Buying time when quakes hit

At 2:46 p.m. on March 11, 2011, an earthquake-detection station on Japan’s northeast coast began rocking back and forth, rattled by a powerful seismic wave racing from deep offshore. Just 5.4 seconds later, the Japan Meteorological Agency issued a notice that a magnitude 4.3 quake had begun.

As the seconds ticked by, however, and more stations picked up the rippling wave, the tremor started looking bigger. Three seconds after the first notice came an official warning: A quake of at least magnitude 7.2 was on its way. That’s a big tremor, even for earthquake-prone Japan. The city of Sendai needed to act quickly.

Televisions, radios and cellphones blared alerts. Trains screeched to a halt. Assembly-line robots froze in place and schoolchildren dived under desks. Fifteen seconds later, the biggest earthquake in Japanese history rocked Sendai
 — a monstrous magnitude 9.0 accompanied by a tsunami that disastrously flooded two nuclear power plants in nearby Fukushima.

Earthquakes are impossible to predict. But in the last few years, officials in various countries have started alerting the public once a quake is under way. Seismic sensors and communications networks have improved to the point that electronic alerts can race ahead of seismic waves. Such notifications give a few crucial seconds during which emergency managers can secure natural gas lines, factory workers can shut down hazardous equipment and surgeons can withdraw their scalpels from patients. In Sendai, those 15 seconds of advance notice may have saved many lives, making the catastrophic Tohoku earthquake at least a little less devastating.

Many other countries have seismic early warning systems. Japan’s nationwide system has been in place since 2007. Mexico City gets public alerts when a big quake begins. Even Romania uses a small network of seismometers to alert nuclear reactor workers when the ground is about to shake.

The United States has no public warning system for incoming quakes — yet. Officials are seriously talking about launching a full-fledged early warning system for the quake-prone West Coast. In September, California Gov. Jerry Brown signed legislation that requires the state to figure out how to fund an earthquake early warning system by 2016. Lawmakers have yet to pony up the money, but even so “we are the closest we’ve ever been,” says Richard Allen, a seismologist at the University of California, Berkeley.

California is close to taking action, in part, because of innovations that are improving the accuracy of warnings. In the last few years, earthquake specialists have begun incorporating real-time data from global positioning system stations, which measure how the ground moves. Traditional earthquake-monitoring systems rely on seismometers, which measure the energy of seismic waves passing through the ground, but do not do a good job measuring big shifts from big quakes. Adding the GPS data produces a better estimate of exactly how a big earthquake propagates across hundreds of kilometers, and therefore what kind of hazard it may pose to people. During the Tohoku quake, for instance, the Japanese early warning system relied mostly on seismometer data, and thus underestimated the strength of shaking very far from where the quake began. The system didn’t accurately warn people in Tokyo, 300 kilometers from Sendai, of their vulnerability.

New discoveries have essentially solved that problem. “We have the technology, we have the science, we have networks that can communicate quickly enough to provide warning,” says Allen, who directs the Berkeley Seismological Laboratory. “It doesn’t make any sense to wait for the next big earthquake. We should go ahead and do it.”

Call to action

The idea of earthquake early warning is far from new. In 1868, a San Francisco physician proposed setting up an alarm bell in the city, strung by telegraph wires to some sort of faraway mechanical device that would generate an electric current if the ground started shaking. “This bell should be very large, of peculiar sound, and known to everybody as the earthquake bell,” J.D. Cooper wrote in the San Francisco Daily Bulletin.

Turning the earthquake bell into reality took more than a century. After the 1989 Loma Prieta earthquake rocked the San Francisco Bay Area, the U.S. Geological Survey rigged a warning system to help protect workers who were trying to restore collapsed portions of an interstate in Oakland. USGS researchers peppered the area around the quake’s epicenter with seismometers. When an aftershock hit, the stations automatically radioed a warning to the workers in Oakland, about 80 kilometers away. “It was a temporary system, but it worked,” says Allen. In one case, it provided workers a warning as long as 20 seconds.

It took more than two decades for the next big advance. In 2011, the USGS and university partners debuted ShakeAlert, a prototype warning system that uses data from about 400 seismic monitoring stations across California. When an earthquake starts, the computer screen of a ShakeAlert user flashes a bright blue-and-yellow warning box. Numbers begin counting down to when ground shaking will begin at the user’s location and how strong it will be. A blaring alarm ensures that nobody misses the message.

Like other countries’ early warning systems, ShakeAlert relies on the fact that earthquakes generate different types of seismic waves. P waves, or primary waves, shake the ground back and forth in the direction of travel. They move faster — at about 6 kilometers per second — and thus arrive first. P waves are noticeable but not particularly damaging. After the P waves come the S or secondary waves, which shake the ground up and down. S waves generate most of the ground movement during a quake and cause most damage to buildings and risk to people.

