Removal of dissolved organic matter from the woodchip bioreactor start-up by foam fractionation Open Access

Woodchip bioreactors are able to achieve continuous nitrate removal over several years from tile drainage and aquaculture effluents (Schipper et al. 2010; Christianson et al. 2012; Lepine et al. 2016; von Ahnen et al. 2018). It is well documented that, as a side effect, leaching of DOM from woodchips occurs during the first year after bioreactor installation with a high initial release of dissolved organic matter (DOM) followed by a rapid decline (Robertson & Cherry 1995; Schipper et al. 2010; von Ahnen et al. 2016a, 2016b, 2018). Fresh wood contains 1–2 wt% soluble organic constituents (Vogan 1993), which, when leached into the aquatic environment, can impair water quality. The leachate lowers pH, contains compounds such as phenols, tannins, lignins, resin acids and terpenes, which can be toxic at high concentrations and can leads to higher biological oxygen demand in recipients (Taylor et al. 1996; Taylor & Carmichael 2003; Tao et al. 2005; Hedmark & Scholz 2008). Lumber yard runoff (which can be considered as analogous to woodchip leachate) has also been shown to exhibit acute toxicity to rainbow trout (Oncorhynchus mykiss) (Bailey et al. 1999). Furthermore, the leached tannins, lignins and other phenolic substances may, if discharged, cause discolouration of the receiving water bodies, increasing their absorbance and thus decreasing light penetration for photosynthesis (Kritzberg et al. 2014; Svensson et al. 2014).

Suggested control measures to reduce DOM and nutrient wash-out during start-up include installing post-bioreactor treatment (e.g. sand filter), collecting the initial effluent for disposal elsewhere (e.g. for irrigation in agriculture) and pre-leaching woodchips prior to use (Schipper et al. 2010). However, more practical solutions still need to be found and properly evaluated, as the aquaculture industry is growing rapidly and new woodchip bioreactors are being installed as end-of-pipe treatment for recirculating aquaculture system (RAS) effluents. DOM leached from woodchip bioreactors typically consists of compounds with low degradability with a BOD₅/chemical oxygen demand (COD) ratio of around 0.2 (Tao et al. 2005; von Ahnen et al. 2016a, 2016b, 2018). A number of studies have recently investigated the effect of different bioreactor setups and carrier materials on bulk carbon concentrations. For example, Feyereisen et al. (2017) and Christianson et al. (2018) conducted laboratory studies to test the performance of anoxic post-bioreactor chambers filled with the inert plastic carriers but found no significant organic matter removal in the post-treatment. These results highlighted the low biodegradability of the leached DOM, due to which removal via physical–chemical treatment may be more effective than aerobic or anaerobic biological treatment (Forgie 1988). Abusallout & Hua (2017) were able to show that a recycled steel chip filter removed on average 44.2% of the dissolved organic carbon (DOC) leached from a laboratory woodchip bioreactor setup. The authors discussed that removal was likely due to the adsorption of the carboxyl/hydroxyl functional groups of the leached DOM onto iron oxide surfaces through ligand exchange reactions (Gu et al. 1994). Furthermore, ozonation has shown to be more effective than biological treatment alone for removing DOM from woodchip filter effluent. For example, in a laboratory study by Svennson et al. (2015), 1.5 L of oak wood leachate was exposed to ozone at specific ozone doses between 0.7 and 7 g/L O₃/g of initial COD removing >90% of polyphenols, up to 73% of COD, 61% of total organic carbon (TOC) and 97% of colour. A positive correlation between biodegradation and ozone pre-treatment was found. Though these approaches may be effective, their practical application to fish farms may be limited due to the additional costs and required process control involved.

