Modelling earthquake ground motions in California using parametric time series methods

Abstract The use of parametric time series analysis to describe strong motion accelerograms has proven to be an efficient method requiring few modelling parameters. In this study, an ARMA model was used to represent accelerograms which were first processed using a variance-stabilizing transformation. By analyzing a large number of recorded accelerograms in the State of California, relationships were found between ARMA parameters and physical variables making it possible to simulate earthquake ground motions throughout the region. The validity of these results was tested by comparing the accelerograms and response spectra of the initial and simulated records.