Appendix A. Additional methods for the large-scale gull exclusion experiment as well as some photographs from the experiment.
Our statistical model examining C. borealis densities as the response variable included two random effects [island and site nested within island], three main fixed effects [tide (high, low); depth (0m, -1m); and the number of crab carapaces per low tide (continuous)], and second and third-order interaction terms. The initial analysis by GLIMMIX ANOVA showed that island and site(island) had no significant effect on C. borealis densities. Therefore, we excluded these factors and re-analyzed the data. There were no significant interaction effects among gull density, depth and tide. We excluded all interaction effects with a P > 0.2, leaving only four fixed effects in the final analysis: depth, tide, gull predation rate, and gull predation rate × depth.
Photos from the experiment: (a) C. borealis eating snail in tidepool, (b) C. borealis in algal canopy at low tide, (c) emergent C. borealis in algal canopy at low tide, and (d) emergent crabs.