Applications: Volcanoes (3)

Lying below Vatnajökull ice cap in Iceland, Bárdarbunga stratovolcano began experiencing wholesale caldera collapse on August 16, 2014.  Associated with this collapse is the initiation of a plate boundary rifting episode extending tens of kilometers northwards of the caldera. The ongoing caldera collapse is the largest such event recorded in the modern instrumental era. We use observations acquired by the international constellation of radar satellites to measure ground deformation associated with the simultaneous caldera collapse and rifting event. A sequence of one-day interval interferograms formed from images acquired by COSMO-SkyMED (CSK), a constellation of X-band satellites operated by the Italian Space Agency (ASI), indicate rapid 50 cm/day subsidence of the glacier surface overlying the collapsing caldera. Longer-term 24- and 48-day interval interferograms formed from images acquired by RADARSAT-2, a C-band satellite operated by the Canadian Space Agency, indicate meter-scale crustal deformation in regions north of Vatnajokull surrounding the active rift zone. Contemporaneous with the InSAR observations are observations of anomalous seismic events around the rim of the caldera that show highly non-double-couple focal mechanisms. While seismic events with similar focal mechanisms have occurred at Bárdarbunga caldera in the past, the availability of surface deformation observations over the caldera with high spatial and temporal resolution provides critical constraints on the collapse sequence within the caldera. We propose a model of the collapse process consistent with available geodetic and seismic observations that suggests that the majority of the observed subsidence occurs aseismically via a deflating sill-like magma chamber. Inferred deformation associated with the anomalous seismic events are likely the manifestation of rapid localized deformation on the borders of the magma chamber, but do not represent a significant fraction of deformation associated with caldera collapse. We also estimate a model of tensile opening along the dike interface in the active rift zone using longer-term CSK and RADARSAT-2 interferograms with independent line-of-sight directions. Quantification of the dike opening allows us to estimate the stress induced by the rifting on nearby fault and volcanic systems, in particular the Askja volcanic system north of the rift zone.

InSAR has over the last two decades become one of the main techniques used to study earth surface deformation. Displacements due to magma movements beneath volcanic systems are one of the applications InSAR has had a major impact on. The high spatial sampling of the deformation shape provides valuable insights into the source parameters of the magma movements. Currently these results typically take weeks or months to obtain, due to the high computational time requirement associated with the timeseries techniques commonly used to identify coherent points. However, the much improved repeat time of current SAR satellites allows us to start using InSAR in different ways. For volcanoes this means we can start using InSAR as a monitoring tool, but only if we can process the large volumes of data in a rapid and efficient manner. Here we present results from a rapid algorithm, able to rapidly and automatically produce deformation maps over volcanic systems and other deforming areas.

Traditional timeseries techniques, like persistent scatterer (PS) and small baselines (SB) methods, work very well on large sets of data, but do not allow for new images to be inserted on the fly. This makes them unsuitable for volcano monitoring, as they are not able to handle these large, rapidly updating volumes of data. We have developed a timeseries technique that selects for every pixel the neighbouring pixels that on average behave in a similar manner, referred to as “cousins”. In stead of re-identifying the cousins every time a new image comes in, we store the cousin information identified using an initial set of interferograms, which we can then apply on new interferograms to calculate the complex coherence. By taking the cousin identification out of the rapid processing chain, we significantly speed up the processing time. The result is an individual estimate for the coherence of every pixel, in each interferogram separately. The coherence estimate is then used to unwrap each interferogram, after which we can invert for the deformation since the last acquisition.

