Statistical modeling of spatio-temporal weather extremes - Inference for serial clustering and homogeneity analysis
The statistical analysis of extreme weather events often ignores a proper accounting for dependencies and heterogeneities induced by the spatio-temporal nature of respective data sets. For instance, taking temporal dependence into account is important for the statistical assessment of the inter-arrival times between extreme events. Furthermore, statistical methods for data collected at various locations and times can be improved substantially if information is available about which locations share similar characteristics and which model parameters can be assumed to be constant over space or time. This project is divided into two parts. The first part, "Statistical inference for serial clustering of intense storms and heavy precipitation“, investigates a generalized extreme value theory model, so-called Continuous Time Random Maxima (CTRM), which allows for heavy-tailed inter-arrival times so that clustering of extreme events occurring in bursts can be captured. The second part, "Homogeneity analysis for spatio-temporal weather extremes", provides statistical tools to detect and make use of homogeneity information, in particular by investigating estimation strategies based on regularization techniques.
Institutions: TU Dortmund, University Düsseldorf
Contact: Roland Fried, Axel Bücher, Katharina Hees