Does active mantle upwelling help drive plate motions?

Earth's lithospheric plates are driven partly by shear tractions exerted on their base by viscous coupling to convective mantle flow. While downwelling flow associated with descent of subducted lithosphere has been well established as a driver of plate motions, the plate-driving role of upwelling flow is more controversial. We used a numerical mantle flow model to predict present-day plate motions driven by various combinations of upwelling and downwelling flow. We obtain a good fit to observed plate motions using flow driven only by downgoing mantle slabs, whose densities can be inferred either from the tectonic history of subduction or from positive velocity anomalies extracted from mantle tomography. Introducing upwelling flow (driven by negative density anomalies inferred from slow seismic velocity anomalies) to this slab-driven flow field generally degrades the fit to plate motions, and an acceptable fit that includes active upwelling requires amplitudes of mantle density heterogeneity that are at the low end of the accepted range (tomography to density conversion factor = 0.05 g cm−3 km−1 s). Because such small amplitudes lead to a deficit of slab material in the mantle, we infer that downwelling flow is a more significant driver of plate motions than upwelling flow. This conclusion can be reconciled with the presence of low-velocity anomalies in the lower mantle if upwelling flow does not couple effectively to plate motions, or if these low-velocity anomalies represent chemical differentiation of the lower mantle and thus do not drive upwelling flow.