Why your KH keeps sliding, and what actually holds it up

KH falls because your tank consumes carbonate faster than water changes put it back. The steadiest drain is biological filtration: every time your bacteria oxidise ammonia to nitrate they release acid, and neutralising that acid burns roughly 7.1 g of CaCO₃-equivalent alkalinity for every gram of ammonia-nitrogen processed. Aqua-soils, driftwood, botanicals and low-KH source water all pull the same way. CO₂ injection, despite the folklore, does not consume KH. Once you slip below about 2 °dKH the buffer is nearly spent and pH can crash overnight — and the fix is either matching your water-change cadence to the loss, or dosing sodium bicarbonate at roughly 1 g per 30 L to add 1 °dKH.

That's the short version. The longer one is worth reading, because the reason "just do more water changes" sometimes fails to hold KH up is the same reason people file this under mystery in the first place.

The buffer only ever goes down

KH is the reserve of carbonate and bicarbonate in your water, and its one job is to absorb acid before that acid can move the pH. It isn't the same thing as GH — GH is the calcium and magnesium your livestock builds shells from, KH is the buffer, and the two move independently (the GH vs KH guide untangles that if you're still conflating them). What matters here is direction: a running tank is a net acid producer. Fish respire CO₂. Bacteria make nitrate. Wood and leaves leach organic acids. Every one of those inputs spends buffer, and nothing in a normal freshwater tank puts carbonate back on its own. So KH, left alone, only ever falls. Water changes are the resupply. When the tank spends buffer faster than the changes deliver it, the line trends down — and that's the whole story, dressed up.

The main drain: nitrification

The biggest steady consumer is the process you spent your first month waiting for. Ammonia oxidation runs, in net terms, like this:

NH₄⁺ + 2 O₂  →  NO₃⁻ + H₂O + 2 H⁺

Two hydrogen ions come off for every ammonia molecule processed. Each of those H⁺ gets neutralised by a bicarbonate ion pulled out of your KH:

H⁺ + HCO₃⁻  →  H₂CO₃  →  CO₂ + H₂O

So two units of buffer vanish for every unit of ammonia the filter handles. Run the arithmetic in CaCO₃ terms — one gram of ammonia-nitrogen is 1/14 of a mole, times two protons, times the 50 g equivalent weight of CaCO₃ — and it lands at 7.14 g of alkalinity consumed per gram of ammonia-N oxidised. That number is fixed by the chemistry, not by your tank.

What varies is how much ammonia your tank actually processes. A heavily stocked, lightly planted community with a busy filter turns over a lot of nitrogen and watches KH tick down week on week. A low-bioload planted tank barely registers it, partly because there's less waste and partly because plants take up ammonium directly, sidestepping some of the acid before the bacteria ever see it. And during a fish-in cycle, when the bacterial colony is still catching up, the KH drain can be sharp enough on its own to trigger the classic new-tank pH crash — especially if you started with soft water and no buffer to spare. If cycling is still fresh in your mind, the nitrogen cycle in plain language covers where that ammonia is coming from.

Three more drains worth knowing

Nitrification is the constant. These three explain most of the rest.

• Active substrates. Aqua-soils like ADA Amazonia and its many cousins strip KH by ion exchange, and they do it on purpose — the whole point of the substrate is to pull the water soft and slightly acidic for plants and shrimp. In its first months a soil tank can drag KH to near zero no matter what you put in. That's not a fault. But you're now running with essentially no buffer, so pH stability rides entirely on the soil and your water-change discipline.
• Driftwood and botanicals. Tannins, humic and other organic acids nudge pH down and slowly spend buffer as they do it. A tank full of catappa leaves and alder cones is a steady, low-grade acid feed. Lovely for a blackwater look; quietly corrosive to your KH.
• Low-KH source water. This is the one people miss. If the water you change with carries little or no carbonate — RO, or naturally soft tap — then a water change stops replenishing KH and merely holds it wherever it already sits, or dilutes it further. So you change water faithfully, and KH still falls, and it feels like the tank is haunted. It isn't. Your fresh water is as soft as the tank.

