Fujitsu Laboratories has announced the development of an AI model, with Fugaku, the world’s fastest supercomputer, that will allow disaster management teams to predict tsunami flooding in coastal areas, in near real time.
The announcement is made in conjunction with Tohoku University’s International Disaster Science Research Institute and the University of Tokyo Earthquake Research Institute. In the event of an actual earthquake, inputting data from tsunami waves observed offshore from this AI model will allow flood prediction.
This will make it easier to obtain, more accurately and quickly, detailed data on the forecast of floods in specific areas, offering a critical view of the effects of localized waves on surrounding infrastructure, such as buildings and roads in urban coastal areas.
Background and challenges
In March 2011, the Great East Japan Earthquake triggered a massive tsunami that caused catastrophic damage in the Tohoku region along the eastern coast of Japan, highlighting the challenges of disaster mitigation efforts.
Tsunami prediction represents a key area in this regard, it requires the development of technologies that allow authorities to quickly obtain accurate and detailed information, to help disaster management teams mitigate damage by ordering appropriate evacuation actions.
About the new technology and initiative developed with Fugaku
To solve this problem, the companies worked together to develop a high-resolution AI technology capable of instantly predicting flooding caused by tsunamis, using the Fugaku assessment environment in the research topic “AI-based tsunami forecasting to exascale to predict the unforeseen ”.
For the AI model, a new deep learning technology was developed, with a two-stage AI setup. It first zooms in on the flood situation on land with approximate resolution and from tsunami waveforms observed offshore. The AI is then used to increase the resolution of the estimated flood conditions and optimize the computational performance for learning in Fugaku.
The supercomputer Fugaku, trained to predict tsunamis
When this technology was applied to the case of predicting tsunami floods in Tokyo Bay caused by a theoretical massive earthquake in the Nankai Basin, it was confirmed that it was possible to make a highly accurate prediction using a normal PC in a matter of seconds. , for a variety of different tsunami scenarios.
If Fugaku’s large-scale, high-speed performance continues to be used in the future, and the system is simultaneously trained with other tsunami scenarios, partners can help realize AI that can deliver predictions of tsunamis and unexpected floods in an area. more espacious.
In the future, the three organizations will contribute to tsunami disaster prevention measures in Japan and internationally, promoting the expansion of AI technology and assessing the accuracy of predictions, to help accelerate the application of the technology. In the real world.