Biodegradation of Nitrogen in a Commercial Recirculating Aquaculture Facility

Commercial production of fish involves high levels of feeding. While digestive breakdown of lipids and carbohydrates yields water and carbon dioxide as waste products, digestion of proteins also yields nitrogenous compounds. In teleost (i.e., bony) fishes, these nitrogenous wastes are excreted predominately as ammonia. Total ammonia-nitrogen (TAN) consists of ionized ammonia (NH4-N) and un-ionized ammonia (NH3-N), the latter of which can prove toxic to fish. The fraction of TAN in the unionized form is dependent upon the pH and temperature of the water (Losordo 1997, Lekang 2007) and to a lesser degree its salinity (Diaz et al. 2012). At pH values less than 7.5, most ammonia is in the ionized form, and high levels of TAN can be tolerated. At higher pH, however, levels of un-ionized ammonia become problematic. Hence, biodegradation of ammonia is critical for the success of fish culture. Nitrifying bacteria, including Nitrosomonas sp., utilize NH3-N as the energy source for growth, producing nitrite, NO2-N. While nitrite-nitrogen is not as toxic as un-ionized ammonianitrogen, it can prove harmful to fish. The most common mode of toxicity is anoxia, as nitritenitrogen crosses the gills into the circulatory system and converts hemoglobin to methemoglobin, rendering it unable to bind and transport oxygen to the tissues (Palachek and Tomasso 1984, Svobodova et al. 2005). Other nitrifying bacteria, including Nitrobacter sp., utilize nitrite as their energy source, producing nitrate, NO3-N. Nitrate-nitrogen concentra‐ tions are not generally of concern to aquaculturists, as most species can tolerate levels as high as 200 mg/L (Russo and Thurston 1991). Nitrate rarely reaches such high levels, as it is removed from the system by water exchanges and by passive denitrification in anaerobic pockets within the production or filtration systems (van Rijn 1996, Tal et al. 2006) or in denitrification reactors (Hamlin et al. 2008, Sandu et al. 2011).


Need for biodegradation of nitrogen species in aquaculture systems
Commercial production of fish involves high levels of feeding. While digestive breakdown of lipids and carbohydrates yields water and carbon dioxide as waste products, digestion of proteins also yields nitrogenous compounds. In teleost (i.e., bony) fishes, these nitrogenous wastes are excreted predominately as ammonia. Total ammonia-nitrogen (TAN) consists of ionized ammonia (NH 4 + -N) and un-ionized ammonia (NH 3 -N), the latter of which can prove toxic to fish. The fraction of TAN in the unionized form is dependent upon the pH and temperature of the water (Losordo 1997, Lekang 2007 and to a lesser degree its salinity (Diaz et al. 2012). At pH values less than 7.5, most ammonia is in the ionized form, and high levels of TAN can be tolerated. At higher pH, however, levels of un-ionized ammonia become problematic. Hence, biodegradation of ammonia is critical for the success of fish culture. Nitrifying bacteria, including Nitrosomonas sp., utilize NH 3 -N as the energy source for growth, producing nitrite, NO 2 -N. While nitrite-nitrogen is not as toxic as un-ionized ammonianitrogen, it can prove harmful to fish. The most common mode of toxicity is anoxia, as nitritenitrogen crosses the gills into the circulatory system and converts hemoglobin to methemoglobin, rendering it unable to bind and transport oxygen to the tissues (Palachek and Tomasso 1984, Svobodova et al. 2005). Other nitrifying bacteria, including Nitrobacter sp., utilize nitrite as their energy source, producing nitrate, NO 3 -N. Nitrate-nitrogen concentrations are not generally of concern to aquaculturists, as most species can tolerate levels as high as 200 mg/L (Russo and Thurston 1991). Nitrate rarely reaches such high levels, as it is removed from the system by water exchanges and by passive denitrification in anaerobic pockets within the production or filtration systems (van Rijn 1996, Tal et al. 2006) or in denitrification reactors (Hamlin et al. 2008, Sandu et al. 2011).
Controlled degradation of nitrogenous wastes in filtration units is a major consideration in design and operation of commercial recirculating aquaculture systems. Among the technologies available (Crab et al. 2007), biological filtration is most commonly used. Biological filters are designed to provide abundant surface area for the attachment of complex microbial communities (Schreier et al. 2010) rich in Nitrosomonas and Nitrobacter species (Chen et al. 2006, Itoi et al. 2007, van Kessel et al. 2010). The nitrification capacity of the water treatment system is often the factor that limits production in a recirculating aquaculture system (Lemarie et al. 2004, Eschar et al. 2006, Diaz et al. 2012).

