ÐÓ°ÉÔ­´´

Late on the evening of Thursday 21 August 1986, a deadly cloud of gas
swept along the valleys north of Lake Nyos in western Cameroon, leaving
a trail of death and devastation in its wake. News started to circulate
the following day and on the Saturday when people from outside the immediate
area visited the market village of Nyos. On Sunday, two scientists from
Cameroon’s Ministry of Mines and Power visited Lake Nyos. By then news of
the disaster was spreading around the world, with eerie descriptions of
villages where every living thing had died – men, women, children, cows,
chickens and even insects. The death toll of more than 1700 people brought
a rapid response from many governments, in the form of aid as well as teams
of scientists to help to find out what had caused so many deaths.

The victims had been overwhelmed by a cloud of poisonous gas from Lake
Nyos. Because the lake lies in a volcanic crater, many of the first investigators
on the scene assumed that this dormant volcano had come back to life and
released a cloud of hot toxic gas beneath the lake. Other investigators
soon realised that the evidence pointed to the slow build-up of carbon dioxide
deep in the lake, followed by its release as a cold, suffocating aerosol.
Initially, the assumption that volcanic gases were responsible hampered
the investigations. But it also proved a bonus. In any natural disaster
it is vital to collect reliable information quickly and volcanologists tend
to react fastest and are most willing to go to potentially hazardous areas.

A team of Italian volcanologists arrived at Lake Nyos only a week after
the disaster, followed within the next few days by teams from France, Japan,
Nigeria, Switzerland and the United States. About a week later the British
team, myself included, arrived.

Devastated vegetation

Lake Nyos is high in the hills to the south of the village of Nyos.
Our first visit involved a long and unpleasant walk up a narrow track strewn
with the carcasses of cows that had been dead for a couple of weeks or so.
When we reached the crest of the final ridge and saw the lake before us,
I was struck by the very precise pattern of damage to the vegetation. From
where we stood, close to the spillway over which water drains during the
wet season, we could look across the lake and see devastated vegetation
on the southern shore up to a height of 25 metres. But there was no obvious
damage to the vegetation on either side of the spillway itself. Another
striking feature was the colour of the lake: its surface was a turbid yellowish-brown.
We later discovered that it was a suspension of hydrated iron oxide formed
when water from deep in the lake mixed with surface waters.

To the south of the lake and in a small cove immediately east of the
spillway, the wave had risen about 25 metres, flattening the vegetation
as it streamed back to the lake and spreading a brown film over the plants.
It had not affected all sides of the lake equally. On the west shore there
is a vertical wall of rock which showed signs of recent rock falls that
could also have been caused by a wave; at the southern end of this wall
water had overtopped a promontory 75 metres high.

From the traces left by the wave, we could work out where on the lake
it had come from. Little or no water had been lost over the spillway, so
this part of the shore must have been out of the direct path of the wave,
shielded by the promontory to the east. This put the source of the wave
close to the northern shore. But a wave had also swept into the cove between
the spillway and this headland; if this wave had reflected off the rock
wall on the western side of the lake, then the disturbance must have started
in the northeast, a position consistent with the effects on the south and
west shores. The shape of the lake would have focussed a wave originating
there so that it could sweep over the 75-metre promontory and across the
delta plain and into the valleys beyond.

Apart from the areas flattened by water, there was surprisingly little
damage to the vegetation around Lake Nyos. Even to the south of the lake,
plants were starting to recover within two or three weeks. A large fig tree
grows on the spillway; neither it, nor any of the vegetation to either side
of the spillway, which must have been in the path of the cloud, showed any
damage, either physical or chemical.

Frost damage

But one species of plant on the delta plain had suffered, in a way that
may be significant. Its young leaves were blackened and shrivelled, although
individual plants had recovered and were continuing to grow. At first, we
interpreted this as chemical damage. But the effect was like that of a late
frost on tender plants. Gas released from deep within the lake would have
cooled by as much as 10 °C as it expanded during its rise to the surface;
‘frost damage’ to the most susceptible plants seems a reasonable explanation.

