EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Combining Direct Simulation and Surrogate Modelling for Simulation‐Consistent Probabilistic Seismic Hazard Analysis

Claudio Schill, Brendon A. Bradley

Earthquake Spectra · 2026-11 · 42 (4)

DOI: 10.1002/esp4.70142 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Earthquake Spectra, online erschienen am 14.11.2026. Verfasst von Claudio Schill und 1 weiteren Autor:innen. Die Arbeit nennt 46 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: University of Canterbury, Marsden Fund.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Physics‐based ground‐motion (GM) simulations are being increasingly adopted in engineering practice. Nonetheless, computational requirements remain a hindrance to explicitly considering the exhaustive number of potential ruptures in typical probabilistic seismic hazard analysis (PSHA) calculations. We present a framework for simulation‐consistent PSHA that combines direct simulation of the ruptures contributing most strongly to the hazard with a surrogate model, trained on physics‐based ground‐motion simulations, to represent the much larger number of ruptures that have individually small, but collectively meaningful, contributions. We illustrate the approach through application to the CyberShake NZ dataset, which collectively includes explicit simulation of approximately 16,000 rupture realisations across New Zealand (NZ) for the fault‐based component of the hazard, while a surrogate model is used to quantify the contributions from the distributed seismicity component of the hazard. The surrogate model is based on a neural network that includes common predictor variables present in contemporary ergodic empirical ground‐motion models, as well as the addition of location dependence. The large number of simulations available for training simulation‐based surrogate models avoids issues with prior attempts to develop machine learning‐based ground motion models with only observational data, and the straightforward model training process means that surrogate models can be iteratively re‐trained as simulation methods and their constituent models and input data continue to evolve. A principal benefit of the use of simulation‐based surrogate models is achieving a unified simulation‐based treatment of the ground‐motion model component in PSHA (within typical computational constraints), in contrast to prior studies that have used simulations for the fault‐based sources but conventional empirical models for distributed seismicity sources.

Förderung

  • University of Canterbury
  • Marsden Fund

Zitieren

APA

Schill, C., & Bradley, B. A. (2026). Combining Direct Simulation and Surrogate Modelling for Simulation‐Consistent Probabilistic Seismic Hazard Analysis. Earthquake Spectra, 42(4). https://doi.org/10.1002/esp4.70142

Vancouver

Schill C, Bradley BA. Combining Direct Simulation and Surrogate Modelling for Simulation‐Consistent Probabilistic Seismic Hazard Analysis. Earthquake Spectra. 2026;42(4). doi:10.1002/esp4.70142

BibTeX
@article{Schill2026,
  author = {Schill, Claudio and Bradley, Brendon A.},
  title = {Combining Direct Simulation and Surrogate Modelling for Simulation‐Consistent Probabilistic Seismic Hazard Analysis},
  journal = {Earthquake Spectra},
  year = {2026},
  volume = {42},
  number = {4},
  doi = {10.1002/esp4.70142}
}
Kostenlos starten →