Rainbow trout (Oncorhynchus mykiss) is one of the most important aquaculture species in Colombia’s inland waters, with a production of 4,617 tons in 2008 (1). Trout requires the greatest amount of water, since producing 1 kg in raceways regularly requires 210.000 L while only 21,000 L are required to produce tilapia (2). To optimize water, it can be reused in culture units arranged in series where the water flows by the force of gravity. However, water quality deteriorates as it moves from one unit to the next, and the disolved and particulate residues progressively accumulate in the water (3). Moreover, partial reuse systems require the incorporation of a certain amount of fresh water to control ammonium accumulation (4).

The increasing scarcity of water resources as well as concerns over environmental pollution management and food security are some of the reasons why recirculation aquaculture is becoming more and more popular (5). Recirculation aquaculture systems (RAS) allow intensive production of high quality fish in close proximity to large markets while reducing the water footprint and environmental concerns. Ideally, RAS are located close to urban areas with markets that have a high demand for sustainably produced protein (2, 6).

RAS use biofilters that treat contaminated water and reduce the amount of liquid required as well as discharges from aquaculture operations (7). In aquaculture production, relative water consumption is defined as the volume of water taken related to the production of biomass or feed used, so that the feedstock, expressed as kg of feed per m3 of water, reflects the intensity of recirculation and affects waste accumulation in untreated systems (8) and is directly related to the biofilter’s ability to nitrify (9).

In aquaculture there is a tendency to use fixed biofilm reactors, preferably those in suspension (10). Within these reactors are fixed granular support filters. Under aerobic conditions, biological filtration includes the autotrophic removal of ammonium and nitrite, as well as the heterotrophic degradation of dissolved and particulate organic matter (11). Dissolved N from the fish is excreted mainly as urea and ammonia, and the ammonium is nitrified to nitrite as an intermediary. In anoxic denitrification, facultative heterotrophic bacteria reduce nitrate and nitrite to gaseous N by capturing energy and electrons from MO (12). NH3 (ammonia) and NH4+ (ammonium ion), the combination of which is called total ammonia nitrogen (TAN), can be toxic to fish at levels that vary according to the species, fish size, the presence of fine solids, metals and nitrate (13). Nitrification as the oxidation of ammonium ion and nitrite occurs in the bacterial film of the biofilter (14). Crab et al (13) state that nitrification is affected by parameters such as substrate and dissolved oxygen concentrations, MO, temperature, pH, alkalinity, salinity and turbulence level.

Biofilters are defined as highly organized biofilm structures that develop due to segregation of members of individual communities in different layers that change according to nutritional conditions (14). Oxidative reactions in nitrification are catalyzed by two groups of microorganisms called ammonium oxidants and nitrite oxidants. However, the mechanisms of this biological process in biofilms are still not fully understood (15).

The aim of this study was to evaluate the performance of RAS in rainbow trout fingerling cultivation to remove N and biofilter solids in fixed support and upflow in which different inoculum were used.

Location. The investigation was done at the Recirculation Laboratory of the Aquaculture Production Engineering Program (APEP) of the Universidad de Nariño−Pasto: Latitude 01° 09’ N, Longitude 77° 08’ W, Altitude 2540 meters above sea level.

Components and operation of the recirculation system. The RAS consisted of a fiberglass culture tank − CT − 0.991 m3, 1.28 m diameter and a height of 0.87 m. The tank had a central drain, external overflow in 1-1/2” PVC piping coupled to a level control with two perforated plates for retention of solids. The effluent passed through a canvas bag filter, the filtered liquid accumulated in a 100 L suction tank that a 1 HP pump sent to a reservoir and then to a constant level box with 0.20m hydraulic loading that distributed the water by gravity to the 6 upflow biofilters (reactors) witha a ½” transparent hose. The reactors were made with 3” PVC sanitary pipe and a total height of 1.0 m. The influent entered the biofilters through an acrylic base perforated with 115 3.57 mm holes and the flow rate was controlled with ball-type valves. The treated effluent left laterally through ½” flexible tubes and hosing located 0.95 m from the base and was collected in a PVC channel that returned it to the CT.

