Applications of extreme value theory to climate science by Anna Kiriliouk (guest UNamur)

Extreme-value theory is the branch of statistics that studies rare events. In the last decades, its popularity within the climate

sciences has been growing, driven by the increasing awareness that extreme climate events will intensify in frequency and magnitude in view

of climate change. The behavior of a climate variable (e.g., rainfall, temperature, …) around its mean does not necessarily give information on

the behavior of that same variable in the tail (i.e., in the extreme part) of its probability distribution. Moreover, in practice, climate

data are observed on a grid of locations, so that variables from nearby grid points exhibit tail dependence, i.e., a tendency of extremes to

occur simultaneously.  Aggregating data without taking such tail dependence into account may lead to severe underestimation of the

probability of an extreme event and may reduce the confidence associated with attribution statements.