Biological assessments of Appalachian streams based on predictive models for fish, macroinvertebrate, and diatom assemblages

We developed empirical models for fish, macroinvertebrate, and diatom assemblages to assess the biological condition of 268 streams sampled from 1993 to 2002 in 7 major river basins in the Appalachian region of the USA. These models estimate the expected taxonomic composition at each site based on observed variation in taxonomic composition at reference sites. The index, O/E, is the ratio of the number of predicted taxa that were observed (O) to that expected (E) to occur at a site and is a measure of taxonomic completeness. We compared how O/E for each assemblage varied among major landuse settings and whether impaired assemblages were associated with particular physicochemical conditions. We also examined concordance among assemblages in their response to stress. Biological, chemical, and physical data were collected following consistent protocols. We used land-cover criteria, published data, and topographic maps to classify sites by major landuse setting. Fish, macroinvertebrate, and diatom assemblages had been sampled at 73, 108, and 52, respectively, of the least disturbed sites used to establish reference conditions. The models accounted for a substantial portion of the natural variation in taxonomic composition across sites that was associated with biogeographic, climatic, and basin-scale factors and generally were unbiased across the range of environmental gradients observed in the region. Assessments at nonreference sites showed that impairment of fish and macroinvertebrate assemblages was most strongly associated with agriculture and urban land uses, whereas impairment of diatom assemblages was most strongly associated with mining in the basin. Concordance in assessments among assemblages was not strong. Assessments based on 2 assemblages differed in 28 to 57% of cases, and assessments were never concordant for cases where all 3 assemblages were sampled. Furthermore, only ½ of these cases would have been assessed as ecologically impaired had only 1 assemblage been sampled. Differences between observed and predicted frequencies of occurrence for individual taxa were generally consistent with known tolerances to environmental stressors and might aid in identifying causes of biological impairment.