Monitoring of seismic signals generated by slow deformation at convergent and transform plate boundaries worldwide, known as tectonic tremor, might provide insights into deformation processes in the source regions of megathrust earthquakes. Tremor signals occur dominantly in the 2–8 Hz frequency band and can last for tens of seconds to several minutes, in contrast to typical earthquakes that produce seismic signals at frequencies up to several tens of hertz and last less than a minute. Because tremor is caused by stochastic processes, the resultant waveforms are represented by a stochastic function and construction of deterministic measures to discriminate tremor signals from earthquakes is very difficult. In this study, we used a convolutional neural network (CNN) to discriminate the signals of tectonic tremor from those of local earthquakes in running spectral images of these signals. We developed a method (seismic running spectra‐CNN [SRSpec‐CNN]) that is sensitive to the absolute frequency of signal appearance, which reflects the physical properties of the signal source, but is insensitive to the time of signal onset. SRSpec‐CNN has 130,211 parameters that were trained by 17,213 images of $64×64pixels$⁠. Based on simultaneous analyses of the frequency contents and durations of the signals, we achieved 99.5% accuracy for our identifications of signals from tectonic tremor, local earthquakes, and noise. Because running spectra clearly differentiate the characteristic features of these signals, we were able to achieve this high accuracy by using a CNN of simple architecture.

Off Cape Muroto area, along the Nankai Trough in southwest Japan, is a typical area with adjacent occurrences of slow and megathrust earthquakes. High‐resolution monitoring of slow earthquakes is necessary to understand tectonic conditions. In the off Muroto area, distributed acoustic sensing (DAS) measurement, which provides high‐density strain data, has been conducted using offshore fiber‐optic cable. We observed shallow tremors, a type of slow earthquakes, using DAS measurement for the first time. The characteristics of the tremor signals recorded by DAS were longer durations than those recorded in seismographs and composed of several phases with apparent velocities of several hundreds of m/s to several km/s that are coherent only in tens of meters. By combining DAS and seismograph data, we located these tremors around a subducted seamount peak. Therefore, spatial relationship between slow earthquakes and structural characteristics is suggested. This is a pioneering study of slow earthquake observation using DAS.

Green’s functions (GFs) for elastic deformation due to unit slip on the fault plane comprise an essential tool for estimating earthquake rupture and underground preparation processes. These estimation results are often applied to generate important information for public such as seismic and tsunami hazard assessments. So, it is important to minimize the distortion of the estimation results on the numerical models used for calculating GFs to guarantee assessment reliability. For this purpose, we here calculated GFs based on a numerical model that is of high fidelity to obtain realistic topography and subsurface structural models of the Earth. We targeted two well-known subduction zones in Japan, the Nankai Trough and the Japan Trench. For these subduction zones, databases for realistic topography and subsurface structural models of the Earth are available in the “Japan integrated velocity structure model version 1”, which was proposed for earthquake hazard assessments conducted by the Japanese government. Furthermore, we eliminated the inconsistency in processing calculated GFs and space geodetic observation data for surface displacements, which is often overlooked, by using the same coordinate system. The ellipsoidal shape of the Earth, which is often approximated with a projected plane or a spherical shape, was also incorporated by faithfully following the definitions of the coordinate systems in Geodetic Reference System 1980, which is the global standard for space geodesy. To calculate elastic GFs based on such high-fidelity subduction zone databases with the ellipsoidal shape of the Earth, we introduced the finite element (FE) method. In the FE meshes, the resolution of the topography and subsurface structure is the same as that of the original databases. Recent development of the state-of-the-art computation techniques for the rapid calculation of crustal deformation using large-scale FE models allows for GF calculation based on such a high-fidelity model. However, it is generally not easy to perform such calculations. Thus, we released a library for the GFs calculated with 1-km grid spacing on the ground surface in this study to the geoscience community on a web server, aiming to contribute more reliable seismic and tsunami hazard assessment.

Motivated by the fact that temporal earthquake aperiodicity was observed off Kamaishi just after the 2011 Tohoku earthquake, we performed numerical simulations of chain reactions due to the postseismic slip of large earthquakes by applying rate‐ and state‐dependent friction laws. If the repeater is composed of single asperity, our results show that, (i) a mixture of partial and whole rupturing of a single asperity can explain some of the observed variability in timing and size of the repeating earthquakes off Kamaishi; (ii) the partial rupturing can be reproduced with moderate frictional instability with the aging‐law and not the slip or Nagata laws; (iii) the perturbation of the activated earthquake hypocenters observed mostly in the ESE‐WNW direction may reflect the fact that the large postseismic slip of the 2011 Tohoku earthquake propagated from ESE to WNW off Kamaishi; (iv) the observed region of repeating earthquake quiescence may reflect the strong plate coupling of megathrust earthquakes.