A common observation during woodchip bioreactor start-up is the formation of foam in the bioreactor outlet, which may be an indication of the nonpolar/hydrophobic nature of some of the surface-active, DOM compounds (e.g. humic substances, fatty acids, lipids and proteins) leached from the woodchips. Abusallout & Hua (2017) found that the DOC released from woodchips was dominated by hydrophobic and high molecular weight (MW) (>10 KDa) carbon compounds. The hydrophobic fraction of DOC leached from woodchips may be particularly high during the initial leaching phase, as it was shown to decrease from 72.2 to 58% between Days 1 and 240 in a laboratory-scale woodchip bioreactor leachate study (Abusallout & Hua 2017). This indicates there is potential to remove larger amounts of the initially leached dissolved organics by utilising foam fractionation (Gregersen et al. 2021). Since hydrophobic natural organic matter is typically rich in aromatic carbon and UV-absorbing compounds (Hua & Reckhow 2007), we hypothesised that the fate and change in the character of leached DOM could be followed using fluorescence excitation-emission matrix (EEM) spectroscopy – a method that has previously been applied to characterise DOM in RAS water (Hambly et al. 2015).

The objectives of this study were to: (1) ascertain whether a fraction of the DOM released from woodchips after bioreactor start-up, could be removed via a novel approach based on continuous recirculation of the water between the woodchip bioreactor and a foam fractionator in a closed system and (2) investigate the fate, change in quantity and quality of the DOM within the entire system using fluorescence EEM spectroscopy.

Grab samples were taken over the course of the 11-day experimental period, from the 7 L reservoirs after 24 h and subsequently on every second day at identical times. Additionally, at each sampling time, the volume of the liquefied foam was recorded. Oxygen, pH and temperature were measured daily within the reservoir (bioreactor inlet water) using Hach Lange HQ40 multimeters (Düsseldorf, Germany).

Subsamples were filtered through 0.2 μm syringe filters (Filtropur S 0.2 μm, Sarstedt, Germany) prior to analyses for total ammonia nitrogen (TAN; DS 224 1975), total dissolved nitrogen N_total (ISO 7890-1 1986; ISO 11905-1 1997), nitrite-N, nitrate-N orthophosphate (PO₄-P) and sulphate (SO₄-S) using an ion chromatograph (Thermo Scientific™ Dionex™ Ion Chromatography, Mettler Toledo). Additional subsamples were also filtered by 0.2 μm and frozen at −20 °C until analysis by EEM spectroscopy.

Further subsamples were filtered through 0.45 μm mixed cellulose ester filters (Whatman, GE Healthcare, UK) and analysed for dissolved COD (COD_diss; 15–150 mg/L test kits, Spectroquant^®, Merck Millipore, Darmstadt, Germany). Finally, unfiltered subsamples were analysed for total COD (COD_total).

Samples were stored in darkness at 3 °C and analysed within the following days after collection. Subsamples for COD_total and COD_diss analyses were conserved until then by adding 1% of sulphuric acid (H₂SO₄; 21.4%, Merck KGaA, Darmstadt, Germany).

Fluorescence EEMs and UV absorbance spectra were obtained at room temperature in a 1 cm quartz cuvette (4 mL volume) using a Horiba Aqualog fluorescence spectrophotometer (Horiba, USA). UV absorbance was measured at 3 nm intervals between 239 and 800 nm. Fluorescence emission was measured at 1 nm increments between 240 and 620 nm, over an excitation range of 240 and 500 nm.

All fluorescence EEM post-processing procedures were carried out using the drEEM toolbox (Murphy et al. 2013) within the Matlab software (The MathWorks Inc.). The EEM correction process consisted of a blank EEM subtraction; Raman and Rayleigh scatter removal, correction for inner filter effects and normalization to Raman units (RU). Inner filter effect correction factors were derived from the absorbance measurements and Raman normalisation factors were calculated using MilliQ blank EEMs at λ_ex = 350 nm (Lawaetz & Stedmon 2009). The resulting corrected and normalised EEM dataset was then subjected to parallel factor analysis (PARAFAC) with the drEEM toolbox (n = 148). The least squares models for 2–7 component models were tested by running 60 iterations, with non-negativity and convergence criteria of 1 × 10⁻⁶. The most appropriate number of PARAFAC components was then found using a combination of split half and residual analysis according to the recommendations in Murphy et al. (2013).

