Machine learning finds quake origin signatures

Combing through historical seismic data with a machine learning model, US researchers have unearthed distinct statistical features that marked the formative stage of slow-slip ruptures in the Earth’s crust months before tremor or GPS data detected a slip in the tectonic plates.

Given the similarity between slow-slip events and classic earthquakes, they suggest, in a paper in the journal Nature Communications, that these distinct signatures may help geophysicists understand the timing of the faster quakes as well.

“The… model found that, close to the end of the slow slip cycle, a snapshot of the data is imprinted with fundamental information regarding the upcoming failure of the system,” says lead author Claudia Hulbert, from the Los Alamos National Laboratory.

“Our results suggest that slow-slip rupture may well be predictable, and because slow slip events have a lot in common with earthquakes, slow-slip events may provide an easier way to study the fundamental physics of earth rupture.”

Slow-slip events gently rattle the ground for days, months, or even years, do not radiate large-amplitude seismic waves, and often go unnoticed by the average person – and they are easier for data-hungry machine learning techniques to study than classic quakes that rupture the ground in minutes.

Hulbert and colleagues looked at continuous seismic waves covering the period 2009 to 2018 from the Pacific Northwest Seismic Network, which tracks earth movements in the Cascadia region of the western US.

In this subduction zone, during a slow slip event, the North American plate lurches south-westerly over the Juan de Fuca plate approximately every 14 months. The data set lent itself well to the supervised-machine learning approach developed in laboratory earthquake experiments by the Los Alamos team collaborators and used for this study.

“For most events, we can see the signatures of impending rupture from weeks to months before the rupture,” Hulbert says.

“They are similar enough from one event cycle to the next so that a model trained on past data can recognize the signatures in data from several years later. But it’s still an open question whether this holds over long periods of time.”

Please login to favourite this article.