Jukuri, open repository of the Natural Resources Institute Finland (Luke) All material supplied via Jukuri is protected by copyright and other intellectual property rights. Duplication or sale, in electronic or print form, of any part of the repository collections is prohibited. Making electronic or print copies of the material is permitted only for your own personal use or for educational purposes. For other purposes, this article may be used in accordance with the publisher’s terms. There may be differences between this version and the publisher’s version. You are advised to cite the publisher’s version. This is an electronic reprint of the original article. This reprint may differ from the original in pagination and typographic detail. Author(s): Jani T. Pulkkinen, Anna-Kaisa Ronkanen, Antti Pasanen, Sepideh Kiani, Tapio Kiuru, Juha Koskela, Petra Lindholm-Lehto, Antti-Jussi Lindroos, Muhammad Muniruzzaman, Lauri Solismaa, Björn Klöve & Jouni Vielma Title: Start-up of a “zero-discharge” recirculating aquaculture system using woodchip denitrification, constructed wetland, and sand infiltration Year: 2021 Version: Published version Copyright: The Author(s) 2021 Rights: CC BY-NC-ND 4.0 Rights url: http://creativecommons.org/licenses/by-nc-nd/4.0/ Please cite the original version: Pulkkinen J.T., Ronkanen A.-K., Pasanen A., Kiani S., Kiuru T., Koskela J., Lindholm-Lehto P., Lindroos A.-J., Muniruzzaman M., Solismaa L., Klöve B., Vielma J. (2021). Start-up of a “zero- discharge” recirculating aquaculture system using woodchip denitrification, constructed wetland, and sand infiltration. Aquacultural Engineering 93, 102161. https://doi.org/10.1016/j.aquaeng.2021.102161 Aquacultural Engineering 93 (2021) 102161 Available online 2 April 2021 0144-8609/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Start-up of a “zero-discharge” recirculating aquaculture system using woodchip denitrification, constructed wetland, and sand infiltration Jani T. Pulkkinen a,*, Anna-Kaisa Ronkanen b, Antti Pasanen c, Sepideh Kiani b, Tapio Kiuru a, Juha Koskela a, Petra Lindholm-Lehto a, Antti-Jussi Lindroos d, Muhammad Muniruzzaman c, Lauri Solismaa c, Bjorn Klove b, Jouni Vielma a a Natural Resources Institute Finland, Survontie 9A, 40500 Jyvaskyla, Finland b Water, Energy and Environmental Engineering Research Unit, Faculty of Technology, 90014, University of Oulu, Finland c Geological Survey of Finland, Neulaniementie 5, 70211 Kuopio, Finland d Natural Resources Institute Finland, Latokartanonkaari 9, Helsinki, Finland A R T I C L E I N F O Keywords: Aquaculture Nutrient load Recirculating aquaculture system (RAS) Research facility Wastewater treatment A B S T R A C T Recirculating aquaculture systems (RAS) discharge management limits the development of the aquaculture sector, because RAS do not automatically result in low nutrient emissions. Research has helped develop discharge management systems such as wetlands and woodchip bioreactors that have been adopted by Danish commercial model trout farms. To further develop the Danish concept, we have modelled and built a novel “zero-discharge” recirculating aquaculture system with an annual capacity of approximately 14 tonnes. The aim of this paper is to describe the entire concept and present the results from the start-up phase of the whole system. The concept includes the treatment of RAS effluent (overflow and sludge supernatant) using a hybrid solution of a woodchip bioreactor, constructed vertical wetland, and sand infiltration. Using this three-step process, the nitrate, phos- phorus, and organic matter effluent are decreased to acceptable levels to reuse the water in the RAS process reducing the need for new raw water. In the first nine months of operation, a water treatment field was used as an end-of-pipe treatment to ensure the water was safe to recirculate for fish. During the winter, the water tem- perature dropped to 2.7 degrees in the sand filter, but the frost did not reach the water levels in any of the treatment processes. It therefore appears that a hybrid solution can operate sufficiently even in winter conditions. In the first year of operation, a woodchip bioreactor can remove 97 % of the nitrate, although the slow start-up of the RAS caused the bioreactor to be N-limited. On average, 79 % and 92 % of the inflow phosphate concentration was removed in the woodchip bioreactor and the entire hybrid treatment field respectively. The wetland and sand filter removed organic matter sufficiently (35 %), but because of the longer than designed actual water residence, it leached from the bioreactor more than was expected. Further experimentation is needed to identify the financial applicability and performance during higher feeding rates. 1. Introduction Most of the aquaculture production in the Nordic countries takes place in sea cages and on flow-through farms, where it has not been economically feasible to treat phosphorus, nitrogen, and organic matter discharges (Silvenius et al., 2017; Pelletier et al., 2009; Gronroos et al., 2006). Land-based recirculating aquaculture systems (RAS) have been developed, in which a lower rate of water usage enables treatment of effluents and a lower fish farming environmental load (Piedrahita, 2003). However, designing well-functioning but cost-effective and low-maintenance treatment units is challenging for low-carbon waste- waters such as RAS effluents. This means that in many cases, the net removal of nutrients and organic matter has not been achieved as effectively as desired (e.g. Dalsgaard et al., 2013). High-technology RAS have been investigated and constructed as a solution, to which more water treatment steps have been added to achieve near-zero water ex- change rates and diminish environmental impacts (Krom et al., 2014; Tal et al., 2009), but they have yet to be widely adopted commercially. In recent years, low-technology solutions such as woodchip denitrifying bioreactors (WB) and constructed wetlands have been applied * Corresponding author. E-mail address: jani.t.pulkkinen@luke.fi (J.T. Pulkkinen). Contents lists available at ScienceDirect Aquacultural Engineering journal homepage: www.elsevier.com/locate/aque https://doi.org/10.1016/j.aquaeng.2021.102161 Received 5 November 2020; Received in revised form 22 March 2021; Accepted 27 March 2021 Aquacultural Engineering 93 (2021) 102161 2 successfully to the aquaculture industry, especially in treating the outflow effluent of RAS farms in Denmark (Dalsgaard et al., 2018; von Ahnen et al., 2018v). Since cold climatic conditions challenge outdoor wastewater treatment, more studies and pilot testing are needed to obtain design parameters for cold climate regions. Woodchip bioreactors have been widely used for nitrogen (N) removal in agricultural (Christianson et al., 2012; Greenan et al., 2009), forestry (Homyak et al., 2008), mining (Nordstom and Herbert, 2018), and aquaculture effluents (von Ahnen et al., 2018v). Nitrate (NO3 ) removal