Constraints on mantle viscosity from convection models with plate motion history

The Earth's long-wavelength geoid and dynamic topography are mainly controlled by the mantle buoyancy and viscosity structure. Previous dynamical models for the geoid provide constraints on the 1-D mantle viscosity, using mantle buoyancy derived from seismic topography models. However, it is a challenge in these studies on how to convert seismic velocity to density anomalies and mantle buoyancy. Furthermore, these studies provide constraints only on relative viscosity variations but not on absolute magnitude of viscosity. In this study, we formulate time-dependent 3-D spherical mantle convection models with imposed plate motion history and seek constraints on mantle viscosity structure for both its radial relative variations and its absolute magnitude (i.e., Rayleigh number), using the geoid from the convection models. We found that the geoid at intermediate wavelengths of degrees 4-9 is mainly controlled by the subducted slabs in the upper mantle and the upper part of lower mantle that result from subduction from the last 50 Myr or the Cenozoic. To fit the degrees 4-9 geoid, we need viscosity contrast β defined as the ratio of the lower mantle viscosity and the asthenospheric viscosity to be larger than 2000 and Ra to be 1e8 (defined by the Earth's radius). The best fit model leads to 57% variance reduction and 76% correlation between the model and the observations. However, the long-wavelength geoid at degrees 2-3 is controlled by the lower mantle structure which requires much longer time scale to develop, as seen from our modeling. The preferred viscosity structure and Rayleigh number as constrained by the Cenozoic plate motion and the degrees 4-9 geoid no longer provide adequate fit to the geoid in models with the plate motion history for the last 450 Myr. The degrees 4-9 geoid amplitude is smaller for the models with longer plate motion history and a smaller Ra is required to fit the observation. In order to satisfy the relative amplitude between degrees 2-3 and degrees 4-9 geoid, either a gradually increase of viscosity in the upper part of lower mantle or larger thermal expansivity in the lower mantle is needed. We also consider thermo-chemical models to examine the effects of the African and Pacific thermochemical piles (i.e., LLSVPSs) on the geoid and the inferred mantle viscosity and Ra.