Oxygen budgeting in RAS: from feed rate to transfer device

Key takeaways

• Fish are not the only oxygen users in a recirculating system. Bacteria in the biofilter, the part that cleans the water, and bacteria living on leftover waste can together use more than the fish.
• Oxygen use climbs for hours after feeding, so size equipment for the busiest hour, not the average one.
• Warm water and salty water hold less oxygen, just when warm fish need more of it.
• Blowing air into warm water can only do so much; past that point, only pure oxygen keeps up.
• Removing waste solids early lowers the oxygen bill in two places at once.
• If the oxygen supply fails in warm water there are minutes to respond, not hours, so backup matters.

What this means for design

Fish breathe the oxygen dissolved in the water around them, and in a recirculating system, where the same water is cleaned and reused, that oxygen has to be put back as fast as it is used. Supply too little and the fish eat less and, in the worst case, die; pile spare capacity onto every part of the system and the plant costs more than anyone will pay to build. For anyone building or running a farm, the practical questions are how much oxygen the system needs at its busiest moment and what kind of equipment can deliver it. This section recommends five steps for answering them.

1. Count every user of oxygen, not only the fish. Write the oxygen budget, the tally of everything that uses oxygen, as three lines rather than one: the fish breathing (fish respiration), the bacteria in the biofilter converting the ammonia that fish excrete (nitrification), and the bacteria that feed on leftover organic waste (heterotrophic bacteria). State the coefficient used for each, meaning the figure that says how much oxygen that line consumes.
2. Find out how much oxygen the water can hold. Water can dissolve only so much oxygen from air, a limit called the saturation concentration, and that limit is lower in warm water and in salty water. Take this saturation ceiling from a published solubility relationship at the design temperature and salinity, because it sets what an air-based device can ever deliver.
3. Decide early between air and pure oxygen. A system is either aerated, with air mixed into the water, or oxygenated, with pure oxygen gas dissolved into it. As a rule of thumb rather than a published threshold, below about 25 °C in fresh water there is usually enough ceiling for aeration, and above it usually not.
4. Judge the device on what it will do on the site. Choose it on utilisation, the share of the oxygen supplied that ends up dissolved in the water, under the head (the water pressure) actually available, not on catalogue capacity.
5. Size for the worst hour, not the average day. Oxygen use peaks some hours after a meal, so size for the hour six hours after the largest meal on the hottest expected day.

The rest of this article is the justification for those five steps.

Where oxygen demand in RAS comes from: three demands, one budget

Oxygen demand in a recirculating aquaculture system (RAS), a farm that cleans and reuses its water rather than replacing it, comes from three sources: the fish, the bacteria in the biofilter, and the bacteria living on waste that the solids filters missed. In the cases examined here, the two bacterial demands together are larger than that of the fish.

The first article in this series, on why every aquaculture design starts with a mass balance, argued that the feed plan sets every load in a fed system. Oxygen is where that argument is easiest to get wrong, because the demand does not come from the fish alone.

Khater and colleagues built and validated a dissolved oxygen model for a tilapia recirculating system that splits consumption three ways: fish respiration, nitrification, and heterotrophic respiration on the organic load solids removal missed (Khater et al., 2021). Fish respiration is the fish breathing; nitrification is the work of the biofilter bacteria that oxidise the ammonia fish excrete; heterotrophic respiration is the oxygen used by other bacteria as they feed on organic waste. At 24 °C and a mean fish weight of 226 g, modelled demand was 30.78 g O₂ m⁻³ h⁻¹ (grams of oxygen per cubic metre per hour), of which fish respiration took 10.42 and nitrification 2.46; the remaining 17.9 is heterotrophic. Two thirds of the oxygen bill at the end of that cycle is bacterial; at 32 °C the fish take 21.57 of 44.54 and the bacteria still take more than half.

The nitrification line is stoichiometric, meaning it follows from the chemistry of the reaction, and also a choice: article one carried 4.18 g of oxygen per gram of ammonia nitrogen oxidised once cell synthesis is allowed for (Ebeling et al., 2006), where Khater's model uses 4.57 g, the figure without that allowance. The second figure is about nine per cent higher than the first (4.57 ÷ 4.18 = 1.09). That is not much, but it should be a stated assumption.

