Observatoire de Grenoble, Université Joseph Fourier, Grenoble, France

We present the results of an analysis of the Michoacan and Landers earthquakes to constrain the kinematic description of the rupture process. The example of the Michoacan earthquake shows that a crack type model is better a dislocation model to describe the displacement in the vicinity of the fault. We also show that this in contradiction with the fact that the faulting appears to be a complex process. We attribute this complexity to instabilities in crack growth. The June 28, Landers earthquake offers an exceptional opportunity to apply a new inversion technique to a major strike slip event. We model the rupture evolution including local differences in slip durations and variations in rupture velocity. The slip distribution shows that this event consists of a series of regions of high slip (sub-events) separated by regions of relative low slip. Our inversion favors the hypothesis that the duration of the slip at each point is of the order of the duration of the rupture of each sub-event and is consistent with a crack type process occurring during each sub-event. For such a large earthquake, the slip duration is however smaller than the total rupture duration.