 300.1: Methods for the Determination of Inorganic anions in Drinking Water Using Ion Chromatography. EPA/815-R-00-014. Cincinnati, OH: EPA Office of Research and Development.

33.

United States Environmental Protection Agency (USEPA). (2000). National Water Quality Inventory. 305(b) List. Washington, DC: EPA Office of Water.

34.

Urban Drainage and Flood Control District (UDFCD). (2010). Urban Storm Drainage Criteria Manual Volume 3. Chapter 4.3.T-6. November. Denver, CO: Water Resources Publications, LLC.

35.

Warneke

, Schipper

L.A.

, Bruesewitz

D.A.

, McDonald

, and Cameron

(2011). Rates, controls and potential adverse effects of nitrate removal in a denitrification bed. Ecol. Eng., 37, 511.

36.

Yamashita

, Yamamota-Ikemoto

, and Zhu

(2011). Sulfate-reducing bacteria in a denitrification reactor packed with wood as a carbon source. Bioresour. Technol., 102, 2235.

37.

Yang

, McCoy

E.L.

, Grewal

P.S.

, and Dick

W.A.

(2010). Dissolved nutrients and atrazine removal by column-scale monophasic and biphasic rain garden model systems. Chemosphere, 80, 929.

38.

Zhang

, Rengel

, Liaghati

, Anoniette

, and Meney

(2011). Influence of plant species and submerged zone with carbon addition on nutrient removal in stormwater biofilter. Ecol. Eng., 37, 1833.

39.

Zinger

, Deletic

, and Fletcher

T.D.

(2007). The effect of various intermittent wet-dry cycles on nitrogen removal capacity in biofilter systems. Presented at the Conference on Rainwater and Urban Design. Sydney, Australia. August 21–23.