There was slight damage to plants and trees in the valleys to the north
of Lake Nyos and several large fig trees were uprooted. The vegetation
damage followed the general line of the valley but was not always worst
by the river. The type of damage was not characteristic of wind or water,
but looked like something in between. It was as if a heavy, narrow stream
of gas had bounced down the valley, flowing straight ahead whenever the
river took a sudden dip or turn and flowing around, rather than through,
particularly dense patches of vegetation.

These observations provided the key to what I now regard as the most
likely explanation for the devastation that came from Lake Nyos. A cold
aerosol of water and carbon dioxide could have damaged vegetation in the
way we saw, a pattern very difficult to explain in any other way. And there
was no shortage of carbon dioxide in the lake, as the investigators soon
found.

In early September 1986, Lake Nyos was highly charged with carbon dioxide.
When the teams first tried to collect samples of deep water they used unpressurised
containers. But the water released so much gas that the containers exploded
as they were brought towards the surface. Only when the investigators recovered
their sample vessels slowly, with a bleed valve left open, did they succeed
in collecting deep water, although they had lost much of the gas dissolved
in it.

Various teams collected samples of gas from the lake, which were mainly
carbon dioxide (99.6 per cent) with a little methane and a minute trace
of helium. We were sure that there was no sulphur dioxide or hydrogen sulphide
to speak of, because the gas that bubbled up through the lake when we took
samples had no smell. Concentrations of these gases as low as one part per
million produce the distinctive smells of gunpowder or bad eggs.

Accurate information on the concentration of gases at depth had to wait
until November 1986, when a German team brought equipment that could be
sealed deep in the lake and returned to the laboratory under pressure. These
sealed samples confirmed what the first samples suggested. And they have
shown that there were approximately 250 million cubic metres (measured at
standard temperature and pressure) of carbon dioxide in Lake Nyos after
the disaster. The isotopes of carbon and oxygen in the gas suggest that
it comes from the interior of the Earth somehow seeping into the lake from
below. Further samples, the most recent of which were taken last April,
show that carbon dioxide is being added to the lake at a rate of nearly
5 million cubic metres per year.

But there were problems with this explanation in eye-witness accounts
circulating in the aftermath of the disaster. Many survivors associated
a smell of bad eggs or gunpowder with the toxic cloud and said that there
had been explosions when the gas was released.

Most of the investigators took these reports at face value, but gave
them varying degrees of emphasis. Those who argued for a volcanic eruption
beneath the lake cited the smell as evidence for sulphurous volcanic gases;
others, who believed that the toxic gas came from stagnant water or sediments
within the lake, explained the smell as an ‘olfactory hallucination’.

The reports of sulphurous smells and explosions pose a problem for the
idea that carbon dioxide caused the disaster. These smells are certainly
to be expected from a volcanic eruption, but this explanation brings its
own difficulties. The bed of Lake Nyos is blanketed by a thick layer of
fine sediment that would be easy to disturb and would settle slowly. None
of the many deep water samples collected soon after the disaster contained
any sediment. If a significant amount of gas had erupted through the lake
bed, then the water would have been cloudy with sediment for months after
the disaster. And there were no sulphurous gases in the samples from the
lake.

Confusing smells

The suggestion that the smell was a hallucination is interesting but
it is difficult, if not impossible, to explain why the survivors would have
imagined the smells of plausible volcanic gases only.

Although carbon dioxide is generally regarded as an odourless gas it
does have a slight acid taste, because it forms carbonic acid in the mouth.
It is possible that some of the survivors interpreted the taste of carbon
dioxide as the smell of an acid gas. Lake Nyos lost a lot of water during
the disaster. Some rained back into the lake but much of it was incorporated
in the toxic cloud that devastated the valleys. Samples of deep lake water
were weakly acid with a pH of 5.6 and contained more than 500 milligrams
of dissolved solids per litre, mainly bicarbonates of sodium, magnesium,
calcium and iron. These bicarbonates give water a very nasty and distinctly
acrid taste. It is hardly surprising that the few people who tasted the
gas cloud and survived spoke of an unpleasant taste, and equally likely
that these reports should have reinforced more general rumours concerning
the smell of the gas.

