Estimation of High-Frequency Empirical Green’s Tensor Spatial Derivatives from Observed Seismic Records

ABSTRACT We investigated the feasibility of estimating high-frequency empirical Green’s tensor spatial derivatives (EGTDs) using aftershocks of the 2016 Kumamoto earthquake, Japan. Before the EGTD estimation, we accurately determined two source parameters (seismic moment and duration of the source time function [STF]) and three fault parameters (strike, dip angle, and rake angle), which are necessary for the EGTD estimation. Further, we investigated the extent to which high-frequency EGTDs can be estimated using five groups consisting of aftershocks in the Japan Meteorological Agency magnitude scale range over 1.5<Mj<3.5, within zones of 1 and 2 km diameter. We found that high-frequency EGTDs with cutoff frequencies between 8 and 16 Hz could be estimated from the aftershocks within the zone of 1 km diameter using optimum values of the source and fault parameters. In contrast, it is difficult to stably estimate the high-frequency EGTDs from the aftershocks within the zone of 2 km diameter, even for the cutoff frequency of 4 Hz. Furthermore, we found that the accuracy of the EGTDs is evaluated by the reproducibility of earthquake motion records as well as the STF. We examined whether there is a difference in the reproducibility of recorded earthquake motions by earthquake motions synthesized using EGTDs with two different STFs. The two types of EGTDs, which were estimated by adopting the two STFs, reproduced the earthquake motion records with equal accuracy but were quite different in waveform and amplitude from each other, indicating that we should adopt the STF that is based on or consistent with the physics of fault fracture when estimating the high-frequency EGTDs. Thus, it is certainly feasible to estimate the high-frequency EGTDs from earthquake motion records radiated from a 1 km diameter zone using an appropriate STF.