Driven to collapse: Volcanic eruptions occur frequently, but only rarely are they large enough to cause the top of the mountain to collapse and form a caldera. Gudmundsson et al. used a variety of geophysical tools to monitor the caldera formation that accompanied the 2014 Bárdarbunga volcanic eruption in Iceland. The volcanic edifice became unstable as magma from beneath Bárdarbunga spilled out into the nearby Holuhraun lava field. The timing of the gradual collapse revealed that it is the eruption that drives caldera formation and not the other way around. Science , this issue p. 262

The 6-month long eruption at Holuhraun (August 2014–February 2015) in the Bárðarbunga-Veiðivötn volcanic system was the largest effusive eruption in Iceland since the 1783–1784 CE Laki eruption. The lava flow field covered ~84 km2 and has an estimated bulk (i.e., including vesicles) volume of ~1.44 km3. The eruption had an average discharge rate of ~90 m3/s making it the longest effusive eruption in modern times to sustain such high average flux. The first phase of the eruption (August 31, 2014 to mid-October 2014) had a discharge rate of ~350 to 100 m3/s and was typified by lava transport via open channels and the formation of four lava flows, no. 1–4,which were emplaced side by side. The eruption began on a 1.8 km long fissure, feeding partly incandescent sheets of slabby pāhoehoe up to 500 m wide. By the following day the lava transport got confined to open channels and the dominant lava morphology changed to rubbly pāhoehoe and ‘a’ā. The latter became the dominating morphology of lava flows no. 1–8. The second phase of the eruption (Mid-October to end November) had a discharge of ~100–50 m3/s. During this time the lava transport system changed, via the formation of a b1 km2 lava pond ~1 km east of the vent. The pond most likely formed in a topographical low created by a the pre-existing Holuhraun and the newHoluhraun lava flow fields. This pond became themain point of lava distribution, controlling the emplacement of subsequent flows (i.e. no. 5–8). Towards the end of this phase inflation plateaus developed in lava flowno. 1. These inflation plateaus were the surface manifestation of a growing lava tube system, which formed as lava ponded in the open lava channels creating sufficient lavastatic pressure in the fluid lava to lift the roof of the lava channels. This allowed new lava into the previously active lava channel lifting the channel roof via inflation. The final (third) phase, lasting from December to end-February 2015 had a mean discharge rate of ~50 m3/s. In this phase the lava transport was mainly confined to lava tubes within lava flows no. 1–2, which fed breakouts that resurfaced N19 km2 of the flow field. The primary lava morphology from this phase was spiny pāhoehoe, which superimposed on the ‘a’ā lava flows no. 1–3 and extended the entire length of the flow field (i.e. 17 km). Thismade the 2014–2015 Holuhraun a paired flow field,where both lava morphologies had similar length. We suggest that the similar length is a consequence of the pāhoehoe is fed from the tube systemutilizing the existing ‘a’ā lava channels, and thereby are controlled by the initial length of the ‘a’ā flows.The work was financed with crisis response funding from the Icelandic Government along with European Community's Seventh Framework Programme Grant No. 308377 (Project FUTUREVOLC) and along with the Icelandic Research fund, Rannis, Grant of Excellence No. 152266-052 (Project EMMIRS). Furthermore, Vinur Vatnajökuls are thanked for support.Peer Reviewe

Geochemical Perspectives Letters is a new journal of the European Association of Geochemistry.

In the Arctic region, Iceland is an important source of dust due to ash production from volcanic eruptions. In addition, dust is resuspended from the surface into the atmosphere as several dust storms occur each year. During volcanic eruptions and dust storms, material is deposited on the glaciers where it influences their energy balance. The effects of deposited volcanic ash on ice and snow melt were examined using laboratory and outdoor experiments. These experiments were made during the snow melt period using two different ash grain sizes (1 ϕ and 3.5 ϕ) from the Eyjafjallajökull 2010 eruption, collected on the glacier. Different amounts of ash were deposited on snow or ice, after which the snow properties and melt were measured. The results show that a thin ash layer increases the snow and ice melt but an ash layer exceeding a certain critical thickness caused insulation. Ash with 1 ϕ in grain size insulated the ice below at a thickness of 9–15 mm. For the 3.5 ϕ grain size, the insulation thickness is 13 mm. The maximum melt occurred at a thickness of 1 mm for the 1 ϕ and only 1–2 mm for 3.5 ϕ ash. A map of dust concentrations on Vatnajökull that represents the dust deposition during the summer of 2013 is presented with concentrations ranging from 0.2 up to 16.6 g m⁻².

Hydrothermal explosions frequently occur in geothermal areas showing various mechanisms and energies of explosivity. Their deposits, though generally hardly recognised or badly preserved, provide important insights to quantify the dynamics and energy of these poorly understood explosive events. Furthermore the host rock lithology of the geothermal system adds a control on the efficiency in the energy release during an explosion. We present results from a detailed study of recent hydrothermal explosion deposits within an active geothermal area at Kverkfjöll, a central volcano at the northern edge of Vatnajökull. On August 15th 2013, a small jökulhlaup occurred when the Gengissig ice-dammed lake drained at Kverkfjöll. The lake level dropped by approximately 30 m, decreasing pressure on the lake bed and triggering several hydrothermal explosions on the 16th. Here, a multidisciplinary approach combining detailed field work, laboratory studies, and models of the energetics of explosions with information on duration and amplitudes of seismic signals, has been used to analyse the mechanisms and characteristics of these hydrothermal explosions. Field and laboratory studies were also carried out to help constrain the sedimentary sequence involved in the event. The explosions lasted for 40–50 s and involved the surficial part of an unconsolidated and hydrothermally altered glacio-lacustrine deposit composed of pyroclasts, lavas, scoriaceous fragments, and fine-grained welded or loosely consolidated aggregates, interbedded with clay-rich levels. Several small fans of ejecta were formed, reaching a distance of 1 km north of the lake and covering an area of approximately 0.3 km2, with a maximum thickness of 40 cm at the crater walls. The material (volume of approximately 104 m3) has been ejected by the expanding boiling fluid, generated by a pressure failure affecting the surficial geothermal reservoir. The maximum thermal, craterisation and ejection energies, calculated for the explosion areas, are on the order of 1011, 1010 and 109 J, respectively. Comparison of these with those estimated by the volume of the ejecta and the crater sizes, yields good agreement. We estimate that approximately 30% of the available thermal energy was converted into mechanical energy during this event. The residual energy was largely dissipated as heat, while only a small portion was converted into seismic energy. Estimation of the amount of freshly-fragmented clasts in the ejected material obtained from SEM morphological analyses, reveals that a low but significant energy consumption by fragmentation occurred. Decompression experiments were performed in the laboratory mimicking the conditions due to the drainage of the lake. Experimental results confirm that only a minor amount of energy is consumed by the creation of new surfaces in fragmentation, whereas most of the fresh fragments derive from the disaggregation of aggregates. Furthermore, ejection velocities of the particles (40–50 m/s), measured via high-speed videos, are consistent with those estimated from the field. The multidisciplinary approach used here to investigate hydrothermal explosions has proven to be a valuable tool which can provide robust constraints on energy release and partitioning for such small-size yet hazardous, steam-explosion events