91av

LIVING in an earthquake zone, you never know when the next one is going to strike. Californians are nervously waiting for the “Big One”. And two years on, the Japanese are still reeling from the Kobe quake, which struck without warning, with devastating consequences. The trouble is that, unlike other natural disasters, earthquakes are almost impossible to predict. Different scientists have claimed to be able to pick up signs of an impending earthquake by monitoring changes in temperature, the composition of groundwater, radon concentrations or the level of the water table. Others have looked to more bizarre sources for clues: the behaviour of cockroaches, say, or the Renaissance writings of Nostradamus.

Lately another prediction technique has been causing uproar among earth scientists. Its supporters claim it works better than any of the methods that have been tried before. Its critics dismiss it as nonsense. But as the argument rages on, the new method appears to keep on coming up trumps. The three scientists that developed it are Panayiotis Varotsos, Kessar Alexopoulos and Konstantine Nomicos, all physicists from the University of Athens-known collectively as “VAN”. In 1995, they successfully predicted three large earthquakes along Greek faults that had not been particularly active before. One of these, on 13 May 1995, was in a region where an earthquake of the predicted magnitude had not occurred for a thousand years. Altogether, the VAN team predicted 10 out of 14 large quakes in the area over the past decade. “Its performance seems much higher than any other methods,” says Hishashi Utada of the Earthquake Research Institute at the University of Tokyo.

The Athens scientists first ventured into the prediction business in the early 1980s, when they were running laboratory experiments to see how rock behaves under stress. Immediately before the rock fractured, the team picked up short-lived electrical currents passing through them. Intrigued, they wondered if the same thing might happen out in the field, when the rock is stressed in advance of a quake.

Their hunch bore fruit. The VAN researchers set up gargantuan voltmeters consisting of several kilometres of wire attached to large electrodes, which they scattered across Greece. These meters duly picked up similar oddball electric signals in regions where earthquakes subsequently occurred. And on the basis of the signals, the VAN team has been making earthquake forecasts ever since.

Though it is far from proven, the mere promise that the VAN method might predict quakes has attracted a flood of attention. In 1995, the International Council of Scientific Unions and the Royal Society convened a conference in London to consider the merits of the VAN approach. In an introduction to the published conference proceedings, mathematician James Lighthill from University College London concluded that the large earthquakes that hit Greece on 13 May and 15 June, after the meeting, were “related” to VAN predictions received on 2 May and 20 May. The entire May 1996 issue of Geophysical Research Letters was dedicated to papers about VAN.

The interest has not only been academic. For the past couple of years, the Greek government has allocated some 5 per cent of its anti-seismic budget (about $100 000 a year) to the study. And there is increasing interest from a number of Japanese officials and researchers. For instance, the Japanese Science and Technology Agency’s Earthquake Frontier Project is funding a multimillion-dollar study led by Seiya Uyeda at Tokai University to monitor similar electrical signals in Japan.

Doubting Thomas

But many geophysicists have been sceptical from the beginning, and criticism of VAN is mounting. Robert Geller, a seismologist at the University of Tokyo who edited the special issue of Geophysical Research Letters, has become increasingly disillusioned. The team’s claims “are utterly without merit” Geller now says. In particular, he asserts that the VAN team’s success rate has been seriously overstated. Geller and other critics say that funds should be directed toward things like hazard mitigation, public education, crisis management and real-time seismic monitoring, rather than on the VAN project. The VAN team seems to be facing opposition on every front. “There isn’t any aspect that isn’t being questioned by somebody,” observes Philip Stark, a statistician at the University of California at Berkeley.

For one thing, it remains very unclear what exactly could be causing the electrical signal. Varotsos’s explanation is that microcracks form in the crystal structure of the rocks as they are strained. This breaks chemical bonds within the rock to leave isolated positive and negative charges. Electrons quickly flow to cancel out the charges, causing the electrical signals that the researchers have detected in the field.

Many geophysicists are not persuaded. Instead of being generated by crust on the brink of breaking, the signals could well originate from human sources such as nearby factories that have nothing to do with earthquakes, critics say. Even if part of the signal does come from the breaking rock, it could be swamped by interference from human-generated signals, especially in countries that are more developed than Greece. VAN “would not be useful in places like California and Japan”, says seismologist Hiroo Kanamori from Caltech.

The Earth’s external magnetic field could also induce currents in the crust, and rain may cause instability in the electrodes used to measure the current, say the critics. Overall, says Uyeda, only about 2 per cent of electrical events in the crust are likely to be of tectonic origin, making it very difficult to pick out the “true” signals. Varotsos maintains that he has developed empirical rules for successfully screening out the noise, and Uyeda independently verified these methods. But others remain unconvinced by such ad hoc rules, which lack any firm theoretical foundation.

