This electronic supplement contains figures of station and event locations, velocity models, and waveform comparisons.
The data in our source inversions are the differences between synthetic and recorded waveforms, which can be attributed to both source and structure. Therefore, in source inversions we need to reduce the structural-model-induced synthetic-record waveform differences as much as possible so that they are not mapped into the source model. Here, we present a waveform comparison between records and synthetics calculated in structural models with different levels of complexity, including a regionally averaged 1D model, path-specific 1D models, a 3D tomography model (Kuo-Chen et al., 2012) for Taiwan, and the 3D model with the consideration of surface topography. We choose three stations in this comparison: one station near the source for which 3D structural effect is small and two more distant stations with different levels of 3D effects. Figure S1 shows locations of the earthquake and stations. The regional 1D model (Fig. S2) is obtained by averaging the 3D model of Kuo-Chen et al. (2012) over the entire island of Taiwan. Path-specific 1D models obtained by averaging the same 3D model horizontally along paths between the earthquake and the three stations are shown in Figure S2. Synthetic seismograms for 1D models are computed by the frequency–wavenumber (f-k) integration method (Zhu and Rivera, 2002). All waveforms are band-pass filtered to 0.05–0.8 Hz.
Figure S3 presents the waveform comparisons. It is clear that at all stations synthetics obtained in regional 1D model (gray traces) have the simplest waveforms and show the largest discrepancy from the records. On the other hand, path-specific 1D models provide remarkably similar waveforms (green traces) to those in the 3D model without consideration of surface topography (red traces). However, there are still obvious additional waveform complications in the records, especially around the peak amplitudes, which can only be seen in synthetics obtained in 3D model with consideration of surface topography (blue traces).
The waveform comparisons at the three stations exhibit different levels of discrepancy between synthetics and records, which can be informative to our selection of data and structural models in finite-fault inversions. At HWA035 (Fig. S3a), the station 20 km away from the hypocenter, all synthetics resemble the record very well, suggesting that at this epicentral distance even the regional 1D model can be as good as the 3D model for finite-fault inversion. However, at station TDCB (Fig. S3b), 78 km from the source, the regional 1D model is no longer sufficient. Path-averaged 1D model is much better and even as good as the 3D model, but some of the waveform complexities can only be modeled when the topography effect is accounted for. At station TPUB (Fig. S3c), across the Backbone Range from the earthquake source, even the current 3D model is clearly insufficient in modeling the recorded waveform at this frequency band. Although there is still much to be done to improve the structural model so that the synthetics fit the records better, these comparisons demonstrate that our synthetic seismograms calculated in 3D model with the consideration of surface topography can effectively reduce danger of mapping the structural-model-induced waveform discrepancies into the source model.
Figure S1. Source (red star) and stations used here for waveform comparison. The source is the 31 October 2013 earthquake (Mw 6.33) that occurred in Ruisui in eastern Taiwan. A point-source focal mechanism (focal mechanism plot) with strike, dip, and rake angles of 197°, 42°, and 26°, respectively, is used. Broadband Array in Taiwan for Seismology (BATS) and Taiwan Strong Motion Instrumentation Program (TSMIP) stations are shown by yellow and black triangles, respectively.
Figure S2. 1D velocity models: black lines are for regional 1D model averaged horizontally from the 3D model of Kuo-Chen et al. (2012) over the entire island of Taiwan; whereas blue, red, and green lines are for the path-specific 1D models averaged horizontally from the same 3D model along the paths between the earthquake and stations HWA035, TDCB, and TPUB, respectively (see Fig. S1 for source and station locations).
Figure S3. Comparison of three-component records at three stations from the Ruisui earthquake (Fig. S1) with synthetic seismograms. The seismograms include records (black), finite-difference synthetics calculated in 3D model with and without consideration of surface topography (blue and red, respectively), and synthetics calculated by f-k integration method in path-specific and regional 1D models (green and gray, respectively). Station name and epicentral distance are given above the east–west component record. All the waveforms are band-pass filtered to 0.05–0.8 Hz and are normalized by the individual peak amplitudes. The vertical dashed lines indicate the predicted P-wave arrival times in the 3D model with no topography. The 1D models are shown in Figure S2.
Kuo-Chen, H., F. T. Wu, and S. W. Roecker (2012). Three-dimensional P velocity structures of the lithosphere beneath Taiwan from the analysis of TAIGER and related seismic data sets, J. Geophys. Res. 117, no. B06306, doi: 10.1029/2011JB009108.
Zhu, L., and L. A. Rivera (2002). A note on the dynamic and static displacements from a point source in multilayered media, Geophys. J. Int. 148, 619–627.
[ Back ]
