J. Boatwright, G. L. Choy, L. C. Seekins (all at the USGS)

We present a spectral technique for estimating the radiated seismic energy using recordings of large earthquakes at distances km, correcting for frequency-dependent seismic impedance (that is,, the density times the S-wave velocity), geometrical and anelastic attenuation as

We classify the stations by their NEHRP site classes (hard rock, rock, stiff soil, and deep soil) and use characteristic velocity structures for each site class to estimate the seismic impedance as a function of frequency, assuming s for the near-surface attenuation. The relative amplification for the deep soil sites exceeds the amplification predicted from the average velocity structure at periods from 2 to 5 s. Regressions of strong motion data indicate that below 0.5 s, from 1 to 3 s, and above 5 s, corresponding to body-wave, Lg-wave, and surface-wave propagation, respectively. These strong motion regressions also suggest that the connection distance for the geometrical attenuation function is bounded as km. with km/s roughly fits the anelastic attenuation. We apply this analysis to 64 regional recordings of the 1999 Hector Mine earthquake, obtained from stations operated by TriNet, CDMG, and USGS. The radiated seismic energy varies as 1.8-3.8xergs for . Directivity enhances the energy estimates along-strike by a factor of 1.5 to the NNW and 2 to the SSE. This estimate of radiated energy agrees closely with the teleseismic estimate of 2.4-3.2x ergs, obtained from 17 stations, computed using the appropriate impedance function for P-waves. The teleseismic energy estimate varies as a function of the assumed teleseismic attenuation.