Even if the signals picked up by the VAN team do turn out to be related to quakes, they may still not be useful predictors. Questions linger as to why the signals can sometimes occur up to 100 kilometres away from the epicentre but not closer to the source or during an actual seismic event. Varotsos explains this with a model he presented to a meeting of the American Geophysical Union in San Francisco in December last year. When the current flows, he says, it travels along the path that has least resistance or highest conductivity. So if in the area near the source there happens to be a channel of high conductivity, overlaid by more resistive rock, the electrical signal would be detectable only where the conductive channel came close to the surface, possibly some distance away, but not near the rupture site. However, Varotsos has yet to demonstrate that this is the case.

This explanation also does nothing to satisfy critics who doubt that the energy of the source is large enough to produce the claimed effects. At a conference convened in London in November 1996 by the Royal Astronomical Society and the Joint Association for Geophysics, Pascal Bernard of the Earth Physics Institute (IPGP) in Paris asserted that the most optimistic physical models would not generate signals of even 1 per cent of the amplitude of those that VAN observed. Another cause for doubt is voiced by Max Wyss, a seismologist from the Geophysical Institute at the University of Alaska in Fairbanks. “Nobody has come up with a way to explain why there is no signal at the time of the earthquake,” he says.

This point at least is dealt with in a model put forward by Kanamori, though it does little to support the VAN group’s claim that the signals make useful predictors. Kanamori argues that local changes in strain could cause a sudden influx into the crust of fluids such as water, magma, carbon dioxide or other gases, which in turn would cause charge separation and an electric current. If the fluid reaches a fault zone, it could weaken the fault and help to trigger an earthquake. But as Kanamori points out, the fault could well be some distance away from the site where the initial fluid flow occurred, and the fluid could take some time to reach it. If so, the signals could be generated at a different time, and in a different place from the eventual rupture-making them much less useful as predictors.

It is not only the VAN group’s failure to come up with a convincing geophysical explanation that is attracting criticism. Also under scrutiny are the methods Varotsos uses to convert the observed electrical signals into predictions, which no one else can really understand. “It’s a black box,” Stark says. Varotsos maintains that the ability to interpret the signals comes with experience. If others don’t get it, he says, they haven’t done their homework. “Critics claiming that they don’t understand did not go carefully through our publications or did not participate in the relevant conferences.”

But detractors say that VAN’s predictions are so vague as to be unverifiable. The most important prediction parameters-time, location and magnitude-are quoted with wide margins for error. The predicted time can span several weeks, the predicted strength 0.7 magnitude units, and the site predicted for the epicentre may lie up to 100 kilometres away. Such vague predictions are hard to assess statistically. “The definition remains too ambiguous to perform a reliable evaluation,” Utada says.

Varotsos insists that the method is of practical use, even with the current margin of accuracy. But David Jackson and Yan Kagan of the University of California at Los Angeles, who have attempted statistical tests of Varotsos’s predictions, claim that the predictions are no better than can be ascribed to chance. And Geller asserts that while success rates of about 70 per cent are claimed in media reports of VAN’s predictions, the real rate is closer to 25 per cent.

Also, when the VAN researchers fax their predictions to the Greek government and to about 40 other institutes internationally, they sometimes fail to specify a key parameter, yet still claim success.

Varotsos’s modus operandi certainly seems to make it difficult to replicate and verify the VAN claims. But many of his critics go further than that. “He’s put forth the notion that he’s battling the conservative establishment, but that’s a terrible misconception,” Jackson says.

House rules

Varotsos counters that the procedure used by method’s critics to evaluate the VAN method is “highly questionable”. In another paper presented at the American Geophysical Union meeting, he contends that his critics have yet to establish how to evaluate any earthquake prediction method reliably, and says that they must put their own house in order before they can reject VAN’s results.

Critics have failed to produce any sound physical argument against the VAN method itself, Varotsos says. And Utada calls for “investigations from different angles” to be done. Kanamori and others would like there to be seismological, hydrological, magnetic and other geophysical investigations in concert with VAN’s giant electric prods. Park wants Varotsos to solidify his prediction rules, to allow them to be fairly tested over the next five years or so. And Geller and others call for standards to be set for predictions, so they can be evaluated more easily.

But with both sides digging in, a rapid consensus on VAN seems unlikely. Varotsos is sure his method will be vindicated. The critics are just as confident it will founder. “It might take another 10 years,” Wyss says. “This is not a question about science, this is a street fight.”