Assessment of mark-recapture models to estimate the abundance of a humpback whale feeding aggregation in Southeast Alaska

The aim of this study was to use photographs of the unique pattern on the ventral surface of the flukes to estimate the abundance of humpback whales (Megaptera novaeangliae) in a discrete feeding aggregation in northern Southeast Alaska. The study was located in northern Southeast Alaska, USA, in the eastern North Pacific Ocean. This study evaluated mark-recapture models, ranging from the simpler models (pooled and stratified, closed Petersen estimators) to more complex multi-strata models (closed Darroch and open Hilborn). The Akaike Information Criterion, corrected (AICc) was used as a model comparison statistic. Our best estimate of whale abundance in northern Southeast Alaska in 2000 is 961 whales [95% confidence interval (657, 1076)]. This estimate comes from the Hilborn open, multi-strata approach with constant migration over time, time-dependent capture probabilities by area, and a fixed survival rate of 0.98. The simpler models were problematic owing to several aspects of whale behaviour, including that (1) the whales did not mix randomly throughout the study area, (2) some whales emigrated temporarily outside the study area and were not available for capture, and (3) whales were not equally identifiable because they did not behave in the same way when they showed their flukes upon diving. This led to heterogeneity in capture probability and a bias in the estimates. The more complex models stratified by area, and using migration movements among areas, compensated for some of these issues when estimating population size. We believe that the Hilborn open, multi-strata model produced the best estimate because: (1) it incorporated the best information about survival, (2) it used detailed information about the various release groups, (3) the analysis provided an integrated environment in which parameters such as migration and capture probabilities are shared, (4) the three strata encompassed a large portion of the areas used by whales, and (5) the Hilborn model selected was superior in terms of model selection criteria and biological realism. These data provide valuable insights into the numbers and movements of humpback whales in three areas of Southeast Alaska.