Seismic attenuation in the eastern Australian and Antarctic plates, from multiple ScS waves

The attenuation of seismic shear waves in the mantle beneath the eastern Australian and Antarctic plates is analyzed using a large data set of multiple ScSn waves, reflected n times at the core-mantle boundary and (n-1) times at the surface. The data are the transverse components of deep earthquakes from the subduction zones north and east of Australia, recorded at stations in Antarctica, Australia, Indonesia, New Caledonia and New Zealand. The data are filtered with narrow band-pass filters at 5 frequencies in the range 0.013-0.040 Hz. The ScSn+1/ScSn amplitude ratios of successive ScS phases are compared to the ratios computed for synthetic seismograms for the same paths and same focal mechanisms, in order to eliminate the effects of source radiation and geometric attenuation. The synthetic seismograms are computed from a summation of toroidal modes for the 1-D reference model PREM (Dziewonski & Anderson 1981). The observed to computed spectral ratios appear consistent for similar paths. They reveal that the attenuation is not frequency dependent, that the contribution of scattering to attenuation is low, and that the PREM model is a valuable reference model for the study region at the considered frequencies. An inversion of the data at 0.026 Hz is performed to retrieve the quality factor Q in the upper mantle, in regions defined using a priori constraints inferred from seismic shear velocities. Q-values close to those of PREM are found beneath the Australian and Antarctic cratons, lower values beneath the Eastern Australian Phanerozoic margin, and very low values beneath the oceanic region between Australia and Antarctica, where ridges and a triple junction are present. The Australian-Antarctic Discordance along the South-Indian ridge appears as an exception with a Q-value close to those of stable continents. The highest Q-values are found beneath the subduction zones, a feature which is not apparent in global attenuation models possibly because of its narrow lateral extension, and because it extends at depths larger than those sampled by surface waves. Despite limitations due to the uneven distribution of the ScSn bounce points at the surface and to the difficulty of collecting a large number of high quality data, our approach appears very promising. It is complementary to the more widely used determination of seismic attenuation using surface waves because it provides increased depth coverage, and a broader spectral coverage. It therefore has a considerable potential in future investigations of mantle structure and dynamics.