Aquarium water changes: why they matter

This page continues

Start with the routine here, then open the special case that matches your tank.

Old tanks and low-change tanks

Rescue changes, established planted systems, and the special cases that do not fit a simple weekly routine.

The accumulation arithmetic

Evaporation, tap-water top-offs, and the ceiling set by a repeated partial water change.

What a water change removes

Alkalinity, dissolved organics, metals, trace depletion, and medications in one reference.

A canister filter with a drain hose — A canister filter connected to a drain hose and bucket.Fish Tank Society · CC BY-SA 4.0 · Resized; uncropped.

The framing

Zero on the kit is a statement about one element, in one form, in one place.

A nitrate reading of zero says that for nitrogen, in the nitrate form, in the water column, uptake is at least as fast as supply. In a well planted tank that is a real achievement and it deserves to be said. It is also very narrow.

It says nothing about the elements plants do not want. Sodium and chloride have no biological sink in a freshwater tank at all. Nothing takes them up and nothing evaporates them out. It says nothing about the nitrogen that never became nitrate, or the carbon that never became carbon dioxide, or the alkalinity that nitrification destroyed on the way to making that nitrate.

A hobby test kit measures about six things. Ammonia, nitrite, nitrate, phosphate, general hardness and carbonate hardness. A laboratory analysis of your tap water measures about thirty. The gap between six and thirty is what a water change is for.

The practical routine farther down works across that whole gap. If you want the chemistry behind each part, the focused guide to what a water change removes keeps it separate from the weekly decision.

What one gram of dry flake a day puts into 100 liters²

22: mg/L of nitrate a week, if all of it mineralises. Reading zero means the plants remove this much²
2.6: mg/L of phosphate a week²
365: mg/L a year of mineral ash. Sodium, chloride, calcium, sulfur. Not a test kit parameter²
0: of that ash is removed by a plant, by a filter, or by evaporation

Calculated from a typical tropical flake at 45 percent crude protein and 10 percent ash². Protein divided by 6.25 gives nitrogen, nitrogen times 4.43 gives nitrate as a test kit reads it³. The ash figure is the one nobody looks at. Fish food is sold on its protein content, and the mineral content is the part that stays.

The practical answer

How much, how often, and on what evidence.

Here is the evidence this rests on, strongest first. Nobody has ever run a controlled study of water change frequency in a planted home aquarium⁴, so what follows is five other kinds of evidence, and they are not equal.

What the recommendation below is actually built on
Source of the number	What it can and cannot tell you
The arithmetic on this page¹	Certain, and the only certain part. Gives a number that depends on your evaporation rate
The recirculating aquaculture literature⁵ ⁶ ⁷	Real experimental work on water re-use, with controls. Run at feed loads far above any ornamental tank, so the direction transfers and the concentrations do not
The nitrate toxicity literature	Gives a target, not a schedule. A recommended maximum of 2 mg of nitrate nitrogen per liter for the most sensitive freshwater species, which is about 9 mg/L on a kit that reads nitrate³. Use that number and not the other one. The 10 mg/L figure the hobby repeats is the United States drinking water standard for human infants⁸
The Estimative Index dosing system	The origin of the 50 percent weekly rule⁹. It is part of a dosing system, not a hygiene recommendation. Dose everything in excess, then reset weekly so nothing builds up. If you are not dosing that way, the number was never addressed to you
Soil and substrate manufacturer instructions	Heavy and front loaded, because a fresh soil leaches ammonium. Correct for weeks one to eight, and not a general schedule
Laboratory husbandry standards	Worth more than any hobby rule, and expressed in a different unit. Published zebrafish facility practice is 8 to 10 percent of system volume per day¹⁰, about 44 to 52 percent a week in equivalent exchange. Run through the continuous version of the model above, that holds a tank at 1.07 to 1.14 times its source water, which is the same ballpark as a 30 to 50 percent weekly change¹. It is the only independent confirmation this page has

Everything else is convention. Measure your evaporation rate with the tape on the glass, decide what multiple of your tap water you are willing to live at, and read the table.