Utility of a nitrogen budget
The production efficiency of an aquaculture system can be evaluated through analysis of the conversion of nitrogen to fish biomass and to biodegradation pathways (Thoman et al. 2001). Nitrogen dynamics can be quantified by a mass balance equation, most simply as the difference between nitrogen in the feed supply and nutrients subsequently fixed as fish biomass. A nitrogen budget can quantify nitrogen fixation in fish biomass at various fish stocking densities (Suresh and Kwei 1992;Siddiqui and Al-Harbi 1999), nutrient release into the water column as dissolved and particulate excretion of fish (Krom and Neori 1989), and deposition of nitrogen into pond sediment (Acosta-Nassar et al. 1994). By estimating total nitrogen budgets for a particular species and culture system, we can evaluate the efficacy of water treatment processes (Porter et al. 1987). Hence, a nitrogen budget provides information crucial for the design and optimization of a production system, feeding strategies, and water and effluent treatment processes.

Biodegradation of nitrogenous wastes in a tilapia production system
Blue Ridge Aquaculture (BRA) in Martinsville, Virginia, USA is a large commercial facility that produces 1300 metric tons of hybrid tilapia Oreochromis sp. per year. To our knowledge, it is the largest recirculating aquaculture facility in existence. Before our study, little information existed about nitrogen budgets in commercial-scale fish production facilities, especially those using freshwater recirculating systems. By deriving a nitrogen budget, we can quantify the forms and proportions of nitrogen ingested as food as it becomes bound in tilapia biomass, excreted as metabolites, biodegraded by microorganisms, lost as gas by denitrification, or released in effluent. Knowledge of the nitrogen budget can help optimize operations, improving facility efficiency and maximizing production. Using Blue Ridge Aquaculture as our study system, our objectives were to: (1) examine nitrogen dynamics for the grow-out systems, (2) relate the nitrogen budget to water quality, (3) evaluate biofilter loading and nitrogen removal efficiency, and (4) predict maximum system carrying capacity. All abbreviations used in this chapter are shown in Table 1.   Wang et al. 2006). Nitrite-nitrogen levels should be kept below 5 mg/L within tilapia culture vessels (Losordo 1997). Knowledge of these toxicity values is useful for setting criteria for the design or evaluating the performance of filters for biodegradation of nitrogenous wastes in aquaculture systems.

