Modeling Uncertainty in Volcanic Hazards with Focus on Pyroclastic Density Currents from Neapolitan Volcanoes

Pyroclastic Density Currents (PDCs) are fast-moving hot mixtures of volcanic gas and particles that can devastate significant areas surrounding explosive volcanoes. Moreover, volcanoes are typically multi-hazard environments where secondary processes, like volcaniclastic mudflows (lahars), can be equally or more dangerous than PDCs. In this view, quantitative volcanic (multi) hazard assessment is among the most pursued goals of modern volcanology. However, the assessment is affected by deep variability-related (aleatory) and knowledge-related (epistemic) uncertainties. De facto, hazard analyses of PDCs and lahars quantifying such uncertainties are uncommon because modeling their complex flow dynamics, in a stochastic scheme, has a high computational cost. In this PhD, different probabilistic approaches to model and quantify uncertainty in volcanic hazard assessment are explored and implemented at two Italian volcanoes: Somma-Vesuvius and Campi Flegrei. First, it is demonstrated that simple PDC models (Energy Cone), coupled with Monte Carlo sampling, are able to: (1) reproduce the spatial extent of past PDC deposits, and (2) quantify epistemic uncertainty comprehensively. Secondly, by merging Energy Cone simulations with more complex statistical models (Bayesian Event Tree for Volcanic Hazard, BET_VH), a cutting-edge hazard product is computed: a multi-volcano, single-target probabilistic PDC hazard assessment over the central Campania region. Thirdly, through assembling more sophisticated PDC models (Titan2D) with Polynomial Chaos Quadrature and BET_VH, hazard curves for dense pumice flows from Somma-Vesuvius are obtained. These curves are extremely valuable for quantitative volcanic risk analyses. Finally, probabilistic volcanic multi-hazard assessment is performed at Somma-Vesuvius by combining a Bayesian Belief Network (Multihaz: which incorporates probabilistic hazard analyses of tephra fallout and dense PDCs and models aleatory uncertainty in lahar triggering) and a numerical model of lahars (LaharFlow: which allows to compute probabilistic hazard footprints). Future steps in probabilistic volcanic hazard assessment will likely require strategies that are multi-disciplinary and explicitly oriented to calculate volcanic hazard.