WHEN IT'S GOOD TO FIND FAULT

Should scientists abandon all dreams of ever accurately predicting earthquakes? New research suggests that while warnings of an imminent quake may be possible, hopes of anything more are fanciful. Martin Redfern investigates

Take about 20 slabs of rock, several of them as big as continents and typically 50 miles thick. Place them on the curved surface of a globe and move them slowly alongside and into one another. The margins are not smooth straight lines. Indeed, on a curved globe, it wouldn't be much use if they were. They do not move smoothly alongside each other, they grate and grind and get stuck as the pressure builds. Eventually something fails at the edge of the rocky slab and you get a sudden jolt running through the whole area. That is an earthquake. Earthquakes are concentrated along the relatively narrow lines where plates on the Earth's surface meet and move past each other and they are responsible, directly or indirectly, for taking thousands of human lives every year.

People have been trying to predict precisely where and when earthquakes will strike since the days of the Chinese emperors, but with comparatively little luck. The only significant success came in February 1975 in China where long-term monitoring of many factors - such as the water level in wells, electrical signals in the ground and even animal behaviour - led the authorities to evacuate thousands of people from Haicheng in Liaoning Province, hours before what turned out to be a very powerful earthquake. There were relatively few casualties, even though over 90 per cent of homes in the hardest-hit area collapsed. However, that was an isolated and possibly fluke incident. Only two years later, another earthquake struck in China, killing 242,000 people who had not been warned.

Will it ever be possible to predict earthquakes reliably? Many delegates at a recent conference on earthquake prediction in London thought not. Professor Stuart Crampin of Edinburgh University said that the branch of mathematics known as chaos theory ruled out such predictions. Speakers at the conference showed how they could study tiny fractures in rock samples put under stress in the laboratory. For pieces of rock, they could begin to see patterns emerging just prior to the rock fracturing. But it's a long way from one piece of rock to a whole continent. The much quoted example in chaos theory of how a single butterfly flapping its wings can create a disturbance which goes on to become a storm is easily misinterpreted. Simply studying one butterfly flapping its wings will not bring you closer to predicting a storm, any more than one piece of rock will lead to an earthquake prediction. We simply don't know enough about all the details of the fracture zones to make accurate prediction feasible.

What seismologists can and do do is work in probabilities. If you live close to a plate boundary, it's likely that you live in an area prone to earthquakes. You can go further than that and look at the frequency of earthquakes in the past (as reported in the Independent on Sunday, 1 September). Then, if you know there is a major quake about once a century, you're reasonably safe in saying that there will be a major quake sometime in the next century. That's not much use for planning your daily life but it may encourage people to build houses that are either strong enough to resist earthquakes or flexible enough to move with them without collapsing.

The middle of next century could see a billion people living in megacities with populations over 20 million each, and half of those cities will be in earthquake-prone regions. Even a handful of nails to tie roof beams firmly into uprights in cheap housing can save lives. Prediction does not have to be accurate for this to be worthwhile, just as in Britain, if you were going out for a walk in March, it might be wise to take an umbrella, even if you hadn't seen a weather forecast.

To improve the probabilities of earthquake prediction further, you need to know a lot more about the plates and how they are moving over the earth's surface. In October 1992, a very unusual satellite was launched to help make such measurements. Instead of containing rocket motors and complex electronics, Lageos 2 is a solid sphere of metal, a core of brass surrounded by a shell of aluminium. It was made in Italy and launched on the Space Shuttle. It is only 60cm in diameter but very heavy, weighing 405kg. Its surface is covered by 426 reflectors. It is 6,000km above the Earth, well clear of the atmosphere. It's small and dense enough not to be affected by the gentle solar wind and it's non- magnetised, so its orbit is stable, affected only by the Earth's gravity. Special lasers in telescopes on the ground fire brief pulses of light at the satellite. The flash is reflected back to Earth in a round trip time of about 400ths of a second. It can be measured so accurately that it reveals the position of the satellite to a precision of less than 1cm. It's not that scientists want to know where the satellite is, more that it provides an independent reference point to measure where they are. If each of their telescopes is on a different continent, they can measure the relative movement of those continents.

