Understanding the acidification risk of cheese whey anaerobic digestion under psychrophilic and mesophilic conditions.

Anaerobic digestion is a suitable technology to treat cheese whey (CW), a high-strength wastewater from cheesemaking. However, CW anaerobic digestion is limited by its high biodegradability, acidic pH, and lack of alkalinity. This publication evaluated the acidification risk of CW anaerobic digestion under psychrophilic and mesophilic conditions, aiming to improve digester design, operation, and decision-making when facing instability periods. To evaluate the acidification risk of CW anaerobic digestion, biochemical methane potential (BMP) tests were carried out at four different organic loads, each under psychrophilic (20 °C) and mesophilic (35 °C) conditions. Besides methane production, pH, soluble chemical oxygen demand, volatile fatty acid and alcohols were also monitored. Experimental results showed that CW can be successfully degraded under both temperature conditions, with methane yields of 389-436 mLCH4/gVS. The organic load had a greater impact on the accumulation of intermediate products than temperature, indicating that process inhibition by overloading is plausible under psychrophilic and mesophilic conditions. However, the degradation rate under mesophilic conditions was faster than under psychrophilic conditions. Experimental results also revealed a higher imbalance between fermentation and methanogenesis rate under psychrophilic conditions, which resulted in higher concentrations of intermediate products (volatile fatty acids and alcohols) and prolonged lower pHs. These results indicate that the degradation of intermediate products is less favourable under psychrophilic conditions compared to mesophilic conditions. This implies that psychrophilic digesters have a lower capacity to recover from process disturbances, increasing the risk of process underperformance or even failure under psychrophilic conditions.

Cheese whey and dairy manure anaerobic co-digestion at psychrophilic conditions: Technical and environmental evaluation.

Cheese whey (CW) and dairy manure (DM) are the main residues from the dairy industry, both of which can led to significant negative environment impacts if not properly managed. However, their combined anaerobic digestion represents an opportunity to obtain bioenergy and a stabilised material as a soil improver on the farm. Biochemical potential of methane (BMP) assays were carried out at psychrophilic conditions (20 °C) to analyse the influence on biomethane production of different CW:DM mixtures (% w/w) at different of inoculum-to-substrate ratios (ISR). Based on the BMP results, a life cycle assessment (LCA) of the cheese manufacturing process was carried out considering two scenarios (i) considering the current process, where propane gas and electricity are used for cheese production (ii) the incorporation of the biogas generated in the cheese production process in the company. BMP results showed that the best mixture between CW and DM was 65:35 (weight basis) at an organic load of 0.6 gVS/L (ISR of X). The LCA showed that CW and DM anaerobic digestion allowed to reduce the cheese manufacturing carbon footprint from through the substitution of propane by the biogas produced, changing from 5.5 to 3.1 kg CO2-eq/kg cheese produced, which indicates that according to the monthly production (633.6 kg) it would stop emitting about 1519 kg CO2-eq, i.e. a saving in terms of emissions of approximately 43,6% of the total currently generated.

Ammoniacal nitrogen recovery from swine slurry using a gas-permeable membrane: pH control strategies and feed-to-trapping volume ratio.

Gas-permeable membrane (GPM) technology is gaining interest to recover nitrogen from residual effluents due to its effectiveness, simple operation and capacity of producing a nutrient rich product with fertilising value. In this study, a GPM contactor was used at 25 °C to recover total ammoniacal nitrogen (TAN) from swine slurry as a concentrated (NH4)2SO4 solution. Firstly, a synthetic solution was tested on a wide pH range (6-12). Results showed that the ammonia mass transfer constants (Km) increased from 7.9·10-9 to 1.2·10-6 m/s as the pH increased. The reagent consumption to control the pH per mole nitrogen recovered had a minimum at pH 9, which showed a Km value of 3.0·10-7 m/s. Secondly, various pH control strategies were tested using swine slurry, including (i) no pH control, (ii) pH control at 8.5, 9.0 and 10.0, and (iii) an initial spike of the NaOH equivalent to the required to control the pH at 9. The test without pH control reached a TAN recovery of around 60%, which could be an interesting strategy when high nitrogen recoveries or short operating times are not required. The pH control at 9 stood out as the most favourable operating condition due to its high Km and lower reagent consumption. Thirdly, several feed-to-trapping volume ratios ranging from 1:1 to 15:1 were tested using swine slurry at pH 9. These assays revealed that a GPM process with a high feed-to-trapping volume ratio fastens the recovery of 99% of TAN as a high purity (NH4)2SO4 solution containing 40 g N/L.

