Mechanism 08 · Corrected 9 September 2026
The big rescue water change is far less dangerous than the folklore says. Here is the arithmetic, and here is the mistake almost everybody makes in it.
The story: someone inherits a neglected tank, is told it needs a big water change, does one, and the fish that were alive that morning are dead by evening. The stated mechanism is that the tank drifted acid, so its ammonia is sitting in the harmless ionised form, and fresh water raises the pH and turns it toxic all at once.
Every step of that mechanism is real. The size of it is not, and the size is the whole claim.
Notice the precondition first, because it does most of the work and is usually skipped. Measurable ammonia in an established tank means the biofilter has failed or been suppressed. In a neglected tank whose filter is still turning over, total ammonia reads zero and nothing below applies. The dangerous object is not a tank nobody changed water in. It is a tank nobody changed water in and whose nitrification has stopped, which is a much rarer thing.
| pH | Fraction of total ammonia present as the toxic form |
|---|---|
| 6.0 | 0.056 %1 |
| 6.5 | 0.178 %1 |
| 7.0 | 0.559 %1 |
| 7.5 | 1.75 %1 |
| 8.0 | 5.32 %1 |
The mistake
A 25 percent water change cannot move the tank to the tap water’s pH. It is still 75 percent old water.1
Almost every version of this calculation, including the first draft of this page and an independently produced check written at the same time, assumes the pH jumps straight to the tap value no matter how little water you change. That assumption produces enormous numbers at small change fractions, and then a conclusion falls out of it backwards.
Below is the same tank under both assumptions. Five milligrams of ammonia nitrogen per liter, tank at pH 6.0, tap water at pH 7.8. With no change at all the un-ionised ammonia sits at 0.0028 mg N/L1.
| Change | Ammonia left | Wrong model: pH teleports to 7.8 | Realistic model: pH moves in proportion | Rise over doing nothing |
|---|---|---|---|---|
| None1 | 5.00 mg N/L | 0.0028 mg N/L | 0.0028 mg N/L | 1x |
| 10 %1 | 4.50 | 0.154, that is 55x | 0.0038 at pH 6.18 | 1.4x |
| 25 %1 | 3.75 | 0.129, that is 46x | 0.0059 at pH 6.45 | 2.1x |
| 50 %1 | 2.50 | 0.086, that is 30x | 0.0111 at pH 6.90 | 4.0x |
| 75 %1 | 1.25 | 0.043, that is 15x | 0.0155 at pH 7.35, the peak | 5.5x |
| 95 %1 | 0.25 | 0.009, that is 3x | 0.0070 at pH 7.71 | 2.5x |
What changes, one
The worst case is in the middle, at about 76 percent.1
Under the realistic model the risk rises with the size of the change, peaks around three quarters, and then falls again as dilution finally overtakes the pH rise. So "small changes are worse" is wrong as stated. So is "bigger is always safer", which is what the first draft of this page said. Neither end of the range is the danger point.
What changes, two, and this is the real finding
The whole thing is an order of magnitude smaller than the story.
The peak un-ionised ammonia in the realistic model is 0.0155 mg N/L, which is 0.019 mg/L expressed as NH31. Tissue damage in fish begins at about 0.05 mg/L as NH33. The worst case of a large rescue water change on this tank does not reach the damage threshold. Work backwards and total ammonia has to be above roughly 13 mg N/L before the peak crosses it1, and a tank reading that is a catastrophe already in progress whether or not anyone touches the water.
Which leaves the other route as the larger one
Ionic and pH shock, which has the weaker evidence.
The neglected tank is also at two or three times the tap water in dissolved solids4, and its fish have spent months osmoregulating against that. A very large change drops the external ionic strength by several hundred milligrams per liter in minutes and moves pH by more than a whole unit. This is standard physiology and it is the universal hobby explanation, and no controlled aquarium study of it was found. With the ammonia route reduced to two fifths of the damage threshold, this is now the bigger of the two, and it is the one nobody has measured. The mechanism as a whole is rated Medium rather than Strong for that reason.
What to do instead of following a percentage
Test for ammonia. That is the instruction, and it replaces the rule.
If total ammonia reads zero, every number above is irrelevant and you can change as much water as you like. That is the case in the large majority of neglected tanks, because the filter is still running.
