Assessing the power to detect systematic change in Adélie penguin foraging trip duration
Models of variability in Adélie penguin foraging trip durations were constructed and fitted to data collected at Béchervaise Island over a 12-year period when only natural variation was known to occur. Variability among trips and penguins was greater in the crèche stage, but variability among years was greater in the guard stage. Estimates of variability were used to explore the power to detect change under particular impact and monitoring scenarios. Power to detect change was greater in the crèche stage than the guard stage. The gain obtained by increasing the number of penguins or trips sampled diminished rapidly when sample sizes were greater than 30 penguins and three trips per penguin. Statistics were developed to test for three forms of change (step, trend and ramp). A test for change based on the difference between pre- and post-impact means generally performed better than a test based on the slope of a trending post-impact change or a joint test of difference and slope. While foraging trip duration is considered to be sensitive to changes in food availability over time scales of days to weeks, because of the high level of natural between-year variation, it would take many years of post-impact monitoring to detect systematic change with high power unless one were willing to relax the Type I error rate to a rate well above the traditional level of 5%. The strategy of including ice cover as a covariate to explain between-year variation in trip duration increased the power to detect change in the guard stage, but the likely dependence between ice cover and fishing activity could confound interpretation and thus, in this case, this strategy is not recommended.