Constructed wetlands with free water surface for treatment of aquaculture effluents

Introduction

Land-based rainbow trout farming systems are commonly operated in open or semi-open ponds, raceways or tanks, with a flow-through water supply from adjacent rivers, brooks or wells. Such open aquaculture systems (the focus of our investigations) discharge their effluents into the environment with enhanced nutrient and solid loads compared with intake water; these effluents may have serious environmental consequences when discharged untreated (Alabaster, 1982; Camargo, 1992; Rennert, 1994). The discharges of nitrogen and phosphorus with particular environmental importance (derived from uneaten feed and excretion products) from German rainbow trout production of 25 000 tonnes a⁻¹ amounts 1300 tonnes (t) total nitrogen (TN) and 150 t total phosphorus (TP) annually.

Nutrients in aquaculture effluents are distributed in particulate and soluble fractions (Ackefors and Enell, 1994). The 7–32% of TN, 30–84% of TP and up to 27% of total carbon are bound in the particulate fraction, and the remainder are dissolved in the effluents (Bergheim et al., 1993a,b). Microscreens or sedimentation tanks can reduce the nutrient load based on the mechanical separation of solids to clarify the effluents (Hussenot et al., 1998; Cripps and Bergheim, 2000; Boyd and Queiroz, 2001). The treatment efficiency of microscreens depends upon the effluent nutrient concentration and mesh size of microscreens. Removal efficiencies of 50–74% for solids, 10–43% for TN and 49–63% for TP, have been reported (Mäkinen et al., 1988; Hennessy, 1991; Bergheim and Forsberg, 1993; Bergheim et al., 1993a,b; Wedekind, 1996; Lekang et al., 2000). Sedimentation tanks or ponds facilitate solid removal of 58–97% (Fladung, 1993; Bergheim et al., 1998). Values for reduction of phosphorus in sedimentation ponds were 34% when performed with a HRT of 30 min. Mechanical treatment methods are not able to reduce dissolved nutrients like NH (up to 90% of total excreted nitrogen), urea, soluble phosphorus or carbon compounds. Therefore, use of systems which reduce both solids and dissolved nutrients should be an aim of sustainable aquaculture.

The use of wetlands in water purification has been used in different parts of the world to successfully treat agricultural, municipal or industrial wastewaters (Verhoeven and Meuleman, 1999). In both natural and constructed wetlands, aquatic plants are an important component of the purification system. Depending on the design and function, wetland treatment systems can be divided into: (i) constructed wetlands with free water surface and no effluent percolation through the soil, (ii) constructed wetlands with subsurface flow and complete effluent percolation through the soil, and (iii) hydrobotanic systems intermediate between ponds and root zone systems.

The objective of this study was to evaluate the removal efficiency for various environmental quality parameters from flow-through aquaculture systems via constructed wetlands operated at three different HRTs.

Materials and methods

Rainbow trout farm and constructed wetlands

The investigations were carried out at a rainbow trout farm operating as a flow-through system in the state of Brandenburg, Germany (Fig. 1). The farm comprised 18 tanks with a total volume of about 70 m³ and a water supply of 150 L s⁻¹ from a small lowland brook. Stocking of fish tanks varied between 70 and 100 kg fish m⁻³, and total fish production was 13 t per year by feeding 1% of fish biomass [extruded diet: 43.8% d. m. protein, 22.9% d. m. lipid, 19.6% d. m. NfE, 9.53% d. m. ash; feed conversion ratio (FCR) of 1].

The farm produces exclusively rainbow trout fingerlings, starting with the first feeding-phase in May and ending with harvesting in October; for the remainder of the year, the fish farm is closed down. Five constructed treatment wetlands with free water surface were attached to the farm; three of these identical wetlands were studied. Each wetland had an area of 350 m² (7 × 50 m) and a hydroperiod of 50% (6 months of water supply). The depth of the wetlands ranged from almost zero at the inlet to 1.2 m at the outlet (slope of 2.4%). The plant community (Phalaris arundinacea, Glyceria maxima, Polygonum hydropiper, Juncus sp., Sparganium sp., Potamogeton natans, Lemna minor, Ceratophyllum demersum, Elodea canadensis) in the wetland established naturally, and was dominated by P. arundinacea. The hydraulic loads of the three wetlands were 5 L s⁻¹, 10 L s⁻¹ to l5 L s⁻¹, resulting in theoretical HRT of 3.5 h, 5.5 h and 11 h, respectively. According to the production period, the wetlands are filled from May to October; during this time they are maintenance-free. After draining, in November the inlet zones of the wetlands are cleaned of settled-out solids and particles.

