Apples, oranges, and ecoregions: on determining pattern in aquatic assemblages

Classification of waterbodies is a necessary step for aquatic biological assessment. Geographic classes, such as physiographic provinces, marine biogeographic provinces, and ecoregions are frequently used as a classification framework for conducting bioassessments. An alternative to such a priori classification of waterbodies is to let the biological data tell us, using various statistical exploratory techniques to identify classes that do not necessarily conform to geographic stratification. In assessment of new sites, it is often necessary to identify or predict which of several classes the new site may belong to. We classified benthic macroinvertebrate assemblages from streams of Wyoming, using 3 alternative methods: a priori establishment of ecoregions, adaptive modification of the ecoregions, and a posteriori clustering. Classification strengths were compared using similarity dendrograms. Clustering separated the sites better than ecoregions, but the clusters showed strong ecoregional affinities. The modified ecoregions were equal to the clusters in classification strength. We conclude that an iterative process that includes generation of hypotheses, exploratory data analysis, and evaluation and modification of hypotheses is most likely to produce robust classifications, regardless of specific analytical approaches used.