Evaluation of a Commercial-Scale Aquaponic Unit for the Production of Tilapia and Lettuce

James E. Rakocy, Donald S. Bailey, Kurt A. Shultz and William M. Cole

University of the Virgin Islands, Agricultural Experiment Station St. Croix, U.S. Virgin Islands

Abstract An outdoor, commercial-scale aquaponic unit for the production of red tilapia and leaf lettuce was in continuous operation at the University of the Virgin Islands for 2.5 years. The unit consisted of four fish 3 3 rearing tanks (4.4 m each, water volume), two cylindro-conical clarifiers (1.8 m each), four rectangular filter tanks (0.7 m3 each) containing orchard netting, six hydro- ponic tanks (11.5 m3 each) and a sump (0.6 m3). The hydroponic tanks were 29.6 m long by 1.3 m wide by 0.4 m deep and had a combined surface area of 214 m2. Total water volume was 91.6 m3. A 1/2-hp in-line pump moved water at an average rate of 378 L/min from the sump to the rearing tanks (mean retention time, 0.8 h), from which effluent flowed by gravity through the solids-removal and hydroponic subsystems. The fish rearing tanks and hydroponic tanks were aerated by air diffusers. A 1/20-hp vertical lift pump was used to supply additional aeration to the fish rearing tanks in the last 12 weeks of each production cycle. Production of red tilapia was staggered so that one tank of fish was harvested every 6 weeks. The final stocking rate was 182 male fingerlings/m3. The fish were fed for 24 weeks with a nutritionallycomplete, floating ration (32% protein) delivered by belt feeders. In the last 11 out of 19 harvests, total harvest weight, net gain, mean weight, growth rate (male fish), survival and feed conversion ratio averaged 81.1 kg/m3, 73.2 kg/m3, 487.2 g, 2.85 g/day, 91.6%, and 1.76, respectively. Mortality in the last 11 harvests resulted primarily from minor episodes of an unidentified bacterial disease.

Lettuce plants were grown in net pots supported by floating polystyrene sheets. Production was staggered so that one fourth of the lettuce in the unit was harvested every week. Three-week old transplants grew to marketable size in 4 weeks. Five varieties (Sierra, Nevada, Montello, Jerhico and Parris Island) and two planting densities (16 and 20 plants/m2) were used. In 112 harvests, marketable production averaged 27 cases per week and ranged from 13 to 38 cases (24-30 heads/cs). Losses occurred as a result of caterpillar and aphid damage, wind damage, tip burn, root damage (caused by zooplankton) and root disease (pythium). Production was interrupted for 9 weeks by two hurricanes and 3 weeks by transplant loss in the greenhouse. The hydroponic tanks maintained good water quality through direct ammonia uptake by lettuce roots and nitrification on the tank surface area. Total ammonia-nitrogen and nitrite-nitrogen values averaged 1.47 and 0.52 mg/L, respectively. Total water consumption and average daily makeup water were 1,444 m3 and 1.5% of system volume. The aquaponic unit was simple to operate, reliable and produced commercial levels of tilapia and lettuce on a continuous basis. Introduction Aquaponics is a new technology that involves the combined culture of fish and hydroponic vegetables, herbs and other plants in recirculating systems. Linking fish and vegetable culture offers several advantages over separate production systems. A large portion of the waste nutrients generated by the fish are recovered by the vegetables rather than being discharged to the environment. A secondary crop of vegetables improves the systems profit potential. The vegetable component receives most of the required nutrients at no cost, purifies the culture water, extends water use and eliminates the need for separate and expensive biofilters. Additional savings are realized by sharing operating and infrastructural costs. During the last 20 years, a large number of experiments, employing mainly small bench-top systems, have demonstrated the technical feasibility of aquaponics, but the commercial feasibility has not been addressed (Rakocy and Hargreaves, 1993). Commercial evaluation requires that a production unit be relatively large in size and operated over a long time period to demonstrate its sustainability. The objective of this study was to evaluate the long-term performance of a commercial-scale aquaponic unit for the production of tilapia and lettuce and to modify the unit as needed to improve its performance. Materials and Methods A commercial-scale, aquaponic unit was constructed on the St. Croix campus of the University of the Virgin Islands in 1994. The unit was placed in operation from January 26, 1995 through June 30, 1997. The unit consisted of four fish rearing tanks (4.4 m3 each, water volume), two cylindro-conical clarifiers (1.8 m3 each), four filter tanks (0.7 m3 each) containing orchard netting, six hydroponic tanks 2