Culture systems
The BRA facility includes systems for broodstock holding, fish breeding, egg incubation/ hatching, fingerling rearing, and food-fish production. The main building houses 42 recirculating aquaculture systems for grow-out to market size ( Figure 1) that were the focus of our study. Each grow-out system ( Figure 2) includes a fish production tank, a sedimentation basin for solids removal, a rotating biological contactor (RBC) for microbial biodegradation including nitrification, and an oxygenation unit. Each fish production system is rectangular in shape, built from concrete, holds 215 m 3 of water, and consists of a fish-rearing tank (119 m 3 ), a multitube clarifier sedimentation basin (37 m 3 ), an air-driven rotating biological contactor (59 m 3 basin volume, 13,366 m 2 surface area per shaft), and an underground U-tube oxygenation system. The total volume of the grow-out unit is 9030 m 3 . The water surface is at the same level in the fish tank, sedimentation and RBC compartments, and water passes freely from one section into another through large pipes or apertures. A pump receives water from the rotating biological contactor compartment and pushes it through U-tubes and then to the far end of the fish production tank, driving the recirculation. The filtration rate is 3.8 m 3 /min, and the system turnover time is about once per hour.  BRA practices partial water exchange daily for controlling solids, dissolved organics and nutrient accumulation in fish grow-out tanks. Water is exchanged daily from the system in the interval between 2:00 p.m. and 8:00 a.m. Management practice is to completely flush the sedimentation basin after each instance that 227 kilograms of feed has been administered to a particular production unit. The exchange rate averages 22.3% per day, but the daily percentage varies among production units as a function of the size of fish, water quality requirements, and the amount of feed delivered to the system. The exchange water originates from wells, and is supplemented with municipal tap water when necessary. Exchange water replaces that used to remove settled particulate material, and thereby dilutes dissolved organic materials, dissolved nutrients, and salts.
Fish are fed commercially-prepared pelleted diets containing 36 or 40% minimum crude protein and 8-16% lipid levels, varying with the age of the fish. The feed is distributed hourly to the tanks over the 24-hour period. Fish production is managed so that 21-27 metric tons of 600-g fish reaches marketable size each week for shipment to a live market.

System boundaries
For the purpose of this study, the 42 recirculating aquaculture systems for grow-out were delimited as a unique system for purposes of quantifying the nitrogen budget. In certain contexts as set out below, N dynamics were quantified in greater detail in four individual systems. Broodstock holding and spawning facilities, a hatchery, and two greenhouses for fingerling production contain only a small part of the facility fish biomass, volume and exchange flow (i.e., they handle 3.0% of the fish biomass and 4.4% of the total nitrogen input).
Because of their small contribution, the nitrogen budgets for these systems are not presented here, but can be found in Sandu (2004).

Inputs, outputs and nitrogen pools
The nitrogen budget is expressed as a mass-balance equation of all nitrogen forms, with total inputs plus generation equal to total outputs plus consumption. We found no measurable amounts of dissolved inorganic nitrogen in the replacement water. Hence, feed provided to fish was the sole nitrogen source in the form of organic nitrogen (N feed ). Multiplication of N feed by the total amount of feed provided the mass of total nitrogen input (TNI). The nitrogen budget was accounted for in five known pools:

1.
Nitrogen fixed in fish biomass as organic nitrogen, N fish ,

2.
Nitrogen fixed in dead fish biomass as organic nitrogen, N mort , 3. Dissolved inorganic nitrogen, N DIN , which included N TAN , N NO2 , and N NO3 ,

4.
Total organic nitrogen in effluent, N TON , and 5. Nitrogen gas removed from the system by passive denitrification, N denit , and by ammonia volatilization, N NH3 vol .
All transformations among pools were assumed to be in a dynamic equilibrium over a defined period of time. We accounted for the mass fractions of nitrogen from Pools 1 to 4 (i.e., the measurable pools) as total nitrogen recovered (TNR), while the difference between total nitrogen input and total nitrogen recovered constituted pool 5, the mass fraction of total nitrogen unaccounted for (TNUA).

Analytical techniques
Analyses of fish and of feed for protein content followed Thiex et al. (2002), who indicated that by dry weight, 16% of protein is nitrogen. Samples were processed at the Forage Testing Laboratory, Virginia Polytechnic Institute and State University, Blacksburg, Virginia. Analyses for inorganic dissolved nitrogen forms (TAN, NO 2 --N, and NO 3 --N) were conducted on site using a Hach DR2400 spectrophotometer (Hach Company, Loveland, Colorado). Total Kjeldall nitrogen (TKN) was determined using macro-Kjeldall Standard Method 4500 -N org B (APHA et al., 1998). Samples were acidified below pH 2 using H 2 SO 4 , refrigerated with ice, and transported to the Department of Civil and Environmental Engineering at Virginia Polytechnic Institute and State University, Blacksburg, Virginia, for analysis. Temperature and pH were measured directly on site using an Acorn Meter (Kit Model pH 6, Oakton, Vernon Hills, Illinois). Alkalinity was determined on-site using the Hach Permachem ® Method. Dissolved oxygen (DO) was measured using a YSI (Model 550, Yellow Springs, Ohio) instrument. We calculated total organic nitrogen as the difference between TKN and total ammonia nitrogen (TAN).