rates in WBs are generally low, being around 5–25 g NO3 m-3 d-1 (Lepine et al., 2020; von Ahnen et al., 2018v), compared to external carbon-fed bioreactors, where much higher nitrate removal rates (1.2–2.8 kg NO3 m-3 d-1) are reported (Letelier-Gordo et al., 2020; Dupla et al., 2006). However, the total cost of N removed in WBs has been estimated to be $2.4–15.2 per kg N (Lepine et al., 2018; Christianson et al., 2012), which is similar to the cost of substrate alone in external carbon-fed denitrifying bioreactors, $1.3…–14.4 kg-1 N (Wang and Chu, 2016; Gutierrez-Wing et al., 2012). In addition, the use of WBs does not require potentially toxic chemicals, indoor facilities, or electricity. Because of their low cost and maintenance, woodchip bioreactors are an excellent choice for passive water treatment for the aquaculture sector. To control carbon leaching and low dissolved oxygen concentration in the outflow of the WB, further treatment steps are needed to recir- culate the water back to the fish culture tanks. A properly designed constructed wetland (CW) as one of the natural-based solutions can serve efficient physical, chemical, and biological nutrient and carbon removal processes, including the aeration of the treated water (e.g. Kadlec and Wallace, 2009; Vymazal, 2007). CWs have also generally been reported to be suitable in cold climatic conditions (Postila and Heiderscheidt, 2020; Heikkinen et al., 2018). Among the many different types of CW, the vertical flow CW remains one of the most effective at aerating the treated water (e.g. Cooper, 1999), especially using the step feeding method (Wang et al., 2020). The managed artificial recharge of groundwater (MAR) is a method in which water is infiltrated through a layer of sand and gravel. During the process, dissolved and particulate organic matter is removed from the water (Kolehmainen et al., 2009; Lindroos et al., 2002). The process is widely used for drinking water treatment in the Nordic countries, and it has also been shown to remove organic matter from recirculating aquaculture effluents (Lindholm-Lehto et al., 2020; Lindroos et al., 2020). To test these low-technology/passive water treatment methods for aquaculture in a cold climate, we have designed and constructed an RAS farm capable of feeding 14 tonnes per year, accompanied by an external hybrid water treatment field that includes a WB, a CW, and a sand infiltration unit (SF). By using this three-step approach, we aimed to remove acceptable levels of effluent nitrate, phosphorus, and organic matter to re-use the water in the RAS process and reduce the need for fresh intake water. When optimising the water treatment field, we hypothesised that sludge (total solids content approx. 1%) was the only outflow discharged from the system, which might further be recycled for other applications such as agricultural fertiliser. This paper presents the construction details and start-up phase of the whole system, and the first year of operation of the water treatment field. To avoid any damages, the start-up phase of the system was carried out carefully, and the full capacity of the water treatment field was not used. 2. Materials & methods 2.1. Recirculating aquaculture system The RAS (FREA Aquaculture solutions, Denmark) was built in the research hall at the Laukaa fish farm of Natural Resources Institute Finland (Luke), where the climate is classified as Dfc (a snowy climate characterised by cold moist winters) in the Koppen climate classification system (Chen and Chen, 2013). Based on the 1981–2010 norm period, the region’s mean annual temperature is 4 C, and the annual precipi- tation is 600 mm (Finnish Meteorological Institute, 2020). The typical snow cover duration is 147 days, whereas the length of the growing season is 165–175 days. Regularly occurring ground frost challenges year-round water treatment in outdoor conditions. The RAS consists of two separate identical units (Fig. 1) that can be operated individually with or without an outdoor hybrid water treat- ment field (Fig. 2). Units can also be used as a joint system in which water circulates through a common pump sump, providing equal water quality for both units. Make-up water is taken from the oligotrophic Lake Peurunkajarvi (62.44886, 25.85201) using a peristaltic pump (Watson Marlow 630, Spirax-Sarco Engineering, UK). A similar type of pump is used to pump the RAS water into the hybrid water treatment field. From the water treatment field, the water is pumped into the research hall to an aerated reservoir tank with a volume of 3.2 m3, from which the water can be further shared with the individual RAS units. One RAS unit has two aluminium 5 m3 raceway fish tanks, which include a 1 m3 space for sludge cones collecting uneaten feed and settleable solids. After the fish tank, the water flows through a 60 μm mesh size drum filter (Hydrotech HDF800, Veolia, France), followed by two parallel 2.5 m3 fixed bed bioreactors (each filled with 1.5 m3 Saddle- Chips, KSK Aqua, Denmark), a 2.24 m2 degassing unit, and a 0.74 m3 pump sump. From the pump sump, the water is pumped (Flygt 3085, Xylem, USA) through a low-head oxygenator (FREA Aquaculture Solu- tions, Denmark) to the fish tanks. The total volume of one unit is approximately 25 m3. The waterflow and degassing rates can be adjusted with frequency converters (Vacon 100 Flow, Grundfos, Denmark). An emergency oxygen diffuser located at the bottom of each fish tank is controlled by a monitoring system (Atlantic, OxyGuard, Denmark). A storing video surveillance system is installed to monitor the system’s operation. Energy consumption is measured by circulating pumps, aeration pumps, and drum filters using an energy meter (DSZ12E, Eltako, Germany). The system’s hydraulic head is 1.4 m. The RAS water temperature is adjusted by controlling the hall air tempera- ture, and pH is adjusted by dosing a dissolved sodium bicarbonate mixture to the pump sumps (EJ-R, Iwaki, Japan). Fish are fed with a commercial feeding system (T Drum 2000, Arvo-Tec, Finland). Rainbow trout (Oncorhynchus mykiss) and European whitefish (Coregonus lavar- etus) were reared in the system during the first year. 2.2. Sludge management Sludge from the sludge cones (manually operated valve, controlled Fig. 1. Raceway-type recirculating aquaculture system containing fish tanks (1), sludge cones (2), drum filters (3), fixed bed bioreactors (4), degassing units (5), pump sumps (6), low-head oxygenators (7), and a mixing tank (8). Water from the treatment field and fresh make-up water is added between the drum filter (3) and the bioreactor (4). Graph by FREA Aquaculture Solutions. J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 3 once per working day), drum filters (water level-based backwash), and the backwash of the four parallel fixed bed bioreactors (backwash approx. 