Bayesian Hierarchical Time Predictable Model for eruption occurrence: an application to Kilauea Volcano

Abstract

The physical processes responsible for volcanic eruptions are characterized by a large number of degrees of freedom, often non-linearly coupled. This extreme complexity leads to an intrinsic deterministic unpredictability of such events that can be satisfactorily described by a stochastic process. Here, we address the long-term eruption forecasting of open conduit volcanoes through a Bayesian Hierarchical Modelling information in the catalogue of past eruptions, such as the time of occurrence, the duration, and the erupted volumes. The aim of the model is twofold: (1) to get new insight about the physics of the process, using the model to test some basic physical hypotheses of the eruptive process and (2) to build a stochastic model for long-term eruption forecasting; this is the basic component of Probabilistic Volcanic Hazard Assessment that is used for rational land use planning and to design Emergency plan. We apply the model to Kilauea eruption occurrences and check its feasibility to be included in Probabilistic Volcanic Hazard Assessment.