The Icelandic mantle contains a range of lithologies associated with the depleted upper mantle, a mantle plume, and recycled oceanic lithosphere but the precise nature of depleted and enriched components in the mantle and their relative contributions to melt production remain poorly constrained. In this study, we collect new olivine- and plagioclase-hosted melt inclusion data and compile this with existing literature data to investigate the relative contributions from different mantle lithologies to basaltic magmas erupted in Icelandic flank zones and neovolcanic zones by modelling the melting of a heterogeneous mantle and subsequent mixing of derived melts. We find that observed melt inclusion compositions from off-axis flank zones are best explained as homogenized mixtures of pyroxenite- and lherzolite-derived melts produced at depths around 80-93 km, by which point lherzolite has only experienced a low degree of melting whereas the pyroxenite lithology has melted extensively. These melts represent the onset of channelization in the mantle and are transported rapidly to the surface without input from shallower melts. Melt compositions from the on-axis neovolcanic zones and off-axis Öræfajökull, are produced by mixing this deep melt component with higher degree lherzolite melts produced at shallower depths, between 57-93 km. Proportions of shallow lherzolite-derived melts and deep homogenized melt vary, but the lowest contribution from the deep homogenized melt is seen in the Northern Volcanic Zone. Ourresults support a model whereby deep melts mix until melt channelization starts in the mantle, after which binary mixing between the homogenized deep melt and shallower fractional melts occurs.

In order to reconcile petrological and geophysical observations in the temporal domain, the uncertainties of diffusion timescales need to be rigorously assessed. Here we present a new diffusion chronometry method: Diffusion chronometry using Finite Elements and Nested Sampling (DFENS). This method combines a finite element numerical model with a nested sampling Bayesian inversion meaning the uncertainties of the parameters that contribute to diffusion timescale estimates can be rigorously assessed, and that observations from multiple elements can be used to better constrain a single timescale. By accounting for the covariance in uncertainty structure in the diffusion parameters, estimates on timescale uncertainties can be reduced by a factor of 2 over assuming that these parameters are independent of each other. We applied the DFENS method to the products of the Skuggafjöll eruption from the Bárðarbunga volcanic system in Iceland, which contains zoned macrocrysts of olivine and plagioclase that record a shared magmatic history. Olivine and plagioclase provide consistent pre-eruptive mixing and mush disaggregation timescales of less than 1 year. The DFENS method goes some way to improving our ability to rigorously address the uncertainties of diffusion timescales, but efforts still need to be made to understand other systematic sources of uncertainty such as crystal morphology, appropriate choice of diffusion coefficients, growth, and the petrological context of diffusion timescales.