Abstract Mantle convection causes the most important contribution to the geoid and dynamic topography. With mantle density inferred from high‐resolution tomography models and numerical methods solving the governing equations of viscous mantle flow, the modeled geoid can fit the observations well. However, there is still a large discrepancy between present dynamic topography predicted by mantle flow and residual topography deduced from observations: Especially, large negative topography is predicted in cratons, contrary to observations. In order to improve the fit of model dynamic topography compared to observations, in this study, we include chemical density anomalies in Earth's lithosphere to model geoid and dynamic topography. We also combine these anomalies with lateral viscosity variations and study the effect on predicted dynamic topography and geoid and investigate which density models would yield a good fit. In the sublithospheric mantle, under the assumption that the density anomalies are induced from temperature variations, we use temperature‐dependent viscosity. Our results show that by adding chemical density anomalies in the continental lithosphere, we can improve the correlation between dynamic topography and residual topography by 37%–59% for the three tomography models compared with one of the residual tomography models considered (RT1) and 1%–11% compared with the other (RT2). Similarly, correlation can also be improved and amplitude ratio brought closer to 1 by downscaling lithospheric density anomalies. Our results could provide a good reference for further studies of the Earth's mantle.
Cui et al. (Thu,) studied this question.