Statistical analyses

To account for the non-independence of individual observations within a year, and the large differences in sample size among groups, analyses were, where possible, performed on yearly means for the relevant sub-group. When appropriate, yearly means were squareroot or arc-sine transformed before analysis. All presented means and parameter estimates are back-transformed. All statistical analyses were performed using the SAS statistical package using the GLM and GENMOD procedures. Non-significant interactions were removed first, starting with the least significant, followed by non-significant single terms, again starting with the least significant. When interactions were significant, the main effects were kept in the model, regardless of their significance. All tests are two-tailed.