Google, Harvard to use machine learning to predict earthquake aftershock locations
Wionews: Researchers from Google’s artificial intelligence department and Harvard University have come up with a model which will be capable of predicting earthquake aftershocks, the size and timing of the aftershocks and also the exact location using deep learning. The researchers published a paper earlier this week to exhibit how this dee leaning model would be able to help predict aftershock locations more accurately.
Scientists trained a neural network to look for patterns in a database of more than 131,000 “mainshock-aftershock” events after which it was tested on 30,000 data points and its accuracy was measured against a model known a “Coulomb failure stress change.”
Coulomb failure stress change was able to predict aftershocks with a 58.3% accuracy, while the deep neural network the researchers developed achieved 84.9% accuracy.
Aftershocks included in the dataset used to train the neural network took place in a perimeter that stretches 50 kilometers vertically and 100 kilometers horizontally from each earthquake epicenter.
Although the timing and size of aftershocks has been understood and explained by established empirical laws, forecasting the locations of these events has proven more challenging, Phoebe DeVries, Post-Doctoral Fellow at Harvard, said in a Google blogpost. “We teamed up with machine learning experts at Google to see if we could apply deep learning to explain where aftershocks might occur… We are looking forward to seeing what machine learning can do in the future to unravel the mysteries behind earthquakes, in an effort to mitigate their harmful effects,” PTI quoted her as saying. DeVries outlined that the teams started with a database of information on more than 118 major earthquakes from around the world, including the one in Bhuj in 2001.
Data used to train the model came from noteworthy earthquakes over the past few decades, such as the 2004 Sumatra earthquake, the 2011 earthquake in Japan, the 1989 Loma Prieta earthquake in the San Francisco Bay Area, and the 1994 Northridge earthquake near Los Angeles.
The “neural net” was then applied to analyse the relationships between static stress changes caused by the mainshocks and aftershock locations, and the algorithm was able to identify useful patterns.
Machine learning-based forecasts may, one day, help deploy emergency services and inform evacuation plans for areas at risk of an aftershock, she added.
“The end result was an improved model to forecast aftershock locations and while this system is still imprecise, it’s a motivating step forward,” she said.
DeVries said the process also helped identify physical quantities that may be important in earthquake generation, which opens up new possibilities for finding potential physical theories that may allow for better understanding this natural phenomenon.