If it reads anything at all, the biofilter has failed, and that is the actual problem. Fix that. While you do, stage the water changes at about 20 to 25 percent every day or two, feeding nothing at all so no new ammonia is made, and test before each one. Stage them not because a small change is safe and a big one is not, but because staging keeps you off the middle of the curve and lets the ionic strength move slowly enough for the fish to follow.
And do not repeat the large multipliers. Any figure in the tens for a small water change is an artefact of assuming the pH teleports.
The general rule survives all of this, and it is the honest one. A water change is safe on a tank that gets water changes. Whatever residual risk there is on a long neglected tank grows with the length of the gap. That is not an argument against water changes. It is an argument for doing them regularly, because the risk is created entirely by not doing them.
Correction, 9 September 2026. An earlier draft of this section published a table built on the teleporting-pH assumption and concluded that a bigger change is safer than a medium one, with 25 and 50 percent crossing the damage threshold and 80 and 90 percent not1. That is withdrawn. An independent fact-check written at the same time made the same modeling error and reached the opposite conclusion, that small changes are the dangerous ones. Both were artefacts of the shared assumption, and neither draft caught it, because the error sat upstream of where the two disagreed. What caught it was re-running the arithmetic rather than comparing the write-ups. The corrected result is stronger than either draft had.
Independently checked, and one of them did not survive
The most exciting claim on this subject is the one that is false.
Three claims were written up as contested and sent to a second researcher working separately. Their report and this page were compared, and the arithmetic was re-run rather than taken on trust. One claim collapsed, one survived in a narrower form than the hobby states it, and one was never in doubt but does not mean what it is used to mean.
This matters for how the rest of the page reads. Mechanisms 1 to 10 never depended on any of these three. Not one line of the arithmetic changes now that the pheromone story has fallen. A page that had rested its case on it would have to be rewritten today. This one does not, and that is the argument for keeping the well supported mechanisms separate from the interesting one.
The growth inhibiting pheromone does not survive. The fish literature traces to essentially one laboratory and stops in 19815. Its own citation trail runs into four unpublished student theses. Fifty five years of subsequent work has not isolated the factor5. And the modern replication points the other way: with fish housed individually, so social effects are removed, small fish grew faster in accumulated water6, which is the same direction an earlier tilapia study found and which a growth inhibitor cannot produce. "Waterborne somatostatin", the molecule the hobby has attached to this story recently, is a fabrication7. What survives is density dependent growth, which is a well established fact in fisheries biology and needs no chemical at all: competition for food, social stress, hierarchy and suppressed feeding account for what aquarists actually see. The amphibian result, published three times between 1958 and 1966, is real, replicated, and in tadpoles8. Carrying it across to fish is the move that started all this.
Something in nearly closed water does suppress feed intake and growth, in one species, and nobody knows what. A 39 day trial ran two recirculating systems differing in one thing, how much accumulated, with 120 individuals each of Nile tilapia, African catfish and European eel9. The tilapia in the high accumulation system ate 42 percent less and grew 83 percent less9. The catfish and the eel showed nothing at all. The authors' own words are that the cause is unclear9. This is real and it is not the pheromone claim: an effect in one species by an unidentified mechanism is a much narrower statement than a signalling molecule, and the second does not follow from the first.
Steroid hormones do accumulate, and whether it matters is unknown. Seven commercial recirculating systems were sampled at three points each10. Steroids were significantly higher in the rearing water than in the make-up water, cortisol averaged 15.7 percent higher leaving the fish than entering them, and the less make-up water a system used the higher the sex steroids ran10. Effluent ranges were 3.8 to 217 nanograms per liter of cortisol and 3 to 12.5 of testosterone10. The accumulation is settled. Whether nanograms per liter do anything, and whether an ornamental tank ever reaches those numbers, is not.
And the nitrate threshold everyone quotes is the wrong number. The 10 milligrams per liter figure repeated across the hobby is the United States drinking water standard, set to protect human infants from methaemoglobinaemia11. Wrong species, wrong endpoint, and quoted in the wrong units about half the time, because 10 mg of nitrate nitrogen per liter is 44 mg/L on an aquarium kit12. The aquatic life number to use instead is 2 mg of nitrate nitrogen per liter, from a review of nitrate toxicity to aquatic animals, which is about 9 mg/L on a kit that reads nitrate12.
The other side, taken seriously
When you genuinely can do fewer, and what that case is quietly assuming.
