Researchers from Google’s AI division and Harvard University – using deep learning algorithms, have developed a system that, while still imprecise, is able to forecast earthquake aftershocks significantly better than using random assignment, reports The Verge.
After an earthquake hits, what often follows are a series of follow up shocks, triggered by the larger initial shock. These aftershocks can last from a few days to as long as several months, further weakening or tumbling structures.
Scientists have been able to predict the size and timing of aftershocks using empirical laws, like Bäth’s Law and Ohmori’s Law, to describe the likely size and timing of those aftershocks, but being able to pinpoint the location where an aftershock might hit has been a problem.
In a paper published in the journal Nature Wednesday, researchers show how deep learning can help predict aftershock locations more reliably than existing models.
Sparked by a suggestion from researchers at Google, Brendan Meade, a Professor of Earth and Planetary Sciences, and Phoebe DeVries, a post-doctoral fellow working in his lab, using deep learning algorithms, created an AI model capable of predicting the location of aftershocks up to one year after a major earthquake.
The researcher’s developed a neural network that trained on more than 131,000 mainshock–aftershock pairs from around the world in an attempt to predict where aftershocks would occur. They then tested the model on a database of 30,000 similar pairs.
The research showed the deep-learning network was significantly more reliable than the existing model being used now, called the “Coulomb failure stress change.”
On a scale of accuracy with a very tight margin running from 0 – 1, with 1 being perfectly accurate and 0.5 as good as flipping a coin, the existing Coulomb model scored 0.583, while the new AI system hit 0.849.
“We found that after feeding these model stress changes into the neural network, the neural network could sort of predict aftershock locations in the testing dataset more accurately than the sort of baseline Coulomb failure stress change criterion that’s used a lot in studies of aftershock locations,” Phoebe DeVries of the Department of Earth and Planetary Sciences at Harvard University told VentureBeat in a phone interview.
And because the neural network was trained on many different types of earthquakes and aftershocks from around the world, Meade said the resulting system worked for many different types of faults.
“Faults in different parts of the world have different geometry,” Meade said. “In California, most are slip-faults, but in other places, like Japan, they have very shallow subduction zones. But what’s cool about this system is you can train it on one, and it will predict on the other, so it’s really generalizable.”
