Ecological Indicators 3 (2003) 195–202 Effects of two classification strategies on a Benthic Community Index for streams in the Northern Lakes and Forests Ecoregion Jason T. Butcher a,∗,1 , Paul M. Stewart b , Thomas P. Simon c a U.S. Geological Survey, Lake Michigan Ecological Research Station, 1100 N. Mineral Springs Road, Porter, IN 46304, USA b Department of Biological and Environmental Sciences, Troy State University, Troy, AL 36082, USA c U.S. Fish and Wildlife Service, 620 S. Walker St., Bloomington, IN 47403, USA Accepted 15 April 2003 Abstract Ninety-four sites were used to analyze the effects of two different classification strategies on the Benthic Community Index (BCI). The first, a priori classification, reflected the wetland status of the streams; the second, a posteriori classification, used a bio-environmental analysis to select classification variables. Both classifications were examined by measuring classification strength and testing differences in metric values with respect to group membership. The a priori (wetland) classification strength (83.3%) was greater than the a posteriori (bio-environmental) classification strength (76.8%). Both classifications found one metric that had significant differences between groups. The original index was modified to reflect the wetland classification by re-calibrating the scoring criteria for percent Crustacea and Mollusca. A proposed refinement to the original Benthic Community Index is suggested. This study shows the importance of using hypothesis-driven classifications, as well as exploratory statistical analysis, to evaluate alternative ways to reveal environmental variability in biological assessment tools. © 2003 Elsevier Ltd. All rights reserved. Keywords: Bioassessment; Biocriteria; Index of biotic integrity; Stream classification; Partitioning; Multivariate analysis 1. Introduction Biological assessments rely on comparison of an expected biological condition to an observed biologi- cal condition. In the case of assessing the impact of a point source in a stream, upstream stations can serve as expected conditions. However, biological assessments in areas impacted by non-point sources require differ- ent methods to attain biological expectations. These expectations must be derived from measures of a simi- ∗ Corresponding author. Tel.: +1-218-626-4307; fax: +1-218-626-4398. E-mail address: jtbutcher@fs.fed.us (J.T. Butcher). 1 Present address: USDA—Forest Service, Superior National Forest, 8901 Grand Ave. Place, Duluth, MN 55808, USA. lar, pristine environment, from historical data, or from reference conditions (Hughes et al., 1986). The later is usually the best-case scenario, since pristine condi- tions are rare for most ecosystems and historical data are usually limited. Indeed, current bioassessment pro- grams rely heavily on reference conditions (Barbour et al., 1999). Without regard for the possibility of further group- ings of reference conditions, for example grouping wetland sites with non-wetland sites within the same biogeographical area, setting biological expectations is futile. Therefore, regional reference conditions may require classification at a finer scale (Hughes, 1995). Factors such as catchment area, stream order, local geology, or stream gradient may be more im- portant in creating homogenous sets of communities 1470-160X/$ – see front matter © 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S1470-160X(03)00043-8