The aim of CoDEx is to develop a detection and attribution (D&A) framework forextreme weather related to moist deep convection. A suitable reduction of degrees of freedom thatimproves the signal-to-noise ratio increases the potential to detect less strong signals such as climatechange signals in high-dimensional and highly variable local weather systems.
This project proposes to focus on a compact statistical-dynamical description ofhigh-dimensional spatial weather extremes: In two subprojects we investigate and develop variousmethods and statistical-dynamical models for information compression of spatially distributed extremes.
In this subproject, we will analyze available strategies for information compression, some of which have been proven very useful in statistical inference in climate science. Most of these methods aim at the description of the bulk of the distribution, while our analysis will focus on the tail behavior of the climate variables. Thus, we will compare different methods and assess their assets and drawbacks with respect to extreme weather. These methods comprise data adaptive decomposition such as principal component analysis, filter approaches using wavelet decomposition, or dynamical decomposition based on reduced dynamical models. We will further use novel spatial statistical models for extremes including the model to be developed in the second subproject. Advances in this project will enlarge the statistical methods tool box for extremes and help to detect changes in extreme frequency and intensity. The methodological advances will be compared in a detection and attribution study for mesoscale weather extremes.
|Principal Investigator:||Petra Friederichs|
|Research Assistant:||Svenja Szemkus|
In this subproject, we will develop spatial statistical models for mesoscale weather extremes, such as heavy precipitation and wind gusts. Based on ideas from extreme value statistics, these models will describe the spatial dependence structure of single extreme events. Here, extreme events are defined as exceedances of certain characteristics (closely related to the “impact”) over high thresholds modeled by Pareto processes rather than compound events modeled by max-stable processes. By the inclusion of an appropriate number of further covariates, such as various meteorological variables at different scales, the model will allow for spatio-temporal non-stationarities. Thus, the statistical model will enable us to describe the variables of interest at a finer spatial resolution (“downscaling”). The model aims at the identification of extreme events and the underlying dynamical patterns, and will serve as one method for information compression.
|Principal Investigator:||Marco Oesting|
|Research Assistant:||Carolin Forster|
"CoDEx" is part of the joint research project "ClimXtreme" funded by the Federal Ministry of Research and Education (BMBF) under the third framework programme Research for Sustainable Development (FONA3).