Formation of Greenhouse Gases in Woodchip Denitrification Treating Aquaculture Effluents – A Case Study

Recirculating Aquaculture Systems (RAS) are used in land-based aquaculture production. In recent decades, they have become increasingly important, as they are based on efficient water recycling and treatment and consume less water per kg of fish produced. In RAS, the rearing conditions are stable throughout the year. Typically, low volumes of effluents are discharged from a RAS, but the effluent is very nutrient-rich and requires end-of-pipe treatment. Nutrients such as nitrate with eutrophication properties in the environment need to be removed.

Various technologies are available for nitrate removal, including ion exchange, adsorption, membrane separation, electrodialysis, and biological denitrification [1,2]. Woodchip bioreactors are passive systems for nitrate removal. In recent years, woodchip bioreactors have been used in treating agricultural water [3], effluents from animal husbandry [4], stormwater [5], and RAS water [6,7]. 

Biological denitrification is carried out by microbes which use nitrate as their electron acceptor and organic substances as an electron donor to get energy from organic matter for microbial growth [8]. In woodchip bioreactors, the diverse communities include e.g., Acidithiobacillales, Desulfovibrionales, Rhodocyclales, and Chlorobiales with the key denitrifying communities Burkholderiales, Pseudomonadales, Rhizobiales, and Rhodobacterales but they vary in different systems, conditions, or by woodchip species [9]. 

Biological denitrification can be classified as heterotrophic and autotrophic. In autotrophic denitrification, hydrogen, iron or sulfur compounds are used as the energy source and inorganic carbon as the carbon source [10]. The heterotrophic denitrifiers use organic carbon compounds as a carbon source [11]. Water-soluble substances (methanol, acetic acid, glucose) are used as a carbon source for denitrification [12-14], but some of them pose risks during storage, transportation, and operation due to their toxicity and inflammability [2]. 

Woodchips consist of a porous matrix with high hydraulic conductivity and a large specific surface area for microorganisms to attach to [15]. During incomplete denitrification, greenhouse gas nitrous oxide (N₂O) can be formed, which has a global warming potential 265-298 times higher than carbon dioxide (CO₂, [16] and ozone-depleting properties [17]. Additionally, methane (CH₄) is a greenhouse gas with 28-fold (including indirect effects) warming potential compared to CO₂ [18]. Both can be released from the denitrifying bioreactors derived from decaying organic matter [15]. 

In woodchip bioreactors, woodchips act as a growth surface for microbial biofilm and as a source of carbon [19]. The aim of denitrification is to convert nitrate (NO₃^-) into nitrogen gas (N₂) and release it into the atmosphere [20]. It is a stepwise process catalyzed by enzymes (nitrate reductase, nitrite reductase, nitric oxide reductase, nitrous oxide reductase) and carried out by denitrifiers, including Pseudomonas, Bacillus, Thiobacillus, and Propionibacterium [21]. Typically, nitrate removal rates of 39 g NO₃-N m^-3 d^-1 [6], 2-22 g NO₃-N m^-3 d^-1 [18], and 7.2 ±9.6 g N m^-3 d^-1 (n=27, [22]) have been achieved. As early as in 1995, woodchips were assessed as a suitable and slowly degrading carbon source for denitrifying microorganisms [23-25]. 

Incomplete denitrification can produce N₂O as a by-product in sub-ideal conditions [15], which can depend on many parameters, e.g., oxygen (anoxic conditions required for the reductase enzymes, [26]) and carbon availability (N and C are needed to ensure the presence of electron acceptors and electron supply, [26]), temperature, redox potential, and microbial population (physical factors affect the availability of e.g., oxygen and carbon, and microbial activity, [26]; [27,28]). The availability of carbon is one of the key factors which regulates the proportion of incomplete denitrification and N₂O formation [29]. The availability of carbon may change with bioreactor´s operation age [9]. 

Environmental factors, such as carbon availability [30] or temperature [6] are crucial in controlling denitrification. Temperature affects all microbial activities, including activity in the hydrolysis of solid substrate and nitrate reduction. At a low temperature, the efficiencies of those activities decrease, leading to a decreased denitrification [31] due to suboptimal conditions. For example, denitrification and NO₃ removal efficiency may reduce at low water temperatures to 10-20% at 5 ºC [32]. In a study by Shen et al. [33], nitrate removal efficiency decreased from 92.5% to 69% (at 15 ºC) and to 50% (at 5 ºC). Additionally, Cameron and Schipper [34] reported nitrate removal that was 1.2-2.3 times higher at 23.5 ºC than at 14 ºC, although a higher temperature may lead to the faster microbial decomposition of the carbon source [34] and the release of CH₄ and CO₂ [15,35]. 

The Hydraulic Retention Time (HRT) of a denitrification reactor has an impact on nitrate removal efficiency. For example, Wang and Wang [16] and Gibert et al. [19] showed that decreased HRT led to increased effluent nitrate concentration, while longer HRT led to higher nitrate removal efficiency. Audet et al. [36] reported that HRT above 60h favored NO₃ removal and low N₂O emissions. However, lower CH₄ production has been reported with decreased HRT (2 h HRT: 0.51 g C m³ day^-1, 8 h HRT: 1.50 g C m³ day^-1, [17]). 

In our previous studies [37,38], high nitrogen removal efficiency was reported in passive woodchip denitrification, with low or absent toxic or harmful compounds. This experiment was motivated by studying the GHGs (CO₂, CH₄, and N₂O) which may be released from the woodchip bioreactor at four sampling points to assess the bioreactor’s efficiency and seasonal changes in concentrations. It was hypothesized that the nitrate removal efficiency and formation of unwanted N₂O would be influenced by the colder winter temperatures.

Experimental setup 

The experiment was conducted at the Laukaa fish farm (April 2021-February 2022) of Natural Resources Institute Finland (Luke), using a pilot-scale RAS (FREA Aquaculture solutions, Denmark). The full description of the RAS has been reported in Pulkkinen et al. [39]. 

