VALIDATION ANALYSIS OF SEAFLOOR STRONG MOTION DATA RECORDED BY S-NET IN THE JAPAN TRENCH AREA

S-net is a large-scale inline-type seafloor observation network for earthquakes and tsunamis in the Japan Trench area and has been in operation since 2016. It consists of 150 observatories and covers about 1000 and 250 km in the directions parallel and perpendicular to the Trench axes. One of the main objectives of the establishment of the network was to enhance the Japan Meteorological Agency earthquake early warning (EEW) and tsunami early warning. The S-net sensors have been housed in cylindrical pressure vessels, which have been buried about one meter beneath the seafloor in the shallow water regions (< 1500 m) while they are laid freely on the seafloors in the deeper water regions. Because of these peculiarities in the settings of these network stations, the seismometers have been found to experience induced rotations even for moderate shakings of about 100 cm/s/s. Moreover, the seismic recordings, especially those from unburied stations, have been suspected to include natural-vibration noises of the pressure vessels at higher frequencies over about 5 Hz. While these observations obviously indicate the limitations of the network in terms of the quality of the strong-motion records compared to those observed on the land, they are the first sensors to detect and transmit the shakings to the data centres for offshore earthquakes in the region. Several significant results have been published by researchers using the S-net waveform data, and methods have been developed for Tsunami disaster prevention and EEW using the S-net data. In the present study, strong-motion data from additional earthquakes were combined to our previous dataset, and we compared the 5% damped acceleration response spectra (ARS) at periods of 0.1-10 s between the land and S-net sites computed from the horizontal component records for earthquakes with magnitudes between 5.3 and 7.4. We discarded records at S-net sites that indicated static rotations of sensors greater than 0.1 degrees. Based on comparisons of data between the land and S-net sites and residuals with respect to a reference ground motion prediction equation, we found that the ARS at shorter periods could be considered reasonable, whereas the values at the longer periods need to be handled cautiously.