| We combine large data sets of surface-wave phase anomalies, long-period waveforms, and body-wave travel times in order to provide new constraints on the anisotropic shear-wave velocity structure of the Earth's mantle. The waveform inversion is performed using a new and more accurate method developed to correct seismograms for non-linear crustal effects. Starting with an isotropic spherically symmetric earth model, we build a new one-dimensional, transversely isotropic reference model by independently constraining variations in five elastic parameters and density. Using this new reference model, we invert the data for a whole-mantle model of shear-wave velocity and investigate lateral anisotropic variations at all depths in the mantle. Finally, we develop a technique that allows us to calculate a high-resolution tomographic model of a specific region as a perturbation with respect to the low-resolution global model, and implement this technique to study the structure beneath Eurasia.; Our new reference model fits the data as well as PREM, although it does not contain the 220-km discontinuity present in PREM. We find the average shear-wave anisotropy to be strongest at a depth of about 125 km and the parameter eta to be very similar to that in PREM. The strong fast-velocity anomalies beneath stable parts of continents, which may represent the continental lithosphere, extend down to a depth of about 200 km if waveform data are corrected for crustal effects using the new non-linear method. In contrast, if the standard, less accurate, linear approach is used, significantly thicker fast-velocity anomalies beneath continents are observed. With the non-linear crustal corrections, the strongest decrease in the absolute shear-wave velocity appears within depths between 150 and 250 km beneath cratons in northern Eurasia. Allowing for radial anisotropy in the transition zone does not improve data fit. The depth of about 650 km is characterized by a significant change in the power spectrum of heterogeneity, which suggests a change in the flow pattern between the upper and lower mantle. We find that allowing for anisotropic variations at the bottom of the mantle improves the data fit. However, constraining such variations is difficult since they strongly trade off with the isotropic variations. |
