Abstract

Tsunami prediction with high accuracy and speed is important to disaster prevention. Traditional tsunami early warning methods are based on source information instantaneously estimated from seismic observations. This information significantly impacts the accuracy of tsunami prediction. However, a limitation of this approach is its inability to respond to tsunamis caused by submarine landslides or volcanic eruptions.

Tsunami early earning based on data assimilation has become feasible now with the deployment of offshore observation networks. This method directly uses offshore tsunami observations to reconstruct the tsunami wavefield and predict coastal tsunamis. Offshore observations can be provided by offshore bottom pressure gauges (OBPGs) or high-frequency (HF) radars. Unlike traditional early warning methods, data assimilation does not require source information. Therefore, in regions where observation networks are deployed, it can be applied to tsunamis from any source, as long as the tsunami signals are detected.

I have been focusing on the application of data assimilation to tsunami prediction, advancing data assimilation algorithms to enable early warnings for all types of tsunamis. I proposed Green’s Function-Based Tsunami Data Assimilation, which precomputes tsunami waveforms between observational points and prediction points and uses their relationship as Green’s function. It has been applied to events such as the 2015 Torishima tsunami, the 2016 Fukushima tsunami, and the 2022 Tonga volcanic tsunami. For example, the 2022 Tonga volcanic tsunami was a non-seismic tsunami triggered by atmospheric waves from the massive eruption. I used the offshore observations of DONET OBPGs to predict the coastal tsunamis in Shikoku Island and Kii Peninsula. It revealed that OBPGv networks are extremely effective tools for tsunami detection and prediction.

Moreover, HF radars monitor the sea surface current velocity. I conducted retrospective analyses of tsunami prediction using HF radar observations. As a case study, the HF radar system in the Kii Channel detected the 2011 Tohoku tsunami. Based on data assimilation approach, I reconstructed the tsunami wavefield and predicted the waveform at Kobe. Comparison with actual observations at Kobe showed that data assimilation can accurately forecast coastal tsunami waveforms at least 50 min prior to tsunami arrival. Hence, it confirmed that data assimilation of HF radar data contributes to tsunami early warning.

In conclusion, tsunami prediction using offshore observations provides a new approach to early warning. As offshore observation systems continue to be developed, this method will play an increasingly significant role in tsunami disaster prevention.