PhD Project Technical Description
Environmental extremes exhibit spatial, temporal and multivariate dependence. The temporal and spatial dependence is typically positive, which means that the probability of an extreme event in one variable in the near future and in the vicinity of another extreme event in another variable are generally increased. The complexity of dependence structures increases with the dimension of environmental variables and the few current higher-dimensional methods are difficult to apply to other problems. Assumptions about the underlying dependence structure using vine copulas or graphical models have the potential to simplify the d-dimensional modelling into a series of lower dimensional models. We use these simplifications to extend the multivariate applications of the conditional extreme value model framework (originally proposed by Heffernan and Tawn, 2004).