Assessing species diagnostic value in large data sets: A comparison between phi-coefficient and Ochiai index

Question: Diagnostic species are useful tools for the identification and ecological interpretation of community types. Vegetation databases facilitate the computation of diagnostic values of regional validity, but it is essential to understand the behaviour of fidelity measures in large data sets. Methods: We focused our study on the phi-coefficient (Phi) of association and its limit value, the Ochiai index. The northeast Spanish releve database was stratified using an arbitrary distance threshold in species composition. Diagnostic species analysis was undertaken using three methods of context selection: I. within a syntaxon of higher rank; II. including releves with similar composition to that of the target unit; III. using the entire stratified database. Species diagnostic values were computed as well as bootstrap percentile confidence intervals. Results: Many species deemed as diagnostic by method I have their optima in vegetation types neighbouring the unit chosen as context. In contrast, method II excluded many of these species. Phi-values and confidence intervals were similar to those obtained by the Ochiai index when using a large dataset (method III) but this similarity was greater for low level syntaxa. Conclusions: The diagnostic value of species in a given region is best assessed using the Ochiai index, since it can be split into two interpretable asymmetrical components. We recommend the determination of context-dependent differential species using the Phi-coefficient, and the assessment of species regional diagnostic value by means of a stratification procedure in combination with the Ochiai index.