10:00 〜 10:15
[MGI26-05] 津波誘導磁場に基づく津波データ同化
キーワード：津波データ同化、津波誘導磁場、津波早期警告
Tsunami data assimilation integrates observed tsunami data into numerical models to enhance the accuracy of tsunami source estimation and improve tsunami propagation predictions. Previous studies have primarily used tsunami pressure data, tidal gauge data, and coastal radar data, all of which are derived from sea level changes caused by tsunami propagation. In addition to sea level variations, tsunamis also generate an induced magnetic field, known as the Tsunami-induced Magnetic Field (TMF), during their propagation. Recent studies have demonstrated that the horizontal velocity field of a tsunami can be obtained from TMF data, even with observations at a single station. Building on this, we propose a novel approach to incorporating TMF data into tsunami data assimilation and investigate its predictability.
To assimilate the TMF data with the tsunami model, we introduce an observation matrix to convert the TMF data to tsunami velocity field, then use a least squares approach to integrate it with a simple linear long-wave equation model, aiming to find an optimal tsunami velocity field for the entire area at the present moment.
Environmental noise in observed TMF data can significantly affect the accuracy of tsunami velocity estimates. Therefore, we first examine how errors in TMF observations influence the predictability of TMF in tsunami data assimilation. To achieve this, we simulate tsunami propagation in the open ocean using a simple linear long-wave equation. Artificial TMF datasets are generated from the simulated tsunami velocity field and contaminated with different levels of random noise to represent various noise conditions. We then apply a data assimilation method to estimate the current tsunami motion states from the artificial TMF data and predict tsunami propagation based on these estimated states. Finally, we analyze how observational errors propagate through the prediction process and assess their growth, which defines the predictability of TMF.
Tsunami pressure data are widely used for tsunami early warning; however, they only provide information on sea level changes at a single station. In contrast, TMF data from a single station can yield information about the tsunami’s horizontal velocity, encompassing both sea level variations and propagation direction. This suggests that TMF data could be more effective for tsunami early warnings. To evaluate this, we compare the predictability of tsunami pressure data and TMF data in tsunami data assimilation. Specifically, we generate four artificial tsunami pressure datasets at different locations and introduce random noise. Four observation points are chosen because a single pressure measurement is insufficient for tsunami data assimilation, as it lacks information on tsunami propagation direction. Additionally, the noise level in the pressure data is set lower than that in the artificial TMF data to reflect real-world conditions. We then assess the predictability of multiple pressure stations (two, three, and four stations) and compare it with the predictability of TMF data.
This study not only introduces a new type of tsunami data into tsunami data assimilation but also explores a potential approach for enhancing tsunami early warning systems in the future.
To assimilate the TMF data with the tsunami model, we introduce an observation matrix to convert the TMF data to tsunami velocity field, then use a least squares approach to integrate it with a simple linear long-wave equation model, aiming to find an optimal tsunami velocity field for the entire area at the present moment.
Environmental noise in observed TMF data can significantly affect the accuracy of tsunami velocity estimates. Therefore, we first examine how errors in TMF observations influence the predictability of TMF in tsunami data assimilation. To achieve this, we simulate tsunami propagation in the open ocean using a simple linear long-wave equation. Artificial TMF datasets are generated from the simulated tsunami velocity field and contaminated with different levels of random noise to represent various noise conditions. We then apply a data assimilation method to estimate the current tsunami motion states from the artificial TMF data and predict tsunami propagation based on these estimated states. Finally, we analyze how observational errors propagate through the prediction process and assess their growth, which defines the predictability of TMF.
Tsunami pressure data are widely used for tsunami early warning; however, they only provide information on sea level changes at a single station. In contrast, TMF data from a single station can yield information about the tsunami’s horizontal velocity, encompassing both sea level variations and propagation direction. This suggests that TMF data could be more effective for tsunami early warnings. To evaluate this, we compare the predictability of tsunami pressure data and TMF data in tsunami data assimilation. Specifically, we generate four artificial tsunami pressure datasets at different locations and introduce random noise. Four observation points are chosen because a single pressure measurement is insufficient for tsunami data assimilation, as it lacks information on tsunami propagation direction. Additionally, the noise level in the pressure data is set lower than that in the artificial TMF data to reflect real-world conditions. We then assess the predictability of multiple pressure stations (two, three, and four stations) and compare it with the predictability of TMF data.
This study not only introduces a new type of tsunami data into tsunami data assimilation but also explores a potential approach for enhancing tsunami early warning systems in the future.