The RAS consisted of two identical units, each unit with two 5 m³ tanks and a 1 m³ space for sludge cones, which collect settleable solid material and uneaten feed. From the tanks, water flowed through a drum filter (60 µm mesh size, Hydrotech HDF800, Veolia, France) and two parallel 2.5 m³ fixed bed bioreactors filled with 1.5 m³ Saddle chips (KSK Aqua, Denmark). Water flowed (water flowrate measured with Fluxus F501, Flexim, Germany) through a 2.24 m³ degassing unit and a 0.74 m³ pump sump. Finally, it was pumped through a low-head oxygenator (FREA, Aquaculture Solutions, Denmark) back to the fish tanks. Dissolved oxygen (Oxi:lyser, s::can, Austria) was monitored online and maintained above 8.0 mg L^-1 in the rearing tanks. The measurement data were stored on an industrial computer (Con::cube, S::can, Austria). 

The water temperature was adjusted to 12.8 ºC by controlling the hall air temperature. The pH was maintained at 7.5 (ProMinent, Germany) by adding dissolved sodium bicarbonate to the pump sumps (EJ-R, Iwaki, Japan). Clean inlet water (Watson Marlow 630, Spirax-Sarco Engineering, UK) was led from the oligotrophic Lake Peurunka (62.44886, 25.85201, area 694 ha, 59 600 m³). The inlet water was a 1:1 mixture of surface water (depth of 4 m) and the aphotic layer (depth of 8 m). Replacement water from Lake Peurunka was taken at 500 L kg feed ^-1 (5.2–7.2 m³ d^-1). 

A selected proportion of the circulating water was treated by a passive water treatment system, which included a woodchip bioreactor for nitrogen removal, wetland to control the Biological Oxygen Demand (BOD) and the Chemical Oxygen Demand (COD), and sand filtration for suspended solids and organic carbon removal [38,39]. The woodchip bioreactor was 14 m x 9 m (=118 m², 50 m³) and 1.5 m deep (1 m+0.5 m dry layer) of birch wood chips with porosity of 0.65. A 0.5m layer of dry woodchips on top were designed to act as an insulating layer during cold winters. Water was channeled from one side of the bioreactor with two perforated pipes 3.3 m and 4.6 m long at the base (Figure 1). The HRT of the woodchip bioreactor was on average 2.2 d (0.9-7.8 d) (Supplementary Figure S1). 

Figure 1 : Schematic figure of the woodchip bioreactor (width 3.3 m, length 9 m) with inlet and outlet pipes (blue), sampling cones (yellow, 1-4), and inlets for maintenance (light blue). The direction of the gradient has been marked with a red arrow.

Supplementary Figure S1: Replacement water from Lake Peurunka (red, L kg^-1 feed) and from the passive water treatment system (gray, L kg^-1 feed), and hydraulic retention time (HRT, blue, d) at the woodchip biofilter.  

The Long-Term Average (LTA) annual temperature (1961–2020) in central Finland (Jyväskylä) is 4.2°C, and the average annual precipitation is 621 mm (Finnish Meteorological Institute 2023). In 2021, annual precipitation was 675.3mm (Finnish Meteorological Institute 2023). Of the total precipitation, approximately 50% falls as snow. Snow cover typically appears in mid-November and melts in late April. The coldest month is February (-7.4 °C), and the warmest is July (17.0 °C). In central Finland (Jyväskylä), the average snow coverage duration is 148 days and snow depth is 40-60 cm (Finnish Meteorological Institute 2023). The local temperature and snow cover data are shown in Supplementary (Figure S2) (Finnish Meteorological Institute 2024).

Supplementary Figure S2: Snow depth (cm), air temperature (ºC) and nitrate (NO₃-N) removal efficiency (%) in central Finland near the area of the experiment (April 2021-February 2022). 

In the experiment, a total of 3164 one-year-old vaccinated full female rainbow trout (Oncorhynchus mykiss) was reared. The fish originated from the Hanka-Taimen hatchery (Hankasalmi, Finland). Their average weight was 494 g, with a total biomass of 1564 kg at 78 kg m^-3. The fish were fed with an automated feeding system (T Drum 2000, Arvo-Tec, Finland) with a commercial fish feed (BioMar Orbit, 3.5 mm and 6 mm), resulting in a Feed Conversion Ratio (FCR) of 1.2. Supernumerary fish were regularly removed to maintain the tank biomass and fish density at a suitable level. The fish were visually inspected on a daily basis to monitor their health and well-being. 

The study followed the protocols approved by the Luke Animal Care Committee, Helsinki, Finland, and EU Directive 2010/63/EU for animal experiments.

Water quality parameters 

At the woodchip bioreactor, the temperature was measured using a YSI EXO probe (Xylem, USA) and a HOBO Pendant (MX2201, Onset Computer Corporation, USA). Oxygen was measured using YSI EXO, YSI ProODO, and Ponsel OPTOD optical probes (Aqualabo Servises SA, France) as explained in Pulkkinen et al. [39]. 

Water samples were taken from the inlet and outlet of the woodchip biofilter to measure the following variables. Water samples were collected in clean high-density polyethene (HDPE) jars and stored frozen until further analysis. The total nitrogen (Procedure 8038 Nessler, TAN, 0.8 mg L^-1), nitrite-N (LCK341, 0.105–0.108 mg L^-1), nitrate-N (LCK340, 5-35 mg L^-1 and LCK342, 44.2–65.4 mg L^-1), COD (LCK1414, 5-60 mg L^-1), sulfate (SulfaVer, Permachem® reagents, 40-150 mg L^-1), and phosphate-P (LCK349, 0.05-1.50 mg L^-1) were analyzed with quick spectrophotometric tests for DS 3900 (Hach, Loveland, USA). The water alkalinity (88.3–113.1 mg L^-1) was measured by a standard titration method ISO 9963-1:1994 (TitraLab AT1000, Hach, Loveland, USA), and turbidity (5.5–6.6 NTU) with a Hach 2100q Turbidimeter (Hach, Loveland, USA).

GHG flux measurements 

The N₂O, CH₄, and CO₂ emissions from the woodchip bioreactor were measured using a static chamber method with Polypropylene (PP) cones (90 cm in height, diameter 15 cm) and PP cover with a rubber collar (n=4, height 10 cm, diameter 15 cm). The cones were pre-installed in the woodchip bed at a depth of 10 cm. The cones were placed on both sides (3 m apart) and different distances from the effluent manifold at the inlet side of the woodchip bioreactor (Figure 1) to give a comprehensive overview of the bioreactor efficiency. 

