Japan’s Fujitsu Labs used AI on the Fugaku supercomputer to develop algorithms that predict tsunami flooding.

A tsunami warning system using dart buoys was badly off the mark when a tsunami struck Japan’s eastern coast a decade ago. A deadly 9.0 magnitude earthquake triggered the tsunami wave, which was as high as 50 meters in some places. Tragically, the buoy system estimated the wave size at only 3 meters. When it struck, the monster wave caused over 20,000 lives. However, Fujitsu (TYO: 6702) Labs used AI and a supercomputer to develop a more reliable system. (The Hindu)

Using AI against tsunamis

Fujitsu’s AI model predicts tsunami-related flooding in coastal areas in real-time. To train the AI model, the scientists created data sets for over 20,000 possible tsunami scenarios. They created these high-resolution simulations on the Fugaku supercomputer, pictured below.

Researchers on coasts can be on high alert in the event of an earthquake. They input offshore waveform data into the model.

The model then predicts the magnitude of the likely flooding and offers insights into the impact on infrastructures such as buildings and roads. With prior warning, relief agencies can also take pre-emptive steps for evacuation if so required.

According to Fujitsu Labs, it is also possible to run the model on ordinary PCs. This makes the model portable, simple and robust for on-spot coastal applications.

Fugaku

In June 2020, the TOP500 list of supercomputers named Fugaku the world’s fastest. The supercomputer is highly energy-efficient and consumes 30 MW at full throttle. It runs on the Fujitsu A64FX microprocessors. These deliver high performance, high throughput, and high performance per watt.

Over 150,000 of these high-performance A64FX CPUs are connected with the Tofu D Interconnect high-speed network. These enable peak performance of 537 PetaFLOPS, i.e. more than 5 * 10 to the power of 17 Floating-Point Operations per second.

Related Story:    AI Can Now Predict Lightning Strikes