BROAD-BAND OBSERVATIONS OF UPPER-MANTLE SEISMIC PHASES IN NORTHERN AUSTRALIA AND THE ATTENUATION STRUCTURE IN THE UPPER-MANTLE

Sources in the earthquake belt through Indonesia and New Guinea, recor ded at a broadband Guralp CMG3 seismometer at the Warramunga array in the Northern Territory of Australia, give good coverage of propagation through the upper part of the mantle. The midpoints of the propagatio n paths lie along the northern margin of the Australian continent. Bey ond 18-degrees there is a significant difference in the frequency cont ent of P and S waves. The P waves remain high frequency, but the S wav es returned from the transition zone and below are of intermediate per iod (0.1-0.5 Hz), which would be difficult to record without a broadba nd instrument and a quiet site. The later branches associated with the 410 and 660 km discontinuities are clearly seen in individual seismog rams. The S waves recorded on the radial (SV) and tangential (SH) comp onents are of comparable quality, because the hard-rock recording site minimizes the influence of coupling to P on the radial component. We interpret the observed difference in frequency content of P and S wave s at transition-zone distances as the effect of a layer under the seis mically fast lid which possesses a large degree of shear dissipation. We quantify the observation in terms of the decay rate of the spectral ratio of S and P waves. That quantity can be interpreted in terms of path integrals of the difference in inverse Q for S waves and P waves, t(s) - t(p)*. We analyse 22 seismograms from the WRA broadband instr ument and a further four from portable broadband instruments deployed near Warramunga. The measured slope of the S/P spectral ratio is consi stently small out to 18-degrees epicentral distance, where it increase s dramatically. Assuming that shear dissipation dominates over bulk di ssipation, these measurements are consistent with an average quality f actor for S waves in the lid of the order of Q(s) = 1400 (Q(p) = 2800) on top of a highly attenuative asthenosphere of 200 km thickness with Q(s) = 100 (Q(p) = 200), which is underlain by a transition zone with Q(s) = 600 (Q(p) = 1200). The low-Q zone in the asthenosphere can equ ally well be modelled with a thinner layer with a lower Q, e.g. a laye r of 100 km thickness with Q(s) = 60. Frequency dependence of Q may re nder the above estimates of Q about 25% too large. These results compa re fairly well with published results from 0.3 Hz body-wave observatio ns under the Eurasian shield. Notable exceptions are that we obtain a somewhat lower value for Q(s) in the asthenosphere and a much higher v alue in the lid. The lid under northern Australia is unusually thick, 210 km, and our data constrain the low-Q zone in the asthenosphere to lie deeper than 210 km. It is difficult to explain the low observed va lues of t(s) - t(p) for paths contained in the seismic lid (lithospher e) in terms of artefacts related to differences in the source radiatio n of P and S waves or scattering effects. We therefore attribute them to intrinsic attenuation. The high value inferred for Q in the lid as compared with other studies in a similar frequency range could represe nt a regional variation and differences between individual shields, bu t could also be explained by a low Q(p)/Q(s) ratio (less than one), i. e. t(p) comparable with t(s). A low Q(p)/Q(s) ratio could be explained by a significant contribution from a thermoelastic dissipation mechan ism.