Machine learning could help predict earthquakes

New machine learning technology could help researchers detect the low rumblings that happen before earthquakes occur.
By James Carlin | Oct 25, 2017
New technology could allow researchers to detect earthquakes days before they happen, a new study in Geophysical Research Letters reports.

Certain rocks, when put under increased pressure before an earthquake, send out low-pitched rumbling sounds. While humans are not able to hear such noises, they can be detected by machines.

To understand this phenomenon, a group of researchers from Cambridge University recreated powerful earthquake-like forces in a laboratory setting and then used high-tech algorithms to pick out specific cues that could signal a pending quake.

Those sounds typically occur a week before an earthquake happens. As a result, deciphering the rumbles could give scientists a way to both predict and pinpoint the timing of a future tremor.

Currently, researchers cancalculate the probability of an earthquake in a particular area. However, they have no way of knowing when such an event will occur. This new technology could be a big step towards that problem, and could help save a number of lives.

"People have said you can't predict earthquakes," said study co-author Colin Humphreys, a professor of materials science at Cambridge University, according to Reuters. "People have tried. We're now saying we believe for the first time we can predict an earthquake in a laboratory."

The team simulated multiple earthquakes by using a combination of steel blocks, pistons, and machine-learning algorithms. All of the tests were conducted at New Mexico's Los Alamos National Laboratory.

The different trials showed that the machine-learning systems are able to identify the low rumblings that get stronger as tremors approach. While the shaking has been detected in past studies, almost all of it was rejected as random noise.

Now that scientists know they can properly detect the shaking, they next plan to apply the new technology to quake-prone areas. However, there is still a long way to go. Even if their method is successful, the team will need to next figure out a way to determine the coming quake's magnitude.

Even so, there is no doubt the system has potential. Earthquakes claim lives all over the world, and being able to predict them could greatly decrease the damage they cause.

"We're at a point where huge advances in instrumentation, machine learning, faster computers and our ability to handle massive data sets could bring about huge advances in earthquake science," said lead authorBertrand Rouet-Leduc, who was a PhD student at Cambridge during the research, according to


Have something to say? Let us know in the comments section or send an email to the author. You can share ideas for stories by contacting us here.

Comments should take into account that readers may hold different opinions. With that in mind, please make sure comments are respectful, insightful, and remain focused on the article topic.