Evaluating ecological indicators: lakes in the northeastern United States

We use data from a survey of several hundred lakes in the northeastern United States by the U.S. Environmental Protection Agency to illustrate an approach to identifying promising indicators of lake condition. We construct a hypothetical gold standard of water quality from the first principal component of 16 chemical variables measured in the lakes, and examine its associations with 71 candidate indicators based on measurements of human activity, birds, fish and zooplankton in the lakes or their watersheds. Nonparametric summaries of these associations--based on rank correlations and receiver-operating-characteristic curves--suggest that variables summarizing the extent of human disturbance are generally the strongest indicators. To the extent that our water-quality variable is a useful proxy for ecological condition, our results suggest that easily-obtained measures of human activity are at least as predictive as many of the harder-to-measure biological indicators that have been proposed.