Using only the cousins in the complex coherence calculation yields a much higher resolution result compared to the classic boxcar approach (see figure). This means that during unwrapping, we can actually take advantage of the high-resolution images, giving better results. The PS and SB techniques tend to select one set of points in all interferograms. This has two drawbacks; Coherent points are removed because of lack of coherence in other interferograms, and noisy points are remain in some interferograms when they are coherent in the majority of the timeseries. Thus, due to the individual coherence estimate per interferogram, our method has the potential to maximize the information extracted from the dataset. By taking the cousin identification out of the rapid processing chain, we significantly speed up the processing. Furthermore, the fact that the coherence is calculated for each interferogram individually means that there is far greater flexibility for computations to be done in parallel. Using 8 processing cores, we can form 15 interferometric combinations (5000 by 5000 pixels), calculate the coherence, unwrap and invert for the deformation in under 2 hours from obtaining the SLC image, with intermediate results being available much faster.

Analysis of a time series of ground deformation measurements at active volcanoes can provide an improved understanding of sub-volcanic and sub-aerial processes; including those related to magmatic, hydrothermal and structural development. Interpreting a long time series may also help determine what is typical “background” behavior, and identify any deviations from this, including the migration of new melt. We use Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) techniques to generate a time series of high-resolution deformation measurements, in the vicinity of the most active volcanoes in Iceland: Bárðarbunga, Askja, Hekla, Katla and Eyjafjallajökull and compare these to other geodetic measurements. A comprehensive network of continuous GPS stations is already deployed at these volcanoes and a series of campaign GPS measurements are routinely undertaken each summer. InSAR observations are complementary to these field based measurements and their high spatial resolution assists in resolving the geometry and location of the source of the deformation.

The Committee on Earth Observation Satellites has recently declared Iceland a Permanent Geohazard Supersite, based on its propensity for relatively frequent eruptions and their potentially hazardous, long ranging effects. The recent Supersite award ensures a considerable amount of SAR data is made available for both past and future satellite acquisitions, including new X-band images (acquired by TerraSAR-X and Cosmo-SkyMed satellites), and historic C-band images from ERS and ENVISAT. We present a series of long-term deformation measurements for Hekla, Katla, Eyjafjallajökull and Askja volcanoes, derived using PS-InSAR techniques, and include recent interferograms spanning the 2014 unrest and eruption within the Bárðarbunga volcanic system.

InSAR and tilt measurements at Hekla indicate renewed melt supply to a sub-volcanic reservoir after the last eruption in 2000. Recent deformation studies utilising data spanning this eruption, have provided insight into the shallow plumbing system, which may explain the large reduction in eruption repose interval following the 1970 eruption. InSAR and GPS observations at Katla volcano prior to 2010 suggest no magma induced deformation.  However, deformation associated with a small flood at Mýrdalsjökull in July 2011, followed by an increase in micro-seismic earthquakes, could be interpreted in relation to magma movements. Post-eruption deformation observations reveal inflation at Eyjafjallajökull possibly related to the influx of new melt or readjustment of crustal stresses following the 2010 eruption. Continued deflation at Askja caldera since 1983 may have led to crustal weakening and the triggering of a mega rockslide and subsequent tsunami occurring on the 21 July 2014. Interferograms spanning the recent unrest and eruption within the Bárðarbunga volcanic system display both pre-eruptive and co-eruptive deformation associated with the initial dyke emplacement and ongoing magma withdrawal from beneath the Bárðarbunga central volcano.

Continued high-resolution geodetic observations at these volcanoes are essential for assessing changes in their behaviour and the associated hazards. Rapid analysis of interferograms combined with GPS and earthquake seismicity measurements assists in tracking the evolution of magmatic activity during volcanic unrest/eruption, and may facilitate the assessment of associated hazards.

An intense earthquake swarm began on 16 August 2014 at Bárðarbunga volcano under the Vatnajökull ice cap in Central Iceland. It marked the beginning of a propagating dyke, over 45 km long. Such major event has not been observed for the last three decades in Iceland, since the Krafla rifting episode 1975-1984. The dyke propagation stopped 15 days after the onset of the seismic activity, with the dyke distal end in the Holuhraun plain north of the Vatnajökull ice cap. A small 4 hour eruption marked the beginning of extrusive activity. A new fissure eruption opened up on 31 August at the northern dyke tip, with lava fountaining and feeding extensive lava flows. In November 2014 the surface covered by the lava had exceeded 60 km², and the eruption did not show significant decline.

