A drinking water catchment area named "Fuhrberger Feld" in the north of Hannover in Lower Saxony is being studied to enable a reliable and continuous evaluation of chemical emissions from agricultural activities. In this research and development project, ENVISAT polarimetric SAR data (provided free of charge by ESA within a pilot project "AO335") are used together with GIS information and ground surveys. Due to only two possible polarisations of the data from the ENVISAT ASAR sensor, their coherence, together with the non- distinguishable response of different vegetation types and the high variance of the backscatter, a classification using single date images will fail or be far too inaccurate. Methods like the use of multi temporal approaches have been tested to increase the classification accuracies. In this paper, the feasibility of a classification method, based on the statistical behaviour of agricultural fields is discussed and an attempt is made to find an optimal combination of preprocessing and classification method. It has been found, that a priori maps or layouts of the agricultural field boundaries are a prerequisite for the method which tries to define the crop type on the base of an existing segmentation. Test results from the years 2004 and 2005 are presented in this paper. An accuracy of 85% is achieved using 11 images of year 2004. However, using only 6 available images in the year 2005 reduces the accuracy down to 64%..