Signals from the benthos: Development and evaluation of a benthic index for the nearshore Gulf of Maine

We developed a benthic index for the nearshore Gulf of Maine to provide researchers and environmental managers a way to make spatial and year-to-year comparisons of benthic condition. The data set used included 248 stations sampled for physical, chemical, and biological variables by the National Coastal Assessment in 2000-2003. We used logistic regression with 49 candidate measures of benthic species diversity, pollution sensitivity-tolerance, and community composition to discriminate sites with high and low benthic environmental quality (BEQ). BEQ was based on the concentrations of metal and organic contaminants in the sediments, total organic carbon, sediment toxicity, and dissolved oxygen level of the bottom water. An analysis of similarity test showed that the community composition of low BEQ stations was significantly different (p < 0.001) from high BEQ stations. Ten of the 49 benthic metrics showed a strong ability to discriminate stations. We developed several candidate benthic indices and tested them with independent data from Massachusetts Bay and Casco Bay to help select and validate the best index. A model using the Shannon-Wiener diversity measure, Rosenberg's species pollution tolerance measure, and the percent capitellid polychaetes (or percent Capitella spp.) strongly discriminated stations, with an area under the receiver operating characteristic (ROC) curve of 0.82 and a classification accuracy of 80%. Signal detection theory (ROC curves and positive-negative predictive value curves) was used to evaluate the index and to predict how well an index developed for one geographic area might work in another area with a different prevalence of the degraded condition. We show how these techniques can also guide decisions by environmental managers about choosing thresholds and weighing costs and benefits of particular actions. (C) 2007 Elsevier Ltd. All rights reserved.