Comparing the effect of carbon media on nutrient removal and greenhouse gas production in laboratory-scale bioreactors Open Access

Natural overland flow is a process that occurs on the landscape when water cannot be absorbed by the soil. In agricultural landscapes, excess water can pool on the soil surface during times of high moisture such as snowmelt or periods of precipitation combined with high antecedent soil moisture conditions. Crop producers often rely on manual surface or subsurface drainage to remove excess water from the soil in order to improve soil moisture conditions and enhance crop production.

Subsurface drainage was introduced in the United States in the 1820s (Irwin & Spencer 1970). In Canada, subsurface drains were used in Alberta and Saskatchewan beginning in the 1970s to reduce soil salinity (Broughton & Jutras 2013). Since then, subsurface drainage has gained popularity in the Canadian prairies. While drainage provides benefits such as reduced soil salinity and improved soil moisture conditions, it also functions as a conduit for nutrient transport. Soils amended with chemical fertilizers or animal manure can contribute nutrients such as nitrogen (N) and phosphorus (P) to nearby streams and waterbodies through drainage water (Palliser Environmental Services Ltd., Alberta Agriculture and Rural Development 2008). The accumulation of nutrients in downstream waterbodies can cause eutrophication, which has increasingly become a water quality concern not only in North America but around the world (Trimbee & Prepas 1987; Turtola & Paajanen 1995; Paterson et al. 2006; Jaynes et al. 2008; Schindler et al. 2008; Jacquet et al. 2014; Bol et al. 2018).

While much of the research done to address nutrient loading to waterbodies has focused on P and its reduction (Schindler et al. 2016), N is also a major contributor to eutrophication. Nitrogen in the dissolved form of nitrate (NO₃-N) is flushed downward through the soil profile during periods of snowmelt, rainfall or when a crop is irrigated (Palliser Environmental Services Ltd., Alberta Agriculture and Rural Development 2008). Thus, a major concern regarding the use of subsurface drainage is the transport of NO₃-N to downstream waterbodies. N-fertilizer management and conservation methods can reduce NO₃-N transport, but additional methods of NO₃-N removal are needed (Jaynes et al. 2008). Nitrate in subsurface drainage water can be removed by denitrification processes under anaerobic conditions, however the natural occurrence of denitrification can be limited by the availability of an energy source, namely carbon, for denitrifying bacteria. Soils can be modified by adding solid organic carbon sources so that NO₃-N passes through carbon which supports denitrification. However, redevelopment of natural soils can take decades, so other in-field or edge-of-field treatment options for subsurface drainage water are needed for NO₃-N removal (Schipper & Vojvodic-Vukovic 1998; Jaynes et al. 2008).

Denitrifying bioreactors are a passive water treatment approach where water is routed through solid carbon substrates to remove dissolved nutrients. Bioreactors can be used to treat nutrient-laden agricultural drainage water for the protection of downstream water bodies.

The potential for carbon-based bioreactors to enhance nitrate removal via denitrification has been studied previously at the laboratory scale. Healy et al. (2012) found that among the four carbon source materials used within their lab-scale bioreactors, the highest NO₃-N removal efficiencies were for cardboard (∼94%), followed by lodgepole pine needles (∼75%), barley straw (∼74%) and lodgepole pine woodchips (∼70%). Woodchips are the commonly used fill media in bioreactor research. Woodchip bioreactors are typically designed to target removal of nitrate, but some studies have included P removal as part of their testing under the assumption that success with P removal would make bioreactors an attractive option for combined N and P management. These studies, however, have had variable outcomes. Several lab-scale bioreactor studies have used woodchips containing other products (metals and water treatment residuals) and have shown reduction of orthophosphate concentrations by 5–10% (Goodwin 2012; Zoski et al. 2013). Sanchez Bustamente-Bailon et al. (2022) found promising results in their study that used discs filled with woodchips that naturally contained a range of metals known to sorb P and obtained P reduction of up to 87%. Rambags et al. (2016) reported reductions of P up to 14% in a full-scale woodchip bioreactor. Choudhury et al. (2016) reported that woodchip filters used in their study exhibited minimal ability to remove soluble reactive P.