Water sampling

After an adjustment period of 1 month (May–June) to allow natural establishment of a substantial plant population and sufficient microbial activity in the wetland, water sampling was begun. Because of diurnal fluctuations in nutrient load in rainbow trout effluents (Rennert, 1994), the samples were taken automatically at 6 min intervals for a 24 h period at the central inlet pipe and at the outlet pipe of each wetland. 24 h samples of each monitoring point were pooled for chemical analysis to obtain daily average values. The 24 h monitoring was repeated every 2 weeks between June 15 (day 0) and October 5 (day 112), 2000. Prior to chemical analysis samples were stored at −17°C and pH 2 (by addition of HCl) as described by Zwirnmann et al. (1999). Seven parameters were analyzed: total suspended solids (TSS), total nitrogen (TN), ammonia-nitrogen (NH-N), nitrate-nitrogen (NO-N), TP, soluble reactive phosphorus (SRP) and chemical oxygen demand (COD). Chemical analysis was carried out following the methods described by the German institute of standardization (DIN). For TSS, DIN 38409 (1987) was utilized by using a fiberglass filter (Whatham GF/F) with a 0.45 μm pore size. The nitrogen fractions NH-N (DIN EN ISO 11732, 1998) and NO-N (DIN EN ISO 13395, 1997) were analyzed photometrically with a flow solution system (PERSTORP) and TP (DIN EN ISO 1189, 1997) and SRP (DIN EN ISO 1189) with standard photometer (CARY 1E). The chemoluminescence method for detecting TN (DIN EN 12260, 1996) was performed with the TN – Analyzer (Abimed). COD was detected with the Dr Lange test kits and the LP 2W photometer (Dr Lange, Düsseldorf, Germany). The physico-chemical water parameters, namely pH, oxygen content, water conductivity, and temperature were monitored with a multimeter sensor (Horiba, Kyoto, Japan).

Statistics

Mean concentrations were calculated for the seven nutrients and four physico-chemical water parameters at the central inlet and at each outlet of the three wetlands. The adjacent levene test scrutinized the homogeneity of the variance (Lozan and Kausch, 1998). The principal influence of the effluent treatment by the wetland and the hydraulic load was tested by one-way anova. The difference of mean values was evaluated for significance by the range test TUCKEY HSD (P = 0.05) for homogenous variance, and by the range tests DUNNETT T3 (P = 0.05) for inhomogenous variances, respectively. Nutritional removal (R), areal loading (AL), and areal removal (AR) were calculated by the mean of nutrient concentrations.

Results

Water parameters and removal performance

Table 1 describes the nutrient values measured at the central inlet pipe and at each outlet of investigated constructed wetlands. The negative influence of a decreasing HRT on the cleaning performance of the purification systems was statistically significant only for the nitrogen parameters TN and NH-N. The concentrations of four nutrients TSS, COD, TP and SRP, were significantly lower in the outlet water of the wetlands than in the inlet water, but there was no significant difference in their concentrations among the three wetlands. Content of COD was reduced approximately 18 mg L⁻¹ to levels of 12 mg L⁻¹ in each wetland. Approximately 10 mg L⁻¹ of inflowing suspended solids could be removed to contents of 2.5–3.2 mg L⁻¹ in the wetlands outlets. The phosphorus parameters TP and SRP decreased from 0.19 and 0.05 mg L⁻¹ after wetland passage to values of 0.09–0.11 and 0.04–0.05 mg L⁻¹, respectively. Reduction of inflowing TN (1.18 mg L⁻¹) was influenced by the effluent residence time, with TN of 0.81 mg L⁻¹ at HRT of 11 h and 0.96 mg L⁻¹ at HRT of 3.5 h. This reduction was mainly based on NH-N metabolism, which decreased inflowing NH-N from 0.27 to 0.11 mg L⁻¹ (at HRT of 11 and 5.5 h) and 0.16 mg L⁻¹ (at HRT of 3.5 h), whereas the inflowing content of NO-N (0.38 mg L⁻¹) was decreased to 0.34 mg L⁻¹ at HRT of 11 h and increased to 0.46 mg L⁻¹ at HRT of 3.5 h.