Masses were calculated based on the measured concentration and the remaining water volume in the experimental system/the collected volume of the liquefied foam, respectively.

Statistical differences of the relative shares of fluorescence compounds 1–3 between water and foam samples were assessed by one-way ANOVA after passing the Shapiro–Wilk normality test, followed by Holm–Sidak pairwise comparisons using SigmaPlot version 13.0 (SystatSoftware Inc., CA, USA).

The initial COD_diss of the RAS water was 17.12 mg/L. The concentration of COD_diss in the water of the experimental setup increased initially over time, reaching concentrations of up to 474 mg/L (1.59 gCOD_diss/m³ woodchips) (Figure 2(a)) at Day 7, after which it levelled off towards the end of the trial. Dissolved COD represented 98 ± 4% of the total COD (n = 18) in the recirculating water at the end of the experiment.

The foam fractionation process removed COD_diss at a constant rate of 75 ± 18 g COD_diss/m³ woodchips/day (n = 3), however, the volumes of the liquefied foam collected showed large variations between sampling days and replicated systems. The mean COD_diss concentration in the liquefied foam was 2.9 ± 2.2 (n = 18) times higher than in recirculating water. COD_diss constituted on average 90 ± 3% (n = 18) of the total COD in the liquefied foam. The total amount of COD_diss removed by foam fractionators amounted to 37.8 ± 4.7% (n = 3) of the COD_diss recovered from the water and foam after 11 days.

As shown in Figure 2(e), nitrate concentrations in the water decreased consistently throughout the trial, from 61.11 mg NO₃-N/L at Day 0 to 0.28 ± 0.05 mg NO₃-N/L at Day 11 (n = 3). The initial constant nitrate removal rate was observed to increase after 3 days, then levelled off after 7 days, where nitrate concentrations in the water remained below 10 mg NO₃-N/L.

Total-N declined in a similar manner as nitrate throughout the trial, decreasing from 62.20 mgN/L (Day 0) to 11.16 ± 1.50 mgN/L (Day 11, n = 3).

TAN increased slightly from 0.93 to 1.87 ± 0.82 mgTAN/L during the initial 3 days, where concentrations increased more rapidly at a near linear rate, reaching 9.50 ± 1.31 mgTAN/L (n = 3) by Day 11.

Nitrite concentrations increased from 0.13 to 21.95 ± 0.54 mg NO₂-N/L after 5 days, thereafter decreasing to 1.38 ± 0.14 mg NO₂-N/L (n = 3) at Day 11.

The foam fractionation process did not fractionate nitrogenous compounds. The concentrations of TAN, NO₂-N, NO₃-N and total-N in the liquefied foam were equivalent to those in the recirculated water (89.0 ± 14.4%, 106 ± 28.4%, 108.9 ± 14.3% and 122.7 ± 18.3%, respectively (n = 18).

Orthophosphate concentrations in the water increased from 3.12 mgPO₄-P/L at Day 0 to 23.55 ± 1.94 mgPO₄-P/L (n = 3) at Day 11 of the experiment. The increase in orthophosphate concentrations was characterised by a steep increase during the initial days that appeared to level off after approximately 5 days (Figure 2(f)). Sulphate concentrations in the water remained stable throughout the experiment, averaging 37.67 ± 0.36 mg SO₄-S/L (n = 37) (Figure 2(f)). The foam fractionation process did not influence orthophosphate or sulphate concentrations. There were no significant differences in the concentrations of PO₄-P and SO₄-S between the liquefied foam and the recirculated water.

