5:15 PM - 6:45 PM
[SSS05-P03] Dependency of waveform inversion results on station distribution
Keywords:Source process, Waveform inversion, Empirical Green's function
Which source model is most realistic? This has been problematic in source kinematics for a long time. The difference in slip models can be due to the difference in the velocity structure, station selection, inversion setting, and methodology. Regarding the station selection, many studies have tried to use high-S/N waveforms with dense station coverage. Although such an effort is required to obtain more accurate slip models, the best way to choose stations has not been established.
In this study, we investigated the effect of the station selection on the waveform inversion results. We utilized the radiation-corrected empirical Green’s function (EGF; Shibata et al. 2022) in the inversion and compared the inversion results among several EGF events for each mainshock. The radiation correction was conducted with the theoretical ray parameter calculated by the TauP package (Crotwell et al. 1999) assuming the velocity structure of the JMA2001 (Ueno et al. 2002). The waveform inversion with the finite fault was performed as the multi-time-window method (Olson & Apsel 1982; Hartzell & Heaton 1983). As we used the trigger-type recordings of the KiK-net stations, operated by the National Research Institute for Earth Science and Disaster Prevention (Aoi et al. 2004), the available stations depended on the EGF events. In this regard, all the available stations within the epicentral distance of 100 km were utilized for the inversion with no arbitrary data selection. Although the waveforms of each EGF event have their own noise, we could estimate the effect of the station coverage by investigating multiple mainshocks.
As a result, we confirmed that the comprehensive characteristic of the slip image was commonly estimated among several EGF events in a case with a better waveform fitting than the variance reduction of 40.0%. In contrast, a subtle difference was confirmed on subfault scale, which can be caused by the station coverage. In addition, when we utilized the same station distribution between two EGF events, the slip images were closer than the original analyses with all the available stations.
In this study, we investigated the effect of the station selection on the waveform inversion results. We utilized the radiation-corrected empirical Green’s function (EGF; Shibata et al. 2022) in the inversion and compared the inversion results among several EGF events for each mainshock. The radiation correction was conducted with the theoretical ray parameter calculated by the TauP package (Crotwell et al. 1999) assuming the velocity structure of the JMA2001 (Ueno et al. 2002). The waveform inversion with the finite fault was performed as the multi-time-window method (Olson & Apsel 1982; Hartzell & Heaton 1983). As we used the trigger-type recordings of the KiK-net stations, operated by the National Research Institute for Earth Science and Disaster Prevention (Aoi et al. 2004), the available stations depended on the EGF events. In this regard, all the available stations within the epicentral distance of 100 km were utilized for the inversion with no arbitrary data selection. Although the waveforms of each EGF event have their own noise, we could estimate the effect of the station coverage by investigating multiple mainshocks.
As a result, we confirmed that the comprehensive characteristic of the slip image was commonly estimated among several EGF events in a case with a better waveform fitting than the variance reduction of 40.0%. In contrast, a subtle difference was confirmed on subfault scale, which can be caused by the station coverage. In addition, when we utilized the same station distribution between two EGF events, the slip images were closer than the original analyses with all the available stations.