As is known, Japan is under threat of high degree seismic activity and tsunami due to its geography. A team of scientists has managed to map with high accuracy flood zones in tsunamis that can occur on the coast with an earthquake, using the world’s fastest supercomputer and an artificial intelligence model that calculates these risks.
Hardware called Fugaku plays a role in the development of this new predictive AI tool. Tsunami simulations were studied on the world’s most powerful supercomputer, developed last year by Riken and Fujitsu. 20,000 possible high-resolution tsunami scenarios have been created thanks to this processing power used by researchers from Fujitsu, Tohoku University, and the Institute for Earthquake Research at the University of Tokyo.
Using a deep learning algorithm, tsunami scenarios that hit the coast when an earthquake occurred were studied, and their accuracy was increased over time. The model created by AI is able to predict the tsunami in close to real time and with high resolution.
The model needs to be trained in Fugaku first, so that after loading it into normal computers, when fed with wave formation data, the prediction can be created in seconds. A simulation of a large earthquake and tsunami that would affect Tokyo Bay has been shown to match the Japanese Cabinet Office’s flood model.
In this way, with simulations to be run on normal computers, Disaster Response Teams can see how buildings, roads and other infrastructure will be affected by flooding in advance. This, in turn, facilitates planning that will save time in disasters.