快猫短视频

ON 16 December 2000, Carlos Vald茅s issued an extraordinary warning.
Popocat茅petl volcano, he said, would explode in exactly two days鈥� time.
And he was right. On 18 December, Popocat茅petl unleashed its largest
eruption for a thousand years.

Thanks to Vald茅s and his colleagues at CENAPRED, Mexico鈥檚 National
Centre for Prevention of Disasters, the army had evacuated thousands of local
residents the day before. No one was hurt. But how did he manage to make such an
accurate prediction?

Volcano prediction has come a long way in the past 20 years. Unlike
earthquakes, which remain erratic and unpredictable, volcanoes make distinctive
creaks and groans in the run-up to an eruption. These allow researchers to
pinpoint the timing of eruptions with extraordinary accuracy, giving people
nearby time to escape before the volcano blows.

But although researchers can now accurately predict when an eruption will
occur, they still have no idea just how big it will be. Without that
information, they can only guess how many people should be evacuated. The 2000
Popocat茅petl eruption may have been huge, but the dramatic fire fountains
that rained on the volcano鈥檚 upper slopes posed relatively little danger to
villages far below. Preventing unnecessary evacuations is crucial, because local
people quickly lose faith in scientists who cry wolf. What鈥檚 really needed is a
way to work out whether a volcano is just clearing its throat, gearing up for a
cataclysmic explosion or鈥攍ike Popocat茅petl鈥攕omething in
between.

The answer may be hidden in the same volcanic rumblings that reveal the
eruption鈥檚 timing. In the past few years, volcanologists have managed to use
records of these growls to explain the size of eruptions after the event. Now
they reckon they can use them to predict the size of eruptions in advance. All
they need is a restless volcano.

Modern skills in predicting eruptions go back to the aftermath of the
cataclysmic eruption of Mount St Helens in Washington state in 1980. The
following year, seismologist Bernard Chouet from the US Geological Survey in
Menlo Park travelled to the volcano with a colleague, Mike Fehler. Mount St
Helens was still very active and the scientists wanted to listen to the
vibrations it was producing in the hope of understanding what was going on
inside. 鈥淰olcanoes are talking to you all the time,鈥� says Chouet. 鈥淭hey give
beautiful signals.鈥�

There was one kind of signal at Mount St Helens that Chouet was particularly
interested in. It was a vibration that built up over several seconds, dominated
by a single low-frequency wave that could last for a minute or more. Chouet
called it a 鈥渓ong-period鈥� event. Looking at the shape of the wave, he realised
what had caused it: resonance.

The phenomenon is familiar enough. Most musical instruments exploit resonance
to produce notes of a particular pitch. Air pumped into one end of an organ
pipe, for example, triggers standing waves in the column of air trapped inside.
Their wavelength鈥攁nd hence pitch of the sound they emit鈥攄epends on
the size and shape of the pipe. Chouet wondered if the same process happens in a
volcano. Each long-period event, he reasoned, could be the sound of some
fluid鈥攎olten magma or gas鈥� resonating in a crack inside the volcano
after a sudden change in pressure, perhaps caused by fresh magma entering the
crack.

At Mount St Helens, Chouet noticed clusters of long-period signals that all
had the same frequency and lasted for the same length of time. He figured that
these must have come from the same crack in the volcano, and that counting these
long-period events would be a way to monitor how fast the pressure was
increasing. The more events from a single crack, the closer you should be to an
explosion.

Chouet looked at research done on other volcanoes around the world and found
that many produced similar long-period events before eruptions. But were these
signals reliable enough for him to use to predict eruptions in advance? He was
about to find out.

In late December 1989, seismometers on Mount Redoubt in Alaska began to
register a series of long-period events. The rate of events increased rapidly
and within a few days merged into a continuous tremor as a sticky lava dome
oozed out at the summit of the volcano. Chouet, working with researchers at the
Alaska Volcano Observatory in Anchorage, began counting the rumbles. On 1
January 1990, the researchers forecast that an eruption would occur within 24
hours, and an oil terminal at the foot of the mountain was hastily evacuated.
The next evening the dome exploded, sending a vast muddy river of melted snow
and ice mixed with ash cascading down the mountain. Chouet was elated. 鈥淚t was a
wonderful feeling,鈥� he recalls. 鈥淭he people at the oil terminal thought we could
walk on water.鈥�

But Chouet was not content to stop there. He wanted to use long-period events
to predict the size of an eruption too. Since he knew that each signal contained
a set of resonant frequencies, he reasoned that the characteristic frequencies
of a volcanic crack would depend on its size鈥攋ust like the pitch of an
organ pipe. Similarly, he realised that the amplitude and duration of the
long-period event would depend on the pressure in the crack. If he could use
long-period events to work out the size of the crack and the pressure it was
under, he would be well on the way to predicting the size of the eruption that
could follow.

Using these ideas Chouet developed what amounts to a virtual volcano鈥攁
mathematical model of how a resonating volcanic crack generates long-period
signals. The model uses factors like the size of a crack and the pressure it is
under to predict the seismic signal that it should produce. Although Chouet
developed the theory during the 1980s, he had to wait until the mid-1990s before
computers were powerful enough to do the number-crunching.

