Predicting earthquakes a week in advance

The research results were published in the peer-reviewed scientific journal Bulletin of the Seismological Society of America.

It was stated that the researchers who developed the algorithm conducted trials in China for 7 months.

In its weekly forecasts, the artificial intelligence algorithm reportedly correctly predicted the location and magnitude of 14 earthquakes within a radius of up to 320 kilometers.

However, the algorithm failed to predict one earthquake and issued 8 false alarms.

"COULD BE A TURNING POINT"

It was stated that the algorithm, which works on the principle of machine learning, was loaded with statistical data based on the researchers' knowledge of seismology and was then instructed to train itself on 5 years of seismic records.

Although the new approach has not yet been tested in other locations, it is noted that the research could be a turning point for AI-based earthquake prediction models.

Sergey Fomel, a member of the research team, said, "Predicting earthquakes is a major goal. We are still not close to making predictions for anywhere in the world. However, the results we obtained show that a problem we once considered impossible is, in principle, solvable."

Alexandros Savvaidis, Director of the Texas Seismological Network Program (TexNet) and one of the participants in the research, spoke as follows:

Even 70 percent is a very significant result. It could reduce loss of life and economic losses. It has the potential to offer a major advancement in earthquake preparedness worldwide.

In the next step, the research team wants to test the algorithm statewide using data from TexNet, which records seismic activity in Texas with more than 300 observation stations.

The researchers aim to improve the algorithm to perform earthquake predictions worldwide by integrating physics-based models that are not specific to geographic regions into their system.