Variability of community metrics: Detecting changes in structure and function

Increased environmental realism in toxicity testing has been advocated to better predict ecological effects of toxic chemicals, and several aquatic microcosm bioassays have been developed that use both natural and synthetic organism assemblages. Additionally, greater emphasis is being placed on community changes as evidence of adverse effects on aquatic ecosystems. The ability to infer structural or functional changes in stressed vs. unstressed communities is linked to statistical power that arises from the combination of experimental design and the underlying variability in chosen metrics. We evaluated several community response variables (metrics) in aquatic laboratory microcosm experiments and field studies to estimate metric variability. Variability of community metrics under laboratory conditions was similar to that observed in the field. Power curves were constructed to estimate the detectability of significant responses with different experimental designs. Using the median C. V. for each measured response, we estimated the minimum detectable distance for power of 0.8 (β = 0.2) and α = 0.05 for each response. Community metrics with C.V.s <30% allow detection of differences between means of 20 to 60%, using as few as three replicates. Structural metrics such as species richness and standing crop are expected to prove more useful in detecting community and ecosystem changes. Process rates are affected by material supplies that may not change under stress.