Earthquake early warnings can come as false alarms — but it’s better to be safer than sorry, researchers conclude in a new study.
Scientists at the U.S. Geological Survey, USC and Caltech worked on a research project, published Thursday, to determine the limits of the accuracy of seismic warnings. They conducted a statistical simulation to determine what strategy would be the best to save lives and protect the economy — is it better to get more warnings but also more false alarms? Or is it better not to receive an alert unless it’s certain there will be shaking, even though that might cause you to miss a warning?
People might have their own personal answer. But the scientists decided to take a different tack to address the dilemma: math.
In research detailed in the journal Scientific Reports, the scientists simulated tens of millions of hypothetical earthquakes in California. They imagined how forecasts made by an earthquake early warning system would compare with how shaking might actually be felt.
First, they found there was no way to have an earthquake early warning system that was 100% accurate. It’s scientifically impossible. Earthquakes that produce the same amount of overall energy — those that have the same magnitude — can produce much different shaking.
Shaking may feel different depending on whether the energy is aimed toward or away from you (like how a train horn sounds different depending on whether it’s speeding toward or away from you). One earthquake could produce intense shaking in a short period of time; another of the same magnitude could do its jiggle slower, resulting in a mellower roll. And the earth’s movement could feel less intense on solid bedrock but worse on soft soil.
Second, the scientists calculated that any particular California location on average had the chance of feeling an earthquake — a minimum of light shaking — just twice per decade.
And here comes their key finding: An early-warning system that gives you an excellent chance of being alerted to those twice-in-a-decade earthquakes will probably deliver four warnings that turn out to be false within that same time period.
“It’s a small price to pay,” said research geophysicist Sarah Minson, “if you’re talking about something where there’s a lot of benefit to be had.” For instance, a potentially deadly derailment could be avoided by slowing a train before the predicted shaking is set to arrive, said Minson, lead author of the study, on Thursday at the USGS office in Menlo Park, Calif.
On the other hand, if your priority is that any warnings you receive are almost certain to precede shaking, you’re far more likely to get no warning at all.
For some, the choice will be obvious.
There are costs to inaction before a big earthquake: You fail to drop, cover and hold on and an unstrapped bookcase tips over and flattens you. A dentist is unable to remove the drill from a patient’s mouth prior to the shaking. A butcher doesn’t have the chance to move away from the deli slicer. Power company workers aren’t warned in time to hang on to a utility pole.
In those cases, it’s an easy call to opt for an early warning even if it means no shaking comes.
When shaking is relatively light, seconds of warning still may provide a sense of comfort for some.
On St. Patrick’s Day in 2014, KTLA-TV anchors were stunned when they felt light shaking — defined as Level 4 on the Modified Mercalli Intensity Scale — at their Hollywood studio from a magnitude 4.4 temblor that began in Encino, 10 miles away.
The quake was scary enough that anchors Megan Henderson and Chris Schauble ducked under their desk.
The announcement — “Earthquake! We’re having an earthquake!” — ended up giving KTLA reporter Eric Spillman in downtown Los Angeles a warning that a temblor had begun and shaking was on its way. Spillman said he felt the ground move seconds later; downtown is farther from Encino than Hollywood is.
(Shaking produced by an earthquake moves at the speed of sound through rock — slower than the speed of today’s communications systems.)
There are situations where it may not be prudent to act on a seismic warning, given the high cost of action during a false alarm.
Starting an emergency shutdown at a nuclear power plant when an alarm sounds probably doesn’t make economic sense “because an emergency shutdown costs more than $250 million, not including other costs such as the resulting decrease in the lifetime of the reactor,” the study said.
Seismic warnings can be difficult to calculate because earthquakes all start the same — tiny. Although many stay small, a few do end up growing to be monsters.
Consider the hypothetical example of an earthquake that begins moving along the San Andreas fault near Eureka and heads toward San Francisco, about 200 miles south.
Four seconds after the shaking begins, the temblor is a magnitude 6 earthquake. If the quake stopped there, San Francisco probably wouldn’t feel a thing.
But if the quake lasts at least 20 seconds or so, becoming a magnitude 7, San Francisco could feel at least light shaking — Level 4 intensity — and would have maybe 48 seconds to prepare for it, Minson said.
But say someone only wanted to get a warning for very strong shaking — Level 7 — in San Francisco. That kind of quake would need to last at least 67 seconds, meaning it has produced enough energy that it has become a magnitude 7.7 earthquake. Waiting for that level of certainty and severity before an alert is sent would give San Francisco only eight seconds to prepare.
The USGS has been building an earthquake early warning system on the West Coast for years, and the system is most advanced in the urban areas of California, particularly in the L.A. region and San Francisco Bay Area. Following an influx of federal and state funding, officials hope that by 2021 all 1,115 seismic sensor stations intended for California will be online.
The city of Los Angeles recently released a public app for iOS and Android devices, ShakeAlertLA, to warn people in L.A. County when at least light shaking is expected. Another app, produced by Santa Monica-based Early Warning Labs, is expected to be made available to 100,000 test users soon.
Besides Minson, the study was coauthored by Annemarie Baltay, Elizabeth Cochran, Thomas Hanks, Morgan Page, and Sara McBride of the USGS, Kevin Milner of USC, and Men-Andrin Meier of Caltech.