3-week interval) are directed to a storage tank in a sludge container (Supplementary Fig. 1, Clewer Technology Oy, Finland). Sludge treatment is operated automatically in batch mode. The treat- ment process starts by pumping water from the storage tank to the coagulation tank. During the pumping, a coagulant agent is dosed. A flocculation chemical may be added, and the pH adjusted in the coag- ulation tank. Sludge is separated by sedimentation and pumped to an outside peat bed or municipal water treatment plant, and the superna- tant to the water treatment field. Pumping volumes are controlled by pressure probes located in both sludge tanks. At the beginning of April 2020, sludge supernatant was directed to the water treatment field. Polyaluminium chloride (PAX-XL100, Kemira, Finland) is currently used as a coagulant. 2.3. Passive water treatment outdoor solution In the first nine months of operation, the water treatment field was used as an end-of-pipe treatment to ensure the water was safe to recir- culate for fishes. Gradually, the water-flow to the field was increased, and after nine months, some of the water was fed to the RAS (Table 1). 2.3.1. Woodchip bioreactor A horizontal subsurface flow WB was constructed as the first unit in the hybrid passive treatment, followed by a vertical flow constructed wetland (CW) and a sand filter (SF) (Fig. 2, Supplementary Fig. 2). A trapezoidal WB with total dimensions of 14 (length)  9 (width)  1.5 m (depth) was designed based on the results of pre-testing in a column experiment at the Laukaa RAS research platform. Based on the pre-tests, a maximum nitrate removal rate of 33 g NO3 - N m-3 d-1 was achieved at a hydraulic retention time (HRT) of approximately 48 h (Kiani et al., 2020), which was further used to size and design the pilot-scale biore- actor unit as part of the hybrid solution. Inlet water was loaded from one side of the WB by placing two inlet perforated pipes (Ø 5 mm) with a length of 3.3 m and 4.6 m at the base, and 75 cm above the base. The maximum capacity of the bioreactor was designed to provide a 50 m3 saturated woodchip layer corresponding to a 1 m depth of active media and 0.5 m of dry woodchips on top. The purpose of the two-level inlet structures and the dry woodchip layer on top was to provide enough dry layer of woodchips to act as an isolator and avoid deep frost during the harsh Arctic winter. The outlet structure mainly regulates the water level in the woodchip bioreactor. However, the location of the inlet pipe has a significant effect on the flow path and HRT, which improves removal efficiency (Suliman et al., 2006). In winter, when more insu- lation is required, the use of a lower inlet pipe provides a deeper layer of insulation in the inlet zone along with a low water level in the outlet structure. Placing the inlet pipe at the bottom of the woodchips provides a longer flow path (higher HRT), which enhances the effect of low temperature on low removal efficiency (Suliman et al., 2006). In addi- tion, in the summer, when the maximum capacity of the WB can be reached, two inlet pipes are used to distribute a high flowrate through the pore media by providing a longer HRT (Suliman et al., 2006). Setting the two inlet pipes at different heights forces most of the flow through the filter bed using the largest inlet zone. In its first year of operation, the volume of the WB used was 26.3 m3. This means that the average HRT was 5.0 days with a range of 2.3–8.5. Because of the slow start-up phase, this was longer than aimed for at full capacity (about 2 days). The treated water was collected with a perforated pipe (diameter 110 mm, length 3.3 m) at the end of the bioreactor and discharged into an adjustable outlet well, which controlled the water level inside the WB by adjusting the height of the PVC outlet pipe (Supplementary Fig. 3). A controlled outlet structure was designed to cope with the variation in the inflow rate, load, and desired HRT as important design criteria in the system. It was hypothesised that alkalinity increased during the deni- trification process, further improving RAS performance and reducing pH Fig. 2. Schematic diagram of the water treatment field, which includes a woodchip bioreactor, vertical flow constructed wetland, and sand infiltration unit. The wetland has an emergency by-pass line in case of freezing. Pipelines are marked with blue lines, frost measurement points with crosses (x), and water sampling locations with red circles. 1 ˆ Inflow to the water treatment system, 2 ˆ bioreactor outflow, 3 ˆ wetland outflow, 4 ˆ sand infiltration outflow. Table 1 Average monthly water-flows to the RAS and the water treatment field. Operating month Average RAS feeding rate (kg d 1) Fresh water flow to the RAS (m3 d 1) Water flow to the treatment field (m3 d 1) Treatment field water flow to the RAS (m3 d 1) July ’19 10.8 Lake 40.0 August ’19 1.5 7.3 Lake 40.0 September ’19 1.3 4.0 RAS 3.5 October ’19 3.8 3.8 RAS 4.3 November ’19 6.0 6.0 RAS 3.0 December ’19 6.4 6.4 RAS 3.0 January ’20 6.8 6.8 RAS 3.5 February ’20 5.6 5.6 RAS 3.5 March ’20 6.6 5.6 RAS 3.7 1.0 April ’20 7.4 3.6 RAS 7.0 3.8 May ’20 5.6 6.7 RAS 5.1 1.2 June ’20 8.3 4.7 RAS 5.6 3.1 J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 4 control costs. 2.3.2. Constructed wetland The constructed wetland was designed to remove reactive P and N, as well as to control biological oxygen demand (BOD) and the dissolved organic carbon (DOC) that can be leached from the WB. Furthermore, the purpose was to aerate water, because the water exiting the WB had a low dissolved oxygen concentration (DO). Based on the literature, the common reed (Phragmites australis) is one of the most commonly used plant species in different types of CW and is found to be efficient at removing nutrients (e.g. Tanner, 1996; Vymazal, 2011). In a gravel-based vertical flow, the common reed CW, oxygen-rich rhizo- sphere, and aerobic microbial degradation processes can also enhance BOD removal. It was hypothesized that a low BOD concentration in the outflow of the CW unit would improve the last SF unit, in which clogging and metal leaching at a very low DO level would be a problem. The CW with a length/width ratio of 1.25 (7.5m  6m) contained three gravel layers, with different particle sizes (Fig. 3) placed on the bentonite mat liner. The water was distributed to the CW using 50-mm perforated pipes located in the root zone of plants, and it was collected using perforated pipes with a 110-mm diameter in the drainage layer (20 cm) from the bases of the unit. The CW was designed using parameters selected from the literature: the BOD removal rate was 7 g BOD7 m 2 d- 1, and the hydraulic loading rate of 278 mm d-1 and water level were adjustable due to the outlet structure. In the first year of operation, the CW water-saturated volume was 41.3 m3 (average HRT 4.9 days, range 2.2–8.3 days). A dry mulch layer (20 cm) was used for soil-frost pro- tection and placed only in the winter as part of the annual maintenance procedure required in cold climatic conditions. 