Harvard, Google AI predicts earthquake aftershocks up to a year later

Google and researchers with Harvard University have teamed up to develop an artificial intelligence model capable of predicting aftershocks that may happen up to a year after the earthquake took place. The AI was trained using a database containing info on more than 118 big earthquakes, that helping researchers predict where an aftershock is likely to happen.

Aftershocks are instances of ground movements following an initial earthquake, which is typically the most destructive part of the disaster. Though aftershocks are expected following the main event, predicting when and where they may happen is difficult, often leaving people caught off guard.

Unfortunately, a surprise aftershock can result in additional injuries or disruption to ongoing recovery efforts, setting everyone back and putting more lives at risk. That makes a tool like the new AI model vital for communities where earthquakes are a frequent reality.

The machine learning technology, using data from past earthquakes and aftershocks, serves as something like a forecasting tool, helping experts predict where an aftershock is likely to take place. Such information could help first responders develop an informed game plan for recovery efforts that takes potential future shocks into consideration.

According to Harvard University post-doctoral fellow Phoebe DeVries writing on Google's Blog, "There was also an unintended consequence of the research: it helped us to identify physical quantities that may be important in earthquake generation." Ultimately, though, the AI is described as "still imprecise" at its current stage.

SOURCE: Google Blog, Nature