Why an Aquarium pH Falls While You Change the Water

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Because a water change does not add pH. It adds whatever alkalinity your tap water happens to carry, and if that is soft water you are refilling a bucket with a thimble. Meanwhile the filter is destroying alkalinity on a perfectly steady schedule, set not by the fish or the tank size but by how much food goes in.

The number is 7.145 grams of alkalinity destroyed for every gram of ammonia nitrogen that passes through the filter, and it is not an estimate. It falls out of the reaction. Once you have it, the whole thing becomes a budget, and the reason the fall looks sudden becomes obvious too.

Nitrification is an acid reaction, and that is the entire story

The bacteria in a filter convert ammonium to nitrate, and people usually describe that as removing a toxin. It does, and it also does something nobody mentions. Write the reaction out:

Ammonium plus two oxygen gives nitrate plus two hydrogen ions plus water.

Two hydrogen ions per nitrogen atom. Each one is neutralised by one equivalent of bicarbonate, which is what your test strip calls alkalinity or carbonate hardness, and once neutralised it is gone. So the filter is a slow, steady acid generator whose output is locked to its input.

Turn that into working numbers. One mole of nitrogen weighs 14.007 grams and consumes two equivalents of alkalinity, each equivalent being 50.04 grams expressed as calcium carbonate. Divide: 100.08 over 14.007 is 7.145 grams of alkalinity per gram of nitrogen. The same reaction eats two moles of oxygen, which is 63.998 over 14.007, or 4.569 grams of oxygen per gram of nitrogen. Both figures match the ones used in wastewater engineering, 7.14 and 4.57, which is a reassuring check on the arithmetic.

Where the nitrogen comes from, and how much

It comes from the food, in a fixed proportion. Dry fish food runs around 40 percent protein, and protein is about 16 percent nitrogen by weight, so a gram of food carries roughly 0.064 grams of nitrogen. Feed a 29 gallon tank, which is 110 litres, two grams a day and you are putting 0.128 grams of nitrogen into it daily.

Multiply through. That nitrogen destroys 0.128 times 7.145, which is 0.915 grams of alkalinity as calcium carbonate every day. Spread through 110 litres, that is 8.32 milligrams per litre per day disappearing from the buffer.

That rate is the answer to the question, because it explains the shape of what you observed. Alkalinity falls smoothly and pH does not move at all while there is buffer left, which is what buffers are for. Then the buffer runs out and pH falls fast. The cause was steady for weeks and the symptom arrived in a day, which is exactly why people look for a recent change and find none.

How long the buffer lasts, which depends entirely on your tap

Soft municipal water might carry 30 milligrams per litre of alkalinity as calcium carbonate, which is about 1.7 degrees of carbonate hardness. Moderately hard water is nearer 100, and hard water around 200.

Divide each by 8.32 milligrams per litre per day. Soft water gives you 3.6 days. Moderate gives 12.0. Hard gives 24.0.

Now put the weekly maintenance interval next to those. On soft water the buffer is exhausted on day four, and every water change arrives three days after the crash it was supposed to prevent. On hard water you have three weeks of headroom and will probably never see the problem at all. Same tank, same fish, same filter, same feeding. Two completely different outcomes, decided by the pipe.

The water change arithmetic, which is where the premise breaks

This is the part worth working out properly, because it says what to do rather than just what is happening.

Over a week the tank destroys seven times 8.32, which is 58.2 milligrams per litre of alkalinity. A water change cannot create alkalinity. It can only import whatever the replacement water has. So the share of the tank you need to exchange each week to break even is 58.2 divided by your source water's alkalinity.

On hard 200 milligram water that is 29 percent of the tank. On 100 milligram water it is 58 percent. On soft 30 milligram water it is 194 percent, which is to say nearly twice the entire tank volume, every week, indefinitely.

Look at the first of those three. Twenty nine percent weekly is the standard maintenance advice, almost to the point. That advice is not a pH treatment and was never derived as one. It is an alkalinity top up that happens to balance exactly on hard municipal water, which is what most of the people who wrote it were using. On soft water it falls short by a factor of 7.8, and no amount of doing it more diligently closes that gap, because the shortfall is in the water and not in the schedule.

Our own guide to testing aquarium and pond water at home is right about the important thing, which is that trends in a logbook beat any single reading, and it names hardness and alkalinity among the parameters to watch. The one line worth correcting is its advice that "off ph may point to a needed adjustment". Adjusting pH directly is the one thing not to do here. pH is the symptom; alkalinity is the quantity, and a tank whose alkalinity is at zero will reject any pH adjustment you make within a day.

If you want a running record rather than a weekly one, a kactoily digital aquarium water tester with app alerts reads pH continuously and flags drift, which is the right instrument for a quantity that moves slowly and then quickly. Its own drawbacks are honest about what it does not tell you: the listing does not state probe lifespan, the calibration routine, or accuracy figures. For spot checks, a pack of Aquatic Eden 7-in-1 pond and aquarium test strips covers alkalinity alongside pH, and its record says plainly that the "Color bands are approximate, not exact". Both of those admissions matter more than they look, because the quantity you are tracking here changes by about 8 milligrams per litre a day and a strip's colour bands are wider than that.

What actually fixes it, and our catalogue already has it

Add carbonate instead of adding water. A Gstrmydz crushed coral filter media bag is 1.1 pounds, or 500 grams, of coral, and its record describes the mechanism correctly: "Coral slowly releases carbonate, which the maker says helps stabilise pH levels".