The results demonstrated that recirculation of woodchip bioreactor leachate through foam fractionators is an effective, novel approach for reducing the amounts of DOM discharged into the environment during bioreactor start-up. The concentration of COD_diss within the water increased rapidly during the first 5 days, after which it levelled off until the end of the study (Figure 2(a)). This may be due to the leaching of COD_diss, which consist of both the instant, short-term release of pre-existing dissolved organic substances in woodchips, as well as COD_diss originating from the degradation of insoluble organic carbon, which then generates further DOC (McLaughlan & Al-Mashaqbeh 2009). A study by Svensson et al. (2014) showed that significant differences in leaching patterns of DOC, phenols, pH, colour, tannins and lignins are found between different tree species. In a pilot-scale woodchip bioreactor operated at a commercial recirculated trout farm and filled with the same type of wood as in the current study, von Ahnen et al. (2016a, 2016b) observed a similar high initial release of COD lasting for about 5 days. The amounts of COD_diss likely levelled off over time due to a biological degradation of COD_diss by the suspended and woodchip associated microbial communities (not quantified in the current study), as well as the constant removal of COD_diss by the foam fractionators (Figure 2(b)).

The COD_diss concentrations were on average 2.9 ± 2.2 times and up to 11.1 times higher in the collected foams than the water. In this laboratory study, the foam fractionators were adjusted to deliver a guaranteed sufficient amount of sample volume per sampling rather than to optimize COD_Diss separation. In contrast, at the field scale, foam fractionators may be adjusted to decrease foam volume and increase COD_Diss concentration in the foam. Similar to the highest concentrated foams of this study, Mills et al. (1996) analysed aquatic foam and stream water samples from two natural streams, observed that the foam had a 10–20 fold higher DOC concentration than the stream water. Furthermore, the authors found that the humic substances in the foam showed increased hydrophobicity, aliphatic character and compositional complexity than the humics within the host stream (Mills et al. 1996).

The collected foam, in particular, contained a higher relative fraction of fluorescence Component 3 compared to the system water (Figures 3 and 4). This component is typically characterised as having high bioavailability and hence is often associated with elevated microbial activity (Hudson et al. 2008). Yao et al. (2020) also detected this protein-like fraction in experimental-scale woodchip bioreactors and found that its relative percentage of the total woodchip leachate decreases over time. Moreover, freshwater RAS have previously been shown to exhibit high levels of this component (Hambly et al. 2015), where it is assumed to originate mainly from the proteins within the fish feed. Degradation studies have demonstrated that this type of fluorescence component represents bioavailable organic matter and, as such, has the potential to boost microbial population growth and the associated consumption of dissolved oxygen (DO) in receiving water bodies (Stedmon & Cory 2014; Logue et al. 2016). Hence, it may play an important role in the balance of heterotrophic microbial communities within the RAS as well as the environmental impact of RAS discharge. Within this study, foam fractionation was shown to selectively remove this bioavailable fraction of DOM. Foam fractionators could therefore be an effective treatment tool in reducing the bioavailability of carbon initially released from denitrifying woodchip bioreactors. Optimisation of the foam fractionation process could also lead to improved DOM removal and achieving higher COD concentrations of the foam at reduced foam volumes. The experimental-scale foam fractionators investigated in this study were relatively short and the air supply was not optimized for DOM removal. Further studies may investigate the effects of foam fractionator size, air/liquid ratio and bubble size on the removal efficiencies of DOM leached from woodchip bioreactors by foam fractionation.

Higher total COD_diss removal by foam fractionation could also likely have been achieved if the process was continued beyond the 11-day experimental period. However, as nitrate had been completely depleted by this time, other transformation processes could potentially take its place. In this situation, if recirculation was continued after nitrate had been depleted, sulphate reduction could then occur, creating toxic hydrogen sulphide and ultimately threaten aquatic life if released to the natural environment. Nitrate depletion may not only lead to sulphate reduction/production of toxic hydrogen sulphide but also to an increased release of COD, which would be counterproductive in this setup. This has been demonstrated in a laboratory-scale woodchip column study by Christianson et al. (2017), who found a correlation between the percentage of COD released and sulphate reduction when NO₃ removal efficiency was above 90%.