The samples were taken on average once a month (4/2021-2/2022) but more frequently in the summer. First, the cones were closed with covers, and gas samples of 20 mL were taken with a 60 mL polypropene (PP) syringe at 2 min intervals from the headspace of the chamber. In each measurement, 20 mL of chamber air was sampled through a rubber septum with a polypropylene syringe (BD Plastipak, Becton, Dickinson and Company, Franklin Lakes, NJ, USA) and transferred into pre-evacuated 20 mL glass vials (Exetainer, Labco Ltd., High Wycombe, UK). The air in the chamber was mixed with one syringe flush before each sampling.

Chemical analysis 

The GHG analyses were performed with a gas chromatograph (GC, Agilent 6890N, Agilent Technologies, Santa Clara, CA, USA), a Flame Ionization Detector (FID), and an Electron Capture Detector (ECD). The method for Agilent 7890A GC with SP1 7890-0468 configuration (Supplementary Figure S3) was used to analyze GHGs (CO₂, CH₄, N₂O), modified from Nieminen et al. (2015). The system included three valves and two columns (G3591-81004 and G3591-81121). The detector FID was used to detect CH₄ and CO₂, and ECD to detect N₂O. The system was modified for using automated headspace sampling (Gilson GX-271).

Supplementary Figure S3: Configuration of SP1 7890-0468 in GC analysis of gaseous samples. 

At the GC, the following conditions were maintained: valve temperature 80 ºC, oven temperature 70 ºC, methanizer temperature 375 ºC, sample loop 2 mL, and column flow (N₂) 21 mL min^-1 (at 60 ºC). At FID, a temperature of 250 ºC, H₂ flow of 48 mL min^-1, airflow of 350 mL min^-1, and make-up (N₂) of 2 mL min^-1 were used. At the ECD, the temperature was 350 ºC, and the make-up gas (Ar/ 5% CH₄) was 2 mL min^-1. 

A standard gas mixture (AGA Gas AB, Lidingö, Sweden) of compressed air with 29.9 mg L^-1 of N₂O, 201 mg L^-1 CH₄, and 3010 mg L^-1 CO₂ was used for the daily calibration and preparation of the calibration curve. Concentrations in the samples were determined based on external calibration (0; 0,5; 1; 5; 15; 20 mL of standard gas mixture). Cumulative fluxes were calculated by assuming linear changes between subsequent measurements. The Limits Of Detection (LOD) and Quantification (LOQ) for the GC analysis were 24.6 µg L^-1 and 162.2 µg L^-1 for N₂O, 5.86 mg L^-1 and 89.3 mg L^-1 for CO₂, and 157.0 µg L^-1 and 718.9 µg L^-1 for CH₄.

Statistical analysis

A statistical analysis of the GHGs was performed using IBM SPSS Statistics (IBM Corp.^©, version 27.0.1.0, Armonk, NY, United States). Statistically significant differences between normalized results of GHGs (N₂O vs. CO₂; N₂O vs. CH₄; CH₄ vs. CO₂, nitrate removal (%) vs. temperature, and nitrate removal efficiency (%) vs. HRT) at different time points were calculated with an independent samples t-test at a statistical significance level of p < 0.05. 

LODs and LOQs of GC analyses were determined with a regression analysis of variance (ANOVA) test. The limit for the statistical significance level was set at p < 0.05.

Water quality 

Nitrate-N decreased from 65 mg L^-1 to below 20 mg L^-1 during the woodchip bioreactor, and pH decreased from 8 to 6 (Figures 2A and 2B). Overall, lower values were detected from the fall of 2021 to the spring of 2022. Concentrations of ammonium and nitrate-N remained very low throughout the experiment. Additionally, alkalinity was higher overall before the biofilter (up to 70 mg L^-1 vs. 20-45 mg L^-1) and fluctuated during the experiment. Concentrations of sulfate and phosphate remained quite constant throughout the experiment (below 30 mg L^-1, (Figures 3A&3B)). However, some increased COD values were detected in the beginning of the experiment.

Figure 2: Axis on the left: concentrations of ammonium (NH₄-N, mg L^-1), nitrite (NO₂-N, mg L^-1), nitrate (NO₃-N, mg L^-1), pH, and turbidity (NTU); axis on the right: alkalinity (mg L^-1) before (A) and after (B) the woodchip bioreactor.

Figure 3: Concentrations (mg L^-1) of sulfate (SO₄^2-), phosphate (PO₄-P), and COD before (A) and after (B) the woodchip bioreactor. 

The nitrate removal rate ranged from 7 to 26 g N m^-3 d^-1 (on average 18 g N m^-3 d^-1) throughout the experiment. HRTs showed a moderate positive correlation with NO₃-N removal efficiency (R²=0.340; Supplementary (Figure S4)).

Supplementary Figure S4: HRT, d from April 2021 to November 2021 plotted against NO₃-N removal efficiency, %. 

The temperature ranged between 5 and 23 ºC at the woodchip biofilter (Supplementary Figure S5A). The results marked in blue were measured in the pipeline back to the RAS. Supplementary Figure S5 B shows that the inlet water from the RAS contained 5-12 mg L^-1 of dissolved oxygen. The oxygen content decreased rapidly and remained very low in the following steps of anaerobic denitrification.

Supplementary Figure S5: Temperature (ºC, A) and dissolved oxygen (mg L^-1, B) before (well 1, green) and after (well 2, red) the woodchip biofilter, and in the pipeline back to the RAS (well 3, light blue). Measurements are missing between October 2021 and May 2022 due to instrumental failure. 

The inflow to the woodchip bioreactor ranged from 1 to 14 m³ d^-1 (Figure 4). The outflow followed the trend of the inflow although it varied occasionally due to the addition of lake water to the inlet from Lake Peurunka and supernatant from the sludge treatment to the outlet flow.

Figure 4: Inflow (m³ d^-1) from the RAS to the passive water treatment system and the woodchip biofilter and outflow (m³ d^-1) back to the RAS.

