Toward understanding slip inversion uncertainty and artifacts

摘要

The main concept of seismic source tomography, the projection lines along which the observed signals are back-projected to the fault, is extended to complete wavefields.The so-called "dynamic projection strips" (DPSs) are defined, and a method to construct the strips from individual waveforms is described. In this way, each individual station role in the inversion can be better understood. Synthetic models with two asperities (two unilateral and one bilateral rupture scenarios) are used as examples. They are analyzed using two independent slip inversion methods with similar results, bias of the rupture speed for all scenarios and a strong false asperity in the middle of the bilateral fault. Both artifacts can be explained by the DPS analysis as inherent nonuniqueness of the inverse problem due to the joint effect of the two true asperities. Removal of some slip artifacts by imposing various constraints is discussed, but most of the constraints are hardly applicable in practice. As such, it is recommended to look at least for possible indications of the most significant inversion artifacts. This seems to be feasible through combining DPSs derived from real data and synthetic models of various rupture scenarios for a given fault and stations. The ideas are applied to the Movri Mountain earthquake in Greece, Mw6.3, 8 June 2008. It appears that the earthquake was predominantly unilateral, however with a nonunique spacetime slip pattern. Few equivalent nonsmooth models, all fitting the data equally well, are illustrated.