Characterizing Common Diatom Species Distributions in the Mid-Atlantic Streams: A Statistical Modeling Approach

Autecological classifications of benthic diatoms in streams have been primarily qualitative. We used several statistical methods to characterize the relationships between some common benthic diatom species and associated environmental variables in the Mid-Atlantic streams. A total of 306 stream riffle sites were sampled for benthic algae, water quality, and physical habitat conditions as part of EPA's Environmental Monitoring and Assessment Program (EMAP). We selected 15 most abundant and widely occurred diatom species for our analyses. Generalized additive model, a semi-parametric method, showed that the species had significant linear or nonlinear relationships with environmental variables such as pH and total phosphorus. However, the percentages of individual species' variance explained by its corresponding environmental variables were in general low (<30%). Regression tree analysis, a binary recursive partitioning method, was used to capture possible interactive effects of these environmental variables on diatom species. The results may provide a quantitative base for improving current diatom autecological classifications.