GHGs 

The N₂O removal rate ranged from 0 to 45 µg m^-2 h^-1 (0-13.4 mg CO₂-eq m^-2 h^-1, Figure 5), and for CO₂ from 10 to 450 mg m^-2 h^-1 (Figure 6). The rates of CH₄ ranged from 0.1 to 2.5 mg m^-2 h^-1 (2.8-70 mg CO₂-eq m^-2 h^-1, Figure 7). For all the studied compounds, the highest rates were observed in the inlet of the woodchip bioreactor (Tubes 1&2) but decreased rapidly at the outlet end of the bioreactor (Tubes 3&4). In addition, the rates varied seasonally. The highest removal rates were observed in the early spring for CH₄ (Figure 7) but in the late summer and fall for N₂O (Figure 5) and CO₂ (Figure 6). 

The normalized results of N₂O positively correlated with those of CO₂ (p=0.029, t₁₉₂=1,899). On the other hand, CO₂ and CH₄ did not show a significant correlation (p=0.286, t₁₉₂=-0.567). This was also the case for N₂O and CH₄ (p=0.073, t₁₉₂=1,457).

Figure 5: Observed N₂O removal rates at the woodchip bioreactor (µg m^-2 h^-1) during the one-year sampling period. Samples were taken from four sampling points (Tubes 1-4; see Figure 1).

Figure 6: Observed CO₂ removal rates at the woodchip bioreactor (mg m^-2 h^-1) during the one-year sampling period. Samples were taken from four sampling points (Tubes 1-4; see Figure 1).

Figure 7: Observed CH₄ removal rates at the woodchip bioreactor (mg m^-2 h^-1) during the one-year sampling period. Samples were taken from four sampling points (Tubes 1-4; see Figure 1). 

The nitrate removal ranged from 66% to 98% (Figure 8, Table 1). It seemed to follow changes in air temperature (Figure 8, Supplementary Figure S2) but without a statistical significance (positive correlation, p=0.217, t=0.795). The N₂O production (based on NO₃-N) ranged in most cases between 1.3% and 17.2%, excluding two events of higher N₂O-N production in August and December (Table 1).

Figure 8: NO₃-N removal (as g NO₃-N m^-3 d^-1 and % removed) with local temperature (ºC) during the study period (4/2021-2/2022).

Table 1: Removal of NO₃-N (%), proportion of produced N₂O-N vs. NO₃-N (%), temperature (ºC), and nitrate removal (g NO₃-N m^-3 d^-1).

n.a.: not analyzed 

• Nitrate removal 

In this study, the nitrate removal rates ranged from 7.2 to 26.2 g NO₃-N m^-3 d^-1 (average 18 NO₃-N m^-3 d^-1). They were somewhat higher than in our previous study [38], where the woodchip bioreactor achieved a nitrate removal rate of 11-15 g NO₃-N m^-3 d^-1. In both studies, the results were higher than those reported in previous studies by Hoffman et al. [40,41]; 1.67-2.22 g N d^-1 m^-3) and Christianson et al. [4]; 0.38-7.76 g N d^-1 m^-3), but in the same range (4.5-16.8 g N d^-1 m^-3) as in von Ahnen et al. [7]. The maximum nitrate removal rate (33 g NO₃-N m^-3 d^-1) of the system was determined by pre-tests at HRT of 48 h [30], but this level was not achieved during the experiment. 

• HRT 

Bioreactor removal efficiency depends on the HRT. In our study, the HRTs ranged from 0.9 d to 7.8 d (Supplementary Figure S2). The NO₃-N removal was moderately correlated with the HRT (R²=0.340), suggesting that longer HRTs might have led to increased nitrate removal (g NO₃-N m^-3 d^-1). This is in line with the results of Jéglot et al. [42] and Nordström and Herbert [43]. Both reported increased removal of NO₃-N and decreased N₂O formation at longer HRTs. The efficient removal of nitrate is crucial for the commercial farmers who use passive systems for denitrification. Poor operation and insufficient denitrification may prevent farmers meeting their discharge allowances and increasing their production [44]. 

Some studies suggest that shorter HRTs favor aerobic conditions and hinder denitrification [44,45]. Audet et al. [36,46] stated that N₂O emissions were higher at HRTs below 60 h (=2.5 days), while Addy et al. [47] detected significantly lower NO₃ removal and increased N₂O production in bioreactors with HRTs below 6 h. Even in this study, short HRTs may have promoted N₂O formation (Supplementary Figure S4). On the other hand, Davis et al. [17] reported lower CH₄ production from 2 h HRTs compared to 8 and 16 hour HRTs. 

• Environmental factors 

Denitrification is directly inhibited by CO₂ [2]. The atmospheric CO₂ concentration has increased from 280 ppm in 1800s to 400 ppm today [47,48]. In our study, even higher CO₂ concentrations were measured suggesting CO₂ formation due to microbial actions derived from decaying organic matter in the woodchip bed [15]. Increased CO₂ concentration inhibits electron transport and the consumption of electrons during denitrification by suppressing the synthesis and activity of the key enzymes [46]. 

Woodchips acted well as a carbon source for the denitrification, although some studies suggest that lignocellulosic material must first be hydrolyzed into soluble compounds before denitrifying bacteria can utilize the material as a substrate [5,49]. However, there are plenty of other microorganisms present in a woodchip biofilter, including saprophytic fungi which can degrade woodchip carbon into suitable form for denitrifiers [50]. 

As previously stated, denitrification is predominantly an anoxic process where N₂O production and reduction to N₂ depend on the local anoxic conditions. Water saturation and soil or woodchip bed structure regulate the oxygen supply. They determine the pathways through which gaseous and dissolved oxygen, nitrogen species, and dissolved organic matter may diffuse to the location of their consumption [26]. In our study, the oxygen content remained from low to moderate levels of 1.4-8.5 mg L^-1 in the inlet and very low 0.00-0.01 mg L^-1 in the outlet throughout the studied year. This suggests anoxic conditions in the system, although a low-oxygenated zones may be possible at the front end of the system. 