(11.5 m3 each) and a sump (0.6 m3) (Figure 1). The hydroponic tanks were 29.6 m long by 1.3 m wide by 0.4 m deep and had a combined surface area of 214 m2. Total water volume was 91.6 m3. Makup water was added by a float valve in the sump. A water meter recorded the daily additions. Rainwater was the sole source of water used in the unit. A 1/5-hp in-line pump was used initially to move water at an average rate of 163 L/min from the sump to the fish rearing tanks (mean retention time, 1.8 h), from which effluent flowed by gravity through the solids-removal and hydroponic subsystems and returned to the sump. The pump was replaced on December 18, 1996 with a 1/2-hp in-line pump, which moved water at an average rate of 378 L/min from the sump to the rearing tanks (mean retention time, 0.8 h). Effluent from the filter tanks was divided into three streams, each of which flowed through a set of two hydroponic tanks and returned to the sump. Each fish rearing and hydroponic tank was aerated by 10 (22.9 cm x 3.8 cm x 3.8 cm) and 24 (7.6 cm x 2.5 cm x 2.5 cm) air diffusers, respectively. A 1/20-hp vertical lift pump was used to supply additional aeration to the rearing tank in the last 12 weeks of the production cycle. Production of red tilapia was staggered so that one fish tank was harvested every 6 weeks. An initial stocking rate of 227 fish/m3 was reduced to 182 fish/m3 beginning with those fish harvested on March 18, 1996 (Table 1). Mixed-sex fingerlings were used in the first four batches of fish. The next 13 batches used male fingerlings that were manually selected. The final two batches used male fingerlings that were produced by sex reversal with 17-methyltestosterone according to INAD (Investigations in New Animal Drugs) protocol. The mean weight of fingerlings per batch was 38.9 g and ranged from 25.9 to 106.1 g. Each batch of fish was fed for 24 weeks with a nutritionally-complete, floating ration (32% protein). The feed was delivered by demand feeders through the harvest on September 3, 1996 and thereafter by 12-hour belt feeders. An initial feeding rate of 6% of body weight per day was gradually decreased to 1.2% by the time of harvest. The fish were kept slightly hungry to avoid feed wastage. Upon harvest, the fish were sorted by sex, counted and weighed, and the rearing tank was immediately restocked with fingerlings. In June, 1995, 400 highfin black tetras (Gymnocorymbus sp.) and 400 tuxedo swordtails (Xiphophorus sp.) were stocked into separate hydroponic tanks. In July, 1995, 100 fantail guppies (Lebistes reticulatus) were stocked into two hydroponic tanks. In June, 1996, the swordtails were removed. In July, 1996, 500 highfin black tetras and 1,000 zebra danios (Brachydanio rerio) were stocked into separate hydroponic tanks. Five varieties of lettuce were cultured during the trial. They included red leaf (Sierra), green leaf (Nevada), romaine (Parris Island and Jerhico) and crisphead (Montello). In general, three varieties (red and green loose leaf and romaine) were cultured simultaneously. Production of lettuce was staggered so that one fourth of the lettuce being cultured was harvested every week. Lettuce transplants were produced from seed in a greenhouse in flats containing peat-based growing media. After 3 weeks in the 3

greenhouse, the transplants were transferred to the aquaponic unit for a four-week growout period. The transplants were placed in net pots (5 cm-diameter and height) that were inserted into holes in floating polystyrene sheets at a density of either 16 or 20 plants/m2. The sheets were 2.44 m long by 1.22 m wide by 3.8 cm thick. Each hydroponic tank contained 12 sheets. The plants were sprayed twice weekly with Bacillus thuriniensis, a bacterial pathogen that controls caterpillars. Every Wednesday 18 sheets of lettuce were harvested at 0600 hours and the marketable plants were packed in cases at a rate of 24 heads/cs. If the lettuce was small, as many as 30 heads were packed in a case. Beginning October, 1996, the lettuce in two cases of each variety were weighed en masse to obtain a mean weight. The solids removal subsystem initially consisted of two clarifiers and two filter tanks containing orchard netting. The effluent from two rearing tanks flowed through one clarifier and one filter tank. Thirty male tilapia fingerlings were placed in each clarifier to graze settleable solids from the sides of the cone and concentrate them at the base. A drain line was opened twice a day to remove sludge from the clarifier. Fine particulate solids in the effluent from the clarifier adhered to the orchard netting mesh in the filter tank. Once a week the netting was washed with a pressure sprayer and the entire water volume in the filter tank was discharged. All solids were discharged through drain lines into two lined 16-m3 ponds, which were continuously aerated with diffused air. As one pond was being filled over a 2 to 4-week period, water from the other pond was used to irrigate and fertilize field crops. On July 27, 1995, two additional filter tanks were installed in series with the initial filter tanks. On May 30, 1996, a microscreen drum filter with 60 micron mesh was installed and the filter tanks were removed from the system. Effluent from the clarifiers flowed through the drum filter to the hydroponic tanks. On June 26, 1996, the clarifiers were removed so that effluent from the rearing tanks flowed through the drum filter to the hydroponic tanks. On July 26, 1996, the filter tanks were reinstalled so that effluent from the drum filter flowed through the filter tanks before entering the hydroponic tanks. On November 11, 1996, the drum filter was removed and the clarifiers were reinstalled. Total ammonia-nitrogen (TAN) and nitrite-nitrogen (NO2-N) were measured biweekly in the rearing tanks, sump and effluent from the clarifiers and second filter tanks using the indophenol and diazotizing methods, respectively (Boyd and Tucker, 1992). Total suspended solids (TSS) and turbidity were measured biweekly at the same locations using the methods described by Boyd and Tucker (1992) and Hach (1990), respectively. Dissolved oxygen (DO) was measured biweekly with a model 51B Yellow Springs Instruments polarographic DO meter at the same locations and at five locations in each set of two hydroponic tanks. Water temperature was measured biweekly in the rearing tanks and sump. pH, total alkalinity and chemical oxygen demand (COD) were measured biweekly in the sump and the effluent from one clarifier and filter tank using a model 340 Corning pH meter, titration with bromcresol green indicator (American Public Health Association et al., 1995) and the heat of dilution technique (Boyd and Tucker, 1992), respectively. pH was also measured several times weekly in the sump. Conductivity (as total dissolved solids, TDS) was measured biweekly in the sump and the effluent from the filter tanks with a model DP4 Myron L Company DS meter. Nitrate-nitrogen (NO3-N) and dissolved orthophosphate (PO4-P) were measured biweekly in the sump using the phenoldisulfonic acid 4

