HIGH FREQUENCY GROUND MOTION SCALING IN THE YUNNAN REGION W. Winston Chan, Multimax, Inc., Largo, MD W. Winston Chan, Multimax, Inc., Largo, MD Robert B. Herrmann, Saint Louis University, St. Louis, MO Robert B. Herrmann, Saint Louis University, St. Louis, MO Wenjie Jiao, Multimax, Inc., Largo, MD Wenjie Jiao, Multimax, Inc., Largo, MD Purpose Pilot study of integrated approach to high frequency ground motion scaling Pilot study of integrated approach to high frequency ground motion scaling Absolute ground motion scaling Moment magnitudes Source inversionSource inversion Crustal models Outline High frequency ground motion processing High frequency ground motion processing Source parameter determination Source parameter determination High frequency ground motion modeling High frequency ground motion modeling Future directions Future directions Background Processing of seismic network data in Yunnan, China Processing of seismic network data in Yunnan, China 325 events recorded at 23 stations yield over 5,000 regional S phases 325 events recorded at 23 stations yield over 5,000 regional S phases 2 Data sets merged 2 Data sets merged regional broadband network portable deployment Station Event Distribution Distribution of Observations STATIONSSTATIONS Data Processing Process raw seismograms Pick P, S Reduce to ground velocity (m/s) Compute tables consisting of - Peak motion Duration Spectra Energy RMS signal shape Apply coda normalization Regression Peak filtered velocities Fourier velocities Model using Random Vibration Theory or Bandpass filtered stochastic white noise Regression Model Use filtered vertical component velocity at 0.33, 0.5, 1, 2, 3, 4, 6, 8, 10 and 12 Hz (DT = 0.02 s) Use Coda normalization as independent test of distance term Perform regression on Peak Velocities and Fourier Velocity Spectra at each frequency for the model log A = D(r) + E(r ref ) + S where A is observed motion, D(r) is distance term, E(r ref ) is the excitation term at the reference distance, and S is a site term Regression Specifics: Sum S i = 0, D(40 km) = 0; D(r) is a piecewise linear continuous function defined by distance nodes Regressions Results Hz Frequency Dependence of D(r) Distance Dependence Duration - 2 HZ Distance Dependent Duration Duration decreases at high frequency Low frequency dominated by direct and scattered surface waves High frequency dominated by simple body wave arrivals Modeling The predicted Fourier velocity spectra for a frequency f and a distance r is a(r, f ) = s( f,M W )g(r)e -3.14fr/Q( f ) V( f )e -3.14f kappa where a(r, f) is the Fourier velocity spectra s( f,M W ) is the source excitation as a function of moment-magnitude g(r) is the geometrical spreading function Q( f) is the frequency dependent quality factor which equals Q ( f /1. 0) eta, Q is the quality factor at 1.0 Hz V( f) is a frequency dependent site amplification, and kappa controls site dependent attenuation of high frequency Modeling Procedure Model Fourier Velocity initially Start with g(r), Q 0 and eta Next look at small event E terms to get kappa 0 Include duration to model time domain D(r) Compare observed and predicted D(r)s Compare observed and predicted E(r ref, f )s Use events with known M W to constrain g(r) from the source to the first observation distance (about 40 km for this study) Time Domain D( r ) Comparison of observed (color) and predicted (black) distance dependence (top figure) for frequencies of HZ. The lower panel shows that the simple model fit is good to about 0.15 log units to a distance of 300 km Propagation Model RVTDCAL1.0 COMMENT TEST OF SOURCE SPECTRUM KAPPA QETA QVELOCITY 3.5 SHEAR 3.5 DENSITY 2.8 DISTANCE SITE FMAX 100 RADIATION 0.55 FREE 2.0 PARTITION SIGMA 100 DURATION Input file to stochastic processing programs of Computer Programs in Seismology Excitation Modeling Motion at 40 km depends on propagation, Source spectra shape Source radiation Unknown geometrical spreading from source to first observation distance Can be resolved using independently determined moment magnitudes, especially for Mw