INFERENTIAL UNCERTAINTY AND MULTIPLE LINES OF EVIDENCE IN DETECTING ECOLOGICAL IMPACTS.

B.J. Downes. School of Anthropology, Geography & Env. Studies, University of Melbourne, Melbourne, Victoria, 3010, Australia

Deciding when an environmental change has been caused by human activities is often difficult. BACI-type survey designs offer the best method for inference, but many situations do not lend themselves to such designs, either because putative human impacts happened long ago and/or there are no control locations that are not also potentially changed by human activities. Lack of data from before human activities began or from control sites creates inferential uncertainty, not (just) a problem in statistical analysis. The solution to this difficulty is to use multiple lines of evidence to try to rule out alternative hypotheses that, on the face of it, can explain any environmental changes equally well. Such lines of evidence can be gathered using a series of circumstantial arguments and published data in the literature, which are combined in a structured way to assess the evidence for and against human impact hypotheses. When a BACI-design is at least partly possible, this approach can still complement the monitoring program by providing a variety of hypotheses that may help either corroborate or disprove different explanations for any environmental changes. A levels-of-evidence approach can also help with critical aspects of design like deciding upon the character of control locations.