Quantification of denitrification potential in carbonaceous trickling filters

ARTICLE IN PRESS

Water Research 37 (2003) 4011–4017

Quantification of denitrification potential in carbonaceous trickling filters Sidney Biesterfelda,*, Greg Farmerb, Linda Figueroac, Denny Parkerd, Phil Russellb a

b

Integra Engineering, 450 Decatur Street, Denver, CO 80204, USA Littleton/Englewood Wastewater Treatment Plant, 2900 South Platte River Drive, Englewood, CO 80110, USA c Environmental Science and Engineering Division, Colorado School of Mines, Golden, CO 80401, USA d Brown and Caldwell, Suite 115, 201 North Civic Drive, Walnut Creek, CA 94596, USA Received 22 October 2002; received in revised form 23 April 2003; accepted 5 May 2003

Abstract Biofilm samples from a carbonaceous trickling filter (TF) were evaluated in bench scale reactors to determine their maximum potential denitrification rates. Intact, undisturbed biofilms were placed into 0.6 L bench-scale reactors filled with sterilized, primary clarifier effluent spiked with nitrate to a final concentration of 16–18 mg/L as N. Dissolved oxygen concentrations were maintained between 2 and 4 mg/L in the bulk aqueous phase. Nitrate loss from the reactors was monitored over a 5 h period. Denitrification rates of 3.09–5.55 g-N/m2 day were observed with no initial lag period. This suggests that the capacity for denitrification is inherent in the biofilm and that denitrification can take place even when oxygen is present in the bulk aqueous phase. There were no significant differences in denitrification rates per unit area of media (g–N/m2 day) either between (a) experimental runs or (b) sampling locations over the trickling filter. This suggests that denitrification potentials are uniform over the entire volume of the full-scale TF. For wastewater treatment plants with TFs that currently nitrify downstream, this approach may be used to meet less stringent permitted discharge concentrations and may allow some facilities to postpone or eliminate construction of additional unit processes for denitrification. r 2003 Elsevier Science Ltd. All rights reserved. Keywords: Denitrification; Trickling filter; Biofilm; Roughing filter; Nitrate; Nitrogen

1. Introduction Many wastewater treatment facilities are facing stricter effluent nitrate-nitrogen (NO3-N) concentration limits as nitrate can contribute to receiving water eutrophication. In arid regions where rivers are effluent dominated, nitrate limitations may also be influenced by safe drinking water standards as one city’s wastewater plant effluent is another’s drinking water plant influent. *Corresponding author. 3085 South Washington Street, Colorado 80110, USA. Tel.: +1-303-825-1802; fax: +1-303825-2322. E-mail address: [email protected] (S. Biesterfeld).

Existing facilities may be upgraded to denitrify by expanding their activated sludge (AS) basins to obtain the longer sludge ages required to nitrify and denitrify or by adding tertiary denitrification filters. Tertiary denitrification filters require a supplemental carbon source and may result in permit violations of biochemical oxygen demand (BOD5) if accidentally overdosed. Methanol, a commonly used supplemental carbon source, is extremely flammable and represents an added expense. Trickling filter (TF) facilities facing low NO3-N reductions to meet new discharge limits might consider a third possibility for denitrification; recycling of nitrified effluent back to existing upstream carbonaceous

0043-1354/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved. doi:10.1016/S0043-1354(03)00302-6