Impact of permeate flux and gas sparging rate on membrane performance and process economics of granular anaerobic membrane bioreactors.

This research investigated the impact of permeate flux and gas sparging rate on membrane permeability, dissolved and colloidal organic matter (DCOM) rejection and process economics of granular anaerobic membrane bioreactors (AnMBRs). The goal of the study was to understand how membrane fouling control strategies influence granular AnMBR economics. To this end, short- and long-term filtration tests were performed under different permeate flux and specific gas demand (SGD) conditions. The results showed that flux and SGD conditions had a direct impact on membrane fouling. At normalised fluxes (J20) of 4.4 and 8.7 L m-2 h-1 (LMH) the most favourable SGD condition was 0.5 m3 m-2 h-1, whereas at J20 of 13.0 and 16.7 LMH the most favourable SGD condition was 1.0 m3 m-2 h-1. The flux and the SGD did not have a direct impact on DCOM rejection, with values ranging between 31 and 44%. The three-dimensional excitation-emission matrix fluorescence (3DEEM) spectra showed that protein-like fluorophores were predominant in mixed liquor and permeate samples (67-79%) and were retained by the membrane (39-50%). This suggests that protein-like fluorophores could be an important foulant for these systems. The economic analysis showed that operating the membranes at moderate fluxes (J20 = 7.8 LMH) and SGD (0.5 m3 m-2 h-1) could be the most favourable alternative. Finally, a sensitivity analysis illustrated that electricity and membrane cost were the most sensitive economic parameters, which highlights the importance of reducing SGD requirements and improving membrane permeability to reduce costs of granular AnMBRs.

The interaction between lipids and ammoniacal nitrogen mitigates inhibition in mesophilic anaerobic digestion.

Ammoniacal nitrogen and long chain fatty acids (LCFA) are common inhibitors of the anaerobic digestion process. However, the interaction between these inhibitors has received little attention. Understanding the interaction between these inhibitors is important to optimise the operation of anaerobic digesters treating slaughterhouse waste or using fat, oil and grease (FOG) as co-substrate among others. To study the interaction between ammoniacal nitrogen and LCFA inhibition, 20 different conditions were trialled in mesophilic batch tests. Experimental conditions included 5 mixtures between slaughterhouse wastewater and LCFA (100:0, 75:25, 50:50, 20:80, 0:100 on a VS basis), each one tested at 4 different ammoniacal nitrogen concentrations (0, 1, 3, 6 gNadded·L-1). Experimental and modelling results showed that ammoniacal nitrogen inhibition was less severe in LCFA-rich mixtures, indicating that LCFA mitigated ammoniacal nitrogen inhibition to a certain extent. However, the positive interaction between inhibitors did not only depend on the LCFA concentration. A protective LCFA coat that limited the diffusion of free ammonia into the cell and/or provided a localised lower pH in the vicinity of the microbial cell could explain the experimental results. However, ammoniacal nitrogen and LCFA inhibition comprise up to 6 different but interrelated inhibitors (i.e. NH3, NH4+, LCFA, VFA, H2 and pH) and therefore the specific mechanism could not be elucidated. Nonetheless, these results suggest that LCFA do not exacerbate TAN-related inhibition and that LCFA-rich substrates can be utilised as co-substrates in mesophilic N-rich digesters.

Considering syntrophic acetate oxidation and ionic strength improves the performance of models for food waste anaerobic digestion.

Current mechanistic anaerobic digestion (AD) models cannot accurately represent the underlying processes occurring during food waste (FW) AD. This work presents an update of the Anaerobic Digestion Model no. 1 (ADM1) to provide accurate estimations of free ammonia concentrations and related inhibition thresholds, and model syntrophic acetate oxidation as acetate-consuming pathway. A modified Davies equation predicted NH3 concentrations and pH more accurately, and better estimated associated inhibitory limits. Sensitivity analysis results showed the importance of accurate disintegration kinetics and volumetric mass transfer coefficients, as well as volatile fatty acids (VFAs) and hydrogen uptake rates. In contrast to the default ADM1, the modified ADM1 could represent methane production and VFA profiles simultaneously (particularly relevant for propionate uptake). The modified ADM1 was also able to predict the predominant acetate-consuming and methane-producing microbial clades. Modelling results using data from reactors dosed with granular activated carbon showed that this additive improves hydrogen uptake.

