Using change-point analysis and weighted averaging approaches to explore the relationships between common benthic diatoms and in-stream environmental variables in mid-atlantic highlands streams, USA

Diatoms are increasingly being used in the bioassessment of aquatic systems. However, autecological information for many common taxa is incomplete. We explored the potential of classification (CT) and regression tree (RT) approaches to identify the hierarchical interaction among water quality variables in predicting the relative abundance of ten common stream diatom taxa in the Mid-Atlantic Highlands ecoregion. RT analysis was also used to identify environmental change points corresponding to major shifts in species abundances. We also used traditional weighted-averaging approaches (WA) to model taxon pH and total phosphorus (TP) optima. RT and WA approaches provided different, yet complementary, information on the complex relationships between common stream diatoms and environmental variables. Both RT and CT highlighted the interaction of stream acidity (pH, acid neutralizing capacity (ANC)), and TP in structuring the stream diatom assemblage. For the RT of taxa, where pH was an important predictor, higher pH predicted higher relative abundances. In contrast, higher TP predicted lower relative abundances for some diatom taxa (e.g., Achnanthidium minutissimum (Kutz.) Czarnecki), while predicting higher relative abundances for other taxa (e.g., Planothidium lanceolatum (Breb.) Round & Bukht., Gomphonema parvulum (Kutz.) Kutz.). The environmental change point for pH derived from RT analysis was lower than WA optima for all species. We suggest that RT change point analysis can be used to complement traditional WA optima approaches, especially when diatom taxa's abundances are affected by interactive environmental factors, to provide more refined information on stream diatom environmental preferences.