One potential side effect for the application of denitrifying bioreactors is the production of the greenhouse gas (GHG) nitrous oxide (N₂O) (Greenan et al. 2009; Moorman et al. 2010; Davis et al. 2019). N₂O has a high global warming potential; 273 times that of CO₂ for a 100-year timescale (International Panel on Climate Change (IPCC) 2013). Incomplete denitrification within the bioreactors can lead to N₂O production (Elgood et al. 2010; Davis et al. 2019). However, several studies (Greenan et al. 2009; Woli et al. 2010; Feyereisen et al. 2016) have shown that the N₂O emissions account for a small percentage of the observed nitrate removal, suggesting little concern for N₂O release from in-field bioreactor systems.

Further, microbial decomposition of the organic carbon media (not quantified in this study) has the potential to release other GHGs such as methane (CH₄), which is the third largest contributor to global radiative forcing, and carbon dioxide (CO₂) which comprises 72% of global emissions (Woli et al. 2010; Healy et al. 2012; Davis et al. 2019).

As such, the objectives of this study were to assess the NO₃-N and PO₄-P removal efficiencies, and to quantify the dissolved N₂O, CH₄ and CO₂ production in laboratory scale bioreactors using three carbon media: woodchips and two locally available agricultural residues, hemp straw (Cannabis sativa L.) and barley straw (Hordeum vulgare L.). The purpose of this study was to contribute to the literature of quantifying GHGs produced by bioreactors using different carbon media and document whether agricultural residues were more suitable than woodchips for removing nitrate. The outcomes of this study were used to validate and implement in-field pilot-scale bioreactors (Kohn et al. 2023a, 2023b).

Tracer tests were done to evaluate the internal hydraulic properties of the bioreactors needed for subsequent removal calculations. Tracer tests are also useful for diagnosing poor performing systems (Hoover et al. 2017). In this study, the tracer was made from deionized water mixed with potassium bromide (KBr), a commonly used laboratory tracer. Although bromide occurs at low natural concentrations, it is considered to be a suitable tracer for water testing within a laboratory setting (Ghane et al. 2019).

Flow was maintained at an estimated 4 h HRT and approximately 100 mg of Br⁻ was introduced to the inlet port by injecting 3 mL of an approximately 35,000 mg L⁻¹ Br⁻ solution using a syringe and a hypodermic needle. After the tracer injection, bromide concentrations at the outlet were assessed with 22 outlet water samples collected over the following 24 h. The samples were analysed in the laboratory for Br⁻ using a Thermo Dionex ICS-5000 Ion Chromatograph (Thermo Fisher Scientific, Waltham, MA, United States).

Hydraulic efficiency values range from 0 to 1.0. According to Hoover et al. (2017), a hydraulic efficiency greater than 0.75 is considered good, while an efficiency below 0.5 is poor. The Morrill Dispersion Index (MDI) is an indicator of dispersion and mixing of the tracer throughout the bioreactor, where 1 represents ideal plug flow (i.e., no mixing or back flow), and values from 1.0 to 2.0 indicate effective plug flow (Metcalf & Eddy 2003). A short-circuiting index (S) value nearer to zero indicates short circuiting is occurring and all water is not effectively being treated and an S value closer to 1.0 indicates more ideal flow.

Approximately 10 days after the tracer tests were completed, tests to determine NO₃-N and PO₄-P removal efficiency were initiated. Simulated drainage water was produced by mixing municipal water with KNO₃ and KH₂PO₄ in a 6,000 L cone-bottom tank to prepare a solution with a NO₃-N concentration of 30 mg L⁻¹ and PO₄-P concentration of 1 mg L⁻¹. A constant head pressure gravity feed system was used to allow the treated water to flow continuously through the bioreactors for 16 weeks at a 4 h HRT. A pump was used to ensure the level of water in the gravity feed tank was maintained. Once a week, water samples were collected at each bioreactor outlet by placing a 125 mL sample bottle in the stream of water flowing between it and the drainage pipe (Figure 2); the bottom (#1), middle (#2), and top (#3) ports were sampled by momentarily removing the cap on the port and allowing the water to flow into a sample bottle. As sampling from these three ports was disruptive to the system flow, the outlet (port #4) was sampled first, followed by the top port then middle, and finally the bottom port. Inlet samples were taken at the nine-port manifold just before the needle flow meters. Samples were filtered through 0.45 μm cellulose membrane and analyzed for NO₃-N and PO₄-P by segmented flow analysis on Astoria Pacific Autoanalyzer2 (Astoria Pacific, Inc., Clackamas, OR, United States).

Time series of NO₃-N and PO₄-P concentrations at different locations along the length of the bioreactors were calculated from samples taken from the bottom, middle and top and outlet ports along the side of the bioreactors.

