Cheap Earthquake Warning Systems

Crowdsourced networks of low-cost sensors and cellphones could have provided life-saving seconds in Nepal


While predicting earthquakes remains a dream, scientists have developed early-warning systems that give people precious seconds to run out of buildings or take cover. Such systems are in place in Japan and Mexico. The U.S. Geological Survey (USGS) is testing a system that gave a 5- to 10-second warning when a temblor hit California’s Napa Valley in 2014. That kind of warning might have saved hundreds of lives when a 7.8-magnitude earthquake devastated Nepal on 25 April.

Earthquake-warning systems come at a high price, though, too high for countries like Nepal and others in quake-prone zones in South Asia, the Caribbean, and Central and South America. But researchers are now working on more affordable, crowdsourced warning systems based on low-cost sensors and cellphone electronics.

Today’s alert systems deploy networks of hundreds of expensive, extremely sensitive seismic sensors that detect energy waves, along with GPS sensors to detect permanent ground movement due to the motion of the geological fault that triggers the earthquake.

Battalgazi Yildirim, founder of Zizmos, based in Palo Alto, Calif., thinks he can get as good—or in some cases even better—earthquake data from a network of cheap sensor packages. These are made up of microelectromechanical accelerometers attached to inexpensive, off-the-shelf cellphone equipment that manages data gathering and communication. “The sensor packages used by the USGS cost about [US] $30,000 each,” he says. “We’re putting out sensors that cost $100 each to build.”

The network should have virtually no installation or maintenance costs, because Yildirim plans to rely on the kindness of the crowd. Zizmos asks for volunteers to donate a tiny bit of interior wall space and a power outlet to host a sensor package, which is about the size of a deck of cards.

With a $150,000 research grant from the National Science Foundation, the company was to launch a trial this May, after this issue went to press, involving 268 sensors distributed in California. When a sensor detects a rumble, it will send information about the time of the event and the magnitude of the shaking to a cloud-based server; algorithms will check reports from neighboring sensors to determine whether the vibration was local—say, from a truck going by—or felt elsewhere. If the latter, the system will calculate a hypothetical epicenter for the earthquake, the original time of the event, and an estimated magnitude. For earthquakes with magnitudes greater than 4.0, it will issue an alert.

Benjamin Brooks, a geophysicist with the USGS, has a different crowdsourcing approach in mind. Why not tap into the GPS sensors in people’s phones and navigation systems, he says. GPS-equipped cellphones are ubiquitous in developing countries, and such a crowdsourced system would offer early warning at practically no cost. “A country like Nepal, with high earthquake hazards and minimal resources, is where a crowdsourcing approach would be most effective,” Brooks says.

In research presented in the journal Science Advances two weeks before the Nepal quake hit, Brooks and his colleagues tested the ability of consumer-grade GPS devices to detect earthquakes. They subjected a Google Nexus smartphone and a commercial GPS module to displacements ranging from 10 centimeters to 2 meters. Both GPS sensors picked up the smallest motion.

Next, the researchers performed simulations using data from a hypothetical magnitude-7 earthquake in northern California and from the real 2011 magnitude-9 earthquake that hit Tohoku-oki, Japan. They simulated smartphone responses based on census data around the earthquake epicenters and recorded a phone as triggered if it and its four nearest neighbors measured more than 5 cm of movement. If at least 100 phones were triggered, the system declared an earthquake.

It took fewer than 5,000 smartphones to detect the simulated California earthquake within 5 seconds, giving enough time to warn San Francisco and San Jose. For the Japan quake, which had an offshore epicenter, detection occurred at just over 80 seconds, too slow for the closest onshore towns but in enough time to issue a warning to Tokyo.

“Sadly, on the Wednesday before the Nepal earthquake I had a discussion with a Nepalese colleague about proposing precisely such an approach,” Brooks says. “There would be challenges in terms of cellphone service in such a mountainous region, but we think it would be doable there eventually.”