Interferometric analysis of SAR data (InSAR) have been conducted since the onset of the unrest. X-band satellite images from COSMO-SkyMed and TerraSAR-X satellites were acquired and analyzed to map ground surface deformation associated with the dyke emplacement. Despite most of the dyke propagation occurred under several hundreds of meters of ice, the last 10 km were outside the ice cap enabling better characterisation of the dyke-induced deformation. Here we focus on the Holuhraun plain, in order to better understand the link between the surface deformation detected in the vicinity of the dyke by InSAR as well as GPS measurements, and the eruptive activity.

The regular SAR acquisitions made over the Holuhraun area since the beginning of the unrest offer a unique opportunity to better understand the evolution of the intrusion feeding the fissure eruption. For that purpose, we focus on the faults and fissures forming the graben borders on the glacier as well as in the Holuhraun plain, initially mapped using high-resolution radar images, acquired by airborne radar. The extraction of movement along these structures in interferograms is performed along profiles centered on them and perpendicularly oriented. During the first weeks of the volcanic unrest, the surface deformation nearby the top of the dyke exceeded one meter in the Line-of-Satellite resulting in a systematic decorrelation in the phase signal observed above the dyke. However the amount of subsidence may be estimated using the estimate of offsets in range and azimuth.

The distribution of slip along these fissures/faults and its space-time evolution is compared to the amount of dyke opening at depth obtained from modeling of geodetic data for different time spans defined by the interferograms. In comparing geodetic and seismic data with observations of the effusive activity, we investigate the development of the dyke and its influence on the surface deformation. This allows us to analyse the plumbing system feeding the eruptive fissure as a clue to better understand the ongoing volcanic eruption.

This study based on multiple data sets is supported by the FP7 FutureVolc project and the CEOS Icelandic Supersite. Funded by the European Commission, FutureVolc project is concerned with improving the monitoring of Icelandic volcanoes and facilitating a multi-disciplinary approach to better understanding volcanic processes.  

Between 16 August and 31 August 2014 a dike propagated from Bárðarbunga caldera. This culminated in an eruption at Holuhraun that is still ongoing at the time of writing. The dyke propagation path is not simple, comprising many segments with differing orientations. We modelled the dyke propagation using deformation data from InSAR and GPS. Initial modelling of the dyke, with no a priori constraints on position, strike or dip, show the deformation data require the dyke to be approximately vertical and line up with the seismicity. We therefore fixed the dip to be vertical and the lateral position of the dyke to coincide with the earthquake locations. We modelled the dyke as a series of rectangular patches and estimated the opening and slip on each patch for each day between 16 August and 6 September. The results suggest that most of the magma injected into the dyke is shallower than the seismicity, which mostly spans the depth range from 5 to 8 km below sea level. Where constraints from InSAR and GPS are good, significant opening is all shallower than 5 km. The total volume intruded into the dyke by 28 August was 0.48-0.51 km3. One-day interferograms and GPS also show rapid subsidence within the caldera, both during dyke propagation and subsequently, at around 40 cm per day. In a simplified approach, the caldera subsidence and deformation surrounding the voclano can be broadly fit with two dip-slip faults at the caldera walls and a flat-topped chamber with a minimum depth of 3.5 km (could also be much deeper). The volume decrease beneath the caldera tracks the volume increase of the dyke for the first week of the activity. The volume decrease then decelerates to less than half the previous rate, although the dyke volume increase continues at the same rate. This suggests inflow of magma from a deeper source after the first week, which is not visible in the geodetic data. We also modelled the expected propagation direction of the dyke considering the regional stress field and the spatially-variable density structure. We find that our model agrees well with the actual propagation path as indicated by the seismicity.

During the round table, seed questions proposed by the chairs will be discussed with the audience.