Abstract

As part of investigations into strategies for biological monitoring of mining impacts in the vicinity of the Kakadu Conservation Zone, statistical procedures were evaluated in nearby Rockhole Mine Creek, a site of past mining activities. The BACI design and associated statistical test is based on temporal replication of some measure of difference between paired control and impact areas, and it requires that the difference values meet certain statistical requirements while providing adequate statistical power. In investigating extensions of the BACI design to community-level data, a simple gradient-based model of control and impact sites is described that yields predictions concerning the relative performance of different community dissimilarity measures. Abundance data for benthic macroinvertebrates based on the first 2 years of our study at Rockhole Mine Creek are then used to evaluate the potential statistical power of these alternative dissimilarity measures. The measures vary widely in their statistical power, with the Bray-Curtis measure providing high power over both years.