On similarity among local communities in biodi ersity experiments

As a potential mechanism to explain how biodiversity loss may influence variability in ecosystem functioning, we examine the hypothesis that biodiversity loss lowers similarity in species composition among local communities and that this decreased similarity in turn lowers ecosystem reliability. Ecosystem reliability refers to the probability that a system will provide a consistent level of performance over a given unit of time. This hypothesis is compared with other hypotheses that make similar predictions, including the sampling effect, insurance, and resource use complementarity hypotheses. We provide evidence for the similarity hypothesis through a reanalysis of a recent experiment and show that a key assumption of the hypothesis may be robust through computer simulations. We also address problems and possible solutions regarding how to separately test the similarity and other hypotheses in biodiversity experiments.