Variation in moisture content influences the microbial communities and biofilter performance [27]. For example, very low moisture content slows down microbial activities [51], while a too high moisture content leads to the filling of woodchip pores and restricts the transportation oxygen and nutrients [20]. In such cases, water competes with N₂O for hydrophilic spots (-OH and -COOH groups, [52]) which can lead to increased N₂O formation [53,54]. 

Only a few studies have been conducted on the role of denitrification at cold temperatures (∼5 °C, [55]). NO₃-N removal may decrease to about 10-20% due to the effects of temperature on microbial action. In this experiment, the woodchip bioreactor was insulated by additional woodchips on the surface of the moist layer (1.0 m depth of active media and 0.5 m of dry woodchips on top, [39]) to prevent the bioreactor freezing during the winter.

Annual mean efficiencies of bioreactors are often in the range of 50% [3,41,55]. However, almost complete nitrate removal (%) can be achieved at high temperatures (98% at 28ºC, [56]). In this study, the NO₃-N removal was above 60% even in the winter and reached above 90% on a few occasions. This suggests that the air temperature did not restrict the biofilter efficiency. 

Jéglot et al. [57] showed that the optimal temperature for NO₃ transformation to N₂ was between 20 and 30 ºC. High temperatures can only occasionally be achieved in northern latitudes in the summer (see Supplementary Figure S2). The pathway of NO₃ transformation to N₂ includes N₂O production [55]. At 5-10ºC (or lower), lag phases can occur, which leads to slow denitrification and more pronounced NO₂, NO, and N₂O accumulation [55]. 

• N₂O formation 

The highest concentrations of N₂O were found at the first sampling point (Figure 5, Tube 1), and the lowest at the fourth point (Figure 5, Tube 4). This suggests that the denitrification occurred progressively along the gradient as the water flowed and proceeded in the woodchip biofilter. Similar changes in concentrations were also observed for CO₂ and CH₄ (Figure 6 and 7). This is likely due to changing conditions (e.g., gradient of O₂ content, varying microbe communities) in the biofilter. 

The N₂O removal rates ranged up to 45 µg N₂O m^-3 h^-1, which corresponds 0.1-2.5 mg N L^-1, while Audet et al. [31] for example, reported lower N₂O rates of 1-200 µg N L^-1 in a woodchip bioreactor treating agricultural water. On the other hand, Maia et al. [20,58] reported higher (0.6-2 mg N₂O L^-1) formation at 60% moisture content. In this study, the main woodchip layer was saturated with water (the moisture content in the woodchips was not measured in this study). This may have been one factor influencing the observed N₂O rates. 

The highest N₂O concentrations were measured in the late summer and fall, which may be due to delayed effects of reactions occurring in the warmer summer period. This includes increased microbial action in the woodchip bed. Overall, N₂O formation seemed to follow the environmental conditions and biological processes in the woodchip bed, and the load in circulating water played a minor role. This is supported by the fact that inlet and outlet water flows, nitrate loads, and removal did not seem to follow the N₂O formation. Additionally, none of the water quality measurements was correlated with the N₂O formation. It is possible that some of the formed and soluble N₂O may have circulated back to the RAS with the effluent water [11]. 

In previous studies, 0.003-9.7% of removed NO₃ was converted to N₂O [59,60], and more recently, 1-6% conversion to N₂O has been reported (250-1250 m³ of aquaculture effluents [9]). Additionally, Audet et al. [36] achieved very low N₂O-N emissions per NO₃-N removal, which were from 0.6% to 2.4% on average. In this study, the N₂O-N formation remained mostly at 1.3-10%, but formation as high as 48% was detected. This suggests that the denitrification was at least occasionally incomplete. Based on IPCC evaluation, over 0.75% NO₃-N conversion to N₂O is considered increased pollution [15], and strict limitations were not achieved in this experiment. Previously, high N₂O formation has been found at low carbon availability and high nitrate concentrations [29]. This study was conducted on the second functioning year of the bioreactor, suggesting that the bioreactor and its processes were fully matured. 

• CH₄ and CO₂ formation 

In woodchip biofilter, even CH₄ can be produced if nitrate is depleted and CH₄ acts as electron acceptor by methanogenic archaea [18,61]. Methanogenic archaea and bacteria produce methane in anaerobic conditions [60,61]. For example, methane-producing microbial communities can include Bacillus sp., Clostridium sp., Desulfotomaculum sp., and Ruminococcus sp. [62]. However, microbial species were not determined in our study.

The highest nitrate concentrations in the inlet water were in the early summer at the time of the highest methane formation. This suggests that the conditions in the woodchip bed favored CH₄ formation, although not due to an insufficient nitrate load in the circulating water. Theoretically, the CH₄ formation at the bioreactor should be low when NO₃ concentrations remain sufficiently high to suppress methanogens, but this concept requires validation [18]. More likely, CO₂ and CH₄ were released from the denitrifying bioreactors derived from decaying organic matter [15].

In this study, the concentrations of N₂O correlated positively with the CO₂ results (p=0.029), while neither CH₄ and N₂O (p=0.286) nor CH₄ and CO₂ (p=0.073) showed significant correlations. A similar connection between CO₂ and N₂O was also observed by Maia et al. [20] and later by Nieminen et al. [63], although the latter studied emissions in soil instead of from a woodchip bioreactor. Even Maia et al. [64] did not observe a positive correlation between CO₂ and CH₄ or between CH₄ and N₂O concentrations. They also suggested that CO₂ was a good indicator of microbial activity in a denitrification bioreactor.

The COD in the inlet water was high in the spring and early summer when methane flux was high. Other gases such as CO₂ and CH₄ are released from the denitrifying bioreactors, both derived from decaying organic matter [18], which may have been the case in this study.

In some studies, CH₄ has been detected during early operation of the bioreactors [25] but has disappeared after a few months, possibly as the highly labile carbon in the woodchips was consumed. In this case, it was the second functioning year of the fully matured bioreactor, although highly labile carbon may still have been readily available.

The formation of CH₄ is connected with methanogens under anaerobic conditions and high moisture content [56]. which was probably the case here in the woodchip bed after a substantial amount of snow had melted (Supplementary Figure S3). Many studies have shown that moisture content and temperature are the main factors influencing methane and N₂O formation [64-66]). Even increased N₂O content is expected when moisture content is high [67-74].

