Main

Many responses to the criticism of the plume model1 have been published3,4 and we shall not reiterate these arguments, which refute a low mantle temperature beneath hotspots based on petrology. First, Anderson and Natland argue that buoyancy beneath ‘hotspots’ could be explained by fertile blobs. In general, buoyancy is driven by contrast in temperature, fertility and melt fraction5. As fertility is usually associated with a higher iron content (and hence a greater density), the effect of melting out the fertile (dense) component will not increase the buoyancy relative to ambient mantle. Thus, the only source of extra buoyancy one can consider in the model of Anderson and Natland1 is buoyancy due to melt retention (or melt fraction φ). In this respect, U-series provide clear clues on the melt fraction during melting and all studies show that it should not exceed a few permil (ref. 2). Hence, the effect of a fertile source on the melt fraction (for φ = 0.003) should in fact be limited and be equivalent to an excess temperature of 10 °C. For these reasons, we do not think a fertile blob can be the source of buoyancy beneath hotspots.

Second, if the increased melting rates were due to the presence of fertile blobs, then there should be a correlation between clear indices of enrichment, such as radiogenic isotopes, and U-series activity ratios. For two localities we used (the Galapagos and the Azores), such a correlation does not exist. Furthermore, as fertility is associated with enrichment in water (at least in the Galapagos and the Azores), this will decrease the melting rate of the blob by at least a factor of ten, which could easily compensate the effect of increased fertility on melting rates. These combined effects would fail to explain the U-series observations. We do not believe that fertile blobs can be a general explanation.

Third, although the mantle upwelling rates determined using U-series may not be entirely reliable in absolute terms, when compared with mid-ocean-ridge settings, they clearly indicate faster upwelling near the centre of the plume than beneath adjacent ridges. This is evident in the Azores region2,6 (where the spreading rate is slow) but also in the Galapagos (A.S. et al., manuscript in preparation), where the spreading rate is faster. The case of the Azores is particularly interesting because it shows that despite the relatively small buoyancy flux, the mantle is upwelling faster than beneath the nearby Mid-Atlantic Ridge. Furthermore, estimates of mantle upwelling velocities beneath Hawaii7 do show that velocities of up to several metres per year are estimated.

Last, the temperatures of hotspot magmas have been widely discussed (see ref. 8 for a critical review of ridge and hotspot mantle temperatures) and these consistently indicate that temperatures are hotter than that of the ambient mantle, even when the variability of normal potential temperatures of mid-ocean-ridge basalts (120 °C; ref. 7) is taken into account. Although we did not focus on a comparison with ridges2, U-series unfailingly indicate higher temperatures beneath hotspots than at mid-ocean ridges9.

Anderson and Natland1 favour the idea that all plumes originate from the core–mantle boundary. But must they? The elegant experiments of Davaille et al.10 show that imagining plumes as narrow conduits from the core–mantle boundary might be a narrow view of convective patterns in the mantle.