Nitrogen budget determination
Under steady-state conditions, fish biomass does not fluctuate significantly over time (i.e., harvest equals growth), and the daily rations of feed are constant. Under these assumptions, we derived the nitrogen budget by determining the nitrogen input with feed and the output of nitrogenous compounds in known pools. We quantified daily amounts of nitrogen in feed, fish, and mortalities using information on feed consumption, fish production, and mortalities provided by BRA management. We measured the components of dissolved inorganic nitrogen and total organic nitrogen pools directly. We extrapolated mean values to the entire exchange volume from a day to determine the mass of nitrogen recovered in these forms. We assumed that the amount of nitrogen missing from the balance was lost by passive denitrification and by ammonia volatilization.
We considered both types of feed used in the system (with 36% or 40% standard protein content) to determine nitrogen fixed in feed, N feed . We collected samples from three different points in storage silos for nitrogen content determination. We calculated N feed as a composite using the equation: where FA = amount of feed, PC = protein content of the feed, and 0.16 = concentration of nitrogen in protein (Thiex et al. 2002). We determined PC by laboratory analyses because protein content may differ from that claimed by the feed producer. We obtained the total mass of nitrogen originating from the feed input, TNI, by multiplying N feed by the amount fed, FA.
To determine fixation of nitrogen in fish, N fish , we analyzed protein content in triplicate samples of muscle tissue from fish from three size-classes. We estimated the proportions of fish in each size-class as 5% juveniles (i.e., newly introduced to the system from the hatchery), 60% intermediate, and 35% marketable size. With data on protein content of each fish size-class, we determined N fish as a composite using the equation: where FB = biomass of fish, and FP = protein content of the fish.
About 3.5% of the fish production (by number) was lost as mortalities. We assumed that nitrogen fixed in dead fish, N mort , had the same nitrogen content as N fish . In order to determine the biomass of N mort , we collected mortalities daily from the production system for a two-week period, sorted them by size, and weighed them. We used these data to determine N mort using equation 2.
Nitrogen load, L N , entered the water column as ammonia and as organic nitrogen bound in feces. We quantified L N as all nitrogen from feed that was not accounted for as living or dead fish as using the equation: Hence, L N quantified the amount of nitrogen that sustained the nitrogen cycle throughout the system, supplying all effluent nitrogen pools.
We quantified total organic nitrogen as the difference between TKN and TAN from the effluent. We obtained values for TKN, TAN, NO 2 --N, and NO 3 --N by analyzing seven samples collected from the effluent discharge pipe at 3-hour intervals between 2:00 p.m. and 8:00 a.m. because effluent originated from the production system only during that interval. We repeated the tests twice (on different days) and averaged the results. We estimated daily production of these nitrogen forms by multiplying the average concentration (mg/L) by the volume of wastewater released from the system during a one-day period.
All nitrogen in feed that was not recovered as living or dead fish or as total organic nitrogen represented the dissolved inorganic fraction that entered the water as TAN. Hence, we determined ammonia production as: The sum of TAN, NO 2 -N, and NO 3 --N found in the effluent represented the fraction of nitrogen recovered as dissolved inorganic nitrogen, N DIN . The summation of N DIN , N fish , N mort , and N TON provided the value for total nitrogen recovered, TNR. We determined total nitrogen unac-counted for, TNUA, by subtracting total nitrogen recovered, TNR, from total nitrogen input, TNI.

