強震観測記録に基づく2016年熊本地震の広帯域震源特性

抄録

 We estimate the broadband source model of the 2016 Kumamoto earthquake (M_J7.3, Mw7.1) based on the empirical Green's function method in order to improve strong motion predictions of crustal earthquakes. Firstly we estimate fmax of 37 earthquakes with M_J4.0 to 7.3 occurring in the source region of the main shock using source spectra estimated by the spectral inversion analysis in our previous paper. The estimated fmax of the main shock is 4.4 Hz, which is smaller than 6 Hz for the strike-slip earthquakes with almost the same magnitude. The fmax tends to increase in accordance with the magnitude decreases. The differences of the fmax between the main shock and Mw5-class two element earthquakes are corrected. The nonlinear amplification factors of strong motions during the main shock are also corrected using horizontal-to-vertical spectral ratios between strong motions and weak motions at K-NET stations. After these corrections we estimate the broadband source model of the main shock using K-NET and KiK-net (borehole) strong motion records at 18 stations.
 We study three cases of outer faults as regions searching strong motion generation areas (SMGAs). In the first case (A-model), two northwest-dipping faults around the Futagawa fault and the Hinagu fault are assumed. In the other cases (Y-model and K-model) only the northeastern part of the Futagawa fault near Mt. Aso is changed to be a southeast-dipping fault based on crustal deformation data and some newly developed source models. Since few aftershocks were occurred there, we assume Y-model and K-model with different strike and dip angles. After estimating the broadband source model for A-model by the forward modeling, the broadband source models for Y-model and K-model are estimated by the grid search method using evaluation functions for envelopes of acceleration and waveforms of velocity filtered in the period range of 0.1 to 5 s.
 The Y-model is the best and the A-model is the second best although the difference of the evaluation-function-value of both models is small. The plane positions of the SMGAs on the fault near Mt. Aso are the almost the same between Y-model and A-model and is in the so-called Aso gap where few aftershocks were occurred. The SMGAs for both models extend to the inside of northwestern region of the Aso caldera. For the Y-model the short-period spectral level is almost the average for crustal earthquakes by Dan et al. (2001) and the total area of SMGAs are 1.16 times of the average for the total area of asperities for crustal earthquakes by Somerville et al. (1999). The 11.5 MPa of stress drop of five SMGAs is almost the average for crustal earthquakes. The source parameters of A-model are also almost same to previous relations for crustal earthquakes. The results on scaling relations are consistent with Irikura et al. (2017) in general but the existence of SMGAs inside Aso caldera is different from Irikura et al. (2017). The rupture velocity of 2.8 km/s for SMGAs on the Futagawa fault is faster than 2.4 km/s for the SMGA on the Hinagu fault. The positions of SMGAs located in the northeast side on the northwest-dipping Futagawa fault agree better to high peak moment rate areas than large slip areas estimated by Yoshida et al. (2017). This result suggests that short-period ground motions generate slightly deeper areas than long-period ground motions. It is qualitatively found that strong motions with large velocity pulses observed at KMMH16 (Mashiki) in borehole are generated by the SMGA located in shallow depth just beneath KMMH16.