Foreseeing Disaster.(Predicting earthquakes.)

Geologists have long known about Japan's vulnerability to the rumbling earth. Back in 1978 the Japanese government enacted a law to ensure that the nation took steps to watch for impending earthquakes; by the end of 1994, it had spent more than $1.3 billion on monitoring programs. Yet in the early morning hours of Jan. 17, 1995, when the Great Hanshin-Awaji Earthquake came rumbling along at 7.2 on the Richter scale, it thoroughly surprised the port town of Kobe. The quake flipped the elevated Hanshin Expressway onto its side like a child's discarded slot-car track, caused skyscrapers to collapse and gas fires to break out among the city's wooden houses. More than 6,400 people were killed and 43,000 injured. "[Earthquake] prediction research betrayed the public," snapped an editorial in Asahi Shimbun.

The Kobe earthquake forced both the Japanese government and researchers to take stock. In the past few years, Japanese scientists have been trying to find out why they failed in predicting Kobe and other earthquakes that have hit Japan over the years. Now that effort is paying off. In a recent issue of the journal Science, Jin-Oh Park and his colleagues at the Japan Marine Science and Technology Center in Yokosuka, near Tokyo, uncovered a major cause of the most unpredictable earthquakes. It was a "splay" fault, a secondary fault that branches off a main one and can cause damage hundreds of kilometers from obvious trouble spots. This was the first time scientists confirmed the existence of a splay fault. By shedding light on the vexing phenomenon, experts think they may be able eventually to predict many types of earthquakes that have tricked geologists in the past. "This brings us a step closer to knowing about earthquake mechanism, and could lead to a better chance of prediction," says Shozo Harada, a Diet member, who has helped draft anti-quake laws.

Even before Kobe, Japanese scientists had come to suspect that predicting earthquakes was not as simple as once thought. They were expecting the next big temblor to arrive at the most obvious place--in the populous Tokai region, west of Tokyo, rather than Kobe, which is even farther to the west. The mistake was understandable, given that the Tokai region lies near the fault produced by two colliding tectonic plates--the Philippines Sea Plate and the Eurasian Plate, which was (and still is) long overdue for an earthquake. In the 1970s, a University of Tokyo researcher warned that a massive earthquake would hit Japan's Tokai region "any day." At the time the warning so spooked Japan's politicians that Tokai almost monopolized Japan's earthquake prediction and disaster prevention efforts for the next 25 years. "While [the government] focused on Tokai, all the recent major quakes, including the Nihonkai-Chubu earthquake (which killed more than 100 in 1983) and Kobe, which was a surprise to everybody, occurred elsewhere," observes Masataka Ando, a prominent seismologist at Nagoya University.

The Kobe disaster gave scientists a good reason to check out those neglected areas. Two winters ago, Korean-born Park was poring over seismic data that had been gathered ...