California issues first quake prediction.

In a drama that could have come from a Hollywood script, the San Andreas fault started showing signs of seismic activity beneath the tiny town of Parkfield, Calif., just in the nick of time. The moderate earthquakes that occurred there last week have raised scientists' hopes that a major jolt will shake this rural region before time runs out on the only longterm quake prediction ever sanctioned by the U.S. seismic community.

Located midway between Los Angeles and San Francisco, the town of Parkfield has suffered strong earthquakes on a remarkably regular schedule almost every 21 years. The small patch of the San Andreas running through this town produced jolts between magnitude 5.5 and 6.0 in 1857, 1881, 1901, 1922, 1934, and 1966 Recognition of this regularity led scientists with the U.S. Geological Survey to predict in 1985 that a magnitude 6.0 earthquake would occur along the 25-kilometer-long Parkfield segment of the San Andreas fault by the end of 1992. The USGS and the State of California funded a multimillion-dollar experiment to monitor the fault in hopes of issuing a shortterm prediction hours to days before the actual quake hit.

In recent years, some seismologists have criticized the original prediction and the decision to pour what has amounted to $19 million into the Parkfield experiment while restricting funds for other earthquake research. This year, with time on the prediction running out, the Parkfield segment of the San Andreas has remained particularly quiet, adding to the feeling that the original prediction had overestimated the quake's chances.

In early October, though, the Parkfield region started popping with a series of three small earthquakes, the largest of which reached magnitude 3.1, says John O. Langbein of the USGS in Menlo Park, Calif., who heads the Parkfield experiment. The three tremors emanated from the San Andreas near Middle Mountain -- a region where past Parkfield earthquakes have started. That activity trigered a C-level alert on a five-level rating system.

On Oct. 19, the same area of the fault produced a much larger jolt, measuring magnitude 4.7, which set off an A-level alert, the highest possible. This was the first A-level alert in the seven-year-long experiment.

As part of the prearranged Parkfield plan, the California Office of Emergency Services issued a public earthquake prediction - the first of its kind in state history. Officials announced there was a 33 percent chance that a magnitude 6 quake would come within three days, a statement that turned out to be a false alarm.

Scientists with the USGS say they are surprised the program went so long without an A-level alert. "We expected maybe two or three by this time," says William Bakun, one of the people who set up the Parkfield experiment.

The magnitude 4.7 tremor caused a stir because past Parkfield earthquakes have been preceded by similar or slightly larger shocks near Middle Mountain. But researchers last week did not see any of the other activity expected before a major quake. In 1966, the San Andreas fault began to creep several hours before the main shock, breaking an irrigation pipeline and producing cracks in the ground. The USGS has installed sensitive creepmeters along the fault, but they did not detect any abnormal activity last week.

After the A-level alert, the fault quieted down for several days. It reawoke on Oct. 25, producing several earthquakes near Middle Mountain, the strongest of which reached magnitude 3.9. This jolt trigered a B-level alert, signifying a 10 percent chance that the expected earthquake would come in 72 hours.

If the quake does not occur by the end of the year, the original Parkfield prediction will be judged a failure. The experiment will continue, but as the years pass it will be harder to keep equipment operating, says Langbein. "Scientifically, though, if the earthquake doesn't come now, but in three years, we'd be pretty happy with the results."