Prediction of earthquake ground motion at rock sites in Japan: evaluation of empirical and stochastic approaches for the PEGASOS Refinement Project

Edwards, Benjamin ORCID: 0000-0001-5648-8015 and Fah, Donat
(2017) Prediction of earthquake ground motion at rock sites in Japan: evaluation of empirical and stochastic approaches for the PEGASOS Refinement Project. GEOPHYSICAL JOURNAL INTERNATIONAL, 211 (2). pp. 766-783.

[img] Text
ggx328.pdf - Published version

Download (10MB)


Strong ground-motion databases used to develop ground-motion prediction equations (GMPEs) and calibrate stochastic simulation models generally include relatively few recordings on what can be considered as engineering rock or hard rock. Ground-motion predictions for such sites are therefore susceptible to uncertainty and bias, which can then propagate into sitespecific hazard and risk estimates. In order to explore this issuewe present a study investigating the prediction of ground motion at rock sites in Japan, where a wide range of recording-site types (from soil to very hard rock) are available for analysis. We employ two approaches: empirical GMPEs and stochastic simulations. The study is undertaken in the context of the PEGASOS Refinement Project (PRP), a Senior Seismic Hazard Analysis Committee (SSHAC) Level 4 probabilistic seismic hazard analysis of Swiss nuclear power plants, commissioned by swissnuclear and running from 2008 to 2013. In order to reduce the impact of site-to-site variability and expand the available data set for rock and hard-rock sites we adjusted Japanese ground-motion data (recorded at sites with 110 m s-1 < Vs30 < 2100 m s-1) to a common hard-rock reference. This was done through deconvolution of: (i) empirically derived amplification functions and (ii) the theoretical 1-D SH amplification between the bedrock and surface. Initial comparison of a Japanese GMPE's predictions with data recorded at rock and hard-rock sites showed systematic overestimation of ground motion. A further investigation of five global GMPEs' prediction residuals as a function of quarter-wavelength velocity showed that they all presented systematic misfit trends, leading to overestimation of median ground motions at rock and hard-rock sites in Japan. In an alternative approach, a stochastic simulation method was tested, allowing the direct incorporation of site-specific Fourier amplification information in forward simulations. We use an adjusted version of the model developed for Switzerland during the PRP. The median simulation prediction at true rock and hard-rock sites (Vs30 > 800 m s-1) was found to be comparable (within expected levels of epistemic uncertainty) to predictions using an empirical GMPE, with reduced residual misfit. As expected, due to including site-specific information in the simulations, the reduction in misfit could be isolated to a reduction in the site-related within-event uncertainty. The results of this study support the use of finite or pseudo-finite fault stochastic simulation methods in estimating strong ground motions in regions of weak and moderate seismicity, such as central and northern Europe. Furthermore, it indicates that weak-motion data has the potential to allow estimation of betweenand within-site variability in ground motion, which is a critical issue in site-specific seismic hazard analysis, particularly for safety critical structures.

Item Type: Article
Uncontrolled Keywords: Earthquake ground motions, Earthquake hazards, Seismic attenuation, Site effects
Depositing User: Symplectic Admin
Date Deposited: 11 Oct 2017 06:46
Last Modified: 19 Jan 2023 06:53
DOI: 10.1093/gji/ggx328
Related URLs: