Nitrate reduction in a simulated free-water surface wetland system

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Nitrate reduction in a simulated free-water surface wetland system Teresa M. Misiti, Malek G. Hajaya, Spyros G. Pavlostathis* School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0512, USA

article info

abstract

Article history:

The feasibility of using a constructed wetland for treatment of nitrate-contaminated

Received 13 June 2011

groundwater resulting from the land application of biosolids was investigated for a site

Received in revised form

in the southeastern United States. Biosolids degradation led to the release of ammonia,

7 August 2011

which upon oxidation resulted in nitrate concentrations in the upper aquifer in the range

Accepted 11 August 2011

of 65e400 mg N/L. A laboratory-scale system was constructed in support of a pilot-scale

Available online 22 August 2011

project to investigate the effect of temperature, hydraulic retention time (HRT) and nitrate and carbon loading on denitrification using soil and groundwater from the biosolids

Keywords:

application site. The maximum specific reduction rates (MSRR), measured in batch assays

Denitrification kinetics

conducted with an open to the atmosphere reactor at four initial nitrate concentrations

Groundwater

from 70 to 400 mg N/L, showed that the nitrate reduction rate was not affected by the initial

Constructed wetland

nitrate concentration. The MSRR values at 22
C for nitrate and nitrite were 1.2  0.2 and

Carbon loading

0.7  0.1 mg N/mg VSSCOD-day, respectively. MSRR values were also measured at 5, 10, 15

Nitrate loading

and 22
C and the temperature coefficient for nitrate reduction was estimated at 1.13.

Temperature effect

Based on the performance of laboratory-scale continuous-flow reactors and model simulations, wetland performance can be maintained at high nitrogen removal efficiency (>90%) with an HRT of 3 days or higher and at temperature values as low as 5
C, as long as there is sufficient biodegradable carbon available to achieve complete denitrification. The results of this study show that based on the climate in the southeastern United States, a constructed wetland can be used for the treatment of nitrate-contaminated groundwater to low, acceptable nitrate levels. ª 2011 Elsevier Ltd. All rights reserved.

1.

Introduction

World-wide, nitrate is among the most common groundwater contaminants, mainly introduced into the environment from agricultural activities related to the excessive use of nitratecontaining fertilizers and manure (Burkart and Stoner, 2002; Murgulet and Tick, 2009; Rivett et al., 2008). In addition, as an alternative to landfilling, biosolids generated by the anaerobic digestion of municipal primary and waste activated sludge or other stabilization processes are in some cases land

applied to serve as a nutrient source for plant growth as well as to enrich the soil in organic matter. In the case of land application of biosolids, ammonia, which is either already present or produced as a result of biosolids degradation on site, is utilized as a nitrogen source for plant growth. When biosolids are land applied at recommended agronomic frequencies and rates, ammonia is not anticipated to be of environmental concern. However, when biosolids are applied in excess of recommended rates or during non-growing seasons, excess ammonia is oxidized to nitrate, which can

* Corresponding author. Tel.: þ1 404 894 9367; fax: þ1 404 894 8266. E-mail address: [email protected] (S.G. Pavlostathis). 0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2011.08.019

5588

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

then leach into the groundwater (Surampalli et al., 2008; US EPA, 2000a). A wastewater treatment plant located in the southeastern United States, which services an area with a total population of approximately 230,000, generates biosolids at approximately 5.4  103 dry metric tons/year. A portion of the biosolids had been land applied to an agricultural field adjacent to the wastewater treatment plant and served as a source of nutrients for crop production. As a result of longterm biosolids application, groundwater nitrate concentrations in the upper aquifer of this site reached levels between 65 and 400 mg NO 3 eN/L, which are above the regulated limit of 10 mg NO 3 eN/L (Maltais-Landry et al., 2009c; US EPA, 2009). Nitrate and nitrite concentrations higher than the regulated limit present health concerns as they are toxic to humans and livestock (US EPA, 2009). As a remediation approach, the wastewater treatment facility proposed groundwater pumping and development of an overland-flow wetland system for the treatment of the nitratecontaminated groundwater before being released into the nearby river. >Wetland-based treatment systems are commonly used for the biological removal of nitrogen, phosphorus, sulfur, heavy metals and other pollutants, acting as a buffer between the pollution source and the natural aquatic ecosystem (Bachand and Horne, 2000; Kadlec and Wallace, 2009; Kjellin et al., 2007; Maltais-Landry et al., 2009c). Among all the nitrogen transformation processes taking place in wetlands, the one related to nitrate removal is denitrification, i.e., the reduction of nitrate to dinitrogen (N2). The effectiveness of wetlands is largely affected by the biological activity and temperature of the wetland location. Treatment wetlands are often constructed in regions with moderate to cold climates that experience large seasonal temperature variations (Kadlec and Reddy, 2001). Wetlands are complex biological systems and their performance depends on a number of chemical, physical and biological processes (US EPA, 2000b). For the biologicallymediated nutrient removal in wetlands, both empirical and mechanistic models have been used. Kadlec and Reddy (2001) assessed the effect of temperature on treatment wetlands by using a simple model where all reaction mechanisms are grouped into a pseudo-first order removal rate. On the other hand, Kjellin et al. (2007) used Menten and first-order kinetics to model nitrate removal in wetland sediment. Denitrification kinetics have also been modeled using both the single- and dual-substrate Monod model (Hajaya et al., 2011; Heinen, 2006; Kornaros et al., 1996). A three-cell pilot-scale, free-water surface wetland system was constructed at the above-mentioned biosolids application site to test the effectiveness of treating the nitratecontaminated groundwater using various carbon sources during the system’s start-up period while vegetation was being established. One cell served as a control (i.e., no external carbon addition) and the other two cells used either MicroC G or hay as carbon source, respectively. The pilot-scale system was designed to demonstrate the efficiency of microbial nitrate reduction under conditions open to the atmosphere and as affected by various environmental and operational conditions. In particular, the high groundwater

nitrate concentrations and the presence of oxygen in the freesurface wetland system presented conditions which may affect nitrogen removal efficiency. In support of the pilot-scale constructed wetland demonstration project, a continuous-flow, laboratory-scale system was built to assess the treatment of nitratecontaminated groundwater as a function of carbon (i.e., electron donor) and nitrate loading, hydraulic retention time, and temperature. The objective of the laboratory study was to determine the feasibility of using a constructed wetland for treatment of the nitrate-contaminated groundwater through a series of batch and continuous-flow nitrate reduction tests using MicroC G as the electron donor, as well as soil and nitrate-contaminated groundwater from the biosolids application site. Nitrate and nitrite reduction rate estimates, resulting from the laboratory batch tests, were used in a mathematical model to simulate the nitrogen removal efficiency of continuous-flow, free-water surface systems as a function of both operational and environmental conditions.

2.

Materials and methods

2.1.

