Robust techniques for regionalization reveal historical and ecological fingerprints on species assemblages

Abstract

Elucidation of beta-diversty indices has led to renewed interest in regionalization methods and applications. The use of Sørensen or Jaccard indices paired with hierarchical clustering has become the preferred method. These indices can be partitioned into turnover (species replacements between areas) and nestedness (the ordered increase of richness among areas) components. Turnover is largely influenced by historical factors (vicariance and endemism), thus being a favoured choice for identifying regionalization. Nevertheless, turnover indices produce a large number of low (zero) dissimilarity values that generate multiple equivalent solutions in most clustering algorithms and can produce highly biased results. We developed consensus techniques to solve this issue that are available in the R package recluster. Using butterflies as a model system we have shown how these novel techniques can unravel complex biogeographic fingerprints for historical and ecological determinants over the Mediterranean and expose how the different components of beta-diversity over the UK have changed in time, probably in response to natural and anthropogenically induced environmental changes.