Weigh that against the consumption rate. Five hundred grams of calcium carbonate, against 0.915 grams destroyed per day, is 546 days. One bag carries roughly eighteen months of this tank's alkalinity loss. Its record's drawback that "500 grams may suit only smaller tanks" is measured against the wrong variable: what consumes alkalinity is the feeding rate, not the water volume, and a lightly fed 75 gallon tank runs through less of it than a heavily fed 20.

The other drawback on that record, "Coral buffers gradually, not instantly", is the most important sentence on the listing and reads like a weakness when it is the whole point. Calcium carbonate dissolves faster as water gets more acidic and effectively stops once pH is up around 8. That is a self regulating supply: it releases buffer when the tank needs it and sits inert when it does not. A measured dose of anything else cannot do that, which is why coral works in a filter chamber and a bottle of pH adjuster does not.

The honest limit on it is that 546 days is a capacity, not a rate. How fast the bag can deliver depends on how much surface area is exposed, how much water passes through it and how low the pH has gone, and no listing we carry publishes grain size, surface area or a dissolution rate. If the pH is already falling fast, more surface area in the flow is the lever, not more mass in the bag.

The second question, answered by the same equation

The companion question people ask is when more biological filtration becomes pointless, and the arithmetic above settles it, because the bacteria are limited by their food supply rather than by their real estate.

Nitrifying biofilm removes something in the region of half a gram of ammonia nitrogen per square metre per day at tank temperatures. Ordinary reticulated filter foam offers of order 1,000 square metres of surface per cubic metre. So handling 0.128 grams of nitrogen a day needs about 0.256 square metres, which is about 256 cubic centimetres of foam. A sponge six and a half centimetres on a side.

That is the whole nitrogen load of a stocked 29 gallon tank, dealt with by a quarter of a litre of ordinary sponge. Every filter in the category has an order of magnitude more than that, so adding media beyond the first block does nothing: the extra surface has no ammonia reaching it to colonise. A biological filter cannot be made stronger by being made bigger, only by being fed more, and you do not want to feed it more.

A fishkeeper 4-in-1 submersible aquarium internal filter describes its own media accurately: a "biochemical sponge that captures waste and grows beneficial bacteria", washable and reusable, with room to add ceramic rings. The sponge is doing the biology and the ceramic rings are optional for exactly the reason above.

What its record does get wrong is the tank range, and it half admits it. It publishes 135 gallons per hour, which is 511 litres per hour, for "80 to 300 gallon" tanks. At the bottom of that range, 80 US gallons is 303 litres, which is 1.69 turnovers an hour. At the top, 300 gallons is 1,136 litres, which is 0.45. Against the conventional four turnovers an hour for an in tank filter those are short by 2.4 and 8.9 times. Its own drawbacks say "135GPH is modest for the top of its range" and its not-for line says "heavily stocked large tanks needing high turnover", so the record flags its own gap, which is to its credit. The 300 in the headline range is still indefensible.

And turnover is the wrong rating anyway, which is the deeper point. An AQQA 581GPH canister filter with 3-stage filtration moves 2,199 litres an hour on 35 watts, four turnovers an hour on a 145 gallon tank, and that figure tells you about particle capture and water movement rather than about nitrogen. Its drawbacks admit that "Canister volume and media amount are unlisted", and media volume is the only figure that could settle a biological filtration question. Not one filter listing in our catalogue states an ammonia load in grams per day, which is the single honest rating a biofilter can carry. They all state gallons and gallons per hour, which are the two things the bacteria do not respond to.

The oxygen side, which our own page already owns

One more line falls out of the same reaction and it is worth seeing. At 4.569 grams of oxygen per gram of nitrogen, this tank's nitrification consumes 0.585 grams of oxygen a day. The whole 110 litres holds about 0.909 grams of dissolved oxygen when saturated at room temperature. So the filter bacteria alone would claim 64 percent of the tank's entire dissolved oxygen inventory every day, before a single fish breathes.

They do not suffocate anything, because the surface resupplies it continuously, and our page on whether bubbling actually adds oxygen to a fish tank works out that resupply properly and is right that the mechanism is surface renewal rather than the bubbles themselves. This page will not repeat it. The useful connection is that the same reaction sets both numbers, so a tank with an alkalinity problem is, by arithmetic, also a tank with an oxygen demand proportional to its feeding.

It is the same structural answer our page on why a 5 gallon and a 15 gallon tank take the same 50 watt heater reaches about heating: the quantity that sets the requirement is not the one printed on the box. There it is surface area rather than volume. Here it is grams of food rather than gallons.

The honest limit

Two soft inputs and one hard one.

The soft ones are the nitrogen accounting and the biofilm rate. Not all the nitrogen you feed reaches the filter as ammonia: a well fed tank excretes perhaps 70 percent of it and retains or settles the rest, so the 0.128 grams a day above is a ceiling and 0.090 is more realistic. That stretches every day count on this page by about a factor of 1.4, turning 3.6 days into 5.2 and 24 into 34. It does not change the conclusion, because the weekly exchange fraction on soft water is still 136 percent of the tank rather than 194. The half gram per square metre per day for nitrifying biofilm and the 1,000 square metres per cubic metre for filter foam are both engineering ranges rather than measurements, and either could move by a factor of two or three. The order of the sponge answer survives that; a quarter litre becoming three quarters of a litre is still a block of foam.

The hard one is the 7.145. It is a ratio of atomic weights in a balanced equation and it does not move for anybody's tank, water or filter.

Which gives a diagnostic that needs no equipment beyond a strip. Test your tap water's alkalinity, not just the tank's. If the tap reads low, no schedule of water changes will hold pH and the whole problem is an import problem. If the tap reads high and the tank reads low anyway, you are feeding more than the exchange brings in, and the lever is the feeding rate, since that is the only term in the equation you control.