Topics

Topics

うなぎセミナー 10/4

うなぎセミナー 10/4

セミナー等

SEMINARS

更新日:2024.10.01

Updated: 2024.10.01

  • 開催場所:京都大学 防災研究所 本館E-232D
  • Place: 京都大学 防災研究所 本館E-232D
  • 開催日時:2024年10月3日(木) 13時30分~
  • Date and Time: 2024年10月3日(木) 13時30分~

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(October, 03).


************** Seminar on Seismology IV B, D /地震学ゼミナールIV B, D (Unagi Seminar) **************

科目:地震学ゼミナールIV B, D / Seminar on Seismology IV B, D(修士・博士)
日時:2024年 10月 3日 (木) 13:30~
場所:京都大学 防災研究所 本館E-232D

Date and Time:2024-10-03, 13:30~
Place:Uji Campus Main Building E232D

---------- ---------- ---------- ----------

1.Part 1 : Seminar Guidance

* * * * * * * * * * * * * *

2. part 2:
Speaker(発表者): Yanhan Chen

Title(題目):
New Insight of Low Frequency Earthquakes (LFEs) at Guerrero Seismic Gap based on matched filter technique and deep learning method

Abstract(要旨):
The Guerrero seismic gap in Mexico subduction exhibits a slip behavior distinct from that of its adjacent segments, which typically experience significant ruptures during large earthquakes. With the acquisition of offshore seismic data in this region and discovery of shallow tectonic tremors, the study of slow earthquakes has gradually increased. It is widely accepted that tectonic tremors comprise a series of LFEs. We use matched filter technique to detect LFEs by constructing templates within the tremors. By detecting LFEs, we can better constrain the depth of tremors, thereby providing improved insights into the location and extent of slow earthquakes along the interface. As a result, we can further elucidate a plausible mechanism responsible for slow slip.
With the successful application of deep learning in seismic phase picking, it has been gradually extended to the study of slow earthquakes in recent years. The use of deep learning to detect LFEs overcomes some of the inherent shortcomings of matched filter methods. Currently, there are few examples of using deep learning to detect LFEs. The two studies successfully detected LFEs, but the problems of insufficient data volume and single station training lead to low accuracy. Therefore, I collected catalogs and waveforms of LFEs from major subduction zones around the world, and from 400,000 original catalogs, I select about 60,000 LFE events with better quality as training samples, and labeled all the events by using the array format.

---------- ---------- ---------- ----------

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(October, 03).


************** Seminar on Seismology IV B, D /地震学ゼミナールIV B, D (Unagi Seminar) **************

科目:地震学ゼミナールIV B, D / Seminar on Seismology IV B, D(修士・博士)
日時:2024年 10月 3日 (木) 13:30~
場所:京都大学 防災研究所 本館E-232D

Date and Time:2024-10-03, 13:30~
Place:Uji Campus Main Building E232D

---------- ---------- ---------- ----------

1.Part 1 : Seminar Guidance

* * * * * * * * * * * * * *

2. part 2:
Speaker(発表者): Yanhan Chen

Title(題目):
New Insight of Low Frequency Earthquakes (LFEs) at Guerrero Seismic Gap based on matched filter technique and deep learning method

Abstract(要旨):
The Guerrero seismic gap in Mexico subduction exhibits a slip behavior distinct from that of its adjacent segments, which typically experience significant ruptures during large earthquakes. With the acquisition of offshore seismic data in this region and discovery of shallow tectonic tremors, the study of slow earthquakes has gradually increased. It is widely accepted that tectonic tremors comprise a series of LFEs. We use matched filter technique to detect LFEs by constructing templates within the tremors. By detecting LFEs, we can better constrain the depth of tremors, thereby providing improved insights into the location and extent of slow earthquakes along the interface. As a result, we can further elucidate a plausible mechanism responsible for slow slip.
With the successful application of deep learning in seismic phase picking, it has been gradually extended to the study of slow earthquakes in recent years. The use of deep learning to detect LFEs overcomes some of the inherent shortcomings of matched filter methods. Currently, there are few examples of using deep learning to detect LFEs. The two studies successfully detected LFEs, but the problems of insufficient data volume and single station training lead to low accuracy. Therefore, I collected catalogs and waveforms of LFEs from major subduction zones around the world, and from 400,000 original catalogs, I select about 60,000 LFE events with better quality as training samples, and labeled all the events by using the array format.

---------- ---------- ---------- ----------

参考ファイル:

Related files:

トピック一覧に戻る

Back to Topic List

© Research Center for Earthquake Hazards.

© Research Center for Earthquake Hazards.