Magma Boiling Underneath Volcanoes: A Key to Massive S Release during Eruption

Thursday, 18 December 2014
Yanqing Su1, Christian Huber1, Olivier Bachmann2 and Zoltán Zajacz3, (1)Georgia Institute of Technology, Atlanta, GA, United States, (2)ETH Zurich, Zurich, Switzerland, (3)University of Toronto, Department of Earth Sciences, Toronto, ON, Canada
Sulfur dioxide, the third most abundant volatile component from volcanic eruptions, can cause both global and regional climate changes once injected into the stratosphere. The quantification of the amount of SO2 emitted during an eruption is a long-term scientific goal in volcanology and climate research. Techniques to estimate sulfur yield to the atmosphere include the petrological method, remote sensing of gas compositions and sulfur measurement in ice-cores. However, the SO2 amounts measured by the latter two methods are often one to two orders of magnitude greater than the petrological estimates based on melt inclusion data. Here, we hypothesize that the inferred “excess sulfur” is a consequence of sulfur loss from the melt to the exsolved volatiles during second boiling. We test this hypothesis with a bubble growth model that includes sulfur kinetics and that allows us to track sulfur partitioning during second boiling and syneruptive decompression. Results show that before eruptions, large amount of S can be stored in gas. Thus the sulfur mass delivered to the atmosphere deduced from melt inclusion data could be greatly hampered by the second boiling process. Our results also explain the loss of S in the melt during crystallization, i.e. melt trapped by melt inclusion late in the crystallization process yield a melt significantly depleted in S, which is often observed in melt inclusion datasets from a single eruption. Our results imply that for most explosive eruptions, the S mass balance derived from melt inclusions could greatly underestimate the S content of the primitive magma. Our degassing model shows that apparent excess S are likely the result of a sampling bias associated with the melt inclusion record