Multivariate Classification of extreme rainfall and discharge events in northern hemispheric latitudes and model-based estimates for the 21st century (already finished)

Start date: 01.01.2003
End date: 31.12.2005
Funded by: Cusanus-Werk
Local project leader: Prof. Dr. J. Jacobeit


Concerning “Global Climate Change” it is of special importance how far particular properties of extreme events will change during the next decades. Extreme precipitation is a prominent example. For individual regions there might be small-scale varieties, which can only be derived by indirect means from large-scale GCM output (General Circulation Models, grid point distances several hundred km). In that respect, deriving regional precipitation from large-scale GCM data (so-called downscaling) has to be critically assessed, since relationships between regional precipitation (spatial scale of Central Europe or smaller areas) and ‘Großwetterlage’ (i.e. pressure fields on a continental scale) are marked by considerable instationarities.
Consequently and in contrast to that, analysing fronts and convective conditions of regional to local dimensions is the starting point of this project. A fundamental advantage could be that fronts and convective phenomena cover areas comparable in extent to those being affected by resulting precipitation and runoff. Based on that, output of the atmosphere-ocean GCM ECHAM4 (Hamburg) can be used for more consistent estimations of 21st century’s precipitation variability.


For the 20th century there are several highly resolved data sets of many meteorological parameters, both gridded and as station time series. This project will focus on selected midlatitudinal areas of the northern hemisphere (Central Europe: about 4-19°E and 46-54°N, northeastern USA: about 68-86°W and 36-45°N). In the transitional zone between the subtropical and westerly belts changes may occur in a comparatively drastic way. Concerning climatic geography the two different continental positions, westside vs. eastside, are of particular importance.
Analyses are based on daily precipitation an discharge time series from the 2nd half of the 20th century. Quality control is followed by the generation of a calendar of singularities for each station. First findings on spatial differences will be based on statistical techniques like cluster analysis. Secondly, selection of extremes will take place by means of statistical criteria (e.g. multiples of standard deviation, appropriate percentiles, return periods). 

Involving features of the daily ‘Großwetterlagen’ (Central Europe) and the Spatial Synoptic Classification (USA), respectively, a specification of meteorological extremes and synoptic regularities can be achieved. Unusual dryness is considered as an extreme, too, by labeling sequences of dry days or days with very light precipitation as a dry period. From the extensive sample of extremes information will be condensed by using multivariate methods aiming at a classification of heavy precipitation fields and flood events in terms of frequency, intensity and duration. Methods applied include regression, discrimination and Principal Component Analyses. Results can be validated with the help of gridded reanalysis data (NCEP/ NCAR) for precipitation rate, convective precipitation, and surface runoff. Water equivalent of snow cover and soil moisture data are available as controlling parameters.
In a next step, instead of large-scale circulation patterns characteristics and frequencies of frontal and convective conditions above the affected regions themselves will be regarded as sources of precipitation and discharge variability. Several cited methods are applied as well as own numerical approaches. Analyses aim at the detection of frontal and convective classes on a daily scale for several tropospheric levels. More sophisticated approaches of classification include meteorological variables like geopotential height, specific humidity, wind direction and wind speed, potential temperature, dew point temperature, and cloud cover. A time series of frontal and convective events independent from precipitation predictors is to be generated. Results are based on spatial gradients – vertical and horizontal ones -, on interdiurnal fluctuations, and on layer thicknesses. Threshold values fitted to the specific problem need to be derived in order to define the existence of a front or of distinct convection, threshold values with regard to an intensity gradation must be fixed.

Finally, the question is raised whether or not and how far a correlation matrix of frontal and convective classes on the one hand and of precipitation plus flooding extremes on the other hand implies reasonable indications of causal relationships. After optimising such a matrix, if necessary, it serves as a basis to draw, in a reverse methodical approach, conclusions from atmospheric GCM (ECHAM4) data to hydrological extremes on regional to local scales. In addition, particular attention will be put on seasonal and spatial changes, both within the individual regions and in comparison between the two regions.




Discharge stations Central Europa
Discharge stations USA
Precipitation stations Central Europe
Precipitation stations USA