Click on image to view larger version.
Fig. 4. Ranges (min.–max. bars) and means (circles) of yield variance accounted for by the classification at Sites A and B. The width of the bars indicates how well a particular classification method performed in accounting for the yield variation in each of the five (Site A) or six (Site B) different years. Methods shown: hierarchical cluster analysis using Ward's methods (WAR), nonhierarchical cluster analysis using k means (KME), unsupervised nonhierarchical Iterative Self-Organizing Data Analysis (ISODATA) clustering method (ISO), and nonhierarchical fuzzy-k-means cluster analysis (FUZ). For each method, values are shown for three different data sources [mean yield (MY), mean and standard deviation of yield (MS), and yields in all years (AY)] and five to seven classes.