The acrylic false bottom of the biofilters was 0.10 m from the base, the sheet held up sublayer 1, followed by sublayer 2, each with a height of 0.05 m. There was 0.60 m of sand on undercoat 2, and on the static level there was a hydraulic load of 0.15 m and 0.05 m free edge on the output of the effluent.

In the CT, air entered from the blower through a PVC pipe, flexible hose and air stone.

The diagram in figure 1 shows the elements of the evaluated recirculation system.

Figure 1 Figure 1 Frontal view of RAS and components. Source: This investigation

Transporting and acclimatizing of the trout . The fish were moved from the cages floating at the ″Inti Yaco” station on Lake Guamués to the laboratory after a 24-hour fast. The fish were packed in groups of 10 in caliber 3 plastic bags with 1/3 water and 2/3 oxygen. In order to homogenize the physicochemical parameters of the transport water with respect to the laboratory water, the bags were introduced into a 1.0 m3 auxiliary tank for a period of 15 min. As a prophylactic treatment, 3 g of NaCl per liter of water was added to each bag. Acclimatized animals had an adjustment period of 7 days, with food supplied to satiety.

Sampling and animal feed. An initial census was performed of 34 specimens of rainbow trout, tranquilized with eugenol to facilitate handling, record weight, and to establish daily feed ration and calculate feed conversion (CA). Feed in a ratio of 3% of the biomass was distributed in three meals. During the experiment five additional samples were taken. The initial system load was 0.83 kg/ m3, the average weight of the animals was 24.11 ± 5.9 g, with minimum and maximum values of 14.4 and 35.7 g. Periodic weight increase was expressed as the average individual weight gain of the population over a period of time.

Monitoring of water quality parameters . In the central area of the constant level box and in the effluent of each reactor composite samples were taken every morning for physical-chemical analysis according to protocols established by APHA-AWWA-WPCF (16). The parameters monitored were nitrite (NO2), nitrate (NO3), ammonium (NH4+), suspended solids (SS), carbon dioxide (CO2), alkalinity, temperature, dissolved oxygen (DO) and the potential of hydrogen ions (pH). Based on the results obtained, removal efficiency of nitrogenous matter and SS were calculated.

Biofilters and growing media. Support media was inoculated as follows: R1-Control, RAS water, sterile distilled water; R2- sterilized sludge from the Guayrapungo station, filtered and sterilized water from the Guayrapungo station; R3-aerated surface lagoon water from the Antanas landfill (AL), culture medium according to Shi et al (17); R4-APEP aquarium sediments, filtered aquarium water; R5- aerated lagoon sludge RAS, culture medium according to Molinuevo et al (18); R6- sludge from sulfidogenic reactor (RAS), sterile distilled water - leachate (1: 1) as modified by Thabet et al (19). The activation period was 2 weeks before starting the system.

Characterization of granular materials in the reactors . The filter consisted of three layers of granular material: torpedo layer one material passed through a 3/8” sieve and held sieve #4; torpedo layer 2 through sieve #4 and retained sieve #8; sand layer through sieve #20 and retained by #40, located in this order in the direction of upflow.

Materials were characterized according to procedures established by the Colombian Institute of Technical Standards ICONTEC and the Colombian Technical Standards -NTC- according to these tests: absorption and density of coarse (NTC 176) and fine (NTC 237) aggregates; sieved for fine and coarse aggregates (NTC 77); determination by washing the material in a 75μ sieve (NTC 78); determination by drying aggregate moisture (NTC 1776).

Coefficient uniformity (CU=D60/D10) and curvature (CC=D302/(D60*D10)) of the composite material (20) were calculated.

RAS flow and hydraulic steering . The hydraulic retention time (HRT) of the biofilter was 11 minutes, and to silt the granular base it was necessary to open the flow control valves to restore the desired HRT.

