The thinning of Iceland's ice caps could be influencing volcanic activity, according to new research. The study developed a model to predict the changes in pressure on magma caused by thinning ice and applied it to the recent activity of two Icelandic volcanoes. The results suggest that the effects of thinning ice may increase volcanic activity in some cases, but reduce it in others. Volcanic eruptions are influenced by the build-up of pressure below the Earth's crust. Increased pressure causes the magma to become more liquefied (or melt), which can then lead to eruption. This means that events that affect the pressure can affect the activity of a volcano, such as earthquakes and nearby eruptions. For those volcanoes below glaciers or ice sheets there is another source of pressure change: the variations in the load on the Earth's surface caused by changes in ice cover. At the end of the last glaciation period, Iceland had a large pulse of volcanic activity that has been linked to the disappearance of an ice sheet more than 300 km wide. Trends toward a warmer climate mean that Icelandic ice caps currently covering areas of volcanic activity have been retreating since 1890. The effect on pressure depends on the extent of the ice cap retreat. A retreating ice cap with a radius of only a few kilometres will influence only the shallow parts of the volcanic system, such as the magma chambers within the crust. A retreating ice cap with a radius of tens of kilometres or more will influence conditions in deeper parts within the mantle, which lies below the crust. The study developed a model to predict both these processes. Considering the impact at the deeper level, the model estimated that since 1890 a significant amount of 'additional magma' has been generated within the mantle under Iceland at a rate of 0.014 km3 per year. Considering the shallower effects on magma chambers within the crust, the model predicted that a decrease in ice pack would inhibit the rupture of certain types of magma chamber, but promote the rupture of others. For example, pressure changes are most likely to rupture spherical chambers. The research applied its predictions of the effects on shallow magma chambers to two Icelandic volcanoes: Gr�msv�ten and Katla. The 2004 eruption of Gr�msv�ten was preceded by the draining of water from a nearby lake beneath a glacier. The model indicated that the removal of this water above a 10 km3 magma chamber at a 2.5 km depth could have initiated the volcanic activity, particularly if the chamber was spherical. For Katla the model predicted that, in the case of spherical or long horizontal chambers, eruptions are more probable when seasonal snow cover is lowest. This appears to be consistent with the fact that the last nine major historical eruptions of Katla all occurred during the summer period. Despite these short-term effects, the model indicated that the long-term thinning of ice might actually reduce the likelihood of eruptions at Katla, although more information is needed on the properties of the different layers of the Earth to explore this further.