Quantifying probabilities of eruption at a well-monitored active volcano: an application to Mt. Etna (Sicily, Italy)

At active volcanoes, distinct eruptions are preceded by complex and different precursory patterns; in addition, there are precursory signals that do not necessarily lead to an eruption. The main purpose of this paper is to present an unprecedented application of the recently developed code named BET_EF (Bayesian Event Tree_Eruption Forecasting) to the quantitative estimate of the eruptive hazard at Mt. Etna volcano. We tested the model for the case history of the July-August 2001 flank eruption. Anomalies in geophysical, geochemical and volcanological monitoring parameters were observed more than a month in advance of the effective onset of the eruption. As a consequence, eruption probabilities larger than 90% were estimated. An important feature of the application of BET_EF to Mt. Etna was the probabilistic estimate of opening vent locations. The methodology allowed a clear identification of assumptions and the monitoring of parameter thresholds and provided rational means for their revision if new data or information are incoming.