Jonathan Rougier - Independent Consultant

Abstract

The Exceedance Probability (EP) of a hazard shows the probability of an event exceeding a specified size, in a specified future interval. For example, if Merapi's 1-year EP for erupting 1Gt of mass were 0.001, then there would be a 0.001 probability of Merapi having an eruption exceeding 1Gt of mass in the next year. EP curves are crucial in hazard and risk assessment. Unfortunately, though, there is no reason to think that they are 'smooth'. For example, the eruptive behaviour of a  volcano is full of non-linearities and phase-changes.  How, then, can we map a sparse historical dataset into an EP curve?  And how do we quantify our confidence in such an assessment?  There are no canonical answers to these questions, but I show how probability and statistics can help.