On a recent trip to the area, I discovered, to my surprise, that the
people who live in the valleys around the lake do not have separate words
for smell and taste. At least six languages are spoken within walking distance
of the lake, so people from different villages tend to communicate with
each other in pidgin English. When people wish to draw the distinction,
they speak of ‘smell for nose’ or ‘smell for mouth’. So someone saying ‘the
cloud had a bad smell’ could as easily mean that ‘the cloud had a bad taste’.
The distinction would be lost on someone simply reporting what was said.
This may explain the reports of a bad smell, but not why it was described
as sulphurous.

There are other signs that some of the reports may not be completely
reliable. When 58 survivors were interviewed in the hospital at Subum a
few weeks after the disaster, the majority described a strong malodour,
mostly as ‘gunpowder’, accompanied by a loud, rumbling noise. That anyone
in Subum, which is 10.4 kilometres from Lake Nyos, should have heard a loud
noise strikes me as unlikely; it is even more curious that they should describe
the noise and smell as simultaneous. A sound wave would have reached Subum
from Lake Nyos in 32 seconds, but the gas would probably have taken about
half an hour to get there. It seems that the reports reflected what by then
had become the accepted version of events rather than the actual experience
of individuals.

Although I do not suggest that all reports from the survivors should
be rejected, I suspect that many, if not all of them, were interwoven with
the expectations of volcanic gases held by many of those who were asking
questions. These ideas, reinforced by the distinctly acrid taste of the
gas cloud, discussed, repeated and retold through partial language barriers,
could easily have created all the ‘eyewitness accounts’ that were circulating
in early September 1986 and have now been absorbed into the accepted accounts.
Any hypothesis which sets out to explain what happened should be judged,
first and foremost, on other, more objective evidence.

An obvious source of such evidence would be how the gas killed. Sadly,
very little is known about the people who died during the disaster; we can
only guess the precise number of dead. I understand, from talking to some
of the soldiers who helped to bury the dead in large communal graves, that
no records were kept – hardly surprising under the circumstances. The official
death toll is 1746 people, obtained by estimating populations from a census
taken about five years previously and then deducting the number of known
survivors. An alternative estimate based on a survey of people who had not
been seen since the disaster came up with a similar though slightly lower
total.

Medical evidence for what happened is sparse and based mainly on photographs
of the dead and the hospital admission records of the survivors. The dead
were buried on the weekend following the disaster, or soon after, and no
autopsies were performed at that time. Linguistic and cultural barriers
ruled out any detailed retrospective medical study, such as questioning
the survivors further. The area is so ethnically and linguistically complex
that even people from the nearest large town would not have been able to
talk to most of the survivors in the language of their own village. For
medical experts from the capital city, the linguistic barrier was almost
complete.

People who visited the area after the disaster noted that many of the
corpses were blistered; some of the survivors in hospital also had blisters.
The first volcanologists to reach the area took this as evidence for gases
that were either hot or acid. But the survivors’ blisters were superficial
and healed rapidly, characteristics that suggest that they were a result
of depriving the skin of oxygen.

The 548 survivors who were admitted to hospital and a further 297 survivors
who were seen as outpatients had symptoms that were compatible with exposure
to a suffocating gas such as carbon dioxide. And although the clinical evidence
does not rule out small amounts of other gases, there is equally no evidence
to suggest that they were there. Virtually all the survivors admitted to
hospital had lost consciousness and many had been unconscious for hours
rather than minutes. This, too, is consistent with exposure to high concentrations
of carbon dioxide.

