Biodiversity-ecosystem function relationships change in sign and magnitude across the Hill diversity spectrum.

Motivated by accelerating anthropogenic extinctions, decades of biodiversity-ecosystem function (BEF) experiments show that ecosystem function declines with species loss from local communities. Yet, at the local scale, changes in species' total and relative abundances are more common than species loss. The consensus best biodiversity measures are Hill numbers, which use a scaling parameter, ℓ , to emphasize rarer versus more common species. Shifting that emphasis captures distinct, function-relevant biodiversity gradients beyond species richness. Here, we hypothesized that Hill numbers that emphasize rare species more than richness does may distinguish large, complex and presumably higher-functioning assemblages from smaller and simpler ones. In this study, we tested which values of ℓ produce the strongest BEF relationships in community datasets of ecosystem functions provided by wild, free-living organisms. We found that ℓ values that emphasized rare species more than richness does most often correlated most strongly with ecosystem functions. As emphasis shifted to more common species, BEF correlations were often weak and/or negative. We argue that unconventional Hill diversities that shift emphasis towards rarer species may be useful for describing biodiversity change, and that employing a wide spectrum of Hill numbers can clarify mechanisms underlying BEF relationships. This article is part of the theme issue 'Detecting and attributing the causes of biodiversity change: needs, gaps and solutions'.