Automated camera systems deployed at Adélie penguin breeding colonies provide daily measurements that allow high resolution temporal availability functions to be estimated, which in turn can be used to correct population estimates for availability bias. However, such frequent data are time consuming and expensive to process, and it is of interest to determine if such data could be subsampled with significant loss of information content. To this end, a simulation study was undertaken to examine how the frequency of sampling attendance at Adélie penguin breeding colonies would affect models of attendance for correcting population counts for potential attendance bias. Generalised additive models of simulated time-series were shown to adequately recover known structure for sampling periodicities up to five days. Most precise estimates of attendance ratios to correct non-optimal population counts for availability bias are obtained from higher frequency sampling, with a trade-off observed between sampling frequency and precision. Subsampling at periodicities of six days or greater did not adequately recover known simulated model structure and cannot be recommended.
Numéro du document:
Point(s) de l'ordre du jour