On linear site amplification behavior of crustal and subduction interface earthquakes in Japan: (1) regional effects, (2) best proxy selection

Abstract

A strong-motion database is compiled from KIK-Net Archive of Japan (kyoshin.bosai.go.jp) with following criteria: shallow crustal and subduction interface events between 2000 and 2016, recorded within 300 km and moment magnitude, Mw range is 4.9–7. Time-based average of shear-wave velocity profiles up to different depths 10 m (VS10), 20 m (VS20), 30 m (VS30) and 50 m (VS50) are computed. Only stations with 200 < VS30 < 1500 m/s and the shear-wave velocity (Vs) profiles of the site reaches 800 m/s (h800) are used. Then ground-motion prediction equations for crustal and interface earthquakes are generated for surface motions, borehole motions, and surface-to-borehole spectral ratios (SBSR). The linear site response effect in surface motion equations are modeled in terms of h800 and site proxies: VS10, VS20, VS30, and VS50. The borehole-motion and SBSR models use Vs at borehole depth, and impedance ratio and h800, respectively. The crustal earthquakes generally produce higher ground-motions than interface events at rock site. The site behavior of crustal and interface earthquakes are found statistically different. The difference in short-period site factor is not more than 5%, whereas, at long-period site factor, crustal motions are amplified 13% more than interface motions. Applying the borehole motion and SBSR models, the estimated surface motions are compared with observed surface motions. The misfit is generally found less than 6%. It is concluded that SBSR can be a good indicator for site-specific amplification. Simple inference test is applied to the between-site residuals of surface motion and SBSR models to understand the regional variations in site amplification. In the point of site amplification, the interface earthquakes are more sensitive to regional effects than crustal earthquakes. The comparisons of between-site residuals and site amplification estimates computed with alternative site proxies reveal that at short periods VS20 is sufficient to predict the site response; however, as the period increases the performance of VS50 proxy becomes the best.