2.3.3. Sand infiltration Sand infiltration was designed to operate as slow sand filtration. The objective of this step was to achieve an infiltration rate of water corre- sponding to those used in Finnish artificial groundwater plants (ARG). In ARG plants, soil infiltration through gravel- and sand-containing sedi- ments is used to remove dissolved organic carbon and suspended solids (Lindroos et al., 2020). The target water velocity was set at 1 4 m d 1, and the retention time in the SF to ca. 10 d, when the passive water treatment field operated at an infiltration rate of 12.5 m3 d 1. Based on RAS-related pre-studies (Lindroos et al., 2020) and the results from the ARG plants (Lindroos et al., 2002), a retention time of at least 10 days would mean a significant TOC reduction. The SF was designed to operate as a saturated water-flow, because this was assumed to result in a stronger DOM reduction than in an unsaturated vertical water-flow (Lindroos et al., 2020, 2002). Based on the retention time criteria, the optimal dimensions of the sand infiltration unit were determined by adopting a fluid flow and solute transport modelling approach (e.g. Muniruzzaman and Rolle, 2015, 2016; 2017; 2019). The latter was performed by considering the SF a water-saturated porous medium, and the transport of a conservative tracer was simulated to identify the optimal design that allowed a minimum residence time of 10 days (e.g. Rolle et al., 2013, 2018; Muniruzzaman et al., 2014, 2020; Sprocati et al., 2019). The sand material for the SF was collected from the nearby esker area, which is composed of glaciofluvial sand and gravel sediments. The sand was relatively well sorted and classified as coarse sand (0.6–2.0 Fig. 3. Cross-section of the constructed wetland. Mulch is only spread during the winter. The substrate includes sand and 10 % peat. J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 5 mm). The mean effective porosity (ne) was determined to be 0.35 (0.32–0.37). The sand material was composed mainly of quartz and feldspars, and the proportion of dark mafic minerals was low. The dimensions of the SF were: width 16 m, length 31 m, and depth 2 m. The total volume was 650 m3. During the starting period, the SF was operated at 165 m3 (average HRT 17 days, range 7.8–28.9). All the water treatment field unit bottoms were lined with a bentonite mat, to prevent water leaching out of the system, and to enable accurate water balance estimation and purification for research purposes. The average HRT for the whole water treatment field was 26.8 days (ranging from 12.3–45.7 days). 2.4. Water quality analyses Carbon dioxide (Franatech, Germany), dissolved oxygen concentra- tion (Oxi:lyser, s::can, Austria), and the water-flow rate (Fluxus F501, Flexim, Germany) were measured online in the four fish tanks. The water pH (ProMinent, Germany), and the inflow water rate (Watson Marlow 630, Spirax-Sarco Engineering, UK) from Lake Peurunka and the water treatment field (Flexim) were measured online from both RAS units. The online measurement data were stored on an industrial com- puter (con::cube, s::can, Austria). The total ammonia nitrogen (TAN), nitrite, nitrate, and phosphate were analysed on site using a spectrophotometer (Procedure 8038 Nessler, LCK341/342, LCK340, and LCK349 respectively, DS 3900, Hach, USA). Alkalinity was analysed on site using a standard titration method (ISO 9963-1:1994) (TitraLab AT1000, Hach, Loveland, USA). The total and dissolved organic carbon were analysed in a commercial laboratory in fresh samples (SFS-EN 1484). Oxidation-reduction poten- tial (ORP), fluorescent dissolved organic matter (fDOM), pH, conduc- tivity, and turbidity (FNU) were measured using an YSI EXO probe (Xylem, USA). The temperature was measured using a YSI EXO probe, and online using a HOBO Pendant (MX2201, Onset Computer Corpo- ration, USA). Oxygen was measured using YSI EXO, YSI ProODO, and Ponsel OPTOD optical probes (Aqualabo Servises SA, France). The measurement intervals for the different water quality parameters are shown in Table 2. Water quality from the sludge dewatering process was analysed in a commercial laboratory, and samples were frozen (-22 C) before analysis (TSS and ash, SFS-EN 872:2005; CODCr, SFS 5504:1988; total nitrogen, EF2021; phosphate, EF2087; total phosphorus, SFS-EN ISO 17294-2; dry matter and ash, SFS 3008:1990). 3. Results 3.1. Removal rates and efficiencies Due to the slow start-up phase and low inflow concentrations, most of the volumetric removal rates were lower than expected (Table 3). At their highest, the nitrate removal rate in the WB was 11.6 g N m 3 d-1 Table 2 Measurement intervals from the water treatment field (W ˆ weekly, B ˆ biweekly, M ˆ monthly, O ˆ online). TAN ˆ total ammonia nitrogen, TOC ˆ total organic carbon, DOC ˆ dissolved organic carbon, ORP ˆ oxidation-reduction potential, fDOM ˆ fluorescent dissolved organic matter. Operating month TAN, Nitrite, Nitrate Alkalinity PO4-P TOC, DOC Temperature, Oxygen ORP, pH, fDOM, Conductivity, Turbidity July ’19 W W August ’19 W W September ’19 W W M M W W October ’19 W W M M O W November ’19 W M M M O W December ’19 W M M M O W January ’20 W M M M O W February ’20 B M M O W March ’20 B M M O W April ’20 B M M O W May ’20 B M M O W June ’20 B M M O W Table 3 Average volumetric removal rates (VRR), removal efficiencies (RE), inflow, and outflow concentrations, and their ranges from the woodchip bioreactor, con- structed wetland, sand infiltration, and total water treatment field. Parameter VRR (g m 3 d- 1) RE (%) Inflow (mg l 1) Outflow (mg l 1) Woodchip bioreactor NO3–N (n ˆ 35) 4.1 (0.5–11.6) 97 % (74–99 %) 28.1 (6.5–48.6) 0.65 (0.15–1.69) NO2–N (n ˆ 33) 0.03 (-0.01–0.36) 63 % (-540–100 %) 0.21 (0.02–2.72) 0.02 (0.00–0.16) TAN (n ˆ 35) 0.07 (-0.22–0.03) 225% (-708–14 %) 0.36 (0.12–1.14) 0.89 (0.37–1.81) PO4–P (n ˆ 10) 0.23 (0.12–0.40) 79 % (60–96 %) 2.07 (0.99–3.99) 0.42 (0.07–0.95) DOC (n ˆ 5) 9.39 (-14.01–-2.38) 1186% (-2240–-549 %) 6.4 (4.7–8.8) 77 (37–110) TOC (n ˆ 5) 9.79 (-16.45–-2.49) 1030% (-1931–-519 %) 7.42 (6.3–10) 81.2 (39–130) Constructed wetland NO3–N (n ˆ 34) 0.02 (-0.01–0.07) 29 % (-13–100 %) 0.65 (0.15–1.69) 0.47 (0.01–1.11) NO2–N (n ˆ 32) 0.00 (0.00–0.01) 28 % (-250–100 %) 0.02 (0.00–0.16) 0.01 (0.00–0.05) TAN (n ˆ 35) 0.03 (0.00–0.11) 32 % (3–64 %) 0.89 (0.37–1.81) 0.61 (0.29–1.18) PO4–P (n ˆ 10) 0.02 (-0.01–0.05) 14 % (-140–49 %) 0.42 (0.07–0.95) 0.24 (0.10–0.52) DOC (n ˆ 5) 4.32 (1.21–7.54) 65 % (45–81 %) 77 (37–110) 26 (12–41) TOC (n ˆ 5) 3.39 (0.22–- 9.24) 52 % (2–84 %) 81.2 (39–130) 37.8 (12–82) Sand infiltration NO3–N (n ˆ 32) 0.00 (0.00–0.01) 20 % (-38–100 %) 0.47 (0.01–1.11) 0.39 (0.09–0.80) NO2–N (n ˆ 30) 0.00 (0.00–0.00) 68% (-1000–100 %) 0.01 (0.00–0.05) 0.02 (0.00–0.05) TAN (n ˆ 33) 0.00 (-0.03–0.02) 9% (-140–100 %) 0.61 (0.29–1.18) 0.52 (0.12–1.14) PO4–P (n ˆ 10) 0.00 (0.00–0.01) 34 % (-67–94 %) 0.24 (0.10–0.52) 0.12 (0.03–0.26) DOC (n ˆ 5) 0.32 (0.09–0.66) 57 % (37–76 %) 26 (12–41) 10.9 (4.3–19) TOC (n ˆ 5) 0.58 (0.08–1.75) 59 % (31–78 %) 37.8 (12–82) 13.3 (5.1–22) Total water treatment field NO3–N (n ˆ 34) 0.48 (0.00–1.31) 98 % (91–100 %) 28.1 (6.5–48.6) 0.39 (0.09–0.80) NO2–N (n ˆ 35) 0.00 (0.00–0.04) 82 % (17–99 %) 0.21 (0.02–2.72) 0.02 (0.00–0.05) TAN (n ˆ 33) 0.00 (-0.02–0.01) 84% (-291–75 %) 0.36 (0.12–1.14) 0.52 (0.12–1.14) PO4–P (n ˆ 10) 0.03 (0.01–0.06) 92 % (78–99 %) 2.07 (0.99–3.99) 0.12 (0.03–0.26) DOC (n ˆ 5) 0.07 (-0.15–0.01) 72% (-171–25 %) 6.4 (4.7–8.8) 10.9 (4.3–19) TOC (n ˆ 5) 0.09 (-0.22–0.01) 80% (-186–19 %) 7.42 (6.3–10) 13.3 (5.1–22) J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 6 (HRT ˆ 2.3 d), and the phosphate removal rate was 0.4 g P m 3 d-1 (HRT ˆ 3.8 d). The WB was the most efficient for N removal, while the CW was the most efficient for carbon, as planned. The total ammonia nitrogen (TAN) was released from the WB, but the CW could remove 32 % and SF 9%. 