Comparison of Ordination Techniques

Eight ordination techniques are described and their performances measured by two statistics under a variety of conditions using simulated coenocline data. The Bray—Curtis ordination using S°rensen's coefficient of community, is found to be best in general, and Principal components analysis the worst. The resemblance functions are found to work from best to worst in the order: coefficient of community, percentage similarity, and Euclidean distance. Beta diversity has a very strong impact upon ordination efficacy, and sampling errors exert some effect; relativizing data or varying alpha diversity has relatively little effect. Causes of these results are implications for improving ordination techniques are discussed.