How often to change aquarium water, and why your nitrate already knows the answer

For most stocked freshwater tanks the answer is 25–50% weekly. The honest version is one level better: the right cadence is whatever keeps your nitrate inside its band — 5–25 mg/L for a planted community tank — and that's a number you measure, not guess. If nitrate sits inside the band on the day before your water change, your routine is working, whatever the percentage on paper says. If it creeps past the top of the band week after week, you need more water or more often, and the rest of this guide is the working-out.

The reason this question produces so many contradictory answers is that every shop, bottle, and forum thread hands out a different sticker number: 10% weekly, 30% fortnightly, 50% monthly. They can't all be right, and in fact none of them are, because they're defaults for a tank the author has never seen. Your cadence depends on two things only: how fast your tank loads the water with waste, and how much of that water you move. Both are measurable. Neither is printed on a sticker.

What a water change actually does

A water change is dilution, nothing more mystical than that. Swap a fraction p of the tank volume and you remove that same fraction of everything dissolved in it that your new water doesn't carry — nitrate, phosphate, dissolved organics, tannins, the lot:

C_after = C_before × (1 − p)

A tank at 40 mg/L nitrate, after a 50% change, sits at 20 mg/L. That's the whole trick.

Nitrate is the clean case because tap and RO water normally arrive nitrate-free — though it's worth testing your tap once, because some agricultural regions run 20 mg/L or more straight out of the pipe, and then your "water change" is quietly a nitrate top-up. Minerals like GH and KH play by a different rule: they don't dilute toward zero, they average toward whatever your source water carries, which is why a change can push them up as easily as down. The full second-term maths lives in the bigger water change guide; for the cadence question, nitrate is the number to watch.

Worth being clear about what a water change is not, too. It doesn't fix an immature filter, it doesn't cure algae whose cause is a light-and-nutrient imbalance, and it doesn't remove ammonia production — it just dilutes the consequences for a few days. It's housekeeping, and housekeeping works on a schedule.

The number that decides your cadence

Everything hangs on one figure: how much nitrate your tank adds per week. Call it A. Measuring it costs two tests — one right after a water change, one just before the next. The difference is your weekly input. A lightly stocked planted tank might add 3–5 mg/L a week; a busy community tank 10–15; a tub of growing cichlids can manage 30.

Once you know A, the long-term behaviour of a weekly routine is fixed. Each week the tank gains A, then a change removes the fraction p, and over a month or two the level converges on a sawtooth bouncing between a steady high and a steady low:

peak ≈ A / p          trough ≈ A × (1 − p) / p

Take a community tank adding 10 mg/L a week. On 25% weekly changes it settles into a 30 → 40 mg/L sawtooth — permanently above the band, no matter how faithfully you keep the schedule. On 50% weekly the same tank rides 10 → 20 mg/L, comfortably inside. Doubling the change size roughly halves where the tank lives, because you're balancing a continuous input, not clearing a fixed pile.

You can also run the formula backwards, which is the actually useful direction: to hold the peak at the top of the band, you need p ≈ A ÷ 25. Ten mg/L of weekly input means a 40% weekly change. Five means 20%. Thirty means either 50% twice a week or a hard look at stocking and feeding. The water-change impact tool runs this arithmetic interactively for your tank volume, if you'd rather slide a number than solve for it.

Small and often beats big and rare

Here's a subtlety the steady-state maths hands you for free. Split the same weekly volume into two smaller changes — say two 25% instead of one 50% — and the nitrate peak barely moves. What changes is the ride: each swing is half the size, the tank never drifts as far from baseline, and every individual change is a smaller environmental step for the livestock to absorb.

That last part is the one fish actually feel. A water change never moves nitrate alone — temperature, GH, KH, and TDS all shift at once, and animals osmoregulate against the water they're sitting in. A slightly elevated but stable nitrate is far kinder than a perfect average delivered in violent weekly lurches. So the two levers separate cleanly: if the peak is too high, increase the fraction; if the swings are too rough, increase the frequency. Most mature tanks I've kept ended up at a boring, moderate weekly change precisely because that's where both levers relax.

