Deep Learning to Extract Earthquakes and Low-Frequency Tremors in Continuous Seismic Waveforms

Applications of artificial intelligence (AI) techniques such as deep learning have been increasing in modern seismology, as big projects related to the integration of state-of-the-art information science and seismology are going on in Japan. Typical applications of AI in seismology are extractions of seismic events buried in continuous seismic waveforms. We show our recent developments in deep learning methods (1) to extract and classify P-/S-waves and noises in digital continuous waveforms and (2) to extract low-frequency tremors in historical paper records obtained by seismometers more than 50 years ago.