Close Mobile Menu

There’s an App for That! Earthquake Early Warning Is Here.

October 31, 2019
by Robin Estrin

On the 30-year anniversary of the Loma Prieta earthquake, Governor Gavin Newsom stood near the Oakland entrance of the Bay Bridge, a portion of which collapsed during the quake, and announced the launch of MyShake, the United States’ first earthquake early warning system.

“The price of admission for living here is preparation,” Newsom said. “Today we are making a big leap forward.”

Had an early warning system existed when Loma Prieta erupted beneath the Santa Cruz Mountains 30 years ago, UC Berkeley seismologists believe that ShakeAlert, the app’s underlying earthquake early warning system, would have detected the shaking and generated a warning within five seconds of its onset. Just four seconds later, MyShake would have pushed the warning to app users across the Bay Area. San Jose would have had five to eight seconds to take cover before shaking intensified. San Francisco’s Marina District, where four people died, would have had 20 seconds before the most destructive tremors wreaked havoc on the neighborhood.

Bay Bridge earthquake damage

Bay Bridge damage from the Loma Prieta Earthquake // Photo courtesy of Joe Lewis, flickr / CC

Earthquake early warning has the potential to provide life-saving seconds to millions of Californians. But the system is imperfect, and statements from its developers have been clear—as users, we need to manage our expectations accordingly.

“It’s inevitable that as we go forward we will make mistakes,” said Richard Allen, MyShake developer and director of the Berkeley Seismological Laboratory. “But I think we can go through a process of educating everybody about what’s possible, and educating [ourselves] about what people want, and then we will have a system that really is effective.”

MyShake is designed to deliver notifications generated by ShakeAlert, a network of seismic sensors developed by the United States Geological Survey (USGS) and its university partners, which include Berkeley, Cal Tech, the University of Washington, and the University of Oregon. ShakeAlert detects shaking, estimates an earthquake’s location and magnitude, and generates an alert—all within a matter of seconds.

Whether or not those alerts are delivered to the public depends on a third-party messenger: an app.

Early this year Los Angeles Mayor Eric Garcetti delivered on his promise to bring early warning to Los Angeles by launching the mobile app ShakeAlertLA. The app made headlines in July for failing to notify users after a magnitude 7.1 earthquake ruptured beneath the small desert town of Ridgecrest, rattling the homes and nerves of many 100 miles south in Los Angeles.

Multiple reports stated that ShakeAlert generated an alert as programmed, but that it wasn’t sent out because the app’s notification threshold was set too high, a story Newsom reiterated at the press conference.

“We are much closer to having a system that people will see as effective now than we were at the time of the Ridgecrest earthquake,” Allen said. But still, there is much to be learned.

But according to the Seismology Lab’s website, the system didn’t send out an alert because ShakeAlert underestimated the magnitude of the second large quake, generating a warning for a magnitude 6.3 earthquake, rather than what it actually was: a 7.1. If the initial magnitude estimate had been accurate, the shaking intensity threshold would have been met, and people in downtown Los Angeles would have received a warning, and 49 seconds to prepare.

That may not seem like a lot of time, but just a few seconds are enough for a person to dive under a sturdy desk or for a surgeon to, say, lift a scalpel from a patient’s eye.

Part of the problem was that seismologists, not social scientists, designed ShakeAlertLA’s notification thresholds. App developers thought people would want to be notified of shaking strong enough to cause harm or structural damage; they learned that people expected notification any time they felt shaking.

“The failure of Ridgecrest was a failure of understanding users’ expectations,” Allen said, reflecting on the Independence Day quakes.

After a strong public outcry, the USGS and the city moved quickly to lower the app’s notification thresholds, a change mirrored with the MyShake app. ShakeAlertLA and MyShake users will receive notifications when an earthquake erupts at a magnitude 4.5 or greater, and produces shaking at an intensity three or higher on the Modified Mercalli Scale (which measures how an earthquake is experienced by people at different distances from its epicenter).

“We are much closer to having a system that people will see as effective now than we were at the time of the Ridgecrest earthquake,” Allen said. But still, there is much to be learned.

Developing a system that can accurately measure an earthquake’s intensity and location is no easy feat; neither is creating a mobile app capable of pushing notifications to millions of users in just a few seconds. The apps developers believe the system is ready for public use, but Ridgecrest may have revealed some of its potential weaknesses.

“People think that ShakeAlert is going to be the panacea, and it’s not. If you have bad building codes, ShakeAlert is not going to save you.”

At the time of the 7.1 quake, the algorithm responsible for estimating its magnitude did so with just four seconds of seismic data—not enough to capture its full intensity. In previous tests, four seconds had been enough to produce an accurate estimate, but Allen said the Ridgecrest miscalculation has pushed researchers to explore the use of more data.

But more seconds of data could come at a cost—fewer seconds of early warning.

The system of seismic sensors that undergirds the early warning app is not fully built. According to USGS national ShakeAlert coordinator Robert de Groot, a dependable system would require almost twice as many sensor stations as there are today. Magnitude estimates will likely improve as more are installed, said de Groot. But until then, errors like the one that occurred during Ridgecrest are just what MyShake users should expect, Allen said. Shaking might be significantly stronger or significantly weaker than the system anticipates. Warnings may arrive after shaking is felt, or if the system underestimates the magnitude, not at all. (And those closer to an earthquake’s epicenter than the nearest sensor won’t receive early warning.)

As we contend with the unpredictability of the next big one, USGS social scientist Sara McBride said that Californians should do their best to reframe their thinking about the early warning system. It may reduce risk, but in earthquake country, it won’t eliminate it.

The USGS predicts with 99.7 percent certainty that a quake of magnitude 6.7 or greater will strike California in our lifetime. No amount of warning can prevent that likelihood, or slow down the speed at which seismic waves travel through the earth’s crust.

The launch of a statewide early warning system is an achievement, McBride said, but it’s one that should be viewed as a part of a wider ecosystem of earthquake resilience tools. Knowing what protective actions to take—and stocking a disaster supply kit—are just as important as the newest technology.

“People think that ShakeAlert is going to be the panacea, and it’s not,” de Groot said. “If you have bad building codes, ShakeAlert is not going to save you.”

So, if you download the app, remember that technology can only go so far. It’s ultimately up to us, the users, to get prepared before the big one strikes.

Share this article