A linear model method for biodiversity-ecosystem functioning experiments

Abstract

Experiments that manipulate species richness and measure ecosystem functioning attempt to separate the effects of species richness (the number of species) from those of species identity. We introduce an experimental design that ensures that each species is selected the same number of times at each level of species richness. In combination with a linear model analysis, this approach is able to unambiguously partition the variance due to different species identities and the variance due to nonlinear species richness, a proxy measure for interactions among species. Our design and analysis provide several advantages over methods that are currently used. First, the linear model method has the potential to directly estimate the role of various ecological mechanisms (e.g., competition, facilitation) rather than the consequences of those mechanisms (e.g., the "complementarity effect"). Second, unlike other methods that are currently used, this one is able to estimate the impact of diversity when the contribution of individual species in a mixture is unknown.