ARTICLE IN PRESS 4012

S. Biesterfeld et al. / Water Research 37 (2003) 4011–4017

TFs. Since dissolved oxygen (DO) typically only penetrates the upper 100–200 mm of a biofilm, thicker biofilms, such as those found in carbonaceous TFs, contain anoxic and anaerobic layers where NO3-N and sulfate may be utilized by bacteria to oxidize organic matter. Parker and Richards [1] previously observed significant denitrification within a low loaded TF during a pilot study at the City of Garland (TX) when internal recycling was employed. The same phenomenon was observed by Mehlhart [2] at several European WWTPs with tertiary nitrifying TFs. The advantages to this approach of recycling flow are that it utilizes existing unit processes thereby allowing some facilities to delay or avoid facility expansion and that it requires less external carbon addition than tertiary denitrification. This process may remove up to 50% of a facility’s NO3-N when a 100% recycle rate is used [3,4] and up to 67% of a facility’s nitrate when a 200% recycle rate is used. This makes it an attractive alternative where discharge limits are relatively high and only partial denitrification is required. This process may be applied to any facility with carbonaceous TFs regardless of their downstream nitrification method. Hydraulic constraints of a given facility may limit the amount of denitrification achievable because of restrictions on the amount of recycle permitted. Hydraulic constraints and diurnal variations in both flow and concentrations make it difficult to estimate the total denitrification potential of TFs from full-scale operating systems. In other words, at the times of day when the highest flows and concentrations of nitrate might be recycled from downstream processes back to the TF influent, the influent wastewater chemical oxygen demand (COD) and soluble chemical oxygen demand (SCOD) concentrations may be comparatively low, thereby preventing complete denitrification from occurring. A laboratory bench-scale procedure was developed to determine the possible amount of nitrate that could be removed, i.e. the denitrification potential, per unit area of TF media. Experimental conditions were selected to mimic the conditions within the full-scale system. Determining the potential denitrification rate under expected field conditions will provide a more accurate estimate of the total mass of nitrogen that may be removed by full-scale TFs. The results presented here represent a preliminary estimate of denitrification potential for plastic media TFs receiving medium strength domestic wastewater.

2. Methodology Full scale system: The Littleton/Englewood Wastewater Treatment Plant, Englewood, Colorado, has two carbonaceous TFs that are operated in parallel. These covered filters were designed with 4.9 m of media depth,

are 32 m in diameter, have an average hydraulic loading rate of 112+ 31 m3/m2 d, and are packed with cross-flow media that has a specific surface area of 121 m2/m3. They are equipped with rotational, four arm distributors that complete one rotation every four minutes. To maintain a relatively constant hydraulic loading rate, flow is recirculated from the composited TF effluents during periods of low plant flow and is bypassed directly to the solids contact basins during periods of very high plant flow. Influent and effluent COD concentrations to the TF were 105721 and 69720 mg/L, respectively, throughout the biofilm sample collection period. Soluble COD concentrations were 4479 mg/L at the TF influent and 1474 mg/L at the TF effluent. TF influent sampling includes recycle flow. Three capped sampling ports are spaced approximately 1.2 m apart from top to bottom and extend approximately 2.0 m into the filter. Biofilm sampling devices: Biofilm samples were grown on standard glass microscope slides inserted into the sampling ports of a TF at the Littleton/Englewood WWTP (Englewood, Colorado) as previously described [5]. Sampling locations are designated as top, middle, and bottom. After 29 days, slide groups were harvested, submerged in process water, and brought back to the laboratory where they were suspended in miniature batch reactors. Approximately 15 min elapsed between the removal of the slide groups from the TF and benchscale reactor start up. Bench scale reactors: Each slide group removed from the TF was suspended in a miniature batch reactor containing 600 mL of primary clarifier (PC) effluent as the COD source. The PC effluent was sterilized by autoclaving prior to use to ensure that only biofilm activity and not PC effluent activity was being measured. Three reactors per sampling location were run. Three control reactors, containing only PC effluent and no microscope slides, were also run. The PC effluent was spiked with sodium nitrate to give a final concentration of 16–18 mg/L of NO3-N. This concentration was chosen as it represented the best possible NO3-N return concentration from the NTFs at this facility. Standard 5-gal aquarium pumps equipped with air stones were used to aerate the reactors in conjunction with magnetic stirring. Needle valves were used to maintain the DO concentrations in the reactors between 2 and 4 mg/L. The full-scale TF bulk aqueous phase DO concentrations at the L/E WWTP fell within this range at the time of sampling. The DO in the bulk aqueous phase at the time of sampling at the top of the tower and at the TF influent pump was about 2 mg/L and the DO at the bottom of the tower and at the TF underdrain was about 4 mg/L. Reactors were run for 5 h. DO, pH, and NO3-N were measured at reactor start-up and once every thirty minutes thereafter. Additionally, initial and final COD, SCOD, ammonia, nitrite, sulfate, alkalinity, and total suspended solids (TSS) concentrations were