3.2. Water quality changes The total ammonia nitrogen concentration increased during the first 10 months of operation in the outflow of the SF (Fig. 4), and the con- centration was higher in the outflow of the entire hybrid water treat- ment field than in the inflow. An unexpected nitrite peak was observed in the RAS, but it was almost completely removed during the water treatment process (Fig. 4). During the coldest part of the winter (January–March), phosphate leaching was observed (average inflow 0.22 mg l 1 and outflow 0.28 mg l 1, Fig. 5A). During the initial start-up period with lake water, con- ductivity increased after all treatment units (Fig. 5B). After the WB, pH decreased substantially: it was below 6 during the first five months of operation, but it increased in the CW and in the SF (Fig. 5C). Due to denitrification and possible sulphate reduction, alkalinity increased in the WB by 89 %, and 19 % in the CW (Fig. 5D). Some turbidity peaks were observed in all treatment units, but in general, turbidity increased in the WB, and decreased in the CW and SF (Fig. 5E). Between July and February, the WB released some dissolved organic matter (fDOM), but an average of 35 % of it was removed in the CW and SF (Fig. 5F). Inflow water was saturated with oxygen, and it was consumed in the WB. However, oxygen was also depleted in the CW and SF, which had reducing conditions (Fig. 6). The water temperature in the treatment system showed some seasonal trends, with few buffer effects against the local air temperature, and below-zero temperature values were not recorded at all (Fig. 7). During the first year of operation, the lowest water temperature was 4.2 C in the wetland unit (CW) and 2.6 C in the SF, although the average air temperature was -0.8 C between November and March (Fig. 7). The winter was unusually mild but long compared to the average Finnish winter. The highest frost depth was 27 cm in the WB, whereas it was somewhat less in the CW (15 cm) (Table 4). In the SF, the frost depth was recorded as deeper (62 cm) than in the control sampling point (46 cm) outside the water treatment area (Fig. 2). This was ex- pected, because it is common that soil frost is typically deeper in sandy soils than in e.g. clay due to its higher heat conductivity and lower heat capacity characteristics (e.g. Rajaei and Baladi, 2015). 3.3. Sludge management An average of 328 l of sludge from the sludge cones, drum filter backwash, and fixed bed bioreactors was produced per kg of feed. On average, 99 % of the total phosphorus and phosphate, and 86 % of total nitrogen, were removed in the coagulation process (Table 5). Thirty per cent of the water fed to the water treatment field was from the sludge supernatant. 4. Discussion 4.1. N balance The water treatment field was able to remove nitrogen, phosphorus, and organic matter efficiently throughout the year. The RAS start-up phase took longer than expected, which led to low loading to the water treatment field, and the water treatment field did not operate at its full purification capacity. The fish feeding rate was 10.1 kg d 1 at its peak, and the woodchip bioreactor (WB) operated at its lowest capacity of 21 kg feed per day (53 % of the designed full capacity). Due to the low feed rate, the WB was N-limited, because the effluent nitrate concen- tration was mostly below 1 mg l-1 (Robertson, 2010; von Ahnen et al., 2018v). The highest measured nitrate removal rate of 11.6 g NO3-N m 3 d 1 is similar to previously reported rates (5.3–25 g N m3 d 1) in pilot or full-scale WBs (Lepine et al., 2020; von Ahnen et al., 2018v; 2016). However, higher removal rates can be expected when the nitrate load increases, because the removal rate has been found to be positively correlated to the inflow concentration in the pre-testing phase for this full system (Kiani et al., 2020). There are also previous studies in which such a correlation was not observed (Robertson, 2010). During the coldest part of the winter (February–March), the removal rate was somewhat lower than during other months, being an average of 4.0 g N m3 d 1, but the removal efficiency remained high (99 %). The results proved that the WB also had a high capacity to remove nitrate during winter conditions. It should be noted that the studied winter was mild, and further studies are needed to investigate how colder air tempera- tures such as those below -10 ℃ affect nitrogen removal efficiency. Some ammonia was released from the WB, and this was probably caused by dissimilatory nitrate reduction to ammonium (DNRA), which is possible in conditions in which nitrogen limits denitrification instead of carbon (Lu et al., 2013; Zhao et al., 2018; Kiani et al., 2020). Some of this ammonia was removed in the CW, and the removal efficiency is expected to be higher when vegetation is fully grown and the microbial communities have matured, because they have been found to improve N Fig. 4. A) nitrate, B) total ammonia nitrogen, and C) nitrite concentrations from the inflow to the woodchip bioreactor, outflow from the woodchip bioreactor, outflow from the constructed vertical wetland, and outflow from the sand infiltration unit (DD.MM.YY). J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 7 removal processes in CWs (e.g. Zhang et al., 2016). Overall, TAN con- centration was higher after the treatment field than in the inflow to the water treatment field. This was caused by the DNRA in the WB, as well as the potential mineralisation of organic matter trapped in the different water treatment compartments. In the RAS context, this will cause elevated nitrification demand in indoor treatment steps if oxygen depletion continues, and nitrification is disabled in the field. 