When preparing for a natural disaster, every second counts. Any tip off to an impending disaster gives residents a second longer to prepare and brace themselves for what’s coming. Some natural disasters, like earthquakes and tsunamis, actually have warning systems built in: localized changes in geomagnetic fields. Now, scientists at Tokyo Metropolitan University are developing a deep neural net to understand changes to magnetic fields as they happen.
An earthquake’s magnetic field is created by phenomena called the piezo-magnetic effect, or piezomagnetism. Earthquakes relieve stresses built up between tectonic plates, and as they do so, the symmetry of elements within the structure of the Earth can become out of sync. When they become asymmetrical, the geomagnetic field around the area shifts.
With tsunamis, their rapid movement across the ocean causes changes in atmospheric pressure. When the atmospheric pressure starts to alter ionosphere, the geomagnetic field alters along with it.
These are two very different processes, but they have the same result. Either can be detected by observation, and they register faster than virtually anything else on the planet—electromagnetic waves move at the speed of light. It’s possible to have a near-instant understanding of any shift to the Earth’s geomagnetic field anywhere across the globe.
However, the Earth is a living planet. That means all sorts of tiny shifts and alterations through the day, leaving scientists with the classic problem of trying to find signal through wide spreads of noise.
Taking measurements around Japan, Professor Yuta Katori and Associate Professor Kan Okubo created estimates of what the normal geomagnetic field looks like in the area. To give an idea of how much geomagnetic flux there is in the world: Just in their Japanese observation sites, just in 2015, Katori and Okubo were able to accumulate 500,000 data points.
They then began feeding these data points into a neural net. With their data points, they can now estimate what a normal magnetic field looks like within the area, thus offering a better understanding of what an abnormal reading due to an earthquake or tsunami would look like.
The hope is to pair up the neural network with high sensitivity detectors, allowing for a lightening-fast emergency warning. Given how the recent tsunami in Indonesia was compounded by faulty emergency warning systems, the sooner the neural net is up and running the better.
Source: Tokyo Metropolitan University
Originally published on Popular Mechanics