A key determinant for nutrient removal is the availability of carbon. As such, total carbon (TC) mass was measured in the initial and final carbon materials to quantify the TC losses of each material.

A portion of the water samples collected each week over the experimental period were used to determine dissolved concentrations of CO₂, CH₄ and N₂O. A 30 mL subsample was drawn directly from the 125 mL sample bottle into a 60 mL closed syringe to prevent sample contact with air. Subsequently, 30 mL of helium was drawn into the syringe. Syringes containing the helium and water were placed on a reciprocal shaker and mixed at room temperature for 5 min at 180 rpm. Following mixing, the first 10 mL of air sample was expelled to ensure there were no water drops in the connector and needle before filling two replicate 5.9 mL evacuated exetainers (Labco) each with 10 mL. Gas samples were then analyzed for N₂O, CO₂ and CH₄ using a gas chromatograph (Varian 3800, Varian Instrument, Walnut Creek, CA) equipped with an electron capture detector (ECD), flame ionization detector (FID), and thermal conductivity detector (TCD).

Dissolved GHG production rates (g CO₂-N m⁻³ d⁻¹, CH₄-N m⁻³ d⁻¹, or N₂O-N m⁻³ d⁻¹) were calculated with the same method using Equation (7). Cumulative N₂O production (g N m⁻³) was determined by trapezoidal integration of N₂O production rate versus time as described in Feyereisen et al. (2016). The cumulative N₂O production relative to the cumulative NO₃-N removal was calculated over the entire 16-week experimental period and expressed as a percentage (%).

Statistical analyses were performed using Microsoft^® Excel^® (Microsoft Corporation 2023) and SigmaPlot (2011). Specifically, Mann-Whitney Rank Sum Tests were used to determine significant differences (P < 0.05) in hydraulic characteristics, nutrient removal efficiencies and rates among the three different fill carbon media. These tests were also used to determine significant differences (P < 0.05) in the GHG production among the different carbon media for each GHG measured. Boxplots and scatterplots were created using the ‘ggplot2’ package (Wickham et al. 2023) in RStudio Team (2020) to provide visual displays of the nutrient concentrations.

Hydraulic efficiency was over 0.5 in all bioreactors and was highest for woodchips. There was a significant difference (P < 0.05) between the median hydraulic efficiency values for bioreactors with woodchips and barley straw fill media.

An MDI less than 2.0 is indicative of effective plug flow and most of the values reported in this study were between 2.7 and 3.9, indicating plug flow characteristics with low dispersion (Hoover et al. 2017). One of the bioreactors filled with hemp straw showed the highest dispersion rate (3.89) and one of the bioreactors filled with woodchips showed the lowest dispersion rate (2.66). As seen in Figure 4, bioreactor I, a woodchip bioreactor, showed a delayed tracer response compared with all other bioreactors. This may have been caused by a higher proportion of small wood fragments which affected flow dynamics within this bioreactor.

S values for all bioreactors were 0.58 or higher, indicating minimal short circuiting. Hemp straw showed lower S values and higher MDI values than the barley straw and woodchips, but there were no significant differences in the median MDI and S values among the three fill media (P < 0.05). Overall, the agricultural residues (hemp straw and barley straw) showed acceptable hydraulic properties and thus were determined to be suitable as carbon media for the bioreactors.

The NO₃-N inlet concentrations ranged from 27 to 30 mg L⁻¹. Flow rates were constant at each bioreactor ranging from 63.7 to 70.5 mL min⁻¹ which achieved the desired HRT of 4 h.

Removal efficiencies at the outlet port varied from 0.5 to 36.9% for the three carbon media. After 16 weeks of operation, mean NO₃-N removal efficiencies were higher for barley straw (20.1 ± 6.1%) than hemp straw (13.8 ± 5.2%) and woodchips (6.8 ± 0.6%). The differences in the median removal efficiencies for NO₃-N among the three carbon media types were statistically significant (P < 0.05) with a performance rank of barley straw > hemp straw > woodchips. This is probably because agricultural straw media contains more soluble carbon than woodchips which can provide more carbon to promote the growth of denitrifying bacteria in the straw bioreactors.

In our study, nitrate removal efficiencies peaked at different times for each carbon media and declined afterwards. This may be because different available carbon compositions (i.e., different cellulose, hemicellulose and lignin content) in three carbon media resulted in the denitrifying bacteria developing and predominating at different periods of time. However, quantification of these bacteria was out of scope of this study.

