Main

Revised estimates of temperature, heat flow and buoyancy at ridges and hotspots, and developments in plume3,13 and plate theory2,9,10 are relevant to the conclusions of Bourdon et al.12. No near-ridge hotspot has anomalous temperature7 and no hotspot has a significant heat flow anomaly5,6. The only active hotspot with a petrology-based temperature higher than mid-ocean-ridge basalt is, arguably, Hawaii14. Hawaiian lithosphere, however, is 140 °C colder than predicted by plume models15. It is no longer generally argued that all, or even many, hotspots are due to deep plumes, but lower mantle conditions are considered by many investigators as essential to the rationalization of observations12. However, tectonics and shallow low-melting heterogeneities, rather than excess temperatures, may be responsible for hotspots2,11.

To satisfy global constraints, narrow plumes must have ascent rates and temperatures much greater than the broad upwellings associated with plate tectonics and normal mantle convection3,13. Required excess temperatures are 200–300 °C and velocities are one-half to tens of metres per year. The absence of evidence for such high values has been rationalized in several ways12, but, taken at face value, supports a plate tectonic and lithologic, rather than thermal, explanation for hotspots.

Uranium-series geochemistry may provide insight that is independent of previous arguments12. Bourdon et al.12 interpret U-series model data as independent evidence for thermal plumes and evidence that hotter plumes are stronger. Plumes are modelled as if they originated in a deep thermal boundary layer12 heated from below, although the data do not constrain anything deeper than about 60–100 km. Shallow processes9,10,11 and homologous temperature (TH) variations2,3 are not considered (plume simulations use a homogeneous mantle); modelling assumptions and parametrizations to date do not permit a shallow or non-thermal interpretation. Shallow chemical buoyancy can mimic effects of temperature, including isotope gradients. Could U-series data discriminate between buoyant decompression melting of shallow fertile blobs1,10 and deep thermal plumes? Model velocities are comparable to plate and passive upwelling velocities, and much less than 0.5 m yr-1.

Melt retention buoyancy in low-melting silicates (fertile blobs) is the equivalent of 300–600 °C temperature excess4; lowering the solidus or raising the temperature provide equivalent buoyancies3. Fertile blobs absorb mantle heat and turn into buoyant diapirs2. High-TH blobs can melt deeper, rise faster and retain melt longer than subsolidus ambient mantle (low-TH) rising passively under ridges. Lateral flow of material beneath the lithosphere has been taken as evidence for plumes5, but is equally consistent with spreading of chemically buoyant or high-TH blobs.

Hawaii has conflicting petrological8,14, fertility and tectonic10,11 interpretations and cannot be understood in terms of temperature alone. Eruption rates peak at the large-offset 300-km-wide Molokai fracture zone11. A plausible explanation involves underplating, decompression melting of ‘eclogite’ at lithosphere steps10 and ascent through Molokai fracture zone conduits. The time between melt extraction and eruption, under these conditions, is unlikely to be similar to ridges.

If hotspots involve tectonic and TH variations, then conventional fluid dynamic constraints on depth, temperature and velocity are removed. Temperatures of thermal plumes cannot be arbitrarily low3,13. Processes that utilize composition, internal heating, lateral flow, stress and ponding to localize volcanism can operate at lower temperatures1,9. The definition of mantle plume, to be useful, should recognize the difference.