Whole ecosystem manipulation: A productive avenue for test system research?

Single‐species, microcosm and mesocosm toxicity tests frequently have been shown to be inadequate in assessing impacts on whole ecosystems, yet effective environmental management requires attention to such impacts. Whole ecosystem manipulation provides community and ecosystem level assessment, but is expensive and logistically demanding. Thus most whole ecosystem experiments are pseudoreplicated, which gives them limited inferential capability. We present the results of two whole ecosystem experiments — dosing of stream channels with chlorine and ammonia and acidification of a lake — and review several other such experiments from the literature. Most whole ecosystem studies are complex, expensive, involve multiple investigators and have a low degree of true replication. Difficulty in replicating whole system manipulations makes quantifying cause‐effect relationships difficult, but this can be partially overcome by carefully developing the inferences from an investigation. In sum, we suggest that such experiments do indeed represent a useful test system, and that they are the only acceptable way of discerning impacts at the community and ecosystem levels.