Last year in May, after a 13-month slumber, the ground beneath Washington’s Puget Sound rumbled to life. The quake started more than 20 miles under the Olympic mountains and, over a few weeks, drifted northwest, reaching Canada’s Vancouver Island. It then briefly reversed course, migrating back throughout the US border before going silent again. All informed, the monthlong earthquake probably released sufficient energy to register as a magnitude 6. By the time it was accomplished, the southern tip of Vancouver Island had been pushed a centimeter or closer to the Pacific Ocean.
Because the quake was so scattered in time and space, however, it’s that nobody felt it. These sorts of phantom earthquakes, which happen deeper underground than standard, fast earthquakes, are generally known as “slow slips.” They occur about once a year in the Pacific Northwest, along a stretch of the mistake where the Juan de Fuca is slowly forcing itself underneath the North American plate. Greater than a dozen slow slips have been detected by the region’s sprawling network of seismic stations since 2003. And for the previous year and a half, these occasions have been the main focus of a unique effort at earthquake prediction by the Paul Johnson (geophysicist).
Johnson’s team is amongst a handful of groups which are utilizing machine learning to attempt to explain earthquake physics and tease out the warning indicators of approaching quakes. About two years ago, using pattern-finding algorithms similar to these following recent advances in image and speech recognition and other types of artificial intelligence, he and his collaborators successfully predicted temblors in a model laboratory system a feat that has since been duplicated by researchers in Europe.