RAS carrying capacity, RBC design, TAN and NO 3 --N removal
We used a simplified version of a model proposed by Losordo and Westers (1994) to determine the carrying capacity of the production system; that is, we considered only the parts of the model concerning maximum system carrying capacity with respect to TAN. Modeling of the flow rate through biofilters was unnecessary because the flow rate was fixed among all recirculating aquaculture systems at 3.78 m 3 /min.
Four recirculating aquaculture systems chosen for intensive study held different age-groups of fish from juvenile to marketable size in order to represent the overall population in the facility. We knew total fish biomass, fish size, feeding rate, crude protein content of feed, daily percent body weight fed, flow rate through the system, and daily rate of exchange for each selected system. We measured other parameters, such as TAN, NO 3 --N, NO 2 --N, pH, temperature, and dissolved oxygen, using standard methods (APHA et al., 1998). We performed these analyses on composite samples collected from the fish-rearing tanks or from the rotating biological contactor's influent and effluent at four-hour intervals. By sampling from appropriate locations, we determined the effects of fish tanks, biofilters or sedimentation basins on each parameter. The experiments extended between consecutive water exchanges. We scaled the data to 24-hour intervals and determined mean and variance for each water quality parameter.
We determined the maximum system carrying capacity with respect to TAN as follows. We calculated the maximum allowable TAN concentration, A TAN, as: where a = the mole fraction of unionized ammonia nitrogen as determined by pH and temperature (Huguenin and Colt, 1989). We calculated maximum feed rate, FR mTAN , by assuming that the TAN concentration of a fish tank equals A TAN , as: where Q f = the recirculating flow rate, or flow rate to the RBC, known to be 227,100 L/hr, and 0.092 = model constant coefficient. We determined the efficiency of the rotating biological contactor for removal of ammonia nitrogen, E a as: We estimated the maximum biomass that could be sustained within the system, SBM mTAN as: where %BW = the feeding rate, expressed as a percent of body weight per day.
P TAN is the rate of production of TAN in the system by metabolism of fish and microbial degradation of uneaten feed. We estimated P TAN as a function of the feed rate and the percentage of protein in feed: ( ) where t = the period of time from the onset of feeding to the next feeding.
This equation is based on the following assumptions and empirical estimates: We determined the mass flow rate of TAN to a rotating biological contactor, or ammonia loading, L TAN , from known (Q f ) and experimentally determined (C TANf ) parameters as: We determined the ammonia removal rate, R TAN , as: The fraction (R TAN x 100) / P TAN represents the percentage of TAN that was removed by means other than the rotating biological contactor.
We estimated the nitrification performance of a rotating biological contactor as areal conversion rate, ACR, representing the amount of TAN oxidized by a unit of surface area in 24 hours: where S, the surface area of an RBC, was 13,336 m 2 .
The mass balance quantifying the partitioning of P TAN removal was: We used a similar approach to determine NO 3 --N partitioning using the equation:

Statistical analysis
We used linear regressions to determine the relationship between daily TAN production (P TAN ) and TAN removal efficiency per pass (E a ), and between fish biomass and percent P TAN transformed by passive denitrification in the four systems tested.

Nitrogen budget
We derived the nitrogen budget for the entire production system for mean conditions of 28.4ºC, pH 7.14, and alkalinity 119.0 mg/L as CaCO 3 . For annual production of 1300 metric tons of fish biomass, BRA administers 2210 metric tons of feeds. These amounts correspond to 6054.8 kg feed consumed per day and 3561.6 kg fish weight gain per day. Of the feed utilized, 95% (5752.0 kg) was nominally 36% protein and 5% (302.8 kg) 40% protein content. However, laboratory analyses showed that the actual protein contents of the two feeds were somewhat lower, 35.0±0.2% and 39.8±0.2%, respectively. The estimated percentages of feed types and the laboratory-determined protein concentrations were used to determine the nitrogen fixed in feed, N feed = 56.38 g/kg feed. By extrapolating N feed , we determined a total nitrogen input of TNI = 341.381 kg/day.
Laboratory analyses showed that the three size-classes of fish from small to large had 18.04±0.16, 20.75±0.02 and 22.26±0.74% protein content, respectively. From these data, we determined that the nitrogen fixed in fish was N fish = 33.83 g/kg produced. Extrapolating to the daily biomass of fish produced, the total nitrogen assimilated in fish was 120.49 kg/day.
Loss of fish represented 3.5% of the total production by number, with weighing of dead fish indicating losses of 2, 1, and 0.5% from the respective size-classes. This was the equivalent of 30.6 kg fish/day or 1.03 kg total N mort /day, representing 0.86% of the total nitrogen assimilated. Hence, 35.3% of nitrogen from feed was assimilated in fish flesh (34.4% harvested and 0.86% removed with mortalities), and 64.7% was unassimilated or excreted in different forms. This latter term included nitrogen in uneaten feed that we accounted for in the overall budget as   Table 2 summarizes the daily nitrogen budget for the production system. The relatively low value of total nitrogen unaccounted for, TNUA, was presumably due to nitrogen lost as nitrogen gas produced by denitrification and as ammonia lost to volatilization. Passive denitrification was likely the primary pathway because recirculated fish culture water passed through the sedimentation basin numerous times. As discussed below, the sediment blanket and associated thick biofilm in the multi-tube clarifier created anoxic conditions favorable for microbially mediated denitrification.