Taking all the reliable evidence together gives a consistent picture
of what happened at Lake Nyos. Before August 1986, the water and sediment
of the lake was highly charged with carbon dioxide. No one knows exactly
how much gas was in the lake because there is no way to tell how much was
lost. But researchers do know the gas concentration shortly after the disaster,
so they can model its distribution beforehand.

During the wet season, which reaches its peak in August, rain falls
on the surrounding hills and flows down streams into the southern part of
Lake Nyos. This water forms a separate layer on the surface of the lake,
but it is a little cooler and denser than the lake water beneath. As the
lake refills in the wet season this water flows in at its southern end and
spreads across the surface of the lake, flowing out over the spillway. The
layer of cool surface water steadily thickens as deep water is lost by seepage
through the porous rocks that confine the lake. Normally, this stratification
decays as the surface water warms up at the end of the wet season.

But late on the evening of 21 August, something disturbed Lake Nyos.
Water from deep in the lake, highly charged with carbon dioxide, rose towards
the surface in the northeastern part of the lake. Why this happened can
never be known for certain. But at that time of year the prevailing winds
are from the northeast; it is possible that in August 1986 they were more
persistent than usual, pushing the cool water at the surface to the southern
part of the lake. Whenever enough of the slightly denser top layer accumulates
in one part of the lake, it will eventually become unstable and sink, pushing
gassy water from deep in the lake towards the surface elsewhere.

The rising water will start to release bubbles of gas as it approaches
the surface. Because the bubbles are themselves buoyant, they will increase
this convective flow, helping to drag upward more deep water with its dissolved
gas.

The release of dissolved carbon dioxide is an endothermic reaction
– in other words it absorbs energy, so that both the gas and the water involved
cool down. The rising bubbles expand and consequently cool further. So the
gas bubbling out of the water and the water itself will be perhaps as much
as 10 °C cooler than the deep parts of the lake. And because carbon
dioxide is much denser than air, the gas would spread across the surface
of the lake as a cold, toxic layer; the cooled water would sink, pushing
more deep water towards the surface.

As the gas release gathered pace, the water would cool more and sink
to greater depths, with the result that water was drawn from deeper in the
lake. This also brought water rich in ferrous bicarbonate to the surface,
where it could mix with oxygenated surface waters and produce the brown
colouring (hydrated ferric oxide) that we saw.

Limited circulation

Klaus Tietze, of the Institute for Geology and Mineralogy in Hanover,
has suggested that this process of gas release would generate a cylinder
of descending cold water surrounding a rising core of gassy water. This
is a particularly interesting possibility because the process could happen
in a limited area of the lake. The rising gassy core draws water into the
descending cylinder of cold water around it, but the surrounding lake is
undisturbed. Such a circulation limited to only part of the lake could explain
why the lake lost only part of its gases in 1986. It might also explain
how stable bottom water was drawn towards the surface without disturbing
the sediment on the lake floor and why the composition of the lake water
below 150 metres was fairly uniform after the disaster.

At the surface the vigorous release of gas generated the wave which
swept across the lake and into the valleys to the south. As the gas was
released, some of the water accompanying it was transformed into a fine
mist, generating the cold aerosol of water and carbon dioxide that swept
down the valleys to the north of the lake through Nyos and on to Subum,
Cha and Fang.

This is a reasonable model for what happened at Lake Nyos. It is certainly
plausible and it does not conflict with any of the reliable evidence. But
what does it suggest about the chances of another tragedy in the future?
There is certainly more than enough gas in the lake now for a similar disaster,
but if the model is correct a major release will be triggered only by a
large disturbance bringing deep water very close to the surface. Water from
as far as 170 metres down, for example, would not begin to release its gas
unless it comes within 30 metres of the surface. We have to hope that these
large disturbances are rare events, and that something will be done to reduce
the amount of gas in Lake Nyos before the next one.

Meanwhile, I suggest that no one should visit the Lake Nyos area without
good reason; anyone who does so at the height of the rainy season is taking
a serious risk.

Freeth is director of the Geological Hazards Research Unit at the University
College of Swansea.