Sample collection and characterization

Groundwater and surface soil were collected at the biosolids application site located in the southeastern United States. The groundwater samples were stored in plastic containers under refrigeration (4
C). The soil sample was passed through a US No. 10 sieve, spread thin to air dry for 24 h at room temperature, and then stored in covered plastic containers at room temperature (22e24
C). All samples were characterized by measuring pH, soluble and total chemical oxygen demand (sCOD and tCOD), dissolved organic carbon (DOC), moisture   content, NHþ 4 eN, NO3 eN, NO2 eN, and other ions. To measure soil pH, DOC, soluble COD, ammonia and ions, a soil filtrate solution was prepared by adding 5 g of dry soil to 300 mL of deionized (DI) water and mixing for 1 day at room temperature. The soil solution was then centrifuged at 10,000 rpm for 30 min. The results of soil and groundwater characterization are shown in Table 1. Both the soil and groundwater samples were slightly acidic, with pH values of 4.4 and 5.7, respectively. The soil sample was mostly inorganic matter (w95%) and did not contribute significant soluble COD or ions to the solution (Table 1). MicroC G, a plant-derived complex carbohydrate mixture, was obtained from Environmental Operating Solutions Inc. (Bourne, MA) and used as the electron donor and carbon source in all experiments. A MicroC G solution was prepared by a 1000-fold dilution in DI water and analyzed for pH, DOC, soluble COD and ions. The concentrated MicroC G stock and the diluted solution were stored in the dark at 22 and 4
C, respectively. The MicroC G solution was acidic with a pH of 3.9 and the measured COD of the undiluted solution was approximately 640 g/L, which agrees closely with the technical specifications provided by the manufacturer (Table 1). MicroC G did not contain any anions or ammonia, is soluble in water and has a freezing point of 8
C (EOS, 2008).

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

Table 1 e Characteristics of soil, groundwater and MicroC G samples used in this study. Parameter pH Water content (%) Dry weight (%) Organic matter (% of dry) DOC (filtrate; mg C/L) Soluble COD (filtrate; mg/L) Total COD (mg/g dry weight) Ions (filtrate) Chloride (mg Cl/L) Nitrite (mg N/L) Nitrate (mg N/L) Sulfate (mg S/L) Phosphate (mg P/L) Ammonia (filtrate; mg N/L)

Soil 4.4 3.9  0.1a 96.1  0.1 4.9  0.1 9.8  0.4 14.5  7

Groundwater

MicroC G

5.7

3.9

9.1  1.5 58.1  4.8

411  32c 642  41c

14.4  0.8 ND 69.3  1.3 28.4  0.1 ND ND

ND ND ND ND ND ND

68.3  5.2

NDb ND 0.3 0.2 1.3 ND

a Mean  standard deviation (n ¼ 3). b ND, not detected. c 1000-fold diluted solution.

2.2.

Continuous-flow laboratory-scale system

Three continuous-flow, laboratory-scale reactors were constructed in order to quantify nitrate removal under different operational conditions including, effect of initial nitrate concentration (Run 1), COD:N ratio (Run 2) and temperature (Run 3). Three 15-L cubic Plexiglas reactors were filled with 10.5 kg of soil and approximately 9 L of nitrate-bearing groundwater and were kept static for 1 day in order to expel all air from the soil and uniformly wet the soil. The water column depth was approximately 12 cm. Run 1 was conducted in a single-compartment continuous-flow reactor; however, in order to more closely simulate the flow regime of a full-scale wetland system, two baffles were inserted into the reactors for Run 2 and 3, thus dividing the liquid volume into three, equal-volume compartments, resulting in a flow regime that simulated 1.5e2 continuous-flow stir tank reactors (CSTRs) in series (see Supplementary Material, Text S1 and Fig. S1). Plastic reservoirs filled with groundwater were attached to peristaltic pumps (Masterflex; Cole-Parmer) and the nitratebearing groundwater was fed to the reactors continuously at a specific flow rate to achieve the target hydraulic retention times (HRTs). MicroC G was used as the electron donor and carbon source and a 200 g COD/L diluted solution was fed using a positive displacement pump (Fluid-Metering, Inc.) at predetermined flow rates to achieve the target COD:N ratios. The first continuous-flow test, Run 1, was designed to investigate the effect of HRT and initial nitrate concentration on system nitrate removal. Site groundwater, with a nitrate concentration of 69 mg NO 3 eN/L, was fed to the system and when the influent nitrate concentration was increased to 150 NO 3 eN/L, the site groundwater was amended with a stock solution of NaNO3. The MicroC G was fed every 2 h with the help of an electronic timer (ChronTrol) at a flow rate dependent on the HRT to maintain a COD:N ratio of 6. Run 1

5589

was conducted at room temperature (22e24
C) and lasted for a total of 121 days. The second continuous-flow test, Run 2, was conducted to investigate the effect of the COD:N ratio on nitrate removal in the system. Similarly to Run 1, site groundwater was fed with a nitrate concentration of 69 mg NO 3 eN/L. The MicroC G was fed every 2 h at a flow rate dependent on the HRT and target COD:N ratios of 6, 5, 4, 3, 2, 1 and 0.5. Run 2 was conducted at room temperature (22e24
C) and lasted for a total of 250 days. The last continuous-flow test, Run 3, was designed to investigate the effect of temperature on the nitrate reduction. The reactor was housed in a controlled temperature room and its temperature was stepwise decreased from 22 to 5
C. The rate of temperature change between the four target temperature values was 2
C/day. The system performance was assessed at four temperature values: 5, 10, 15 and 22
C. Site groundwater and MicroC G were fed at flow rates dependent on the HRT and to maintain a COD:N ratio of 6. At 5
C, the MicroC G had low solubility and a mixer was installed to incorporate the feed into the groundwater (see Section 3.1.3 for more details). Run 3 lasted for a total of 220 days. In all continuous-flow runs, overflow reactor effluent was periodically collected and analyzed for nitrate, nitrite, ammonia, pH, DOC, and soluble COD.

2.3.

Batch assays

A 15-L cubic Plexiglas reactor was used in all batch assays open to the atmosphere, filled with 10.5 kg of soil and approximately 9 L of nitrate-bearing groundwater. Batch assays were performed to investigate the effect of initial nitrate concentration and temperature on denitrification kinetics. To investigate the effect of initial nitrate concentration, the site groundwater, containing approximately 69 mg NO 3 eN/L, was used, and in selected batch assays was amended with a volume of a NaNO3 stock solution to achieve initial concentrations of 150, 300, and 400 mg NO 3 eN/L. This batch assay was conducted at room temperature (22e24
C). To investigate the effect of temperature on nitrate reduction, the site groundwater was amended with a volume of NaNO3 stock solution to achieve an initial concentration of 150 mg NO 3 eN/L at all temperature values tested: 5, 10, 15 and 22
C. The reactor was housed in a controlled temperature room and its temperature decreased stepwise from 22 to 5
C. The rate of temperature change between the four target temperature values was 2
C/day. This batch assay was conducted simultaneously with the above-described continuous-flow Run 3. MicroC G was used as the electron donor and carbon source in all batch assays at a COD:N ratio of 6. In order to maintain similar initial biomass concentrations in the soil layer, after each batch assay was complete, the reactor was drained, backflushed with deionized water three times and nitrate-bearing groundwater once before being refilled with nitrate-bearing groundwater for the next batch assay. After the first batch assay, which assessed the effect of initial nitrate concentration on nitrate reduction, was completed at room temperature, approximately 1 inch of the top soil layer was replaced with fresh site soil and the reactor was transferred to the controlled temperature room to assess the effect of temperature on denitrification kinetics. Similarly to the first batch assay, the

5590

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

reactor was drained and back-flushed in between each batch assay at the various target temperatures. In all open batch assays, liquid samples were taken daily and nitrate, nitrite, pH and periodic COD concentrations were measured. In order to assess the effect of oxygen on the denitrification kinetic rates measured in the open to the atmosphere reactor, a closed batch assay was conducted using duplicate 160-mL serum bottles sealed with rubber stoppers and aluminum crimps and flushed with helium. In each serum bottle, 5 g dry soil and 100 mL nitrate-containing groundwater were added. A volume of different NaNO3 stock solutions were added to each serum bottle resulting in initial nitrate concentrations ranging from 70 to 400 mg NO 3 eN/L. Different volumes of a MicroC G stock solution were added to each bottle resulting in a COD:N ratio of 6. Incubation was carried out at room temperature (22e24
C). During the incubation period, the following parameters were measured: nitrate, nitrite, gas production and gas composition (CO2, NO, N2O and N2).

