DETERMINATION OF SITE-DEPENDENT SCATTERING-Q FROM P-WAVE CODA ANALYSIS WITH AN ENERGY-FLUX MODEL

The seismic coda of deep events is mainly generated by near-receiver scattering in the lithosphere and therefore contains information about its small-scale structure. A theoretical model which describes coda excitation and decay rate in terms of scattering and anelastic attenuation of the medium is the energy-flux model. Here, a version of the model appropriate for a wavefront propagating through a scattering layer above a homogeneous half space is used to interpret short period data from several stations of the Global Digital Seismic Network which are available on CD-ROMs. Data from events with depths greater than 300 km and epicentral distances less than 60-degrees allow P-coda observations for about 70 s after the first onset undisturbed by later arrivals. At nine stations in the circumpacific area estimates of scattering Q (Q(s)) are obtained for the frequency range 0.5 Hz to 4 Hz. Q(s) differs significantly between sites: strong scattering (Q(s) almost-equal-to 100-200) was found under stations MAJO, TATO, GUMO and AFI. Relatively weak scattering (Q(s) > 500) occurs at NWAO and CTAO, which are located on stable continental areas. The variation of Q(s) with frequency qualitatively agrees with Q(s) obtained from single-scattering theory for von Karman type autocorrelation functions of the heterogeneities. At several sites correlation distances a and rms velocity fluctuations epsilon could be determined. At other sites only the ratio epsilon/a(nu) was obtained, where nu is the order of the von Karman autocorrelation function. Anelastic Q (Q(i)) could only be estimated at TATO, GUMO and SNZO. However, resolution is poor, because the coda decay is dominated by the leaking of scattered energy into the mantle and not by anelastic effects. There is an indication that scattering (except at TATO) is limited mainly to the uppermost 30 to 50 km.