Aftershock asymmetry/rupture directivity among central San Andreas fault microearthquakes

Using a waveform cross-correlation technique, we have obtained precise relative locations for nearly 75% of the Northern California Seismic Network catalog (4300 earthquakes) occurring between 1984 and 1997 along 50 km of the San Andreas fault. Errors in relative location are meters to tens of meters for events separated by tens to hundreds of meters. We find that consecutive earthquakes in the relocated catalog occur no closer than a distance approximately equal to the radius of the first rupture, as estimated from the moment-magnitude relationship of Abercrombie [1996] assuming a 10-MPa stress drop. When the relative position vectors between consecutive events are normalized by this distance and projected onto the fault surface, they define a hole whose shape suggests that typical microearthquakes are elongate in the mode II (slip-parallel) direction by several tens of percent. Moreover, of the 100 immediate aftershocks occurring closest to the mode II edges of the prior rupture, more than twice as many occur to the northwest than to the southeast. We interpret this asymmetry as resulting from the large contrast in material properties across the fault. Models of dynamic rupture between dissimilar media predict that ruptures in this region may run preferentially to the southeast, in the direction of motion of the lower-velocity material. If so, then the barriers that stop rupture fronts moving to the southeast should initially be farther from failure, on average, than the barriers that stop rupture fronts moving to the northwest. Once the rupture stops, the induced stress change is more symmetric but the fault remains farther from failure (on average) to the southeast. This interpretation receives some support from pulse width measurements on a localized set of 72 magnitude 0.6 to 3.6 earthquakes.