The recommendation, by kind of tank
Tank	Typical evaporation	Change	Holds the tank at
High tech, injected carbon dioxide, Estimative Index dosing, high light	5 to 8 % a week¹	50 % weekly, and it is not optional⁹	1.05x tap, and a bounded fertiliser build-up¹
Ordinary planted community tank, lidded	1 to 2 % a week¹	20 to 25 % every two weeks, or 30 % monthly¹	About 1.2 to 1.3x tap¹
Ordinary planted community tank, open top	about 5 % a week¹	20 to 25 % weekly¹	About 1.2x tap¹
Heavily planted, lightly stocked, low tech, lidded	1 to 2 % a week¹	25 to 30 % monthly¹	About 1.3x tap¹
Soil tank, first eight weeks	any	25 to 50 % weekly, sometimes more	Handles the soil leaching, not the ions
Caridina shrimp	any	Whatever holds dissolved solids steady, and top up with reverse osmosis water only	Stability matters more than the number
Long neglected tank	any	Test total ammonia first. Zero means change as much as you like¹¹. Above zero means the filter has failed, so fix that and stage at 20 to 25 % every day or two	The staging is for the ionic shock, not for the ammonia. See the section above

The single most defensible general answer, for a normal planted tank on Yerevan tap water, is 20 to 25 percent weekly, or 30 to 40 percent fortnightly.¹ Not 50 percent, unless you are dosing the system that requires it⁹. Not zero, unless you top up with reverse osmosis water.

And a caveat on the whole table, because it is the honest one. Every row is a percentage per event, which is the hobby’s unit and nobody else’s. The two communities that actually measure, aquaculture and laboratory husbandry, both express exchange as a rate: liters per kilogram of feed per day⁶, or percent of system volume per day¹⁰. There is a reason for that. A rate is what the arithmetic on this page is about, and a percentage per event only becomes a rate once you also state the interval. The table above is a translation into the unit you already use, not a finding in that unit. If these numbers are ever tested properly, they will be tested as rates.

Top up with distilled or reverse osmosis water

Worth more than doubling your water change frequency, and it is the least known intervention in this hobby. It removes the largest mechanism on the page completely, and it costs the price of a bottle of distilled water.

Put a lid on the tank

A close fitting glass lid can take evaporation from 5 percent a week to 1 or 2. That is a factor of three to five on every number above¹. The cost is that an open rimless tank looks better, which is a real trade-off and should be stated as one rather than ignored.

Feed less

Feeding is the largest controllable input of nitrogen, phosphorus and mineral ash all at once. Halving the food halves the load on every mechanism on this page simultaneously, including the alkalinity one.

In Yerevan, watch July to September

Hardness rises in summer, because harder artesian water is blended into the softer spring supply when demand rises¹². The utility and its critic agree on that mechanism¹³ ¹² and disagree only about whether it matters for health. Between July and September a water change is a larger chemical event than it is in March, and it is also the season when evaporation runs fastest, so the two effects multiply rather than add. Test the new water, not only the tank. From 2027 the operator is obliged to publish hardness and dissolved mineral data for consumers¹⁴, and until then you need a test kit or a meter.

A pothos cutting with exposed roots — A rooted pothos cutting, photographed out of the water.Ddra5202 · CC BY-SA 4.0 · Resized; uncropped.

Pale at the left, dark at the right, week by week. That gradient is the subject of this page, and it is the one thing in a tank a camera cannot see. Media boxes on this site are 5 to 2 and this banner is 16 to 5, both inside the range measured on grand prize contest photographs, which run from 1.9 to 1 up to 3.4 to 1¹⁵.

A water change is the cheapest and least precise intervention in the hobby, and that is exactly why it works.

Every other tool fixes one named thing. A phosphate remover removes phosphate. An algaecide kills algae. Activated carbon adsorbs a defined class of organics for a few weeks and then stops. A buffer raises carbonate hardness. Each of them requires you to already know what is wrong, which means you had to measure it, which means it only ever works on the six things a hobby test kit measures.

A water change removes a fixed fraction of everything dissolved in the tank that sits above the concentration in your tap water. All of it, at once, with no diagnosis, no test kit, and without you knowing the name of a single compound involved. It works on the twenty-four things you cannot measure exactly as well as on the six you can. That is not a crude version of a better tool. It is a different kind of tool, and the imprecision is the whole point.

