On dendrogram‐based measures of functional diversity

TLDR

Euclidean distance is commonly used to calculate functional diversity via dendrogram branch lengths, but Gower's measure is preferable because it handles mixed scale types and missing values. The authors propose using Gower's formula with UPGMA clustering as a standard method for computing functional diversity and suggest evaluating species' effects by removing them and comparing resulting tree lengths. The methodology combines Gower's distance with UPGMA clustering, evaluates species influence through removal experiments, and assesses trait importance by examining changes in dendrogram length and topology. They show that Euclidean distance is inappropriate in many cases, that clustering choice matters more than previously thought, and that UPGMA clustering is more robust with a better goodness‑of‑fit to dissimilarities than complete or single linkage.

Abstract

Euclidean distance is commonly involved in calculating functional diversity (FD), for example, in measures based on dendrogram branch lengths. We point out that this function is inappropriate in many cases and that the choice of clustering method is more crucial than earlier thought. Gower's formula and UPGMA clustering are suggested here as a standard combination of techniques for calculating FD. The advantage of Gower's measure is its suitability to a mixture of scale types and its tolerance to missing values. Examples demonstrate that UPGMA clustering is more robust and has a better goodness of fit to dissimilarities than complete and single linkage classifications. In addition, we propose that the effect of individual species on FD is best evaluated by species removals and subsequent comparisons of tree length values. The influence of each functional trait is optimally judged by considering both dendrogram length and topological changes.