Compound earthquakes on a bimaterial interface and implications for rupture mechanics

Abstract

Earthquake ruptures on the San Andreas Fault are affected by the material contrast across the fault. Previous observations of microearthquakes at the northern end of the creeping section have found strong signals of asymmetry in both rupture directivity (preferential propagation to the SE), and aftershock asymmetry (many more to the NW, on timescales from 10 s to 9 hr). To understand the aftershock asymmetry, Rubin & Ampuero simulated slip-weakening ruptures on a bimaterial interface and observed differences in the timescales for the two edges to experience their peak stress after being slowed by barriers. This is suggestive of the possibility of asymmetry of subevents in compound earthquakes. A second possible source of subevent asymmetry is that when slowed by barriers, a significant tensile stress pulse is predicted to propagate in the SE but not the NW direction. To study the possible asymmetry of subevent distribution, we search for compound events using an empirical Green's function method. Three sections on the northern San Andreas and part of the Calaveras faults were selected where the events have high spatial density and similar focal mechanisms. About 677 candidate compound events were identified in a 28 869-event catalogue from 1984 to 2009. Most delays between the two subevents cluster around the shear wave transit time over the subevent separation, although with considerable scatter. For subevents on the San Andreas Fault separated by 0.7–2 times the estimated radius of the first subevent (the same spatial separation found to exhibit strong asymmetry of longer term aftershocks), nearly twice as many second subevents occurred to the SE of the first than to the NW. This asymmetry of second subevent distribution is not present on the Calaveras Fault, which does not have a significant across-fault contrast in wave speed in this region. One interpretation is that the extra SE subevents on the San Andreas Fault are representative of the events ‘missing'from the ‘longer term’(10 s–9 hr) aftershock population because they became part of the main shock.