Volcanic Eruptions: Stochastic Models of Occurrence Patterns

Abstract

Depending on the definition, there are approximately 1,300 Holocene (last 10,000 years) active volcanoes (Simkin 1993), some 55–70 of which are typically active in any given year. Being tectonic phenomena, they are spatially overrepresented along littorals and in island arcs where populations congregate, and millions of lives are potentially at risk. Hence, the quantitative forecasting of hazard has long been a key aim of volcanology. This requires models to quantify the likelihood of future outcomes, which may involve the size, location, and/or style of eruption, in addition to temporal occurrence. While a wide range of complex deterministic models exist to model various volcanic processes, these provide little in the way of information about future activity. Being the (partially) observed realization of a complex system, volcanological data are inherently stochastic in nature and need to be modeled using statistical models.