ARTICLE IN PRESS S. Biesterfeld et al. / Water Research 37 (2003) 4011–4017

measured. NO3-N reduction rates in g N/m2 day were calculated. Analytical methods: At the completion of reactor runs, slide groups were removed from the reactors and were transferred to a drying oven at 103 C. Aqueous phase samples were collected from each reactor. All samples were analyzed in duplicate by accepted United States Environmental Protection Agency (USEPA) methods and/or in accordance with Standard Methods for the Examination of Water and Wastewater (1995). DO and pH measurements were taken with Thermo-Orion (Beverly, Maryland) meters, model numbers 850A+ and 525A+, respectively. Both meters were calibrated immediately prior to use (two point calibrations). COD and SCOD concentrations were determined by the dichromate method, Hach (Hach Company, Loveland, Colorado) method number 8000, using the 0–150 mg/L range COD vials. Samples for soluble COD were filtered through 2–5 mm glass fiber filters prior to analysis. The ammonia-nitrogen concentration of each reactor was determined by the Nesslerization method [6] without distillation. Nitrate- and nitrite-nitrogen concentrations were determined by Hach (Loveland, Colorado) method numbers 10020, chromotropic acid method for nitrate, and 8507, diazotization method for nitrite. Alkalinity was determined by titration with 0.02 normal sulfuric acid to the bromo-cresol green, methyl-red indicator endpoint in accordance with Standard Method number 2320B/23210B4b [6]. Sulfate was determined by ion chromatography in accordance with USEPA Method

4013

number 375.1. Samples that could not be analyzed immediately were preserved by storing at 4oC and were analyzed within 24 h of collection. Dry weight biomass: Accumulated biomass was quantified as dry weight biomass in milligrams per microscope slide as previously described [7].

3. Results Three separate sampling campaigns were conducted to establish the biofilm denitrification potential. Changes in NO3-N concentration with time for a typical set of samples is plotted in Fig. 1. No lag periods were observed for denitrification to begin and the rate of denitrification was zero order with respect to nitrate concentration. Denitrification rates in g-N/m2 day are presented in Table 1 for each reactor for each determination. The last group of data points from the second experimental run were not included in rate calculations as the reactors may have become SCOD limited. Analysis by ANOVA at a 95% confidence interval did not detect any significant differences (po0.05) between removal rates per unit area of media (g-NO3-N/m2 day) either between (a) experimental runs or (b) slide incubation location. Denitrification rates were then compared against dry weight biomass accumulations for each reactor (Fig. 2). There was no correlation between denitrification rates and accumulated biomass.

Fig. 1. First determination of denitrification rates.

ARTICLE IN PRESS S. Biesterfeld et al. / Water Research 37 (2003) 4011–4017

4014 Table 1 Nitrate removal rates Top

Top

Top

Middle

Middle

Middle

Bottom

Bottom

Bottom

5.10 4.49 4.61

4.67 4.33 4.77

5.23 4.49 5.29

4.88 4.64 4.82

4.10 4.38 5.55

3.88 4.33 4.87

4.75 3.09 5.44

2

Rates are given in g-NO3-N/m day 07/17/01 4.71 5.10 08/22/01 4.64 4.02 09/22/01 5.13 4.30

Fig. 2. Dry weight biomass per reactor versus denitrification rate.

Table 2 Bench reactor characteristics First determination

COD SCOD NH3-N NO2-N NO3-N SO4 Alkalinity TSS pH

Second determination

Third determination

Initial

Final

Initial

Final

Initial

Final

227 171 17.8 o0.5 16.3 151 206 106 9.1270.02

162739 11075 15.970.3 o0.5 5.471.0 15973 23476 67732 8.4070.05

181 94 13.8 o0.5 16.4 221 205 73 9.1170.02

13079 6575 15.870.3 o0.5 8.071.0 22074 23776 5178 8.5170.08

231 96 15.1 o0.5 17.8 161 200 93 9.0270.11

151714 7476 16.270.4 o0.5 8.470.8 15972 22374 44713 8.4070.17

All results are reported as mg/L plus or minus one standard deviation. Initial values are from single measurements of the PC effluent stock solution used to fill all of the reactors for that sampling campaign, therefore, standard deviations are not reported for these measurements with the exception of pH which was measured for each reactor after set-up. Alkalinity is reported as total alkalinity as calcium carbonate.

ARTICLE IN PRESS S. Biesterfeld et al. / Water Research 37 (2003) 4011–4017

Initial and final reactor characteristics are presented in Table 2. On average, 3.871.8 mg of soluble COD and 10.074.4 mg of total COD were consumed for every mg of NO3-N reduced. Additionally, 3.2+1.0 mg of alkalinity, expressed as total calcium carbonate alkalinity, were generated for every mg of NO3-N reduced. Sulfate concentrations did not change from start to end of the bench-scale reactor runs. For each of these reported values, a total of 27 measurements were made—one for each reactor in each sampling campaign.

