This electronic supplement includes two animations, one of the temporal evolution of the San Simeon aftershock sequence, and one of the spatial distribution of the aftershocks. It also includes validation of the inversion for the 3D velocity model used to relocate the aftershocks.
Movie showing the temporal evolution of the San Simeon aftershock sequence.
or
Download: movie02.mp4.zip: [~800 KB Zipped H.264 MP4 Video File].
Movie showing the 3D spatial distribution of the San Simeon aftershocks. The shapes of the seismicity features are represented by the 8 events/km3 isosurface, in yellow.
or
Download: movie01.mp4.zip: [~2 MB Zipped H.264 MP4 Video File].
Tech Note: The movie files linked above are compressed using the H.264 video codec and may require Apple's free QuickTime software to view. QuickTime is included on all Apple Macintosh and many Microsoft Windows computers. If you need to install QuickTime software it can be downloaded from http://www.apple.com/quicktime/. Users of Linux or Unix operating systems may be able to play these movies using open source software such as Mplayer, Totem, or VLC.
The velocity model was determined first on a coarse grid (10 km spacing), then on a medium grid (4 km spacing), and finally on a fine grid (2 km spacing.) Eight iterations were performed for each grid, with the final result from one grid used as the starting model for the next. The final RMS arrival time residual is 0.08 sec, compared to a residual of 0.17 sec for the 1D starting model. The following figures show the source-station geometry (Figure 1), the ray coverage (Figures 2, 3, and 4), the trade-off curves used to find the optimal damping values (Figure 5), and the results of a checkerboard resolution test (Figures 6 and 7).
The most prominent feature of the 3D Vp model is a region of low velocity in the center of the aftershock zone at depths of ~4-8 km. We test how strongly this low velocity region is required by the travel time data by constructing several alternative models. The low Vp anomaly results in somewhat shallower earthquake depths than found using a 1D velocity model, so we fixed the earthquakes to their 1D-model depths during the inversion to force a higher-velocity model. This model also exhibits a lower Vp region in the center of the aftershock zone, similar to that of the preferred model (Figure 8). We also determine a Vp model from the pre-mainshock events and active source data only, to test whether the low-velocity zone in the middle of the aftershock region could be an artifact of the dense sampling in this area. This model is considerably less well constrained, but still exhibits a similar low Vp zone (Figure 9). Therefore the low Vp anomaly appears to be a robust feature of the velocity model.
Figure 1. Source-station geometry for the velocity model inversion, the black box outlines the model region. The earthquake arrival time dataset is generated using an algorithm that assigns weights to each source-station path based on the hit count along the path, with higher weighting for lower hit-count, and selects events with the highest total weight. The algorithm therefore identifies events with large numbers of picks and/or rays through undersampled parts of the model region, and eliminates "duplicate" near-by events. P-wave and S-wave hit count are tabulated separately, and the S-wave data are more sparse, so the algorithm favors events with more S-wave picks. The data quality and quantity may be lower in areas of sparse seismicity, but ray coverage uniformity is maximized and repetitive data is minimized.
Figure 2. Total available ray coverage for P-waves and S-waves, expressed as hit count, for the CISN phase catalog alone, and the combined CISN and CCSN catalog. The black box is the model region.
Figure 3. The actual ray coverage for the P-wave velocity model, expressed as the derivative weight sum (DWS), the weighted hit count computed by SIMUL2000 [Thurber, 1993; Eberhart-Phillips, 1993; Evans et al., 1994].
Figure 4. The actual ray coverage for the Vp/Vs model, expressed as the derivative weight sum (DWS), as in Figure 3.
Figure 5. Trade-off curves used to select the damping parameters for the inversion. The best damping parameter jointly minimizes the model length and the data variance (misfit). The selected damping parameter indicated by a filled square. (a) For the coarse-grid (10 km spacing) Vp model inversion. (b) For the medium-grid (4 km) Vp inversion. (c) For the fine-grid (2 km) Vp inversion. (d) For the coarse-grid (10 km) Vp/Vs ratio inversion. (e) For the medium-grid (4 km) Vp/Vs ratio inversion. There is not adequate S-wave coverage for a fine-grid Vp/Vs inversion.
Figure 6. The results of a checkerboard test of the resolving power of the Vp model. The assumed "true" Vp model consisted of the average 1D model with perturbations of ±5% in alternating 10 km x 10 km squares. The synthetic arrival time dataset was generating using ray-tracing through this "true" model and the source-station geometry of the real dataset. The inversion was carried out exactly as the inversion of the real arrival time data. Note that the best resolution is achieved between the coast and the Nacimiento Fault, from 4 km to 8 km depth.
Figure 7. The results of a checkerboard test of the resolving power of the Vp/Vs model. As in Figure 6. The resolution is poorer than the Vp model (Figure 6), but is also best between the coast and the Nacimiento Fault, from 4 km to 8 km depth.
Figure 8. An alternative Vp model. For this model, we did not jointly invert for relocated hypocenters, but left the event locations fixed to their best 1D locations. This demonstrates that the low Vp zone in the middle of the model is not an artifact of a trade-off between velocity and event depth.
Figure 9. An alternative Vp model, computed using only pre-San Simeon earthquakes and active source data. This model also exhibits the low Vp zone in the center of the region at 6 km depth.
[ Back ]
