Prediction of riverine fish assemblages through the concept of environmental filters

Abstract

Although the taxonomic composition and richness of fish assemblages are important properties to be considered in freshwater bioassessment, conservation and rehabilitation, it can be difficult to establish a natural benchmark for these properties because of widespread human impact and a lack of pristine reference sites or pre-impact data. As an alternative to the reference site approach, the concept of multiple environmental filters was used to predict the assemblages of fish taxa expected in the absence of anthropogenic stress at 85 sites on rivers in north-eastern New South Wales, Australia. The predicted native fish assemblages were compared with the assemblages recorded by backpack and boat electrofishing at each site. The number of native species predicted by the filters model at each site was highly correlated with the observed number of native species (R2 = 0.75; P < 0.001) but the observed number was generally lower. The model had an average sensitivity of 93% and specificity of 87%, but sensitivity and specificity were considerably less for a few species, including some that are known to have suffered historical declines or been translocated outside of their natural ranges. Comparisons between predicted and observed richness and composition can be used to identify areas of high conservation value and areas where native fish assemblages have been adversely affected by anthropogenic impacts.