On January 1, 2008, at 1:59 a.m., an earthquake occurred in Calipatria, California. You haven't heard about this earthquake; even if you lived in Kalipatria, you would not feel anything. Magnitude -0.53, about the same shaking as a passing truck. However, this earthquake is notable not because it was large, but because it was small, and yet we know about it.
Over the past seven years, artificial intelligence tools based on computer images have almost completely automated one of the fundamental tasks of seismology: detecting earthquakes. What was once a task for human analysts and later simpler computer programs can now be done automatically and quickly using machine learning tools.
These machine learning tools can detect smaller earthquakes than human analysts, especially in noisy places like cities. Earthquakes provide valuable information about the composition of the Earth and what hazards may arise in the future.
“At best, when you apply these new methods, even with the same old data, it's like putting on glasses for the first time and being able to see leaves on trees,” said Kyle Bradley, co-author of the study. Earthquake Information newsletter.
I spoke with several earthquake scientists, and they all agreed that machine learning techniques were better at replacing humans in these specific tasks.
“It’s really remarkable,” Judith Hubbard, a Cornell professor and Bradley’s co-author, told me.
What's less certain is what comes next. Earthquake detection is a fundamental part of seismology, but there are many other data processing challenges that remain to be solved. The biggest potential impacts, even earthquake predictions, have yet to materialize.
“It really was a revolution,” said Joe Byrnes, a professor at the University of Texas at Dallas. “But the revolution continues.”
When an earthquake occurs in one place, shaking travels through the ground, much like sound waves travel through air. In both cases, inferences can be made about the materials through which the waves pass.