Diana Walstad's published figure is that her own tanks often go six months or more without a water change, and that routine maintenance is topping up evaporation and pruning13. The "no maintenance" myth grew out of that sentence by dropping the two verbs at the end. She also changes water when there is a reason: after disturbing a substrate in April 2026 she did a 30 percent change the same day14.
Where she is right
A water change is not the only export route. Nitrogen and phosphorus leave a tank three ways: a water change, denitrification, and plant tissue physically carried out of the room. Pruning is an export. Her whole nutrient strategy is that the plants take the nutrients out of the water and the aquarist takes the plants out of the tank. That works, and nobody measures it.
Plants prevent the acidification, and the arithmetic supports her. Nitrification releases two protons per nitrogen. A plant taking that nitrogen back up as nitrate consumes roughly one, for charge balance. So a well planted tank pays about half the alkalinity cost of an unplanted one. She reported exactly that pattern, drift in her unplanted tanks that stopped once plants grew well2.
The substrate is a buffer, and buffers make systems slow. A soil with real cation exchange capacity holds ammonium, potassium, calcium, magnesium and reduced iron and releases them to roots rather than to the water. Gravel holds almost nothing. Comparing a soil tank's water change requirement to a bare gravel tank's is comparing two different machines.
What the figure is assuming
An evaporation rate. This is the assumption nobody examines and it is the decisive one. Six months with no water change ends at 1.26 times tap at 1 percent evaporation a week, 1.52 times at 2 percent, 2.30 times at 5 percent, and 3.08 times at 8 percent4. Her figure is defensible at a low rate and is not defensible at a high one, and she has never had to think about it because her climate never made her. An open rimless tank in a dry Yerevan August is a different regime from the one that number came out of.
A source water. Her chapter treats hard tap water as a nutrient input rather than a problem. That holds when the top-up water is moderate. It stops holding when the arithmetic above takes a Davtashen tank past 20 dGH4. The assumption should be checked against your own tap, not inherited.
A stocking level. Moderate bioload, generous feeding, heavily planted, low light, shallow tanks. The break-even feeding table is the quantitative version, and a tank with twice the fish is outside it.
A mature substrate. A fresh soil releases ammonium and dissolved organic carbon, and her own description of the first month is a race against time13. The first six to eight weeks of a soil tank need more water changes than a conventional tank, not fewer, and the method is usually sold the other way round.
The exact place the two positions meet
Plant uptake exports nutrients. It does not export ions.
Pruning a bag of Vallisneria out of a tank removes nitrogen, phosphorus, potassium, magnesium, iron, manganese and a little calcium. It removes no sodium, no chloride, no sulfate, no accumulated humic matter, no steroid, no medication residue and no metal the plant did not take up.
So a heavily planted, heavily pruned tank can genuinely run at zero nitrate and zero phosphate for years, and be entirely healthy on the nutrient axis, while the ion axis climbs on the straight line above the whole time, unmeasured The kit reads zero because the plants are winning at the one thing the kit measures.
Which makes the advice short. If you want to run a low water change planted tank, the thing to fix is not the nitrate. It is the top-off. Top up with distilled or reverse osmosis water and the whole curve flattens. That single change makes the six month interval defensible in a climate it was never tested in, and it is in none of the standard advice.
