Offshore real‐time ocean bottom networks of seismometers and ocean bottom pressure (OBP) gauges have been recently established such as DONET and S‐net around the Japanese islands. One of their purposes is to practice rapid and accurate tsunami forecasting. Near‐fault OBP records, however, are always contaminated by nontsunami components such as sea‐bottom acceleration change until an earthquake stops its fault or sea‐floor motions. This study proposes a new method to separate tsunami and ocean bottom displacement components from coseismic OBP records in a real‐time basis. Associated with the Off‐Mie earthquake of 2016 April 1, we first compared OBP data with acceleration, velocity, and displacement seismograms recorded by seismometers at common ocean bottom sites in both time and frequency domains. Based on this comparison, we adopted a band‐pass filter of 0.05–0.15 Hz to remove ocean‐bottom acceleration components from the OBP data. Resulting OBP waveforms agree well with the tsunami components estimated by a 100‐s low‐pass filter with records of several hundred seconds in length. Our method requires only an early portion of a given OBP record after 30 s of an origin time in order to estimate its tsunami component accurately. Our method enhances early tsunami detections with near‐fault OBP data; that is, it will make a tsunami forecasting system faster and more reliable than the previous detection schemes that require data away from source regions or after coseismic motions are over.

A back‐projection method has been applied to many earthquakes in seismology due to its simple and low computational cost, and it can estimate complex fault rupture processes without any specific a priori information. In this study, we applied the back‐projection method to the tsunami records observed using an ocean‐bottom pressure gauge array and demonstrated it to be a powerful new tool other than the familiar waveform inversion. The obtained back‐projection image was consistent with the initial tsunami height distributions estimated by previous waveform inversions, and its spatial resolution appeared to be even better. Our result suggests that the fault size of the 2016 Off‐Fukushima earthquake was about half, different from the scaling law of standard earthquakes. The present tsunami back‐projection analysis can also estimate the feature of early tsunami propagations. In addition, the estimated image seems to be reliable even 30 min after the origin time, so the back‐projection analysis will be useful in an early detection of the location and spatial extent of a tsunami source. In the present case, the number of available stations in the analysis was found to be affected by the diffraction of tsunami propagation caused by the refraction by a high velocity zone near the Japan Trench. In other words, the further the source is from the coast, the more stations to be analyzed are available. Since most tsunami‐generating earthquakes occur near the subduction axis or its outer‐rise region, the back‐projection analysis should be effective for source estimation of the majority of tsunami‐generating earthquakes.

Summary On 15 January 2022, a large eruption of the Hunga Tonga-Hunga Ha’apai volcano in Tonga triggered globally observed tsunami waves. While the first arrival in the observed wavetrains is now widely known to be related to the atmospheric Lamb wave generated by the eruption, large later phases, whose amplitudes were comparable to the first ones, were also recorded. In this study, we estimated the source of the later phases based on the Vespa analysis and proposed a new numerical scheme to reproduce them. The Vespa analysis estimates the arrival time and incident angle of each signal by a slant-stack process using its theoretical travel time. The Vespa analysis revealed that small atmospheric waves excited the large later tsunamis. For the numerical experiments, we used two types of synthetic methods: finite difference method and normal mode theory. We found that both a good atmospheric wave model and bathymetric effect were important to generate the atmospheric-induced tsunamis corresponding to the later phases. A hybrid method calculating tsunamis by the finite difference method with the atmospheric waves by the normal mode theory as the input successfully reproduced the observed records, particularly in amplitude over the entire records.