Intra-event spatial correlation model for the vertical component of response spectral accelerations

Abstract

Seismic risk assessment of lifeline networks or building portfolios requires the quantification of spatial correlation of ground motion intensity over a region during an earthquake event. In this paper, the spatial correlation of vertical spectral accelerations (SAs) at single and multiple periods is studied. In the first part, the spatial correlations of vertical SAs at eight periods in the range of 0.0–3.0 s are investigated. The geostatistical analysis is performed to compute the spatial correlations of SAs using records compiled from ten earthquake events occurred in California, Japan, Taiwan, and Mexico. The results show that the spatial correlations of the vertical SAs at short periods are dependent on the regional site conditions indicated with the spatial correlation of shear wave velocity in the top of 30 m (Vs30). However, the effect of the correlation of Vs30 on the correlation of the vertical SAs decreases as the spectral period increases. A simple predictive model is proposed to quantify the correlation ranges of the vertical SAs considering regional site conditions. The proposed model is compared with the models presented for horizontal SAs. The comparison shows that the correlation ranges of horizontal SAs are generally larger than those of the vertical ones. As an illustrative example, seismic performance of a hypothetical bridge network is studied with different ranges of spatial correlations. The results indicate that ignoring spatial correlation may significantly affect the exceedance probability for spatially distributed engineering demand parameters of bridges in the network. In the second part, the multivariate spatial correlation is investigated using linear model of coregionalization (LMC) approach. Based on the correlation estimates obtained from geostatistical analysis, a predictive model is proposed to simulate vertical SAs at multiple periods and different sites.