The International Research Institute of Disaster Science, Tohoku University, the Earthquake Research Institute, the University of Tokyo, and the Fujitsu Laboratories have succeeded in constructing an AI model that predicts tsunami inundation in coastal areas with high resolution and in real time by utilizing the supercomputer "Tomitake".
Since the Great East Japan Earthquake, the tsunami observation network in the waters near Japan has been strengthened to reduce damage through appropriate evacuation behavior, and technological development of highly accurate tsunami prediction in coastal areas using real-time tsunami observation data offshore has been promoted. ing.Conventional tsunami forecasting requires simulation calculations and database searches using a large-scale supercomputer at the time of a disaster, making it difficult to construct and operate a forecasting system.
This time, the research group conducted a large number of high-resolution tsunami simulations using the supercomputer "Tomitake".Using the obtained offshore tsunami waveforms and coastal inundation conditions as teacher data, we constructed a new AI model that extends deep learning technology.By inputting the tsunami waveform observed offshore at the time of an earthquake into the constructed AI model, it will be possible to predict the inundation situation in the coastal area with a high spatial resolution of 3 m before the arrival of the tsunami.
This makes it possible to grasp detailed inundation prediction information for each section, such as the rise of local tsunamis that incorporates the effects of social infrastructure such as buildings, structures, and roads in coastal urban areas, and more appropriate evacuation. Can support actions.In addition, the AI model learned in advance using "Tomitake" can be executed in a few seconds even on a general personal computer, which greatly facilitates the construction of a real-time inundation prediction system that previously required a supercomputer.
In the future, it is said that it will contribute to tsunami disaster prevention measures in Japan and overseas by expanding AI technology and evaluating prediction accuracy toward the practical application of the developed technology.