method (Boyd, 1979) and the ascorbic acid method (Boyd and Tucker, 1992), respectively. Sulfate (SO4-S) and chloride (Cl) were measured monthly in the sump using the turbidimetric (American Public Health Association et al., 1995) and mercuric nitrate (Boyd and Tucker, 1992) methods, respectively. Calcium (Ca), magnesium (Mg), potassium (K), sodium (Na), iron (Fe), manganese (Mn), copper (Cu), zinc (Zn), boron (B) and molybdenum (Mo) were m easured monthly in the sump using the inductively coupled plasma (ICP) method (American Public Health Association et al., 1995). pH was maintained in the range of 6.4 to 7.6 by frequently adding quantities (generally ranging from 300 to 1,000 g per addition) of potassium hydroxide (KOH), calcium oxide (CaO), or calcium hydroxide [Ca(OH)2] to the unit. Calcium oxide was replaced with calcium hydroxide beginning August 22, 1995. Potassium and calcium bases were added alternately in equal amounts. Iron chelate (10% Fe) was added to the unit every 3 to 4 weeks at a rate of 1,360 g through July 17, 1995 and thereafter at a rate of 1,832 g. There was an addition of 2,500 g of dipotassium phosphate (K 2HPO4) on August 13, 1996 and 265.0 g of boric acid (H3BO3), 161.0 g of manga- nese sulfate (MnSO44H2O), 10.8 g of copper sulfate (CuSO45H2O) and 9.0 g of ammonium mo- lybdate [(NH4)6Mo7O24] on August 29, 1996. The entire water volume of the unit was exchanged once on August 12-13, 1996. Results and Discussion Fish Production The initial stocking rate of 227 fish/m3 and the use of a mixed-sex population did not produce sufficient numbers of marketable-sized fish (454 g) (Table 1). Female red tilapia were only half as large as males at harvest and the mean weight during the first four harvests was 314 g/fish. A reduction in the stocking rate to 182 fish/m3, starting with the harvest on March 18, 1996, and the use of male populations significantly increased the harvest weight, which averaged 487 g/fish in the last 11 harvests. Manual selection of males fingerlings produced variable and unsatisfactory results. Based on the sex distribution at harvest, the average number of male fish in the manually-sexed populations was 87.2% (range = 63.3 to 98%). It was difficult to distinguish sexual characteristics in this strain of red tilapia, which was obtained from Sunfish Hatcheries in Jamaica in 1986 (Rakocy et al. 1993a). The use of sex-reversal in the last two batches produced excellent results (mean = 99.2% males). Total harvest weight and net gain in the last 11 harvests averaged 81.1 kg/m3 (range = 61.0 to 91.8 kg/m3) and 73.2 kg/m3 (range = 53.8 to 81.4 kg/m3), respectively. Production was substantially greater than the general production range (30 to 40 kg/m3) for tilapia using diffused aeration (Losordo, 1997) due to the high aeration rates in relatively small water volumes and the high quality of incoming water. Annual production, based on the last 11 harvests, was 3,096 kg. The average growth rate during the final 11 harvests was 2.85 g/day (range = 2.55 to 3.29 g/day), which compares favorably with a range of results reported by Losordo (1997).

