A doubly stochastic model for pyroclastic density current hazard assessment: the example of Campi Flegrei caldera

Abstract

Pyroclastic density currents (PDCs) represent one of the most dangerous volcanic hazards for people living in proximity of explosive volcanoes. The zonation of areas potentially affected by this threat is therefore of paramount importance and is the first step needed to set up appropriate mitigation measures. Campi Flegrei (CF) caldera represents a high-risk volcano with a remarkable PDC hazard due to the frequent occurrence of this phenomenon in its eruptive history. Despite the fact that CF caldera has been the object of many studies in recent decades, the mapping of PDC hazard there remains particularly challenging due to the remarkable variability of potential vent locations and eruption scales, and the complex dynamics of PDC propagation over the caldera topography. In this study we have produced, through the application of a doubly stochastic model, quantitative background (also called long-term or baseline) probabilistic maps of PDC invasion able to incorporate some of the main sources of epistemic uncertainty that influence the models for aleatoric (physical) variability. The new method developed combines the spatial probability distribution of vent opening locations, the density distribution of PDC invasion areas, and a simplified PDC model able to describe the main effect of topography on flow propagation. Our results indicate that the entire caldera has the potential to be affected (with a mean probability of flow invasion higher than about 5%) and the central-eastern area of the caldera (i.e. Agnano-Astroni-Solfatara) has invasion probabilities above about 30% (with local peaks of mean probability of about 50% in Agnano). Significant mean probabilities (up to values of about 10%) are also computed in some areas outside the caldera borders. Our findings are quite robust against different assumptions about several of the main physical and numerical parameters adopted in the study. In addition to mean values of probability of PDC invasion, the study provides the estimates of the credible uncertainty ranges associated with such probabilities in relation to some key sources of epistemic uncertainty. From our analysis, uncertainty spreads on invasion probabilities inside the caldera typically range between ±15 and ±35% of the local mean value, with an average of about ±25%; wider uncertainties are found outside the caldera, with an average above ±50% and a significantly larger range of variability from place to place.