Spatial correlations of ground-motion intensity measures (IMs) are essential for seismic analysis of spatially distributed systems. In this paper, geostatistical analysis is conducted to calculate the spatial correlations for cumulative absolute velocity (CAV), Arias intensity (Ia), and spectral accelerations (SA) using a total number of more than 1500 earthquake records from nine recent earthquakes occurred in Taiwan, California, and Japan. The results indicate that the spatial correlations for these IMs are closely related to the regional site conditions, and they can be predicted based on the spatial correlations of shear-wave velocity in the top 30 m (V S30). In general, an IM recorded from a relatively homogeneous regional site condition tends to have a larger spatial correlation range than that from a heterogeneous site condition. Due to their intrinsic similarity to represent the integration of acceleration time histories, CAV and Ia have similar spatial correlation coefficients. Besides, the range of spatial correlation of SA generally increases as the spectral period increases. Simple predictive equations are proposed in this study to quantify the spatial correlations of CAV, Ia, and SA based on regional site conditions. Methods for data correction are also proposed to eliminate artificial correlations due to biased distance scaling and V S30 estimation in the database. Finally, Monte Carlo method is used to generate spatially distributed IMs. The results demonstrate that the annual frequency of exceedance curves for spatially distributed IMs differ significantly if different ranges of spatial correlations are used.

This study investigates spatial cross-correlation models for two sets of vector intensity measures (IMs) considering the influence of regional site conditions. The first set of the vector IM consists of the peak ground acceleration, Arias intensity, and the peak ground velocity; the second set is for spectral accelerations at multiple periods. Geostatistics analyses are performed using 2686 strong-motion data from 11 recent earthquakes that occurred in California, Japan, Taiwan, and Mexico. The results indicate that the spatial cross correlations of the vector IMs are strongly influenced by the spatial distribution of regional site conditions, which can be quantified using R V S30 , the correlation range of shear-wave velocity in the top 30 m. The linear model of coregionalization is proposed to construct a permissible spatial correlation model, and the short-range and long-range coregionalization matrices is specified to vary linearly with R V S30 . The proposed model demonstrated excellent performance in quantifying the influence of regional site conditions on the spatial cross correlations for the vector IMs, meanwhile the model guarantees a positive-definite covariance matrix for any reasonable value of R V S30 , a mathematical condition required for stochastic generation of the spatially correlated random fields. The spatial cross-correlation models proposed in this study can be conveniently used in regional-specific seismic risk analysis and loss estimation of spatially distributed infrastructure using vector IMs.

No abstract