Community ecologists are often confronted with multiple possible partitions of a single set of records of species composition and/or abundances from several sites. Not only must ecologists choose the classification algorithm and parameters, but they also have to make a choice of the number of clusters to be interpreted. The question is how many clusters are appropriate for the description of a given system. Several methods that intend to locate optical clusters have been developed so far. However, no criterion for determination of the optimal partition has received general acceptance. 227 sites of Yellow River Delta were classified by Ward's hierarchical clustering method. We simultaneously evaluate several criteria while varying the number of clusters (from 2 to 15 clusters), to help determine the most appropriate index and number of clusters. Validation indices (based on comparison of within-cluster and between-cluster heterogeneity), species composition and environment, and number of species with high fidelity to clusters in a cluster were used to validate optimal number of clusters. Validation indices including average silhouette width, Goodman and Kruskal's Gamma coefficient, Dunn index, entropy of the distribution of cluster memberships, wb.ratio(average.within/average.between), Calinski and Harabasz index, Cindex, partana, biserial. We used multiple response permutation procedure (MRPP) to compare the cluster with respect to the dissimilarity of their vegetation composition and environmental variables. The ecological meaning of clusters of sites was assessed by indicator species. The statistical significance of the species indicator values is evaluated using a randomization procedure. The optimal number of clusters is different from different evaluator criteria, which including 2, 5, 7, 11 and 15. Most of the evaluators are agreement at cluster level of seven. There is an optimal level of the vegetation classification. A too high level lead to a small number of large, unspecific clusters, and that a too low level will on the other hand lead to more specific but very small and too many clusters with environmental variables not significant different. When data of Yellow River Delta vegetation were classified into seven clusters, the species composition and environmental variables are significantly different, and the species have high fidelity. Evaluator criteria such as dunn、silhouette、Calinski and Harabasz and indicator species provide useful information about the level of vegetation classification in the field. We used Ward's method to classify abundance vegetation field data to demonstrate the character of evaluators, an alternative approach (e.g. K-means methods) would be to study other data.