Title: Spatio-temporal extremes based on single-location conditioning
Authors: Emma Simpson - Lancaster University (United Kingdom)
Jenny Wadsworth - Lancaster University (United Kingdom) [presenting]
Abstract: Conditional approaches for modelling multivariate extreme events have benefits in terms of computational efficiency and their ability to capture both asymptotic dependence and asymptotic independence. We present such a model for the spatio-temporal setting, conditioning on the occurrence of an extreme value at single location in space, and a single point in time. Combining information across all conditioning sites allows for inference via a composite likelihood approach. We apply our model to Red Sea surface temperature data, and present and discuss diagnostic approaches to assess the model fit.