Abstract Multivariate methods for the detection of impacts on communities are: (i) explored in the context of a disturbance experiment at a site of previous uranium mining in Kakadu National Park, Northern Territory; and (ii) applied to the design of a monitoring programme for a proposed mining development at nearby Coronation Hill. The disturbance experiment in Rockhole Mine Creek used a gravity-fed redirection of polluted water from the mine adit to create a point-source pollutant. Weekly samples of the macroinvertebrate communities were taken both before and after disturbance at the control and‘impact’ site, as part of a multivariate BACIP (Before-After-Control-Impact-Pairs) design. The results of the experiment show that the multivariate approach provided high sensitivity in detecting impacts on the community, with little difference in sensitivity between species and family level analyses. A simple procedure for deleting those taxa that show poor discrimination between pristine and presently disturbed areas further increased the sensitivity of the tests. In addition to the BACIP analyses, robust ordination of the community data suggested a time/treatment interaction effect.
Application of the multivariate BACIP approach also was explored using baseline data for benthic communities from the South Alligator River near Coronation Hill. The method provides high statistical power for monitoring; a relatively small change in community dissimilarity would be detectable in a programme with 5 years of observations before and after commencement of mining. A robust ordination of the yearly samples for the control and potential-impact sites showed that, although samples varied along identifiable gradients over time, the two areas changed in parallel so that the difference between the areas was approximately constant over years. This result demonstrates that the multivariate BACIP approach can provide relative constancy of pre-impact difference values, a key requirement of BACIP designs that is difficult to satisfy using abundance data for a single species.