Detection of Deep Low-Frequency Tremors from Continuous Paper Records at a Station in Southwest Japan About 50 Years Ago Based on Convolutional Neural Network

The establishment of the High Sensitivity Seismograph Network (Hi-net) in Japan has led to the discovery of deep low-frequency tremors. Since such tremors are considered to be associated with large earthquakes adjacent to tremors on the same subducting plate interface, it is important in seismology to investigate these tremors before establishing modern seismograph networks that record seismic data digitally. We propose a deep-learning method to detect evidence of tremors from seismogram images recorded on paper more than 50 years ago. In this study, we trained a convolutional neural network (CNN) based on the Residual Network (ResNet) with seismogram images converted from real seismic data recorded by Hi-net. The CNN trained by fine-tuning achieved an accuracy of 98.64% for determining whether an input image contains tremors. The Gradient-weighted Class Activation Mapping (Grad-CAM) heatmaps for visualizing model predictions indicated that the CNN successfully detects tremors without being affected by teleseisms. The trained CNN was applied to the past seismograms recorded from 1966 to 1977 at the Kumano observatory, in southwest Japan, operated by Earthquake Research Institute, The University of Tokyo. The CNN showed potential for detecting tremors from past seismogram images for broader applications, such as publishing a new tremor catalog, although further training using data including more variables such as the thickness of the pen would be required to develop a universally applicable model.