SCIENCE AND TECHNOLOGY
Nature rarely repeats itself
T O K Y O
Japan’s earthquake-prediction programme, the last in the world, is about to get the chop
IN THE 1960s the study of earthquakes entered a golden age. The discovery that the earth’s crust is composed of giant rocky “plates”, which wander slowly over the planet’s surface, gave geophysicists a coherent explanation of what it is that causes tremors and why they tend to happen more in some places than in others. As the plates slide into, past or away from each other, considerable strain is put on the rocks at their edges. Every so often, this strain is suddenly released as an earthquake.
Armed with such knowledge, a number of geophysicists felt that with enough research it ought to be possible to predict where, and more importantly when, major earthquakes would occur. Alas, this was not so. Most of the earthquake-prediction programmes that blossomed in those rich countries that straddle plate boundaries have since been recognised as failures, and closed. Japan’s programme, however, has lived on—fuelled by an annual budget of ¥17 billion ($145m) and the knowledge that the country’s capital sits over the junction of three tectonic plates.
But that programme will not last much longer. A report just adopted by the Ministry of Education’s Geodesy Council finally admits that Japanese geophysicists have no special insight denied to others, and that their programme, too, is a complete waste of money. Having spent ¥160 billion over the past 30 years, the council engages in some justifiable self-criticism—rebuking itself for its own naive optimism, and urging the government to divert resources that would otherwise have been wasted on the programme into designing safer buildings and making better preparations for dealing with disasters after they have happened.
The Japanese government’s efforts to predict earthquakes date back to 1965. Since then, the Geodesy Council has completed six five-year forecasting programmes; it is now in the middle of a seventh. The hope was that with enough data about precursor phenomena—abnormal bulging of the earth’s crust, variations in its electrical resistance and in the local magnetic field, changes in the level and chemical composition of ground water, and so on—some pattern would surely emerge, just as particular patterns in the atmosphere are more-or-less reliable predictors of rain. This theory was given something of a boost in 1975, when the Chinese claimed to have predicted a big earthquake in Haicheng on the basis of earth movements beforehand.
In addition to the idea that earthquakes might give a few hours’ notice of their arrival, there was also a feeling that, in some places at least, a grander pattern was detectable, and that earthquakes came in regular cycles. Two things fuelled this theory. Seismologists in California had noticed that a section of the San Andreas fault near the town of Parkfield tended to slip every 20 years, generating a magnitude 6 earthquake; as a result, they confidently predicted the next one before 1993. And Japan’s own earthquake history led the late Hiroshi Kawasumi, who was head of Tokyo University’s Earthquake Research Institute, to the conclusion that the shock which killed 142,000 people in the greater Tokyo area in 1923 was one in a series of quakes that occur every 69 years or so.
Performing badly under stress
Unfortunately for the science of earthquake forecasting (though fortunately for the residents of Parkfield and Tokyo), neither of the predicted shocks has yet happened. And the more that geophysicists learn about seismic activity, the more they realise that earthquake behaviour is essentially chaotic, and that the “elastic-rebound” model—the basis of practically all seismic studies since the great San Francisco earthquake in 1906—is far too simple to be used for predictions, whether in the short or the longer term.
The classic theory breaks down because the faults where quakes are spawned are rarely found in isolation. Most are part of a cat’s cradle of interconnecting cracks, which redistribute the energy that builds up in a plate as it is squeezed by its neighbours. This dissipation often causes stresses to accumulate elsewhere in the fault system, triggering tremors far from the place where the rocks first slipped. Fathoming where and when such events will occur is too complicated for any conceivable computer model.
This leaves the Geodesy Council in an awkward position. In 1978 the Japanese government passed a law requiring adequate warnings to be given of major earthquakes so that people could be evacuated and emergency services put on full alert. The council is supposed to make such warnings possible.
The nearest that it has got to predicting a big tremor was in 1983 when a foreshock pointed a finger at a subsequent magnitude 7.7 earthquake within the Sea of Japan. The only problem was that the two events occurred 12 days apart. The Japan Research Institute, a private-sector think-tank associated with the Ministry of International Trade and Industry, reckons that if emergency measures had been introduced on the basis of such a precursor—with trains being stopped, banks, post offices and department stores closed, hospitals emptied of all but the most serious of cases, and school children sent home—the cost would have run to more than ¥700 billion a day. If they had been kept in place for 12 days, the cost of the measures could have amounted to more than the cost of any damage done when the earthquake actually came.
In any case, this “prediction” proved to be a one-off. Japan’s extensive network of detection equipment completely missed four big tremors in the early 1990s. These failures could be played down because the events occurred in sparsely populated areas of Hokkaido and northern Honshu. Where the forecasters really fell over was in failing to give any warning of the Kobe earthquake in 1995. This killed 6,400 people in an area that was supposed to be seismically inactive.
In the wake of Kobe one of the doyens of Japanese seismology, Kiyoo Mogi, noisily quit his post as the government’s top earthquake adviser because of the futility of the task. Dr Mogi’s job was to advise the prime minister on whether or not to declare a national emergency in the hours before a major earthquake was due to hit Tokyo, something he felt he could not possibly do.
The Geodesy Council’s decision is expected to meet resistance from the other three government agencies with an interest in earthquake prediction and its associated budget—the National Land Agency, the Meteorological Agency and the Ministry of Construction. But the council is expected to make their objections redundant by cancelling its eighth prediction programme, which would otherwise be due to get under way in 1999.
There is, however, one further task that needs to be performed. For decades, the Japanese have been drip-fed the idea that big earthquakes are predictable—and, despite the horrifying experience of Kobe, polls show that half of them still expect to be given adequate warning of any future earthquake. A lot of unlearning still needs to take place. Amplifying the anguished apologies of the Geodesy Council throughout the land would be a beginning.
© Copyright 1997 The Economist Newspaper Limited. All Rights Reserved