The Lake Nyos Disaster

Summary

On the evening of 21 August 1986 a cloud of carbon dioxide rose out of Lake Nyos in Cameroon and killed about 1,700 people in the valleys below it. The gas had been dissolved in the lake's deep water, and the explanation, a limnic eruption, is now the accepted one.

There was no fire, no lava and no blast that anyone saw. People heard rumbling from the lake, smelled something like rotten eggs or gunpowder, and lost consciousness. Survivors woke hours later to find their lamps out, their animals dead, and members of their families dead. Two years earlier, beside Lake Monoun, 95 kilometres away, 37 people had died in the same way.

The answer came out of the lake itself. Water drawn up from its bottom fizzed like an opened bottle. The lake had been storing gas that seeps up from below, and in one night it let the gas go. Since 2001, pipes have been letting it go slowly.

What is documented

The lake. Lake Nyos fills a volcanic crater (a maar) in the Oku Volcanic Field, on the Cameroon Volcanic Line, about 310 kilometres north-west of Yaoundé. The United States scientific team sent after the disaster measured it at 1,925 metres long, 1,180 metres wide and 208 metres deep, with a surface area of 1.58 square kilometres. Its outlet spillway, the team noted, is formed of "relatively weak pyroclastic ash beds".

The evening. 21 August 1986 was a Thursday. The US team's reconstruction, built from survivor interviews, is that by about 21:30 the weather was calm and most people were engaged in their usual evening activities when a series of rumbling sounds, lasting perhaps 15 to 20 seconds, came from the lake. People near the lake came out of their houses, felt a warm sensation, smelled rotten eggs or gunpowder, and rapidly lost consciousness. Survivors woke 6 to 36 hours later. Many found that their oil lamps had gone out with oil still in them.

The dead. The US team reported that about 1,700 people and 3,000 cattle died near the lake and along drainages up to 10 kilometres north of it, over an area of about 29 square kilometres. Subum, about 10 kilometres from the lake, was among the villages hit; Time named Nyos, Subum and Cha as the worst affected. The Smithsonian Institution's Scientific Event Alert Network bulletin for August 1986 traced a western tongue of gas along the Cha valley for about 15 kilometres. Within days the UN Disaster Relief coordinator in Geneva put the toll at 1,746, and most of the dead were buried in mass graves (Time, 8 September 1986). Time also reported perhaps 3,000 more people displaced from the fringes of the affected area.

The lake afterwards. The US report describes the water after the event as rusty red, with mats of vegetation floating on it, and its level about a metre lower. A water surge 6 metres high had overflowed the spillway, and water had washed up the southern shore to about 25 metres. The Smithsonian bulletin gives a larger figure for one spot, a surge onto a small peninsula on the south side that F. LeGuern measured at 80 metres above the lake. The sources seen by the registry do not reconcile the two. From the distribution of dead cattle, the US team judged that the gas cloud first rose about 100 metres above the crater rim.

What the water held. The US team arrived in Yaoundé on 27 August 1986. Water brought up from the bottom of the lake effervesced as it reached the surface: it was supersaturated with dissolved gas, and the gas was carbon dioxide. Depth soundings found no crater or disturbance on the lake bed. The team wrote that the water temperature, the composition of the dissolved gases and the low sulphur content of the water and sediments did not support recent, direct injection of volcanic gas, and estimated that the lake could have released about 1 cubic kilometre of carbon dioxide. The Smithsonian bulletin recorded no evidence of hydrogen cyanide or carbon monoxide.

The medical findings. Peter Baxter and two colleagues reported in 1989 that the clinical findings in 845 survivors seen at or admitted to hospital were compatible with exposure to an asphyxiant gas. Reddened skin and blisters on 161 of them (19 per cent) had at first been taken for burns from acid gases; further investigation suggested they were associated with coma states caused by the carbon dioxide. The authors did not exclude other toxic factors.

Monoun, two years earlier. In August 1984, 37 people died beside Lake Monoun, about 95 kilometres south-east of Nyos, after walking into a visible cloud around the lake. A team led by the volcanologist Haraldur Sigurdsson traced the gas to carbon dioxide of volcanic origin, stored in the lake's deep water and released after a landslide from the eastern crater rim slumped into it. Their paper dates the event to 15 August 1984; a Smithsonian bulletin report places the release before dawn on 16 August. According to Kevin Krajick, writing in Smithsonian in 2003, Science had rejected Sigurdsson's paper as far-fetched. It was published in March 1987, seven months after Nyos.

