How good is earthquake early warning?

By Elisabeth Nadin, Ph.D., Associate Professor, University of Alaska Fairbanks

Scientists can’t predict when earthquakes will occur. They can however, with reasonable certainty, identify regions at elevated risk for damaging earthquakes. A recent magnitude-6.7 earthquake in Turkey and subsequent recovery efforts prompted the country’s interior minister to warn citizens of the potential for an even larger event in the region.

How many lives could be spared or injuries prevented during such an event if people were alerted to imminent shaking, even just seconds before seismic waves hit? Researchers have exactly this question in mind when designing Earthquake Early Warning systems. Early warning systems utilize local or regional seismic networks to detect initial shaking from an earthquake. Alerts are then sent to the public through television or radio broadcasts and increasingly through mobile phones. Several such systems exist around the world, including ShakeAlert, a U.S. West Coast-wide system that became public in California last October, but exactly how effective they are has not been well quantified. New results from a test of existing warning algorithms tells us that alerts are received in a timely manner in a large number of cases.

An amazing tool or too slow?

Significant resources have gone towards developing these systems and researchers have touted the injury and death prevention and economic benefits of early earthquake warnings. But is this money and time well spent?

“There’s a debate about the potential usefulness of early warnings,” says Men-Andrin Meier, a seismologist at Caltech and the lead author of a recent study published in the Journal of Geophysical Research focused on this subject. “In our community, proponents say this is an amazing tool; we can reduce risk by a lot with this technology. Critics say it’s too slow for the close sites that have the most damage.”

This is why Meier wanted to test how well early warning systems perform where the shaking is greatest. “The biggest potential for early warning is in vulnerable situations, for example a construction site. So many things can collapse on a construction site. You don’t need a lot of warning to take a step back and put yourself in a safe situation,” he notes. Or to “drop, cover and hold on” as the U.S. Geological Survey (USGS) advises.

Who benefits from an early warning?

The intensity of an earthquake varies with location—the area around the epicenter is the first to feel an earthquake and will often experience the most intense shaking. For Earthquake Early Warning systems, this is usually a “late alert zone” because seismic waves arrive before an alert. Unfortunately, this is the area that needs the alert the most. In contrast, sites with moderate and weak shaking, which are much farther away, are successfully alerted before shaking but may not have been at risk of significant damage anyway.

Herein lies the debate about early warning utility: Is it really useful in the places that need it the most? Meier and his colleagues decided to test exactly that. Using data from more than 200 earthquakes in Japan that occurred over the last 20 years, the group compared observed ground motion to the ground motion predicted by three different early warning algorithms. They evaluated who would have gotten an alert and when. Similar previous tests used synthetic data or hypothetical early warning algorithms.

“We should have done this type of study 10 years ago, but we were busy developing the algorithms,” says Meier.

The team used the Japanese data because “they have amazing networks. And they make the data freely available,” Meier says. Meier and his team discovered that about 50 percent of all sites with the strongest shaking levels receive timely alerts. That is, they receive alerts at least five seconds before the shaking becomes noticeable. The other half are in the late alert zone. This means that currently existing early warning algorithms can alert about half of all sites that are at greatest risk from damaging seismic waves. In detail, the number of sites that can be successfully alerted varies depending on how large the earthquake is and how deep it occurs. For sites with medium and lower shaking intensities, the current algorithms can alert up to 90 percent of sites before shaking is felt.

Hope for the future of Early Earthquake Warning

Alerts may improve as algorithms improve as well, notes Meier. Also, people may have additional time to take cover, even after the shaking is already noticeable. “It can take quite some time until the really strong shaking starts.” But he points out that it will remain difficult to provide early warning for those within the first 30 to 50 kilometers (19 to 30 miles) of the epicenter. The seismic waves simply get to them too quickly. “The biggest limitation is the incredible speed with which seismic waves propagate. We’re in a race against this really fast wave,” he says. But Meier is optimistic, noting that “it turns out that it’s a race we can often win.”

This test is critical for people who live in earthquake-prone areas and might rely on early warning systems. Annemarie Baltay, a USGS scientist and member of California’s ShakeAlert team, is positive about the study’s results. “Critics will say that we can’t send early warnings to the most critical sites. But the results give us the confidence that some percent of the time we can send useful warnings.”

Studies like this help ShakeAlert managers think about how to fine-tune their system, Baltay says. For example, a surgeon might not want to interrupt a procedure in the operating room for low-level shaking but would definitely stop with high shaking, so maybe the hospital only wants alerts about higher-intensity quakes. In contrast, schools or public transportation systems might want an alert for any earthquake as a way to test their emergency responses. “We need to describe the potential performance to them,” says Baltay. “Then we’ll have a better idea of how to further develop warnings,” she says.