Analysis of the Eyjafjallajökull 2010 eruption through TerraSAR-X InSAR Time Series

Abstract

After a period of quiescence since a sill intrusion in 1999-2000, a subtle deformation signal was again detected at Eyjafjallajökull, beginning in the summer of 2009, at a continuous GPS station on the southern flank. At our request the German Space Centre (DLR) immediately began tasking the TerraSAR-X satellite to acquire three SAR images every 11 days, giving a time series of SAR images prior to the eruption with unprecedented temporal sampling (although interrupted by snow during the winter).
Previously we modelled individual interferograms that showed that there was a large uplift signal. We modelled this as a series of sills and a dike with a total volume of ~0.05 km3. During the flank eruption, beginning on 20 March, no significant deformation is detected, but coinciding with the start of the explosive eruption on April 14, we detected subsidence centred on the caldera. What we modelled showed us that Eyjafjallajökull was an unusual which we modelled …. This deformation does not relate to pressure changes within a single magma chamber.
Here we extend our analysis InSAR time series covering full eruptive period. After correcting for DEM errors and reduction of atmospheric signal, we have found a number of signals that we interpreted in terms of magma movement. These magma movements are separately analysed in 3 phases: pre-eruptive (inflation), co-eruptive (no deformation) and post-eruptive (deflation).
The displacement time series from June 2009 to 4 February 2010 (pre-eruptive-phase) shows line-of-sight shortening on the southwest flank of about 2 cm. The displacement signal is present in a set of interferograms and it has a consistent behaviour in time, implying that it is not due to atmospheric contamination. We performed atmospheric stratification over the entire Interferogram (4 Feb 2010) to check how much of the signal correlated with the topography would disappear when removed. The correlation coefficient over the southwest flank is very small compared with the signal from the entire interferogram. We can say that in this area not much of the atmospheric effects related with the topography are present, suggesting that the signal could be deformation.
For the co/post-eruptive phases we calculated phase difference between nearby points to check their evolution in time. In the southeast flanks we observe deflation through all analyzed period, while in western flanks of the volcano we observe Inflation during effusive eruption, followed by deflation during explosive eruption, and a new inflation pattern between 05 June and 19 July that we cannot explain. In preliminary modelling we fit this post-eruptive phase with a pressure decrease of an ellipsoidal source, equivalent to a volume reduction of ~0.03 km3.
The limitations when analysing this dataset are mainly concerning the phase unwrapping performance through ice- and ash-covered areas. This is caused by decorrelation owing to ash cover where there is almost complete loss of coherence. We applied new methods to overcome these limitations. To improve point density over the scene, we combined PS (Persistent Scatterers) and SB (Small Baselines) methods. By combining highly coherent interferograms, the increase of distributed scatterers is clear and the phase unwrapping performance improved. To detect and correct non-systematic unwrapping errors, we calculated azimuth and range offsets. Additionally, because of the fact that L band has higher penetration, we processed ALOS images trough single interferogram analysis. By these means we were able to extract more of the deformation signal around decorrelated areas.