Table 1. Mean () concentrations (c) ± SD, reduction (R), areal loading (AL) and areal removal (AR) for total suspended solids (TSS), chemical oxygen demand (COD), total phosphorus (TP), soluble reactive phosphorus (SRP), total nitrogen (TN), nitrate-nitrogen (NO-N), ammonia-nitrogen (NH-N) and physico-chemical water parameters pH, oxygen (O₂), conductivity (C) and temperature (T) in the inlet and the effluent waters of the three constructed wetlands with different hydraulic residence times (HRT).

	Inlet	HRT
		11 h	5.5 h	3.5 h
TSS
c (mg L−1)	9.9a ± 3.6	2.5b ± 1.2	3.1b ± 1.7	3.2b ± 1.3
R (%)		72a ± 17	66a ± 23	67a ± 10
AL (g m−2 day−1)		12.3 ± 4.4	24.6 ± 8.8	36.8 ± 13.2
AR (g m−2 day−1)		9.2a ± 4.8	16.8a,b ± 9.2	25.0b ± 9.5
COD
c (mg L−1)	17.6a ± 4.9	12.3b ± 4.1	12b ± 2.6	12.1b ± 3.1
R (%)		30a ± 12	30a ± 13	31a ± 7
AL (g m−2 day−1)		21.7 ± 6.1	43.5 ± 12.2	65.2 ± 18.3
AR (g m−2 day−1)		6.6a ± 2.9	13.9a,b ± 7.9	20.5b ± 8.0
TP
c (mg L−1)	0.19a ± 0.05	0.09b ± 0.02	0.09b ± 0.02	0.11b ± 0.03
R (%)		53a ± 9	49a ± 11	41a ± 12
AL (g m−2 day−1)		0.23 ± 0.06	0.47 ± 0.12	0.70 ± 0.18
AR (g m−2 day−1)		0.12a ± 0.05	0.23b ± 0.09	0.30b ± 0.14
SRP
c (mg L−1)	0.05a ± <0.01	0.04b ± 0.01	0.04b ± 0.01	0.05a,b ± 0.01
R (%)		18a ± 14	23a ± 18	3a ± 20
AL (g m−2 day−1)		0.06 ± <0.01	0.12 ± 0.01	0.18 ± 0.01
AR (g m−2 day−1)		0.01a ± 0.01	0.03a ± 0.02	0.01a ± 0.04
TN
c (mg L−1)	1.18a ± 0.12	0.81b ± 0.13	0.89b,c ± 0.08	0.96c ± 0.09
R (%)		30a ± 14	24a ± 8	19a ± 5
AL (g m−2 day−1)		1.46 ± 0.15	2.91 ± 0.31	4.37 ± 0.46
AR (g m−2 day−1)		0.45a ± 0.24	0.71a,b ± 0.30	0.82b ± 0.31
NO-N
c (mg L−1)	0.38a ± 0.02	0.34a,b ± 0.11	0.43a,b ± 0.05	0.46b ± 0.05
R (%)		12a ± 30	−13a ± 11	−19a ± 8
AL (g m−2 day−1)		0.47 ± 0.02	0.95 ± 0.05	1.42 ± 0.07
AR (g m−2 day−1)		0.06a ± 0.14	−0.13b ± 0.10	−0.27b ± 0.12
NH-N
c (mg L−1)	0.27a ± 0.03	0.11b ± 0.04	0.11b,c ± 0.05	0.16c ± 0.05
R (%)		60a ± 15.28	59a ± 14.52	41b ± 14.16
AL (g m−2 day−1)		0.33 ± 0.04	0.66 ± 0.07	1.00 ± 0.11
AR (g m−2 day−1)		0.20a ± 0.07	0.38b ± 0.09	0.40b ± 0.12
pH	7.6a ± 0.1	7.3b ± 0.3	7.3b ± 0.2	7.5a,b ± 0.2
O2 (mg L−1)	6.7a ± 0.4	3.5b ± 1.4	4.1b ± 1.0	5.6c ± 0.7
C (μs cm−2)	250a ± 12	249a ± 11	250a ± 13	249a ± 13
T (°C)	12.9a ± 1.3	13.2a ± 1.5	12.8a ± 1.3	12.7a ± 1.4

• Values (each based on n = 9) with same superscript letter do not differ significantly.

