Accurate Prediction of High-Frequency Earthquake Motions Using Empirical Green’s Tensor Spatial Derivatives

ABSTRACT

Using two aftershock groups with 1 km diameter zones and three groups with 2 km diameter zones, we investigated three research issues on the prediction of high-frequency earthquake motions using empirical Green’s tensor spatial derivatives (EGTD): determination of the cutoff frequency on the high-frequency side with reference to the size of aftershock zones, conditions for accurate prediction, and selection of the source time function giving the best prediction accuracy from among the currently available source time functions. Our results indicate that: (1) the EGTD derived from the aftershock zones of 1 km diameter can predict high-frequency earthquake motions accurately up to the cutoff frequency of 8 Hz. Although the EGTD derived from 2 km aftershock zones cannot predict ground motions accurately up to the cutoff frequency of 4 Hz, they can still predict high-frequency motions at observation sites with similar earthquake records. (2) Two conditions are imposed on the estimation of the EGTD having such good prediction performance: earthquake motion records at individual observation sites are similar in waveform to each other, and the aftershocks have various focal mechanisms. (3) The selection is difficult because there are no significant differences in the prediction accuracy between the source time functions considered. Conclusions (1) and (2) mean that after confirming the similarities in waveforms of earthquake motion records, radiated from aftershocks with various focal mechanisms, observed at individual sites, we can estimate the EGTDs that can predict high-frequency earthquake motions accurately. Therefore, we think that the EGTDs are promising tools for the prediction of high-frequency earthquake motions and for the kinematic source inversion. Moreover, our results can be used to identify high-frequency earthquake motion radiation areas on fault surfaces and to estimate dynamic fault rupture parameters.