Re-analysis of species associational data using bootstrap significance tests

Re-analysis with the bootstrap technique of species associational data obtained in 1973 at a waste discharge site showed that stations previously accepted as apparently similar could actually be significantly different. Two station pairs with almost identical similarities (Bray-Curtis Coefficients of 0.85 and 0.84) were found to be significantly different (P 0.040) in the first case, but not significantly different in the second (P 0.78). The significantly different pair reflected differences that can arise between sampling stations impacted by light deposits of mine tailings. The other clustered pair (not significantly different) comprised two depauperate stations that had received heavy tailings deposition. A similarity of 0.67 between samples obtained using two types of samplers (Van Veen and Ponar) at one station showed a significant difference (P 0.0002). This supported previous suspicion that the samplers were collecting differently, although this had not been demonstrated by the cluster analysis used in 1973. However the use of either sampler's data in the cluster analysis had only trivial effect on similarity levels, and no effect on the significance tests.