Biological assessment of freshwater ecosystems using a reference condition approach: comparing predicted and actual benthic invertebrate communities in Yukon streams

1. The reference condition approach to bioassessment is based on comparing a biological community found at a test site to the range of communities observed at a set of reference sites. A community descriptor (e.g. number of taxa) is estimated for the test site. If the value of the descriptor falls outside of a given boundary, or biocriterion, from the distribution of the descriptors for the reference sites, the test site fails. 2. The sensitivity of the reference condition approach can be increased by modelling and explaining variation in the community descriptor among the reference sites, and then using the predictive model to refine the expectation of the descriptor’s value at a test site. 3. This study applied the reference condition approach, with predictive modelling, to the bioassessment, using benthic macroinvertebrate (BMI) communities, of streams exposed to placer gold mining effluent in central Yukon Territory, Canada. The major changes to the stream caused by mining are increased turbidity and metal concentrations. 4. Among reference sites sampled from 1993 to 1995, a predictive model using year of sampling and simple geographical characteristics (distance to source of the stream, upstream catchment area, altitude, two‐way interactions) explained over half of the variation in each of five BMI community descriptors including richness (number of families), Simpson’s diversity, Simpson’s equitability, family biotic index and Bray–Curtis distance to the median reference community. 5. Biocriteria (other than Bray–Curtis distance to the median community) based on the predictive models failed a far greater proportion of sites currently exposed to placer mining (50–100%) than biocriteria not based on predictive models (7–71%). A similar increase in the sensitivity of the bioassessment was seen when evaluating sites previously exposed to placer mining effluent. The simplest, most sensitive bioassessment of the effects of placer gold mining effluent on stream ecosystems used richness (number of families) of the BMI community together with a predictive model.