Construction and validation of a simulated ground motion dataset for Turkey

Abstract

A dataset of simulated ground motions is created for seven recorded and previously validated, along with three hypothetical earthquakes in Turkey. This dataset has potential uses in engineering practice and research by both seismological and engineering communities. The simulated ground motion dataset with extensive information on the simulations and ground motion intensity parameters for each simulated motion is presented in an open-access online repository. A two-level randomization scheme is proposed to account for the uncertainties in input parameters and source-to-site geometries. An investigation of the magnitude-distance ranges in the simulated dataset, as well as the distribution of ground motion intensity measures, showed that the created dataset fills the gaps observed in recorded ground motion datasets. Pulse-like motions in the dataset are identified, and the relationship between pulse periods and earthquake magnitude is shown to agree with other relationships in the literature which are derived from recorded ground motions. The effects of source-to-site geometry and uncertainties in the following four input parameters are investigated: magnitude ($Mw)$, stress drop, ($\Delta \tau$), time-averaged shear-wave velocity in the upper 30 m (${V}_{S30}$), and high-frequency attenuation parameter (${\kappa }_{0}$). The dataset is validated by investigating the variability and inter-period correlation of normalized residual spectral acceleration values ($\epsilon )$, calculated using a ground motion model (GMM). The variability of $\epsilon$ is found to be consistent with the variability of GMMs. However, inter-period correlations of $\epsilon$ are shown to be larger than predictions of empirical models based on recorded earthquakes.