4. Discussion The biofilms harvested from the TF were able to denitrify at high rates in the presence of 2–4 mg/L of DO in the bulk aqueous phase without an acclimatization period. Thus, the capacity for denitrification is inherent in the biofilm and can proceed at a rapid rate under typical field conditions. This suggests that the enzymes required for nitrate reduction are typically present, but are inhibited at higher concentrations of oxygen. This is consistent with observations by others for AS systems that practice on/off aeration or that cycle through anoxic and oxic zones [8,9]. In these systems, denitrification has been observed to begin when the DO concentration within the AS flocs drops below 0.6 mg/ L [10]. Revsbech et al. [11] used a microelectrode to determine DO concentrations in biofilms scraped from a rock TF and found that even at bulk aqueous phase DO concentrations of 6.4 mg/L, DO could not be detected by 1500 mm of biofilm depth. Since carbonaceous TF biofilms may be up to 10,000 mm thick [4], a large fraction of the biofilm may be available for denitrification. Denitrification rates averaged 4.65 g-N/m2 day with a standard deviation of 11% over 27 rate measurements. This suggests that the denitrification potential is uniform over the entire volume of the full-scale TF in the presence of excess COD and SCOD and NO3-N above 2 mg/L as N. Denitrification potentials measured in this study were higher than full-scale denitrification rates observed under similar conditions reported in other investigations. Increases in nitrate concentration have been observed to increase biofilm denitrification rates [2,11–13]. Researchers have also reported a direct correlation between the amount of dry weight biomass present in a system and the denitrification rate [14,15]. In these situations, nitrate was likely the rate limiting substrate in the biofilm and thus controlled the active biofilm thickness. As more nitrate was added, the biofilms became acclimated and their active depths increased. Denitrification rates up to 9.12 g-N/m2 day were observed by Rudiger and Sekoulov [13] when nitrate concentrations in the bulk aqueous phase were gradually increased to

4015

90 mg/L as N. For the biofilms evaluated here, there was no direct correlation between the dry weight biomass and the denitrification rates. Since the test period was only 5 h and the biofilms were harvested from a nitratefree environment, the biofilms tested were not acclimated. The active thickness of the harvested biofilms was defined by the growth conditions in the full-scale TF. Thus, the lack of correlation between biomass and denitrification rate in this study, may be an artifact of the testing protocol. This suggests that full-scale plastic media TFs may be capable of higher rates of denitrification than those reported here once they have been acclimated. Three studies in the literature report denitrification rates (per unit of media surface area) that were observed under similar, but not identical conditions, to those evaluated here. In the first study, Rudiger and Sekoulov [13] recycled nitrified effluent to a full-scale TF filled with slag media. DO concentrations over the TF were between 4 and 7 mg/L and the influent nitrate-nitrogen concentration was 20 mg/L. PC effluent was the sole carbon source. They reported maximum denitrification rates of 3.61 g-N/m2 day. This is lower than the average denitrification rate of 4.65 g-N/m2 day found here and may reflect nitrate limiting conditions and/or inaccuracies in surface area calculations. Nitrate has been reported to become rate limiting for denitrification below 1.2 mg/L [16] and 2.0 mg/L [3]. For this system, the effluent nitrate concentrations were below 1.0 mg/L. Consequently, denitrification rates in the bottom portion of this TF may have been negatively impacted. Because overall denitrification rates were calculated for this system, it is possible that denitrification was occurring at a higher rate in the upper portion of this TF. Some of the difference between rates may also be accounted for by the assumed surface area for the slag material may have been greater than the actual area. In the second study, Bosander and Westlund [17] reported denitrification rates of 2.1 g-N/m2 day for a full-scale fluidized bed reactor. For this system, influent and effluent nitrate concentrations were 18 and 1.9 mg/L as N, respectively. DO concentrations were not reported. In the third study, Aspegren et al. [18] reported denitrification rates between 1.2 and 2.5 g-N/m2 day for a moving bed biofilm reactor (MBBR). For this system, the influent nitrate concentration was 13 mg/L as N and the effluent nitrate concentration was o1 mg/L as N. DO was 7 mg/L. Both Bossander and Westlund [17] and Aspegren et al. [18] used methanol in addition to wastewater as their carbon source. The rates reported for both of these systems are lower than the average denitrification rate of 4.65 g-N/m2 day found here and may reflect differences in biomass accumulation and thus active thickness as discussed above. Although both fluidized bed reactors and MBBRs are biofilm processes, the media in each is typically subjected to greater shear

