Empirical Predictive Equations for Strong-Ground Motions in the 2006 Pingtung Earthquake Sequence

The December 26, 2006 Pingtung earthquake is the largest event observed in the southwest Taiwan. In order to provide the basic data for reevaluating the seismic hazard of this power plant and the southern Taiwan, we derive empirical predictive equations for peak ground acceleration(PGA), peak ground velocity(PGV), peak ground displacement (PGD) and pseudo-spectral acceleration（PSA）at 0.01~10 periods using the strong-motion data obtained from the Taiwan Strong Motion Instrumentation Program(TSMIP). We select moment magnitude as the source parameter, epicentral distance as the wave propagation parameter and consider three types of site effects terms in our predictive model. These site parameters include the traditional site classification，VS30 and site amplification factor estimated in this study. For obtaining coefficients of these predictive equations, we apply genetic algorithm for global searching. The complete listings of coefficients for of all the predictive equations are given in Tables 4.1 and 4.2. According to the residual analyses, we can evaluate the performance of various site-effect terms introduced in our predictive model. Results show that the VS30 gives a best estimation in PGA but it fails in PGV estimation. On the other hand, the traditional site classification reduces a significant uncertainty in the PGV estimation. The amplification factor does not improve much in the overall estimation, but it improves the PGA estimations in Taipei basin and Ilan plain. Therefore, we conclude that no single site-effect term can give best estimation for all types of ground motions in this study. Our predictive equations are also applied to an independent observation in the Power Plant No. 3. One free-field observation inside the power plant is 156.8cm/s/s while our predictions are about 103.3 cm/s/s. However, this observation still falls within one standard error of our predictive equations. Comparing with other independent study in Taiwan, we found that the attenuation in our predictive model is less than that of an island-wide shallow crustal model or that of the subduction zone in the northeastern Taiwan.