The nitrate removal efficiency in woodchip bioreactors peaked during the second week of operation (up to 14.8%) at the outlet sampling port (Figure 5, panel 4) and remained below 10% after the fifth week of operation at all sampling ports. For barley straw, removal efficiency peaked at 40.6% during the fifth week at the top sampling port (#3) and remained above 10% at the outlet sampling port (#4) after the second week of operation (Figure 5, panel 3). For hemp straw, the removal efficiency peaked during the ninth week of operation, up to 25.7% (Figure 5, panels 3 and 4).

Nitrate removal rates differed significantly (P < 0.05) between carbon media types and declined after the first 8 weeks for all carbon media. They were consistently higher for barley straw than the other two media over the 16-week experiment with values up to 64 g N m⁻³ d⁻¹. Nitrate removal rates for hemp straw showed values up to 42 g N m⁻³ d⁻¹ while woodchips showed values no higher than 23 g N m⁻³ d⁻¹. These results show that agricultural residues are attractive alternatives as carbon media for denitrifying bioreactors. Moreover, under the conditions of this experiment and time frame of study, barley straw is the most effective at removing nitrate.

The results from this study are consistent with findings presented in the literature, which indicate lower performance for woodchip bioreactors compared to other carbon media. For example, Cameron & Schipper (2010) conducted a study over 23 months and reported nitrate removal rates ranging from 0 to 52 g N m⁻³ d⁻¹ depending on carbon media and ranked their performance in the order of maize cobs > green waste > wheat straw > wood for treatments at 14 and 23.5 °C. In a laboratory study by Li et al. (2018), three hydraulic retention times ranging from 0.67 to 4.0 h were targeted, which resulted in total nitrate removal rates for woodchips between 40 and 49 g N m⁻³ d⁻¹, and 49–85% of the nitrate concentration being removed. Zoski et al. (2013) reported nitrate removal rates ranging from 0.11 to 0.29 g N m⁻³ d⁻¹ for lab-scale woodchip bioreactors, which equate to removal efficiencies ranging from 8 to 60%. The authors suggest that the geometry of the woodchips could have created preferential flow conditions (i.e., carved out flow paths) within the bioreactors which resulted in fast flow through the chambers and consequently less opportunity for nitrate removal.

The calculated TC losses in water from the carbon materials were 5.7 ± 1.4, 27.1 ± 2.9 and 35.6 ± 1.1% for woodchips, hemp straw and barley straw, respectively. These results show that the agricultural residues have more carbon depletion than the woodchips, meaning that the agricultural residues would need to be replenished more often due to the higher depletion rate. These results are consistent with the findings reported by Schipper et al. (2010).

Orthophosphate removal efficiencies in this study were lower than some of the values reported in the literature. Husk et al. (2018) found that a mixed-media bioreactor showed approximately 19 times more P removal than three woodchips-only bioreactors, however this study was also performed as a full-scale field experiment in the cold, humid climate of Quebec, Canada, during a three-year period. The soluble reactive P fraction concentration was reduced by 53% in the woodchip bioreactors, and by 97% in the mixed-media (activated alumina and gravel) bioreactor. Bock et al. (2015) performed a laboratory scale experiment with nine bioreactors – one woodchip control, and eight different woodchip-biochar treatments – and reported an average P removal of 65% by 18 h from the biochar treatments. Comparatively, Li et al. (2018) used a laboratory bioreactor design in which a 3-m long horizontal column was filled with woodchips and a 1-m long column was filled with fly ash pellets. The tests were performed for 28 h. The authors reported an average phosphate removal of 71.4% from the fly ash pellet column and an orthophosphate removal ranging from 10 to 13% from the woodchips column, despite their primary purpose being for nitrate removal. Further, Allred (2010) tested six different media fill materials to remove orthophosphate within lab-scale bioreactors and found that a high calcium oxide-high carbon fly ash material and surfactant-modified zeolite in combination had the greatest removal of PO₄-P at 50% or higher.

Orthophosphate removal rates differed slightly between the three carbon media with significant differences between both agricultural residues and woodchips (P < 0.05) and no significant difference between barley straw and hemp straw (P < 0.05). Removal rates were consistently higher for barley straw, with values up to 3.0 g P m⁻³ d⁻¹. Orthophosphate removal rates for hemp straw had values up to 2.3 g P m⁻³ d⁻¹ while removal rates for woodchips had values no higher than 2.7 g P m⁻³ d⁻¹. Phosphate removal rates declined rapidly after the first week for the three carbon media making it difficult to recommend any for phosphate removal.

