Insights from an analogous eruption: Bárðarbunga caldera collapse and re-inflation in Iceland during 2014-2019

Abstract:

Bárðarbunga volcano is situated beneath the Vatnajökull ice cap in the Icelandic highlands. In August 2014, melt intruded 48 km from the caldera in a lateral dike, before erupting at Holuhraun for 6 months until February 2015. During the eruption, the caldera collapsed as melt flowed out from beneath it, closely mirroring the events at Kīlauea in 2018. Intense seismicity accompanied the process, both along the dike path and within the caldera, recorded by a dense local network of seismometers deployed by the University of Cambridge and the Icelandic Meteorological Office (IMO). Since the eruption ceased, seismicity has continued in the caldera, and geodetic observations indicate that there has been significant re-inflation. Global Centroid Moment Tensor (GCMT) solutions for the largest earthquakes in the caldera show a corresponding reversal in polarity. Intriguingly – as at Kīlauea – these are all vertical Complex Linear Vector Dipole (vCLVD) solutions.

To investigate further, automatic hypocentre locations for earthquakes detected since 2015 have been manually refined through picking phase arrivals and P-wave first motion polarities. These observations, recorded by a significantly denser seismic network than was operational in 2014, have been inverted to obtain tightly constrained focal mechanisms, providing further insight into the style of faulting and therefore the geometry of the caldera fault. Focal mechanisms for large magnitude events along the northern rim of the caldera showed mostly normal faulting during the eruptive period as the caldera collapsed. However, from 2017 reverse faulting dominates. Most focal mechanisms in this study can be fitted by a double-couple moment tensor solution, in contrast to the vCLVD solutions presented in global moment tensor catalogues. The fault plane orientation is similar for eruptive and post-eruptive events, with a steeply inwards-dipping fault on the Northern rim, suggesting that the same faults are reactivated post-eruption in the opposite sense. This can be attributed to re-intrusion of melt into the magma storage region beneath the caldera, thereby causing re-inflation. These results have implications for our understanding of caldera collapse and re-inflation cycles and the interpretation of vCLVD moment tensors in caldera collapse events worldwide.