His first subject was Mount Redoubt. Using his model, Chouet tried various
combinations of crack sizes and pressures. Eventually he created synthetic
long-period events that bore a striking resemblance to the signals he had
observed on Redoubt. What鈥檚 more, Chouet鈥檚 crack sizes and pressures would
produce the same size of eruption that had occurred on Redoubt. 鈥淚t鈥檚 a defining
moment because suddenly you realise the volcano is speaking to you and you
understand the language,鈥� he says.

But guessing all the different possible combinations of crack size and
pressure was infuriatingly laborious. Chouet needed to get the computer to do
the donkey work. His idea was to take the data on long-period events from real
volcanoes and run it backwards through the model. That way, the computer would
work out the size of the crack that generated those events and how much pressure
it was under.

It took a further advance in computing power before Chouet could turn his
idea into reality. Now he has managed to use his automated virtual volcano to
work out crack sizes and pressures for a few minor explosions that took place at
Stromboli in Italy. And he believes his volcano model is now ready to be used to
forecast the size of eruptions in advance. 鈥淲e鈥檙e one step away from the big
answer,鈥� he says.

But things may not be that simple. J眉rgen Neuberg of the University of
Leeds agrees with Chouet that long-period events are the key to predicting the
timing of eruptions鈥攂ut predicting their size is a different story.
Neuberg has carried out a detailed study of the rumblings from the
Soufri猫re Hills volcano on the Caribbean island of Montserrat. Just as at
Redoubt, long-period events at the Soufri猫re Hills are a sign that
pressure is building up in the lava dome. For Neuberg, this pattern sounds the
alarm that the dome is about to blow鈥攁s happened in July 2001, showering
the nearby village of Salem with cherry-sized pumice stones.

According to Neuberg, Chouet鈥檚 model of long-period events omits an important
characteristic of magma inside volcanoes. Deep down, the magma is just thick and
gloopy. But as the molten rock rises, the pressure exerted by the weight of the
rock above decreases, and gases dissolved in the magma begin to bubble out of
solution. The magma becomes bubbly, and then higher still the gas overwhelms the
magma, until sticky droplets of liquid magma become suspended in a matrix of
gas. Finally the droplets of magma may shatter into pieces of ash. So the way
magma behaves depends critically on how deep it is in the volcano
(快猫短视频, 26 October 1996, p 28).
And Neuberg points out that the nature of the fluid in the crack could have a significant
effect on the rumbles that fluid produces.

Using his observations at Montserrat, Neuberg has constructed his own virtual
volcano based on the same principles as Chouet鈥檚, but incorporating the idea
that the 鈥渂ubbliness鈥� of the magma varies with depth (Journal of Volcanology and
Geothermal Research, vol 101, p 83). Neuberg found that lots of different
combinations of fluid type, crack length and pressure could generate similar
long-period events. That makes it impossible to run the model backwards and use
the long-period events to calculate the size of the crack and the pressure it is
under.

Chouet agrees that depth is important but believes it is not as big a problem
as Neuberg suggests. Chouet points to a physical model of a volcano developed by
Steve Lane at the University of Lancaster. Lane uses pine sap together with
acetone or ether to represent magma and the gas dissolved in it. To simulate a
magma chamber with a crack above it, he puts this mixture in a sealed chamber
below a glass tube. When the seal is broken, the mixture explodes into a foam
that races up the tube, becoming increasingly bubbly the higher it
goes鈥攋ust like real magma.

Lane found that the foam produced resonating signals similar to the
long-period events detected from within volcanoes. But when he pinpointed where
the resonance was coming from, he realised that it only occurred where the
bubbliness was at a roughly constant level鈥攏ot throughout the flow
(Journal of Geophysical Research, vol 106, p 6461). Chouet reckons that if magma
resonance works in the same way, his virtual volcano still provides an accurate
model of what happens in real volcanic cracks.

For the moment this dispute will be difficult to resolve because researchers
don鈥檛 understand exactly what the fluid that resonates in volcanoes is, or why
it starts resonating in the first place. Neuberg hopes that some answers will
come from a joint European-American project he has helped set up on Montserrat.
Next month, he and his colleagues will install a full array of monitoring
equipment on the Soufri猫re Hills volcano, all collecting data in real
time. Later in the year, they plan to drill four boreholes into the heart of the
volcano to collect information down where the action is happening. The enormous
range of data Neuberg and colleagues will gather should give them an
unparalleled view of what magma and gases are doing inside the mountain.

Meanwhile, Chouet is confident about his model. He believes that we now know
enough about volcanoes to be able to 鈥減eel back鈥� the surface of an active
volcano and peer inside. In 10 years鈥� time, he hopes that tourist visitors to
Kilauea volcano in Hawaii will be greeted by a giant TV screen. There they will
have a virtual view right inside Kilauea鈥檚 caldera and be able to see in real
time what is happening under their feet. 鈥淲e will remove the mountain and look
directly at the fluid,鈥� says Chouet.

But the big test will be when Montserrat or another well-monitored volcano
next erupts. Will Chouet, Neuberg or anyone else be prepared to stick their neck
out and publicly predict in advance how big the eruption will be? It鈥檚 a tough
call, especially when thousands of lives may depend on the answer. But given
that volcanoes killed a hundred thousand people in the past century, there鈥檚 a
strong incentive to try.