CONSTRAINING UPPER-MANTLE ANELASTICITY USING SURFACE-WAVE AMPLITUDE ANOMALIES

We present an iterative method to constrain lateral variations in surface wave attenuation using long-period surface wave amplitude anomalies. The method acts to isolate the anelastic signal from elastic focusing effects, yielding largely unbiased estimates of lateral variations in the inverse seismic quality factor, q(omega) = Q-1(omega), of the surface wave. In the zeroth iteration, linearized ray theory motivates the construction of a reduced datum, using measurements from four consecutive surface wave orbits, which is insensitive to elastic heterogeneity, an operation which requires no a priori knowledge of elastic structure. Synthetic experiments using both ray theoretic formalisms and normal-mode calculations reveal that significant levels of elastic bias remain in the reduced data due to deviations from linearized ray theory. In further efforts to eliminate elastic bias, the remaining elastic signal in each reduced datum is predicted and subtracted using accurate forward theory and existing aspherical elastic mantle models. This operation is the first iteration of a non-linear inversion in which the data at each iteration are the residuals between the observed anomalies and the anomalies predicted for a fixed phase velocity model and updated attenuation model. The zeroth and first iterations are performed with 1610 vertical and 790 longitudinal component seismograms from 144 events. Heterogeneity maps of Rayleigh wave attenuation deltaq(omega, theta, phi) are retrieved for the even degrees 2, 4 and 6 of a spherical harmonic expansion in the period range of 150-300 s (S-0(25)- S-0(60)). These surface wave attenuation maps are analogous to elastic phase velocity in their radial averaging of, and linear relationship to, intrinsic heterogeneity and are inverted for upper mantle anelastic structure. Under the physically plausible assumption that intrinsic elastic and anelastic structure are correlated at every depth, the surface wave attenuation maps are well explained by a source region of anelastic heterogeneity that is radially localized in the shallow mantle (100-300 km) in a region we infer to be near the solidus temperature.