A new aquarium is not ready for fish on the day it is filled. It has to grow a population of bacteria first, and until that population exists the fish are swimming in their own ammonia. This is the single most common way a first aquarium kills its fish, and it happens on a schedule that is well documented.
Fish excrete ammonia across their gills, and uneaten food and waste release more. In an established tank two groups of bacteria living in the filter deal with it: the first turns ammonia into nitrite, the second turns nitrite into nitrate. Nitrate is much less harmful and you remove it by changing water.
In a new tank neither group is there in useful numbers yet, so ammonia climbs first. Merck's veterinary reference dates it: "NH3 levels rise during the first 2-3 weeks after a new system is set up." Nitrite then climbs as the first group gets going and the second has not caught up. Only when both readings sit at zero, while you are still feeding, is the tank actually running.
| Who | How long |
|---|---|
| Merck Veterinary Manual, management of aquarium fish | "It can take up to 8 weeks for a biofilter in a tropical fish tank to become established." |
| RSPCA Australia | Four to six weeks |
| Merck Veterinary Manual, environmental diseases | Ammonia rising over the first two to three weeks |
| Merck Veterinary Manual, disorders and diseases of fish | New tank syndrome within one to three weeks |
Merck disagrees with Merck, on three of its own pages. That is not a scandal, it is what happens when different pages are written about different things: how long until the filter is fully mature is a longer answer than how soon the fish start suffering. The practical version is that the dangerous window opens in the first week or two and the tank is not finished for a month or more, so the only honest instruction is to test rather than to count days.
Almost everything written about aquariums names Nitrosomonas and Nitrobacter as the two groups doing the work. Extension fact sheets still print it that way, including the one on starting a biofilter.
In 1998 a group of researchers did something nobody had done: they sampled the filters of actual freshwater aquariums and looked for the bacteria. Their finding, published in Applied and Environmental Microbiology, was that nitrite-oxidising bacteria of the kind that includes Nitrobacter "were not detected in these samples". What they found instead were Nitrospira-like bacteria. Oklahoma State's extension service now lists four genera rather than one.
This is worth knowing for one practical reason. Products sold to start a tank quickly are not all carrying the same organisms, and the older literature has not caught up with the aquarium research. It changes nothing about what you do; it does tell you how much of the standard account has ever been checked.
The bacteria live on the filter media, not in the water. So washing the media properly is the usual way people restart the whole process by accident. OATA puts it in one line: "Never rinse them under a tap, since this will wash away and kill the beneficial bacteria." Rinse media in water taken out of the tank, and do not replace all of it at once.
The other way to lose it is slower. Nitrification produces acid and uses up the water's buffer, so a tank that never gets a water change eventually loses its ability to hold pH steady, and then the pH collapses. Merck describes that as "a sudden and drastic drop in pH, often below 6.0, which kills bacteria in the biofilter". The tank then behaves like a new one, except with a full load of fish in it.
| Who | What they publish |
|---|---|
| RSPCA, UK | "You must also carry out weekly 30% water changes." |
| OATA | Up to 25% every week |
| RSPCA Australia | 10 to 25%, weekly |
| Merck Veterinary Manual, during an ammonia problem | At least 50% of the water changed |
Every one of those is a reasonable weekly number and none of them is derived from anything you can check. What the readings say matters more than which figure you picked. The UK government's guidance for pet sellers requires that water quality "must be tested and recorded ... at least weekly", which is the closest thing to a rule anyone has written down.
Stop feeding for a few days, because the food is where most of it is coming from. Change water. Test before and after rather than guessing. And put the reading through the ammonia calculator with your pH and temperature, because if your pH is high the same number is far more dangerous than it looks.