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, 14th).

************** Seminar on Seismology IV A, C /地震学ゼミナールIV A, 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：2026年 5月 14日 (木) 13:30~
場所：京都大学 防災研究所 本館E-232D

Date and Time：2026-05-14(Thursday), 13:30~
Place：Uji Campus Main Building E232D

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

Speaker 1（発表者）:　小出　鯉太朗（Koitaro Koide）
Title（題目）:
Characteristics of displacement fields due to fault motion in an elastic-viscoelastic layered half-space and preliminary development of PINNs toward crustal deformation modeling

Abstract（要旨）:
The displacement field due to a fault motion is often difficult to understand intuitively. In the case of an infinite elastic medium, the displacement field is symmetrical to the fault, which can be easily understood by double-couple force system. However, the surface displacement field caused by a reverse fault in an elastic half-space shows an asymmetrical pattern. This behavior can be understood by considering the influence of the free surface. When viscoelastic medium is assumed, displacement field will be more complex.
In this study, I compute the vertical displacements resulting from reverse faults in the shallow and deep part of the elastic layer. When a reverse fault is put in the shallow part, the resultant viscoelastic relaxation is downward displacements around the fault. In contrast, when a reverse fault is put in the deep part of the elastic layer, upward displacements occur around the fault. Generally, the elastic layer behaves like a plate after viscoelastic relaxation because the viscoelastic layer behaves like a fluid. When a reverse fault is put only in the either the shallow or deep part of the elastic layer, unbalanced shortening occurs, which causes the elastic plate to bend, resulting in the deformation field mentioned above.
I apply this result to landscape development in Echigo Plain. Echigo Plain is the largest plain in the Japan Sea side with very thick sedimentary layer. In addition, Echigo Plain is located in Niigata-Kobe Tectonic Zone (Sagiya et al, 2000), which is significant compression field. Therefore, it is difficult to understand why the subsidence field is formed.
Niitsu Hills is located at the center of Echigo Plain with west dipping-fault on the east(e.g. Sato, 2013). When I assumed elastic half-space, uplift occurs dominantly, and it is different from landscape. When I assumed elastic-viscoelastic layered half-space, subsidence occurs, which is better fit to the landscape than the case of the elastic medium. However, it is difficult to explain overall pattern of Echigo Plain, which can be explained considering heterogeneity.
Physics-Informed Neural Networks (PINNs; Raissi et al., 2019) can be strong tool to calculate the displacement field considering heterogeneous medium. PINNs is a type of deep learning using partial differential equations for training. Okazaki et al. (2022, 2024, 2025) developed PINNs program calculating the displacement fields due to a fault motion in the elastic half-space. My research plan for doctoral studies is to compute deformation fields due to fault motion in elastic-viscoelastic layered system by PINNs. There are some strengths of using PINNs; easily and rapidly calculating the deformation field considering heterogeneous medium and topography, not to make mesh and not to set boundary conditions at the end of calculating area.
As the first step, I solved some partial differential equations, such as load deformation of viscoelastic medium, using PINNs. I will present the method of PINNs and results of solving PDE.

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

Speaker 2（発表者）:　福田　公平（Kohei Fukuda）
Title（題目）:
Ambient Noise Detection through Array Processing with a Dense Seismic Network

Abstract（要旨）:
Seismic waves (especially surface waves) are generated by various phenomena. Analysis of wave sources, magnitude, and frequency of these events is important for disaster prevention, improving the accuracy of observed seismic waveforms, and understanding ecosystems. AELUMA method can detect the epicenter of such events of ambient noise by using array analysis for surface waves. It is reported that AELUMA detected stormquake, landslide, and tsunami event of submarine volcano.
In this study, we aim to automate and enhance the accuracy of epicenter determination using the AELUMA method, with the ultimate goal of detecting submarine phenomena such as landslides. This presentation details specific improvements made to the methodology, motivated by challenges identified during our previous analysis of earthquakes near the Tokara Islands. We have worked on optimizing the grid search process to improve source localization, and have also incorporated an STA/LTA-based triggering system to enhance detection sensitivity. We will discuss how these refinements resolve previous issues and contribute to more robust seismic analysis.