An algorithm to better anticipate tsunamis

Paris (AFP) – Tsunamis would be better anticipated if we were able to assess in real time the power of the earthquakes that cause them: an artificial intelligence algorithm, described in a study on Wednesday, could meet this challenge and help limit the damage.

On March 11, 2011, one of the world’s strongest earthquakes ever recorded shook the depths of the Pacific Ocean off the northeast coast of Japan, causing a tidal wave that killed nearly 18,500 dead and led to the Fukushima nuclear disaster.

The magnitude of the quake was 9 on the Richter scale, but at the time of the tremor, warning systems had only rated it at 8.1. A big underestimate, but almost unavoidable when it comes to mega-earthquakes.

“Current instruments for measuring seismic waves are limited,” emphasizes Quentin Bletery, co-author of the study published in the journal Nature. They are too slow, and beyond a certain threshold – a magnitude of 8 – they tend to “saturate and make a prediction at maximum 8, whatever happens”, explains to AFP this geophysicist at Research Institute for Development (IRD).

In 2017, signals faster than seismic waves were discovered: PEGS (“Prompt Elastro-Gravity Signals”), which are disturbances of the Earth’s gravity field caused by earthquakes. And move at the speed of light.

350,000 virtual earthquakes

But these signals are too tenuous to be exploited directly. A team of IRD researchers based at the Université Côte d’Azur therefore decided to develop a deep learning model based on the modeling of these “PEGS”.

The artificial intelligence (AI) algorithm was trained from 350,000 virtual earthquake scenarios along major fault lines in Japan, calculating all predicted gravitational signals. “For each virtual earthquake, we trained the AI, giving it the response each time, to find the magnitude and location from the expected PEGS”, develops Quentin Bletery.

The model was then tested on real data, those of the March 2011 earthquake. Result: it predicted the correct location and the correct magnitude – to within 0.3 points – after 50 seconds. And gave the perfectly exact magnitude of 9 after just two minutes.

“In 2011, it took several hours to assess it exactly, well after the arrival of the tsunami. However, with an estimate of 8.1, we expected waves of a maximum of three meters”, recalls the researcher.

A size that would not have been a problem for the dikes erected on the Japanese coast. “But an earthquake at 9 causes waves of more than 15 meters… If we had known it before, we would have evacuated the populations even more”, he underlines.

The algorithm is currently in the test phase in Peru, and is intended to be integrated into the warning systems of regions of the world threatened by tsunamis.


We would love to say thanks to the writer of this post for this outstanding content

An algorithm to better anticipate tsunamis

Check out our social media accounts as well as other pages related to it.