Explaining maximum variation in productivity requires phylogenetic diversity and single functional traits.

Many community experiments have shown a positive relationship between plant biodiversity and community productivity, with biodiversity measured in multiple ways based on taxonomy, function, and phylogeny. Whether these different measures of biodiversity and their interactions explain variation in productivity in natural assemblages has rarely been tested. In a removal experiment using natural alpine assemblages in the Tibetan Plateau, we manipulated species richness and functional diversity to examine how different measures of biodiversity predict aboveground biomass production. We combined different biodiversity measures (functional, phylogenetic, richness, evenness) in generalized linear models to determine which combinations provided the most parsimonious explanations of variation in biomass production. Although multivariate functional diversity indices alone consistently explained more variation in productivity than other single measures, phylogenetic diversity and plant height represented the most parsimonious combination. In natural assemblages, single metrics alone cannot fully explain ecosystem function. Instead, a combination of phylogenetic diversity and traits with weak or no phylogenetic signal is required to explain the effects of biodiversity loss on ecosystem function.