Evaluating photographic capture-recapture estimates of abundance of North Atlantic humpback whales

The Years of the North Atlantic Humpback (YoNAH) project collected photographs of flukes of humpback whales to identify individuals for estimating abundance using capture-recapture techniques. Data were collected during the winter breeding and summer feeding seasons in 1992 and 1993. Smith et al. (in prep) identified possible violations in three of the assumptions needed for unbiased abundance estimators: closed population, equal probabilities of capture, and the identification of marks. To evaluate the assumption that all animals were recognized when photographed again, methods for coding photographic quality and the distinctiveness of whales were developed and applied to the YoNAH photographic database. Analytical methods were then developed to evaluate the effects of photographic quality and whale distinctiveness on the abundance estimates produced from this database. To evaluate the assumptions of a closed population and equal probabilities of capture, a model of the probability of sampling individual humpback whales was developed. The error rate of the photographic quality coding process for the YoNAH database was found to be a function of the quality of the photographs, such that poor quality photographs were less likely to change code on a second coding. Poor quality photographs were found to cause a potential positive bias in the abundance estimates. Analyses indicated that removing photographs of the lowest quality level and images of partial and right or left half flukes provided the optimum balance between precision and bias. The coding of whale distinctiveness independently of photographic quality proved difficult for some coders. The distinctiveness of whales represented by at least one high quality photograph did not significantly affect their recapture rate. Simulations of the YoNAH sampling project indicated that estimators that compared data collected during a breeding season to data collect during a feeding season within the same year were more robust than other estimators to violations of the equal probability of capture assumption and the closed population assumption. However, variability in the individual probability of sampling significantly affected all estimators. Areas where additional data are needed to measure the bias in these abundance estimates are identified