Near-fault ocean-bottom observation of an offshore earthquake by DONET: Array-based measurement of tsunami phase speed and early estimation of tsunami height

Dense and wide-span ocean-bottom networks have been recently deployed in the offshore Japan such as DONET. Important features of these networks are: (1) seismometer and ocean bottom pressure (OBP) gauge installed at a same site, and (2) high sampling rate 10 Hz of OBP gauge. They enable us to analyze OBP records by traditional array-based methods in seismology, to retrieve coseismic ground motions, and to obtain near-fault information for early warning of earthquakes (EQ) and tsunamis.

During an EQ, a near-fault OBP record changes not only by tsunamis but also by depth change of an instrument, vertical acceleration of the seafloor, and ocean acoustic waves (e.g. Saito and Tsushima, 2016), which makes any near-fault OBP record highly complicated.

In this study, we analyzed the OBP and strong-motion accelerogram records of DONET associated with the 2016 Off-Mie EQ (Mw 6.0) that occurred inside the network. First, we estimated the phase speed of tsunami using the multi-station method (Aki, 1961). Secondly, we compared the OBP records with acceleration, velocity and displacement waveforms in both time and frequency domains. To obtain velocity and displacement waveforms from the original accelerograms, we applyed the baseline correction method based on Iwan et al. (1985) because a simple integration leads to unstable results. Finally, we estimated tsunami heights at near-fault stations in a near real-time basis.

Our main results are as follows: (1)the estimated phase speed for < 0.03 Hz agree well with the theoretical one, but larger by the order of two for > 0.03 Hz, which may correspond to seismic surface waves (3-4 km/s) or scattered waves away from each great circle (over 10 km/s) (Fig.1). (2) The OBP records are proportional to ocean-bottom velocity data especially for their first motions, while proportional to accelerations at around 0.1 Hz, consistent with the numerical simulations by Saito (2017). (3) We could retrieve tsunami and displacement signals at 10-s intervals using a 60-s time window and a 0.05-0.15 Hz filter to remove the effect of accelerations, agreeing well with the low-pass filtered (< 0.01 Hz) OBP data of a long record length (Fig.2).

In summary, DONET OBP records enable us not only to trace tsunami propagation for < 0.03 Hz but also to detect reasonable tsunami heights with 3 minutes for early warning.