Berkeley's New Early Warning System Detected SF's Quake 10 Seconds Before It Hit

Ten seconds before a magnitude six earthquake hit the San Francisco Bay Area at 3:20 AM yesterday, a Star Trek-style red alert klaxon went off, giving scientists and a handful of beta testers a heads up of the impending natural disaster.

"Earthquake! Earthquake! Light shaking expected in three seconds," the automated message exclaimed from any computer running the Berkeley Seismological Laboratory beta software.

The early warning system uses existing seismometers to trigger an algorithm that crunches the quake's size and point of origin. The computers in Berkeley's lab then arbitrage the time-lag between an earthquake's P waves, and destructive S waves. The P waves, which travel 18,000 miles per hour, send the warning out.

The tremor was the shakiest I've ever experienced, and had me out of bed, and cowering under a table in about 30 seconds (admittedly, it's hard to pinpoint since I was fast asleep when the quake hit). The American Canyon quake—centered in wine country Napa county—is the most serious Bay Area seismic event in 25 years, the last being Loma Prieta in 1989.

That was a 6.9 magnitude ground-shaker which wreaked havoc on the region during a World Series game in Oakland. And this one, while less severe, still sent nearly 200 people to the hospital, knocked out power to tens of thousands, and caused millions of dollars in damage.

Earlier warnings could have softened the blow. And the earlier the better. Which is why minimizing all potential time lags is critical to the system's success: The system takes advantage of the fact that communications move at light speed communications, or 671 million MPH, much faster than the P or S waves, according to Jennifer Strauss, a spokeswoman for the Berkeley Seismological Laboratory. Computation and the lack of sensor density—the time it takes a P wave to hit a seismic sensor can cause a delay of a couple seconds or more—all factor into the scientist's system.

Also, the further away from the epicenter, the greater duration of the warning. For this quake, Berkeley was 10 seconds from the epicenter in the American Canyon in Napa county. If I were in San Francisco or Silicon Valley, it would have given me a couple more seconds, Strauss said. The Loma Prieta quake would have generated a 20 second warning for San Francisco, she explained.

Since the warning system is still in development, only a handful of people and institutions actually received a heads up—one of which is the regional rail network called the Bay Area Rapid Transit (BART).

"BART is really fully integrated, with all of the pipelines to use the early warning system laid down," Strauss explained over the telephone. "They got the alert, but at the time there weren't any trains running, so there was no trigger." If it had been during operating hours, it might have slowed or stopped trains—especially those going about 70 MPH—the only exception being trains traveling through the tunnel under the bay connecting Oakland and San Francisco.

Other parties interested include Deutsche Telekom, Google, and San Francisco's Department of Emergency Management.

Once the beta test is more evolved the few seconds provided by the warning system could significantly reduce a quake's impact. For example, Strauss said, if elevators had a 10 second warning before a quake, they could automatically stop and open their doors at the next floor. Or, she said, if a hospital were part of the early warning network, its systems could release a bed count to all emergency services—allowing ambulance drivers, and other first responders to more rapidly get the injured to hospitals with free space.

Once the system is ready for mass messaging, it would give people a chance to get under cover, and out of danger, she said. It's not ready for mass distribution at the moment, because the scientists are still working out the kinks in the system. There's also the red tape to clear with US Geological Survey, and ample testing that has to be done before mass distribution could prove feasible.

As the future continues to barrel onward, barriers to implementing a more widespread earthquake warning system are less and less significant. "Because we're so comfortable with computers and robots now, automated controls have become possible," Strauss said. "And we're always parsing the data on how we can make the alert better, constantly running diagnostics on this system, and replying events, testing the algorithms."