Empirical nonlinear site effects in Japanese megathrust earthquakes

In the course of developing ground-motion models (GMMs) of peak ground acceleration (PGA) for Japanese megathrust earthquakes, we discovered that empirical nonlinear site effects from these events are very different from those found for shallow crustal earthquakes in California and other global active tectonic regions. In particular, nonlinearity in site response is much less prominent in Japan than these other regions, which is consistent with other empirical and theoretical ground-response studies using recordings from both crustal and subduction earthquakes. In this study, we used the residuals from Japanese megathrust GMMs that account for linear site response to develop a nonlinear site-response term that uses the value of PGA on rock as a measure of ground-motion severity and the value of time-averaged shear-wave velocity in the top 30 m of a site ( V S 30 ) as a measure of site properties. The resulting residuals are unbiased and trendless over all PGA values of interest. We also investigated whether soil nonlinearity can be modeled using proxies for shear stress and shear strain. We found deviations from expected linearity unrelated to nonlinear site effects at small values of strain—believed to be due to network accelerograph trigger thresholds—and at large earthquake magnitudes, believed to be due to the shape of the source spectrum.