Survival was low for the harvests on October 2, 1995 through March 18, 1996 due to Hurricane Marilyn. The system was without power for 12 hours, during which time mortality of more than 40% occurred in the two tanks with the largest fish. Additional mortality occurred due to periodic generator failure in the three months after the hurricane. Survival in the last 11 harvests averaged 91.6% (range = 78.6 to 99.2%). The primary cause of mortality was an unidentified bacterial disease. During disease episodes, fish died at a rate of one or two per day while the majority of the fish in the tank fed vigorously and was not affected. Often fish died in just one of the four rearing tanks, which indicates that the disease was transmitted by direct contact. Mortality generally occurred during the warmer months (June - November) when water temperatures were above 280C. Table 1. Production characteristics of red tilapia over 2.5 years in a commercial-scale aquaponic system. _____________________________________________________________________________ Stocking Harvest Net Final Growth a Rate Weight Gain Weight Males Rateb Survival 3 3 3 Date (#/m ) (kg/m ) (kg/m ) (g/fish) (%) (g/day ) (%) FCRc _____________________________________________________________________________ 05/24/95 227 74.1 50.0 341 57.8 2.82 95.7 1.92 07/10/95 227 75.6 63.7 340 52.7 2.27 97.8 1.92 08/21/95 227 49.6 42.0 326 51.4 2.40 66.8 2.24 10/02/95 227 30.1 22.4 248 39.1 1.97 53.5 4.35 11/13/95 227 49.5 42.7 364 96.0 2.02 59.8 2.54 12/26/95 227 46.0 37.8 291 63.3 1.78 69.5 2.68 02/05/96 227 45.0 39.1 436 94.0 2.52 45.4 2.05 03/18/96 182 44.0 36.2 417 95.7 2.34 58.1 2.05 04/30/96 182 88.6 81.4 551 98.0 3.12 88.4 1.71 06/10/96 182 81.2 71.4 460 96.1 2.55 97.0 1.73 07/23/96 182 82.8 76.3 534 83.0 3.29 85.4 1.80 09/03/96 182 75.3 66.7 453 86.2 2.66 91.4 1.76 10/14/96 182 78.0 71.2 478 80.8 2.99 89.8 1.76 11/26/96 182 82.7 76.2 482 68.7 3.19 94.2 1.65 01/07/97 182 61.0 53.8 427 88.7 2.51 78.6 1.94 02/18/97 182 80.3 72.9 483 90.7 2.82 91.5 1.73 04/01/97 182 91.8 79.6 534 92.2 2.96 94.5 1.81 05/13/97 182 90.0 81.3 498 99.1 2.75 99.2 1.75 06/24/97 182 81.5 74.3 459 99.2 2.55 97.5 1.75 _____________________________________________________________________________ a Rearing tank water volume was 4.4 m3. b Male fish. c Feed conversion ratio.

Based on data from the last 11 harvests, the average feed conversion ratio was 1.75 with demand feeders (four harvests) and 1.77 with manual feeders (seven harvests). Demand feeders were discontinued because they were less reliable. Often clumps of feed would partially block the flow of feed or strong winds would trigger the release of too much feed. Belt feeders provided a steady release of feed over 12 hours and any failure, which seldom occurred, was quickly detected. With belt feeders, it was also easier to assess feeding response. With the initial pump (1/5 hp), the flow rate (163 L/min) and the water exchange rate (0.55 times/h) in the rearing tank appeared to be too low to achieve optimum production characteristics. However, after a new pump (1/2 hp) was installed on December 18, 1996, a higher flow rate (378 L/min) and exchange rate (1.25 times/h) did not improve total weight (81.4 kg/m3 before vs. 80.9 kg/m3 after), net gain (73.9 kg/m3 vs. 72.4 kg/m3), growth rate (2.97 g/day vs. 2.72 g/day), survival (91% vs 92.3%) or feed conversion ratio (1.74 vs. 1.80), based on the last 11 harvests. A reduction in density may be necessary to attain a substantial improvement in production characteristics. Lettuce Production Total annual lettuce production averaged 1,248 cases. Using improved production levels during the last six months (Figure 2), projected annual production would increase to 1,694 cases. In 112 lettuce harvests, marketable production averaged 27 cases per week and ranged from 13 to 38 cases. Production was interrupted for 9 weeks by two hurricanes. During the September through November period in 1995, Hurricane Marilyn (100 mph winds) stopped production for 8 weeks (Figure 2). One week of production was lost to Hurricane Hortense in September, 1996. Three weeks of production were lost in December, 1996 when rodents destroyed the lettuce transplants in the greenhouse. Of the 3,044 cases of lettuce harvested during the trial, there were 1,552 cases of Sierra, 1,030 cases of romaine (Parris Island and Jerhico), 410 cases of Nevada and 71 cases of Montello. Montello was discontinued on July 31, 1996 and replaced with Nevada. The average weekly harvest weight of marketable plants since October, 1996 was 269 g for Sierra (range = 182 to 340 g), 327 g for Parris Island (181 to 446 g), 314 g for Jerhico (267 to 344 g) and 265 g for Nevada (149 to 360 g). Production was greater in the cooler months (February-March, 1995; January-February, 1996; January-April, 1997) (Figure 2). During these periods, water temperature averaged 25.10C. Production of lettuce, which is a cool season crop, decreased during the warmer periods (AprilAugust, 1995; March-November, 1996; March, 1997) when water temperature averaged 27.50C and peak daytime canopy temperatures often reached 390C. High production in June, 1997 was due in part to the good performance of Nevada, a heat tolerant variety. During the April-August period of 1995, production also declined due to root damage caused by zooplankton blooms, comprised mainly of cladacera and ostracods. Plant growth was normal in areas near diffuser-induced currents, which dispelled the zooplankton, and was stunted in quiescent areas 7

