3:00 PM - 3:15 PM
[HDS10-06] Optimization of Filters for Real-Time Tsunami Detection Using Ocean Bottom Pressure Data
Keywords:Tsunami, Observation waveform, Ocean bottom pressure observation
Offshore tsunami observation is really effective in terms of real-time tsunami forecasting because it enables direct observation of tsunamis generated by earthquakes and other phenomena before the tsunami reaches the coastal region. The waveform analysis of far-field tsunamis observed by the Deep-ocean Assessment and Reporting of Tsunamis (DART), which is widely installed in the open ocean, reveals the effect of dispersive waves that cannot be reproduced by the linear long wave approximation (Baba et al., 2017).
DART is often installed in deep water near the trench axis, where huge earthquakes are predicted to occur. On the other hand, the Seafloor observation network for earthquakes and tsunamis along the Japan Trench (S-net), the Dense Oceanfloor Network System for Earthquakes and Tsunamis (DONET), and the Nankai Trough Seafloor Observation Network for Earthquakes and Tsunamis (N-net) are installed not only in deep sea but also in shallow water near the land side. These observation networks record sea level changes as pressure changes using pressure gauges installed on the seafloor to monitor tsunami. In many cases, hydrostatic equilibrium is assumed, therefore wave heights at the sea surface are assumed to be proportional to pressure changes on the seafloor. On the other hand, theoretically, it is known that the seafloor pressure attenuates by 1/cosh(kh) at a sea depth of h with respect to the spatial wavenumber k of the sea level change, and its effect on the tsunami waveform has also been confirmed (Chikasada et al. (2018) and Chikasada (2019)).
In order to detect tsunamis using data from the seafloor pressure network in real-time, a trigger that a tsunami is occurring is required. For seismic tsunamis, the time of earthquake occurrence can be used as a tsunami detection trigger, but for non-seismic tsunamis, we should consider a trigger to detect the tsunami using only ocean bottom pressure changes. As the simplest method, the ratio of the short-term average to the long-term average of the waveform amplitude is commonly used. In such method, tsunami occurrence is recognized when the ratio exceeds a threshold level given in advance. There are other detection methods, but most of them are based on amplitude alone and do not based on wavelength of tsunami. Therefore, using data from a wide-area observation network causes different sensitivity due to the difference in water depth at the observation location. In addition, it is necessary to change the threshold value because wind waves, which generally have shorter wavelengths than tsunamis, cause false detection at stations with shallower water depths, but quantitative discussions have not been conducted.
In this study, we propose a method for setting the threshold independent of water depth by applying a filter to the waveform at a hypothetical uniform water depth, based on the wave number and sea depth dependent attenuation rate. We also report the results of our study on appropriate filters for tsunami detection, such as a method to deal with ocean bottom pressure steps in an in-line ocean bottom pressure gauge and waveform deformation due to tidal change removing using a Butterworth filter, etc.
DART is often installed in deep water near the trench axis, where huge earthquakes are predicted to occur. On the other hand, the Seafloor observation network for earthquakes and tsunamis along the Japan Trench (S-net), the Dense Oceanfloor Network System for Earthquakes and Tsunamis (DONET), and the Nankai Trough Seafloor Observation Network for Earthquakes and Tsunamis (N-net) are installed not only in deep sea but also in shallow water near the land side. These observation networks record sea level changes as pressure changes using pressure gauges installed on the seafloor to monitor tsunami. In many cases, hydrostatic equilibrium is assumed, therefore wave heights at the sea surface are assumed to be proportional to pressure changes on the seafloor. On the other hand, theoretically, it is known that the seafloor pressure attenuates by 1/cosh(kh) at a sea depth of h with respect to the spatial wavenumber k of the sea level change, and its effect on the tsunami waveform has also been confirmed (Chikasada et al. (2018) and Chikasada (2019)).
In order to detect tsunamis using data from the seafloor pressure network in real-time, a trigger that a tsunami is occurring is required. For seismic tsunamis, the time of earthquake occurrence can be used as a tsunami detection trigger, but for non-seismic tsunamis, we should consider a trigger to detect the tsunami using only ocean bottom pressure changes. As the simplest method, the ratio of the short-term average to the long-term average of the waveform amplitude is commonly used. In such method, tsunami occurrence is recognized when the ratio exceeds a threshold level given in advance. There are other detection methods, but most of them are based on amplitude alone and do not based on wavelength of tsunami. Therefore, using data from a wide-area observation network causes different sensitivity due to the difference in water depth at the observation location. In addition, it is necessary to change the threshold value because wind waves, which generally have shorter wavelengths than tsunamis, cause false detection at stations with shallower water depths, but quantitative discussions have not been conducted.
In this study, we propose a method for setting the threshold independent of water depth by applying a filter to the waveform at a hypothetical uniform water depth, based on the wave number and sea depth dependent attenuation rate. We also report the results of our study on appropriate filters for tsunami detection, such as a method to deal with ocean bottom pressure steps in an in-line ocean bottom pressure gauge and waveform deformation due to tidal change removing using a Butterworth filter, etc.