4.2. P balance A high removal efficiency of phosphate (80 %) was observed in the WB, which was not observed in previous studies of the WB treatment of aquaculture effluent. Lepine et al. (2020) reported a minor production of dissolved reactive phosphorus from the WB, and von Ahnen et al. (2016) did not observe a net release or net production of phosphate from the start-up period of the woodchip bioreactor. When using agricultural drainage, phosphate reduction was observed (Hua et al., 2016), which might be explained by the intake of phosphorus by the growing bacterial biomass, as well as extracellular polymetric substances that can absorb phosphate (Li et al., 2015). This may be related to the high hydraulic retention time in the WB (between 2.5 and 5.7 days), which enhanced the woodchip/wastewater contact time (Sharrer et al., 2016). In addi- tion, the inflow phosphate concentration in the WB decreased in the first six months of operation, which might be explained by the biofilm growth and maturation in the RAS, which assimilated some of the phosphorus released from the fish faeces. The sludge dewatering process effectively removed phosphorus and nitrogen. However, the dewatering process requires further optimisa- tion, because the relative amount of sludge leaving the process was still fairly high at 114 l kg 1 feed. There was a large variation in the Fig. 5. Phosphate concentration, conductivity, pH, alkalinity, turbidity and fluorescent dissolved organic matter (fDOM) from the inflow to the woodchip bioreactor, outflow from the woodchip bioreactor, outflow from the constructed vertical wetland, and outflow from the sand infiltration unit (DD.MM.YY). The arrow indicates the time point when the lake water flushing period ended and RAS water was introduced to the water treatment field. Fig. 6. Oxidation-reduction potential (ORP, A) and oxygen concentrations (B) from the inflow to the woodchip bioreactor, outflow from the woodchip bioreactor, outflow from the constructed vertical wetland, and outflow from the sand infiltration unit (DD.MM.YY). J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 8 properties of the sludge, which affected the quantity of solids settling in the coagulation tank. The gradient between clear supernatant water and thick sludge was not automatically detected, so as a precaution, more water was fed to the municipal water treatment plant, because sludge escaping to the water treatment field might have caused the WB to be clogged. 4.3. Organic matter and turbidity The main purpose of the slow sand infiltration was the removal of organic carbon and suspended solids, as well as the decrease in turbidity. During the first seven months of operation, it was able to remove 36 % of the organic matter (fDOM) when it was operated at an HRT of 19 days. After the flow rate was increased to an HRT of 10 days (7.8–10.3), the organic matter removal efficiency dropped to 6%. It has been reported that the removal of organic matter from the infiltration water is based on the physical filtering and chemical retention processes in the soil par- ticles and biological degradation of natural organic matter in the arti- ficial recharge of ground water, which is widely used in Finland for household water production (Kolehmainen et al., 2009; Lindroos et al., 2002). In ARG, lake water is infiltrated through sand- and gravel-containing soil layers to remove TOC from the infiltration water, and the sand filtration has also been shown to remove TOC from the RAS effluents in small-scale experiments (Lindroos et al., 2020). The mo- lecular size distribution of TOC is also reflected in the TOC removal; small molecular TOC fractions are not removed as effectively as large molecular fractions (Kolehmainen et al., 2009; Lindroos et al., 2002). All these processes are undoubtedly partly responsible for the organic matter removal in the SF. There are several possible reasons for fDOM removal being weaker during the spring and summer of 2020 than 2019. The fDOM values in wetland outflow dropped in the spring and summer of 2020 on two occasions, and the SF outflow did not respond to this drop as strongly. After the decline, the fDOM values soon increased again in the wetland outflow. Generally, the oxygen concentration was low in the wetland outflow and the SF. Oxygen is consumed by the biodegradation of organic matter, and it is likely that the higher oxygen content in the SF in 2019 contributed to the better organic matter removal during that period. The possible accumulation of organic matter in the SF may also have affected the removal efficiency of organic matter from the infil- tration water. Turbidity decreased significantly in the SF, as has also been reported in small-scale RAS experiments (Lindroos et al., 2020). 4.4. Future perspectives and challenges The low oxygen concentration in the water entering the CW and SF appeared to be the main challenge when treating RAS effluent. Winter conditions mean that the distribution pipes need to be placed under- ground to prevent freezing when the oxygenation of water is limited. In addition, snow and soil frost decrease the transport of oxygen from the atmosphere, posing challenges for water oxidation. However, when vegetation and microbial communities in the CW are fully grown, it is expected that the water oxygen level will improve, because the common reed has a good capacity to enhance microbiological activity (e.g. Gagnon et al., 2007) and increase the oxygen level in the media (e.g. Nivala et al., 2013). In addition, the high reducing conditions in the SF caused the dissolution of potentially toxic metals such as arsenic (As), cobalt (Co), and manganese (Mn) in the water (manuscript in progress). However, these metals can be converted back to solid form by vigorous aeration and recovered in the fixed bed bioreactor. Artificial aeration or recirculation of water within the CW unit has been found to be efficient at improving removal processes (Poldvere et al., 2009; Ji et al., 2020), and they may be solutions for increasing the oxygen level in the outflow of the CW and preventing metal leaching in the SF. In the start-up phase, waterflow rates and the N load on the WB were limited, which led to much longer HRT than planned. Due to this, a high amount of organic matter leached from the WB, further aggravating the oxygen depletion in the CW. To fully control the removal rates of ni- trogen in the WB, the water level adjustment wells need to be rede- signed, because the outlet wells were operated already operating at their lowest level. This means that there was no option to shorten the HRT with the available flow-rate. Experience has shown the importance of proper HRT and the adjustability of the system hydraulics for efficient removal rates and the prevention of carbon leaching from the woodchips. When designing a full-scale hybrid water treatment field, it would be Fig. 7. Air and water temperature from the inflow to the woodchip bioreactor, outflow from the woodchip bioreactor, outflow from the constructed vertical wetland, and outflow from the sand infiltration unit (D.MM.YY). Table 4 Frost depths (cm) in different locations of the water treatment field and average air temperature (C) from two weeks prior to the frost measurements. Place \ Date Dec. 2 Jan. 2 Feb. 3 March 2 April 14 May 4 