Khaberni - A small earthquake shook the city of Califatria in California, USA, in the early hours of January 1, 2008.
Its magnitude was only -0.53 on the Richter scale, weaker than the vibration caused by a passing truck, and it went unfelt by anyone. Despite its small size, scientists accurately detected it, thanks to artificial intelligence.
Artificial intelligence has completely changed the field of seismology, as it now can automatically and quickly detect very small earthquakes, a task that previously required human analysts and specialized equipment, according to a report published on "arstechnica".
Kyle Bradley, one of the authors of the Earthquake Insights newsletter, said: "Adopting these technologies is like wearing glasses for the first time...suddenly you see small details you couldn't see before."
From wave analysis to deep learning
Traditionally, earthquake monitoring depended on human analysts who read data from seismometers. Then came traditional algorithms that allowed partial automation of this task.
But the problem always was with small earthquakes that get lost among the daily noise in cities.
Here came the role of artificial intelligence, especially a model called Earthquake Transformer, developed by researchers from Stanford University in 2020.
This model relies on the same ideas used in image or sound recognition, but instead of analyzing pixels, it analyzes ground vibrations over time.
This model can differentiate the various types of seismic waves (P and S), and pinpoint the exact moment of their initiation with extreme accuracy.
Thanks to these models, monitoring devices are now able to detect a much larger number of small earthquakes, many times more than what was previously detected.
Joe Burns, a professor of geology at the University of Texas, said: "It has been a real revolution... but it is a revolution that is still ongoing."
This technology has contributed to creating accurate pictures of the earth's interior and volcanoes. In 2022, researchers from Hawaii used artificial intelligence data to confirm the existence of a hidden magma channel linking Mauna Loa volcano with the deep Pahala area, a discovery that had long been a subject of debate among scientists.
Earthquake prediction
Despite the huge success in monitoring and analysis, predicting earthquakes before they occur remains elusive.
Artificial intelligence is currently used to improve understanding of patterns of seismic activity, but it has not yet reached the stage of being able to determine the timing of the next earthquake accurately.
Scientists believe that this step is just the beginning, as deep learning technologies have changed the way images and speech are understood, they may soon open the door to a new generation of early warning systems for earthquakes.