That motion is typically two or three centimetres a year, about the rate at which fingernails grow, but something which is impossible to measure on the scale of continents by any other means. Although the stable cores of the continents are drifting at a smooth and relentless pace, the edges, as we've seen, are moving in fits and starts. A new network of about 200 receivers for the military global-positioning satellite system have just been installed in quake-prone California. Though not quite as accurate as the laser measurements, they can nevertheless reveal places where the continent has got stuck and other places where it is starting to move rapidly, perhaps prior to a big earthquake.

Earthquakes are taking place practically all the time in seismic zones such as California, where the Pacific is sliding north along the American continent. Although the principle crack, or fault, is famous - the San Andreas fault - the whole state is crazed by a network of secondary faults. Quakes often represent movement along these, but it's not always possible to tell from the vibrations which fault moved, or by how much. That, too, can now be done by satellite. The European remote-sensing satellites carry radar scanners which map the surface even through cloud. By combining images taken before and after a small quake, the way the ground has moved is revealed by coloured interference patterns.

Another line of approach is not prediction but detection. If you detect minor earthquakes, there's a good chance they may be the foreshocks to a major quake. That's particularly true in sections of fault that have been quiet recently, where minor tremors could signal that they are about to re-activate. The United States Geological Survey has set up a network of 300 monitoring stations in southern California. Whenever there is a minor tremor, the stations can get a fix on its exact location. Computers then analyse the information to filter out things like heavy trucks going past and to exclude areas of fault that seem to be active most of the time. If the computer concludes that a previously quiet fault zone is experiencing tremors, then it will automatically send a warning to geologists and bleep them on radio pagers. If the alarms go off, it could mean that the risk of a major quake within the next 24 hours has gone up from one in 10,000 to one in 20, but that's still the same as saying that there's a 95 per cent chance that there won't be a quake today. That's hardly a reason to issue public warnings that could lead to panic and mass evacuation, particularly in a country as litigious as the US.

What could be done, however, is alert emergency services to move fire engines into the open; stop pumping hazardous chemicals; and shut down computer systems to protect financial data. In cities such as San Francisco and Los Angeles where buildings are designed to withstand earthquakes, evacuation would not be recommended, even if the odds were more certain.

After the Loma Prieta earthquake near San Francisco in 1989, an automatic quake detection system was set up to protect rescue workers. They were searching beneath the huge slabs of concrete of the collapsed Nimitz freeway. Quake monitors were set up near the fault zone almost 100km away. If three of them registered a tremor simultaneously, a radio signal was sent out directly to the rescue workers where it would ring an alarm. Because it was so far, the radio signal travelling at the speed of light would get there 25 seconds before the earthquake waves. Then it wasn't a matter of probabilities; the shock waves were on their way. Twenty-five seconds may not seem long, but it's long enough to jump clear of a partially collapsed slab of concrete.

Ultimately, similar detection systems might give brief warnings of the big quakes themselves. Los Angeles, for example, is more than 200km from a fault that is due for a devastating quake.

While people can't be expected to live their lives in permanent readiness for a one-minute warning, computers can. The biggest fear in Los Angles is not of building damage, but that banks might lose their computer records of the day's transactions A computerised early-warning system could prevent that, as well as, for example, shutting safety valves on pipes and parking lifts and opening their doors. The trouble is that such systems are expensive, but perhaps Californians will find the money to implement them.

As far as the rest of us are concerned, quakes are as unpredictable as ever. Beyond building homes with quakes in mind there seems little we can do. On geological timescales, the Atlantic is getting wider; Africa has not quite finished crashing into Europe, and India's road-traffic accident with the rest of Asia is still in full swing. In a few more million years there will be new mountains, new oceans and also new casualties.

Martin Redfern is an executive producer with the BBC World Service science unit