Earthquake-detecting AI algorithm finds success in early trial

Using real-time seismic data paired with previous earthquakes, the release says researchers trained the AI algorithm to identify statistical bumps which it used over a seven-month trial period in China to forecast 70% of earthquakes a week before they happened.  

AUSTIN (KXAN) — Researchers at the University of Texas at Austin have developed an artificial intelligence algorithm which aims to one day predict earthquakes before they happen, according to a press release by UT’s Jackson School of Geosciences.  

Using real-time seismic data paired with previous earthquakes, the release said researchers trained the AI algorithm to identify statistical bumps which it used over a seven-month trial period in China to forecast 70% of earthquakes a week before they happened.  

The international competition held in China pitted the UT-developed AI against 600 other designs. It came in first place by successfully predicting 14 earthquakes. The predictions estimated where the earthquakes would occur within approximately 200 miles and almost exactly calculated their strength, with only one missed earthquake and eight false warnings, according to researchers.  

Although it is unknown if the technology will work elsewhere, the release says researchers hope it will eventually minimize the impact of earthquakes on individuals and economies.  

“We’re not yet close to making predictions for anywhere in the world but what we achieved tells us that what we thought was an impossible problem is solvable in principle,” Professor Sergey Fomel with UT’s Bureau of Economic Geology said.  

Researchers believe the AI may improve the accuracy of its results in locations with more expansive seismic tracking. The UT researchers said one of the next steps will be testing it in Texas due to the high rate of seismic activity here.  

The research was supported by TexNet, the Texas Consortium of Computational Seismology and Zhejiang University and, according to the release, the trial’s findings can be found in the journal Bulletin of the Seismological Society of America.