*Doctoral Dissertation 2003*

###
Study on Dynamic Rupture Process and Near Source Strong Motion Simulation

-Case of the 1999 Chi-Chi, Taiwan, Earthquake-

Wenbo Zhang

Earthquake source dynamics provides key elements for the prediction of strong ground motion and for understanding the physics of earthquake processes. This thesis addresses the characteristics of dynamic source rupture process of a large earthquake by using a 3D finite difference method with variable grid spacing. A new algorithm is proposed to deal with a non-planar fault model. We apply this approach to the 1999 Chi-Chi earthquake with a curved fault surface and rebuild the dynamic source rupture process for this larger earthquake.

Our analyzing procedure is as follows, 1) Obtain the spatial-temporal stress
distribution on the fault surface from the kinematic model inverted from strong
motion data. Then estimate the strength excess (yielding stress) and the frictional
stress level for each subfault. 2) Analyze the relationships between stress and slip, and
between stress and slip-rate. Then the critical slip-weakening distance *D _{c}* will be
estimated based on the revealed friction-laws (the slip-weakening law). 3)
Reconstruct the dynamic source rupture process using those dynamic source
parameters with the friction law and simulate the near source ground motions from
the dynamic source model.

For the dynamic source parameters estimated from the kinematically inverted
slip distribution, our results show that the distributions of the dynamic source
parameters of the Chi-Chi earthquake are quite heterogeneous. The stress drop varies
greatly over the fault and its distribution is positively correlated with the final slip
distribution. The peak value of the static stress drop is about 35 MPa. The estimated
strength excesses are generally small suggesting that tectonic shear stress had reached
to the level of the fault strength before the main shock. The analyzed relationships
between stress and slip, and between stress and slip-rate suggest that the behaviors of
the most of the subfaults follow a slip-weakening friction law during rupture.
Applying a simple slip-weakening law, we find that the critical slip-weakening
distances, *D _{c}*, does not seem to be depth dependent, but rather spatially heterogeneous
and appear to be proportional to the final slip. The aftershock activity is correlated
with the distribution of dynamic source parameters. Usually, the aftershocks
concentrated in the areas with small or negative stress drop. Large value of Dc intends
to prevent the seismic nucleation process and therefore reduces aftershock activity in
the areas.

For the dynamic rupture process, this study reveals the rupture propagation "jumping" phenomenon which is difficult to be simulated in kinematic modeling. That is when the propagation front encountered a zone with a high strength excess, the rupture would pause to accumulate more energy to break it. Meanwhile, if there are low strength excess zones around the barrier, the propagation front would jump over the barrier to break the low strength excess zones and leave the high strength barrier unbroken. Such phenomenon of the high strength excess barriers intend to delay the propagation front can be seen clearly in the dynamic model. So, the distribution of the rupture starting time is much more inhomogeneous than that of the kinematic model. Using a thick fault zone model, the dynamic model discovers that the slip on the hanging-wall side is larger than that on the food-wall side. Based on the dynamic source rupture model, the strong ground motions near the fault surface breaks are simulated in frequency range of 0.05 to 0.5 Hz. In general, the synthetic velocity waveforms agree well with the observed records for most stations. The dynamic source model successfully simulates the distinctive velocity pulse for the stations in the forward rupture direction. Also our dynamic source model successfully reproduced the waveforms as well as the distinctive velocity pulses for the station nearby or on the fault surface breaks. These results demonstrate that our dynamic source model can reproduce the main features of long period ground motions; hence, lead us to a better understanding on the source rupture process of the Chi-Chi earthquake. We expect that the results of this present study would help us to yield more realistic ground motion prediction for a large earthquake.