Units
Nitrogen pool

Carrying capacity, RBC design, TAN and NO 3 --N removal
The carrying capacity model indicated that recirculating aquaculture systems at Blue Ridge Aquaculture could support biodegradation of up to 3.15 mg TAN/L. This value corresponds to 0.025 mg/L maximum allowable unionized ammonia (A TAN ) at conditions of pH 7.0 and temperature of 30ºC (Huguenin and Colt 1989); our average values of these parameters for the four recirculating aquaculture systems monitored in greater detail were pH 7.09 and 27.8ºC. At 0.025 mg/L TAN, a recirculating system should be able to receive a maximum feeding rate of FR max TAN = 269.8 kg feed/day, which would support a fish biomass of SBM max TAN = 10,287.4 kg fish/system. Estimates of these parameters for each selected RAS are presented in Table 3.
Comparison with actual feeding rates at the time of experiment (Table 4) showed that system loadings were 56.7 -91.5% of the maxima estimated (Table 3, Figure 3). Over the four tanks examined in detail, TAN removal efficiency per pass, E a , averaged 54.4%. We determined the rate of TAN production (P TAN , Table 3). We determined P TAN per kg of feed consumed by dividing these values by the daily amount of feed introduced into a system: i.e., 40.6 g/kg feed for feed with 40% crude protein content, and 36.7 g/kg feed for feed with 36% crude protein content. We found a positive, linear relationship between P TAN (which also was proportional to the feeding rate) and E a (slope = 0.0013, r 2 = 0.72), thereby showing that the rotating biological contactors efficiently removed various loadings of ammonia. None of the RBCs tested were working at maximum capacity.   Table 4. Characteristics of the recirculating aquaculture systems selected for evaluation. The mass flow-rate of TAN to a rotating biological contactor, L TAN , averaged 467.7 g/hr, which was removed at an average rate of R TAN = 255.2 g/hr. Per-system values are presented in Table  3. The ratio of R TAN to P TAN showed that rotating biological contactors removed an average of 84.0% of total ammonia nitrogen from the selected systems. From the difference, 1.1% of TAN was recovered from exchanged water and 15.0% remained unaccounted for, probably transformed to NO 2 --N and NO 3 --N by passive nitrification or lost by volatilization of ammonia. Data in Table 3 show that fish biomass in the system was positively correlated with the percentage of total ammonia nitrogen transformed by passive nitrification (slope = 0.0015, r 2 = 0.69); although the correlation was not strong, it shows that systems with higher biomass had lower water quality and larger microbial populations, including nitrifiers that promoted in-situ biodegradation of ammonia.
The rotating biological contactors removed between 378.4 and 534.2 mg TAN/m 2 /day (442.5 mg TAN/m 2 /day on average, Table 3). The areal conversion rate, ACR, increased with the loading of total ammonia nitrogen. Average ACR under conditions of maximum system biomass was estimated at 634.0 mg TAN/m 2 /day. We note that the difference between existing ACR and predicted maximum ACR is consistent with that between the existing fish biomass and predicted maximum fish biomass.
We derived a daily nitrogen budget partitioning the total ammonia nitrogen removal from each RAS (  4 TAN removed with exchanged water. Table 5. Partitioning of total ammonia nitrogen removal for each recirculating aquaculture system studied.
We conducted tests on the same recirculating aquaculture systems to determine the fate of NO 3 --N following its production by nitrification. We regarded P NO3 --N as approximately equal to P TAN by assuming that TAN lost from the systems by water exchange and volatilization was negligible. Data on P NO3 --N , water exchange rates, and NO 3 --N concentrations before and after water exchange allowed us to determine the total mass of NO 3 --N in the systems at these times and the amounts of NO 3 --N lost by water exchange and passive denitrification. That is, we derived a daily mass balance quantifying P NO3 --N removal pathways from each RAS (Table 6). Results indicated that NO 3 --N accumulation was in the range of 9.1 -17.2 mg/L in each RAS over a 24-hour period. On average, 44.1% of NO 3 --N was removed by water exchange, and the difference of 55.9% was removed by passive denitrification. NO 3 --N in effluent could be subject to microbial denitrification if water reuse is implemented (Sandu et al. 2008 Table 6. Dynamics and partitioning of P NO3 --N removal for each recirculating aquaculture system studied.
NO 2 --N always remained at concentrations lower than 0.3 mg/L in the fish tanks. Its concentration increased slightly as water passed through the sedimentation basin, but decreased again to concentrations lower that those in fish tanks after contact with the RBC, creating an equilibrium concentration. Because NO 2 --N concentrations were generally stable and below levels considered a threat to fish, we pursued no further determination of NO 2 --N dynamics.

