Attached figure compares observed peak ground velocity (PGV) with values from obtained from my initial 3D ground motion simulation. There are a total of 114 sites (CI network only) extending from just north of the international border to beyond the San Fernando basin (region covered by the simulation can be seen in the animation posted earlier today).

The rupture model used in this simulation (800-v1.0) has a moment of 7.8e+26 dyne-cm which gives a magnitude of Mw 7.22.

In the figure, I plot the natural logarithm of the ratio of simulated to observed PGV as a function of observed PGV. The results are displayed for two bandwidths: f < 0.2 Hz (5 sec) and f < 0.5 Hz (2 sec). These peak values are measured from the observed and simulated three-component waveforms after low pass filtering.

In general, the simulated values tend to overpredict the observations by about 30% for both frequency bands. This suggests that the moment used in the simulation is too high. Scaling the moment down by 30% would give a value of 6.0e+26 dyne-cm and thus reduce the magnitude to Mw 7.15, which is entirely plausible.

There is also a clear trend of decreasing over-prediction for increasing observed PGV values (i.e., simulation has larger over-prediction for low observed PGV compared to the higher observed PGV values). This may represent a distance trend where the more distant sites (lower observed PGVs) have larger overprediction by the simulation compared to the closer sites (higher observed PGV). This would suggest the Q used in the simulation might be too high (I use Qs = 50 * Vs, Vs in km/s; Qp = 2 * Vs). However, since the trend is about the same for both frequency bands, anelastic attenuation may not be the only culprit. Other factors (e.g., basin amplification, wave guide effects, etc.) may certainly be coming into play as well, so this needs to be investigated further.