Blind predictions using a hybrid of 1D multiple reflection theory/equivalent linear method and 3D FDM

Abstract

Abstract We have participated in the blind predictions held in ESG6 and predicted the weak ground motions during the aftershock (BP2) of the 2016 Kumamoto earthquake sequence, and the strong ground motions during the foreshock and mainshock (BP3). First, broadband incident waves in the engineering bedrock at the target site (KUMA) were estimated. Then, the ground motions on the ground surface were predicted by the 1D linear analysis in BP2 and the equivalent linear analysis in BP3. Incident waves less than 0.5Hz were estimated by the 3D finite difference method using the 3D structure model. Ground motions during the target earthquakes have been recorded on the ground surface at the observation station EEB of Japan Meteorological Agency located 1.2km from KUMA. The incident waves in the engineering bedrock at EEB were estimated by the 1D linear analysis in BP2 and the equivalent linear analysis in BP3, and then the incident wave above 0.5Hz in the engineering base layer at KUMA were estimated by correcting with the hypocentral distance. The broadband incident waves at KUMA were synthesized by superposing the incident waves above 0.5Hz and below 0.5Hz. In BP2, the predictions were excessive from 3 to 6Hz. The reason was considered to be modeling error of velocity structure shallower than engineering bedrock of EEB. In particular, NS components became significantly excessive at 5Hz. The reason was considered to be the frequency response of the propagation path localized in EEB. In BP3, both foreshocks and mainshocks were underpredicted below 0.5 Hz and above 5 Hz, and overpredicted between 1.5 Hz and 3 Hz. The underestimation below 0.5Hz was considered to be mainly caused by the source model. The reason for the overprediction of 3Hz from 1.5 Hz was the possibility of underestimation of the lowering of the S-wave amplification at KUMA due to the nonlinear soil response. The overestimation and the underestimation of the reduction of the theoretical S-wave amplification due to the nonlinear soil responses in each of EEB and KUMA were considered to be responsible for the underprediction above 5 Hz. Compared with the predicted results of other participants, except for the overpredicted frequency range, our predicted results were almost within the average +/- standard deviation of the Fourier spectra predicted by the participants. For both the foreshock and mainshock, the average of the participants underpredicted the large Fourier spectrum around1 Hz, but this study reproduced it.