In the paper there is presented the method designed for vector fields classification. It is based on the algorithm proposed by Lund (1963). The replacement of Pearson correlation coefficient with vector correlation coefficient defined by Crosby (1993) is the essential change that makes this technique suitable for treatment of vector fields. The correlation threshold, set a priori in scalar analyses, is calculated objectively, with the help of PVCR index introduced by Huth (1996). PVCR index expresses the proportion of average similarity between objects classified into different categories to objects grouped in the same classes. PVCR is computed for selected range of threshold values and minimum value indicates the ,,best" correlation threshold. The classification was carried out for wind fields at the 250 hPa surface over Europe for winter season during 1958-2003 period. 20 classes of jet stream patterns were distinguished, which constitute 60.7% of the sample. Basic statistics, i.e. average and maximum frequency and duration time, were calculated for the first six key jet types. The main features of upper circulation field as well as temperature and pressure distribution in the lower troposphere associated with the jet types were also described.