2.4. Batch denitrification kinetics and parameter estimation For this work, a two-step denitrification model (nitrate to nitrite to dinitrogen) was used. Monod kinetic equations were used to describe microbial growth utilizing nitrate and nitrite in all batch assays. Assuming that nitrate and nitrite are the limiting substrates (electron donor in excess with a COD/N ratio of 6), the following differential equations were used: !

dSNO3 kNO3 SNO3 XNOx ¼ dt KSNO3 þ SNO3 dSNO2 ¼ dt

! kNO3 SNO3 XNOx  KSNO3 þ SNO3

dXNOx ¼ dt

! YNO3 kNO3 SNO3 XNOx þ KSNO3 þ SNO3

(1) ! kNO2 SNO2 XNOx KSNO2 þ SNO2

(2)

! YNO2 kNO2 SNO2 XNOx  bXNOx KSNO2 þ SNO2 (3)

where SNO3 , SNO2 , and XNOx are nitrate, nitrite and denitrifiers  concentrations (mg NO 3 eN/L, mg NO2 eN/L and mg VSSCOD/L, respectively); t is time (days); kNO3 and kNO2 are the nitrate and nitrite maximum specific reduction rates (MSRR; mg N/ gVSSCOD day); KSNO3 and KSNO2 are the nitrate and nitrite halfsaturation constants (mg N/L); YNO3 and YNO2 are the theoretical yield coefficients (g VSSCOD/mg N); and b is the microbial decay coefficient (day1). Based on bioenergetic calculations, the yield coefficients for nitrate (YNO3 ) and nitrite (YNO2 ) used for all simulations were calculated to be 1.14 and 1.72 g VSSCOD/g N, respectively (Rittmann and McCarty, 2001). The decay rate values for denitrifiers are generally in the range of 0.05e0.15 day1 (Rittmann and McCarty, 2001; Tchobanoglous et al., 2003). A microorganism decay rate of 0.1 day1 was chosen for all simulations; however, preliminary simulations using values of 0.05 and 0.15 day1 resulted in small variations in nitrate concentration patterns. Given the fact that the initial, active denitrifiers concentration in the soil (XNOx ) was not measurable, for each set of

experimental data, an initial biomass concentration was chosen to fit the nitrate experimental data. Typical KS values for denitrification have been reported in the range of 0.8e153 mg N/L (Kjellin et al., 2007; Tugtas and Pavlostathis, 2007; Zumft, 1997). The half-saturation constants for nitrate and nitrite (KSNO3 and KSNO2 ) were estimated for each batch assay based on reported value ranges to best fit the experimental data. Parameter estimation was conducted following a previously reported procedure (Hajaya et al., 2011). Parameters sensitivity and identifiability analysis was performed using the Berkeley Madonna Software Version 8.3 (Macey and Oster, 2006) and Matlab ode15 solver (MATLAB 7.0.1; The Mathworks, Natick, MA) following the procedure described by Gujer (2008). The Fit ODE toolbox in Igor Professional v.5.057 (WaveMetrics, Inc., Lake Oswego, OR) was used to calculate the standard deviation values for the evaluated parameters. The effect of temperature on the MSRR was quantified by fitting the resulting MSRR values at each temperature to the modified Arrhenius model using nonlinear regression (SigmaPlot, Version 10.0 software; Systat Software Inc., San Jose, CA, USA): kT2 ¼ kT1 qðT2 T1 Þ T2

(4) T1

where k and k are MSRR values (mg N/mg VSSCOD-day) at two different temperatures (
C) and q is the dimensionless temperature coefficient.

2.5.

Continuous-flow model

The denitrification kinetic rates estimated in the batch assays can be used to simulate and predict the performance of the continuous-flow wetland system under various operational conditions. Based on a modified version of the dual-substrate Monod model presented by Kornaros et al. (1996) and a system mass balance, the continuous-flow system was modeled using the series of differential equations (5) through (8) for nitrate and nitrite reduction, cell growth and electron donor utilization as follows: !    SNO3 ;o  SNO3 dSNO3 kNO3 SNO3 C  ¼ XNOx s dt kSNO3 þ SNO3 KC þ C !    SNO2 ;o  SNO2 dSNO2 kNO3 SNO3 C  ¼ XNOx s dt kSNO3 þ SNO3 KC þ C !  kNO2 SNO2 C XNOx  KSNO2 þ SNO2 KC þ C !  dC ðCo  CÞ YNO3 kNO3 SNO3 C XNOx ¼  kSNO3 þ SNO3 dt s KC þ C !  YNO2 kNO2 SNO2 C XNOx  KSNO2 þ SNO2 KC þ C !  dXNOx ðXNOx ;o  XNOx Þ YNO3 kNO3 SNO3 C ¼ XNOx þ dt kSNO3 þ SNO3 s KC þ C !  YNO2 kNO2 SNO2 C XNOx  bXNOx þ KSNO2 þ SNO2 KC þ C

(5)

(6)

(7)

(8)

5591

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

DOC, COD, pH, dissolved oxygen (DO), water content, NHþ 4 eN,  NO 3 eN, NO2 eN, and other ions were measured following procedures outlined in Standard Methods (APHA, 2005). Soluble COD was measured by the HACH colorimetric method and the total COD was measured by the Open Reflux Method. DOC was measured using a Shimadzu Total Organic Carbon (TOC) Analyzer equipped with an infrared detector for CO2 measurement (Shimadzu Scientific Instruments, Inc., Columbia, MD). DO was measured using the polarographic method with a YSI Model 58 oxygen meter in conjunction with a YSI 5750 oxygen probe (Yellow Springs Instruments, Yellow Springs, OH). Anions were measured using a Dionex DX-100 ion chromatography unit (Dionex Corporation, Sunnyvale, CA) equipped with a conductivity detector, a Dionex IonPac AG14A (4  50 mm) precolumn, and a Dionex IonPac AS14A (4  250 mm) analytical column. The unit was operated in autosuppression mode with 1 mM NaHCO3/8 mM Na2CO3 eluent at a flow rate of 1 mL/min. The minimum detection limit for nitrate and nitrite was 0.05 and 0.1 mg N/L, respectively. Gas composition was determined by a gas chromatography (GC) unit (Agilent Technologies, Model 6890N; Agilent Technologies, Inc., Palo Alto, CA) equipped with two columns and two thermal conductivity detectors. Dinitrogen (N2) was separated with a 15 m HP-Molesieve fused silica, 0.53 mm i.d. column (Agilent Technologies, Inc.). Carbon dioxide (CO2), nitric oxide (NO) and nitrous oxide (N2O) were separated with a 25 m Chrompac PoraPLOT Q fused silica, 0.53 mm i.d. column (Varian, Inc., Palo Alto, CA). Helium was used as the carrier gas at a constant flow rate of 6 mL/min. The 10:1 split injector was maintained at 150
C, the oven was set at 40
C and the detector temperature was set at 150
C. The minimum detection limit for CO2, NO, N2O and N2 was 800, 500, 7 and 50 ppmv, respectively.