The heterotrophic line is the one that moves. In commercial tanks of Atlantic salmon smolt (juvenile salmon) carrying an oxygen demand of 7.4 to 10.4 mg/L (milligrams per litre), Gorle and colleagues found the estimated respiration rate roughly doubling, from about 0.3 to about 0.7 kg of O₂ per kg of feed, as suspended solids (fine waste particles carried in the water) rose from 3.0 to 10 to 12 mg/L (Gorle et al., 2018). Solids control is an oxygen control.

The peak, not the mean

Oxygen demand in a fed tank is not steady through the day: it climbs after feeding and peaks hours later. The equipment has to meet that peak, so a budget built on the daily average (the mean) undersizes it.

Stejskal and colleagues measured oxygen consumption and ammonia excretion in Eurasian perch (Perca fluviatilis) at 23.3 °C, in fed and fasted (unfed) groups across six size classes from 44.8 to 336.2 g (Stejskal et al., 2009). Mean daily consumption ran from 288.3 down to 180.6 mg O₂ kg⁻¹ h⁻¹ (milligrams of oxygen per kilogram of fish per hour) in the fed fish, and from 181.1 down to 110.5 in the fasted fish. Feeding lifts the daily mean to roughly 1.6 times the fasted rate, before the shape of the day is considered.

Consumption peaked six hours after feeding began in every size class, while ammonia excretion peaked at ten hours and fell away by sixteen. The two peaks are four hours apart, which cuts the helpful way: one device does not carry both maxima at once, so sizing on their sum overbuilds the plant. Find the largest coincident total instead, meaning the highest combined load that falls in any one hour. How far that peak rises above the daily mean is the figure a designer actually wants, and the source gives timing without magnitude, so no peak-to-mean multiplier is offered here. That factor has to come from the farm's own dissolved oxygen logging.

Temperature moves the whole curve up: in Khater's model, demand at 226 g rose 45 per cent between 24 and 32 °C. The design condition is the hot day, at the biomass carrying the most feed, hours after the largest meal, with the solids at their worst.

Key figures

The table gathers the figures this article relies on most, with the conditions under which each was measured or modelled and its source.

What the water can hold

Water can hold only a limited amount of dissolved oxygen, and that limit is lower in warm water and in salty water. A warm, salty system therefore starts with less oxygen to work with than a cool, fresh one.

Everything a transfer device (the aerator or oxygenator that dissolves gas into the water) does is bounded by the saturation concentration, the most oxygen that water in contact with air will hold, and that ceiling falls with both temperature and salinity. On the solubility relationship that underpins the standard tables (Benson and Krause, 1984), fresh water at 1 atm (standard atmospheric pressure) falls from 14.62 mg/L at 0 °C to 9.09 at 20 °C and 7.56 at 30 °C; sea water at 35 ppt (parts per thousand, the usual measure of salinity) gives 11.45, 7.40 and 6.24.

The two effects compound, and in an Australian context they compound badly. A brackish system (one between fresh water and sea water in salinity) at 15 ppt run at 30 °C rather than 15 °C loses 24 per cent of its ceiling, from 9.20 to 6.96 mg/L, and taking that water on to full sea water costs about 10 per cent more, down to 6.24 mg/L. Fresh to sea water at 30 °C is the 17 per cent fall marked on the chart. Colt treats reuse-system criteria as interdependent rather than separate (Colt, 2006), and oxygen is the clearest case: ceiling, demand and floor (the lowest oxygen level the fish can tolerate) all move with temperature at once.

Dissolved oxygen at air saturation. At 30 °C sea water holds well under half the cold fresh water value. Original chart, AquaBlueprint.

Aeration cannot get you there

Aeration, which mixes air into the water, can only push dissolved oxygen towards the saturation ceiling, and it transfers less the closer the water already is to that ceiling. In warm, salty water held at the oxygen level the fish need, air has very little left to work with, and pure oxygen or pressure is the way to get it back.