Through the 3” reactors with 0.80 m liquid column and a useful volume of 3.53 L, 21.2 L circulated every 11 min. Since the CT volume was 991 L, it underwent a replacement of its total volume every 8.58 h, longer than the range between 0.5 and 1.67 h recommended by Lekang (21) and well above the 0.23 h adopted by Good et al (22).

The clogging of the biofilter made it necessary to wash it by interrupting the filter and removing and washing the granular layers to remove any retained solids. Then the sand was separated in a sieve and was returned to the biofilters.

Experimental design and statistical analysis . For daily monitored parameters (pH, CO2, alkalinity, temperature and OD), the Kruskal-Wallis nonparametric test was chosen. For parameters monitored twice a week (ammonium, nitrite, nitrate and SS), an experimental block design with sub-sampling was used because it was expected to need to wash the layers due to clogged filters, considering the time between washing as a blocking factor. The level of significance was considered at α=0.05, so that values of p <0.05 indicated significant differences. The SAS statistical package version 9.0 was used.

Particle size characteristics of compound granular media. With the granulometric curve of the composite material (Figure 2), the following values were determined: D60=0.70mm; D10=0.45mm and D30=0.52mm. Based on these, the CC and CU coefficients were 0.858 and 1.55, respectively.

Figure 2 Figure 2 Granulometric curve for composed granular means

Physical properties of granular media . The values of natural moisture, absorption, and density of materials used in the biofiltration units are presented in table 1.

Table 1 Table 1 Physical properties of granular material used

Periodic weight increase and feed conversion. The initial biomass loading was 0.83 kg/m3 and at the end 4.23 kg/m3. An average daily of individual weight increase during the test was 1.58 g/d and an average FC value of 1.41. Figure 3 shows the values of the CT bioburden during the experiment.

Figure 3 Figure 3 Growth curve for cultivated biological load

Daily measurement of water quality parameters . Table 2 presents the mean values as well as standard deviations calculated for pH, alkalinity (Alk.), CO2, temperature (Temp.) and OD, both in the constant level box that corresponded to the crude entrance or influent (CE) and each reactor’s output, where the letter R with the respective number refer to the reactor evaluated.

Table 2 Table 2 Mean values and standard deviation of the parameters measured daily

pH . Highly significant differences between medians corresponding to reactors 2, 3, 4, 5 and 6 with respect to the raw entrance or influent were found. Meanwhile, R1 showed no difference with respect to input. The average pH value recorded in the reactors was 5.1 and the minimum and maximum values ​​were 4.3 and 7.0

Alkalinity and CO2 . For alkalinity there were significant differences in reactors 2, 3, 4 and 6 with respect to entrance but not between them. Alkalinity values ​​were generally low, with an average in the reactors of 2.9 mg/L of CaCO3, showing peaks after partial replacements. The average CO2 concentration recorded was 16 mg/L.

Temperature. No significant differences were found in the recorded values during the investigation.

Dissolved oxygen . There were no differences between the reactors, but they all differ with respect to the input whose OD concentration ranged from 3.8 to 4.8 mg/L. In reactor effluent an average range of values between 1.0 and 2.0 mg/L with an average concentration of 1.6 mg/L was recorded.

Water quality parameters measured 2 times per week. Table 3 shows variance analysis values corresponding to the effects of the reactors and their respective P value.

The results for the parameters listed in table 3 are presented below, supported by a series of diagrams of boxes in which the letter E refers to the corresponding input or raw influent data, the letter R with its respective number to the reactor reported and the letter B next to the number associated with it indicates the block to which the data belongs.

Table 3 Table 3 ANOVA for parameters taken twice a week

Ammonium. There were no significant differences (Table 3). Figure 4 illustrates the concentration of ammonium measured for the two blocks established by time

Figure 4 Figure 4 Diagram of box and whisker plots for ammonium concentrations.

Nitrites. No significant differences (Table 3) were found. Figure 5 shows the values registered for this parameter during the study.

Figure 5 Figure 5 Diagram of box and whisker plots nitrite concentrations

Nitrates . For this parameter significant differences between reactors (Table 3) were found. Based on the multiple range test four groups were established: biofilter R5; biofilter R6; biofilters R2, R1 and R4, with equal means, and biofilter R3. Figure 6 illustrates nitrate concentrations recorded for the two blocks.