Effect of HRT and influent nitrate concentration

Run 1 was performed in a single-compartment, continuousflow reactor which was operated at HRT values of 2.8, 3.5 and 5 days, while being continuously fed with groundwater at 67 mg NO 3 eN/L and MicroC G at a COD:N of 6 after the first week, during which external carbon was not added. The effluent nitrate concentration over the entire run period is shown in Fig. 1, along with other operational parameters. Upon addition of MicroC G directly to the reactor on day 8, the effluent nitrate concentration decreased and reached non-detectable levels within 3 days. For the remainder of this run, the effluent nitrate concentration did not exceed 20 mg NO 3 eN/L at any of the three hydraulic retention times tested. To illustrate that the nitrate removal follows Monod kinetics, according to which the effluent nitrate concentration is not a function of influent nitrate concentration, the influent groundwater concentration was increased on day 88e130 mg NO 3 eN/L and the concentration of the MicroC G solution changed accordingly to maintain a COD:N ratio of 6. The effluent nitrate concentration increased slightly to approximately 7 mg NO 3 eN/L until steady-state was

INFLUENT NITRATE (mg N/L)

Analytical methods

3.1.1.

150 125 100 75 50 25 0 6

HRT (Days)

2.6.

The effluent pH ranged from 6.5 to 7.5 and the DO remained below 1 mg/L after addition of MicroC G. Ammonia was not detected in any effluent samples. In all reactors, the effluent COD and DOC remained constant and low (DOC below 40 mg/L and COD below 75 mg/L). Sulfate was also periodically measured and was consistently in the range of 20e30 mg S/L in both influent and effluent samples, indicating that significant sulfate reduction was not occurring in the open to the atmosphere reactors.

EFFLUENT NITRATE (mg N/L)

where SNO3 ;o , SNO2 ;o , and XNOx ;o are influent nitrate, nitrite and biomass concentrations (mg N/L, mg N/L and mg VSSCOD/L, respectively); Co and C are the electron donor (MicroC G) concentrations in the influent and effluent (mg COD/L); KC is the half-saturation constant for the electron donor (mg COD/L); and s is the HRT (¼V/Q) (days). The KC value used in all simulations was 20 mg COD/L, a value reported for MicroC G as the electron donor for denitrification (Cherchi et al., 2009). The effluent nitrate and nitrite concentrations were simulated in Matlab at various HRT, COD:N and temperature values using the previously estimated biokinetic constants (kNO3 , kNO2 , KSNO3 , KSNO2 , YNO3 , YNO2 , and b) and initial biomass concentrations based on the denitrification kinetics described in Section 2.4, above.

4 2 0 100 90 80 70 60 50 40 30 20 10 0 0

3.

Results and discussion

3.1.

Continuous-flow reactor system performance

The effluent streams from the three continuous-flow reactors were periodically analyzed for pH, ammonia, COD and DOC.

10

20

30

40

50

60

70

80

90 100 110 120 130

TIME (Days)

Fig. 1 e Effluent nitrate concentration in a continuous-flow reactor operated at room temperature (22e24
C) at various HRT values, influent nitrate concentrations and with MicroC G at a COD:N ratio of 6:1 at/after day 8 (MicroC G was not added from day 92 to 104).

5592

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

achieved, after which the effluent concentration quickly returned to non-detectable levels (Fig. 1). In order to qualitatively evaluate the effect of carbon source on the nitrate removal in this reactor, at approximately 92 days, the MicroC G pump was turned off and only the groundwater at approximately 130 mg NO 3 eN/L was fed to the reactor at an HRT of 5 days. As shown in Fig. 1, the effluent nitrate concentration increased and reached about 88 mg NO 3 eN/L within 10 days, further demonstrating the necessity of a continuous addition of a degradable carbon source. On day 104, the groundwater pump was turned off to simulate a batch system while the MicroC G pump continued supplying carbon. A gradual decrease of the nitrate concentration to about 20 mg NO 3 eN/L in 20 days was observed. Thus, even at an elevated influent nitrate concentration, continuous addition of biodegradable organic carbon in excess of stoichiometric levels achieved high nitrate removal efficiency.

3.1.2.

Effect of COD:N ratio

Run 2 was conducted in a three-compartment, continuousflow reactor with an influent groundwater nitrate concentration kept constant at 70 mg NO 3 eN/L, an HRT of 2 and then 5 days, while the COD:N ratio was stepwise decreased to lower values. Fig. 2 shows the reactor effluent nitrate concentration along with other operational parameters. For the first 18 days, the reactor was operated at an HRT of 2 days and a COD:N ratio of 6, during which the effluent nitrate concentration decreased sharply to less than 10 mg NO 3 eN/L. The HRT was then increased to 5 days to achieve more stable operation and consistent effluent nitrate concentration of less than 10 mg NO 3 eN/L. The COD:N ratio was stepwise decreased from an initial value of 6:1 to the lowest value of 0.5:1. On day

HRT (Days)

6 4 2

COD:N

4

EFFLUENT NITRATE (mg N/L)

0 6

70

2 0

25, the COD:N ratio was decreased to 5:1, during which the effluent nitrate concentration remained below 4 mg NO 3 eN/L. When on day 75, the COD:N ratio was further decreased to 4:1, the effluent nitrate concentration increased sharply to about 27 mg NO 3 eN/L. A decrease of the COD:N ratio to 3:1 and then to 2:1 resulted in an effluent nitrate concentration ranging between 32 and 40 mg NO 3 eN/L. A further decrease of the COD:N ratio to 1:1 and then to 0.5:1 resulted in a gradual increase of the effluent nitrate concentration to 42 mg NO 3 eN/L. On day 232, the COD:N ratio was increased to 5:1, which resulted in a rapid decrease of the effluent nitrate concentration to below 4 mg NO 3 eN/L. Based on these results, for a system open to the atmosphere at ambient temperature between 22 and 24
C, influent nitrate concentration of 67 mg NO 3 eN/L, and an HRT value of 5 days, the minimum COD:N ratio is approximately 5:1 in order to achieve an effluent nitrate concentration of less than 10 mg N/L. The theoretical requirement for complete denitrification, ignoring microbial growth, is 2.85 mg COD/mg nitrateeN reduced to N2. At relatively low COD:N values, incomplete denitrification is possible, which could lead to the formation of nitric oxide (NO) and nitrous oxide (N2O), both potent greenhouse gases (Maltais-Landry et al., 2009a, 2009b). It has been reported that N2O emissions in wetlands are highly dependent on the COD:N ratio, as well as the pH, dissolved oxygen, and temperature among other parameters (Inamori et al., 2008; Wu et al., 2009). Wu et al. (2009) found that significant amounts of N2O were released from constructed wetlands at very high and very low COD:N ratios (2 > COD:N > 10), with minimum emissions at a ratio of 5:1. In the present study, in order to investigate if NO and N2O were released in the laboratory reactor due to incomplete denitrification, on days 104, 143, 192 and 215 when the continuousflow reactor was operated with a COD:N ratio of 3:1, 2:1, 1:1 and 0.5:1, respectively, gas bubbles and water were collected biweekly from the soil/water interface by using an inverted glass vial fully submerged in the water and partially imbedded into the soil. Gas bubbles released from the surface soil were collected in the vial by water displacement. Then, the vial was sealed with a stopper while under water, positioned upright and its headspace analyzed by gas chromatography after 30 min equilibration at room temperature. NO and N2O were not detected at any of the COD:N ratios tested, confirming that complete denitrification occurred, leading to the production of nitrogen gas (N2) as the main nitrate reduction process in the laboratory reactor.