ARTICLE IN PRESS 4016

S. Biesterfeld et al. / Water Research 37 (2003) 4011–4017

forces than the TF media. As a result, the biofilms developed in these processes may not have been as thick as those developed in TFs and would have correspondingly lower denitrification rates [19] for the same nitrate concentration. The ratio of 3.81 g of SCOD consumed per gram of NO3-N reduced is consistent with previously published values of 3.61 [17] and 2.86 [3], however, the ratio of 9.95 g of total COD consumed per gram of NO3-N reduced is greater and may be attributed to oxidation of COD with oxygen in the outer layers of the biofilm. Higher than stoichiometrically required COD requirements were also observed by Aspegren et al. [18] for a denitrifying MBBR with an influent DO concentration of 7 mg/L. The ratio of 3.16 g of alkalinity produced per gram of NO3-N reduced is consistent with the stoichiometric ratio of 3.57 [3]. Full-scale TFs typically produce H2S through the reduction of sulfate present in the influent wastewater. This odorous off-gas may be neutralized by scrubbing with sodium hypochlorite. Its formation may also be mitigated through the addition of ferric chloride. Ferric chloride reacts with sulfides to form the solid iron sulfide, which precipitates and ultimately ends up in the wasted sludge. Heukelekian [20] noted that in oxygen deficient biological systems, NO3-N is reduced preferentially over sulfate and that its presence may mitigate H2S formation. Sulfate concentrations in the batch reactors did not change for any of the three sampling campaigns. Therefore, hydrogen sulfide was not formed by the reduction of sulfate. This implies a potential odor control benefit and a potential reduction in chemical usage for odor control in full-scale systems when nitrate is recycled. Heukelekian [20] proposed the addition of nitrate to meet fifty percent of the oxygen demand from BOD5 to obtain complete protection against odors from wastewater. For the Littleton/Englewood WWTP, the potential rate of NO3-N reduction per unit area of media was greater than the downstream generation rate when the reactor nitrate concentration was 18 mg/L as N. Higher nitrate concentrations are expected to result in higher denitrification rates. It is therefore theoretically possible to achieve nearly complete denitrification by this method. Hydraulic constraints of a particular facility, e.g. surface overflow rates on clarifiers, will clearly limit the amount of denitrification achievable because of restrictions on the amount of recycle permitted. However, recycle may be used in combination with traditional denitrification processes to minimize costs. The removal rates observed here may be extrapolated to plastic media filters at other facilities where the media surface area is known and allow for preliminary estimates of the quantity of nitrate that might be removed at a particular facility. These estimates may

then be used for plant capacity assessments and for decisions regarding facility expansions to add additional unit processes for denitrification. Care should be taken to ensure that the conditions of the facility being evaluated approximate those used in these batch reactor tests with respect to influent nitrate concentrations and that neither COD nor SCOD will be rate limiting.

5. Conclusions 1. Denitrification is due to biofilm activity since the control reactors did not exhibit NO3-N reduction. 2. There is no lag time for the development of NO3-N reduction capability in TF biofilms. 3. NO3-N reduction occurs in the presence of oxygen at levels up to 5 mg/L in the bulk aqueous phase suggesting that even at high bulk aqueous phase DO levels, an anoxic biofilm layer exists. 4. The rate of NO3-N reduction is independent of the slide incubation location within the TF. Therefore, the entire TF is capable of NO3-N reduction when COD and SCOD are not rate limiting. 5. The removal rates reported here are for smooth surfaces and are representative of plastic media TFs receiving medium strength domestic wastewater. Thus, preliminary estimates of a particular facility’s nitrate removal capacity may be calculated using this testing protocol. 6. Recycling NO3-N should increase active biofilm thickness and thus TF capacity as more COD and SCOD may be removed per unit area. 7. For this facility, the rate of NO3-N reduction per unit area of media was greater than the downstream generation rate. It is therefore theoretically possible to achieve nearly complete denitrification by this method. However, hydraulic constraints will clearly limit the amount of denitrification achievable because of restrictions on the amount of recycle permitted through the TF and through downstream processes. 8. Recycling NO3-N to upstream tricking filters may also provide some odor control benefit by preventing the formation of hydrogen sulfide.