Figure 6 Figure 6 Diagram of box and whisker plots for nitrate concentrations.

Suspended solids . Significant difference for this parameter (Table 3) were found. From the multiple range test the formation of two groups of biofilters were noticed: the one formed by biofilters R3, R2, R5, and R6; and the one consisting biofilters R2, R5, R6, R4 and R1. Figure 7 illustrates the SS concentrations recorded for the two blocks.

Figure 7 Figure 7 Diagram of box and whisker plots for SS concentrations

Uniformity properties of granular material. From the CC and CU values it can be concluded that although the sand showed high uniformity since sieves #20 and #40 were used, the composite presented a lack of uniformity due to the presence of a variety of sizes and a CC close to the unit.

Periodic weight increase . The average weight increase in this investigation for the phase studied was 1.58 g/d, which shows that the water treatment system guaranteed water quality conditions suitable for cultivating trout. This value was higher than that reported by Mocanu et al (23) who for the same phase of initial and final growth and biomass densities of 2.64 and 6.73 kg/m3, respectively, obtained an average individual weight gain of 1.49 g/d. However, the specific growth rate of this study was lower than that reported by d’Orbcastel et al (24) for trout in recirculation systems but in a more advanced phase.

FCR. FC obtained in this study was 1.41, lower than typical values reported by van Rijn (25) of 0.8 to 1.1, and less than the average conversion recorded by Mocanu et al (23), which under similar conditions of biomass density, growth phase and feed rate was 0.83. This indicates that although in this investigation a greater weight gain was recorded per time unit, high FC values show that there was low efficiency of feed utilization (26).

pH . Since the R1 was a control, it was colonized primarily with microorganisms from the culture unit. The microbiological community that developed within the filter was similar to that already prevailing within the tank itself, which can explain why the pH in the effluent of the biofilter was very similar values to what was recorded in the effluent.

Generally, the reactors tended to acidify over time; however, the effluents did not affect trout survival, since mortality rate was 11.76%, which was within the ranges for the species and cultivated phase. The low pH values meant that the non-ionized fraction of ammonium remained in low ranges (2).

In spite of the reactors’ predominantly acid medium, bacterial populations were able to carry out their removal functions in the measured pH, although several authors recommend an optimum range of 7.2 to 7.8 for reactors inoculated with bacteria (13, 27).

Alkalinity and CO2 . According to Timmons and Ebeling (2), alkalinity in water is linked directly to the pH of the system and CO2 concentrations. Since in this experiment a normal range of CO2 was maintained for prolonged cultivation and low pH values, relatively low levels of alkalinity were found. These were also associated with microbial processes, especially concerning nitrification, since for every gram of nitrified ammonium 4.57 g and 7.14 g of oxygen alkalinity (2) are required. The highest alkalinity values recorded in R1 and R5 reactors relative to those measured in the other treatment units could be due to incipient denitrification processes, since according to Tsukuda et al (6), denitrification produces alkalinity.

Temperature . Registered values were within the temperature ranges recommended by Magerhans and Hörstgen-Schwark (28) for cultivating rainbow trout in the evaluated phase, which directly influenced the physiological processes such as growth and feed efficiency.

Dissolved oxygen . Oxygen consumption due to the biological processes of wastewater treatment was evident in the effluent from each of the reactors. The biofilters need oxygen for the heterotrophic bacteria, since 4.57 g of oxygen are required for each gram of ammonia nitrogen oxidized to nitrate nitrogen. Timmons and Ebeling (2) stated that levels of 2 mg/L of OD in the effluent biofilter are sufficient to maintain a maximum rate of nitrification.