Discussion
We quantified nitrogen fixation and biodegradation through the recirculating tilapia production system at Blue Ridge Aquaculture, a large commercial production facility. The 34.4% of total nitrogen input assimilated by the fish indicated excellent nitrogen utilization relative to other production systems. For example, Suresh and Kwei (1992) found that less than 20% of nitrogen was assimilated by tilapia using feed with 22% crude protein content and much lower fish stocking densities than those at BRA. Using feed with 34% crude protein content, Siddiqui and Al-Harbi (1999) reported 21.4% nitrogen assimilation by red tilapia. Although Suresh and Kwei (1992) found decreasing nitrogen assimilation with increasing fish density, Refstie (1977), Rakcocy and Allison (1981), and Vijayan and Leatherland (1988) reported the opposite finding. We attribute the high nitrogen assimilation in our study to higher protein content in feeds used at BRA, well-managed water quality, and to production of selectively bred fish (Hallerman 2000). Also, most earlier studies reported higher mortality rates, diminishing total nitrogen accumulated in fish.
The small amounts of nitrogen recovered as TAN and NO 2 --N likely were due to biodegradation in rotating biological contactors, which oxidized them effectively to NO 3 --N. Most of the nitrogen recovered as total organic nitrogen (18.93%) was probably due to feces, noting that feed was consumed by fish almost instantly at distribution, and that only fine particulates could escape as wasted feed. Assuming that some organic nitrogen in feces dissolved upon contact with water, our results with tilapia, which accounted for nitrogen from the entire organic pool, broadly agree with those of Porter et al. For total nitrogen unaccounted for (14.31%), removal of N 2 gas through passive denitrification is the most reasonable explanation. Although denitrification may seem surprising given the relatively high dissolved oxygen in the recirculating systems, development of anoxic microsites in sediment provides likely sites for denitrification (Brandes and Devol 1997). Anoxic microsites could arise in fish tanks where particles accumulated, or more likely, in the sedimentation basin, where a blanket of sediments developed for 19 -36 hours before removal. We observed that large amounts of gases rapidly collected beneath the water surface in the sedimentation basin; however, samples we collected were contaminated with oxygen, precluding evaluation of biologically-generated nitrogen production. A thick biofilm on the tanks' walls also could have provided anoxic microsites, contributing to NO 3 --N removal. This explanation was supported by our results for Tank A12, where fish were harvested and the biofilm removed from the walls less than two weeks before our monitoring began. The time for regrowth of the biofilm to a thickness that could allow denitrification was limited. Subsequently, less than 35% of NO 3 --N production was removed by passive denitrification from this particular system, considerably less than in the other three systems monitored. In-situ denitrification has been reported by other authors. For example, Bovendeur et al. (1987) found that 40 -80% of TAN oxidized by nitrification then was reduced by denitrification. Thoman et al. (2001) attributed 9 -21% losses of systems' nitrogen to denitrification. The 56% removal of NO 3 --N by passive denitrification in our study represented an important, positive outcome, because it could reduce by more than half the investment necessary for nitrogen removal should the effluent be treated and reused as suggested by Sandu et al. (2008Sandu et al. ( , 2011.
Our results indicated that despite high fish densities maintained at BRA, the systems are not being operated at their maximum carrying capacity. Our results showed that an average of 73% of the recirculating systems' productive potential was utilized, although utilization approached 92% in systems holding fish close to harvest size. In particular, much productive potential can be realized in systems holding smaller fish for long periods. By better distributing fish biomass among systems via more frequent grading, net production could be increased within existing space. Our suggestion for increased production is supported by the excellent average removal efficiency for rotating biological contactors (54.4%) at a recirculation rate of almost one pass per hour, and by an average areal conversion rate of 442.5 mg TAN/m 2 /day, which maintained an average TAN of 2.06 mg/L in fish tanks. Up to 2830 mg TAN/m 2 /day can be removed by a rotating biological contactor (Rogers and Klemeston 1985), suggesting that the biofilters could function successfully under the maximum conditions of 3.15 mg/L TAN and 634 mg TAN/m 2 /day areal conversion rate that we predicted. Additionally, reusing water using a treatment train such as that described by Sandu (2004) and Sandu et al. (2008Sandu et al. ( , 2011 with 1.6 mg/L TAN, only 0.84% of total ammonia nitrogen produced would be reintroduced to the recirculating systems. This additional loading would be removed easily by the rotating biological contactors, without significant increase of TAN throughout the systems.