Acknowledgements Special thanks go to the staff at the Littleton/ Englewood Wastewater Treatment Plant for supporting and funding this work. Special thanks also go to Brown and Caldwell whose innovative approach to achieving denitrification at the Littleton/Englewood Wastewater Treatment Plant provided the inspiration for this work.

ARTICLE IN PRESS S. Biesterfeld et al. / Water Research 37 (2003) 4011–4017

References [1] Parker DS, Richards T. Nitrification in trickling filters. J WPCF 1986;58(9):896–902. [2] Mehlhart GF. Upgrading of existing trickling filter plants for denitrification. Water Sci Technol 1994;30(6):173–9. [3] USEPA Office of Research and Development, Office of Water. Nitrogen Control Manual EPA/625/R-93/010. Washington DC, USA: United States Environmental Protection Agency; 1993. [4] Metcalf and Eddy Inc. Wastewater engineering: treatment and reuse, 4th ed. New York, NY, USA: McGraw Hill, 2002 [Revised by Tchobanoglous G, Burton FL, Stensel HD]. [5] Biesterfeld S, Figueroa L, Hernandez M, Russell P. Quantification of nitrifying bacterial populations in a full-scale NTF using fluorescent in situ hybridization. Water Environ Res 2001;73(3):329–39. [6] APHA/AWWA/WEF. Standard methods for the examination of water and wastewater, 19th ed. Washington DC, USA: American Public Health Association/American Water Works Association/Water Environment Federation; 1995. [7] Biesterfeld S, Farmer G, Russell P, Figueroa L. Effect of alkalinity type and concentration on nitrifying biofilm activity. Water Environ Res 2003;75(3). [8] Schuyler, R. Personal communication. Senior Project Manager, Bug Dr. Rothberg, Tamburini, and Winsor Professional Engineers and Consultants, Inc. Denver, CO, USA, 2002. [9] Rittmann BE, Langeland WE. Simultaneous denitrification with nitrification in single-channel oxidation ditches. J WPFC 1985;57(4):300–8. [10] Schramm A, Santegoeds CM, Nielsen HK, Ploug H, Wagner M, Pribyl M, Wanner J, Amann R, de Beer D. On the occurrence of anoxic microniches, denitrification, and

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

4017

sulfate reduction in aerated activated sludge. Appl Environ Microbiol 1999;65(9):4189–96. Revsbech NP, Christensen PB, Nielsen LP, Sorensen J. Denitrification in a trickling filter biofilm studied by a microsensor for oxygen and nitrous oxide. Water Res 1989;23(7):867–71. Patureau D, Bernet N, Delgenes JP, Moletta R. Effect of dissolved oxygen and carbon–nitrogen loads on denitrification by an aerobic consortium. Appl Microbiol Biotech 2000;54:535–42. Rudiger A, Sekoulov I. Denitrification in trickling filters. Proceedings of the European Conference on Nutrient Removal from Wastewater, University of Leeds, Wakefield, UK, 1992. Oh J, Yoon SM, Park JM. Denitrification in submerged biofilters of concentrated-nitrate wastewater. Water Sci Technol 2001;43(1):217–23. Strand SE, McDonnell AJ, Unz RF. Concurrent denitrification and oxygen uptake in microbial biofilms. Water Res 1985;19(3):335–44. Beccari M, Passino R, Ramadori R, Tandoi V. Kinetics of dissimilatory nitrate and nitrite reduction in suspended growth culture. J WPCF 1983;55(1):58–64. Bosander J, Westlund AD. Operation of full-scale fluidized bed for denitrification. Water Sci Technol 2000;41(9): 115–21. Aspegren H, Nyberg U, Andersson B, Gotthardsson S, la Cour Janson J. Post denitrification in a moving bed biofilm reactor process. Water Sci Technol 1998;38(1): 31–8. Van Loosdrecht MCM, Eikelbom D, Gjaltema A, Mulder A, Tijhuis L, Heijnen JJ. Biofilm structures. Water Sci Technol 1995;32(8):35–43. Heukelekian H. Effect of the addition of sodium nitrate to sewage on hydrogen sulfide production and BOD reduction. Sewage Works J 1943;15(2):255–61.