Ammonium . The similarity in ammonium removal efficiency could be due to the ripening time of inoculum in the biofilters, which was insufficient to demonstrate the effects of each one. After washing, removal decreased; authors like Suhr and Pedersen (29) state that the removal rates of ammonia nitrogen in biofilters are highly dependent on environmental and operational parameters such as flow distribution and backwash regimes. Average removal values ​​in reactors ranged from 2.8 to 7.9%, with an overall average of 4.7%. Removal is a main function of ammonia oxidizing bacteria under aerobic or anaerobic conditions. Malone and Pfeiffer (10) indicate that heterotrophic bacteria contribute to defining biofilm thickness.

Under anoxic conditions, the anaerobic oxidation of ammonium - anammox - allows the ammonium to be oxidized to nitrite, which produces molecular nitrogen and requires about 50% less oxygen compared with nitrification-denitrification processes (30). Anammox bacteria in RAS are probably simultaneously activated by nitrifying bacteria, which have been found in natural and artificial environments. This could mean that in aquaculture systems the BOA or ammonium oxidizing archaea (AOA) convert the ammonia excreted by fish under oxic conditions into nitrite. This nitrite and the additional ammonium is then used by anammox bacteria in anoxic reactor parts to form nitrogen gas, removing both nitrite as well as ammonium from the system without the need for denitrification (31).

Nitrites . The maturation time of inoculum in the biofilters could be a reason why there was similar behavior in the reactors. However, it was observed that nitrite removal was different in each block, which allows the conclusion that washing affected nitrite removal (Figure 5). Nitrite removal is attributed to nitrite oxidizing bacteria − BON (14), and according Malone and Pfeiffer (10) BON can control the nitrification rate under certain conditions. Some negative efficiency values were recorded, suggesting a temporary prevalence of nitrification in relation to removal of NH2+. Average removal efficiencies in the reactors ranged between 8.9 and 45.2%, reporting an overall average efficiency of 27.2%.

Nitrates . Each reactor had a different nitrate removal efficiency, which was affected by washing the reactors, mainly in R3 and R4. In both blocks R5 had the highest nitrate removal at 47.4 and 47.2% respectively, followed by R6 (44.9 and 40.7% respectively). R3 presented the lowest efficiency in the second block with a value of 1.8%.

The overall average nitrate removal efficiency was 32.3%. This can be attributed to denitrifying bacteria that reduce nitrate to dinitrogen gas, metalloenzymes nitrate reductase, nitrite reductase, nitric oxide reductase and nitrous oxide reductase (31).

Washing affected nitrate removal efficiency since it higher in Block 1, the function of denitrifying bacteria might have affected the second block since washing improved water circulation, which carries OD. This explanation can be applied to R3, which could have been more aggressively washed and caused a loss of biofilm where this community of bacteria was found.

Suspended solids . There were significant SS removal differences between reactors (Table 3). The results obtained with the LSD test indicate that R1 and R4 reactors behaved similarly but differ from R3 (Figure 7). SS were efficiently removed, retained by the filters and due to MO decomposition due to microorganisms. Reactor R3 had the best performance with an average efficiency of 58.2%. Overall, removal efficiencies in the blocks ranged between 1.9 and 81.9% with average total removal of 37.5%. These values are within the ranges reported for both trickling filters (32) and submerged aerated biofilters (33).

The filters of the evaluated biofilters clogged often, since plugging interstices altered hydraulic operation, mainly in TRH control. Filter washing is a valid solid removal mechanism that clogs the reactor; however, it is a disturbing factor because it affects the performance efficiency of water treatment from a biological point of view.

There were no differences between the reactors concerning the removal of ammonia and nitrites; washing the biofilter affected the performance of treatments.

Statistically significant differences in nitrate removal were presented, with reactors R5 and R6 showing a better performance, with average efficiency exceeding 42%.

Finally, it can be concluded that the best reactor in terms of removal of various forms of nitrogen was the reactor inoculated with sludge from the aerated Antanas Landfill lagoon.

Acknowledgments

To the Department of Hydrobiological Resources of the Universidad de Nariño, Empresa Municipal de Aseo de Pasto EMAS, Professor Marco Antonio Imuez Figueroa, and students of the Aquaculture Production Engineering Program: María Fernanda Montenegro, Jessica Hernández and Danilo Duarte.