Inferring the coseismic and postseismic stress changes caused by the 2004 Mw=6 Parkfield earthquake from variations of recurrence times of microearthquakes

Kinematic models of coseismic stress, inverted from ground motion data, do not usually find good correlation between the location of aftershocks and high-stress patches. In particular, numerous earthquakes are recorded in areas of the fault where the stress decreases. However, most of coseismic slip distributions have limited spatial resolution (typically not better than similar to 1 km). Here we investigate the stress changes produced by the 2004 M-w = 6 Parkfield earthquake on and near its rupture zone, at the scale of magnitude 2 earthquake asperities (approximately tens of meters). After relocating earthquakes in this zone between 1984 and 2007, we form repeating, highly similar earthquake sequences and study how the quasiperiodicity of occurrence at each sequence, observed during the 20 years preceding the 2004 main shock, is perturbed by this event. We apply a simple model of the seismic cycle to infer the coseismic and postseismic stresses experienced by the repeatedly failing asperities. Despite being spatially sparse, these stress distributions have resolutions only limited by the typical scale of an asperity. We propose that the high spatial variability of the seismicity patterns following the M-w = 6 earthquake, results from an heterogeneous coseismic stress field. The emergence of the Omori-Utsu law observed at large-scale (greater than kilometers) at Parkfield is simply the outcome of averaging such quasi-deterministic patterns over many sequences. The fact that the coseismic stress can significantly change over distances of the order of 100 m adds credence to the hypothesis that earthquake rupture is intrinsically very heterogeneous.