Bioreactors were a source of GHG production across the three different carbon media. In our study, woodchips generally produced the lowest GHG among the three media types (Table 2) which is consistent with the results from Bock et al. (2018) for laboratory scale bioreactors. The authors used three forms of woodchips as bioreactor fill media; woodchips, woodchips amended with 10% biochar – used as an agricultural soil amendment – and woodchips amended with 30% biochar. They found that biochar addition, particularly at 30% rate, was found to increase N₂O and CO₂ emissions.

Healy et al. (2012) reported high initial levels of CO₂ and a subsequent shift to CH₄ production under steady-state conditions in laboratory scale denitrification bioreactors using different media: lodgepole pine woodchips, cardboard, lodgepole pine needles and barley straw. They recorded the highest emissions for the cardboard bioreactors, followed by barley straw and then by lodgepole pine woodchips.

After 16 weeks of bioreactor operation, the mean cumulative N₂O production in woodchip, hemp straw and barely straw bioreactors was 1.79, 8.03 and 24.16 g N m⁻³, respectively (Table 3). Cumulative N₂O production was higher for barley straw than woodchips, although there was no significant difference (P < 0.05). After 16 weeks, the percentages of NO₃-N converted to N₂O while passing through the bioreactors were 0.16% for woodchips, 0.35% for hemp straw and 0.75% for barley straw (Table 3), indicating that N₂O production was negligible.

Our results are consistent with the findings reported by Feyereisen et al. (2016) who found that cumulative N₂O production in lab-scale bioreactors filled with barley straw (56 g N m⁻³) remained higher than those filled with woodchips (15.3 g N m⁻³) over five months at a 12 h HRT. One difference, however, is that lower N₂O production rates and percentage of NO₃-N converted to N₂O were obtained in our study. This may have been because we used a shorter operation time and shorter HRTs which would not have allowed for as much opportunity for denitrification in comparison to the five months used in Feyereisen et al. (2016).

As mentioned, less than 1% of the NO₃-N removed was in the form of N₂O (0.16–0.75%). This is comparable with the values reported for woodchips by David et al. (2016) (0.44–0.89%) and by Christianson et al. (2013a) (<0.32%). Our values were much lower, however, than the values reported by Feyereisen et al. (2016) (9.7%). Woli et al. (2010) also reported negligible N₂O fluxes, suggesting that the denitrification process resulted in the production of N₂, rather than becoming N₂O and indicating little concern for N₂O release from their bioreactors. Greenan et al. (2009) also found in their laboratory study using columnar bioreactors with woodchips, that the N₂O released ranged from 0.003 to 0.033% relative to NO₃-N removal, indicating that near-complete denitrification occurred.

This study evaluated the performance of three carbon media including two agricultural residues – barley straw and hemp straw – and woodchips for removing NO₃-N and PO₄-P from simulated subsurface drainage waters. Bioreactor replication allowed for the estimation of the variability of the results. Findings from the two agricultural residues show adequate hydraulic properties for denitrification and thus were suited for field-scale bioreactor applications. Removal efficiencies varied from 0.5 to 36.9% for NO₃-N and from 1.1 to 53.5% for PO₄-P, with greatest to poorest performance shown by barley straw > hemp straw > woodchips, and the durability of carbon materials was in the order of woodchips > hemp straw > barley straw. Earlier peak time of NO₃-N removal occurred in woodchips than straw bioreactors, suggesting agricultural straw materials might need more flushing time after bioreactor installation to establish microbial populations for acceptable NO₃-N removal. This study shows that barley straw could remove twice as much nitrate as woodchips. PO₄-P removal efficiencies and rates declined rapidly after the first week for the three carbon media meaning none were superior. This could be a co-benefit for reduction of PO₄-P concentrations in tile drainage water considering the relatively low P concentrations as compared to N.

In this study, GHG production rates were dominated by CO₂, with less CH₄ and N₂O produced. Although the woodchip bioreactors produced the smallest quantities of N₂O, the respective production related to NO₃-N removal in all nine bioreactors were very low at 0.16–0.75%, so there was little concern of N₂O release from these bioreactors.

We would like to thank the staff of Alberta Agriculture and Irrigation for their contributions to this study. We also would like to thank Dr Xiying Hao and Jessica Stoeckli of Agriculture and Agri-food Canada for their roles in conducting GHG measurements.

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

The authors declare there is no conflict.