where the zooplankton flourished. Most damage was caused by ostracods, which clung to roots as they ate detritus and root hairs. More than 23,000 ostracods were dislodged from the roots of just one medium-sized lettuce plant. With the introduction of ornament fish (tetras, swordtails, guppies, zebra danios), beginning in June of 1995, the problem gradually dissipated and the plants grew uniformly in size. As the swordtail population increased, they caused some minor root damage and were removed in June, 1996. Two species of pathogenic root fungi (Pythium myriotylum and P. dissoticum) caused production to decline noticeably from March through November of 1996 (Figure 2). P. myriotylum causes root rot while P. dissoticum causes general retardation in the marturation rate of the plant. The problem with Pythium may have been aggravated by the use of a microscreen drum filter to remove suspended solids from May 30 through November 11, 1996. The drum filter was more efficient at removing solids quickly (the wash cycle occurred every two minutes) than the clarifier (twice a day) and filter tanks (once or twice a week). The rapid removal of particulate organic matter from the system may have reduced mineralization and the accumulation of micronutrients. Although nutrient deficiences were not detected, suboptimal nutrient levels could have increased susceptibility to fungal infection. Unfortunately, the ICP method for metal analysis (macro- and micronutrients) gave erroneous results during this period. Efficient removal of solids by the drum filter may have reduced levels of microbes associated with organic decomposition. There was evidence that microbes in the system were antagonistic to plant root pathogens. The Pythium infection was worse in four of the hydroponic tanks after removal of organic matter from the tank bottom. After removal of the drum filter in November, 1996 and return to the original setup of clarifiers and filter tanks, production was consistently high in 1997. However, Pythium was still present as the plants were relatively small at harvest (265 to 327 g) compared to previous unpublished data (>500 g). The use of resistant varieties and antagonstic organisms offer the best potential for Pythium control in aquaponic systems. Minor losses of lettuce were caused by caterpillar (fall armyworm and corn earworm) and aphid damage. Caterpillars were controlled by twice-weekly sprays with Bacillus thuriniensis. During wet periods (September-November), more frequent spraying was sometimes required. Red aphids affected only a few plants for very limited periods. Since the lettuce plants were grown outdoors in an unprotected environment, they were hardened and were not affected by heavy rainfall (less than hurricane strength) or extended periods of wetting. Fungus never occurred on the leaves. The plants were affected by intense solar radiation and air temperature during the warmer months. Wilting often occurred between 1000 and 1600 hours. Nutrient Dynamics There were no observable nutient deficiences during the 2.5-year trial. Direct excretion by the fish, mineralization of organic waste and supplementation with K, Ca and Fe provided sufficient amounts of the essential nutrients for normal plant growth. One complete water exchange and one minor 8

supplementation with P, Mn, Cu, B and Mo in August, 1996 was based on erroneous water quality data. Nutrient levels generally increased during the first 8 months of the trial (Tables 2 and 3). After the system was impacted by Hurricane Marilyn in September, 1995, nutrient levels decreased due to a major reduction in fish biomass and feeding rate. After the poststorm period, the concentration of several nutrients remained relatively constant until a drum filter was added in May, 1996, with the exception of NO3-N, SO4-S, Mn, Cu and Zn, which increased. Levels of K and Fe were variable and reflected supplementation rates. The addition of the drum filter reduced nutrient levels as indicated by the conductivity (as TDS), which decreased from 677 in May to 482 in July, 1996. The drum filter increased daily makeup water from 1.5% of system volume, the overall average for the trial, to 3.2%. In addition, rapid removal of solids from the system reduced mineralization, a process that generates nutrients. Furthermore, two failures of the drum filter float switch led to massive water loss (78% of system volume) in October, 1996. Therefore, the drum filter was removed in November, 1996. A complete water exchange in August, 1996 and a partial accidental water loss (50%) in March, 1997 caused only a moderate reduction in nutrient levels. After installation of a larger pump on December 18, 1996, the flow rate increased by 132%, which significantly altered solids removal and nutrient levels. There was a decrease in clarifier retention time from 22 to 9.5 minutes. As a consequence, less solids were removed by the clarifier and significantly more solids accumulated in the filter tanks, which increased rates of denitrification and mineralization. Nitrate-nitrogen levels decreased dramatically to a low of 1.8 mg/L in May, 1997. When the frequency of solids removal from the first filter tank was increased to twice a week, the NO3-N concentration increased to 11.9 mg/L in June. With the faster flow rate, K and Mn levels decreased while concentations of Mg, PO4-P, SO4-S, Cu, Zn, B and Mo increased, an indication of increased mineralization. Conductivity (as TDS) and Ca remained realtively constant. The levels of most nutrients remained well below the initial concentrations of nutrients in hydroponic formulations, which are high because nutrients are steadily depleted in hydroponics unlike aquaponics where nutrients are constantly regenerated. Manganese levels were often quite low compared to hydroponic levels while Zn concentrations were substantially higher. There was no problem with nutrient accumulation in the commercial-scale unit as found in earlier studies with experimental systems (Rakocy et al., 1993b). Total feed application during the trial was 10,440 kg and the daily feed input averaged 12.0 kg. The average feed input was equivalent to 56 g/m2 of plant growing area/day, very close to the design ratio of 57 g/m2/day, which was determined to be optimum for lettuce production (Rakocy, 1997). During the trial, 168.48 kg of KOH, 34.48 kg of CaO, 142.9 kg of Ca(OH)2 and 62.668 kg of iron chelate (10% Fe) were added to the system, which was equivalent to the addition of 16.1, 3.3, 13.7 and 6.0 g, respectively, for every kilogram of feed added to the system. The use of CaO was discontinued because it was too expensive.

