Frequency-dependent Effects of Rupture for the 2004 Parkfield, California, Mainshock at UPSAR

We processed the accelerograms for the Sept. 28, 2004 Parkfield mainshock (M6) from the Parkfield Dense Seismograph Array (UPSAR) using beam forming techniques to determine coherence and slowness of incoming arrivals and we found that the frequency-dependent effects of rupture propagation of can be seen in the acceleration records in at least two ways. The first is the effect of directivity. During the 4-6s interval after nucleation, UPSAR lies in the forward azimuth of the rupture front and coherence and amplitudes are relatively high. In this interval, the records are rich in frequencies as high as 40 Hz. As the rupture front progresses, UPSAR becomes perpendicular to the location of the rupture front on the fault and then passes into the back azimuth. During this 5-7s interval, the coherence drops off, amplitudes are typically smaller and the traces are depleted in high frequencies compared to when UPSAR was in the forward azimuth. We model the rupture process of the Parkfield mainshock using realistic distributions of slip and show that they also have significant changes in frequency as the rupture passes from the forward to the back azimuth.

The second effect is the frequency dependence of the rupture front. That is observations at UPSAR were used to address the question of whether low frequency energy emerges from the same parts of the fault as high frequency energy. Acceleration records of the Parkfield mainshock from UPSAR were filtered in 5 different narrow frequency bands (0.25-0.5 Hz, 0.5-1 Hz, 1-2 Hz, 2-4 Hz, and 4-8 Hz) and beam forming was repeated. Above 1 Hz the estimates of back azimuth and apparent velocity are about the same and are independent of frequency. Below 1 Hz, however, the estimates appear to be significantly different with a change of about of 0.34 km/s in apparent velocity and 22 degrees in back azimuth at 2s after the S wave arrival, for example, between the middle and lowest band. From previous work on mapping values of apparent velocity and back azimuth to specific locations on the San Andrea fault (Fletcher et al., 2006, Bull. Seism. Soc. Amer.), the values suggest that the low frequency energy comes from shallower parts of the fault and trails the high frequency energy radiated by the rupture front, but we have not investigated the frequency-dependence of this correction.