Areal nutrient loading and areal nutrient removal differed among the wetlands. Loading varied from 12 to 37 g TSS m⁻² day⁻¹ and removal varied from 9 to 25 g TSS m⁻² day⁻¹. COD and TP loadings of the wetlands amounted to 22–65 g m⁻² day⁻¹ and 0.23–0.70 g m⁻² day⁻¹, which were reduced at rates of 7–21 g COD m⁻² day⁻¹ and 0.1–0.3 g TP m⁻² day⁻¹, respectively. Areal TN loading and removal fluctuated among the wetlands from 1.46–4.37 g m⁻² day⁻¹ and 0.47–1.42 g m⁻² day⁻¹, respectively. According to removal percentage, NO-N was only removed in the wetland with the highest HRT of 11 h. Wetlands with HRT of 5.5 and 3.5 h released NO-N with rates of 0.13 and 0.27 g m⁻² day⁻¹. Elimination of TN was mainly influenced by the removal of NH-N, which amounted from 0.20 g m⁻² day⁻¹ with HRT of 11 h up to 0.40 g m⁻² day⁻¹ with HRT of 3.5 h.

The inflowing pH of 7.57 was reduced with increasing HRTs of the wetlands from 7.46 to 7.32. Oxygen levels of 6.69 mg L⁻¹ in the inlet also decreased with increasing HRTs in wetlands from 5.59 to 3.52 mg L⁻¹. Mean temperature and conductivity fed to and within the effluents of the constructed wetlands were not significantly different, being 13°C and 250 μs cm⁻², respectively.

Time course of removal

The temporal changes in reduction of TSS, COD, TN and TP in the constructed wetlands with different HRT are shown in Fig. 2. It was found that the nutrient removal fluctuated in every wetland during the 5-month investigation. The reduction of suspended solids increased in the first months (June–August) from 30% up to approximately 80% in all wetlands and then stabilized at 60–80%. The COD removal was found to perform with increasing values up to 40–50% from June to July and then decreased. High fluctuations were also found for TP removal in all wetlands. From June to July, TP elimination amounted in all wetlands to 44–60% and then decreased especially in the wetland with HRT of 3.5 h to 20%. In September, TP reduction increased up to 63% (at HRT of 11 and 5.5 h) and decreased again in October. The TN reduction showed smallest fluctuations on a minor level in all wetlands. In contrast to the mentioned nutrients, nitrogen elimination showed no increase during the first month of observation. The wetland with HRT of 11 h decreased the TN content with lowest rates of a maximum 30%, whereas the wetland with HRT of 3.5 h removed on levels of 52%.

Discussion

Reports from different investigations on nutritional composition of untreated aquaculture effluents resulted in wide ranges for TSS 5–50 mg L⁻¹, for TP 0.05–0.50 mg L⁻¹, for SRP 0.06–0.15 mg L⁻¹, for TN 0.2–3.3 mg L⁻¹, for NO-N 0.7–2.2 mg L⁻¹ and for NH-N 0.5–1.1 mg L⁻¹ (Bergheim et al., 1993a; Ackefors and Enell, 1994; Kelly et al., 1994; Dumas et al., 1998; Bergheim and Brinker, 2003). It has been shown that the nutritional composition of aquaculture effluents depends on various parameters concerning hydraulic management, oxygen and feeding management (Summerfelt et al., 1999; Cripps and Bergheim, 2000). Because of enhanced feed formulas, improved feeding systems and optimization of the rearing equipment over the years, FCR of (German) rainbow trout production declined from 1.5–2.0 in 1985 to 0.7–1.0 in 2000 (Bergheim and Brinker, 2003). Therefore, relative effluent load decreased from 120 kg TN and 15–20 kg TP in 1985 to 40 kg TN and 7 kg TP per tonne of rainbow trout produced. Compared with cited investigations, the nutritional load found in the rainbow trout effluents of this study was low and reflects the described improvements in feed and farm management.

