Japan Geoscience Union Meeting 2023

Presentation information

[J] Online Poster

S (Solid Earth Sciences ) » S-TT Technology & Techniques

[S-TT43] Creating future of solid Earth science with high performance computing (HPC)

Tue. May 23, 2023 10:45 AM - 12:15 PM Online Poster Zoom Room (17) (Online Poster)

convener:Takane Hori(Japan Agency for Marine-Earth Science and Technology), Yuji Yagi(Graduate School of Life and Environmental Sciences, University of Tsukuba, Tsukuba), Katsuhiko Shiomi(National Research Institute for Earth Science and Disaster Resilience), Takanori Matsuzawa(National Research Institute for Earth Science and Disaster Resilience)

On-site poster schedule(2023/5/22 17:15-18:45)

10:45 AM - 12:15 PM

[STT43-P03] Accuracy Evaluation for Finite Element Calculation of Long-period Ground Motions in the Sagami Trough Megathrust Earthquake

*yoshihisa HIYOSHI1, Ryoichiro Agata1, sayako hirobe1, Takane Hori1 (1.Japan Agency for Marine-Earth Science and Technology)

Keywords:long-period ground motion, Finite element method, Finite difference method

Numerical simulations of seismic ground motions considering the complex three-dimensional underground structure are required for detailed damage estimation in buildings and structures in urban areas due to long-period seismic ground motions caused by subduction zone megathrust earthquake. The finite difference method (FDM) is widely used for such calculations because of its relatively simple algorithm of modeling and calculation. Although the finite element method (FEM) can introduce surface topography and layer boundary geometry more flexibly and accurately, it has been considered to be computationally demanding to introduce FEM to such calculation because of high computation cost including preprocessing such as mesh generation. On the other hand, recent research and developments, such as the large-scale FEM calculation program “E-wave FEM” (Ichimura et al., 2009, 2014, 2015), have allowed for automatic generation of unstructured finite element meshes and accelerated calculation of seismic ground motion using massively parallel supercomputers such as “Fugaku”. Because of its high practical utility, E-wave FEM has been introduced to damage estimation due to long-period ground motions conducted by a national agency. It is important to confirm the level of numerical errors included in the calculation results used in such a damage estimation in order to ensure its reliability. In this study, we evaluate the accuracy of E-wave FEM results obtained in the same setup of long-period ground motion calculation used in damage estimation conducted by a national agency, focusing on the mesh size of the unstructured mesh.
In the target E-wave FEM calculation setup, a source fault model consisting of four asperities is placed in a calculation domain of 182 km in the east-west direction, 168 km in the north-south direction, and 140 km in the vertical direction, which includes the Kanto Plain. This setup models the Sagami Trough megathrust earthquake. The seismic wave velocity structure is set as a three-dimensional heterogeneous velocity structure consisting of 41 layers, each of which has uniform seismic wave velocity, based on the J-SHIS database of the National Research Institute for Earth Science and Disaster Prevention. The finite element mesh composed of unstructured tetrahedral elements were generated for this domain to meet the requirement of at least five elements per wavelength. Seismic ground motions were calculated for a frequency band up to 0.5 Hz with a time step length of 0.02 second x 15000 steps for a period of 300 seconds. In order to evaluate the setup of the mesh size in this calculation, we newly generated a finer unstructured mesh, which meets the requirement of at least of 10 elements per wavelength. Using this finer mesh and keeping other conditions the same, we performed long-period seismic motion calculations for comparison. The computation time was about 7 hours using 1024 computation nodes of "Fugaku".
We compared the pseudo-velocity response spectra of the above two calculations of difference mesh sizes for multiple evaluation points in the target area. The differences were generally low at the level of a few percent, suggesting that the mesh size of the original setup was small enough. In the presentation at the conference, we will also show the cross sections of subsurface velocity structure between the evaluation points and source fault models. We then discuss the relationship the cross sections with the comparison results.

Acknowledgments: This work was supported by MEXT as “Program for Promoting Researches on the Supercomputer Fugaku” (Large-scale numerical simulation of earthquake generation, wave propagation and soil amplification, Project ID: hp200126, hp210171, hp220171) and used computational resources of the Earth Simulator provided by the Japan Agency for Marine-Earth Science and Technology. This work used the E-wave FEM code which was originally provided by the Earthquake Research Institute, The University of Tokyo and modified by JAMSTEC.