By triggering when a P wave arrives, an earthquake warning system can provide seconds to tens of seconds of warning before the S waves show up. The farther you are from the quake, the more warning you get.

The real science of ShakeAlert is in how it converts seismic signals into a warning. It uses three different algorithms, or sets of calculations, to analyze the seismic data. Each algorithm has its strengths and weaknesses; together they are meant to add up to the best possible prediction of how much the ground will shake.

The first algorithm, called Onsite, issues an alert once one seismic station has experienced three seconds of P wave shaking, followed by a second station detection. Onsite generally spits out alerts faster than the other two algorithms. Its drawback is that it also sends out a lot of false alarms, particularly when the research team simulates large earthquakes.

The second algorithm, ElarmS, offers a nice balance of speed and accuracy, says Allen. It kicks into gear on just 0.5 seconds of P wave data. But it won’t send out an alarm until at least three other stations also sense the P waves. The idea is to cut down on the rate of false alarms by verifying motion at more than one station, and it seems to work, says Allen.

The third algorithm, with the less catchy name of Virtual Seismologist, triggers on three seconds of P waves, like Onsite. It’s the slowest of the three because it runs through a series of complex analyses (taking into account factors such as the region’s known fault hazards and the health of each station) before it issues any warning.

ShakeAlert combines all three algorithms to generate an estimate of how large a quake is and where its energy is radiating. The system’s organizers tweak it regularly, sometimes giving more weight to ElarmS while dialing back on the Virtual Seismologist. Then they test it by seeing how well each combination performs on the small quakes that frequently rattle the California coast.

So far, the only groups who receive ShakeAlert alarms are carefully screened organizations that want to use the information in emergency planning and are willing to tolerate a few false alarms to help improve the system. They include the biotech company Amgen in Thousand Oaks, the Disneyland Resort in Anaheim, Los Angeles County’s emergency managers and the BART rail system in the San Francisco Bay Area.

Earthquake early warning is particularly important for high-speed transportation, says John McPartland, a director of BART. On any given workday there could be 45 BART trains speeding along at up to 112 kilometers per hour. Those trains automatically start to decelerate if they receive a ShakeAlert alarm. “Within 24 seconds we can get the train to a complete stop,” McPartland said at a meeting of the American Geophysical Union in San Francisco in December. “That’s a huge advantage.”

Simulations show that if the Loma Prieta quake were to happen again and an early warning system were in place, the cities where BART runs would get about 20 seconds warning. That’s enough time to halt most trains, saving them from derailing.

(Story continues below the image)

Up the coast

California is the capital of U.S. earthquake early warning because it is the state with the highest seismic activity (after sparsely populated Alaska) and it already has the densest network of seismometers. The Pacific Northwest, however, faces the prospect of an earthquake that could dwarf even a Big One in California.

Here, the greatest seismic risk is the subduction zone known as Cascadia, where the Juan de Fuca crustal plate dives under the North American continent. The zone has been essentially locked in place since its last great quake, in the year 1700, when it ruptured in an estimated magnitude 9.0.

A similar quake today could devastate coastal communities as well as the major cities of Seattle, Portland and Vancouver. On the bright side, the fault is at least 75 kilometers offshore, far enough that an earthquake early warning system could give those cities up to 30 seconds of warning before the shaking hit.

Quakes in Oregon and Washington are monitored by a University of Washington–led group called the Pacific Northwest Seismic Network. The group joined ShakeAlert in 2012.

Because a Cascadia quake would be so enormous, any early warning system would have to accurately assess its magnitude from the start. And that’s the problem with the early generation of ShakeAlert algorithms. Essentially, they treat an earthquake as occurring at a single point rather than recognizing that very large earthquakes can rupture for long distances, unzipping faults along the way. The Japanese system didn’t warn Tokyo appropriately of the Tohoku quake because it failed to account for the quake’s sheer physical size. It takes an entirely different set of calculations to account for seismic energy radiating from a fault plane rather than from a single point.

“We had known this was a limitation, but we were just focusing on getting the point-source algorithms working,” says Allen. The Tohoku quake “was a push for us to solve the problem.” In the last few years, seismologists have radically improved their algorithms to accurately capture these large quakes.

One new algorithm, called FinDer, maps a rupture in real time by comparing ground shaking measured by a seismic network with a set of precalculated values, says Maren Böse, a seismologist at Caltech who helped develop the system. Böse tests FinDer by simulating how it would have performed in hypothetical or past quakes. It does a good job, she says, at figuring out the real-time rupturing during a magnitude 7.8 test scenario on southern California’s San Andreas fault.

