CONVERTING MANTLE TOMOGRAPHY INTO MASS ANOMALIES TO PREDICT THE EARTHS RADIAL VISCOSITY

In this study, several mantle tomography models are used as input to constrain the amplitude of mass anomalies and viscosity variations with depth in the mantle. For this purpose, we compute the dynamical state of an incompressible mantle overlaid with rigid plates. The best profiles for the viscosity and for the seismic velocity (upsilon) to density (rho) conversion factor, partial derivative rho/partial derivative upsilon, are obtained through a Tarantola-Valette inversion by matching the predicted and observed geoid and the poloidal and toroidal surface velocities, and by minimizing the dynamical topography induced by the lower mantle. The recently available second generation of tomography models yields the best results, i.e. 70% of variance reduction for the geoid. An unusual trend of partial derivative rho/partial derivative upsilon with depth is found in the lower mantle. We show that the inversion is more dependent on a priori parameters than on the model, and that there is a strong trade-off between partial derivative rho/partial derivative upsilon and viscosity structures in this kind of modeling. Finally, we discuss the origin of deep mass anomalies in terms of past subduction.