It is estimated that for a tsunami to be generated by an earthquake, it must be of a strong magnitude, at least 6.5 on the Richter scale. A team of researchers from IRD, CNRS, Université Côte d’Azur, Observatoire de la Côte d’Azur, Los Alamos National Laboratory and Kyoto University have used artificial intelligence to instantly estimate the magnitude of large earthquakes from “Prompt Elasto-Gravity Signals” (PEGS). Their study, titled ” Instantaneous tracking of earthquake growth with elasto-gravity signals,” was published in the journal Nature on May 11.

Fortunately, tsunamis are rare natural disasters but they can kill many people living in coastal areas or island states. For example, the tsunami that occurred on December 26, 2004 in the Indian Ocean, following an earthquake of magnitude 9.1, affected 14 countries, including India, Indonesia, Sri Lanka and Thailand, leaving about 230,000 people dead or missing.

Warning systems based on seismic waves have been put in place to limit the human and material losses of these disasters. The populations are warned only a few seconds before the tremors, but as the tsunamis move more slowly, this leaves them a few tens of minutes to take refuge in a safe place.

However, these warning systems are not able to estimate quickly the magnitude of very large earthquakes. For example, the Japanese warning system estimated a magnitude of 8 instead of 9 during the 2011 Tōhoku Pacific coast earthquake (commonly referred to as the Fukushima earthquake), predicting a 3-meter wave instead of 15, an error with dramatic consequences (18,079 dead or missing). According to the research team, geodesy-based approaches provide better estimates, but are also subject to large uncertainties and latency associated with the slow speed of seismic waves.

The study

The team demonstrated that it is possible to exploit the gravitational signals (PEGS), which, although very weak, were discovered in 2017 in the 2011 earthquake data, to instantaneously estimate the magnitude of large earthquakes.

PEGS (Prompt Elasto-Gravity Signals) are gravitational waves generated by the motion of a huge mass of rock during large earthquakes. These signals propagate at the speed of light, much faster than seismic waves, and can thus provide additional time for populations to protect themselves.

The very low amplitude of PEGS made it impossible to use them in warning systems. The researchers got around this problem thanks to an AI algorithm. They developed PESGNet, a deep learning model, the CNNs, which exploits the information provided by PEGS recorded by regional broadband seismometers in Japan before the arrival of seismic waves. After training on a database of 500,000 synthetic waveforms augmented with empirical noise, the algorithm can instantly track a time function of the source of an earthquake on real data. According to the researchers, “ Our model unlocks ‘real-time’ access to the rupture evolution of large earthquakes by using a portion of the seismograms that is systematically treated as noise and can be immediately transformative for tsunami early warning.”

Andrea Licciardi, geophysicist at GEOAZUR and first author of the study, states:

“ Tested in Japan, the algorithm is shown to be able to estimate the magnitude of the Fukushima earthquake faster and more reliably than any existing system, without using seismic waves. “

Quentin Bletery, who initiated the project, and in the same unit, adds:

“ Implementation in operational warning systems remains to be done, but our results indicate that PEGS could significantly improve tsunami warning systems. “

Article sources Licciardi, A., Bletery, Q., Rouet-Leduc, B. et al. Instantaneous tracking of earthquake growth with elastogravity signals. Nature (2022). https://doi.org/10.1038/s41586-022-04672-7.

Translated from Estimer en temps réel la magnitude des grands séismes grâce aux ondes gravitationnelles