Development of ground motion prediction equations using median distance

In comparing the existing ground motion prediction equations (GMPEs) with the observed records of the 2011 Tohoku earthquake, it is becoming clear that the results of the equations may vary significantly according to which definition of distance is applied. The problems arising from the definition of distance, such as the hanging wall effect, are also becoming more evident. Based on these observations, we proposed a new distance measurement called median distance (MED) defined as the shortest distance from an observation point to the straight line extended from the midpoint of the fault plane in the along-strike direction, accounting for the characteristics that the main strong motion generation areas (SMGAs) are often configured to be placed at the central part of an earthquake source fault. We have confirmed that by using MED the geometric issues such as the hanging wall effect will be resolved to some extent.

Furthermore, based on the strong motion recordings from 35 Japanese earthquakes with moment magnitudes of 5.6 to 9.1, we developed a new GMPE of acceleration response spectra (GMRotI50) using MED. We corrected the data to be at the bedrock with a shear-wave velocity of 1500 m/s by using an empirical evaluation equation of spectral amplification factor proposed by Si et al. (2013). We confirmed that there are no obvious distance-saturation at close distances based on the records from the 1995 M_w 6.9 Kobe earthquake and 2008 Mw 6.9 Iwate-Miyagi Nairiku earthquake. We then used a simple equation of log GMRotI50 = b - log X_med - k X_med as the first-stage regression model. In the second stage, we made regression analyses for the dependence of the coefficient b on moment magnitude, earthquake type (crustal, intra-plate, or inter-plate), and focal depth. We used a bilinear equation with respect to moment magnitude, thus our GMPE can be applied to M_w 9 megathrust earthquakes.