Conclusion
Routine aquaculture production generates waste products for which controlled biodegradation in treatment units is a major consideration in design and operation of recirculating aquaculture systems. Biodegradation of nitrogenous wastes is critical, especially for unionized ammonia and nitrite, which are toxic to fish. We quantified the dynamics of nitrogen through a large commercial recirculating aquaculture facility producing hybrid tilapia Oreochromis sp. Our nitrogen budget evaluated total ammonia nitrogen (TAN) production and removal in biofilters, quantifying the fate of nitrate-nitrogen (NO 3 --N) and determining the systems' maximum carrying capacity under steady-state conditions. Most of the recovered nitrogen was in fish, nitrate-nitrogen, and total organic nitrogen pools, with relatively small proportions as total ammonia nitrogen, mortalities, and nitrite-nitrogen, totaling 86%. The remaining 14% of the nitrogen budget unaccounted for likely was lost by passive denitrification to nitrogen gas and by volatilization of ammonia. Our nitrogen biodegradation model predicts that the systems could operate safely at up to 3.15 mg/L total ammonia nitrogen. Under current production conditions, system loading was 57-92% of the maximum fish biomass that could be supported. The biofilters' areal conversion rate could be increased by half under conditions of maximum biomass loading. NO 2 --N was not a parameter of concern, always remaining below 0.3 mg /L. Our results showed that microbial biodegradation of fish wastes was more than adequate and that fish production could be increased within the existing farm infrastructure, especially by more frequent grading of fish in order to stock production systems at densities approaching carrying capacity. With denitrification, discharged culture water could be reused to realize savings in operating costs. Beyond the narrow interest in our study system, our approach can be applied more broadly to other fish culture systems.