Modelling of Source Terms Separated From Observed Response Spectra To Reduce Variability in Gmpe

Abstract eng:
We investigated the characteristics of strong ground motions separated from acceleration Fourier spectra and acceleration response spectra of 5% damping calculated from weak and moderate ground motions observed by K-NET, KiK-net, and the JMA Shindokei Network in Japan using the generalized spectral inversion method. As a reference our separation method uses the outcrop motions at a rock site, YMGH01, where we extracted the site response due to shallow weathered layers. We include events with JMA magnitude equal to or larger than 4.5 observed from 1996 to 2011. From the corner frequencies of Fourier source spectra and CMT seismic moment values, we calculate Brune’s stress parameters to find a clear magnitude dependence, in which smaller events tend to spread over a wider range while maintaining the same maximum value. We confirm that this is exactly the case for several mainshock-aftershock sequences. The average stress parameter for crustal earthquakes, ~0.8MPa, is much smaller than those of subduction zone earthquakes, either plateboundary or intraplate, ~5MPa, which can be primarily explained by their depth dependence. Next we compare the strong motion characteristics based on the 5% damping acceleration response spectra and find that the separated characteristics of strong ground motions are different, especially in the lower frequency range less than 1Hz. These differences comes from the difference between Fourier spectra and response spectra in the observed data; that is, predominant components in high frequency range of Fourier spectra contribute to increase the response in lower frequency range with small Fourier amplitude because strong high frequency component acts as an impulse to a SingleDegree-of-Freedom system. By using the separated source terms for 5% damping response spectra we can obtain regression coefficients with respect to the moment magnitude, which lead to a new Ground Motion Prediction Equation. Although stress drops for crustal earthquakes are 1/7 of the subduction-zone earthquakes, we found that linear regression works quite well. After this linear regression we correlate residuals from this linear relationship as a function of obtained stress drops of corresponding events from Fourier spectra. We find quite a good linear correlation in a wide frequency range from 0.3 Hz to 20 Hz, which yields reduction of variability higher than 50 % from the original one in terms of the standard deviation. When we applied the same kind of correction to Fourier source terms we can see reduction in high frequency range but cannot see any reduction in the frequency range below 1 Hz, as expected. To use derived correction based on the stress drop for a future event we can use average stress drops for different types of events together with the source depth correction.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket