A historical analysis of Plinian unrest and the key promoters of explosive activity.

Abstract:

Plinian eruptions are the largest historically recorded volcanic phenomena, and have the potential to be widely destructive. Yet when a volcano becomes newly restless we are unable to anticipate whether or not a large eruption is imminent. We present the findings from a multi-parametric study of 42 large explosive eruptions (29 Plinian and 13 Sub-plinian) that form the basis for a new Bayesian Belief network that addresses this question. We combine the eruptive history of the volcanoes that have produced these large eruptions with petrological studies, and reported unrest phenomena to assess the probability of an eruption being plinian. We find that the 'plinian probability' is increased most strongly by the presence of an exsolved volatile phase in the reservoir prior to an eruption. In our survey 60% of the plinian eruptions, had an excess SO2 gas phase of more than double than it is calculated by petrologic studies alone. Probability is also increased by three related and more easily observable parameters: a high plinian Ratio (that is the ratio of VEI≥4 eruptions in a volcanoes history to the number of all VEI≥2 eruptions in the history), a repose time of more than 1000 years, and a Repose Ratio (the ratio of the average return of VEI≥4 eruptions in the volcanic record to the repose time since the last VEI≥4) of greater than 0.7. We looked for unrest signals that potentially are indicative of future plinian activity and report a few observations from case studies but cannot say if these will generally appear. Finally we present a retrospective analysis of the probabilities of eruptions in our study becoming plinian, using our Bayesian belief network. We find that these probabilities are up to about 4 times greater than those calculate from an a priori assessment of the global eruptive catalogue.