Nitrate removal (%) remained high in the woodchip biofilter throughout the studied year, demonstrating the efficient operation of the biofilter. N₂O formation was mostly at a low level, but relatively high formation was occasionally observed. This suggests that denitrification may have occasionally been incomplete. 

Unlike the hypothesis, no reduction in the nitrate removal rate (g NO₃-N m^-3 d^-1) or increase in N₂O formation induced by the cold winter temperatures was observed. The formation of the studied GHGs did not follow the conditions in RAS or nitrogen load in the circulating water. It is more likely that the environmental conditions in the woodchip bed played the main role in determining the formed and released gases, and some effects may have occurred at a delayed pace. Although a thorough study of microbial processes in the woodchip bed were not the goal of this experiment but they definitely are an interesting subject for further study to fully understand the processes behind the formation of GHGs in denitrification.

Financial support provided by the European Maritime and Fisheries Fund is gratefully acknowledged. The fish husbandry and system management of the staff at Laukaa fish farm is greatly appreciated, as are the laboratory staff in Jokioinen for conducting the GC analyses.

The supplementary material includes (Supplementary Figures S1-S5): the amounts of replacement water and HRTs at the woodchip bioreactor (S1); air temperatures, snow coverage and nitrate removal efficiencies from April 2021 to February 2022 (S2); Configuration in GC analysis for the GHG samples (S3); a correlation plot of HRTs and nitrate removal efficiency from April 2021 to November 2021 (S4); temperature and dissolved oxygen before and after the woodchip biofilter, and in the pipeline back to the RAS (S5).

Ethics approval 

Not applicable.

Consent to participate 

Not applicable.

Consent to publish 

Not applicable.

Data availability statement 

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Author’s contribution 

Petra Lindholm-Lehto: Data curation, Formal analysis, Investigation, Methodology, Resources, Validation, Writing –original draft preparation

This work was supported by the European Union through the European Maritime and Fisheries Fund.

No potential conflict of interest was reported by the author. There are no relevant financial or non-financial competing interests to report.