60

3.1.3.

50 40 30 20 10 0 0

20

40

60

80 100 120 140 160 180 200 220 240 260

TIME (Days)

Fig. 2 e Effluent nitrate concentration in a continuous-flow reactor operated at room temperature (22e24
C), mean influent groundwater nitrate concentration of 70 mg N/L and with MicroC G at several COD:N ratios.

Effect of temperature

The three-compartment, continuous-flow reactor was housed and operated in a temperature-controlled room to simulate the effect of temperature on nitrate reduction. For the first 15 days, the continuous-flow reactor was operated at 22
C with an HRT of 2 days, during which the effluent nitrate concentration decreased sharply to less than 13 mg NO 3 eN/L (Fig. 3). The HRT was then increased to 5 days to achieve more stable operation and an effluent nitrate concentration of less than 10 mg NO 3 eN/L. On day 30 the room temperature was decreased and by day 32 reached 15
C. While at 15
C, the reactor performance did not change and the effluent nitrate concentration was kept at non-detectable levels. On day 51,

EFFLUENT NITRATE (mg N/L)

HRT (Days)

o

TEMP. ( C)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

25 20 15 10 5 0 12 10 8 6 4 2 0 70 60 50 40 30 20 10 0 0

20

40

60

80

100

120

140

160

180

200

220

TIME (Days)

Fig. 3 e Effluent nitrate concentration in a continuous-flow reactor operated at a range of temperature (22e5
C), mean influent groundwater nitrate concentration of 70 mg N/L and with MicroC G at a COD:N ratio of 6:1 (arrow indicates start of complete MicroC G and groundwater mixing; see text).

5593

for approximately another 25 days, during which period the effluent nitrate concentration remained below 2 mg N/L (Fig. 3). During this period at 5
C, biofilm formation on the reactor walls was more noticeable than during operation at temperature values above 5
C. Based on the experimental results of the temperature study, an effluent nitrate concentration of 10 mg N/L or less can be achieved even at a water temperature as low as 5
C, as long as sufficient degradable carbon is provided and well incorporated into the groundwater. It is noteworthy that the conditions used in this test differed from what was typically observed at the biosolids application site where the groundwater temperature did not change significantly throughout the year (it ranged between 18 and 20
C). In addition, during winter with ambient air temperature between 4 and 15
C, the temperature at the sediment/water interface in the pilot-scale wetland ranged from 5 to 20
C. Therefore, even during winter, the impact of temperature on the wetland performance of the pilot-scale wetland at the biosolids application site, where the lowest average monthly air temperature, usually in January, is about 8e9
C, is expected to be less drastic. Therefore, the results of the laboratory study at a water temperature as low as 5
C are conservative, but show the resilience and efficiency of the denitrification process at low temperature values.

3.2.

Denitrification kinetics

3.2.1. Effect of initial nitrate concentration under open and closed conditions the room temperature was decreased again and reached 10
C by day 55. There was a slight increase in the effluent nitrate concentration at this time, but within 24 h it returned to nondetectable levels. After the room temperature was reduced to 5
C by day 82, the reactor effluent concentration increased rapidly to a maximum 47 mg NO 3 eN/L, and then started to decrease. However, for over 35 days at 5
C, the reactor’s performance was not stable and the effluent nitrate concentration fluctuated between 10 and 35 mg NO 3 eN/L, albeit with a downwards trend (Fig. 3). In an attempt to achieve a stable effluent concentration, on day 120 the HRT was increased to 10 days. Although the effluent nitrate concentration decreased significantly at an HRT of 10 days, it continued to fluctuate between 5 and 20 mg NO 3 eN/L. Upon further observation, it was realized that MicroC G was not well mixed with the groundwater in the reactor as it was delivered intermittently by a micro pump every 2 h at the point where the groundwater was constantly pumped into the reactor (head of reactor). It appears that MicroC G has a low solubility at 5
C. Therefore, the unstable and poor performance of the reactor was attributed to lack of uniform electron donor distribution and thus availability. On day 172, a mixer was installed in the influent portion of the reactor and turned on by an electronic timer every 2 h while the MicroC G was fed, and for an additional 10 min after feeding was stopped. With intermittent mixing, even at 5
C, MicroC G was well incorporated into the reactor groundwater, which resulted in stable reactor performance with non-detectable effluent nitrate concentrations. On day 190, the HRT was returned to 5 days and the reactor operated

Denitrification at initial nitrate concentrations of 70, 140, 300 and 400 mg N/L was tested under conditions open to the atmosphere. The nitrate and nitrite concentrations in each assay over the incubation period are shown in Fig. 4. A lag period of approximately 1 day was observed in the first batch assay performed with site soil and groundwater, which was attributed to the very low active denitrifying population size of the surface soil used in these assays. Nitrate reduction proceeded immediately in all subsequent assays, indicating that some active biomass was retained in the soil despite the rinsing procedure performed in between each batch assay. Transient nitrite concentrations were observed in all assays, but after the complete removal of nitrate, nitrite was completely removed in less than 4 days, except in the first assay, in which nitrite reduction was slower and nitrite was removed in approximately 5 days. Similarly to the open batch assays, closed batch assays were conducted in serum bottles in the absence of oxygen (data not shown). A relatively low nitrate removal rate was also observed in the first 20 h of incubation, which is attributed to the low population size of active denitrifying bacteria in the soil. Significant nitrite levels were observed in series with an initial nitrate concentration of 140 mg NO 3 eN/L and above. Nevertheless, the nitrite reduction rate was fast and all series achieved complete denitrification in less than 5 days. Using the nitrate and nitrite reduction data and assuming a two-step model (nitrate to nitrite to dinitrogen), the MSRR values were estimated. Based on the data from the open batch assays and applying Monod kinetics with the biokinetic parameter values described in Section 2.4 above, the MSRR