The origin of waste activated sludge affects the enhancement of anaerobic digestion by free nitrous acid pre-treatment.

Anaerobic digestion is a common stabilization method for treating primary sludge (PS) and waste activated sludge (WAS). However, its application is often limited by the degradation of WAS. Recent studies have demonstrated FNA to be an effective pre-treatment for enhancing WAS degradability, while having limited effect on PS degradability. WAS characteristics are impacted by wastewater treatment plant (WWTP) configuration and this study is the first to compare the effectiveness of FNA pre-treatment on WAS from WWTP with and without primary treatment. In this study, WAS samples were collected from four full-scale WWTPs with or without primary treatment. Sludge characterization, biomethane potential tests and mathematical modeling were conducted to assess the impacts of FNA pre-treatment on anaerobic digestion. The results showed that FNA pre-treatment was consistently effective for WAS from different WWTPs, while the extent of enhancement varied between WWTPs. For WAS from WWTPs without primary treatment, FNA pretreatment increased the rate of hydrolysis by 54-66% compared to 22-33% increase for WAS without primary treatment. In contrast, WAS from WWTPs with primary treatment experienced greater increases in methane potential (22-24%) compared to WAS from WWTPs without primary treatment (14-16%). These variances could be associated with primary treatment impacting the wastewater COD/N ratio and thus portion of extracellular polymetric substances (EPS) and cells in WAS. FNA pre-treatment targets the destruction of polymetric substances and cells, therefore WAS with a higher proportion of cells (i.e., WAS with primary treatment) experienced greater improvements in methane yield. Similarly, greater improvements in hydrolysis rate were observed for WAS from WWTP without primary sedimentation which contain higher proportions of large EPS molecules. Despite its consistent effectiveness on WAS samples, FNA pre-treatment was ineffective for improving the digestibility of high-rate activated sludge (HRAS).

Co-digestion of sewage sludge and food waste in a wastewater treatment plant based on mainstream anaerobic membrane bioreactor technology: A techno-economic evaluation.

The implementation of anaerobic membrane bioreactor as mainstream technology would reduce the load of sidestream anaerobic digesters. This research evaluated the techno-economic implications of co-digesting sewage sludge and food waste in such wastewater treatment plants to optimise the usage of the sludge line infrastructure. Three organic loading rates (1.0, 1.5 and 2.0 kg VS m-3 d-1) and different strategies to manage the additional nutrients backload were considered. Results showed that the higher electricity revenue from co-digesting food waste offsets the additional costs of food waste acceptance infrastructure and biosolids disposal. However, the higher electricity revenue did not offset the additional costs when the nutrients backload was treated in the sidestream (partial-nitritation/anammox and struvite precipitation). Biosolids disposal was identified as the most important gross cost contributor in all the scenarios. Finally, a sensitivity analysis showed that food waste gate fee had a noticeable influence on co-digestion economic feasibility.

Towards a standardization of biomethane potential tests: a commentary.

Inter-laboratory reproducibility of biomethane potential (BMP) is dismal, with differences in BMP values for the same sample exceeding a factor of two in some cases. A large group of BMP researchers directly addressed this problem during a workshop held in Leysin, Switzerland, in June 2015. The workshop resulted in a new set of guidelines for BMP tests published in 2016, which is the subject of the present commentary. The work has continued with two international inter-laboratory studies and one additional workshop held in Freising, Germany, in 2018. The dataset generated by the two inter-laboratory studies were used to refine the validation criteria for BMP tests. Based on these new results an update to the original guidelines is proposed here.

Anaerobic membrane bioreactor performance at different wastewater pre-concentration factors: An experimental and economic study.

