A new approach for tsunami early warning using tsunami observations in a source region

Abstract:

After the 2011 devastating Tohoku tsunami, improvement of tsunami early warning system is one of key issues in Japan. Japanese government was decided to install 125 ocean bottom pressure sensors and seismometers with a cable system along the Japan and Kurile trench. Each sensor is separated by 30km. We should develop a new approach for real-time tsunami forecast using those newly available data combined with GNSS data or seismic data.  A well-recognized problem to use tsunami data at pressure sensors on the top of tsunami source area is a fact that a large vertical coseismic deformation due to a large earthquake cannot be observed at those sensors. The sensors observe a tsunami wave when it starts to propagate. Because of that problem, GSNN data or seismic data are typically used to estimate the coseismic deformation for the tsunami numerical simulation. In this paper, we develop a new technique, which solve the problem. Our technique uses the observations at pressure sensors on the tsunami source area as an input to compute the tsunami directly. Actual tsunami heights at the sensors on the source area is unknown because the cosismic vertical deformation is unknown. However, we can observe directly the time derivative of tsunami heights at those sensors. Time derivatives of tsunami heights at each point are used as inputs to compute the tsunami height distribution in the calculated area. Then we can numerically compute a tsunami using a traditional finite difference technique from the tsunami height distribution computed. For numerical test, first, we compute the synthetic tsunamis using the fault model with 1 minute grid system. The computed tsunami waveforms at 15 minutes x 15 minutes grid points are used as the observed data for this new technique. Each observed point is separated by 15 minutes, about 30km. The result show that the accuracy of tsunami computation is good enough for tsunami forecast. Tsunami generation with a long duration, such as tsunami earthquake, is also tested for this technique. The results are also good enough for tsunami forecast. Without information of an earthquake, tsunami can be computed from observed tsunami data near the source area using this new technique. The technique is useful for a tsunami early warning system. It is important to be combined with other approaches using other data such as GSNN data.