Read our COVID-19 research and news.

The magnitude 7.8 Kaikōura earthquake in 2016 ripped up highways and hillsides on New Zealand’s South Island.

REUTERS/Anthony Phelps

New Google effort uses cellphones to detect earthquakes

Google is getting further into the business of saving lives. Today, the internet giant announced that users of its Android phones in New Zealand and Greece will receive warnings of damaging earthquakes about to strike their locations. And those earthquakes will be detected not by the usual seismometers, but by the phones themselves.

Earthquakes are a well-known threat in both countries. In New Zealand, the Pacific Plate collides with the Australian Plate and their grinding regularly causes large quakes, including a 2011 shock in Christchurch that killed nearly 200 people. Greece is spread across three tectonic plates, with damaging quakes a near-annual occurrence. But neither country has deployed an operational warning system. That created an opportunity to make a difference, says Marc Stogaitis, the project’s lead engineer at Google. “We have two big problems we want to solve: detecting earthquakes as quickly as possible and sending out alerts as quickly as possible.”

Such earthquake early warning systems take advantage of a simple fact: The speed of light travels faster through the fiber optic cables of the internet than an earthquake’s waves. Traditional warning systems use seismometers to detect an earthquake’s size and magnitude, then relay a warning, via smartphone or loudspeakers, to residents likely to feel the quake. Such warnings, even if they only come seconds before a quake hits, can buy people enough time to drop to the floor, take cover beneath a desk, or hold on until the shaking stops. These systems are robust, but they are difficult and expensive to develop. One system, known as ShakeAlert, took 15 years to create and deploy in California, Oregon, and—starting next week—Washington state. It cost $60 million to build and needs more than $30 million annually to operate.

An earthquake warning system that can forgo this expense and rely instead on smartphones has tremendous potential globally, says Men-Andrin Meier, a seismologist who studies earthquake warning systems at ETH Zurich. “If you can do it with just the phones, via Google, that’s an enormous shortcut.”

A phone-based network is possible because modern devices are equipped with accelerometers that monitor movement, such as when a user picks up or rotates the phone. But those motion sensors can also be programmed to act as rudimentary seismometers, detecting the distinctive shaking caused by the pressure and shear waves of earthquakes. But getting users to download the apps that would make such phone-based networks a success has been a challenge. So Google’s interest in building quake-sensing capabilities directly into its more than 2 billion active Android phones, downloaded in a standard system update, presented an enormous opportunity, says Richard Allen, a seismologist at the University of California, Berkeley. “For us, it was a complete no brainer,” says Allen, who helped build ShakeAlert and has advised Google on its earthquake program.

Android phones have been publicly detecting earthquakes since last year, but those results have so far only been available in Google search results. False alarms haven’t been too hard to weed out: Early on, some phones thought AMBER Alerts issued in the United States, which cause a phone to vibrate to warn users of a missing or abducted child, were quakes; the shaking caused by the sound of thunderstorms also caused false reports.

When a phone detects an earthquake signal, it sends word, along with a rough location, to a central server. And whereas a traditional earthquake warning system only needs four seismic stations to detect a quake, more than 100 phones need to sound the alarm before Google believes it. Once such a quake is detected, loud, full-screen alerts go out to phones located in regions that, based on traditional physics-based seismic projections, should receive shaking strong enough to be felt and break windows or dishes.

So far, Android phones have detected more than 1000 quakes worldwide. And the alerts have performed well in terms of speed and accuracy when compared with ShakeAlert and Japan’s warning system, Allen says. The phones also detected a magnitude 7 quake that originated just north of Samos, a Greek island close to Turkey, in late October 2020. “We could really show [that], if we had alerting in place, we would have provided alerts in Greece,” Allen says.

But the Android system still has a lot to prove, outside researchers say, including its ability to detect quakes that start in sparsely settled regions. New Zealand will pose a particular challenge, Meier says, because much of its population is clustered in a few cities. That means phones might not warn of a quake that starts far from downtown until after the trembling has already begun.

Many earthquakes in New Zealand also originate offshore where few phones are found, notes Caroline Francois-Holden, an independent seismologist who until recently worked at GNS Science, the country’s lead geological institute. That makes the phone system less than ideal for warning of lethal tsunamis. “Any earthquake early warning system needs to be designed with that in mind,” she says. Although the cost savings presented by the Google system sounds ideal, it needs to be thoroughly vetted, she says. “As a Kiwi scientist and ex–public servant, I would love to see quantitative measures of performance.”

Until the system’s performance is clear, Google should be wary of promising too much, she adds. New Zealand’s wealth, size, and high seismicity make it a great playground for testing such new technologies. But the scars of the Christchurch earthquake are still raw. “Everyone knows someone who lost a loved one,” she says. “Any promise to reduce the loss of life should be approached with much compassion.”