The CO₂ myth, put to bed

Because injected CO₂ drives pH down through the photoperiod, a lot of keepers assume it must be eating KH. It isn't. CO₂ forms carbonic acid, shifts the carbonate equilibrium and lowers pH — but the bicarbonate is still sitting in the water. Turn the gas off and pH climbs straight back up with KH unchanged. What CO₂ genuinely does is widen the daily pH swing, and a low KH lets that swing run wider than the fish would like. So CO₂ isn't a cause of the drop, but a tank that's already low on KH is a worse place to inject — the CO₂ from pH and KH guide shows exactly how tightly the three are locked together.

What the slide looks like

The symptom ladder runs slow, then fast. First KH creeps down over weeks, which you'll only notice if you're logging it. Then pH — which the KH was holding up — starts to sag: the overnight low drifts first, then the whole baseline. Below roughly 2–3 °dKH the buffer is effectively spent, and the next ordinary acid input — a big feeding, a filter that clogs, a CO₂ solenoid stuck on overnight — can drop pH a full point or more in a matter of hours. It's that speed, not the low pH itself, that harms fish. The crash is the symptom. The buffer that drifted down over the preceding weeks is the cause, and it's visible well ahead of time on the trend chart if you've been writing KH down. The Forecast is built to make exactly that slow slide obvious before it becomes an event.

The fix, part one: match water changes to the loss

If your source water carries KH, the simplest fix is the boring one: change enough water, often enough, to replace what nitrification burns. To size it, log KH just before a change and again a few days later — the drop between those two readings is your consumption rate. Then make the change big enough, or frequent enough, to cover it. The water-change impact tool shows how far a given swap moves KH, and it makes the source-water trap visible too: a change only lifts KH if the new water is harder than the tank. If your tap or RO mix is soft, water changes alone will never raise the number — they'll just stop it dropping. At that point you have to add carbonate yourself.

The fix, part two: the bicarbonate maths

Sodium bicarbonate — plain baking soda, NaHCO₃ — raises KH cleanly and cheaply without touching GH. The anchor figure:

About 1 g of NaHCO₃ per 30 L raises KH by roughly 1 °dKH (technically closer to 1.1, near enough for dosing).

Scaled to a round tank size:

KH rise	NaHCO₃ per 100 L
+1 °dKH	~3 g
+2 °dKH	~6 g
+3 °dKH	~9 g

Dissolve it in a cup of tank water first, add it over the course of a day, and don't chase big jumps — hold to no more than 1–2 °dKH per day so pH doesn't lurch as fast as it would in a crash. If you run RO or soft tap, a KH+ or all-in-one GH/KH+ remineraliser does the same job dosed to a target, and it's easier to repeat consistently on every change than weighing out baking soda. The passive option is crushed coral or aragonite in the filter: it dissolves as pH falls and self-limits once the water stabilises, which suits a tank that drifts slowly rather than one that's already crashing. If you need to move between °dKH and ppm along the way, the hardness converter does the arithmetic.

One caveat, because it catches well-meaning people: if you deliberately keep soft-water species — discus, caridina shrimp, wild Apistogramma — don't reflexively pump KH up. There is no universal target; a discus tank runs happily at 1–3 °dKH, a Tanganyika tank wants 12–18, and both are correct for what lives in them. The danger isn't low KH as such. It's KH low enough that pH stops being stable. For a deliberate soft-water tank the fix is soft source water plus a cadence that keeps KH from riding all the way to zero — not a spoon of baking soda that undoes the biotope you set out to build.

Catch it before the crash

All of this closes with the same habit. Test KH weekly, or at least on every water change, and watch the slow line rather than the single reading. Match a water-change reminder to the cadence your logged consumption rate actually calls for. The whole point of writing the number down is that a drifting buffer announces itself weeks before it becomes a pH event — the drop is legible long before the crash, and legible is fixable.