[S-SS02] Seismological advances in the ocean

[SSS02-07] Data-driven seismic signals detection: the variety of OBS events in the SW offshore Taiwan and their implications

★Invited Papers

*emmy TY CHANG¹ (1.Institute of Oceanography, National Taiwan University)

Keywords:ocean bottom seismometer, gas emission, machine learning

The seafloor OBS records every ground motion energy reaching the stations, which can be caused by the tectonic origins or by the transformed signals converting from the acoustic waves propagating in ocean water. Except for the ambient noises which are the random vibrations in amplitudes (but maybe frequency-dependent), most of the OBS signals can exhibit the specific waveforms corresponding to the different mechanical causes. However, it is a tedious task to recognize all these waveforms by the conventional way of human identification, and the quantification of the occurrences of those waveforms is never been complete. Our study applies the machine Learning package FAST to the continuous OBS seismograms. The OBS data are acquired from two OBS arrays deployed in the SW offshore Taiwan, where is well known to be the plate convergent boundary and the gashydrate storage field. The FAST (Fingerprint and Similarity Thresholding, Yoon et al. 2015) is a novel method for large-scale event detection. FAST is an unsupervised detector. It does not require any examples of known event waveforms or waveform characteristics for detection. This allows FAST to discover new waveforms with the OBS data. The preliminary FAST result shows that the tectonic earthquakes are very numerous in the offshore area; most of them are microseisms that are not detectable for the inland stations. Taking the occurrence of the tectonic earthquakes as references, we found that a few T-wave signals seem to occur solely. And the short-duration events (SDEs, in a duration of less than 2 seconds) are also abundant in the SW offshore Taiwan. The SDEs can be classified into two groups, one with a specific monotone vibration attenuating with time; the other in an abrupt emerges without any specific waveforms. We consider the monotone SDEs are the methane bubbles bursting at the sediment /water interface. The other SDEs are mostly detected at the mid-slope stations. They may be caused by the animal touch or the unknown local geological events. Our study demonstrates that with good identification and classification, the OBS seismic records can be a proxy in studying the marine environment.