Predicting diatom communities at the genus level for the rapid biological assessment of rivers

1. Diatoms were sampled in the spring of 1994 and the autumn of 1995 at 137 pristine or near-pristine reference sites on large and small streams at various altitudes in eastern New South Wales and Victoria. Scrapings were taken from five firm substrata across a range of microhabitat conditions at each site on each occasion. For each substratum, 100 valves were identified to genus level. 2. Multivariate statistical models were constructed to predict the probability of occurrence of each genus at a given site under near-pristine conditions on the basis of physical features of the site that are not affected by human activity. Model predictions were compared with the flora observed at the reference sites and at 55 test sites affected by varying degrees of human disturbance. 3. Test sites were characterized more by the presence of genera not predicted by the model than by the absence of predicted genera. The degree of departure from model predictions was related mainly to increasing alkalinity, electrical conductivity, hardness and pH of river water. We therefore conclude that the main effect of human activity on the composition of diatom communities in the rivers of coastal south-eastern Australia has been to increase genus richness through enrichment with alkaline salts. 4. The models did not perform as well as similar models applied to river macroinvertebrates at the family level, perhaps because of greater temporal variability in diatom communities and differences in the environmental variables to which diatoms and macroinvertebrates respond. Means of improving the current models are discussed.