In 2017, Ross experienced a breakthrough that could reshape the field of seismology. Observing the prowess of machine learning programs in processing and categorizing vast sets of images with unmatched speed and accuracy, he wondered: could this technology be adapted to detect earthquakes?
Ross and his team began their innovative journey by collecting seismic waveforms from across Southern California, using data previously identified by scientists as genuine earthquakes. They created templates—essentially, snapshots of each earthquake’s unique seismic wave pattern. With these templates in hand, they developed an algorithm designed to comb through seismic records, searching for quakes that matched these specific patterns.
However, the initial program had its limitations. It was a precursor to true artificial intelligence, capable of identifying only those earthquakes it had been specifically trained to recognize. As a result, any novel seismic events went unnoticed, leaving a significant gap in earthquake detection.
Recognizing this drawback, Ross pivoted toward more advanced self-learning algorithms. These programs were designed to analyze existing data and predict what a broader range of earthquakes might sound like, rather than relying solely on pre-defined templates. This shift allowed the software to recognize unfamiliar seismic events that had previously eluded detection.
The results were impressive. The self-learning programs rapidly identified numerous unfamiliar quakes, which were subsequently verified by human scientists. “You just see so many things that were completely missed,” Ross remarked, highlighting the transformative potential of integrating machine learning into seismology.
Through this innovative approach, Ross and his colleagues have not only enhanced the accuracy of earthquake detection but also opened the door to understanding the complex dynamics of seismic activity in ways previously thought impossible. As technology continues to evolve, the combination of machine learning and traditional seismological methods may provide critical insights that improve earthquake prediction and preparedness efforts worldwide.
Reference: https://www.preventionweb.net/news/ai-helping-seismologists-find-next-monster-earthquake