Feeding large eruptions : crystallisation, mixing and degassing in Icelandic magma chambers

Abstract

Iceland straddles the Mid-Atlantic Ridge and overlies a mantle hotspot. This tectonic setting produces voluminous tholeiitic magmas. Volcanism in Iceland is focussed along three neovolcanic spreading ridges. During the Holocene, the Eastern Volcanic Zone (EVZ) in southeast Iceland has been the most volcanically active and has been the site of several large (>6 km3) eruptions, including the only floodbasalt type eruption in recorded history, the 1783-84 Laki eruption. Three eruptions of large volume have been sampled for this study: the 1783-84 Laki eruption (15.1 km3); the 3,000-4,000 yBP Thjórsárdalur eruption (probably >4 km3); and the ~8,600 yBP Thjórsá eruption (>21 km3). The products of these eruptions have been analysed using a range of analytical techniques, with the specific aim of investigating crystallisation, degassing and mixing processes in the magma reservoirs that feed large eruptions. The Laki eruption has been the particular focus of this study. Samples from different parts of the lava flow show fine-scale variations in trace element concentrations and ratios. This compositional variation is not fully explained by fractional crystallisation processes, but is strongly controlled by crystal accumulation as whole-rock incompatible trace element concentrations show a linear, negative correlation with the mass fraction of crystals in the sample. Simple crystal accumulation models, however, fail to explain the compositional variation, and one explanation is that the homogeneous Laki melt mixed with varying proportions of a crystal mush that contained its own liquid. The results of thermobarometry calculations indicate that the erupted Laki liquid was in equilibrium with olivine, plagioclase and augite at 1-3 kb. Most of the crystals carried by the flow are too primitive to have crystallised from the erupted liquid and barometry calculations indicate that clinopyroxene crystallised at 3-7 kb. The majority of the large crystals hosted in the Laki basalt samples are therefore antecrysts that grew within the same magma plumbing system as the Laki carrier melt but are not in direct chemical equilibrium with it. This finding is verified by the fact that olivine crystals that are too magnesian to be in chemical equilibrium with the Laki whole-rock composition contain melt inclusions with average La/Yb values that are the same within error as the whole-rock values. The wide range of La/Yb values in melt inclusions hosted in the most magnesian (Fo86) olivine crystals in comparison to the least magnesian (Fo<74) indicates the initial variability of the Laki magma prior to concurrent crystallisation and extensive mixing, which acted to homogenise the carrier melt composition. The preservation of a wide range of La/Yb within the melt inclusions in comparison to the whole-rock composition, and a range of La/Yb values in different inclusions from the same crystal, indicates short timescales between melt inclusion entrapment and quenching during eruption. Melt inclusion studies also reveal the dissolved volatile content of the Laki magma at the onset of olivine crystallisation, although the majority of H2O concentrations have almost certainly been reset by low pressure diffusive exchange with the host crystal or surrounding magma. Comparison of the behaviour of volatiles with that of incompatible elements in the melt inclusions indicates that CO2 was degassing during olivine crystallisation, but S, F and Cl were not. New estimates of total volatile loading to the atmosphere during the eruption based on melt inclusion volatile concentrations show SO2 and HCl loading comparable to previous estimates, but higher HF loading. Mass balance calculations show that the observed H2O and CO2 concentrations of melt inclusions hosted in olivines in chemical equilibrium with the Laki whole-rock composition are ~50% and ~93% lower respectively than would be expected if no pre-eruptive degassing of the magma reservoir had occurred, meaning that pre-eruptive degassing of H2O and CO2 from the magma must have been significant. Lava flows from Thjórsá are more compositionally variable than those from Laki, and have different average major and trace element compositions. Compositional variation within the Thjórsá whole-rock composition is explained by a combination of source variation and fractional crystallisation, and, unlike Laki, is not strongly controlled by crystal accumulation.