[pp.11-20]

Does hierarchical cluster analysis perform better than TWINSPAN? A comparative study using field and simulated data sets

Kazuhiro KATOH （The University of Tokyo）

Abstract:
Though TWINSPAN is a popular classification method in ecological community analysis, it has sometimes been criticized as that the performance of detecting species compositional change pattern is poorer than that of hierarchical cluster analysis. In the present study the performance was compared between the two methods, considering compositional change pattern with nestedness. Nestedness is the structure where species present at species-poor sites are subsets of those present at species-rich sites. It is widely observable and has recently been in the spotlight in ecological studies. The results indicate that TWINSPAN is more robust against noise in species occurrence. Among methods of hierarchical cluster analysis, complete linkage clustering performed better than group average method, though the use of the latter is commonly recommended. When the data are qualitative, or the expected data structure is rather simple (e.g., with nestedness), the use of TWINSPAN is more appropriate than that of hierarchical cluster analysis.

Key Word:
classification, hierarchical cluster analysis, nestedness, noise, species composition, TWINSPAN

PDFダウンロード（会員限定）