On 16 August 2014 an unusual sequence of earthquakes began near the southeastern rim of the ice-covered Bárðarbunga caldera in central Iceland. Over the course of two weeks a dyke propagated 48 km beneath the glacier northeastwards and into the Holuhraun lava field, where it erupted for six months, becoming the largest eruption in Iceland for over 200 years. During this time, a gradual, incremental caldera collapse took place in the central volcano. The rifting episode was captured both geodetically and seismically. In this thesis, I analyse the seismic response to the event, both due to the dyke propagation, and the subsequent caldera collapse. This gives an insight into the underlying processes controlling rifting events, and the nature of the responding crust.

The Cambridge seismic network recorded the 2014-15 Bárðarbunga-Holuhraun rifting episode in exceptional detail. I discuss the deployment and operation of this dense seismic network in the remote Icelandic highlands, as well as the campaign deployments on the volcano caldera, on the glacier (above the dyke path) and around the eventual eruption site, as a first response to the crisis. Using this dataset I have accurately located, and analysed, 47,000 earthquakes during the pre-intrusive, intrusive, eruptive and post-eruptive periods. Approximately 4,000 of the recorded earthquakes are associated with the caldera collapse, delineating faults accommodating the subsidence and showing good correlation with geodetic data. The seismicity reveals activation of both inner and outer caldera faults with 60 inward dipping planes on the northern and southern side, indicating a symmetric caldera structure. Detailed analysis of the earthquake source mechanisms shows that 90% can be explained by a double-couple solution, which is in contrast to results from previous studies of Bárðarbunga. I find the dominant failure mechanism during the collapse to be steep normal faulting, with sub-vertical P-axes, striking sub-parallel to the caldera rim. The northern and southern sides of the caldera, however experienced very different seismicity rates, highlighted by the order of magnitude difference in the cumulative seismic moments. The southeastern part of the caldera, whilst experiencing less activity, shows a mixture of failure mechanisms, owing to the interaction of the caldera collapse and the dyke exit. Therefore, this thesis presents evidence of a complex asymmetric caldera collapse, not controlled by a single caldera ring fault.

Of the 47,000 earthquakes located, 31,000 delineate the segmented, lateral dyke intrusion as it fractured a pathway through the crust, utilizing pre-existing weaknesses. Despite the extensional rift setting, the dyke emplacement generated exclusively doublecouple earthquakes. At the leading edge of the propagation, earthquake source mechanisms show exclusively strike-slip faulting, in contrast to the conventional model of normal faulting above a propagating dyke. I observe right-lateral strike-slip faulting as the dyke propagates to the NE, and an abrupt change to left-lateral strike-slip faulting as the dyke turns and propagates in a more northerly direction into the northern volcanic zone. This shows that the direction of fault motion is determined by the opening of the dyke, rather than by the regional extension.

I am also able to define the thickness of the seismogenic crust under Bárðarbunga as 7 km, based on the depth extent of observed seismicity. The bulk of the seismicity in the volcano is located at 1-4 km below the surface, whereas the dyke exited the caldera at 4-6 km depth, propagating at 6-8 km b.s.l. I hypothesise that the magma storage region is likely located at 4-6 km b.s.l. (6-8 km below the caldera surface), just below the most active caldera seismicity and at similar depth levels to the dyke. Thus, this thesis details the melt distribution and movement at depth from a large basaltic central volcano, and the coupled deformation of the subsiding caldera with the dyke intrusion and eruption.