Water Quality The aeration, solids removal and hydroponic components maintained good water quality. DO levels averaged 6.2 mg/L in the rearing tanks (range = 2.4 to 8.6 mg/L), 4.0 mg/L in the effluent from the filter tanks and 6.9 mg/L in the effluent from the hydroponic tanks. TAN

Table 2. Conductivity (as TDS) and concentrations (mg/L) of macronutrients over 2.5 years in a commercial-scale aquaponic unit. ____________________________________________________________________________ Macronutrient _______________________________________________________ a Date Conductivity Ca Mg K NO3-N PO4-P SO4-S _____________________________________________________________________________ 01/95 62 10.7 0.7 0.7 0.4 0.4 0.0 02/95 74 13.7 0.9 0.3 2.6 0.6 0.1 03/95 134 20.0 2.7 1.7 13.1 2.0 0.1 04/95 218 33.6 5.2 10.7 23.2 2.2 0.2 05/95 554 64.9 9.6 80.8 65.9 5.4 10.7 06/95 701 82.1 12.1 104.4 79.0 6.7 13.2 07/95 779 81.3 12.8 98.1 59.0 6.2 14.1 08/95 706 66.8 12.7 90.6 82.2 5.1 13.8 09/95 687 66.9 10.8 101.9 31.8 1.7 12.0 10/95 532 49.3 7.4 76.9 46.8 2.9 5.5 11/95 674 58.7 8.7 139.8 51.2 2.2 8.3 12/95 645 58.8 10.7 192.1 55.2 1.8 9.6 01/96 598 48.0 9.8 110.8 58.2 2.5 9.8 02/96 682 56.8 11.5 108.1 64.7 2.1 12.6 03/96 725 69.4 0.9 14.2 04/96 704 64.6 1.7 18.3 05/96 677 59.2 2.0 20.4 06/96 610 58.1 1.6 1.6 07/96 482 50.0 0.9 9.8 08/96 287 21.4 5.0 0.2 09/96 304 19.3 3.6 0.4 10/96 308 43.7 6.2 54.0 26.8 5.1 8.3 11/96 178 12.3 7.0 0.5 12/96 300 24.9 5.9 42.0 15.6 7.2 8.1 01/97 374 26.0 8.8 51.6 15.6 10.7 13.3 02/97 354 27.0 10.9 49.9 14.3 9.7 16.2 03/97 300 23.8 10.7 33.6 3.9 11.5 16.0 10

04/97 279 22.4 9.4 7.0 1.8 10.3 16.4 05/97 353 2.3 14.6 22.4 06/97 390 24.3 12.9 4.8 11.9 15.3 23.0 b HNF 1900 250 36 200 177 60 129 c HNF 1200 150 50 150 115 50 52 _____________________________________________________________________________ a Monthly values are the average of two or three biweekly samples for conductivity, N and P. b Hydroponic nutrient formulation for lettuce grown in the tropics, wet season (Resh, 1995). c Hydroponic nutrient formulation for lettuce grown in the tropics, dry season (Resh, 1995). Table 3. Concentrations (mg/L) of micronutrients over 2.5 years in a commercial-scale aquaponic unit. ____________________________________________________________________________ Micronutrient ________________________________________________________________ Date Fe Mn Cu Zn B Mo ____________________________________________________________________________ 01/13/95a 1.17 0.01 0.01 0.11 0.02 0.01 02/23/95 0.17 0.01 0.01 0.17 0.01 0.00 03/22/95 0.15 0.01 0.03 0.21 0.03 0.00 04/18/95 0.46 0.01 0.04 0.33 0.02 0.00 05/16/95 0.77 0.16 0.04 0.46 0.07 0.01 06/14/95 0.98 0.13 0.06 0.60 0.09 0.01 07/25/95 1.23 0.03 0.07 0.74 0.11 0.01 08/22/95 1.18 0.03 0.07 0.62 0.10 0.00 09/12/95 0.13 0.02 0.05 0.45 0.08 0.01 10/10/95 0.03 0.01 0.02 0.12 0.07 0.00 11/07/95 2.40 0.14 0.03 0.26 0.06 0.00 12/05/95 2.37 0.19 0.05 0.34 0.06 0.00 01/02/96 2.44 0.05 0.05 0.35 0.07 0.00 01/30/96 3.26 0.07 0.08 0.54 0.08 0.00 10/09/96b 0.46 0.05 0.06 0.48 0.11 0.02 12/18/96 3.26 0.05 0.05 0.39 0.05 0.01 01/17/96 3.35 0.10 0.11 0.64 0.08 0.07 02/17/97 3.02 0.07 0.10 0.78 0.09 0.01 03/11/97 3.24 0.05 0.10 0.80 0.09 0.01 04/11/97 4.30 0.04 0.08 0.75 0.23 0.17 06/03/97 1.31 0.02 0.07 0.75 0.19 0.03 c HNF 5.0 0.5 0.03 0.05 0.5 0.02 HNFd 5.0 0.5 0.1 0.1 0.5 0.05 _____________________________________________________________________________ a Initial water sample before fish were stocked. b Water quality results between 01/30/96 and 10/09/96 are not given due to an analytical error. 11

