Icecap and subglacial crustal deformation inferred from SAR pixel tracking: the 2014 dike intrusion episode at the Bárðarbunga volcanic system

The Bardarbunga dike intrusion episode started to occur on 16 August 2014 in the central Iceland, which is located at a subaerial divergent plate boundary between the North American plate and the Eurasian plate. Seismicity propagation during the episode indicated dike propagation from Bardarbunga caldera to the Holuhraun plain, indicating 40 km of subglacial dike intrusion. Due to the subglacial dike intrusion, aerial photographs reported three a few km in diameter of ice cauldrons were formed along the inferred dike path.

Previous geodetic studies have reported crustal deformation at the ice-free region and dike opening/faulting model using interferometric synthetic aperture radar (InSAR) and SAR pixel tracking data (e.g., Sigmundsson et al., 2015; Ruch et al., 2016). These geodetic data detected graben formation signal with 6 m of subsidence associated with the dike intrusion, but these data did not tell any information about the crustal deformation beneath the icecap.

Pixel tracking approach is suitable for detecting signal of both ice surface movement and large-scale crustal deformation. In this study, we applied the approach to Cosmo-SkyMed (CSK), TerraSAR-X (TSX), and RADARSAT-2 (RS2) data to infer the subglacial crustal deformation due to the dike intrusion. We estimated the subglacial crustal deformation by subtracting the scaled pre-diking data from the co-diking data, assuming that the pre-diking data contain only the signals of steady-state of icecap surface movement. The strategy allowed us to infer the subglacial crustal deformation at the northern half of dike. The subtracted RS2 data implied aligned signal along the dike path, indicating subglacial graben formation. On the other hand, the data could not detect signals at the southern half of dike due to decorrelation problems. We confirmed that the inferred subglacial crustal deformation signals contribute to improve the dike opening distribution. Our model suggested plausible opening distribution at more southern part of dike compared to previous models.