Flowmodel | Platevelocities* | Free-airgravity | Geoid | Dynamictopography | CMBellipticity** |

S20RTS + L4^{†} | 44% | −7% | −60% | −50% | 1440 m |

S20RTS_{f−s} + L4^{‡} | N/A | 12% | 64% | −17% | 1430 m |

S20RTS_{inv}+ V2^{§} | 48% | 28% | 57% | 40% | 450 m |

TX2008 + V2^{¶} | 80% | 74% | 90% | 73% | 540 m |

TX2008 + L4^{||} | 60% | 50% | 41% | 69% | 690 m |

*Fits between the predicted and NUVEL-1A vector field of plate velocities (*48*) in the NNR frame are calculated on a global 5° × 5° grid.

†Fits are calculated for the flow model using a posteriori density conversion of tomography model S20RTS used by Behn *et al*. (*45*) and L4 viscosity model (Fig. 6). Plate motions are viscously coupled to underlying mantle flow.

‡Fits are calculated for a theoretical free-slip surface boundary condition (no surface plates) using identical density and viscosity inputs as in the previous footnote (see previous model above).

§Fits are calculated for the flow model using optimized density-velocity scaling of model S20RTS from Occam inversion of surface geodynamic data (*42*) and using V2 viscosity model (Fig. 5). Plate motions are viscously coupled to underlying mantle flow.

¶Fits are calculated for the flow model using the TX2008 density heterogeneity model and the V2 viscosity model. Plate motions are viscously coupled to underlying mantle flow.

||Fits are calculated for the flow model using the TX2008 density heterogeneity model and the L4 viscosity model. Plate motions are viscously coupled to underlying mantle flow.

**The space-geodetic inference of excess CMB ellipticity is 400 m (*76*).