Preliminary Validation of a Spectral-Based Stochastic Ground Motion Model with a Non-Parametric Time-Modulating Function

Abstract

This study presents a site-based parameterized stochastic model for simulation of far-field synthetic ground motions. The model employs a modulated and filtered white-noise process defined via spectral representation. The modulating function is a recently proposed non-parametric function based on a monotonic cubic spline interpolation. As for the time-frequency modulating function, two slightly different versions are explored. The two versions of the model are fitted to a catalog of recorded ground motions and synthetic catalogs are generated using the fitted model parameters. To validate the model, some characteristics of the synthetic catalogs, namely the median, logarithmic standard deviations, and correlations of the elastic response spectra, are compared with those of the recorded catalog. These comparisons show that both versions of the model are able to adequately capture the spectral amplitudes, variability and correlations of recorded ground motions. The addition of a parameter to describe the rate of change of bandwidth with time does not result in any noticeable improvement and is therefore not warranted. Moreover, comparison with synthetic motions generated from the model by Rezaeian and Der Kiureghian (2010) shows that the proposed model results in an improved estimation of the correlations. Further studies are required to identify which feature(s) of our model are behind this improvement.