The system could do even better by incorporating measurements from GPS stations. Whereas seismic stations measure waves of seismic energy passing through the ground, GPS measures the large physical displacement caused by the ground moving. Such measurements provide a direct look at how far and how fast a fault is ripping apart.

In California, data from dozens of GPS stations near San Francisco and Los Angeles flow into ShakeAlert along with the traditional seismic
information. Combining the two makes for more accurate warnings, says Allen; ShakeAlert is testing several algorithms, including a version of ElarmS that incorporates GPS data. Once these are fine-tuned to catch earthquakes as best as possible, they will likely become a permanent part of ShakeAlert.

The blind zone

The team is also working to reduce the blind zone, or the area so close to a quake that there is essentially no time to provide a warning. In California, the distance between seismic stations varies dramatically; in cities they can be less than 5 kilometers apart, whereas in the sparsely populated northern counties, 70 kilometers may separate them. The farther apart the stations, the sparser the information flow in the case of an earthquake.

Allen and his colleague H. Serdar Kuyuk, also of UC Berkeley, recently studied how much they could shrink the blind zone if stations were closer together. Their calculations showed that a typical California earthquake, occurring 8 kilometers deep, would mean a blind zone about 32 kilometers across no matter how closely spaced the stations. Within that circle, there’s simply no time to disseminate an alert based on P waves before the more damaging S waves strike. Still, to give as many people as possible the best warning, stations should be placed less than 10 kilometers apart in urban areas on known faults, the team wrote last year in Seismological Research Letters.

All such suggestions remain purely theoretical unless the state of California decides to fund the next step for ShakeAlert. To become truly operational, the state will need to build hundreds more seismic and GPS stations across the state, as well as roll out a huge public education effort to tell people what to do once they get an earthquake alert. “An early warning that you don’t know what to do with is not an early warning,” says Peggy Hellweg, operations manager at the Berkeley Seismological Laboratory.

The cost to build and operate a California-wide system for five years would be $80 million; a Pacific Northwest system would require another $40 million. After the initial five years, operating costs would run $16 million annually for the entire West Coast. That’s about twice what the region currently spends on earthquake monitoring.

California’s Office of Emergency Services has been tasked with finding money for the state’s contribution; it is due to report back with options for what a system might look like by the end of June. If money isn’t found by January 2016, the law requiring an earthquake early warning system expires.

Yet between the legislative push and the new technological developments, scientists hope to have an early warning system in place to buy crucial time for West Coast inhabitants before the next big quake. “It’s amazing to see such rampant progress both in terms of developing new ideas and in turning those ideas into things that are actually working,” says Thomas Heaton, a seismologist at Caltech. “It’s been a very exciting time for us.”

---

Personal quake detectors

Chances are, you’ve got an earthquake detector in your pocket.

Earthquake specialists are hunting for new, nontraditional sources of seismic data to flesh out traditional seismic monitoring. Any device that can measure shaking and is hooked into a network that tracks its location has the potential to become a quake-catcher. In California, there are about 16 million smartphones fitting that description.

At the University of California, Berkeley, Richard Allen and his colleagues are developing a smartphone app, called MyShake, that uses a phone’s built-in accelerometer to sense shaking from earthquakes.

“This is never going to replace our traditional seismic networks,” says Allen. “It’s just an additional source of data.” Accelerometers inside today’s phones can detect a magnitude 5.0 earthquake that strikes within 10 kilometers, and soon should be able to detect a magnitude 3.0 within 100 kilometers, he says. Because an earthquake ripples over such a large area, the scientists can distinguish the distributed pings of seismic shaking from an individual iPhone lurching around in a backpack.

People without smartphones can become part of the Quake-Catcher Network, which uses small accelerometers bolted to the floors of houses and offices and wired to Internet-connected computers. Quake-Catcher software monitors signals from its 2,000-plus sensors, and by combining them can rule out sources like trucks rumbling by or doors slamming. Because geology can vary on the scale of city blocks, the shaking measured by neighborhood Quake-Catchers can reveal which addresses are particularly vulnerable in a given earthquake. “It just makes a lot of sense,” says seismologist Elizabeth Cochran of the USGS Pasadena office.

Crowdsourcing earthquake sensors is one approach; another is to monitor the flow of quake chatter online. Seismologist Rémy Bossu watches traffic on the website of the European-Mediterranean Seismological Centre, in Strasbourg, France, to see who might have just felt an earthquake and runs to the Internet to check it out. And Paul Earle of the USGS in Golden, Colo., uses Twitter to confirm ground shaking by searching for the keyword “earthquake” in tweets.