The Lakes That Exhale, and Who Watches Them

By the time the sun rose on August 22, 1986, every neighbor in the village of Nyos, in Cameroon's northwest highlands, was dead, and so were the cows, the goats, the chickens, and the birds in the surrounding valleys [S1]. The bodies bore no wounds [S1]. There was no smoke, no fire, no flood line [S1]. The number eventually counted was 1,746 [S1]. The killer was the lake itself [S1].

The forensic break came when a joint scientific team lowered sample bottles into the crater's deep water and watched what they brought up fizz violently — the bottom of the lake was charged with carbon dioxide the way an unopened soda is charged with carbon dioxide [S1][S2]. Isotope work pointed the gas back to a magmatic source: a volcanic vent beneath the lake had been pumping CO2 into the bottom layers for centuries [S1].

The physics is the physics of any soft drink. Henry's law says a gas's solubility in a liquid scales with the partial pressure above it; under 200 meters of water, a great deal of CO2 will stay dissolved [S1]. A lake that is both deep and meromictic — permanently stratified, with no seasonal turnover — will keep that loaded bottom layer locked under its own weight indefinitely [S1]. Disturb a parcel of saturated deep water enough to lift it a few tens of meters, and the pressure drop lets it fizz; the fizz makes it buoyant; the buoyancy lifts more; within an hour the whole lake has turned itself inside out [S1].

The trigger at Nyos remains contested — landslide on the inner crater wall, a small earthquake, cold rainwater downwelling, even nothing at all if the stratification was already at its limit [S1][S2]. Survivors interviewed by John Lockwood and colleagues described a sound like a distant explosion around 9:30 p.m. on the 21st, a strange smell some called rotten eggs and others gunpowder, and then the floor [S9]. The popular "silently exhaled" framing is a journalistic tidying; what people who lived through it described was a bang, a stink, and the dark [S9].

CO2 is heavier than air. The cloud — somewhere between 100,000 and 1.6 million tonnes, depending on which estimate you trust — flowed downhill as an invisible river, pooled in three valleys, and asphyxiated everything breathing [S1][S2].

That is the part that has been told. The part that hasn't is what has been done about it since, and by whom.

A French physicist named Michel Halbwachs proposed something almost embarrassingly simple: stand a pipe vertically in the lake, prime it once, and let physics do the rest [S3]. A parcel of deep water entering the pipe begins to fizz as it rises and depressurizes; the bubbles make the column lighter than the surrounding water; the column rises faster; the faster rise pulls in more saturated water below it [S3]. Self-siphoning, indefinitely, on no power [S3]. The pipes are high-density polyethylene, chosen because their density is close to water — eliminating the heavy support rigging earlier metal designs had required [S3].

The first pipe was commissioned at Nyos on January 30, 2001, and immediately threw a fountain roughly fifty meters into the air [S4]. Two more were added in 2011 [S4]. The pipes run on no fuel; the monitoring telemetry runs on a single solar panel [S4]. Maintenance is a small rotating team of French and Cameroonian geologists who fly in periodically to confirm the lake is still burping itself safely [S4].

Nyos is not safe in the way a defused bomb is safe. It is safe in the way an open valve is safe — only as long as someone keeps the valve open. The hazard requires permanent attention, forever.

And there is a second hazard the pipes do nothing about. The lake is impounded by a natural pyroclastic dam roughly fifty meters wide, and that dam is eroding [S7]. A USGS-affiliated hazard assessment in the Bulletin of Volcanology judged a breach likely within ten years, possibly within five, with downstream flooding reaching as far as Nigeria, 108 kilometers away [S7]. A breach would also drop the water level enough to depressurize whatever CO2 is left in the deep layer — the limnic mechanism, re-triggered by the engineering's failure to address a different problem [S7]. Cameroonian reinforcement work has since been carried out; independent verification that the risk is "averted" is thin [S7].

Only three lakes on Earth are known to be capable of this trick. Nyos is one. The smaller Lake Monoun, also in Cameroon, killed thirty-seven people the same way in 1984 and is now degassed by the same Halbwachs system [S3]. The third is Lake Kivu, on the border of Rwanda and the Democratic Republic of Congo.

Kivu holds, by current estimates, between 256 and 300 cubic kilometers of dissolved CO2 and another 40 to 65 cubic kilometers of dissolved methane [S5][S6]. It sits beneath an active volcano. About two million people live around its shores and immediate drainage basin [S5].

The methane has no Nyos analogue. Roughly two-thirds of it is produced by archaea reducing the lake's own geogenic CO2 — a slow microbial conversion of the volcanic inventory into a far more flammable form — and the remaining third is ordinary acetoclastic methanogenesis from sediment organics [S5]. The CO2 problem is becoming, year by year, a methane problem.

That methane is also why the world's most populated limnic-eruption lake is being defused commercially rather than by a public-safety budget. The KivuWatt floating barge anchored off Kibuye, Rwanda, has been pulling deep water up, separating the methane, and burning it for electricity in a 26-megawatt plant since December 2015, returning the degassed water to a non-bioactive depth [S10]. A second 37.5-megawatt plant, Shema Power Lake Kivu, is following [S10]. Energy on one side of the ledger; gas-inventory drawdown on the other [S10].

Whether the inventory is actually growing is the live scientific argument. A 2005 paper by Schmid, Halbwachs, Wehrli and Wüest measured Kivu's stratification and concluded gas concentrations were rising and the eruption risk increasing [S5]. A 2020 intercomparison reanalyzed the same datasets, attributed the earlier trend to inconsistent sampling methods, and concluded the lake has been close to steady state since 1974 [S6]. Two peer-reviewed claims, directly opposed [S5][S6].

A 2016 PNAS analysis of Kivu's carbonate sediment layers traced the onset of major hydrothermal input to roughly 3,100 years before present — the same forcing that maintains the lake's permanent stratification today and the same plumbing that loads its deep water with gas [S8]. Kivu's stratification is not a stable inheritance; it is an ongoing geological process, sitting on top of an active vent, under a live volcano, beneath two million people [S5][S8].

The standpipes at Nyos cost a few hundred thousand dollars and are maintained by a handful of geologists [S3][S4]. Whatever the equivalent intervention at Kivu turns out to be — more KivuWatt-class barges, an international monitoring station that does not yet exist, a political agreement between Kigali and Kinshasa that does not yet exist — it will be larger by something like three orders of magnitude [S5][S10]. The closest analogue is Nyos times a thousand. That is the scale of the babysitting job nobody has yet signed up for.