Upwelling and melting of the Iceland plume from radial variation of ²³⁸U-²³⁰Th disequilibria in postglacial volcanic rocks

New ²³⁸U-²³⁰Th disequilibria data by thermal ionisation mass spectrometry are presented for a comprehensive set of postglacial basaltic lavas from the neovolcanic zones in Iceland. The new data show a striking systematic decrease in ²³⁰Th excess towards central Iceland and the presumed centre of the Iceland plume. This finding would appear paradoxical if source composition was the main factor responsible for generating the ²³⁸U-²³⁰Th disequilibria, because generally main rift lavas erupted proximal to the plume should be generated from a melting column that initiates deeper in the garnet stability field, compared to the marginal rift zones. Preferential crustal interaction in central Iceland, where the crust is thickest, involving either old (>350 kyr) Icelandic crust or lower crustal melts, may provide a viable explanation for only part of the data variation, namely the moderately low ²³⁸U-²³⁰Th disequilibria found in the more evolved SE rift lavas. Moreover, there is no variation of ²³⁰Th excesses with degree of differentiation (Mg# or ppm Th) overall, or within individual rift systems, to indicate that crustal contamination causes the radial variation in ²³⁰Th excess. The ²³⁸U-²³⁰Th disequilibria variation is therefore ascribed to variable dynamic parameters in the melting regime induced by interaction of the Iceland plume with the rift systems. The higher ²³⁰Th excesses in alkalic off-rift lavas (Snæfellsnes Peninsula) (24±3%) compared to the main rift lavas (15±3%) is consistent with more garnet-rich lithologies dominating the bulk melt compositions away from the main rifts and indicates small-scale source heterogeneity beneath Iceland. The data are reconciled within a model in which mantle upwelling rates in the centre of the plume are significantly faster than at the margins, consistent with fluid dynamic predictions for a plume head. The radial variation observed in (²³⁰Th/²³⁸U) provides independent support that the centre of the Iceland plume is located beneath SE Iceland, as has been proposed from seismic tomographic studies. For a reasonable range of mantle porosities (Φ =0.05-0.2%) we can explain the Iceland data with a dynamic melting model, by relatively fast mantle upwelling rates in the centre (∼5-20 cm/yr), compared to those at the margins (∼1-4 cm/yr). The radial variation is also shown to be consistent with, though not requiring, a model of deep dehydration melting [Ito et al., Earth Planet. Sci. Lett. 165 (1999) 81-96]. In such a scenario, the generation of (moderately) low ²³⁸U-²³⁰Th disequilibria will be confined to the lowermost part of the melting column, which is characterised by fast upwelling and low porosity. For Φ values down to 0.05% in the lower part of the hydrous melting zone, moderately low ²³⁰Th excesses (5-10%) are likely to result, whereas higher ²³⁰Th excesses may arise for lower values of Φ.