5594

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

80

values were determined to be very similar in all assays conducted with different initial nitrate concentrations (Table 2). The best model fit to the experimental data is shown in Fig. 4. The root mean squared deviation (RMSD) at initial nitrate concentration of 70, 140, 300 and 400 mg N/L was 7.0, 11.1, 36.4 and 44.8, respectively. Sensitivity and identifiability analysis (Section 2.4, above) showed that the highest degree of sensitivity was for the MSRR, while the lowest degree of sensitivity was for the halfsaturation constant (see Supplementary Material, Fig. S2). Although the sensitivity analysis indicates that the halfsaturation constant cannot be uniquely identified from the provided data sets, the KSNO3 and KSNO2 values were estimated at 20.7  6.3 and 18.8  7.5 mg N/L for nitrate and nitrite, respectively. These values are apparent half-saturation constants, as opposed to intrinsic, which take into consideration any mass transfer limitations (Tugtas and Pavlostathis, 2007). Such limitations are expected for a system where most of the microbial activity occurred at the lower water layers, soil/water interface and within the soil matrix. Visual inspection lead to the observation that most N2 gas production was occurring in the top soil layer, indicated by the entrapment of gas bubbles and thick biofilm layer on the soil/water interface compared to a relatively clear reactor water column. When estimating denitrification kinetics for each initial nitrate concentration of 70, 140, 300 and 400 mg N/L, the initial biomass was assumed to be 5, 10, 10 and 15 mg VSS/L, respectively, due to increased retention of active biomass in the sediment despite flushing of the reactor in between each batch assay. Similarly to the open assay, the MSRR values for the closed batch assays for nitrate and nitrite were estimated to be 1.5  0.2 and 0.4  0.1 mg N/ mg VSSCOD-day, respectively. The KSNO3 and KSNO2 values were 16.1  2.8 and 20.5  6.2 mg N/L, respectively. Oxygen affects (increases) the electron donor requirement for denitrification; however, comparing the nitrate reduction rate achieved in closed systems (1.5  0.2 mg NO 3 eN/ mg VSSCOD-day) to that achieved in open to the atmosphere systems (1.2  0.2 mg NO 3 eN/mg VSSCOD-day) at an initial COD:N ratio of 6, the kinetics of nitrate reduction were not severely affected in the open to the atmosphere reactor. Therefore, as long as a bioavailable carbon source is supplied in excess of that required for the complete nitrate reduction, the nitrate reduction kinetics are not impacted by other alternative electron acceptors (e.g., oxygen for open systems) for similar systems (e.g., low mixing intensity and reaeration).

A Nitrate

60

Nitrite

40

20

0 150

B

125 100 75

NITROGEN (mg N/L)

50 25 0 300

C

250 200 150 100 50 0 400

D

300

200

100

0 0

1

2

3

4

5

6

TIME (Days) Fig. 4 e Measured (data points) and simulated (lines) nitrogen species in batch assays conducted with an open to the atmosphere reactor at an initial nitrate concentration of (A) 70, (B) 140, (C) 300 and (D) 400 mg N/L using MicroC G as the carbon source at an initial COD:N ratio of 6:1.

Table 2 e Estimated maximum specific reduction rate (MSRR; k) and half-saturation constant (KS) values for batch assays conducted with an open to the atmosphere reactor and four initial nitrate concentrations (70, 140, 300 and 400 mg N/L). Parameter kNO3 (mg N/mg VSSCOD-day) KSNO3 (mg N/L) kNO2 (mg N/mg VSSCOD-day) KSNO2 (mg N/L) a Estimate  standard deviation. b Range.

Value 1.2  0.2a 20.7  6.3 0.7  0.1 18.8  7.5

(1.0e1.5)b (16.5e30.0) (0.6e0.9) (15.0e30.0)

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

A

The laboratory reactors were static and mixing was minimal. In both the continuous-flow and batch assays, most of the microbial activity was in the soil/water interface and top soil layer where dissolved oxygen concentrations were typically less than 1 mg/L and maintained only by diffusion from the free-water surface. Alternatively, under field conditions, a higher rate of aeration is expected (e.g., wind action), which may negatively impact the nitrate reduction rate as a result of a higher competition for the carbon/electron source between oxygen and nitrate-reducing processes.

3.2.2.

B

C

D

Fig. 5 e Measured (data points) and simulated (lines) nitrogen species in batch assays conducted with an open to the atmosphere reactor maintained at (A) 22, (B) 15, (C)

5595

Effect of temperature

In order to assess the effect of temperature on denitrification kinetics, batch assays were performed at target temperatures of 5, 10, 15 and 22
C in a reactor open to the atmosphere (Fig. 5). Similarly to the previous batch assays, a lag of approximately 1 day was observed for nitrate removal. The rates of both nitrate and nitrite reduction were very similar at 22 and 15
C, with a maximum transient nitrite concentration slightly higher at 15
C. At 10
C, a significantly lower rate of nitrate and nitrite reduction was observed; however, at 5
C, the time required for the complete removal of nitrate and nitrite was more than double of that under 10
C. The nitrate and nitrite reduction rates at 5
C were much lower than at the other three temperature values, and the maximum transient nitrite concentration was the lowest. Therefore, the nitrate reduction rates were not severely affected until the temperature dropped below 10
C, which agrees with the findings of other studies on denitrification at low temperature values (Burgoon, 2001; Darbi and Viraraghavan, 2004; Hajaya et al., 2011; Lee et al., 2009). Hajaya et al. (2011) reported a decrease of more than 50% in the nitrate and nitrite reduction rates for a temperature decrease from 15 to 10
C. Studies investigating the seasonal temperature effects on denitrification in wetland systems have reported significant decrease in denitrification rates at temperature values below 15
C (Kadlec and Reddy, 2001; Poe et al., 2003). Using the nitrate reduction data, estimated biokinetic parameters and assuming a two-step denitrification model as previously discussed in Section 2.4 above, the MSRR values were estimated and are summarized in Table 3. The half-saturation constants were similar at all temperature values; KSNO3 and KSNO2 were estimated as 57.0  4.8 and 16.5  2.4 mg N/L, respectively. As expected, the MSRR increased with increasing temperature; however, the MSRR values were very similar at 15 and 22
C, which agrees with previous reports stating that the optimum temperature for nitrate reduction is closer to 15 than 22
C, and that in this temperature range, variations in temperature only slightly affect denitrification rates (Hajaya et al., 2011; Kristiansen, 1983; Lee et al., 2009). The model accurately simulated the nitrate reduction at all four temperature values; however, as temperature decreased, the discrepancy between the model and experimental nitrite concentrations increased. At lower temperature values the model overestimated the 10, and (D) 5
C with groundwater at an initial nitrate concentration of 150 mg N/L using MicroC G as the carbon source at an initial COD:N ratio of 6:1 (note the different xaxis scale of panel D).

5596

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

Table 3 e Estimated maximum specific reduction rate (MSRR; k) values for batch assays conducted with an open to the atmosphere reactor maintained at different temperatures. Temperature (
C)

kNO3

kNO2

RMSDa

mg N/mg VSSCOD-day 5 10 15 22

0.41 0.87 1.61 1.64

   

0.01b 0.01 0.02 0.03

0.20  0.31  0.50  0.53 

0.02 0.03 0.04 0.50

23.7 19.7 26.1 15.5

a Root mean square deviation qX ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ ðmeasured value  estimated valueÞ2 : b Estimate  standard deviation.

experimentally measured nitrite concentrations, with the deviation increasing with decreasing temperature, indicated by the increasing RMSD values (Table 3). The best model fit for the four temperature batch assays is shown in Fig. 5. Based on the MSRR values at the four temperature values tested and the modified Arrhenius model (equation (4)), the dimensionless temperature coefficient (q) was estimated. By taking 15
C as the basis, the temperature coefficient was 1.13 for nitrate reduction, a value which is within the temperature coefficient range of 1.04 and 1.16 reported by Kadlec and Reddy (2001) for removal of nitrate in wetlands. Cherchi et al. (2009) reported a temperature coefficient of 1.11 for nitrate reduction in a chemostat using MicroC G as the carbon source. Kadlec and Wallace (2009) also reported a mean temperature coefficient of 1.11 for nitrate reduction in wetland systems with temperatures ranging from 6 to 24
C. Temperature coefficient values of 1.07e1.14 were estimated for bacteria in biofilms and artificially encapsulated at temperatures as low as 3
C, indicating that the temperature dependence of microbial activity is not affected in immobilized bacteria (Vackova et al., 2011; Welander and Mattiasson, 2003). Thus, the q value found in the present study agrees well with those previously reported.