1. Wang JL, Kang J (2005) The characteristics of anaerobic ammonium oxidation (ANAMMOX) by granular sludge from an EGSB reactor. Process Biochem 40: 1973-1978.
2. Wang J, Chu L (2016) Biological nitrate removal from water and wastewater by solid-phase denitrification process. Biotechnol Adv 34: 1103-1112.
3. Christianson L, Bhandari A, Helmers M, Kult K, Sutphin T, et al. (2012) Performance evaluation of four field-scale agricultural drainage denitrification bioreactors in Iowa. Trans ASABE 55:2163-2174.
4. Addy K, Gold AJ, Christianson LE, David MB, Schipper LA, et al. (2016) Denitrifying bioreactors for nitrate removal: a meta-analysis. J Environ Qual 45:873-881.
5. Lynn TJ, Yeh DH, Ergas SJ (2015) Performance and longevity of denitrifying wood-chip biofilters for stormwater treatment: A microcosm study. Environ Eng Sci 32: 321-330.
6. Lepine C, Christianson L, Sharrer K, Summerfelt S (2016) Optimizing Hydraulic retention times in denitrifying woodchip bioreactors treating recirculating aquaculture system wastewater. J Environ Qual 45: 813-821.
7. von Ahnen M, Bovbjerg Pedersen P, Dalsgaard J (2018) Performance of full-scale woodchip bioreactors treating effluents from commercial RAS. Aquacult Eng 83:130-137.
8. Ghafari S, Hasan M, Aroua MK (2008) Bio-electrochemical removal of nitrate from water and wastewater - a review. Bioresour Technol 99:3965 3974.
9. Aalto SL, Suurnäkki S, von Ahnen M, Siljanen HMP, Bovbjerg PP, et al. (2020) Nitrate removal microbiology in woodchip bioreactors: A case-study with full-scale bioreactors treating aquaculture effluents. Sci Total Environ 723: 138093.
10. Karanasios KA, Vasiliadou IA, Pavlou S, Vayenas DV (2010) Hydrogenotrophic denitrification of potable water: A review. J Hazard Mater 180:20-37.
11. van Rijn J, Tal Y, Schreier HJ (2006) Denitrification in recirculating systems: theory and applications. Aquacult Eng 34: 364-376.
12. Bill KA, Bott CB, Murthy SN (2009) Evaluation of alternative electron donors for denitrifying moving bed biofilm reactors (MBBRs). Water Sci Technol 60: 2647-2657.
13. Moghaddam R, Torres-Rojas D, Louis Schipper L (2022) Enhanced nitrate removal and side effects of methanol dosing in denitrifying bioreactors Ecol Eng 185: 106818.
14. Feyereisen GW, Wang H, Wang P, Anderson EL, Jang J, et al. (2023) Carbon supplementation and bioaugmentation to improve denitrifying woodchip bioreactor performance under cold conditions. Ecol Eng 191: 106920.
15. White SA, Morris SA, Wadnerkar PD, Woodrow RL, Tucker JP, et al. (2022) Anthropogenic nitrate attenuation versus nitrous oxide release from a woodchip bioreactor. Environ Pollut 300: 118814.
16. Wang XM, Wang JL (2013) Nitrate removal from groundwater using solid-phase denitrification process without inoculating with external microorganisms. Int J Environ Sci Technol 10: 955-960.
17. Davis MP, Martin EA, Moorman TB, Isenhart TM, Soupir ML (2019) Nitrous oxide and methane production from denitrifying woodchip bioreactors at three hydraulic residence times. J Environ Manage 242: 290-297.
18. Schipper LA, Robertson WD, Gold AJ, Jaynes DB, Cameron SC (2010) Denitrifying bioreactors An approach for reducing nitrate loads to receiving waters. Ecol Eng 36: 1532-1543.
19. Lopez-Ponnada EV, Lynn TJ, Peterson M, Ergas SJ, Mihelcic JR (2017) Application of denitrifying wood chip bioreactors for management of residential non-point sources of nitrogen. J Biol Eng 11: 16.
20. Maia GDN, Day VGB, Gates RS, Taraba JL (2012) Ammonia biofiltration and nitrous oxide generation during the start-up of gas-phase compost biofilters. Atmos Environ 46: 659-664.
21. Wrage N, Velthof GL, van Beusichem ML, Oenema O (2001) Role of nitrifier denitrification in the production of nitrous oxide. Soil Biol Biochem 33:1723 1732.
22. Christianson LE, Cooke RA, Hay CH, Helmers MJ, Feyereisen GW, et al. (2021) Effectiveness of denitrifying bioreactors on water pollutant re-duction from agricultural areas. Trans ASABE 64: 641-658.
23. Robertson WD, Cherry JA (1995) In situ denitrification of septic-system nitrate using reactive porous media barriers: Field trials. Groundwater 33: 99-111.
24. Robertson WD, Vogan JL, Lombardo PS (2008) Nitrate removal rates in a 15-year-old permeable reactive barrier treating septic system nitrate. Gr Water Monit Remediat 28: 65-72.
25. Moorman TB, Parkin TB, Kaspar TC, Jaynes DB (2010) Denitrification activity, wood loss, and N2O emissions over 9 years from a wood chip bioreactor. Ecol Eng 36: 1567-1574.
26. Rohe L, Apelt B, Vogel H-J, Well R, Wu G-M, et al. (2021) Denitrification in soil as a function of oxygen availability at the microscale. Biogeosci 18: 1185-1201.
27. Sakuma T, Jinsiriwanit S, Hattori T, Deshusses MA (2008) Removal of ammonia from contaminated air in a biotrickling filter-denitrifying bioreactor combination system. Water Res 42: 450-4513.
28. Walker JT, Geron CD, Vose JM, Swank WT (2002) Nitrogen trace gas emissions from a riparian ecosystem in southern Appalachia. Chemosphere Global Change Sci 49: 1389-1398.
29. Spinelli M, Eusebi AL, Vasilaki V, Katsou E, Frison N, et al. (2018) Critical analyses of nitrous oxide emissions in a full scale activated sludge system treating low carbon-to-nitrogen ratio wastewater. J Clean Prod 190: 517-524.
30. Kiani S, Kujala K, Pulkkinen JT, Aalto SA, Suurnäkki S, et al. (2020) Enhanced nitrogen removal of low carbon wastewaters in denitrification bioreactors by utilizing industrial waste toward circular economy. J Clean Prod 254: 119973.
31. Canziani R, Vismara R, Basilico D, Zinni L (1999) Nitrogen removal in fixed-bed submerged biofilters without backwashing. Water Sci. Technol. 40: 145-152.
32. Shen ZQ, Hu J, Wang JL, Zhou YX (2015) Biological denitrification using starch/polycaprolactone blends as carbon source and biofilm support. Desalin Water Treat 54: 1-7.
33. Cameron SG, Schipper LA (2010) Nitrate removal and hydraulic performance of organic carbon for use in denitrification beds. Ecol Eng 36: 1588-1595.
34. Tatti E, Goyer C, Chantigny M, Wertz S, Zebarth BJ, et al. (2014) Influences of over winter con-ditions on denitrification and nitrous oxide-producing microorganism abundance and structure in an agri-cultural soil amended with different nitrogen sources. Agricult Ecosyst Environ 183: 47-59.
35. Audet J, Jéglot A, Elsgaard L, Maagaard AL, Sørensen SR, et al. (2021) Nitrogen removal and nitrous oxide emissions from woodchip bioreactors treating agricultural drainage waters. Ecol Eng 169: 106328.
36. Lindholm-Lehto P, Pulkkinen JT, Kiuru T, Koskela J, Vielma J (2020) Water quality in recirculating aquaculture system using woodchip denitrification and slow sand filtration. Environ Sci Pollut Res 27: 17314-17328.
37. Lindholm-Lehto PC, Pulkkinen JT, Kiuru T, Koskela J, Vielma J (2021) Efficient water treatment achieved in recirculating aquaculture system using woodchip denitrification and slow sand filtration. Environ Sci Pollut Res 28: 65333-65348.
38. Pulkkinen JT, Ronkanen A-K, Pasanen A, Kiani S, Kiuru T, et al. (2021) Start-up of a “zero-discharge” recirculating aquaculture system using woodchip denitrification, constructed wetland, and sand infiltration. Aquacult Eng 93: 102161.
