Key Findings on the Role of Corner Frequency in Site-Based Stochastic Ground Motion Models

Site-based stochastic ground motion models (GMMs) are essential for representing seismic input in uncertainty quantification methods for earthquake engineering. A key element controlling the long-period spectral behavior of these models is the high-pass filter’s corner frequency (fc), which must thus be included as a free parameter. In fact, our previous study demonstrated the need to optimize fc to effectively control the long-period response spectra. This paper builds on and complements our previous study by presenting three key findings. First, for site-based stochastic GMMs, not only fc’s values but also fc’s correlation with other GM parameters (e.g., Arias intensity, predominant frequency, and duration) significantly impact the correlation of linear-response spectra Sa(T) at different spectral periods. Second, using a modulated filtered-white-noise GMM with 11 parameters, including the optimized fc, we effectively fit a large data set comprising 3,243 strong motion records. Specifically, a visual classification indicates that circa 93% of the recorded spectra Sa(T) closely fall within the 2-sigma confidence bounds of the simulated S^a(T). Third, the fitted fc values contain unrealistic information when the records downloaded from strong-motion databases are overfiltered to remove signal noises.