A new effort in the field of earthquake prediction using artificial ielligence has raised hopes for the application of this technology in reducing the effects of earthquakes on human lives and economies. The artificial ielligence algorithm designed by researchers at the University of Texas at Austin, during a seven-moh experime in China, managed to predict 70% of earthquakes a week before they occurred.
Test results and global implications
This artificial ielligence has been trained to ideify statistical patterns in seismographic data instaly and in relation to previous earthquakes. The result of this process was a weekly forecast in which the AI was able to predict 14 earthquakes within approximately 200 miles of the predicted location with remarkable accuracy. The system missed one earthquake and issued eight false alarms.
It is not yet clear whether this method will be effective in other places as well, but this effort is considered a milestone in the research related to earthquake prediction using artificial ielligence.
Challenges and future directions
“Earthquake forecasting is like the Holy Grail,” says Sergey Fumel, a professor of economic geology at UT and a member of the research team. “Although we are not yet close to predicting earthquakes anywhere in the world, our findings show that what we previously thought of as an iractable problem is theoretically solvable,” he coinues.
The experime was part of an iernational competition in China where UT’s advanced artificial ielligence won first place out of 600 other designs. The project was led by Yangkan Chen, the bureau’s seismologist and lead AI developer. The results of this experime were published in the journal Bulletin of the Seismological Society of America.
Implications for preparation and further research
“You can’t predict earthquakes,” says Alexandros Savavidi, senior research scieist and program director of the Texas Seismological Network (TexNet). “It depends on the millisecond and the only thing you can corol is your preparation. “Even with 70% preparedness, significa results can be achieved that can help reduce economic and human losses and create the poteial for significa improvemes in earthquake preparedness globally.”
The researchers announced that they achieved success using a relatively simple method in machine learning. The AI was trained using statistical features designed based on the team’s knowledge of earthquake physics and then instructed to train on a five-year database of seismic recordings.
After passing the training course, the artificial ielligence provided its prediction by analyzing the signs of incoming earthquakes among the background sounds of the earth.
Office Manager Scott Tinker stated, “We are extremely proud of this team and their first place finish in this prestigious competition. Of course, importance is not only related to location and magnitude, but time also plays an importa role. “Earthquake prediction is an insurmouable challenge and we cannot overemphasize its difficulty.”
The researchers concluded that in areas with strong seismic tracking networks, such as California, Italy, Japan, Greece, Turkey, and Texas, the AI was able to increase its success rate and limit its predictions to within a few tens of miles.
One of the next stages of artificial ielligence testing will be done in the state of Texas, because this region sees the occurrence of small earthquakes and some moderate earthquakes with high frequency. With 300 seismic stations and more than six years of coinuous data, TexNet has become a good place to validate this method.
Finally, the researchers plan to combine this system with physics-based models. This can be especially importa in areas where data is limited, or in places like Cascadia, where the last major earthquake occurred hundreds of years ago and before seismographs were installed.
“Our future goal is to combine physics-based and data-based methods to provide a comprehensive chatGPT-like system that can be used anywhere in the world,” Chen said.
The new research is a fundameal step towards achieving this goal.
“It may be a long road, but many of these kinds of advances collectively drive science forward,” Tinker said.
