What fraction of species do we need to maintain a functioning ecosystem?

Background/Question/Methods While ecologists have historically focused on explaining the causes of biological diversity in nature, a growing body of contemporary research has focused on detailing the ecological consequences of biodiversity. For example, 150+ experiments performed in the last two decades have manipulated the number of species of bacteria, fungi, plants and animals in model ecosystems around the globe to ask how species richness impacts the rates at which organisms capture limiting resources and convert those into new biomass. Here I fit data from these studies to several mathematical functions (log, power, and hyperbolic curves) that allow one to derive a first approximation of the fraction of species that must be conserved if we are to maintain key ecological processes in ecosystems. Results/Conclusions I begin by confirming a pattern that has been shown in many individual studies, which is that the impact of species loss on measured processes is highly non-linear such that, on average, experimental systems require just a tiny fraction (usually < 20%) of the species pool for a community to achieve 95% of the estimated maximal function. However, an important caveat to this conclusion is that most studies have been performed in experimental units that are just a few orders of magnitude larger than the organisms themselves, and the experiments have typically lasted for less than one generation of the focal organisms. When parameter estimates from these models are scaled to account for the spatial and temporal extent of each experiment, estimates for more realistic ecosystems suggest that 80% or more of species may be needed to maintain processes within 95% of their maximal values. Using select examples from the literature, I detail what appears to be a common explanation for these results, which is that small-scale short-duration experiments generally do not permit the expression of niche differences that allow diversity to matter in the first place. These results not only provide a preliminary estimate of how much diversity is needed to maintain a focal ecological processes, they also give a cautionary note for the interpretation of biodiversity studies. When experiments are performed at scales smaller than those at which the focal species coexist, and when a study does not specifically mimic the conditions that allow for niche opportunities, one can grossly underestimate the functional role of biodiversity and over-estimate the apparent level of ‘redundancy’ among species.