39. Lindroos A-J, Lindholm-Lehto P, Pulkkinen J, Kiuru T, Vielma J (2020) The Effect of filtration with natural esker sand on the removal of organic carbon and suspended solids from the effluent of experimental recirculating aquaculture systems. Water Air Soil Pollut 231: 209.
40. Hoffmann CC, Larsen SE, Kjaergaard C (2019) Nitrogen removal in woodchip-based biofilters of variable designs treating agricultural drainage discharges. J Environ Qual 48: 1881-1889.
41. Jéglot A, Miranda-Velez JF, Plauborg F, Elsgaard L (2023) Nitrate removal and environmental side-effects controlled by hydraulic residence time in woodchip bioreactors treating cold agricultural drainage water. Environ Technol 44: 4324-4333.
42. Nordström A, Herbert RB (2017) Denitrification in a low-temperature bioreactor system at two different hydraulic residence times: laboratory column studies. Environ Technol 38: 1362-1375.
43. Ahnen M, Bovbjerg Pedersen P, Dalsgaard J (2020) Improving denitrification in an aquaculture wetland using fish waste a case study. Ecol Eng 143: 105686.
44. Dalsgaard J, von Ahnen M, Naas C, Pedersen PB (2018) Nutrient removal in a constructed wetland treating aquaculture effluent at short hydraulic retention time. Aquacult Environ Interact 10: 329-343.
45. Wan R, Chen Y, Zheng X, Su Y, Li M (2016) Effect of CO2 on microbial denitrification via inhibiting electron transport and consumption. Environ Sci Technol 50: 9915-9922.
46. Wan R, Chen Y, Zheng X, Su Y, Huang H (2018) Effect of CO2 on NADH production of denitrifying microbes via inhibiting carbon source transport and its metabolism. Sci Total Environ 627: 896-904.
47. Zhang J-Z, Fischer CJ (2014) Carbon dynamics of Florida Bay: spatiotemporal patterns and biological control. Environ Sci Technol 48:9161-9169.
48. Desvaux M (2006) Unravelling carbon metabolism in anaerobic cellulolytic bacteria. Biotechnol Prog 22:1229-1238.
49. Aldossari N, Ishii S (2020) Isolation of cold-adapted nitrate-reducing fungi that have potential to increase nitrate removal in woodchip bioreactors. J Appl Microbiol131:197-207.
50. Yang L, Kent AD, Wang X, Funk TL, Gates RS, et al. (2014) Moisture effects on gas-phase biofilter ammonia removal efficiency, nitrous oxide generation, and microbial communities. J Hazard Mater 271:292-301.
51. Bratasz L, Kozlowska A, Kozlowski R (2012) Analysis of water adsorption by wood using the Guggenheim–Anderson-de Boer equation. Eur J Wood Wood Prod 70: 445-451.
52. Wallenstein MD, Myrold DD, Firestone M, Voytek M (2006) Environmental controls on denitrifying communities and denitrification rates: insights from molecular methods. Ecol Appl 16: 2143-2152.
53. Avrahami S, Bohannan BJA (2007) Response of Nitrosospira sp. strain AF-like ammonia oxidizers to changes in temperature, soil moisture content, and fertilizer concentration. Appl Environ Microbiol 73: 3121.
54. Jéglot A, Sørensen SR, Schnorr KM, Plauborg F, Elsgaard L (2021a) Temperature sensitivity and composition of nitrate-reducing microbiomes from a full-scale woodchip bioreactor treating agricultural drainage water. Microorganisms 9: 1331.
55. Li Y, Lu S, Zhang W, Liu X, Jiang H, et al. (2023) A Simultaneous denitrification and fermentation process for nitrate removal in a whiteleg shrimp (Litopenaeus vannamei) recirculating aquaculture system: Using endogenous carbon from a biofilm. J Mar Sci Eng 11: 2329.
56. Jéglot A, Audet J, Sørensen SR, Schnorr K, Plauborg F, et al. (2021b) Microbiome structure and function in woodchip bioreactors for nitrate removal in agricultural drainage water. Front Microbiol 12: 678448.
57. Warneke S, Schipper LA, Bruesewitz DA, McDonald I, Cameron S (2011) Rates, controls and potential adverse effects of nitrate removal in a denitrification bed. Ecological Engineering 37: 511-522.
58. David MB, Gentry LE, Cooke RA, Herbstritt SM (2016) Temperature and substrate control woodchip bioreactor performance in reducing tile nitrate loads in East-Central Illinois. J Environ Qual 45: 822 829.
59. Elgood Z, Robertson WD, Schiff SL, Elgood R (2010) Nitrate removal and greenhouse gas production in a stream-bed denitrifying bioreactor. Ecol Eng 36: 1575 1580.
60. Grießmeier V, Bremges A, McHardy AC, Gescher J (2017) Investigation of different nitrogen reduction routes and their key microbial players in wood chip-driven denitrification beds. Sci Rep 7: 17028.
61. Yang Y, Tsukahara K, Sawayama S (2008) Biodegradation and methane production from glycerol-containing synthetic wastes with fixed-bed bioreactor under mesophilic and thermophilic anaerobic conditions. Pro-cess Biochem 43: 362-367.
62. Nieminen M, Hurme T, Mikola J, Regina K, Nuutinen V (2015) Impact of earthworm Lumbricus terrestris living sites on the greenhouse gas balance of no-till arable soil. Biogeosci 12: 5481-5493.
63. Werner C, Kiese R, Butterbach-Bahl K (2007) N2O, CH4 and CO2 flux measurements from tropical rainforest soils in western Kenya. J Geophys Res 112: D03308.
64. Teh YA, Dubinsky EA, Silver WL, Carlson CM (2008) Suppression of methanogenesis by dissimilatory Fe(III)-reducing bacteria in tropical rain forest soils: implications for ecosystem methane flux, Glob Change Biol 14: 413-422.
65. Yashiro Y, Kadir WR, Okuda T, Koizumi H (2008) The effects of logging on soil greenhouse gas (CO2, CH4, N2O) flux in a tropical rain forest, Peninsular Malaysia. Agr Forest Meteorol 148: 799 806.
66. Chu LB, Wang JL (2016) Denitrification of groundwater using PHBV blends in packed bed reactors and the microbial diversity. Chemosphere 155:463-470.
67. Bohn HL, Bohn KH (1999) Moisture in biofilters. Environ Prog 18: 156-161.
68. Council Directive 2010/63/EU (2010) Directives on the protection of animals used for scientific purposes. Official Journal of the European Communities L 276: 33-79.
69. Finnish Meteorological Institute (2023) Statistics of temperature and rainfall in Finland from the year 1961. Online: https://www.ilmatieteenlaitos.fi/tilastoja-vuodesta-1961 (accessed December 20th, 2023).
70. Finnish Meteorological Institute (2024) Statistics of weather in Finland, open data. Online: https://www.ilmatieteenlaitos.fi/havaintojen-lataus (accessed March 4th, 2024).
71. Friedl J, Scheer C, Rowlings DW, Deltedesco E, Gorfer M, et al. (2020) Effect of the nitrification inhibitor 3,4-dimethylpyrazole phosphate (DMPP) on N-turnover, the N2O reductase-gene nos Z and N2O:N2 partitioning from agricultural soils. Sci Rep 10: 2399.
72. Popper R, Niemz P, Croptier S (2009) Adsorption and desorption measurements on selected exotic wood species. Analysis with the Hailwood–Horrobin model to describe the sorption hysteresis. Wood Res Slovakia 54: 43-55.
73. Rivas A, Barkle G, Stenger R, Moorhead B, Clague J (2020) Nitrate removal and secondary effects of a woodchip bioreactor for the treatment of subsurface drainage with dynamic flows under pastoral agriculture. Ecol Eng 148:105786.
74. van Lith C, Leson G, Michelsen R (1997) Evaluating design options for biofilters. J Air Waste Manag Assoc 47: 37-48.