Kyoto, Japan -- We all want to be warned about giant earthquakes as early as possible.
Now, a team from Kyoto University and Géoazur has developed a new approach, based on a deep learning AI for detecting prompt elasto-gravity signals, or PEGS. These are gravitational changes generated by large-mass motion in megaquakes and can be recorded by seismometers.
PEGS carry information about an ongoing earthquake at the speed of light, arriving much faster than even the fastest seismic waves.
"These signals can be used in real-time to track earthquake growth instantaneously after the event reaches magnitudes above 8," says author Bertrand Rouet-Leduc.
Until now, however, the minute amplitude of PEGS has prevented their use for earthquake and tsunami alert systems.
Current early warning systems based on seismic waves rely on estimating magnitude directly from the shaking. Although a magnitude 9 earthquake is thirty times more energetic than a magnitude 8, early warning systems tend to saturate and fail to estimate magnitude beyond 8. Even GPS-based methods are limited by large uncertainties, latency, and dependence on prior assumptions.
"When our team developed a deep-learning AI model making use of the information carried by PEGS, we were pleasantly surprised by our success with the initial prototype, which classifies earthquakes according to their magnitude," explains Rouet-Leduc.
The team was then able to demonstrate their deep-learning model's ability to instantaneously track an earthquake in real time after it reaches a certain size. The event was sourced from actual Japanese seismic data after training the AI to process simulated waveforms.
Although the algorithm still needs to be tested on live data, the scientists believe the results have the potential to improve earthquake and tsunami alert systems.
"Our new model could help alert communities critically faster of a seawall-breaching tsunami following a megaquake," the author concludes.