This research evaluated the performance of a lab-scale anaerobic membrane bioreactor (AnMBR) treating municipal sewage pre-concentrated by forward osmosis (FO). The organic loading rate (OLR) and sodium concentrations of the synthetic sewage stepwise increased from 0.3 to 2.0 g COD L-1 d-1 and from 0.28 to 2.30 g Na+ L-1 to simulate pre-concentration factors of 1, 2, 5 and 10. No major operational problems were observed during AnMBR operation, with COD removal efficiencies ranging between 90 and 96%. The methane yield progressively increased from 214 ± 79 to 322 ± 60 mL CH4 g-1 COD as the pre-concentration factor increased from 1 to 10. This was mainly attributed to the lower fraction of methane dissolved lost in the permeate at higher OLRs. Interestingly, at the highest pre-concentration factor (2.30 g Na+ L-1) the difference between the permeate and the digester soluble COD indicated that membrane biofilm also played a role in COD removal. Finally, a preliminary energy and economic analysis showed that, at a pre-concentration factor of 10, the AnMBR temperature could be increased 10 °C and achieve a positive net present value (NPV) of 4 M€ for a newly constructed AnMBR treating 10,000 m3 d-1 of pre-concentrated sewage with an AnMBR lifetime of 20 years.

Transition of microbial communities and degradation pathways in anaerobic digestion at decreasing retention time.

Tuning of operational variables is a common practice to control the anaerobic digestion process and, in advanced applications, to promote the accumulation of fermentation products. However, process variables are interrelated. In this study, the hydraulic retention time (HRT) was decoupled from the organic loading rate (OLR) in order to isolate the effect of HRT as a selective pressure on: process performance, metabolic rates (hydrolytic, acetogenic, and methanogenic) and the microbial community. Four mesophilic anaerobic digesters were subjected to a sequential decrease in HRT from 15 to 8, 4 and 2 days while keeping the OLR constant at chemical oxygen demand of 1 gCOD L r-1 d-1. The results showed that HRT alone was insufficient to washout methanogens from the digesters, which in turn prevented the accumulation of volatile fatty acids (VFA). Methanosaeta was the dominant genus in the four digesters at all HRTs. Metabolic rates showed that process performance was controlled by hydrolysis, with a clear shift in acetogenic rates, from butyrate and propionate degradation to ethanol degradation at 4 and 2d HRT. The change in acetogenic pathways was attributed to a shift in the fermentation pathways co-current with changes in fermentative bacteria. At 2d HRT, biofilm was formed on the walls and paddles of the digesters, probably as a survival strategy. Most of the taxa in the biofilm were also present in the digester media. Overall, it is the combination of HRT with other operational parameters which promotes the washout of methanogens and the accumulation of VFA.

Unravelling the economics behind mainstream anaerobic membrane bioreactor application under different plant layouts.

This research evaluated the economic feasibility of anaerobic membrane bioreactor (AnMBR) as a mainstream technology for municipal sewage treatment. To this end, different wastewater treatment plant (WWTP) layouts were considered, including primary settler, AnMBR, degassing membrane, partial nitritation-Anammox, phosphorus precipitation and sidestream anaerobic digestion. The net treatment cost of an AnMBR-WWTP decreased from 0.42 to 0.35 € m-3 as the sewage COD concentration increased from 100 to 1100 mg COD L-1 due to revenue from electricity production. However, the net treatment cost increased above 0.51 € m-3 when nutrient removal technologies were included. The AnMBR and partial nitritation-Anammox were the costliest processes representing a 57.6 and 30.3% of the treatment cost, respectively. Energy self-sufficiency was achieved for high-strength municipal sewage treatment (1000 mg COD L-1) and a COD:SO42--S ratio above 40. Overall, the results showed that mainstream AnMBR has potential to be an economically competitive option for full-scale implementation.

Volatile fatty acids production from biowaste at mechanical-biological treatment plants: Focusing on fermentation temperature.

The impact of temperature (20, 35, 45, 55, 70 °C) on volatile fatty acid (VFA) production from biowaste collected at a mechanical-biological treatment plant was analysed. Additionally, relevant streams of the treatment plant were characterised to assess seasonality effects and conceive the integration of a fermentation unit. Batch fermentation tests at 35 °C showed the highest VFA yields (0.49-0.59 gCODVFA/gVS). The VFA yield at 35 °C was 2%, 6%, 10% and 14% higher than at 55, 45, 20 and 70 °C, respectively. The VFA profile was not affected by the fermentation temperature nor seasonality and was dominated by acetic, propionic and butyric acid (75-86% CODVFA). The concentration of non-VFA soluble COD and ammoniacal nitrogen in the fermentation liquor increased with temperature. The fermentation unit in the treatment plant was conceived after the pulper and hydrocyclones and before the anaerobic digester, while the fermenter temperature depends on the VFA application.