Gas transfer is proportional to the deficit between saturation and the concentration already in the water, so an air-based device working a tank held at 6 mg/L in 30 °C brackish water has under 1 mg/L of driving force. Summerfelt, Vinci and Piedrahita set out the theory and the equipment: raising the partial pressure of oxygen in the gas phase (in practice, supplying a gas richer in oxygen than air), or the pressure at which contact happens, is the only way to buy that driving force back (Summerfelt et al., 2000).

Aerators are compared on the oxygen they transfer for each unit of electricity, in kg O₂ per kWh. Across aerator types at 20 and 35 ppt, an impeller unit managed 2.098 kg O₂ per kWh, a paddlewheel 1.436 and a submersible unit only 0.380 (Jayanthi et al., 2020). Those are controlled-test figures; transfer into a tank already close to saturation is lower, for the reason above.

Pure oxygen devices are compared on utilisation, the share of the oxygen supplied that ends up dissolved in the water. A survey of trout farms in Serbia, North Macedonia and Bosnia and Herzegovina, a practitioner source rather than a controlled trial, reports low head oxygenators at 65 to 91 per cent utilisation; down-flow bubble contactors, the Speece cone family, pressurised and delivering 25 to 50 mg/L at the outlet; surface oxygenators reaching only 110 to 120 per cent of saturation; and ceramic diffusers at 10 to 20 per cent, against 3 to 7 per cent per metre of depth for perforated hose (Čanak et al., 2025). Contact time under pressure buys utilisation; a bubble released at the bottom of a shallow tank does not get long enough.

A worked example: the design day for a 100 t/yr barramundi RAS

This example works out a daily oxygen budget for a barramundi farm by two independent routes, one starting from the feed and one adding up each user of oxygen separately. The totals overlap, at 209 to 466 kg and 369 to 440 kg of oxygen a day, and the gap between them shows what each route leaves out.

Take the design day from article one, 480 kg of a 45 per cent protein diet for a nominal 100 t per year barramundi (Lates calcarifer) system, and add what that article did not need: 40 t standing biomass (the weight of fish in the system at one time) at a mean weight of 600 g, at 30 °C and 15 ppt. Only the feed figure carries over; the rest are assumptions, not measurements. Two routes give the daily budget, and the table sets them side by side, one demand term to a row.

Totals are computed before rounding; sources follow.

Route A takes tank respiration from feed at 0.3 to 0.7 kg of O₂ per kg (Gorle et al., 2018) and nitrification on the 15.6 to 31.1 kg of ammonia nitrogen a day from article one. Route B builds the fish term from physiology, using an equation that estimates the oxygen one fish consumes in an hour from the water temperature and the weight of the fish. Glencross and Felsing give gross oxygen consumption as (−20.7818 + 1.4017T − 0.0227T²) × W⁰·⁶⁷³ mg O₂ per hour per fish, T in °C and W in grams (Glencross and Felsing, 2006). At T = 30 the bracket evaluates to 0.8392, so a 600 g fish takes 62.2 mg an hour and 66,700 of them take 100 kg a day. Two caveats ride on that. Their 26 to 32 °C range covers the design condition, but carrying the weight exponent out to 600 g is an extrapolation. More importantly, what they fitted is a standard metabolic rate, by definition the resting rate of fasted fish, so 100 kg a day is a floor, not a working figure; the perch fed-to-fasted ratio of about 1.6 puts the fed term nearer 160 kg. Route B then adds the heterotrophic term as 2.3 times a BOD₅ excretion rate of 2,160 mg O₂ per kg of fish per day, the approach in Khater's model, giving 199 kg at 40 t. BOD₅, the five-day biochemical oxygen demand, is a standard measure of the oxygen bacteria consume in breaking down organic waste.

Three things fall out. The routes are incomplete in opposite places: Route A's tank term excludes the biofilter, Route B's biofilter term excludes tank heterotrophy. That is most of the disagreement, and Route A's range is wider because the solids assumption inside the 0.3 to 0.7 coefficient swings 192 kg a day on its own. Under Route B the fish are the smaller part of the budget, 100 kg against 270 to 341 kg for the bacterial terms, and they stay smaller at the fed figure of about 160 kg. And both totals are daily means for a plant that has to survive its worst hour.

