Lessons from a quake that came 16 years late

Most earthquakes make history for the damage they inflict. But last week's quake at Parkfield, Calif., is likely to be remembered for insights that could improve earthquake-hazard assessments, quake-resistant construction techniques, and, perhaps, efforts to forecast significant temblors.

Over the past 18 years, scientists turned the area, which straddles the San Andreas fault, into one of the most heavily instrumented fault segments on the planet. As a result, they have amassed enormous amounts of information.

When the quake happened Sept. 28, researchers recorded it as it happened. Now they are poring over all the data to unravel the specific factors that led to the quake.

Past efforts to test different theories about earthquakes and gauge their usefulness for forecasting have been stymied "because our instruments were too far from the earthquake to see anything," notes Lucile Jones, scientist in charge of the US Geological Survey's southern California earthquake-hazards team. "We were in the Catch-22 position of needing to predict an earthquake to collect the data to be able to learn how to predict earthquakes. Many of those questions may be answered" with insights from last week's event, she says.

For example, she notes that buildings near faults are designed for earthquake resistance using computer simulations of ground motion 10 to 100 kilometers from a quake's epicenter. Results vary widely inside 10 kilometers because instruments have never recorded an actual quake inside that radius. "We just got 100 records from within 10 kilometers" of the Parkfield quake's epicenter, Ms. Jones says.

In addition, quake experts hope to gain a better understanding of how quickly a fault releases its energy. Caltech seismology and earthquake engineering professor Tom Heaton had proposed that a sudden break in a fault such as the San Andreas can generate "fault fling" - a single large pulse of motion that can shove a building's foundation into the building itself. It appears as though fault fling occurred at Parkfield, giving engineers a close look at another source of stress to compensate for as they design new buildings.

Parkfield has been a mecca for seismology ever since researchers noted that the hamlet (pop. 37) has endured a magnitude 6.0 earthquake on average every 22 years since 1857. The culprit: the San Andreas fault, which forms the boundary between two large crustal plates on the Earth's surface - the Pacific and North American plates.

Before last week's event, the last magnitude 6 temblor to hit the region occurred in 1966. If past patterns held, researchers figured the next quake would happen in 1988. So in 1986, they started placing a dense array of instruments to measure changes in well-water level, crustal strain, magnetic fields, electrical properties, changes in the relative positions among sites on the surface, and creep along the fault itself.

Although last week's quake occurred much later than predicted, other aspects of the temblor lived up to expectations. According to the US Geological Survey, the location, magnitude, endpoints of the rupture, and the relative motion of crust on each side of the fault held true to forecasts.

Other aspects of the quake, which was felt from San Francisco to Los Angeles, present conundrums. Besides the timing issue, the quake began at the southern end of the segment and ruptured north - unlike several past quakes there. Last week's event also lacked light foreshocks, such as those felt prior to the quakes in 1934 and 1966.

The lack of foreshocks and the long wait for the latest quake could be tied to a phenomenon first uncovered at the site, notes Paul Silver, a staff scientist at the Carnegie Institution of Washington's department of terrestrial magnetism. Essentially, it's a slow earthquake.

In early 1993, slip along the San Andreas at Parkfield began to speed up by 30 to 40 percent. The change, observed over three years, was accompanied by four light quakes. Typically, seismologists had thought in terms of extremes, he explains. At the benign end is movement called aseismic slip, where the plates slide past each other in a continuous motion.

Indeed, Dr. Silver says, a section of the San Andreas north of Parkfield creeps along "with rarely any large earthquakes at all. It's slipping happily all the time." At the more violent end of the spectrum, the sides of the fault get locked in place. Stress builds up until the segment snaps and releases its pent-up energy at once.

The hybrid "slow quake" in 1993, by contrast, represented "an event we hadn't had much experience with before," Silver says. The discovery was "very important both for understanding the fundamentals of faults but also for the tantalizing possibility that some event like this might be a precursor to an earthquake, leading us toward the holy grail of earthquake prediction."