ACCURACY OF SOME GROUND MOTION PREDICTION MODELS FOR PPI STATION-WEST SUMATRA USING OBSERVED STRONG MOTION DATA
Supervisor: Tsuyoshi TAKADA
This study proposes the ground motion prediction models applicable to PPI Station located in Padang Panjang, West Sumatra, at which 120 components of strong motion data of 40 earthquake events with sampling rate 100 Hz have been recorded, in order to find site-specific parameters used to investigate the site class and the accuracy of the prediction models for this station. The H/V spectral acceleration ratio scheme is adopted to find the dominant period of ground beneath this station since there are no quantitative subsurface soil properties available. The results of H/V analysis show that the mean peak of H/V ratio of EW, NS and total horizontal components to vertical component is around 0.2 second. The peak period of H/V ratio is in good agreement with the geologic information from geological surface map published by Geological Research and Development Center (GRDC, 1973). The geologic formation for the ground beneath and around this station is dominated by quaternary volcanic rock (Qast). Both results indicating the site class for the ground beneath the station could be categorized as rock site with AVS30 more than 600 m/s. The existing prediction models for peak ground acceleration (PGA) and 5% damped spectral acceleration (Sa) at rock site are tested in this study. The accuracy of the selected models is discussed on the basis of the statistical distribution of the logarithmic deviation between the prediction and the observed value. The mean residuals of all prediction models of Sa are found to have the same tendency that they underestimate Sa at period T = 0.2 second and overestimate Sa at T ³ 0.4 second. It is found that all prediction models show significant period-dependent mean errors and are not allowed to be applied to this site without site correction factors, and among them, Youngs et al., 1997, Kanno et al., 2006, and Zhao et al., 2006 provide the smallest prediction error. Based on these results, it can be concluded that all prediction models may be applied with each site correction factor to reflect the site-specific condition and among them three models proposed by Youngs et al., 1997, Kanno et al., 2006, and Zhao et al., 2006a are the best for this station. However, the error is still large; therefore a further study is needed to check it.
|Citation: ||Bulletin of IISEE, 43, 73-78.|