Event trees for eruption forecasting at Vesuvius volcano

Abstract

A probabilistic approach is used to forecast a future eruption at Vesuvius volcano. Such approach, differently from a deterministic one, allows to account for spatial and temporal variability of eruptive style (effusive, explosive), event magnitude (VEI), and environmental impact (dispersion, runout) (Newhall and Hoblitt, 2002; Marzocchi et al., 2004; Neri et al., 2008). This variability is quantified by means of Event Trees and conditional probabilities (Newhall and Hoblitt, 2002). To better constrain uncertainty, different sources of information should be considered and integrated with each other: geological record, historical observations, monitoring activities, results from scenario modelling. The integration of the different data is important to provide a robust characterization of the state of Vesuvius over geological vs. historical times, also in light of its current state as inferred from monitoring data and conceptual models. Different techniques exist to carry out this integration. For Vesuvius, available studies are based on the application of the Bayesian Event Tree (BET) model (Marzocchi et al., 2008; Sandri et al., 2009; Selva et al., 2014), and on the development of an Event Tree informed by expert elicitations (Neri et al., 2008), making possible to set up probabilistic eruption forecasting models both at long- (years) and short-term (hours to days), based on the current vs. past states of the volcano.