今週のうなぎセミナーについてお知らせいたします。
Here is information of the Unagi-seminar(May 20).
************** Seminar on Seismology IV A,C /地震学ゼミナールIVＡ,C (Unagi Seminar) **************
科目：地震学ゼミナールIV A, C / Seminar on Seismology IV A, C（修士・博士）
日時：2021年 05月 20日 (木) 14:00~
場所：オンライン(Zoom)
Date and Time：2021-05-20, 14:00~
Place：Zoom
====
Presenter 1 (発表者 1)：
MPUANG Admore
Title (題目)：
Mass extinction in genetic algorithms applied to receiver functions inversion: method development
Abstract (要旨)：
The use of genetic algorithms in receiver functions inversion for crustal and uppermost mantle velocity-depth structure is well established. Their operation is based on the process of evolution of biological species, by using (pseudo)random numbers to control the selection, crossover and mutation processes in searching a model space for an optimal model. Despite their robustness, one drawback of the standard genetic algorithms is that towards the end of a run, only a few 'new ideas' are explored which may lead to the stagnation of the optimization process. This can be an especially major drawback for large model dimensions, such as in the inversion of receiver functions for dipping and anisotropic structures. To avoid this problem, mass-extinction, in which major parts of the explored model population are regularly replaced, is introduced to genetic algorithm inversion in this study to exploit highly fit models and random explorations of other domains. The concept of self-organized criticality is applied to control the size and frequency of the extinction events. Different configurations and model replacement strategies are tested against the standard genetic algorithm (Shibutani et al., 1996) for performance comparison. Preliminary results show that while mass-extinction may slow down the convergence speed of optimization through increased exploration, the control parameters can be tuned for great performance that can outperform the standard genetic algorithm at no significant additional computational costs.
............................................................................................................................................................
Presenter 2 (発表者 2)：
Shukei Ohyanagi
Title (題目)：
A synchronized activity of offshore seismicity and shallow tremor revealed by application of deep learning module to an OBS dataset
Abstract (要旨)：
The Japan Trench subduction zone is a cradle for various types of seismic activities, such as Mw 9.0 Tohoku-Oki megathrust event, VLFE, or SSE. In past few years, ocean bottom seismic observations by S-Net (cable network of OBSs) and other dense networks of free-fall type OBS reveal astonishing slow earthquake activities of this subduction zone. Unlike depth-varying transition of a megathrust earthquake fault to a slow earthquake fault in a Nankai subduction zone (e.g. Kato and Obara, 2016), slow earthquake regions illuminated by tectonic tremors transit to a megathrust earthquake region in along strike direction in the shallow subduction margin [Nishikawa et al., 2019; Tanaka et al., 2019; Ohta et al. 2019]. However, although acquired OBS datasets have used to search for tectonic tremors, they have not fully utilized to understand microseismicity of the same region. They are often used to relocate earthquakes cataloged in the JMA unified catalog which developed with recordings of onshore seismic network [e.g. Shinohara et al., 2012], but attempt to detect microseismicity solely visible from offshore seismic network is uncommon. As Obana et al. (2021) identify ordinal earthquakes that occur simultaneously with tectonic tremors based on a dense OBS observation, a microseismicity catalog developed from an OBS data would play a key-role on understanding interplay between ordinal earthquakes and shallow tremors.
During 2016 to 2017, we performed OBS Array of Arrays observation in offshore Fukushima region near the trench. This dataset reveals shallow tremor activity in this region [Ohta et al., 2019]. To investigate ordinal seismicity during the tremor activity, we applied Earthquake Transformer (EQT) [Mousavi et al., 2020], the newly developed deep learning module to perform detection and phase picking of microearthquakes to the OBS dataset. As a result, EQT detects more than 5000 microearthquakes. The detected events are located using hypomh [Hirata and Matsu’ura, 1987] with station corrections estimated from PS converted phases [Iwasaki et al., 1991]. The developed catalog reveals dense microseismicity in the down-dip side of a shallow tectonic tremor region identified by Nishikawa et al. (2019) and Ohta et al., (2019). A synchronized activity of ordinal earthquakes and shallow tremors are observed in the down-dip side of the tremor region, while Obana et al. (2021) reports such synchronization in the up-dip side of the tremor region. Our study suggests migrating tremors stimulate seismicity on both of the down-dip seismogenic zone and the shallower portion of the subduction zone near the trench.
