Of the forces of nature that threaten humans, earthquakes may be the most worrisome, in large part because they always seem to strike without warning. For years geologists have been trying to figure out how to predict these often-catastrophic events. Forecasts in a number of temblor-prone regions--including California, Japan and New Zealand--currently rely on the so-called time-predictable recurrence model, which holds that an earthquake occurs when the geological fault recovers the stress released in the most recent quake. Findings published today in the journal Nature, however, place that model on shaky ground.

According to the time-predictable model, after an earthquake, strain accumulates steadily as the shifting tectonic plates continuously grind together. Thus "if you know the size of the most recent earthquake and the rate of strain accumulation afterwards, you should be able to forecast the time that the next event will happen simply by dividing the strain released by the strain-accumulation rate," explains Jessica Murray, a geophysics graduate student at Stanford University. To test the theory, she and Paul Segall, also at Stanford, applied it to data collected from a town in central California known as Parkfield, which rests on the infamous San Andreas Fault (see image). Parkfield, it turns out, has suffered five quakes measuring magnitude 6 on the Richter scale at regular intervals since 1857, with the last one striking in 1966. "Parkfield is a good place to test the model because we have measurements of surface ground motion during the 1966 earthquake and of the strain thats been accumulating since," Murray observes. "Its also located in a fairly simple part of the San Andreas system because its on the main strand of the fault and doesnt have other parallel faults running nearby."

Murray and Segall determined that the strain released in the 1966 quake had actually recovered by 1987. The earthquake that one would expect based on the time-predictability model has not come, however. "In fact, 15 years have gone by," Murray notes. "Our results show, with 95 percent confidence, that it should definitely have happened before now, and it hasnt, so that shows that the model doesnt workat least in this location." And if it doesnt work in a tectonically simple setting like Parkfield, she adds, its even less likely to make accurate predictions in more complex locales, like the Bay Area or Los Angeles.