Bioreactor 6 7 10 26 0 0 Bioreactor 0 0 0 27 0 0 Bioreactor 5 4 6 23 0 0 Wetland 0 0 0 15 0 0 Wetland 2 0 0 10 0 0 Sand infiltration 7 10 25 56 0 0 Sand infiltration 0 13 36 62 0 0 Sand infiltration 8 12 36 57 0 0 Control 2 3 25 46 42 0 Air temperature 0.6 1.6 2.6 2.0 1.9 5.8 Table 5 Mean water quality values ( SD) from the inflow, sludge, and supernatant from the RAS sludge and removal efficiency treated with polyaluminium chloride and sedimentation (n ˆ 3). TS ˆ total solids, TSS ˆ total suspended solids, TVS ˆ total volatile solids, TVSS ˆ total volatile suspended solids, CODcr ˆ chemical oxygen demand. Parameter Inflow Sludge to the municipal water treatment plant Supernatant to the water treatment field Removal efficiency TS (g l 1) 3.29  1.28 37  12 TVS (g l 1) 3.16  8 33.3  19.1 TSS (mg l 1) 36  12 TVSS (mg l 1) 32.2  10.8 Total phosphorus (mg l 1) 46  6 142  128 0.48  0.12 98.9 %  0.4 PO4-P (mg l 1) 3.53  2.03 6.80  5.81 0.06  0.03 99.4 %  0.1 Total nitrogen (mg l 1) 240  22 230  89 27  6 86.5 %  2.8 CODCr (mg O2 l 1) 5167  942 8667  3399 243  104 93.5 %  2.5 J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 9 useful to look for sites where water can be distributed between units by gravity. In addition, a sand filter requires hundreds of cubic meters of sand, so it should be built in areas with natural sand deposits. It must also be carefully examined whether the CW could achieve sufficient P and OM removal, making the SF unnecessary. Author statement Jani Pulkkinen: Writing - Original Draft, Formal analysis, Meth- odology, Visualization Anna-Kaisa Ronkanen: Writing - Original Draft, Conceptualization Antti Pasanen: Resources, Writing - Review & Edit- ing, Conceptualization Sepideh Kiani: Formal analysis, Writing - Re- view & Editing, Tapio Kiuru: Conceptualization, Resources, Project administration Juha Koskela: Writing - Review & Editing Petra Lind- holm-Lehto: Writing - Review & Editing Antti-Jussi Lindroos: Formal analysis, Writing - Review & Editing Muhammad Muniruzzaman: Methodology, Software, Writing - Review & Editing Lauri Solismaa: Writing - Review & Editing, Bjorn Klove: Conceptualization, Resources Jouni Vielma: Conceptualization, Resources, Project administration. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements This study was supported by the Finnish Ministry of Agriculture and Forestry, Natural Resources Institute Finland, and the European Union through the European maritime and fisheries fund. The staff at Laukaa fish farm are greatly appreciated for their fish husbandry and system management. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.aquaeng.2021.102161. References Chen, D., Chen, H.W., 2013. Using the Koppen classification to quantify climate variation and change: an example for 1901-2010. Environ. Dev. 6, 69–79. Christianson, L.E., Bhandari, A., Helmers, M.J., 2012. A practice-oriented review of woodchip bioreactors for subsurface agricultural drainage. Appl. Eng. Agric. 28, 861–874. Cooper, P., 1999. A review of the design and performance of vertical-flow and hybrid reed bed treatment systems. Water Sci. Technol. 40, 1–9. Dalsgaard, J., Lund, I., Thorarinsdottir, R., Drengstig, A., Arvonen, K., Pedersen, P.B., 2013. Farming different species in RAS in Nordic countries: current status and future perspectives. Aquacult. Eng. 53, 2–13. Dalsgaard, J., von Ahnen, M., Naas, C., Pedersen, P.B., 2018. Nutrient removal in a constructed wetland treating aquaculture effluent at short hydraulic retention time. Aquacult. Environ. Interact. 10, 329–343. Dupla, M., Comeau, Y., Parent, S., Villemur, R., Jolicoeur, M., 2006. Design optimization of a self-cleaning moving-bed bioreactor for seawater denitrification. Water Res. 40, 249–258. Finnish Meteorological Institute, 2020. Open Database. https://en.ilmatieteenlaitos.fi/ open-data. Gagnon, V., Chazarenc, F., Comeau, Y., Brisson, J., 2007. Influence of macrophyte species on microbial density and activity in constructed wetlands. Water Sci. Technol. 56, 249–254. Greenan, C.M., Moorman, T.B., Parkin, T.B., Kaspar, T.C., Jaynes, D.B., 2009. Denitrification in wood chip bioreactors at different water flows. J. Environ. Qual. 38, 1664–1671. Gronroos, J., Seppala, J., Silvenius, F., Makinen, T., 2006. Life cycle assessment of Finnish cultivated rainbow trout. Boreal Environ. Res. 11, 410–414. Gutierrez-Wing, M.T., Malone, R.F., Rusch, K.A., 2012. Evaluation of polyhydroxylbutyrate as a carbon source for recirculating aquaculture water denitrification. Aquacult. Eng. 51, 36–43. Heikkinen, K., Karppinen, A., Karjalainen, S.M., Postila, H., Hadzic, M., Tolkkinen, M., Marttila, H., Ihme, R., Kløve, B., 2018. Long-term purification efficiency and factors affecting performance in peatland-based treatment wetlands: an analysis of 28 peat extraction sites in Finland. Ecol. Eng. 117, 153–164. Homyak, P.M., Yanai, R.D., Burns, D.A., Briggs, R.D., Germain, R.H., 2008. Nitrogen immobilization by wood-chip application: protecting water quality in a northern hardwood forest. For. Ecol. Manag. 255, 2589–2601. Hua, G., Salo, M.W., Schmit, C.G., Hay, C.H., 2016. Nitrate and phosphate removal from agricultural subsurface drainage using laboratory woodchip bioreactors and recycled steel byproduct filters. Water Res 102, 180–189. Ji, B., Zhao, Y., Vyzamal, J., Qiao, S., Wei, T., Li, J., Mander, Ü., 2020. Can subsurface flow constructed wetlands be applied in cold climate regions? A review of the current knowledge. Ecol. Eng. 157, 105992. Kadlec, R., Wallace, S., 2009. Treatment Wetlands. CRC Press, Boca Raton, Florida, USA. Kiani, S., Kujala, K., Pulkkinen, J.T., Aalto, S.A., Suurnakki, S., Kiuru, T., Tiirola, M., Kløve, B., Ronkanen, A.-K., 2020. Enhanced nitrogen removal of low carbon wastewaters in denitrification bioreactors by utilizing industrial waste toward circular economy. J. Clean. Prod. 254, 119973. Kolehmainen, R.E., Kortelainen, N.M., Langwalt, J.H., Puhakka, J.A., 2009. Biodegradation of natural organic matter in long-term continuous-flow experiments simulating artificial groundwater recharge for drinking water production. J. Environ. Qual. 38, 44–52. Krom, M.D., David, A.B., Ingall, E.D., Benning, L.G., Clerici, S., Bottrell, S., Davies, C., Potts, N.J., Mortimer, R.J.G., van Rijn, J., 2014. Bacterially mediated removal of phosphorus and cycling of nitrate and sulfate in the waste stream of a “zero- discharge” recirculating mariculture system. Water Res. 56, 109–121. Lepine, C., Christianson, L., Davidson, J., Summerfelt, S., 2018. Woodchip bioreactors as treatment for recirculating aquaculture systems’ wastewater: a cost assessment of nitrogen removal. Aquacult. Eng 83, 85–92. Lepine, C., Christianson, L., Mclsaac, G., Summerfelt, S., 2020. Denitrifying bioreactor inflow manifold design for treatment of aquacultural wastewater. Aquacult. Eng 88, 102036. Letelier-Gordo, C.O., Huang, X., Aalto, S.L., Pedersen, P.B., 2020. Activated sludge denitrification in marine recirculating aquaculture system effluent using external and internal carbon sources. Aquacult. Eng 90, 102096. Li, W.