| dc.description.abstract | We use the GPS data collected from the 64 continuous stations in the Taiwan area, set up and operating by Academia Sinica, Ministry of Interior Affairs, and Central Weather Bureau, and the repeated surveys of 103 campaign sites conducted by Academia Sinica after the Chi-Chi earthquake. The data period is from January 1st, 2001 to the end of 2005. GPS data are processed with the Bernse software V.4.2. To obtain the velocity field of continuous GPS stations, the position time series are corrected for the coseismic offsets and the seasonal periodic motions. We also use the repeated GPS data observed from 1992 to 1999 and apply the linear interpolation to obtain the preseismic velocities for the new sites established after Chi-Chi earthquake. Then the preseismic velocity, is subtracted from the observed velocity for each station to obtain the surface displacement after the earthquake. The maximum horizontal and vertical postseismic displacements for the 5-year period from 2001 to 2005 are 17 cm and 12 cm, respectively.
To realize the impact of viscoelastic relaxation on the lower crust and the upper mantle caused by Chi-Chi earthquake, we use the Cubit software by Sandia National Laboratory to establish grids, and the models are constructed by using the finite-element code, PYLith. The appropriate viscosity is adopted for each layer to estimate the surface deformation caused by the viscoelastic relaxation, and further comparisons are made between the model results and the GPS observation. To verify the consistency of the finite-element model with the analytic solution, we start with a two-layer model with a horizontal viscoelastic layer under an elastic layer. In general, the differences between the model values and analytic solution are quite small, indicating the good reliability of finite element models. Afterward we apply a three-layer model to compare with the GPS observation, and we find that the tendency of model curve in the horizontal component is similar to the observations, however, the values are in much difference. Thus we consider to add a block of low viscosity underneath the Central Range following the results of Lin (2000).
We change the position, viscosity, and size for the block of low viscosity to examine the relation between the viscosity and the model curve. The results show that the variation of position has the most influence upon the model curve. As the block of low viscosity contacts the fault, the model curve expresses a very large amplitude. We find that model with the vertical low viscosity block located at 40-60km from the fault (Visc1), and another model with the horizontal low viscosity block located at 40-80km from the fault (Visc10) have the minimum RMS values. Moreover, comparing the model curves with GPS data, we find that models Visc1 and Visc10 are more or less consistent with the observation. Accordingly, we consider these two models are the best solutions in this research. We inferred that a low viscosity material may exist beneath the Central Range, and it does not contact with the Chelungpu fault. However, the best models are still not fitting the vertical data well. A more sophisticated three-dimensional model is needed in the future studies. | en_US |