Uncertainty of site period on the variability of the site period-based ground motion prediction equations

Abstract

Site period has been widely acknowledged to be a reasonable site parameter for ground-motion modeling and site effect modeling. Its value could be analytically calculated using the shear-wave velocity (VS) profile by assuming that soil layers overlie a rigid bedrock, or empirically estimated from the horizontal-to-vertical spectral ratio (HVSR). In this study, we tested the uncertainty of site period on the variability of the site period-based ground-motion prediction equations (GMPEs). The ground-motion records and site data in Japan were used. We first used different VS values as engineering bedrock to derive a set of analytical site periods, and tested the uncertainty because of the bedrock VS selection on the GMPE variability. The results show that adding information about the stiff-soil or soft-rock layer beneath the surface soft-soil layers would significantly improve site effect prediction accuracy. An engineering bedrock with VS equal to about 800 m/s is recommended. We then tested the uncertainty of site period because of the choice between the analytical one and the empirical one on the GMPE variability. The empirical site period is better than the analytical site period at spectral periods from 0.06s to 0.4s. At spectral periods longer than 0.4s, the empirical site period produces larger variability than the analytical one using VS ≥ 800 m/s bedrock and has a better performance than that using VS < 600 m/s bedrock. The larger variability is partly contributed by the uncertainty in identifying site period value from the HVSR curve. Using the site period from the first peak of the HVSR curve generally produces a smaller variability than that from the highest peak. For sites with more than one comparable peak in the HVSR curve, the site period from the first peak is recommended. For the regions out of Japan, cautions may be paid for these recommendations because of the potential region dependence of site response.