Extreme Value Theory (EVT) is a useful tool to describe the statistical properties of extreme events. Its underlying assumptions include some form of temporal stationarity in the data. Previous studies have been able to treat long-term trends in datasets, to obtain the time dependence of EVT parameters in a parametric form. Since there is also a dependence of surface temperature and precipitation to weather patterns obtained from pressure data, we determine the EVT parameters of those meteorological variables over France conditional to the occurrence of North Atlantic weather patterns in the summer. We use a clustering algorithm on geopotential height data over the North Atlantic to obtain those patterns. This approach refines the straightforward application of EVT on climate data by allowing us to assess the role of atmospheric variability on temperature and precipitation extreme parameters. This study also investigates the statistical robustness of this relation. Our results show how weather regimes can modulate the different behavior of mean climate variables and their extremes. Such a modulation can be very different for the mean and extreme precipitation.
Nonlinear Processes in Geophysics
- Pub Date:
- May 2008