Ground-Motion Prediction Equations for the Vertical Component from Shallow Crustal and Upper-Mantle Earthquakes in Japan Using Site Classes as the Site Term

This study presents ground-motion prediction equations for the vertical component of the strong-motion records from shallow crustal and upper-mantle earthquakes in Japan. A large dataset from earthquakes with a reliable earthquake category (the tectonic locations of earthquakes) up to the end of 2012 was used. Following a previous study by Zhao, Zhou, et al. (2016), we used a bilinear magnitude scaling function hinged at a moment magnitude (M-w) of 7.1. The scaling rate for events with M-w > 7.1 is smaller than that for the other events. We find that the magnitude scaling rate for large events increases with increasing spectral periods at periods above 0.1 s, in sharp contrast to the magnitude scaling rate for the horizontal components. We modeled the effect of volcanic zones in the same manner as in the Zhao, Zhou, et al. (2016) study. This study uses site classes (SCs) based on site period computed from shear-wave velocity profile and the engineering bedrock depth as the site terms without accounting for nonlinear site effects. We found that the site information quality makes a significant difference for all spectral periods; that is, the exclusion of records from sites with inferred SCs improves the model fit significantly. Similar to a previous study for the horizontal ground motions in Japan, we also found that the normal events in both shallow crustal and upper-mantle categories produce significantly larger spectrum than the strike-slip or reverse-faulting events; this is not consistent with many other studies. The total model standard deviations for the vertical component are larger than those for the horizontal components at spectral periods shorter than 0.1 s and are considerably smaller than those for the horizontal components above 0.1 s. The within-event residuals were approximately separated into within-site and between-site components, and the corresponding standard deviations are presented.