The Relationship Between Sample Similarity and Ecological Distance

Similarity measures for samples from natural communities show a complex, curvilinear decrease with increasing separation of samples along environmental gradients. The form of this decrease for samples without sampling errors has been analyzed and found to be a complement of a non—standardized error function for percentage similarity, and similar functions for coefficient of community and Euclidean distance. For samples affected by sampling error, altered and somewhat flattened curves result. These relationships are important in many ordination techniques. In particular, it is demonstrated that Bray—Curtis ordination can be improved by application of the inverse function transforms with moderate beta diversity (up to 5 half—changes). Such transforms produce similarity measures which are linear with respect to separation along the gradient, up to the point beyond which samples have practically no species in common and simlarity measures of any sort are consequently meaningless. See full-text article at JSTOR