Summary:

In the decision making process concerning numerous objects researchers frequently face the problem of selecting an appropriate method to link them into groups of similar objects. This study is a proposal of the application of agglomeration methods to search for relationships between individual forest ecosystems in terms of their water retaining capacity. It was assumed that most parameters describing water retaining capacity of a specific ecosystem should come from forest management plans. Selected characteristics, adopted as those describing water retaining capacity of forest ecosystems, i.e. mean fall of land [%]; distance from watercourse network [m]; distance from standing waters, wetlands (lakes, ponds, swamps) [m]; site moisture content variant [-]; stand density index [-]; type of soil cover [-]; undergrowth cover index of the experimental plot [-]; canopy index [ha∙ha^-1] and potential available water retentiveness of the soil profile [mm] were tested using seven data clustering methods, i.e. the single linkage method, group average method (UPGMA – Unweighted Pair-Group method using arithmetic averages), weighted average method (WPGMA – Weighted Pair-Group Method using arithmetic Averages), the furthest neighbor method, the unweighted centroid method (UPGMC – Unweighted 169 Pair-Group Method using Centroids), weighted centroid method (WPGMC – Weighted Pair-Group Method using Centroids) and the Ward method. The applied clustering methods group analyzed objects, showing to a larger or smaller degree the existence of a considerable dependency between parameters selected for the purpose of the characteristic of water retaining capacity and the course of the clustering process itself. The main factor affecting the formation of dendrites was the age of tree stands and - in case of alder carrs – the forest site. The furthest neighbor method and the Ward method seem to be the most advantageous for the search between objects in terms of the water retaining capacity defined in this study. The single linkage, unweighted centroid and weighted centroid methods, due to the tendency to form rather unclear clusters with the structure of long chains, do not seem suitable for such analyses.