Katla is a subglacial basaltic volcano and one of Iceland’s most notorious. It last erupted in 1918; the eruption was five times larger than Eyjafjallajökull 2010. However, little is understood on why this eruption was so powerful.

 

We have collected jökulhlaup and air-fall samples from this eruption. In both locations, layers were observed with varying grain sizes, which we have sampled from.

 

FTIR data suggests that the jökulhlaup samples quenched under elevated pressure. This could either represent ~130 m of water, ~120 m of ice or ~40 m of rock. In general the jökulhlaup clasts are very bubbly and have good quality glass. By comparison air-fall samples are generally more microlite and bubble rich, suggesting they had a slower cooling rate and longer to degas. However, many clasts have a bubble free margin, suggesting that the clast interiors were able to carry on degassing post-fragmentation. A magma-water heat transfer model (Woodcock et al., 2012) suggests that the clasts would have taken seconds to cool had they done so in water. Hotstage experiments have revealed a bubble growth rate of ~1 μm s-1 for typical eruptive temperatures, implying that bubbles would have been able to grow at most a few micrometers had the clasts quenched in water. We therefore suggest that clasts with a significant margin-core variation in bubble size have fragmented in air allowing longer cooling times with little/no interaction with water.

 

Many clasts, particularly from the jökulhlaup deposit, show evidence of there being clasts within them, often with significantly different bubble textures. We use these clasts to suggest that there were multiple episodes of fragmentation and vesiculation which could be explained by fragmentation taking place within the conduit and/or within close proximity to the vent and may explain the elevated pressure recorded by the FTIR data.

 

There are, however, a large variety of clast types which is apparent in both thin sections and SEM images of external textures. Some clasts show evidence of magma-water interactions whilst others are incredibly vesicular.

 

Geochemical data has determined that typical Katla 1918 clasts have a SiO2 content of ~47 wt.%. However there is a range of data with some clasts having up to ~74 wt.%. The distribution is bimodal which is supportive of historic and prehistoric Katla data and implies that the rhyolite was formed by late reheating and partial melting of basaltic crust (Lacasse et al., 2007). Some of the clasts show visual evidence of magma mingling, however this is less obvious in both the major and trace element chemistry.

 

Melt inclusions reveal significantly higher pre-eruptive S contents than matrix glass. The matrix glass data is varied but jökulhlaup samples generally have more retained than air-fall samples consistent with reports that phreatomagmatic samples retain higher S concentrations than magmatic tephra (Óladóttir et al., 2007). Rhyolite samples generally have lower S but higher concentrations of F and Cl.

 

Further work will involve SEM on thin section images to quantify bubble textures, produce bubble number densities and hopefully use these to calculate decompression rates.

 

Woodcock D. et al., (2012) J Geophys Res, 117(B10)

Lacasse C. et al., (2007) Bull volcano, 69: 373-399

Óladóttir B. et al., (2007) Ann of Glaciol, 45: 183-188