Sensitivity and variability of metrics used in biological assessments of running waters

We evaluated the variability and sensitivity of a suite of biological metri cs for detecting the ecological effects of metals in streams. The variabili ty of these metrics was evaluated by partitioning the total variance in a t hree-way analysis of variance among spatial, seasonal, annual, and temporal -spatial interaction components using data from 6 years of biomonitoring on the Arkansas River, Colorado, USA. We then calculated the statistical powe r of these metrics given a likely experimental design aimed at detecting me tal-pollution effects in streams and using estimates of variability from fi eld data. Finally, we experimentally tested the sensitivity of these metric s to a metal mixture in stream microcosms. More than one half of the variat ion in richness and scraper functional feeding group metrics was explained by differences among sampling sites, which were presumably due to the prese nce of metal pollution. Statistical power was highest for richness measures and low for all other metrics examined. Experimental exposures revealed th at richness measures were also more sensitive than functional group metrics . Our results support those of previous, comparative studies that show the superiority (in terms of sensitivity, variability, and statistical power) o f taxa richness measures. Most abundance, ratio, and functional group metri cs were either insensitive to metal pollution, highly variable, or both. We conclude that similar systematic testing on a variety of metrics with othe r stressors will greatly enhance the utility of biological metrics in asses sing the ecological effects of contaminants and establishing biological cri teria.