Study on spatial correlation of ground-motion: a case study of Napa earthquake

The probability of joint exceedance of ground motion intensities at multiple sites during the same earthquake needs to be quantified in aggregated seismic hazard analysis or loss assessment of spatially distributed infrastructure systems. The spatial correlation of ground motion at multiple sites cannot be considered in conventional specific-site seismic hazard analysis method. In this paper, the spatial correlation of ground motion is preliminarily studied using 344 sets of strong earthquake records from the MW6.0 Napa earthquake, California, US on August 24, 2014. The results of finite fault inversion for Napa earthquake(Dreger et al., 2015)is used as the surface projection of fault rupture. The ground motion attenuation relationship of Boore in NGA-West2 in the Pacific Seismic Engineering Research Center, which is referred to as BSSA14 in this paper, is selected in this research. The local site amplification effect of ground motion is obtained by using the correlation between topographic slope and VS30(average shear wave velocity of rock and soil layer from surface to 30 meters underground)to obtain the site amplification coefficient of amplitude and frequency. The ground motion parameters actually observed by the station are converted to the reference surface of bedrock by using the site amplification coefficient, and the fault projection distance between each station and the surface projection of fault rupture is calculated. Comparison of strong ground motion recordings of the Napa earthquake with BSSA14 in the Next Generation Attenuation(NGA)-West2 Ground-Motion Models, indicates that the ground motion of the high frequency components of the Napa earthquake was underestimated in the attenuation relationship of BSSA14. And the peak ground acceleration residuals are mostly negative. The residuals of acceleration response spectrum for the 3-second period are basically distributed around the 0 value. The spatial correlation functions of the geometric mean for the peak ground acceleration(PGA), the peak ground velocity(PGV)and spectral acceleration at three specific periods(0.3s, 1.0s and 3.0s)of the two horizontal components are derived using the method of semivariogram function. 2km and 1km of the distance between stations is used in this paper respectively, which can guarantee the reliability of data calculation statistics. However, we found that the decrease of the distance between stations did not significantly improve the statistical results of the spatial correlation of strong ground motion in the observed data of Napa earthquake. The corresponding continuous spatial correlation function is fitted with exponential model and compared with the past studies of ground-motion correlation, which has been widely researched in the past. The analysis results show that the ground motion parameters are spatially correlated, and these spatial correlations approximate exponential decay as the distance increases. Secondly, the spatial correlation of ground motion increases with the increase of response spectrum period. It may be because the similarity is reduced by the scattering of waves during propagation, and that this reduction is greater for high-frequency waves. The short-wavelength waves tend to be more affected or changed by small-scale heterogeneities in the process of propagation, so the spatial correlation of high-frequency seismic waves is smaller than that of long-period seismic waves. Finally, there is a regional characteristic in the spatial correlation of seismic ground motion. The spatial correlation of peak ground acceleration is weaker in southern California than in Japan and Taiwan. Results of this research in this paper can provide theoretical basis and reference for aggregated seismic hazard analysis or loss assessment. And uncertainty will be reduced in ShakeMap considering the spatial correlation of ground motion. © 2020, Editorial Office of Seismology and Geology. All right reserved.