Enhancement and Validation of Ground Motion Simulations

Accurate prediction of strong ground motion is central to seismic hazard analysis in order to estimate losses during major earthquakes. Ground motion simulations are essential to seismic ground motion prediction, especially for locations of infrequent observations, such as large magnitude and short distance events, where simulations can provide a viable alternative to data. Therefore, enhancement and validation of ground motion simulations, the primary goal of this dissertation, are highly desirable. In Chapter 2, we quantify the effects of four important factors on ground motions from large normal-faulting earthquakes on the Wasatch fault in the Salt Lake Basin: rupture direction, location on the hanging wall versus the footwall, deep 3D basin structure, and the distance from the rupture in the near field range. In Chapter 3, we attempt to validate the presence of several proposed waveguides in the Los Angeles area using 3D simulations and observed data from ambient noise. Here, we compare the numerical and empirical surface-to-surface Green tensors for virtual sources located on the San Andreas Fault. The regions of large peak motions caused by waveguide focusing in the simulations show generally good agreement with increases in the Green tensor amplitudes, supporting the presence of two separate waveguides in greater Los Angeles. In Chapters 4 and 5, we develop an empirical frequency-dependent spatial ground motion correlation model and methods to rectify simulation techniques that otherwise produce synthetic time histories deficient in inter-frequency and spatial correlation structure. The methods are tested using a hybrid deterministic-stochastic broadband ground motion generation module, where our method reproduces the empirical correlations well for a large number of realizations without biasing the fit of the median of the spectral accelerations to data. We find that the best fit of the inter-frequency correlation to data is obtained assuming that the horizontal components are correlated with a correlation coefficient of about 0.7