We performed 3-D nonlinear-viscoelastic ground motion simulations using a finite difference modeling method in a frequency range of 0.1 to 1 Hz in the Kinburn basin, Ottawa, Canada, for large earthquakes. Comparing the records and simulated velocity time series showed that regular viscoelastic simulations could model the ground motions at the rock and soil sites in the Kinburn basin for the Ladysmith earthquake (Mw=4.7). Using nonlinear-viscoelastic ground motion simulations for the scaled Ladysmith earthquake (Mw=7.5) significantly reduced the amplitude of the horizontal components of the Fourier spectrum and the predicted PGA and PGV values compared to regular linear viscoelastic simulations. Further, using a finite fault source (Mw=7) for the nonlinear-viscoelastic simulation decreased PGAs of the horizontal components. Our sensitivity analysis of simulations for different seismic moments showed that the PGV values exponentially increased with moment magnitude. Using a Gaussian source function with a short half duration increased the PGVs and the amplitude of velocity Fourier spectrum. Relaxation times and relaxation coefficients for viscoelastic simulation significantly increased PGV, the amplitude of the PSA ratio, and the velocity Fourier spectrum for a small earthquake. Employing a small soil Q model reduced PGV, PSA of soil/rock ratios, and the amplitude of velocity Fourier spectrum. Using finite fault model for a large earthquake (Mw =7) significantly reduced the PGV values relative to a point source model.
We increased the maximum frequency to 2.5 Hz in FD modeling using a dual grid size method for two basins (Kinburn and Orleans basins in Ottawa, Canada). The simulated velocity time series from the dual grid size method provided better results compared to the results of the single grid size, although there were large differences between the amplitude of the velocity Fourier spectrum of the simulations and the amplitude of the records, particularly at low frequencies (<1 Hz). However, the velocity Fourier spectrums of the records and simulations showed a similar variation versus frequency. Further, using the dual grid size method increased the PGVs and amplitudes of the velocity Fourier spectrum for both the Kinburn and Orleans basins compared to the results of the single grid size simulation.