Hydroponic nutrient formulation for lettuce grown in the tropics (Resh, 1995). Hydroponic nutrient formulation for lettuce grown in Florida and California (Resh, 1995).

averaged 1.47 mg/L in the rearing tanks (0.08 to 4.25 mg/L), 1.25 mg/L in the effluent from the filter tanks and 0.61 mg/L (a 51% reduction) in the effluent from the hydroponic tanks. Nitrite-nitrogen averaged 0.52 mg/L in the rearing tanks (0.07 to 1.51 mg/L), 0.68 mg/L in the effluent from the filter tanks and 0.42 mg/L (a 38% reduction) in the effluent from the hydroponic tanks. COD averaged 48.8 mg/L in the effluent from one clarifier (28.2 to 69.4 mg/L), 48.0 mg/L in the effluent from one filter tank (the second one) and 45.3 mg/L (a 5.6% reduction) in the effluent from the hydroponic tanks. TSS averaged 9.0 mg/L in the rearing tanks (2.0 to 34.0 mg/L), 8.2 mg/L (a 9% reduction) in the effluent from the clarifers, 4.5 mg/L (a 45% reduction) in the effluent from the filter tanks and 3.4 mg/L (a 24% reduction) in the effluent from the hydroponic tanks. Turbidity averaged 31.2 formazin turbidity units (FTU) in the rearing tanks (6 to 55 FTU), 30.5 FTU in the effluent from the clarifiers and the filter tanks and 27 FTU in the effluent from the hydroponic tanks. The hydroponic tanks treated the water through direct ammonia uptake by the lettuce roots, nitrification on the tank surface area, and through sedimentation and filtration (by the roots) of solids. A previous study showed that raft hydroponics, using lettuce, provides adequate and sustainable waste treatment up to a feeding rate of 180 g/m2 of hydroponic growing area/day, which is three times higher than the average feeding rate (56 g/m2/day) in this trial (Gloger et al., 1995). Therefore, there was a wide margin of safety in regards to water quality. The average water temperature was 26.60C (23.8 to 29.00C). pH averaged 7.2 in the effluent from one clarifier, 7.1 in the effluent from one filter tank (the second tank) and 7.2 (6.3 to 8.3) in the effluent from the hydroponic tanks. Total alkalinity averaged 56.5 mg/L as CaCO3 in the effluent from one clarifier, 56.5 mg/L in the effluent from one filter tank and 51.9 mg/L (6.0 to 193.0 mg/L) in the effluent from the hydroponic tanks. A decrease in total alkalinity by 4.6 mg/L on passage through the hydroponic tanks indicates the occurance of nitrification, an acid-producing process that reduces alkalinity. In response to nitrification, pH and total alkalinity constantly declined. pH was monitored several times weekly, and whenever pH decreased to less than 7.0, base was added. It is important to maintain pH near 7.0 because nitrification efficiency decreases at lower pH values and nutrient solubility decreases at higher pH values. The average Na concentration was 36.0 mg/L (5.6 to 62.6 mg/L) and the average Cl level was 55.3 mg/L (8.5 to 102.2 mg/L). The accumulation of Na is a concern in aquaponic systems because high Na levels in the presence of Cl are toxic to plants (Resh, 1995). The maximum Na concentration should not exceed 50 mg/L (Verwer and Wellman, 1980). Only one Na value during the trial was greater than 50 mg/L. High Na levels will interfere with the uptake of K and Ca (Douglas, 1985). In lettuce, reduced Ca uptake leads to tipburn, resulting in an unmarketable plant (Collier and Tibbitts, 1982). Tipburn has occurred during the warmer months. Soluble salt (NaCl) levels in fish feed, associated particularly with the fish meal fraction, are relatively high. Special feed formulations with reduced salt levels are needed for aquaponics. Until low-salt diets are developed, partial dilutions may be necessary to reduce Na levels. Rainwater was used in this trial because the groundwater of semiarid islands such as St. Croix 12

