A neon blue curve is labeled “Tohoku” on the screen. Here, in the simplest geophysical essence, the 2011 earthquake in Japan, which was responsible for the Fukushima disaster, is represented. Worldwide, earthquakes have caused nearly one million deaths over the past three decades, particularly in Indonesia and Haiti. At Sophia Antipolis in the hinterland of Nice, Quentin Bletery and his team use artificial intelligence to try to predict and prevent these disasters.
“Earthquake forecasting is the holy grail of seismology”, says the geophysicist at the Research and Development Institute (IRD) in the introduction. The research fellow received a European grant from the ERC program (European Research Council) and has an amount of 1.5 million euros over five years. “Detection of early seismic signals thanks to artificial intelligence”. Study published mid-May in the journal Nature, based on a recent discovery: gravitational waves associated with an earthquake, in this case Fukushima. Gravitational waves correspond to “folds” in space-time caused by the displacement of large masses, much like a bowling ball stabbing into a stretched net.
Gravitational waves, previously detected during massive collisions such as two black holes swallowing each other in space, are also born during major earthquakes on Earth, when large masses are in motion. “The idea came during a seminar at Caltech in the United States where a person in charge of a space gravity wave detector was investigating how to precisely control seismic noise to prevent the instrument from failing.”He recalls Jean-Paul Ampuero, director of research at IRD and who discovered the signal.
It’s too early but the signals are hard to detect
It turns out that the gravitational signals from an earthquake known as Pegs (“Prompt Elastogravity Signals”) are particularly discreet, but detectable by quality seismometers. “All earthquakes produce Pegs, but these signals are only observable above magnitude 7.5 or 8.describes geophysicist Kevin Juhel. In total, there are less than ten events for which we were able to record them. “The scientific community was quite skeptical because the signal was very weak compared to seismic noise in a very loud environment”, Knows Jean-Paul Ampuero.
These gravitational waves travel at the speed of light as soon as the fault breaks, where P waves, the first seismic waves to appear, “only” propagate at about six kilometers per second. During a major earthquake, a seismometer therefore records a Peg before receiving the P waves. “Using gravitational waves saves warning systems about ten secondsSays Gabriela Arias, a doctoral student on Quentin Bletery’s team. It doesn’t seem like much, but when combined with automated systems, it could shut down a power station or an airport. » Or simply stopping a surgical procedure in hospitals that could cause the shock to slip.
Saturation seismometers for large earthquakes
The project is even more beneficial for tsunamis. In Japan, in 2011, a warning was issued by authorities three minutes after the earthquake. It took nine minutes for the American-run Pacific Warning Center to respond. This new approach could accelerate the triggering of sirens even in developing countries without infrastructure. Because gravitational waves are not only instantaneous but also felt all over the world. Countries well equipped with seismometers can thus act as sentinels for others.
Typically, the occurrence of a tsunami after an earthquake depends on two factors: the magnitude and the focal mechanism, i.e. the way the rupture occurs. At sea, subduction mechanisms, in which one plate passes under the other, cause tsunamis as the seafloor rises sharply.
But “Most systems are saturated and very poor at determining the true magnitude of an earthquake and therefore the magnitude of the associated tsunami”, explains Gabriela Arias. The wave predicted in Japan was much smaller than the one that finally broke. “Systems expected a tsunami corresponding to a magnitude 8 earthquake, not a magnitude 9”, Explains Quentin Bletery. Backwards, “The latches are not getting enough”, Gabriela Arias continues. That’s why the team developed an AI-powered algorithm that uses these early signals to instantly determine the size and location of the tear.
Early but not early warnings
“Fortunately, there aren’t too many big earthquakes. So we simulated over half a million times to have enough data to train a neural network.says Quentin Bletery. AI can then give the magnitude of any earthquake by observing the propagating gravitational waves. »“We’ve tested the algorithm against previous earthquakes, it works.Welcome to Andrea Licciardi, who developed the computer program within the team. Now, we need to try this in real time and in context. That’s how I imagine science: useful for societies. »
That’s not enough to read the future. No matter how momentary, gravitational waves appear only when there is a rupture, not before. Quentin Bletery, the second part of the research grant, is therefore cautious about estimation. “At this time we don’t know if there is anything, a signal or something else that could indicate a future earthquake.accepts. If you know what you are looking for, artificial intelligence can help. »
Recent major earthquakes
• August 2021. A 7.2 magnitude earthquake struck Haiti and killed 2,000 people. The disaster is reminiscent of the disaster of 2010, when more than 200,000 people died.
• March 2011. The 9.1-magnitude Tohoku earthquake causes a massive tsunami that ravages the coast of Japan and affects the Fukushima nuclear power plant.
• December 2004. A 9.1 magnitude earthquake and ensuing tsunami off Sumatra killed more than 200,000 people. The devastating wave reaches as far as Africa and Australia.