Abstract
Regional seismic loss and engineering applications generally require the simulation of spatially distributed ground motions using multiple intensity measures (IMs), which can be described by spatial correlations. In this study, a geostatistical analysis is conducted to obtain spatial correlations for peak ground acceleration and spectral acceleration using more than 2000 measured recordings from eight earthquake events that occurred in California recently. In general, the spatial correlations of IMs are higher under uniform site conditions than in areas with varying site conditions. Considering the effects of site conditions on spatial correlations, we establish a predictive equation linking spatial correlation range and Vs30Vs30 values. This facilitates the estimation of spatial correlation under varying site conditions in regions with limited observational records. Subsequently, we directly obtain the spatial correlations based on geological information. In addition, considering the high similarity of IMs within a small scale of less than 1km, the spatial correlations are replaced by lagged coherence at a small separation distance instead of IMs to describe the variation in the Fourier phase. The spatial variation in any region can be described by combining coherence and spatial correlations.