Résumé
We analysed the surface displacements estimated by Global Positioning System (GPS) measurements and Système Probatoire d'Observation de la Terre (SPOT) image correlation and the seismicity change induced by the Mw =7.6 1999 September 20 Chi‐Chi earthquake. We first modelled the seismogenic zone for which we introduced a realistic geologically constrained ramp and flat geometry, and estimated by joint inversion of the GPS and SPOT data the coseismic slip distribution, stressing the resolution of the model parameters. The fair adjustment indicates that the ramp and flat geometry is consistent with the geodetic measurement. The calculated coseismic slip increases both towards the surface and northwards with maximum slip of approximately 15 m, in good agreement with the observed surface rupture and previously obtained slip models based on seismic waveforms. This distribution may be represented to a good approximation by a linear increase of slip with a constant azimuth in the plate convergence direction. Special emphasis is given to the analysis of the model resolution, which sharply decreases with depth in spite of the exceptionally well distributed and high‐quality inverted data. This resolution decrease, expected for any slip distribution based on surface observations only, led us then to examine the seismicity as independent data to test our model following the static stress triggering hypothesis. The model is accurately defined in terms of Coulomb stress change in the central part of the rupture where the fault geometry is well constrained by seismic profile. In contrast, at the southern and northern ends it is less appropriate, however the static stress triggering approach is not precise enough to determine whether fault geometry or slip distribution is less suitable. We thus conclude that the analysis of seismicity in terms of Coulomb stress change might not allow us to examine small‐scale features of seismogenic zones. Finally, we discuss post‐seismic relaxation from the temporal variation of seismicity rate. In contrast to the relaxation time of 0.23–0.90 yr inferred from post‐seismic permanent GPS measurements, we found that the decay of aftershock sequence follows Omori's law with an 8–14 yr relaxation time. We suggest that this discrepancy results from the coexistence of several different processes, not all associated with the production of aftershocks.