-W., Zhang, H.-L., Sheng, G.-P., Yu, H.-Q., 2015. Roles of extracellular polymetric substances in enhanced biological phosphorus removal processes. Water Res 86, 85–95. Lindholm-Lehto, P., Pulkkinen, J., 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. Lindroos, A.-J., Kitunen, V., Derome, J., Helmisaari, H.-S., 2002. Changes in dissolved organic carbon during artificial recharge of groundwater in a forested esker in southern Finland. Water Res. 36, 4951–4958. 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 effluents of experimental recirculating aquaculture systems. Water Air Soil Pollut. 231, 209. Lu, W.-W., Zhang, H.-L., Shi, W.-M., 2013. Dissimilatory nitrate reduction to ammonium in an anaerobic agricultural soil as affected by glucose and free sulfide. Eur. J. Soil Biol. 58, 98–104. Muniruzzaman, M., Rolle, M., 2015. Impact of multicomponent ionic transport on pH fronts propagation in saturated porous media. Water Resour. Res. 51, 6739–6755. Muniruzzaman, M., Rolle, M., 2016. Modeling multicomponent ionic transport in groundwater with IPhreeqc coupling: electrostatic interactions and geochemical reactions in homogeneous and heterogeneous domains. Adv. Water Resour. 98, 1–15. Muniruzzaman, M., Rolle, M., 2017. Experimental investigation of the impact of compound-specific dispersion and electrostatic interactions on transient transport and solute breakthrough. Water Resour. Res. 53, 1189–1209. Muniruzzaman, M., Rolle, M., 2019. Multicomponent ionic transport modeling in physically and electrostatically heterogeneous porous media with PhreeqcRM coupling for geochemical reactions. Water Resour. Res. 55, 11121–11143. Muniruzzaman, M., Haberer, C.M., Grathwohl, P., Rolle, M., 2014. Multicomponent ionic dispersion during transport of electrolytes in heterogeneous porous media: experiments and model-based interpretation. Geochim. Cosmochim. Acta 141, 656–669. Muniruzzaman, M., Karlsson, T., Ahmadi, N., Rolle, M., 2020. Multiphase and multicomponent simulation of acid mine drainage in unsaturated mine waste: modeling approach, benchmarks and application examples. Appl. Geochem. 120, 104677. Nivala, J., Wallace, S., Headley, T., Kassa, K., Brix, H., van Afferden, M., Müller, R., 2013. Oxygen transfer and consumption in subsurface flow treatment wetlands. Ecol. Eng. 61, 544–554. Nordstom, A., Herbert, R.B., 2018. Determination of major biogeochemical processes in a denitrifying woodchip bioreactor for treating mine drainage. Ecol. Eng. 110, 54–66. Pelletier, N., Tyedmers, P., Sonesson, U., Scholz, A., Ziegler, F., Flysjo, A., Kruse, S., Cancino, B., Silverman, H., 2009. Not all salmon are created equal: life cycle assessment (LCA) of global salmon farming systems. Environ. Sci. Technol. 43, 8730–8736. Piedrahita, R.H., 2003. Reducing the potential environmental impact of tank aquaculture effluents through intensification and recirculation. Aquaculture 226, 35–44. Poldvere, E., Karabelnik, K., Noorvee, A., Maddison, M., Nurk, K., Zaytsev, I., Mander, Ü., 2009. Improving wastewater effluent filtration by changing flow regimes—investigations in two cold climate pilot scale systems. Ecol. Eng. 35, 193–203. J.T. Pulkkinen et al. Aquacultural Engineering 93 (2021) 102161 10 Postila, H., Heiderscheidt, E., 2020. Function and biomass production of willow wetlands applied in the polishing phase of sewage treatment in cold climate conditions. Sci. Total Environ. 727, 138620. Rajaei, P., Baladi, G., 2015. Frost depth: general prediction model. Transp. Res. Rec. 2510, 74–80. Robertson, W.D., 2010. Nitrate removal rates in woodchip media of varying age. Ecol. Eng. 36, 1581–1587. Rolle, M., Muniruzzaman, M., Haberer, C.M., Grathwohl, P., 2013. Coulombic effects in advection-dominated transport of electrolytes in porous media: multicomponent ionic dispersion. Geochim. Cosmochim. Acta 120, 195–205. Rolle, M., Sprocati, R., Masi, M., Jin, B., Muniruzzaman, M., 2018. Nernst-planck-based description of transport, coulombic interactions, and geochemical reactions in porous media: modeling approach and benchmark experiments. Water Resour. Res. 54, 3176–3195. Sharrer, K.L., Christianson, L.E., Lepine, C., Summerfelt, S.T., 2016. Modeling and mitigation of denitrification ‘woodchip’ bioreactor phosphorus releases during treatment of aquaculture wastewater. Ecol. Eng. 93, 135–143. Silvenius, F., Gronroos, J., Kankainen, M., Kurppa, S., Makinen, T., Vielma, J., 2017. Impact of feed raw material to climate and eutrophication impacts of Finnish rainbow trout farming and comparisons on climate impact and eutrophication between farmed and wild fish. J. Clean. Prod. 164, 1467–1473. Sprocati, R., Masi, M., Muniruzzaman, M., Rolle, M., 2019. Modeling electrokinetic transport and biogeochemical reactions in porous media: a multidimensional Nernst–planck–poisson approach with PHREEQC coupling. Adv. Water Resour. 127, 134–147. Suliman, F., Futsaether, C., Oxaal, U., Haugen, L.E., Jenssen, P., 2006. Effect of the inlet–outlet positions on the hydraulic performance of horizontal subsurface-flow wetlands constructed with heterogeneous porous media. J. Contam. Hydrol 87, 22–36. Tal, Y., Schreier, H.J., Sowers, K.R., Stubblefield, J.D., Place, A.R., Zohar, Y., 2009. Environmentally sustainable land–based marine aquaculture. Aquaculture 286, 28–35. Tanner, C.C., 1996. Plants for constructed wetland treatment systems – a comparison of the growth and nutrient uptake of eight emergent species. Ecol. Eng. 7, 59–83. von Ahnen, M., Pedersen, P.B., Dalsgaard, J., 2016v. Start-up performance of a woodchip bioreactor operated end-of-pipe at a commercial fish farm–a case study. Aquacult. Eng 74, 96–104. von Ahnen, M., Pedersen, P.B., Dalsgaard, J., 2018v. Performance of full-scale woodchip bioreactors treating effluents from commercial RAS. Aquacult. Eng 83, 130–137. Vymazal, J., 2007. Removal of nutrients in various types of constructed wetlands. Sci. Total Environ. 380, 48–65. Vymazal, J., 2011. Plants used in constructed wetlands with horizontal subsurface flow: a review. Hydrobiologia 674, 133–156. Wang, J., Chu, L., 2016. Biological nitrate removal from water and wastewater by solid- phase denitrification process. Biotechnol. Adv. 34, 1103–1112. Wang, Y., Shen, L., Wu, J., Zhong, F., Chen, S., 2020. Step-feeding ratios affect nitrogen removal and related microbial communities in multi-stage vertical flow constructed wetlands. Sci. Total Environ. 721, 137689. Zhang, S., Xiao, R., Liu, F., Zhou, J., Li, H., Wu, J., 2016. Effect of vegetation on nitrogen removal and ammonia volatilization from wetland microcosms. Ecol. Eng. 97, 363–369. Zhao, J., Feng, C., Tong, S., Chen, N., Dong, S., Peng, T., Jin, S., 2018. Denitrification behavior and microbial community spatial distribution inside woodchip-based solid- phase denitrification (W_SPD) bioreactor for nitrate-contaminated water. Bioresour. Technol. Rep. 249, 869–879. J.T. Pulkkinen et al.