Icecap and crustal deformation associated with the 2014-2015 Bardarbunga dike intrusion episode revealed by pixel-offset approach

The 2014-2015 Bardarbunga dike intrusion episode is one of the largest events in Iceland. During the event, a seismic swarm, fissure eruption and caldera collapse were reported by previous studies. The seismic observation in the previous studies indicated a migration of hypocenters from Bardarbunga caldera and Holuhraun where the fissure eruption was observed at 5 km north of the northern edge of the Vatnajokull icecap. The airplane altimetry has indicated 66 m subsidence of the caldera. These results indicate dike intrusion caused by the caldera collapse of Bardarbunga. Previous GNSS and InSAR revealed the crustal deformation associated with the event, showing 2 m horizontal displacements toward WNW-ESE that are nearly identical to the direction of the plate motion. However, InSAR observations have been lacking data across the center of the displacement field due to the decorrelation problem. Moreover, the crustal deformation under the icecap remained unclear. In this study, we applied pixel offset approach to Cosmos-SkyMed and TerraSAR-X images in order to detect the displacement signal at both the ground and the icecap. The pixel offset data revealed the crustal deformation signals that were not shown by interferograms due to the unwrapping error. Moreover, the icecap deformation at the northern edge was also detected, which was interpreted as a combination of the co-rifting deformation and the steady-state ice flow. The 3D displacements inferred from pixel offset data indicate over 8 m subsidence and 1 m rift-parallel displacement toward the vent of effusive fissure eruption at the graben floor, where we can find the maximum subsidence. Based on these data, we estimated the distributions of the fault slip and pure opening were allowed to be derided, and we also applied smoothing operator and the inferred slip/opening distribution. The inversion shows the maximum normal slip of 10 m and the maximum dike opening of 6.5 m. Standard deviations derived from iterative calculations with 2D random noise show normal slip of 50 cm and dike opening of 1m.