Exploring the potential of co-fermenting sewage sludge and lipids in a resource recovery scenario.

In this study, co-fermentation of primary sludge (PS) or waste activated sludge (WAS) with lipids was explored to improve volatile fatty acid production. PS and WAS were used as base substrate to facilitate lipid fermentation at 20 °C under semi-aerobic conditions. Mono-fermentation tests showed higher VFA yields for PS (32-89 mgCOD gVS-1) than for WAS (20-41 mgCOD gVS-1) where propionate production was favoured. The principal component analysis showed that the base substrate had a notable influence on co-fermentation yields and profile. Co-fermentation with WAS resulted in a greater extent of oleic acid degradation (up to 4.7%) and evidence of chain elongation producing valerate. The occurrence of chain elongation suggests that co-fermentation can be engineered to favour medium-chain fatty acids without the addition of external commodity chemicals. BMP tests showed that neither mono-fermentation nor co-fermentation had an impact on downstream anaerobic digestion.

Techno-economic analysis of combining forward osmosis-reverse osmosis and anaerobic membrane bioreactor technologies for municipal wastewater treatment and water production.

The economic feasibility of combining forward osmosis (FO), reverse osmosis (RO) and anaerobic membrane bioreactor (AnMBR) technologies for municipal wastewater treatment with energy and water production was analysed. FO was used to pre-concentrate the AnMBR influent, RO for draw solution regeneration and water production, and AnMBR for wastewater treatment and energy production. The minimum wastewater treatment cost was estimated at 0.81 € m-3, achieved when limiting the FO recovery to 50% in a closed-loop scheme. However, the cost increased to 1.01 and 1.27 € m-3 for FO recoveries of 80% and 90%, respectively. The fresh water production cost was estimated at 0.80 and 1.16 € m-3 for an open-loop scheme maximising water production and a closed-loop scheme, respectively. The low FO membrane fluxes were identified as a limiting factor and a sensitivity analysis revealed that FO membrane fluxes of 10 LMH would significantly improve the competitiveness of FO-RO + AnMBR technology.

Anaerobic co-digestion of food waste and waste activated sludge: ADM1 modelling and microbial analysis to gain insights into the two substrates' synergistic effects.

The reasons for the acidification problem affecting Food Waste (FW) anaerobic digestion were explored, combining the outcomes of microbiological data (FISH and CARD-FISH) and process modelling, based on the Anaerobic Digestion Model n°1 (ADM1). Long term semi continuous experiments were carried out, both with sole FW and with Waste Activated Sludge (WAS) as a co-substrate, at varying operational conditions (0.8-2.2 g VS L-1 d-1) and FW / WAS ratios. Acidification was observed along FW mono-digestion, making it necessary to buffer the digesters; ADM1 modelling and experimental results suggested that this phenomenon was due to the methanogenic activity decline, most likely related to a deficiency in trace elements. WAS addition, even at proportions as low as 10% of the organic load, settled the acidification issue; this ability was related to the promotion of the methanogenic activity and the consequent enhancement of acetate consumption, rather than to WAS buffering capacity. The ability of the ADM1 to model processes affected by low microbial activity, such as FW mono-digestion, was also assessed. It was observed that the ADM1 was only adequate for digestions with a high activity level for both bacteria and methanogens (FISH/CARD-FISH ratio preferably >0.8) and, under these conditions, the model was able to correctly predict the relative abundance of both microbial populations, extrapolated from FISH analysis.

Benefits and drawbacks of food and dairy waste co-digestion at a high organic loading rate: A Moosburg WWTP case study.