The nutrient concentrations in the wetland outlets indicate that sludge removal and stabilization occurred within every constructed wetland. Especially for TSS, high removals of 67–72% in every wetland were observed, but with no statistical differences among the three wetlands. Presented results are in agreement with previous studies (Bahlo and Wach, 1993; Schwartz and Boyd, 1995; Kadlec and Knight, 1996; Sansanayuth et al., 1996), where a constant mechanical screening and sedimentation in constructed wetlands with free water surface or subsurface water flow was observed, and varying HRTs of effluents did not alter the removal of TSS. Cripps and Bergheim (2000) reported that overflow rates in settling basins of 1.5–3 m h⁻¹ produce adequate settlement of aquaculture solids. In this study, hydraulic loading of each wetland was much lower, because the HRT did not effect TSS removal. As a result of similar TSS removals of 67–72% in the wetlands, areal TSS removal of 9.2–25 g m⁻² day⁻¹ increased with increasing loadings of 12.3–36.8 g m⁻² day⁻¹. Tanner et al. (1995) found in constructed wetlands microcosms for treatment of municipal sewage similar correlations between areal loading and removal but at minor loadings of 1.9–8.5 g TSS m⁻² day⁻¹. Müller (2000) advised a maximum of 15 g TSS m⁻² day⁻¹ for sustainable removal efficiencies in constructed wetlands for treatment of municipal sewage. Higher areal TSS loading at HRT 3.5 and 5.5 h (36.8 and 24.6 g m⁻² day⁻¹) in this study did not result in resuspension of settled out solids. Therefore, wetlands management (hydroperiod of 50% with draining and cleaning the inlet zones) seems to support the constant removal efficiencies observed after plant establishment (up to day 28).

After sedimentation and filtration of suspended solids, a number of processes occurred in constructed wetlands. Kadlec and Knight (1996) identified microbial respiration of carbon components to CO₂ in the aerobic zones of wetlands to be a major process of the carbon cycle. In anaerobic zones, nitrate, iron, and sulfate reduction and methanogenesis occurred under evaporation of CO₂ or CH₄ (methanogenesis), respectively. In contrast to constructed wetlands for purification of municipal effluents with low-hydraulic loadings (10–100 mm day⁻¹), in the present study aerobic conditions were found in the effluents of all wetlands (3.5–5.6 mg O₂ L⁻¹). Therefore, aerobic microbial decomposition of organic matter was promoted. In agreement with Kickuth (1981), presented treatment efficiencies for COD of 30–31% showed no significant differences between the different HRTs. Because of higher effluent interactions with microbial soil flora, treatment of catfish effluents in subsurface constructed wetlands at HRT of 1–4 days reduced 54–67% of organic matter (Schwartz and Boyd, 1995). Additionally, higher internal growth and build-up of biomass in constructed wetlands with free water surface decreases COD removal in comparison with COD reduction in subsurface systems (Gersberg et al., 1986; Bavor et al., 1988). In agreement with previous studies of Burgoon et al. (1991a) and Tanner et al. (1995), areal removal of organic matter rises in relation to increasing areal loadings. As reported by Kadlec et al. (1997), constructed wetlands with free water surface performing at low HRT with high COD areal loading are able to reduce adequate amounts of COD (up to 20.5 g m⁻² day⁻¹).