contains too much salt for aquaponics. Accumulation of nitrates is a concern with aquaponic systems. The discharge from one experimental system contained 180 mg/L of NO3-N (Rakocy, 1994). In an experiment with bibb lettuce, NO3-N accumulated at a rate of 14.9 g/kg of feed (dry weight) input at the same design ratio (57 g/m2/day) used in this trial (Rakocy et al., 1993). With an average daily feeding rate of 12.0 kg (10.8 kg dry weight), NO3-N should have accumulated at a rate of 161 g/day (1.8 mg/L/day) in the commercialscale unit. However, the highest accumulation rate during the trial, between April 3 and May 30, 1995, was only 1.1 mg/L/day, and in the last 10 months of trial, NO3-N concentrations generally declined. On the final sampling day the unit contained only 1.8 kg of total inorganic nitrogen (approximately 0.6 kg on nitrogen was added to the system daily in the feed). The main difference between the experimental system and the commercial unit was the presence of filter tanks in the commercial unit. Large quantities of organic matter accumulated on the orchard netting between cleanings. Denitrification most likely occurred in anaerobic pockets that developed in the sludge. The entire water column moved through the accumulated sludge, which provided good contact between nitrate ions and the denitrifying bacteria. In other areas of the system where organic matter accumulates, such as the clarifier and hydroponic tanks, it forms a layer on the tank bottom, has much less contact with the overlying water and mediates less denitrification, as in the experimental unit. Rakocy (1995) calculated a hypothetical mass balance for this unit and determined that optimally-growing plants were expected to remove no more than 37% of the inorganic nitrogen entering the water. Now it appears that through manipulation of organic matter in the filter tanks (i.e., adjusting the cleaning frequency) nitrate levels can be controlled and excessive buildup can be avoided, thereby reducing the environmental impact of water discharged from the system. Water Consumption Total water consumption during the trial was 1,444 m3. Average daily makeup water was 1.4 m3 or 1.5% of the system volume. In the last 6 months of the trial, water loss through sludge removal from the clarifiers averaged 36 L/day. When all the filter tanks were cleaned once per week, average water loss from the filter tanks was 301 L/day. Therefore, sludge removal accounted for 24% of the water loss from the system. Splashing, evaporation and evapotranspiration accounted for the remaining water loss. In the last month, the filter tanks adjacent to the clarifiers were cleaned twice per week, which brought average water loss from the filter tanks to 468 L/day. Water use was 0.25 m3/kg of total tilapia production and 0.29 m3/kg of net production. This aquaponic unit ranks among the top systems for water use efficiency (Cole et al., this publication; Losordo, 1997). In addition, each cubic meter of water produced 2.1 cases of lettuce or approximately 50 heads. Conclusion This aquaponic unit proved to be reliable, simple to operate and very productive. It can be con-figured to occupy as little as 0.04 ha of land. A hectare of units (25) would produce 77,000 kg of tilapia and 13

42,000 cases of lettuce annually in the tropics. Modifications to the unit improved its performance, and with some minor alterations and additional management experience, further gains in efficiency and productivity are expected. References American Public Health Association, American Water Works Association and Water Pollution C nr l o to Federation. 1995. Standard Methods for the Examination of Water and Wastewater. 19th ed. APHA, Washington, D.C., USA. Boyd, C.E. 1979. Water Quality in Warmwater Fish Ponds. Agricultural Experiment Station, University, Alabama, USA. Abr uun

Boyd, C.E. and C.S. Tucker. 1992. Water Quality and Pond Soil Analysis for Aquaculture. Alabama Agricultural Experiment Station, Auburn University, Alabama, USA. Douglas, J.S. 1985. Advanced guide to hydroponics. Pelham Books, London, England. Collier, G.F. and T.W. Tibbitts. 1982. Tipburn of lettuce. Horticultural Reviews, 4:49-65. HACH. 1990. DR/2000 Spectrophotometer Instrument Manual and Procedures. HACH Company, Loveland, Colorado, USA. Losordo, T.M. 1997. Tilapia culture in intensive recirculating systems. In press in B.A. Costa- Pierce and J.E. Rakocy, eds. Tilapia Aquaculture in the Americas. World Aquaculture Society,Baton Rouge, Louisiana, USA. Gloger, K.G., J.E. Rakocy, J.B. Cotner, D.S. Bailey, W.M. Cole and K.A. Shultz. 1995. Waste treatment capacity of raft hydroponics in a closed recirculating fish culture system. Book of Abstracts, World Aquaculture Society, Baton Rouge, Louisiana, USA. Rakocy, J.E. 1994. Waste management in integrated recirculating systems. Bull. Natl. Res. Inst. Aquaculture, Suppl., 1: 75-80. Rakocy, J.E. 1995. The role of plant crop production in aquacultural waste management. Pages 349-364 in M.B. Timmons, ed. Aquacultural Engineering and Waste Management. Northeast Regional Agricultural Engineering Service, Ithaca, New York, USA. Rakocy, J.E. 1997. Integrating tilapia culture with vegetable hydroponics in recirculating systems. In press in B.A. Costa-Pierce and J.E. Rakocy, eds. Tilapia Aquaculture in the Americas. W rd ol Aquaculture Society, Baton Rouge, Louisiana, USA.

14

Rakocy, J.E. and J.A. Hargreaves. 1993a. Integration of vegetable hydroponics with fish culture: a review. Pages 112-136 in J.-K. Wang, ed. Techniques for Modern Aquaculture. American Society of Agricultural Engineers, St. Joseph, Michigan, USA. Rakocy, J.E., J.A. Hargreaves and D.S. Bailey. 1993b. Comparison of tilapia species for cage culture in the Virgin Islands. UVI Research, 5: 13-17. Rakocy, J.E., J.A. Hargreaves and D.S. Bailey. 1993. Nutrient accumulation in a recirculating aquaculture system integrated with vegetable hydroponic production. Pages 148-158 in J.-K. Wang, ed. Techniques for Modern Aquaculture. American Society of Agricultural Engineers, St. Joseph, Michigan, USA. Resh, H.M. 1995. Hydroponic food production: a definitive guidebook of soilless food-growing methods. Woodbridge Press Publishing Company, Santa Barbara, California, USA. Verwer F.L. and J.J.C. Wellman. 1980. The possibilities of Grodan rockwool in horticulture. International Congress on Soilless Culture, 5:263-278.

15