Degassing. Test experiments began at Monoun in 1992 and at Nyos in 1995. A single permanent pipe went into operation at Nyos in January 2001 and at Monoun in 2003, and two more pipes were added at Monoun in 2006 and at Nyos in 2011 (Tanyileke and colleagues, 2019). Each pipe is self-siphoning: once deep water has been started up it, the gas coming out of solution lightens the column and keeps the flow going without a pump. Krajick reported the first Nyos jet at 148 feet (about 45 metres). In 2005 George Kling and colleagues reported that both lakes had been recharging with carbon dioxide at up to 80 moles per square metre per year, and that a single pipe would remove only about 30 per cent of the remaining gas at Monoun before recharge balanced it, and about 25 per cent at Nyos by 2015. They called for more pipes. In 2008 Minoru Kusakabe and colleagues reported that the carbon dioxide in Nyos still amounted to 80 per cent of its maximum level.

The scientific legacy. An international working group on crater lakes was formed at a conference in Yaoundé in 1987. In 1993 it became the Commission on Volcanic Lakes of the International Association of Volcanology and Chemistry of the Earth's Interior (Rouwet, Tanyileke and Costa, 2016).

Leading explanations

Limnic eruption: carbon dioxide stored in the deep water and released when the lake overturned. Established.

Carbon dioxide of magmatic origin seeps into the bottom of the lake and dissolves in the deep water, which stays down because it is denser than the water above it and is held under pressure (Kling and colleagues, 1987). Over years the deep water approaches saturation. If some of it is lifted, the pressure on it falls, gas comes out of solution, and the bubbling water becomes lighter and rises faster, drawing more deep water up behind it. The process runs away until much of the stored gas has escaped. Carbon dioxide is about one and a half times as dense as air, so the cloud spilled over the crater rim and ran down the valleys to the villages.

The case for it is cumulative: the fizzing deep water, the undisturbed lake bed, the temperature and sulphur evidence against an injection of volcanic gas, and above all the lake's behaviour afterwards. Its deep water filled with carbon dioxide again, as a lake that stores gas should. Rouwet, Tanyileke and Costa (2016) note that it was this natural recharge that made the limnic explanation the accepted one. Every degassing design since has been built on it.

Phreatic eruption through the lake (Haroun Tazieff). Not accepted.

The volcanologist Haroun Tazieff argued that the gas came from an exceptionally CO2-rich phreatic eruption, as at the Dieng Plateau in Java in February 1979, where the erupting crater held no lake. In a 1989 paper he listed six observed facts that, he wrote, the eruptive hypothesis easily accounts for and the limnic one ignores. By his own account his was the first hypothesis put forward. It is now a minority position: the recharge measurements and three decades of controlled degassing have followed the limnic model.

The trigger (unknown).

What disturbed the lake that night is not known. The US team wrote that if the water was supersaturated before the event, "any small disturbance of the water column would initiate degassing". Proposals include a landslide or rockfall into the lake, an earthquake, and cold rain blown to one side of the lake by strong winds. The San Diego State University volcano site notes that there is little or no evidence that a landslide or an earthquake started it. At Monoun, by contrast, Sigurdsson's team did identify a landslide. The trigger is the open question. The mechanism is not.

Current status

Later explained. The mechanism, a limnic eruption of carbon dioxide stored in stratified deep water, is established, and the same mechanism explains Monoun. The trigger at Nyos is unknown and may stay unknown, and that does not reopen the case.

The lake is no longer the hazard it was. Cameroonian scientists reported in 2019 that the degassing had brought the carbon dioxide in both lakes down to safe levels, and that reinforcement works had averted the feared collapse of the Nyos dam and the flood that would follow. Halbwachs, Sabroux and Kayser concluded in 2020 that a single pipe, running continuously, could match the natural recharge. A 2021 study stressed that the lake is still being supplied with carbon dioxide, and that regular surveys are needed to keep it safe. The registry found no new published assessment from 2025 or 2026. What would change this record is not a new theory but a failure: of the pipes, of the monitoring, or of the dam.

Last reviewed: September 2026. Records are provisional. Where the evidence changes, the entry changes. Found an error? Tell us.