Researchers at the Institute of Industrial Science, The University of Tokyo (UTokyo-IIS) have developed a real-time data assimilation system that substantially improves streamflow and flood forecasting accuracy across Japan. This new data assimilation system outperforms Japan’s current early warning system. Compared with the previous early warning system, the new data assimilation system yields improved forecasts for 80% of the major river reaches in Japan and enables reliable flood prediction across more than 60% of the observation stations in Japan.
The data assimilation system was evaluated against major flood events: Typhoon Hagibis in 2019, the Northern Japan Flood in 2022, and the Akita flash flood in 2024. Across all three cases, the data assimilation framework demonstrated potential to improve the one-day-ahead flood forecasts, with the most striking gains observed in events where the existing early warning system had historically performed poorly.
“What surprised us most was the scale of the improvement in the flood forecasts. We kept the model exactly the same and just used gauge observations to correct the model initial state. This action alone was sufficient to start capturing peak flows in flash flood cases that the previous early warning system had missed, such as in the Akita 2024 flood events,” said Yingying Liu, lead author of the study and former master’s student at UTokyo-IIS.
Data assimilation is a powerful technique that systematically reduces discrepancies between model simulations and observed conditions by integrating real-world measurements into the modeling cycle. In this study, the system ingested hourly water level observations from 1800 in situ gauges within the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) monitoring network. The broad spatial coverage of the MLIT gauge network allows the data assimilation system to deliver improved forecast performance across Japan’s major river basins. By initializing the model with a more accurate representation of current conditions, the system consistently produced better predictions over lead times ranging from several hours to one day ahead.
“With more reliable one-day-ahead flood forecasts, emergency managers and communities have more time to act before the water arrives. This critical additional time window can save lives and reduce economic damage,” added Kei Yoshimura, professor at UTokyo-IIS. “And the exciting thing is, this data assimilation system is not just for Japan. The same approach can work in any other flood-prone region and offers a scalable pathway to improved flood management globally.”
The research team expects the data assimilation framework to contribute to improved flood forecasting across an expanding range of extreme events. It is anticipated that the approach will be adopted in operational early warning systems across Japan and applied in other flood-prone regions around the world.
###
The article, “Application of real-time data assimilation system to improve streamflow forecasts in Japan” was published in Journal of Hydrology at DOI:10.1016/j.jhydrol.2026.135680.
The Institute of Industrial Science, The University of Tokyo (UTokyo-IIS) is one of the largest university-attached research institutes in Japan. UTokyo-IIS is comprised of over 120 research laboratories—each headed by a faculty member—and has over 1,200 members (approximately 400 staff and 800 students) actively engaged in education and research. Its activities cover almost all areas of engineering. Since its foundation in 1949, UTokyo-IIS has worked to bridge the huge gaps that exist between academic disciplines and real-world applications.
Published: 09 Jun 2026
10.1016/j.jhydrol.2026.135680
Bringing discovery to light
Leave a Reply