Where the margin belongs

Safety margin, the spare capacity beyond the calculated need, does most good in three places: a device that can be turned up and down, backup equipment, and solids removal. Spreading it evenly over every line of the budget is the costly alternative.

Margin stacked on every line produces a plant nobody will build, so the question is which failure it protects against. Tran-Duy and colleagues fed Nile tilapia (Oreochromis niloticus) to satiation (as much as they would eat) at four oxygen levels and found intake for fish under 100 g levelling off at 3 mg/L, while for fish over 200 g it was still climbing as oxygen rose from 2.6 to 6.0 mg/L, implying an incipient concentration, the level below which feed intake starts to suffer, near 5.5 mg/L (Tran-Duy et al., 2012). The economic floor sits well above the survival floor, and the survival floor moves too: across five sub-populations of barramundi, critical oxygen saturation, the level below which a fish can no longer take up the oxygen it needs at rest, was 15.44 ± 3.20 per cent at 26 °C and 21.07 ± 3.92 per cent at 36 °C (Collins et al., 2013). Demand rises, the ceiling falls and the lethal threshold climbs, all on the same afternoon.

So the margin belongs in three places rather than in a blanket multiplier. In the device's turndown (its ability to run well below full output), so one unit meets the post-feeding peak without running wastefully overnight. In redundancy (backup equipment), because a transfer failure at 30 °C is measured in minutes, not hours: 40 t of barramundi at 40 to 80 kg/m³ occupies 500 to 1,000 m³, and at Route B's mean demand the 4.5 mg/L of headroom between a working 5.7 mg/L and the critical concentration implied by Collins' data is gone in seven to eighteen minutes. And in solids removal. Solids left in the water raise the oxygen consumed in the culture tank (the Gorle result above) and feed the heterotrophic bacteria downstream (the 199 kg line in Route B), so removing them early is the only measure that lowers two demand terms at once.

In practice

AquaBlueprint is a web-based design platform for aquaculture systems that works through mass balance, component sizing and financial feasibility to produce a design report summary and a bill of materials. Its oxygen module follows the five steps set out at the top of this article: it takes design-day feed, diet, temperature and salinity, returns the three demand terms separately with the saturation ceiling for the stated conditions, and keeps every coefficient visible and editable. The methods implement widely accepted industry design standards, and the output is a design aid for the professional doing the work. What it cannot supply is the diel oxygen curve (the pattern of oxygen use across the day) for your species on your site, because for most Australian species nobody has published one.

Questions practitioners ask

How much oxygen does a RAS need per kilogram of feed?

There is no single figure, because it depends on how much waste stays in the water. In commercial Atlantic salmon smolt tanks, the estimated respiration rate in the tank ran from about 0.3 to about 0.7 kg of O₂ per kg of feed as suspended solids rose (Gorle et al., 2018). The biofilter adds to that: 4.18 g of oxygen for each gram of ammonia nitrogen oxidised (Ebeling et al., 2006).

When does oxygen demand peak after feeding?

Some hours after the meal, not during it. In Eurasian perch at 23.3 °C, oxygen consumption peaked six hours after feeding began in every size class, and ammonia excretion peaked later, at ten hours (Stejskal et al., 2009). That study gives the timing but not how far the peak rises above the daily average, so that figure has to come from the farm's own dissolved oxygen logging.

Aeration or pure oxygen: how do I decide?

Start with how much oxygen the water can hold at the design temperature and salinity, because air can only push water towards that limit. As a rule of thumb rather than a published threshold, below about 25 °C in fresh water aeration usually has enough room to work, and above it usually not. Beyond that point, a gas richer in oxygen, or contact under pressure, is the only way to keep up (Summerfelt et al., 2000).

Why does salinity matter for oxygen supply?

