A New Algorithm Offers the Possibility of Faster Tsunami Warnings

Shaving crucial minutes off the time it takes to send tsunami warnings would give people more time to evacuate.  

Published January 28, 2016

By the time the warning went out, the waves were already landing.

Last September, a fault slipped just west of Illapel, Chile—a magnitude 8.3 earthquake that launched waves ripping across the Pacific. Chile’s coastal communities, just 50 kilometers away, were on the front lines.

Having learned from previous earthquakes, Chilean officials quickly ordered an evacuation. Sixteen minutes after the fault ruptured, some one million Chileans were told to flee. But by then, the waves had already begun to crash onshore—and even with the delay, the warning was incomplete. Officials had underestimated the earthquake’s strength and they couldn’t hazard a guess about the height of the waves bearing down on them. By the time they got it right and added a regional forecast, 15 more minutes had passed. By the end of the day, 13 people had died.

The tsunami warning was fast, but not fast enough. Seismologist Diego Melgar thinks he can do better.

In a new paper, Melgar and his colleagues argue that by using a new analytical technique they could have issued a forecast for the Illapel tsunami just two minutes after the ground stopped shaking—a full 12 minutes before the waves hit the shore. Similarly speedy warnings would have been possible for Chile’s 2010 and 2014 quakes and Japan’s 2011 Tōhoku quake, they say, had their system been in place.

Crucially, Melgar’s method doesn’t require any expensive new equipment: it’s simply a new way of analyzing and interpreting existing data. That means the technique could be rolled out quickly, and with little cost. The method uses common onshore GPS sensors, which track slight movements in the surface of the earth. More complex tsunami detection networks, such as the one operated by the US National Oceanographic and Atmospheric Administration (NOAA), rely on seismometers and expensive floating buoys.

“What we’re proposing … can be done now, today,” Melgar says. “With only a few stations, you can make a very rapid assessment of the earthquake’s size, and those [stations] exist everywhere now.”

By watching GPS sensors jiggle, Melgar’s new algorithm estimates what happened when the fault slipped, then models how the quake warped the seafloor, and simulates the ensuing tsunami. Within a few minutes, it pops out a color-coded map of the predicted at-risk coastal regions. Green means waves may be lapping at your knees, yellow predicts waves under a meter tall, and orange could signal a large tsunami up to three meters. Illapel, Chile and nearby cities would have seen red, indicating a wall of water three or more meters tall could be coming their way.

Melgar’s team thinks showing the public a quick estimate of where waves will likely be dangerous is warranted. In studies of Japan’s Tōhoku tsunami, the biggest factor determining who survived was simply when they started to move.

But the speed-for-accuracy tradeoff in the new method is also a weakness, according to Vasily Titov, head of NOAA’s Center for Tsunami Research. When comparing first-glimpse guesses of the magnitude of historical quakes calculated with the new algorithm against the known magnitude, the new approach’s estimates were only accurate to within 0.3. That leaves a lot of room for misleading predictions, Titov says: “Plus or minus point three is the difference between magnitude 8.4 and magnitude 9,” he says. “It’s a strong tsunami and a catastrophic tsunami.”

While Titov agrees that incorporating GPS data may help forecasts, he stresses that more testing will be needed before the new algorithm can make its way into NOAA’s alert systems. And it will have to prove its mettle not just against historical waves, but in real-time situations, too.

In the meantime, Melgar has reached out to NOAA’s tsunami warning centers. And, his Chilean collaborators are in talks with that nation’s navy, which runs Chile’s alert network and plans to adopt a similar system.

The “audience that we’re trying to reach here are developing countries across the Pacific Rim,” Melgar says, also naming Mexico, Columbia, and Peru. These are countries that face a significant tsunami risk, but may not have the funds to support a more elaborate detection network.

The tsunami warnings Melgar can issue using his new algorithm may not be perfect, but at least they would be sent before you can see the waves with your own eyes.