As application of the foam fractionation process on woodchip biofilters has been shown to result in an enriched solution of DOM, this could potentially be used as a resource. Both the foam extract and the remaining water in the system could be applied as fertilizer on fields, as humic substances have been shown to provide valuable nutrients and bio-stimulants for plant growth (Nardi et al. 2007; Canellas et al. 2008). Moreover, at fish farms, both the foam extract and remaining water could be directed into the sludge basins. Here, the complex DOM from woodchips could be transformed to simpler organic molecules through anaerobic digestion, which could thereafter be used to fuel heterotrophic denitrification in a subsequent denitrification reactor, constructed wetland or woodchip bioreactor. Field et al. (1988) found that 30–50% of the soluble COD from bark water could be acidified into volatile fatty acids (VFAs) and between 7 and 36% could be converted by methanogenesis to methane after 8–9 days of anaerobic digestion. In the case that near complete removal of the remaining DOM was required, post-treatment by combined ozone and biological treatment may be effective at treating woodchip leachate (Zenaitis et al. 2002; Svennson et al. 2015).

A slight initial increase in TAN concentrations within the water was observed (Figure 2(e)). This may be due to TAN readily leaching from woodchips. From Day 3 until the end of the trial, TAN concentrations consistently increased, which was also accompanied by an accelerated reduction in nitrate concentrations (Figure 2(e)). It is therefore likely that TAN was produced by dissimilatory nitrate reduction to ammonia (DNRA); a microbial process that is favoured over denitrification in excessive reductive environments (i.e., high C/N ratios) (Tiedje 1988; Washbourne et al. 2011), as was the case in the current study. The occurrence of the DNRA process is a downside of such a recirculated system, as the produced TAN may require additional post-treatment before discharge. Further evidence for the hypothesis that TAN production was due to microbial processes rather than leaching from woodchips, is shown by a comparison with the PO₄-P release pattern (Figure 2(f)). In contrast to the measured TAN concentration curve pattern, the PO₄-P showed a typical saturation curve for pure concentration gradient-driven leaching, without significant microbial removal or production.

The low DO concentrations measured in the woodchip bioreactor outlets may indicate that the decline in nitrate concentrations in the water throughout the study period was due to denitrification and DNRA activity. Both the alkalinity produced by denitrification, as well as the CO₂ stripping effect of the foam fractionation may have caused the pH to rise after Day 2, following the initial pH drop, that was likely driven by instant, high releases of organic acids (Tao et al. 2005).

This study demonstrated that the release of DOM from newly installed woodchip bioreactors could be reduced when woodchip bioreactors were recirculated through foam fractionators during the start-up period. This novel approach would remove large fractions of the harmful DOM initially released and even the simplest foam fractionators can achieve meaningful reductions of the DOM, which would otherwise be discharged into the environment. These can be built consisting of a tube and an air blower/diffuser and hence can be constructed at relatively low cost on-site. Beyond the start-up period, foam fractionation may also find a useful application in single-pass configurations. This would help to further polish woodchip bioreactor effluents for DOM with little extra cost, as re-aeration of the water is already necessary in most cases before discharge to a receiving body. Future studies may investigate this method during start-up of full-scale bioreactors and may optimize the design and operation of the foam fractionation process to increase DOM removal from woodchip bioreactors during start-up.

The study was partly funded by the Ministry of Environment and Food of Denmark and by the European Maritime and Fisheries Fund (EMFF) as part of project ‘Miljøvenlig brug af træflisfiltre på dambrug’ (J. no. 33111-I-17-058). We thank Ulla Sproegel and Brian Møller (DTU Aqua) for valuable technical assistance in the laboratory.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.