Predicting phytoplankton composition in the River Thames,England

Phytoplankton were sampled fortnightly in summer and monthly in winter for 33 months to determine temporal and spatial variations in taxonomic structure and relate these to the physical and chemical environment. Using canonical correspondence analysis, environmental variables selected by stepwise regression were found to be correlated with the major variable(s) determining phytoplankton composition. An improved proportion of explained variance was achieved by classifying data into three seasonal subsets. The highest percentage of explained variance in phytoplankton composition was achieved for the months March to May (36·5%), the lowest variance explained was for the period June to September (28·6%). Within each season, general linear modelling was used to predict taxa responses to variations in particular environmental variables. Taxa responses to discharge are presented as examples for determining river management policies. The rationale of conceptual models and the empirical models used for this study are discussed. © 1997 John Wiley & Sons, Ltd.