====
今週のうなぎセミナーについてお知らせいたします。
Here is information of the Unagi-seminar(May 20).
************** Seminar on Seismology IV A,C /地震学ゼミナールIVＡ,C (Unagi Seminar) **************
科目：地震学ゼミナールIV A, C / Seminar on Seismology IV A, C（修士・博士）
日時：2021年 05月 20日 (木) 14:00~
場所：オンライン(Zoom)
Date and Time：2021-05-20, 14:00~
Place：Zoom
====
Presenter 1 (発表者 1)：
MPUANG Admore
Title (題目)：
Mass extinction in genetic algorithms applied to receiver functions inversion: method development
Abstract (要旨)：
The use of genetic algorithms in receiver functions inversion for crustal and uppermost mantle velocity-depth structure is well established. Their operation is based on the process of evolution of biological species, by using (pseudo)random numbers to control the selection, crossover and mutation processes in searching a model space for an optimal model. Despite their robustness, one drawback of the standard genetic algorithms is that towards the end of a run, only a few 'new ideas' are explored which may lead to the stagnation of the optimization process. This can be an especially major drawback for large model dimensions, such as in the inversion of receiver functions for dipping and anisotropic structures. To avoid this problem, mass-extinction, in which major parts of the explored model population are regularly replaced, is introduced to genetic algorithm inversion in this study to exploit highly fit models and random explorations of other domains. The concept of self-organized criticality is applied to control the size and frequency of the extinction events. Different configurations and model replacement strategies are tested against the standard genetic algorithm (Shibutani et al., 1996) for performance comparison. Preliminary results show that while mass-extinction may slow down the convergence speed of optimization through increased exploration, the control parameters can be tuned for great performance that can outperform the standard genetic algorithm at no significant additional computational costs.
............................................................................................................................................................
Presenter 2 (発表者 2)：
Shukei Ohyanagi
Title (題目)：
A synchronized activity of offshore seismicity and shallow tremor revealed by application of deep learning module to an OBS dataset
Abstract (要旨)：
The Japan Trench subduction zone is a cradle for various types of seismic activities, such as Mw 9.0 Tohoku-Oki megathrust event, VLFE, or SSE. In past few years, ocean bottom seismic observations by S-Net (cable network of OBSs) and other dense networks of free-fall type OBS reveal astonishing slow earthquake activities of this subduction zone. Unlike depth-varying transition of a megathrust earthquake fault to a slow earthquake fault in a Nankai subduction zone (e.g. Kato and Obara, 2016), slow earthquake regions illuminated by tectonic tremors transit to a megathrust earthquake region in along strike direction in the shallow subduction margin [Nishikawa et al., 2019; Tanaka et al., 2019; Ohta et al. 2019]. However, although acquired OBS datasets have used to search for tectonic tremors, they have not fully utilized to understand microseismicity of the same region. They are often used to relocate earthquakes cataloged in the JMA unified catalog which developed with recordings of onshore seismic network [e.g. Shinohara et al., 2012], but attempt to detect microseismicity solely visible from offshore seismic network is uncommon. As Obana et al. (2021) identify ordinal earthquakes that occur simultaneously with tectonic tremors based on a dense OBS observation, a microseismicity catalog developed from an OBS data would play a key-role on understanding interplay between ordinal earthquakes and shallow tremors.
During 2016 to 2017, we performed OBS Array of Arrays observation in offshore Fukushima region near the trench. This dataset reveals shallow tremor activity in this region [Ohta et al., 2019]. To investigate ordinal seismicity during the tremor activity, we applied Earthquake Transformer (EQT) [Mousavi et al., 2020], the newly developed deep learning module to perform detection and phase picking of microearthquakes to the OBS dataset. As a result, EQT detects more than 5000 microearthquakes. The detected events are located using hypomh [Hirata and Matsu’ura, 1987] with station corrections estimated from PS converted phases [Iwasaki et al., 1991]. The developed catalog reveals dense microseismicity in the down-dip side of a shallow tectonic tremor region identified by Nishikawa et al. (2019) and Ohta et al., (2019). A synchronized activity of ordinal earthquakes and shallow tremors are observed in the down-dip side of the tremor region, while Obana et al. (2021) reports such synchronization in the up-dip side of the tremor region. Our study suggests migrating tremors stimulate seismicity on both of the down-dip seismogenic zone and the shallower portion of the subduction zone near the trench.
====
