Fig. 4
From: Classifying development stages of primeval European beech forests: is clustering a useful tool?
Stem position maps of the primeval forest Mirdita with k-means clustering solutions of the structural data highlighted (2 to 5 clusters, panels a to d). Coloring of the background images indicates areas which were assigned to the same cluster (gray tone) and how well a point is represented by its cluster (silhouette coefficient, red tone). A moving window of an observation scale of 500 m2 was used to aggregate the structural datasets (7 attributes, Table 2). A uniform kernel was used for the moving window (equal weighting of all objects within the window). For results of a bivariate normal kernel (weighting of objects by their distance to the window center) see Additional file 4: Figure S4