Fig. 2
From: Classifying development stages of primeval European beech forests: is clustering a useful tool?
Between-clusters variance of stand structural data (7 attributes, abbreviations see Table 2) of the primeval beech forests Mirdita (A1–A4) and Rajca (B1–B4). K-means clustering was used to detect clusters (2 to 5 clusters, panels 1 to 4). A moving window approach with a uniform kernel (equal weighting of all objects within the window) of several observation scales was used to aggregate the datasets (X-axis). For the results of a bivariate normal kernel see Additional file 2: Figure S2 (weighting of objects by their distance to the window center)