Freiburger Geographische Hefte, Vol 52

Summary

In the present work possibilities for the regionalization of precipitation data are investigated, which are given by the use of fields of meteorological data derived from mesoscale flow models (e.g. wind direction and velocity, height of the atmospheric boundary layer etc.) and the combination of these data within an Geographic Information System (GIS). The basic idea behind this concept is, that the spatial distribution pattern of precipitation is strongly determined by the two factors wind and relief. Examples may be orographically induced rainfall and the asymmetry between windward- and leesides of mountains.

The method presented has been tested in the southwestern part of the German Bundesland Baden-Württemberg, including small parts of the neighbouring countries France and Switzerland. The research area covers a rectangle of 98 km in W-E- and 74 km in N-S-direction. In this study five days (01.01.87, 09.04.87, 04.06.87, 24.08.87 and 06.12.87) and a period of one month (May 1987) are considered. The data sets necessary for the investigation are primarily a digital terrain model, a landuse classification, precipitation data measured at 94 stations within the research area and meteorological data such as the vertical profile of temperature and geostrophic wind.

Using the mesoscale flow model REWIMET threedimensional windfields for all selected time periods are calculated. In a second step, rates for the rising and falling of air particles ( so called shifting values) at each point of the calculation grid are derived from these wind fields using special GIS-methods. To evaluate this procedure, correlation coefficients between shifting values and precipitation data are calculated. In the next step, the hight of the atmospheric boundary layer is extracted from the datasets given by REWIMET and - like the shifting values - compared with the precipitation data. To consider aspects, which result from the wind induced drift of falling raindrops a method has been developed, with which the shifting values and the hight of the atmospheric boundary layer can be driftcorrected.

In combination with the terrain hight these variables serve as a basis for the development of a regression model, with which area covering datasets of precipitation can be derived. To combine the different data levels a GIS is an essential help. The concept of this procedure is the regionalization model FREIM developed at the "Institut für Physische Geographie" of the University of Freiburg (Germany). It has been created for the derivation of area covering data sets from geoecological point data, applying statistical methods.

One way to evaluate this procedure is the visual interpretation of the precipitation maps. The other way is given by the regression model, which derives the percentages the different predictors contribute to the explained variance of the precipitation. Last not least, the regression procedure creates multiple correlation coefficients to obtain a measure for the precision of the whole model. These coefficients vary between 0.75 and 0.83 for the five days and reach 0.88 for May, 1987.