Key points are not available for this paper at this time.
With recent progress in resolution in global seismic mantle imaging provided by numerical wavefield computations using the Spectral Element Method and full waveform inversion, Jason Morgans suggestion from over 50 years ago that mantle plumes may be rooted at the core-mantle boundary (CMB) has been confirmed. Yet the imaged plumes present intriguing features that contrast with the classical thermal plume model and should inform our understanding of mantle dynamics. Among other features, they are broader than purely thermal plumes, and do not extend straight from the CMB to the corresponding hotspot volcanoes, but they are frequently deflected horizontally in the extended transition zone (400-1000 km depth), so that their lower mantle location can be significantly offset (as much as a 1000 km) from their surface expression. They appear to be thinner in the upper mantle. This, together with similar horizontal flattening observed in subduction zones suggests a change in the radial viscosity structure of the mantle that may occur deeper than usually assumed to be related to the 660 km phase change. The fattest plumes have been shown to be anchored within the perimeter of the large low shear velocity provinces (LLSVPs) and an increasing number of them appear to house mega-ultra low velocity zones within their roots. Moreover, in the upper mantle, they appear to be associated with regularly spaced low velocity channels aligned with absolute plate motion. We discuss these features in the light of recent regional imaging updates in the south Atlantic and beneath Yellowstone, contrasting the corresponding mantle plumes, and in particular showing mounting evidence that the LLSVPs are not compact piles extending high above the CMB, but rather a bundle of thermo-chemical plumes feeding secondary scale convection in the top 1000 km of the mantle.
Romanowicz et al. (Fri,) studied this question.