[1L3-OS-17] 地震研究と人工知能

[1L3-OS-17-05] Detection of Low-Frequency Tremors in Seismic Waveform on Paper Records Based on Residual Learning

Ryosuke Kaneko¹, 〇Hiromichi Nagao¹, Shin-ichi Ito¹, Hiroshi Tsuruoka¹, Kazushige Obara¹ (1. The University of Tokyo)

Keywords:Earthquake, Deep Low-Frequency Tremor, Residual Learning, Image Recognition

Low-frequency tremors, a type of slow earthquake, were discovered in southwest Japan owing to the establishment of spatially-dense seismic observation networks in Japan. Tremors are considered to occur along plate boundaries in areas slightly deeper or shallower than ordinary earthquakes, and are therefore expected to be associated with massive plate boundary earthquakes. Current tremor catalogs, which list tremor occurrence times and hypocenter locations, contain only tremor events after 2001. Considering that plate boundary earthquakes periodically occur with an interval of approximately 100 or 200 years in southwest Japan, it is important to catalog tremors recorded in historical seismograms before the establishment of the modern seismic observation networks. In this study, we developed a convolutional neural network based on the ResNet to detect tremors in a large amount of historical seismograms directly recorded on paper sheets with pens about 50 years ago. The trained network successfully identified many previously unknown tremors in the past.