Researchers believe that artificial intelligence could help them predict earthquakes. Is it really possible? Currently, this research is being applied for crowdfunding in order to be continued.
A piece of Canada moved
As Wired explains in an article on September 22, 2019, the earth shook throughout the month of May, from Washington State (United States) to Vancouver Island (Canada). These places are located right in the subduction zone of Cascadia, the place where the Juan de Fuca plate sinks under the North American plate.
However, despite a magnitude of 6 on the Richter scale, it turns out that the inhabitants did not feel anything since it was about a slow earthquake. This is a discontinuous displacement comparable to an ordinary earthquake, however, releasing the elastic energy in hours (or days) instead of usually minutes. Despite the fact that locals did not notice anything, the southern tip of Vancouver Island still moved a centimeter!
Geophysicist Paul Johnson and his team at Los Alamos National Laboratory (USA) have used these slow-moving earthquakes to develop a new technology based on artificial intelligence. The goal? Predict earthquakes and reduce the risk of loss of life.
Waiting for funding
Paul Johnson and his team were interested in the slow earthquakes that would occur once a year in the Vancouver area. As part of their research, scientists have established no less than 80 criteria for analysis to find a common point to earthquakes. According to the researchers, the sound would be the key element, announcing the first shocks. Their research shows that seismic energy continues in the subduction zone following specific patterns leading to failure. Thus, artificial intelligence has been trained with audio recordings to distinguish warning signs of earthquakes.
Published in the arXiv platform on September 15, 2019, the study needs to be continued. The researchers are asking for help to finance the development of a complete system capable of predicting earthquakes of a certain magnitude. In sum, according to the team, the next step is to establish a link between slow earthquakes considered harmless and ordinary earthquakes that can be very deadly. We can imagine that predicting earthquakes in advance would, for example, calmly organize the evacuation of the population.
Finally, let’s mention the fact that this is not the first time that I.A is used for earthquakes. In 2018, MIT researchers developed a neural network designed to detect and locate earthquakes using seismograms. However, this system has detected 17 times more earthquakes than those recorded by traditional methods!