[S-SS05] Fault Rheology and Earthquake Physics

[SSS05-03] Exploring the strong heterogeneities in the seismogenic environment using seismic later phase and machine learning

★Invited Papers

*Yuta Amezawa¹ (1.Tokyo Institute of Technology)

Keywords:later phase, crust, heterogeneous structure, machine learning

In the upper-middle crust, which is the environment for crustal earthquakes, the existence of significant heterogeneous structures due to crustal fluid reservoirs and differences in crustal components etc. have been suggested by low-seismic wave velocity zones and low-electric resistivity zones. High-resolution and comprehensive investigation of crustal fluid behavior is important for understanding the mechanism of crustal earthquakes occurring in the vicinity of such reservoirs. Seismic later phase is one of the useful tools for these objectives. We determine later phases, in this talk, are reflected and scattered waves generated by subsurface seismic velocity contrasts and localized strong scatterers. They observed in the coda part of direct waves. Later phases are particularly sensitive to their origins. Thus, we can estimate the characteristics of the later phase origin by examining the arrival time and/or the temporal changes in the later phase shape. Furthermore, if we detect later phases from the seismogram big-data accumulating by the permanent dence observation network such as Hi-net of the National Research Institute for Earth Science and Disaster Prevention, we can investigate the later phase origins comprehensively with high spatiotemporal resolution. In this talk, I introduce the topics about [1] the existence of crustal fluid at the origin of seismic later phase suggested by the temporal change in the waveform shape of later phase in S coda, [2] our machine-learning-based automatic detection model of the later phase in S coda, and [3] the results of comprehensive detection of the later phase in S coda in the inland area of Japanese archipelago.

Acknowledgements: We used seismic waveform records from NIED Hi-net, JMA unified hypocenter catalog, and arrival time. Studies [2] and [3] were supported by MEXT Project for Seismology toward Research Innovation with Data of Earthquake (STAR-E) Grant Number JPJ010217.

Reference:
[1] Y. Amezawa, M. Kosuga, and T. Maeda. Temporal changes in the distinct scattered wave packets associated with earthquake swarm activity beneath the Moriyoshi-zan volcano, northeastern Japan. Earth Planets Space, vol.71, 132 (2019) https://doi.org/10.1186/s40623-019-1115-6
[2] Y. Amezawa, T. Uchide, and T. Shiina. Automatic detection of S-wave later phase by 1D convolutional neural network. SSJ Fall Meeting (2022)
[3] Y. Amezawa, T. Uchide, and T. Shiina., J.Ogata, S. Fukayama, and H. Kuroda. Comprehensive detection of later phase in S coda throughout the inland Japanese archipelago with deep learning. SSJ Fall Meeting (2023)