3.3.

Continuous-flow system simulation

Based on the estimated denitrification kinetic rates, the continuous-flow wetland system was simulated using the series of differential equations and parameters described in Section 2.5, above. Model conditions were chosen based on the results of both the batch and continuous-flow laboratoryscale experiments. The obtained MSSR values indicated that nitrate reduction rates were not affected by initial nitrate concentration and trial simulations showed that the effluent nitrate concentration at high COD:N ratios was not a function of influent nitrate concentrations, agreeing with Monod kinetics. As a result, for all simulations the value for SNO3 ;o was kept constant at 150 mg N/L. Nitrite was not detected in the site groundwater and the groundwater biomass was assumed to be negligible (i.e., SNO2 ;o and XNOx ;o were assumed to be 0). The influent electron donor concentration, Co, was adjusted to simulate different COD:N ratios.

In addition to HRT, COD:N ratio and temperature, microbial biomass retention is another variable parameter for a continuous-flow system. Full-scale wetland systems are generally rich with various types of biomass, such as trees, cattails, algae and bacteria, most of which is retained within the system, providing nutrients and serving as a carbon source to drive biological processes and cell growth (Kadlec and Wallace, 2009). Retention of the denitrifying biomass ultimately controls the solids retention time (SRT) of the system, which is usually very high and not easily controlled in fullscale wetland systems. To simulate a wetland system with a high degree of microbial biomass retention, the biomass retention factor (b) was introduced into the system biomass mass balance equation as follows: !  dXNOx ðXNOx ;o  bXNOx Þ YNO3 kNO3 SNO3 C ¼ XNOx  dt kSNO3 þ SNO3 s KC þ C !  YNO2 kNO2 SNO2 C XNOx  bXNOx þ KSNO2 þ SNO2 KC þ C

(9)

The factor b takes values from 0 (i.e., 100% biomass retention) to 1 (i.e., 0% biomass retention). The pilot-scale wetland system at the biosolids application site was open to the atmosphere and plant and vegetation growth was significant in all cells, likely contributing significant biodegradable carbon to the system and increasing retention of the denitrifying biomass. Alternatively, the laboratory-scale system was constructed with only soil and groundwater, with MicroC G serving as the sole carbon source in all laboratory assays. Although there was no vegetation, most of the biomass was retained in the laboratoryscale continuous-flow reactors, with minimum amounts released in the effluent streams. Because the fraction of biomass retained could not be quantified accurately, a value of 75% biomass retention (i.e., b ¼ 0.25) was assumed for all continuous-flow simulations. Using this biomass retention value and an HRT of 5 days at 22
C, the effluent steady-state nitrogen concentration (nitrate and nitrite) was predicted to be approximately 5 mg N/L using model simulation. Alternatively, assuming no biomass retention (i.e., b ¼ 1) and operating under the same conditions (HRT of 5 days at 22
C), the effluent steady-state nitrogen concentration was predicted to be approximately 13.5 mg N/L. The steady-state effluent nitrate concentration in the laboratory reactor, Run 1, under the same conditions ranged from 0 to 3 mg N/L, which is slightly lower than the simulation results. Therefore, these results verify that the assumption for biomass retention is reasonable and that the model closely predicts the performance of the laboratory continuous-flow system. Using the continuous-flow design equations (equations (5) through (7) and (9)), the system performance was evaluated as a function of HRT, temperature and electron donor availability (i.e., COD:N ratio). System performance is expressed as nitrogen removal efficiency (%), which includes both nitrate and nitrite in the effluent. Other possible denitrification intermediates, NO and N2O, which are not included in the model and were not detected in any of the continuous-flow runs, are assumed to react rapidly and therefore cannot be detected (Kornaros et al., 1996). Thus, the nitrogen removal efficiency was evaluated using the following relationship:

5597

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

0

REMOVAL (%)

A 100 80 60 40 20 06

5

CO 4 3 2 D:N 1

NITROGEN REMOVAL (%) 20 40 60

80

B

C

100 80 60 40 20 06 5 15 CO 4 3 2 1113 9 7 D:N 1 ) 5 s ay 13 T (D

HR

100

100 80 60 40 20 06 5 15 1113 CO 4 3 2 9 7 D:N 1 5 ) s 13 (Day

HRT

23 1720 ) 1114 5 8 MP. (°C E

T

Fig. 6 e Model simulation of the effect of COD:N ratio and HRT on nitrogen removal in a continuous-flow system at (A) 5 and (B) 15
C; (C) effect of COD:N ratio (1e6) and temperature (5e22
C) on nitrogen removal at an HRT of 5 days (75% denitrifying biomass retention is assumed in all simulations).

Nitrogen Removal Efficiency ð%Þ ¼

  SNO3 ;o  SNO3 þ SNO2  100 SNO3 ;o (10)

Fig. 6A and B shows the effect of COD:N and HRT on nitrogen removal efficiency of the system at 5 and 15
C, respectively. Assuming that 75% of biomass is retained (i.e., b ¼ 0.25), changes in the HRT only slightly affect nitrogen removal efficiency. Due to the high SRT, only at low HRT values, below 3 days, is the system performance impacted. Similarly, Fig. 6C illustrates the system nitrogen removal efficiency (%) as a function of COD:N ratio and temperature for values ranging from 1 to 6 and 5 to 15
C, respectively. System performance is only slightly impacted at a decreased temperature and is more severely impacted at COD:N ratio values below 3. The experimental results of the laboratory system and simulations indicate that denitrification can be successful in free-water surface wetland systems at low temperature and HRT values as long as there is enough biodegradable carbon to achieve complete denitrification.

4.

Based on results obtained with the laboratory-scale continuous-flow system and model simulations, high nitrogen removal efficiencies (>90%) can be consistently achieved and maintained by free-water surface wetland systems while treating nitrate-contaminated groundwater with high nitrate levels with an HRT of 3 days or higher and at temperature values as low as 5
C, as long as there is sufficient biodegradable carbon available to achieve complete denitrification. Influent COD:N ratios should be maintained at 3:1 and higher. The results of this study show that constructed wetland technology is a technically feasible and attractive alternative for the treatment of nitrate-bearing groundwater at the biosolids application site.

Acknowledgments This research was supported by the Columbus Water Works (CWW), Columbus, GA through Jordan, Jones and Goulding, Inc. (JJG), Norcross, GA. Special thanks to Camp, Dresser and McKee, Inc. (CDM) for a graduate fellowship to T. Misiti.

