Oxygen Demand Math for Ponds and RAS

Aeration on most farms is sized the way it was sized on the farm next door. Horsepower per acre, a number that worked, and a paddlewheel added the year something went wrong. It is not irrational — it encodes real experience — but it prices installed capacity rather than delivered oxygen, and it cannot tell an operator what happens when the feed rate goes up.

There is a better basis, and it is already in the office. Oxygen demand tracks what the farm feeds, and the feed table is the one document on a farm that is written down daily and rarely wrong.

What actually drives oxygen demand in a pond?

The proportions move with the system. An earthen pond carries heavy sediment and a dense plankton bloom, and the non-animal share can rival the animals. A lined pond carries less, a tank almost none. Arithmetic that ignores this is roughly right on a raceway and badly wrong on a fertilized pond.

Timing is what makes the pond version hard. Photosynthesis puts oxygen in through the day and takes none out; respiration runs around the clock. Dissolved oxygen therefore peaks in the late afternoon and bottoms out just before dawn, and the dawn minimum is what kills fish. Equipment sized to average demand is sized for an hour that does not matter.

Temperature pushes both sides the wrong way at once. Metabolic rate rises with temperature while solubility falls: fresh water holds about 9.1 mg/L at 20 °C and about 8.3 mg/L at 25 °C under air at one atmosphere. The hot week is the week demand is highest and the ceiling is lowest.

Which is why a pond cannot bank oxygen. Take a 5,000 m³ pond purely as an arithmetic illustration: at 8.3 mg/L, the entire water column holds about 41 kg of dissolved oxygen, and that is the whole standing reserve against a demand that runs continuously.

Why is feed rate a better predictor than biomass?

Those inferences compound. A stocking count with its own error, a mortality estimate nobody can verify without draining the pond, and a mean weight from a net that does not catch representatively — multiplied together, the biomass figure can be off by a wide margin either way. The feed table has one error term: whether the recorded amount is the amount that went in.

Feed also captures the demand biomass misses. Uneaten pellets and metabolic waste settle and become sediment oxygen demand; ammonia excreted after a meal is nitrified, and nitrification consumes oxygen of its own. All of it scales with what was fed, not with what is swimming.

That gives the method a useful asymmetry. Overfeeding does not make the estimate optimistic — feed that is not eaten still creates demand, just later and lower in the water column. Underfeeding a stressed population, on the other hand, will understate the demand of the biomass present, which is why the feed table should be read alongside a stocking record rather than instead of it.

How do I size aeration from my feed table?

1. The heaviest feeding day. This takes nothing on trust. Feeding is heavily back-loaded, so the last weeks of a grow-out carry far more biomass and far more feed than the cycle average, and the plant has to survive that day rather than a typical one. Your own feed table already contains that number: find the largest single day and use it. Adding a peaking factor to an average is a way of estimating something you can look up.
2. The coefficient. This is the one number this post will not print. Take it from your own feed manufacturer or a standard aquaculture engineering reference, where it arrives attached to the feed formulation and the species it was derived for. The value inside our aquaculture calculator covers the excretion and assimilation load that follows a meal rather than respiration alone, and the tool shows that basis alongside the number. Use it there, on your own feed table, rather than lifted out of context.
3. Non-animal demand. Add what is not a fish. In a fertilized earthen pond, sediment and plankton respiration are a large multiplier on the animal figure; in a lined intensive pond, smaller; in a recirculating tank, close to negligible, though the biofilter carries a nitrification demand of its own. This is a system property to be measured or estimated locally, and it is the term most often left out entirely.
4. From daily mass to a rate. The denominator decides everything. If the arithmetic returns 30 kg of oxygen a day, dividing by 24 gives 1.25 kg per hour — the flat average, and the wrong number to size on, because the pond does not fail on an average hour. The denominator you want is the window you have to survive: the hours from the pre-dawn minimum until light and warming lift the pond back out of it. That window comes off your own dissolved oxygen record. Take a six-hour window purely as an arithmetic illustration, on the same footing as the 5,000 m³ pond above, and 30 kg over six hours is 5 kg per hour — four times the flat average.
5. From rate to equipment. Oxygen gas weighs about 1.33 g/L at 20 °C and one atmosphere, so 5 kg per hour is roughly 3,760 L per hour, or about 63 L per minute of pure oxygen. A PSA concentrator does not deliver pure oxygen; units are typically specified around 90–95% O₂, so divide by the purity on your own spec sheet — at 90%, about 70 L per minute at the machine. That is the number a concentrator has to actually deliver.
6. The reality check. Aerators are rated in kilograms of oxygen per kilowatt-hour in clean water at standard conditions, and your pond is neither. Use your manufacturer's rating for your unit, then derate it for temperature, salinity, and the dissolved oxygen deficit you are actually working against. A rating used without derating promises oxygen the machine cannot deliver on the day it matters.

What do the usual rules of thumb get wrong?

The rule is not arbitrary. It was calibrated on extensive ponds at modest feed rates, where surface area and stocking moved together. Intensify the system and the correlation breaks: same pond, twice the demand.

• Sizing to a target concentration. A minimum dissolved oxygen level is a useful alarm and tells an operator when the pond is in trouble. It does not say how much equipment closes the gap, because concentration is a stock and what the equipment supplies is a flow.
• Averaging. Average daily demand is a real quantity and it sizes nothing, because the animals do not die on an average night. Peak day, worst hour, warmest week: size against all three or accept that the system is a fair-weather system.
• Nameplate capacity as delivered oxygen. A paddlewheel rated in clean water at standard conditions delivers less in warm, saline, organically loaded water at a high dissolved oxygen concentration, and the gap is not small. Two farms with identical installed horsepower can have materially different delivered oxygen.

How does this differ between ponds and RAS?

In a pond the arithmetic above is a daily mass balance with a bad hour in it. The controlling questions are how deep the dawn minimum goes, how fast a bloom crash strips the plankton's daytime contribution and leaves only its respiration, and what reserve exists for the night that goes wrong.

In a recirculating system the same arithmetic is done per pass. Demand per hour divided by the loop flow gives the oxygen that has to be added per volume of water each time it goes around, and that is a concentration the water has to be able to hold and give up. It is a tighter, more controllable problem, and it fails faster when the pump stops.

The biofilter deserves its own line in a RAS. Nitrifying bacteria consume oxygen to oxidize the ammonia the stock excretes, and that demand rides on the same feed number as everything else. A design that oxygenates the fish and starves the filter has moved the constraint rather than removed it.