Two limits, so it is not oversold. It cannot take the tank below your tap water, because the equation has a floor at the tap value. If the problem is in your source water, no amount of changing helps and the answer is reverse osmosis. And it removes nothing that is not dissolved in the water column: not the substrate reserve, not the detritus, not what is bound in the biofilm, not what is already inside the animals.

And one last time, because it is the point of the page. A test kit reading zero is good news about one thing. Treating it as good news about the tank is a mistake of logic rather than of chemistry. Having no evidence about twenty-four unmeasured substances is not evidence that they are absent. The water change is what you do because you cannot measure them.

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 1; the hardness unit conversion from 10; 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
Our own calculation, not a published finding. Feed load, and the alkalinity break-even. Named inputs: an assumed tropical flake at 45 percent crude protein and 10 percent ash, which is not a named product’s published analysis; protein divided by 6.25 for nitrogen; the 4.43 nitrogen to nitrate conversion from 3; 7.1 mg of alkalinity destroyed per mg of ammonium nitrogen from 5; 17.8 mg/L per German degree from 10; and the assumption that all dietary nitrogen mineralises, which makes every figure a worst case. The halving for plant nitrate uptake is standard plant physiology and is tied to no citation. How we handle our own arithmetic
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
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
Yamin, G., Borisover, M., Cohen, E. and van Rijn, J. (2017). Accumulation of humic-like and proteinaceous dissolved organic matter in zero-discharge aquaculture systems as revealed by fluorescence EEM spectroscopy. Water Research 108: 412 to 421. The measured accumulation of one protein-like and four humic-like components, faster than in a flow-through control, and their appearance in fish blood. DOI 10.1016/j.watres.2016.11.028
Martins, C.I.M., Eding, E.H. and Verreth, J.A.J. (2011). The effect of recirculating aquaculture systems on the concentrations of heavy metals in culture water and tissues of Nile tilapia. Food Chemistry 126(3): 1001 to 1005. Arsenic, iron, manganese, nickel and zinc rising as water exchange fell, and the counter-finding that tissue levels stayed below food safety limits. DOI 10.1016/j.foodchem.2010.11.108
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
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
Barr, T. The Estimative Index of Dosing. Barr Report. The 50 percent weekly change, and the statement that with it the maximum build-up is about twice the weekly dose. The origin of the hobby's default number. barrreport.com
Guibert, N., Mocho, J.-P., Bally-Cuif, L. and Bedu, S. (2025). Balance between daily water renewal and nitrate levels in a recirculating zebrafish housing system. Zebrafish 22(3): 189 to 194, with Aleström, P. and colleagues (2020), Zebrafish: housing and husbandry recommendations, Laboratory Animals 54(3): 213 to 224. The published laboratory figure of 8 to 10 percent of system volume renewed per day. DOI 10.1177/15458547251370938
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 5; 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 6. Nobody has peer reviewed this, and an earlier version of it was wrong. How we handle our own arithmetic
Golos Armenii, 18 July 2026, and Armenia Today, 20 July 2026. The account that harder artesian water is blended into the softer spring supply when summer demand rises. Cited only for the mechanism, on which the utility and its critic agree. A separate claim in the same dispute, of hardness ten times the norm, is recorded and discounted in the Armenia research file. armeniatoday.am
Veolia Djur (20 July 2026). Statement on drinking water quality monitoring. The utility's own confirmation that it supplies different areas from different sources and that hardness therefore differs between areas. veolia.am
Republic of Armenia sanitary standards. Norms N 2-III-A2-1 for the 1,000 mg/L total mineralisation limit, and Rules 2.1.4.001-23 of 2023, which drops the hardness limit but obliges the operator to publish hardness and dissolved mineral data for consumers from 2027. arlis.am
Our own measurement, from a primary artefact. Contest photograph aspect ratios. , 8 September 2026, Intrinsic width and height read off the shipped img tags of the IAPLC grand prize archive: twelve images running from 1024 by 538, which is 1.90 to 1, out to 540 by 157, which is 3.44 to 1, clustering at 2.4 to 2.9 to 1. Measured, not asserted, and re-measurable from the same markup by anyone. The 5 to 2 ratio used for the image boxes on this site comes from it.

Continue