Conclusions

A laboratory-scale system was designed and developed to simulate a pilot-scale, free-water surface constructed wetland system proposed for the treatment of nitrate-contaminated groundwater at a biosolids application site. Although oxygen can increase the electron donor requirement for denitrification, this study showed that fast nitrate reduction rates can be achieved even in systems open to the atmosphere, as long as the electron/carbon source is not limiting. The kinetics of nitrate reduction in open to the atmosphere reactors were not severely affected by oxygen competition at initial nitrate concentrations as high as 400 mg N/L. The rate of nitrate reduction was not affected by nitrate concentrations as high as 400 mg N/L, but decreased with decreasing temperature; however, even at temperature values as low as 5
C, complete denitrification occurred in both batch and continuous-flow systems as long as sufficient biodegradable carbon was bioavailable (dissolved).

Appendix. Supplementary data Supplementary data related to this article can be found online at doi:10.1016/j.watres.2011.08.019.

references

American Public Health Association, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. APHAAWWA-WEF, Washington, D.C. Bachand, P.A.M., Horne, A.J., 2000. Denitrification in constructed free-water surface wetlands: I. Very high nitrate removal rates in a macrocosm study. Ecological Engineering 14 (1e2), 9e15. Burgoon, P.S., 2001. Denitrification in free water surface wetlands receiving carbon supplements. Water Science and Technology 44 (11e12), 163e169.

5598

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

Burkart, M.R., Stoner, J.D., 2002. Nitrate in aquifers beneath agricultural systems. Water Science and Technology 45 (9), 19e28. Cherchi, C., Onnis-Hayden, A., El-Shawabkeh, I., Gu, A.Z., 2009. Implication of using different carbon sources for denitrification in wastewater treatments. Water Environment Research 81 (8), 788e799. Darbi, A., Viraraghavan, T., 2004. Effect of low temperature on denitrification. Fresenius Environmental Bulletin 13 (3B), 279e282. Environmental Operating Solutions, 2008. Material Safety Data Sheet e EOS MicroC GTM. http://www.eosenvironmental.com/ product/microc_g.htm (accessed 01.12.10). Bourne, Massachusetts. Gujer, W., 2008. Systems Analysis for Water Technology. Springer, Berlin, Germany. Hajaya, M.G., Tezel, U., Pavlostathis, S.G., 2011. Effect of temperature and benzalkonium chloride on nitrate reduction. Bioresource Technology 102 (8), 5039e5047. Heinen, M., 2006. Simplified denitrification models: overview and properties. Geoderma 133 (3e4), 444e463. Inamori, R., Wang, Y., Yamamoto, T., Zhang, J., Kong, H., Xu, K., Inamori, Y., 2008. Seasonal effect on N2O formation in nitrification in constructed wetlands. Chemosphere 73 (7), 1071e1077. Kadlec, R.H., Reddy, K.R., 2001. Temperature effects in treatment wetlands. Water Environment Research 73 (5), 543e557. Kadlec, R.H., Wallace, S.D., 2009. Treatment Wetlands, second ed. CRC-Press, Boca Raton, FL. Kjellin, J., Hallin, S., Worman, A., 2007. Spatial variations in denitrification activity in wetland sediments explained by hydrology and denitrifying community structure. Water Research 41 (20), 4710e4720. Kornaros, M., Zafiri, C., Lyberatos, G., 1996. Kinetics of denitrification by Pseudomonas denitrificans under growth conditions limited by carbon and/or nitrate or nitrite. Water Environment Research 68 (5), 934e945. Kristiansen, S., 1983. The temperature optimum of the nitrate reductase assay for marine-phytoplankton. Limnology and Oceanography 28 (4), 776e780. Lee, C.G., Fletcher, T.D., Sun, G.Z., 2009. Nitrogen removal in constructed wetland systems. Engineering in Life Sciences 9 (1), 11e22. Macey, R.I., Oster, G.F., 2006. Berkeley Madonna, 8.3. Berkeley Madonna Inc. Maltais-Landry, G., Maranger, R., Brisson, J., 2009a. Effect of artificial aeration and macrophyte species on nitrogen cycling and gas flux in constructed wetlands. Ecological Engineering 35 (2), 221e229. Maltais-Landry, G., Maranger, R., Brisson, J., Chazarenc, F., 2009b. Greenhouse gas production and efficiency of planted and artificially aerated constructed wetlands. Environmental Pollution 157 (3), 748e754.

Maltais-Landry, G., Maranger, R., Brisson, J., Chazarenc, F., 2009c. Nitrogen transformations and retention in planted and artificially aerated constructed wetlands. Water Research 43 (2), 535e545. Murgulet, D., Tick, G.R., 2009. Assessing the extent and sources of nitrate contamination in the aquifer system of southern Baldwin County, Alabama. Environmental Geology 58 (5), 1051e1065. Poe, A.C., Piehler, M.F., Thompson, S.P., Paerl, H.W., 2003. Denitrification in a constructed wetland receiving agricultural runoff. Wetlands 23 (4), 817e826. Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology: Principles and Applications. McGraw-Hill, New York, NY. Rivett, M.O., Buss, S.R., Morgan, P., Smith, J.W.N., Bemment, C.D., 2008. Nitrate attenuation in groundwater: a review of biogeochemical controlling processes. Water Research 42 (16), 4215e4232. Surampalli, R.Y., Lai, K.C.K., Banerji, S.K., Smith, J., Tyagi, R.D., Lohani, B.N., 2008. Long-term land application of biosolids e a case study. Water Science and Technology 57 (3), 345e352. Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater Engineering: Treatment and Reuse, fourth ed. McGraw-Hill, Boston. Tugtas, A.E., Pavlostathis, S.G., 2007. Electron donor effect on nitrate reduction pathway and kinetics in a mixed methanogenic culture. Biotechnology and Bioengineering 98 (4), 756e763. United States Environmental Protection Agency (US EPA), 2000a. Biosolids Technology Fact Sheet e Land Application of Biosolids. Office of Water, Washington, D.C. United States Environmental Protection Agency (US EPA), 2000b. Constructed Wetlands Treatment of Municipal Wastewaters. Office of Research and Development, Cincinnati, Ohio. United States Environmental Protection Agency (US EPA), 2009. Drinking Water Contaminants. http://water.epa.gov/drink/ contaminants/index.cfm (accessed 31.05.11). Vackova, L., Srb, M., Stloukal, R., Wanner, J., 2011. Comparison of denitrification at low temperature using encapsulated Paracoccus denitrificans, Pseudomonas fluorescens and mixed culture. Bioresource Technology 102 (7), 4661e4666. Welander, U., Mattiasson, B., 2003. Denitrification at low temperatures using a suspended carrier biofilm process. Water Research 37 (10), 2394e2398. Wu, J., Zhang, J., Jia, W.L., Xie, H.J., Gu, R.R., Li, C., Gao, B.Y., 2009. Impact of COD/N ratio on nitrous oxide emission from microcosm wetlands and their performance in removing nitrogen from wastewater. Bioresource Technology 100 (12), 2910e2917. Zumft, W., 1997. Cell biology and molecular basis of denitrification. Microbiology and Molecular Biology Reviews 61 (4), 533e616.