The reduction mechanisms for phosphorus in wetlands with free water surface comprise sedimentation, precipitation and adsorption reactions through ligand exchange of inorganic dissolved phosphorous in aluminum-, iron-, manganese hydroxides/oxides, calcium and clay minerals, and biochemical nutrient removal by plants (Bahlo and Wach, 1993). The removal of phosphorus in constructed wetlands generally involves rapid adsorption processes and slower chemical reactions, leading to the formation of a solid phosphate phase. Sorption at iron and aluminum compounds of the soil seems to be the most important process under presented neutral pH conditions (Richardson, 1985). Protonation of aluminum and iron surfaces increased with decreasing pH and phosphate-binding capacity declined (Bergheiser et al., 1980). Generally a lowering of the redox-potential increases the soil solution phosphorus content because of the reduction of Fe³⁺ to Fe²⁺, resulting in a release of iron-bound phosphorus (Gumbricht, 1993). Apart from oxygen content, nitrate was found to increase the redox-potential and hinder phosphorus release from the sediments (Gumbricht, 1993). Therefore, plants play a crucial role in constructed wetlands, because the aerenchymatous oxygen supply of rhizomes oxidizes the soil body, thus promoting decomposition of organic material, further aerobic microbial degradation of nutrients and phosphorus-binding capacity in the soil. Nutrient removal in root zone treatment systems by plant assimilation and uptake is reported to have minor importance, accounting for 10–30% of TN and TP retention (Burgoon et al., 1991b; Gumbricht, 1993; Koottatep and Polprasert, 1997). Nevertheless, the macrophytes can serve as windbreakers near the water surface, thus reducing resuspension of settled material (Brix, 1997). The increasing plant coverage in the first month after flooding seemed to influence removal in each wetland, especially for TSS and COD. A pilot-scale aquaculture effluent treatment system consisting of a free water surface and a subsurface flow constructed wetland also showed increasing removal performances with growing plant coverage in the first months of operation (Lin et al., 2002). Therefore, to avoid low removal performances of constructed wetlands in the start-up phase, establishment of a high percentage of plant coverage should be a goal.

Nevertheless, TP amounts were decreased (not significantly) with declining HRT in the wetlands consistently at levels of 41–53%. Sansanayuth et al. (1996) treated shrimp farm effluents in subsurface constructed wetlands at HRT of 1–3 days, with TP removals of 44–77%. Higher soil effluent contact in subsurface wetlands promotes phosphorus sorption processes, resulting in higher TP removal (Summerfelt et al., 1999) as is reported in this study. Inclining areal TP loadings (0.23–0.70 g m⁻² day⁻¹) increased areal TP removals of 0.12–0.30 g m⁻² day⁻¹, comparable with 0.05–0.22 g TP m⁻² day⁻¹ observed in treatment of dairy wastewater in subsurface constructed wetlands (Burgoon et al., 1991b), although performing with short HRT constructed wetlands with free water surface in this study showed sufficient TP removals.

Removal of nitrogen in this study was highest (31%) at HRT of 11 h and lowest (19%) at HRT of 3.5 h, indicating that HRT directly influences removal processes. Nitrogen elimination in wetlands begins with microbial ammonification of organic bound nitrogen, which can be either aerobic or anaerobic (Hiley, 1995). Nitrification or oxidation of ammonia (ammonificated and excreted ammonia) to nitrate, as an oxygen-demanding process, occurs in two steps involving microbial species such as Nitrosomonas and Nitrobacter (Schlegel, 1985). The decrease in quantity of inflowing NH-N can be attributed to high microbial ammonification and nitrification within each wetland, promoted by mean oxygen levels of 3.5–5.6 mg L⁻¹. Due to the release of H⁺ during microbial nitrification processes, reduction of pH after wetland passage refers to high nitrification activities. Significantly reduced removals of TN with decreased HRT of investigated wetlands showed that denitrification activity, as a major nitrogen retention process, varied among the three wetlands. Even if different denitrifiers have various demands, several reports indicated that denitrification begins at a threshold level of 0.25 mg O₂ L⁻¹, with increasing activity as oxygen content declines (Chan and Cambell, 1980; Rönner and Sörensson, 1985; Terai et al., 1987). Decreasing HRT supported the development of aerobic conditions in the investigated wetlands and hindered denitrification processes. Highest HRT with lower oxygen contents promoted denitrification and highest NO-N elimination. TN areal removal of 0.45–0.82 g m⁻² day⁻¹ was found to increase with rising loadings (1.46–4.37 g m⁻² day⁻¹). According to Burgoon et al. (1991b), the declining (%) removal at lower HRT decreases wetlands efficiency for TN reduction. Nevertheless, due to the short HRT in the constructed wetlands of this study, TN removal was generally low. Higher TN removal could be expected when performing with HRT of minimum 7 days, but leading to an inapplicable area demand on the constructed wetlands (Bahlo and Wach, 1993).

It can be concluded that constructed wetlands with free water surface represent a potential technology for effective effluent treatment of aquaculture flow-through systems. Compared with common treatment facilities of flow-through systems, such as microsieves or settling basins, the removal performance of free water surface systems was similar or even higher. The high effluent volumes of flow-through systems imply using short HRTs of 3.5 h, with highest nutrient areal removals.