Anaerobic co-digestion (AcoD) is a key technology in reframing organic waste as a viable energy source. A lack of documented experience on full-scale AcoD at wastewater treatment plants (WWTPs) has created a bottleneck in AcoD implementation, which is further tightened by the focus of existing AcoD studies being on low co-substrate loading (<50%) and the obtainable benefits. This study aims to fill this gap by investigating the drawbacks and benefits of high-ratio co-substrate dosing of food and dairy wastes at the Moosburg WWTP (Germany) from 2014 to 2017. The Moosburg WWTP co-digests sewage sludge, food waste, and dairy wastes at a 35:47:18 ratio by volatile solids (organic loading rate (OLR) of 3.0 kgVS/(m3·day)). During the study period, this high co-substrate dosing increased the methane potential by 300 ±â€¯50%. The corresponding high methane yield significantly increased the on-site electricity production, resulting in energy neutrality in 2014-2015. The corresponding economic gain from gate fees was 48,000 ±â€¯5,000 € per year. The observed drawbacks included solids accumulation inside the digester (5 m3/month), high nitrogen backload (65% increase from co-substrate addition), reduced retention time (loss of 1.18 days/year from solids accumulation), and reduced dewaterability. The high nitrogen content in the centrate is treated by sequential batch reactors (SBRs), using lactose as the carbon source for denitrification. This study presents an alternative approach for determining gate fees based on the economic gains from inherent methane content, which identified waste milk, lactose and grease trap sludge as the most profitable co-substrates.

Systematic error in manometric measurement of biochemical methane potential: Sources and solutions.

This work focused on identification and quantification of systematic sources of error in manometric measurement of biochemical methane potential (BMP). Error was determined by comparison to gravimetric measurements and direct measurement of leakage. One out of three types of septa leaked above 1 bar (gauge) headspace pressure, losing 25 to 30% of biogas produced. But manometric BMP showed a negative bias even in the absence of leakage. Maximum error was 24% from 160 mL bottles with 40 mL of headspace (headspace fraction of 0.25). Error decreased with increasing headspace fraction, and was small (3%) for a headspace fraction of 0.75, showing that a high headspace volume is the best approach for minimizing error. Relative error in CH4 production measurement increased with headspace pressure as well, but controlling pressure alone is not sufficient for minimizing error. Calculations showed that observed error may be due to volatilization of CH4 during venting as well as inaccurate headspace volume determination, although these sources do not completely explain the magnitude of error observed. Measurement of biogas composition before and after venting showed that CO2 volatilization can occur, but is probably a minor source of error. Calculations showed that error in estimation of ambient pressure or headspace temperature had only minor effects (<3%). Gravimetric measurements, which were unaffected by leakage and insensitive to error in estimation of headspace pressure, temperature or volume, can provide a simple check on manometric results, or a complete replacement.

Biogas production coupled to repeat microalgae cultivation using a closed nutrient loop.

Anaerobic digestion is an established technology to produce renewable energy as methane-rich biogas for which microalgae are a suitable substrate. Besides biogas production, anaerobic digestion of microalgae generates an effluent rich in nutrients, so-called digestate, that can be used as a growth medium for microalgal cultures, with the potential for a closed nutrient loop and sustainable bioenergy facility. In this study, the methane potential and nutrient mobilization of the microalga Scenedemus dimorphus was evaluated under continuous conditions. The suitability of using the digestate as culture medium was also evaluated. The results show that S. dimorphus is a suitable substrate for anaerobic digestion with an average methane yield of 199 mL g-1 VS. The low level of phosphorus in digestate did not limit algae growth when used as culture medium. The potential of liquid digestate as a superior culture medium rather than inorganic medium was demonstrated.

The role of COD/N ratio on the start-up performance and microbial mechanism of an upflow microaerobic reactor treating piggery wastewater.

This study investigated the role of COD/N ratio on the start-up and performance of an upflow microaerobic sludge reactor (UMSR) treating piggery wastewater at 0.5 mgO2/L. At high COD/N ratio (6.24 and 4.52), results showed that the competition for oxygen between ammonia-oxidizing bacteria, nitrite-oxidizing bacteria and heterotrophic bacteria limited the removal of nitrogen. Nitrogen removal efficiency was below 40% in both scenarios. Decreasing the influent COD/N ratio to 0.88 allowed achieving high removal efficiencies for COD (∼75%) and nitrogen (∼85%) due to the lower oxygen consumption for COD mineralization. Molecular biology techniques showed that nitrogen conversion at a COD/N ratio 0.88 was dominated by the anammox pathway and that Candidatus Brocadia sp. was the most important anammox bacteria in the reactor with a relative abundance of 58.5% among the anammox bacteria. Molecular techniques also showed that Nitrosomonas spp. was the major ammonia-oxidiser bacteria (relative abundance of 86.3%) and that denitrification via NO3- and NO2- also contributed to remove nitrogen from the system.