*Doctoral Dissertation 2000*

### CONSTRAINTS FOR DYNAMIC MODELS OF THE RUPTURE FROM KINEMATIC SOURCE INVERSION

Nelson Pulido

The present thesis is divided in two parts. In the first part I studied the source process of the 1995 Tauramena (Colombia) earthquake by performing a kinematic source inversion. In the second part I developed a methodology for estimating the dynamic rupture parameters from kinematic models of the source. The methodology is applied to the kinematic source models of the 1995 Tauramena, 1995 Hyogo-ken Nanbu and 1992 Landers (California) earthquakes.

In the first part I studied the rupture characteristics of the 1995 Tauramena (Mw=6.5) (Colombia) earthquake by performing a multi-time window linear source inversion using strong motion and teleseismic waveforms. The rupture plane of the Tauramena earthquake was estimated from the aftershocks distribution, which shows two perpendicular fault planes, one corresponding to the mainshock, dipping to the northwest with a reverse mechanism, and the other corresponding to a big aftershock and dipping to the southeast. The resolved slip distribution of the mainshock shows the existence of three asperities with a maximum slip of about 3 meters. The slip vector on the fault plane showed a strong change in orientation during the rupture, suggesting a low level of absolute initial stress with a highly heterogeneous orientation.

In the second part I developed a methodology for retrieving the Critical Slip Weakening Distance (Dc) of an earthquake by studying the spatio-temporal distribution of the apparent stress calculated from the spatio-temporal distribution of the slip velocity function in the fault plane obtained from a kinematic model of the source. The idea is based on the fact that the apparent stress that is simply the earthquake radiated energy normalized by the seismic moment, can be related to the difference between the earthquake average stress and the frictional dynamical stress in the fault plane during the rupture process. Under the assumption that the earthquake stress release process is mainly characterized by the Slip Weakening Friction Law, the above observation allowed me to give an estimate of the Critical Slip Weakening Distance Dc, from the relationship between the cumulative slip and the apparent stress evolution in the fault plane while the rupture proceeds. Detailed kinematic models of the source, in which the fault plane has been discretized into a large number of subfaults, allow the calculation of the spatio-temporal evolution of the apparent stress at every subfault and then its distribution over the fault plane.

From the cumulative slip-apparent stress relationship I estimated the critical slip Dc of the 1992 Landers and 1995 Hyogoken-Nanbu (Kobe). I found that the Critical Slip Weakening played a very important role in controlling the rupture velocity. In the case of Landers earthquake the very low rupture velocity (1 km/sec) in the central segment could be explained by the large Dc value obtained for the main asperity on that segment (Dc=3.5 m). I also found that the Critical Slip Weakening Distance is larger in the shallow crust compared with the deeper crust for both Landers and Hyogoken-Nanbu earthquakes. The previous observation is in agreement with the fact that a large Dc value in the shallow crust prevented the nucleation process in that region.

I conclude that very useful information about the dynamics of the rupture process can be obtained directly from the kinematic inversions of the source. The procedure I propose gives a high bound estimate of the Critical Slip Weakening Distance Dc, and therefore can be useful in constraining the earthquake friction law parameters to be used for dynamic models of the source. The applicability of the procedure is however limited to the low frequency range of most of the kinematic source models available. Kinematic inversions with a higher frequency range would allow more accurate estimations of the radiated energy and therefore better estimations of the Critical Slip Weakening Distance.