The Implications of Experimental Design for Biodiversity Manipulations.

I report a simulation study that tested the ability of a variety of experimental designs to achieve two fundamental goals: (1) to determine the association between loss of biological diversity and responses such as ecosystem functioning and (2) to determine which components of biodiversity, such as number of species, functional diversity, or a keystone species, were most responsible for that association. For the goal of reliably detecting an overall association, all designs I tested performed well and were unlikely to misidentify predominant patterns. Thus, this study affirms the common conclusion of many published biodiversity experiments that loss of biological diversity is often associated with a reduction in ecosystem functioning. However, for the goal of identifying the components of biodiversity that are most responsible for the effects, designs differed markedly. Some designs performed well in detecting number-of-species effects but poorly in detecting effects of unique species or functional groups. No design tested was able to discriminate both numeric effects and compositional effects in all situations. Thus, this study demonstrates that interpreting results about mechanisms from biodiversity experiments will be critically dependent on an experiment's design.