Big-and-rare fails on a second front: life gets in the way. A 60% monthly routine that slips a fortnight leaves the tank six weeks deep in accumulation; a weekly 30% that slips a week is barely a blip. Routines survive on being small enough that you actually do them.

When the big bucket is right

There are genuine cases for a large change, and they're corrections, not routines. A test that comes back at 80+ mg/L on a stocked tank. A dead fish found late, an overfeeding accident, a stirred-up substrate. The end of a medication course. In all of those, dilute the problem generously, then return to the steady cadence — the bigger water change guide covers the sizing and the traps in detail.

The one warning worth repeating here: a tank that's gone months without changes has drifted far from your tap in pH, KH, and TDS, and the livestock has drifted with it. One heroic 80% change slams them back across that whole gap in an afternoon, and that shock kills fish that the bad water hadn't. Re-approach a neglected tank with a series of 10–15% changes over a week or two, not one purge.

The special cases

Heavily planted, lightly stocked tanks can stretch the cadence — sometimes to a fortnight or beyond — because the plants are consuming nitrogen faster than the fish produce it. But watch the floor as well as the ceiling: below 5 mg/L nitrate the plants themselves start starving, and plenty of planted keepers end up dosing nitrate in rather than changing it out. And note the odd exception: high-tech tanks fertilised by the Estimative Index do a 50% weekly change as part of the method, to reset accumulating fertiliser — that change is about the dosing, not the fish.

Shrimp tanks want small and slow. Dwarf shrimp are more sensitive to the swing than to the average, and a big fast change is a classic trigger for failed moults. 10–15% weekly with temperature-matched water, refilled gently, is the usual shrimp-keeper's rhythm — the shrimp-safe parameters guide has the full picture.

New tanks don't get a cadence at all — they get a trigger. In the first four to eight weeks the filter isn't mature, and the numbers that matter are ammonium and nitrite, not nitrate. With livestock in the tank, any reading above 0.1 mg/L NH₄ or NO₂ means change water today; 0.25 mg/L NH₄ or 0.5 mg/L NO₂ is an emergency, not a diary entry. That can mean changes several times a week early on, tapering as the nitrogen cycle establishes. Only once both sit at zero for a couple of weeks does the nitrate-based routine above take over.

Starting points, not commandments

Until you've measured your own weekly input, these are sane places to start:

Tank	Starting cadence	What to watch
Heavily planted, light stock, no CO₂	10–20% every 1–2 weeks	the nitrate floor — don't starve the plants below 5 mg/L
Standard community, moderate stock	25–30% weekly	nitrate the day before the change
Heavily stocked, big or messy feeders	40–50% weekly, or 2 × 25–30%	nitrate and phosphate peaks
Shrimp-focused tank	10–15% weekly, slow refill	the per-change GH/TDS swing
High-tech planted, EI dosing	50% weekly (part of the method)	the fertiliser reset, not the fish
New tank, first 4–8 weeks	on demand, test-driven	NH₄/NO₂ above 0.1 mg/L → change today
Grow-out or fry tank	30–50%, two or three times weekly	growth and appetite as much as nitrate

Treat the table as day one, not gospel. Two months of before-and-after tests replaces every row with a number that's actually yours.

Write it down

The cadence question answers itself embarrassingly quickly once the readings are in one place. A dozen logged nitrate tests show you the sawtooth, the sawtooth shows you the peak, and the peak either sits inside the band or it doesn't. I log mine in reefnotes — it's free and runs in the browser — where the trend chart draws the Forecast a few weeks ahead and the scheduled water-change tasks keep the cadence honest on the weeks when motivation doesn't. If reading that chart is new territory, the trend and Forecast guide walks through what the line and the band around it mean.

However you record it: measure the rise, size the change, and let the sticker numbers stay on the sticker.