Quotes of the Day
Can scientists forecast major earth-quakes months before they happen? Most experts are dubious, but that hasn't stopped Vladimir Keilis-Borok from trying. And for a while the mathematical geophysicist, 83, seemed to be on a roll. In September 2003, within a time frame Keilis-Borok and colleagues had stipulated in advance, a pair of powerful earthquakes struck northern Japan, setting off tsunamis and injuring several hundred people. In December, another anticipated temblor rattled central California, killing two and damaging dozens of buildings. Then, in July, yet another quake, near the border of Austria, Italy and Slovenia, came close to fulfilling a third prediction (though there is good reason to think it was weaker than expected).
Last month, however, Keilis-Borok's winning streak very publicly ended. After nearly nine months of suspense, an earthquake warning that he and his team had issued for a 12,000-sq.-mi. area of Southern California expired without incident, marring what, up to then, had seemed to be a pretty impressive record.
From his office at UCLA, where he is a professor in residence, Keilis-Borok appears to be taking the setback in stride. At some point, he says, he had expected this to happen, adding, "If we made no mistakes, that would mean the problem of earthquake prediction is easy, which it is not."
The difficulty, in fact, is something Keilis-Borok seems to relish. He started wrestling with quake prediction in the 1960s and persevered well past the point where many others had given up. The leader of an international team of some 20 researchers, KB, as he is known to colleagues, now divides his time between UCLA, a physics center in Italy and his native Moscow, where he founded an institute for studying earthquake prediction.
Keilis-Borok calls his approach "tail wags the dog," the tail referring to patterns of seismic activity that appear to presage large tremors. (He does not try to forecast smaller events, like the earthquake swarms that rumbled beneath Mount St. Helens before it erupted last week, or the more significant quakes that perturbed Parkfield, Calif.) At first he and his colleagues looked for strong quakes that had already occurred, then scrolled backward through years of seismic data. More recently they have been working with current seismic records as well. Their computer programs home in on small quakes that occur in temporal and spatial proximity, linking up in a mathematically defined chain. Only when a chain is preceded by longer-term precursory patterns does the group issue an earthquake alarm. But it will take a long, sustained effort, Keilis-Borok concedes, before he or anyone else can claim to have cracked the puzzle.
- J. Madeleine Nash