Arithmetic- Our own calculation, not a published finding. Ammonia speciation and the rescue water change. Named inputs: the standard relation fraction as NH3 = 1 / (1 + 10^(pKa - pH)) with pKa 9.25 at 25 degrees, checked against the published fractions at pH 7 and pH 8 in 2; a tank at pH 6.0 holding 5 mg N/L of total ammonia; tap water at pH 7.8; and the assumption that pH after a change moves in proportion to the fraction of water replaced. Compared against the tissue damage threshold in 3. Nobody has peer reviewed this, and an earlier version of it was wrong. How we handle our own arithmetic
- Walstad, D. (2017, reissued 2023). Nitrogen Uptake by Aquatic Plants. Open access, with its own reference list to the primary literature. The nitrification stoichiometry, the 7.1 mg of alkalinity destroyed per mg of ammonium nitrogen, the ammonia fractions at pH 7 and pH 8, and her report that plant growth stopped the acidification. PDF
- Francis-Floyd, R., Watson, C., Petty, D. and Pouder, D.B. Ammonia in Aquatic Systems. University of Florida IFAS Extension, FA031. The hundredfold toxicity difference between the two ammonia forms and the 0.05 mg/L tissue damage threshold. ask.ifas.ufl.edu
- Our own calculation, not a published finding. The evaporation, top-off and ceiling model. Two lines: C(n) = C_tap (1 + n e) for topping up only, and C* = C_tap (1 + e (1 - f) / f) for the ceiling under a repeated change, with f = e / (M - 1 + e) inverted from it. Named inputs: Yerevan district tap water from 4; the hardness unit conversion from 19; and evaporation rates of 2, 5 and 8 percent a week, which are assumed parameters and not measurements, because no measured evaporation rate for an aquarium in Yerevan exists. Nobody has peer reviewed this. Re-run it rather than trust it. How we handle our own arithmetic
- Roales, R.R. (1981). The effect of growth inhibiting factors on the total lipid content of the zebrafish. Journal of Fish Biology 18(6): 715 to 721. The late end of the fish growth inhibiting factor literature, which traces to essentially one laboratory and stops here. An independent check of that literature found four unpublished student theses in its citation trail and no isolation of the factor in the fifty five years since. DOI 10.1111/j.1095-8649.1981.tb03813.x
- Martins, C.I.M., Ochola, D., Ende, S.S.W., Eding, E.H. and Verreth, J.A.J. (2009). Is growth retardation present in Nile tilapia cultured in low water exchange recirculating aquaculture systems? Aquaculture 298(1-2): 43 to 50. The size dependent result, including small fish growing better in high accumulation water. DOI 10.1016/j.aquaculture.2009.09.030
- This hub's own research files. Materials and water chemistry for the carbonate ceiling, the units, the ammonia and nitrate figures and the interaction summary. The bacteria comparison for heterotroph and nitrifier yields, the Nogueira burial result and the Michaud carbon to nitrogen numbers. The Armenia file for tap water, plumbing and the polluted river catchments. All references, by topic
- Richards, C.M. (1958), West, N.B. (1960) and Akin, G.C. (1966). Three papers in Physiological Zoology on growth inhibition in crowded Rana pipiens tadpoles and the nature of the inhibitory material. The amphibian result is real and replicated. It is in tadpoles, not in fish, and borrowing it for fish is where the hobby claim came from. DOI 10.1086/physzool.31.2.30155387
- Mota, V.C., Limbu, P., Martins, C.I.M., Eding, E.H. and Verreth, J.A.J. (2015). The effect of nearly closed recirculating systems on the feed intake and growth of Nile tilapia, African catfish and European eel. Aquacultural Engineering 68: 1 to 5. The 42 percent feed intake and 83 percent growth reduction in tilapia, its absence in the other two species, and the authors' statement that the cause is unclear. DOI 10.1016/j.aquaeng.2015.06.002
- Mota, V.C., Martins, C.I.M., Eding, E.H., Canario, A.V.M. and Verreth, J.A.J. (2014). Steroids accumulate in the rearing water of commercial recirculating aquaculture systems. Aquacultural Engineering 62: 9 to 16. Seven commercial systems, the cortisol and sex steroid ranges, and the correlation between low make-up water use and higher sex steroids. DOI 10.1016/j.aquaeng.2014.07.004
- United States Environmental Protection Agency. National Primary Drinking Water Regulations, nitrate maximum contaminant level, 10 mg/L as nitrate nitrogen. Cited as the origin of the figure the hobby repeats, and as the reason it is the wrong number for a fish tank: wrong species, wrong endpoint. epa.gov
- Camargo, J.A., Alonso, A. and Salamanca, A. (2005). Nitrate toxicity to aquatic animals: a review with new data for freshwater invertebrates. Chemosphere 58(9): 1255 to 1267. The 2 mg of nitrate nitrogen per litre recommendation and the 4.43 conversion to what a test kit reads. DOI 10.1016/j.chemosphere.2004.10.044
- Walstad, D. (2023). Ecology of the Planted Aquarium, 4th edition. ISBN 979-8-9875741-0-2. The six month figure and its two named ongoing tasks, the submerged soil chapter including phosphate release from dissolving iron oxides, cation exchange capacity, and the fish food nutrient argument. Read through this hub's own Walstad research file, which cites the book's page ranges. dianawalstad.com
- Walstad, D. (20 April 2026). "Fertilizing Substrates." dianawalstad.com blog. Open access. The account of pushing fertiliser into an established substrate and doing a 30 percent water change the same day. dianawalstad.com
Continue