Salty water holds less dissolved oxygen than fresh water at the same temperature, so there is less to work with before a single fish is fed. At 30 °C, fresh water saturates at 7.56 mg/L and sea water at 35 ppt at 6.24 mg/L, a fall of 17 per cent (Benson and Krause, 1984). Warm water lowers the limit as well, and the two effects compound.

References

Benson, B.B. and Krause, D. (1984). The concentration and isotopic fractionation of oxygen dissolved in freshwater and seawater in equilibrium with the atmosphere. Limnology and Oceanography 29, 620-632. https://doi.org/10.4319/lo.1984.29.3.0620

Čanak, S., Radosavljević, V., Gligorević, K., Trajchovski, A., Cvetkovikj, A., Pavličević, J. and Savić, N. (2025). Performances of oxygenation devices on trout farms in Republic of Serbia, Republic of North Macedonia, and Bosnia and Herzegovina. Poljoprivredna tehnika (Agricultural Engineering) 50(4), 99-111. https://doi.org/10.5937/poljteh2504099c

Collins, G.M., Clark, T.D., Rummer, J.L. and Carton, A.G. (2013). Hypoxia tolerance is conserved across genetically distinct sub-populations of an iconic, tropical Australian teleost (Lates calcarifer). Conservation Physiology 1, cot029. https://doi.org/10.1093/conphys/cot029

Colt, J. (2006). Water quality requirements for reuse systems. Aquacultural Engineering 34, 143-156. https://doi.org/10.1016/j.aquaeng.2005.08.011

Ebeling, J.M., Timmons, M.B. and Bisogni, J.J. (2006). Engineering analysis of the stoichiometry of photoautotrophic, autotrophic, and heterotrophic removal of ammonia-nitrogen in aquaculture systems. Aquaculture 257, 346-358. https://doi.org/10.1016/j.aquaculture.2006.03.019

Glencross, B.D. and Felsing, M. (2006). Influence of fish size and water temperature on the metabolic demand for oxygen by barramundi, Lates calcarifer (Bloch), in freshwater. Aquaculture Research 37, 1055-1062. https://doi.org/10.1111/j.1365-2109.2006.01526.x

Gorle, J.M.R., Terjesen, B.F., Mota, V.C. and Summerfelt, S.T. (2018). Water velocity in commercial RAS culture tanks for Atlantic salmon smolt production. Aquacultural Engineering 81, 89-100. https://doi.org/10.1016/j.aquaeng.2018.03.001

Jayanthi, M., Anathaikamatchi, B.A., Sakkarai, S., Thiagarajan, R., Muthusamy, D., Kuppusamy, M., Moturi, M., Bera, A., Ramasamy, P. and Shanmugam, S. (2020). Assessment of the new generation aeration systems efficiency and water current flow rate, its relation to the cost economics at varying salinities for Penaeus vannamei culture. Aquaculture Research 51, 2112-2124. https://doi.org/10.1111/are.14562

Khater, E.S., Bahnasawy, A.H., El-Ghobashy, H., Shaban, Y., Elsheikh, F., Abd El-Reheem, S. and Aboegela, M. (2021). Mathematical model for predicting oxygen concentration in tilapia fish farms. Scientific Reports 11, 24130. https://doi.org/10.1038/s41598-021-03604-1

Stejskal, V., Kouřil, J., Valentová, O., Hamáčková, J. and Policar, T. (2009). Size-related oxygen consumption and ammonia excretion of Eurasian perch (Perca fluviatilis L.) reared in a recirculating system. Aquaculture Research 41, 135-142. https://doi.org/10.1111/j.1365-2109.2009.02316.x

Summerfelt, S.T., Vinci, B.J. and Piedrahita, R.H. (2000). Oxygenation and carbon dioxide control in water reuse systems. Aquacultural Engineering 22, 87-108. https://doi.org/10.1016/S0144-8609(00)00034-0

Tran-Duy, A., van Dam, A.A. and Schrama, J.W. (2012). Feed intake, growth and metabolism of Nile tilapia (Oreochromis niloticus) in relation to dissolved oxygen concentration. Aquaculture Research 43, 730-744. https://doi.org/10.1111/j.1365-2109.2011.02882.x