MENU

Machine Learning Technology From Google And Harvard Could Lead To More Accurate Earthquake Predictions

Close

Researchers have found a way to predict where aftershocks will strike after an intense earthquake through an artificial intelligence machine.

The study analyzed more than 131,000 mainshock and aftershock earthquakes including the 9.1-magnitude earthquake in Japan and other major earthquakes in history. By using data on past earthquakes, researchers harnessed a better predicting method which made use of a learning machine.

About The Study

The team trained a neural network to understand earthquakes more and identify a pattern. The machine could also immensely aid in finding out new methods to assess seismic risk. Through the study, the researchers surpassed the standard method of predicting the next aftershock's location.

According to Phoebe DeVries, a seismologist at Harvard University and a part of the team of researchers who presented their findings in the journal Nature on Aug. 29, they have just scratched the surface of what the device can contribute for aftershock forecasting.

"The application of machine learning to high-quality earthquake datasets is a big step beyond what has been done in the past," said Professor Mark Stirling, who serves as the chair of earthquake science at the University of Otago.

"With evolving methods like this, we stand to gain a better understanding of how this method can contribute to the ensemble of existing earthquake forecasting methods," Stirling added.

Aftershocks can be similarly damaging as the main earthquake. For example, New Zealand was hit by a magnitude 7.1 earthquake in 2010. The earthquake did not kill anyone; however, the 6.3-magnitude earthquake, which hit 5 months after the initial earthquake, killed 185 people.

Current Methods

Currently, seismologists can predict how intense the aftershock can be; however, they cannot figure out where it will happen. This is calculated by using a method that finds out how earthquakes change the stress in rocks.

Following this, seismologists assess if it would result in an aftershock. Although this method can explain the patterns of aftershocks for intense earthquakes, it is not foolproof.

The scientists tested their findings on 30,000 mainshock and aftershock events where the neural-network forecast managed to predict the locations of the aftershock more accurately than the past method.

Additionally, the network also found out the physical changes that happened in the ground after the first earthquake. Though the study is a fresh way to look at aftershocks, it would not be the final study about the mainshock and aftershock relationship.

ⓒ 2018 TECHTIMES.com All rights reserved. Do not reproduce without permission.
Real Time Analytics