Published July 10, 2025 | Version 2.0
Report Open

Exploring the SW4 Synthetic Seismic Performance of the ARCH Rockfish Cluster

Authors/Creators

  • 1. ROR icon Southern California Earthquake Center
  • 2. ROR icon University of Nevada, Reno
  • 3. ROR icon Victoria University of Wellington
  • 4. Terēan

Description

A preliminary research report:

Exploring the SW4 Synthetic Seismic Performance of the ARCH Rockfish Cluster 

MORGAN NEWTON, Santa Monica College, SCEC SOURCES, Univ. of Southern California, morganoliviaevans@gmail.com 

JOHN LOUIE, Nevada Seismological Laboratory, University of Nevada Reno, louie@unr.edu 

TIM STERN, Victoria University of Wellington, New Zealand, tim.stern@vuw.ac.nz 

AASHA PANCHA, Aurecon, Wellington, New Zealand, AASHA.PANCHA@AURECONGROUP.COM 

20 March 2023 

For seismic research, high-performance computing enables us to perform earthquake simulations at higher source frequencies and lower minimum shear-wave velocity parameters than on desktop or laptop systems, delivering higher-accuracy seismic ground motion estimates. We initially computed 27 low-resolution, low-frequency (<0.6 Hz) 3D shaking models using SW4 2.01 software and a MacBook Pro. Eleven high-resolution scenarios were computed using the ARCH Rockfish supercomputing cluster located at the John Hopkins High Performance Computing Center. We compared low-resolution MacBook Pro simulations to both low- and high-resolution Rockfish simulations, and analyzed Rockfish’s overall performance using SW4. We found that low-resolution simulations were slightly faster on Rockfish. We achieved high efficiency for the high-resolution simulations on Rockfish; for all configurations tested, efficiency was above 99%. High-resolution scenarios on Rockfish show a substantial improvement in detail, with prominent basin-edge amplifications. Notably, in high-resolution scenarios we see wave resonance within the narrow Wainuiomata Valley, an effect not visible in the low-resolution models. Scenarios built using Rockfish showed a wide variety of performance, highly dependent on Slurm job directive values. With just a portion of our CPU allocation, we were able to successfully compute a non-ergodic set of Wellington shaking scenarios valid from 0.2 to 1.5 Hz. These scenarios will provide the city with basin-amplification maps and spectra. Results enable increased accuracy for seismic analysis, improving seismic hazard estimates and preparedness for expected seismic events within the Wellington region. This work also provides a starting point for any subsequent seismology research that may choose to utilize high-performance computing. 

CCS CONCEPTS • Parallel algorithms • Modeling and simulation • Earth and atmospheric sciences 

Additional Keywords and Phrases: high-performance computing, seismic modeling, parallel performance analysis, ground motion simulation 

Files

Exploring the SW4 Synthetic Seismic Performance of ARCH Rockfish Cluster-Zenodo.pdf

Files (470.8 kB)

Additional details

Funding

U.S. National Science Foundation
REU Site: SCEC Undergraduate Studies in Earthquake Information Technology (SCEC/UseIT) 1005235

Dates

Submitted
2023-03-20
Originally submitted to the 2023 PEARC Conference
Available
2025-07-10
Uploaded to Zenodo

References

  • Rafael Benites, Kim B. Olsen (2005). Modeling Strong Ground Motion in the Wellington Metropolitan Area, New Zealand. Bulletin of the Seismological Society of America; 95 (6): 2180–2196. doi: https://doi.org/10.1785/0120040223
  • Eric Eckert, Michelle Scalise, John N. Louie, and Kenneth D. Smith (2022). Exploring basin amplification within the Reno metropolitan area in Northern Nevada using a magnitude 6.3 ShakeOut scenario: Bulletin of the Seismological Society of America, 112(1), 457-473, doi: 10.1785/0120200309
  • Gold, D. (2021, June 7). Scaling Experiments: How to Measure the Performance of Parallel Code on HPC Systems. Water Programming: A Collaborative Research Blog. Retrieved from: https://waterprogramming.wordpress.com/2021/06/07/ scaling-experiments-how-to-measure-the-performance-of-parallel-code-on-hpc-systems/
  • Newton, M. O., Louie, J. N., Stern, T., & Pancha, A. (2022, 09). Low Frequency Non-Ergodic Synthetic Modeling of Earthquake Basin Effects in Wellington, New Zealand. Poster Presentation at 2022 SCEC Annual Meeting. SCEC Contribution 12218
  • Petersson, N.A.; Sjögreen, B. (2017). SW4, version 2.01 [software], Computational Infrastructure of Geodynamics, doi: 10.5281/zenodo.1063644, url: https://doi.org/10.5281/zenodo.1063644
  • Petersson, N.A.; Sjögreen, B. (2017), User's guide to SW4, version 2.0, LLNL-SM-741439 (LLNL-SM-741439)
  • Princeton Research Computing, (n.d.). Choosing the Number of Nodes, CPU-cores and GPUs | Princeton Research Computing. Princeton University. Retrieved from: https://researchcomputing.princeton.edu/support/knowledge-base/scaling-analysis
  • Alistair Stronach, Tim Stern (2021). A new basin depth map of the fault-bound Wellington CBD based on residual gravity anomalies, New Zealand Journal of Geology and Geophysics, doi: 10.1080/00288306.2021.2000438