A study of 518 individually identified humpback whales (Megaptera novaeangliae) in the southern Gulf of Maine between 1979 and 1988 showed high rates of both within-season occurrence and annual return. On average, 77.3% of whales were resighted on more than 1 day during the year (maximum 83 days), with occupancies ranging from 1 to 324 days (mean 88.1 days). The mean rate of return of individuals in consecutive years was 73.2% (range 59.1–87.4%), and the majority of whales were still being resighted up to 10 years later. Only 13 humpbacks were recorded during 50 midwinter surveys. Variation in the occurrence of individuals appears to be related to variability in the abundance of prey and perhaps to regional preferences.

Annual survival probability of bowhead whales, Balaena mysticetus, was estimated using both Bayesian and maximum likelihood implementations of Cormack and Jolly-Seber (JS) models for capture-recapture estimation in open populations and reduced-parameter generalizations of these models. Aerial photographs of naturally marked bowheads collected between 1981 and 1998 provided the data. The marked whales first photographed in a particular year provided the initial 'capture' and 'release' of those marked whales and photographs in subsequent years the 'recaptures'. The Cormack model, often called the Cormack-Jolly-Seber (CJS) model, and the program MARK were used to identify the model with a single survival and time-varying capture probabilities as the most appropriate for these data. When survival was constrained to be one or less, the maximum likelihood estimate computed by MARK was one, invalidating confidence interval computations based on the asymptotic standard error or profile likelihood. A Bayesian Markov chain Monte Carlo (MCMC) implementation of the model was used to produce a posterior distribution for annual survival. The corresponding reduced-parameter JS model was also fit via MCMC because it is the more appropriate of the two models for these photoidentification data. Because the CJS model ignores much of the information on capture probabilities provided by the data, its results are less precise and more sensitive to the prior distributions used than results from the JS model. With priors for annual survival and capture probabilities uniform from 0 to 1, the posterior mean for bowhead survival rate from the JS model is 0.984, and 95% of the posterior probability lies between 0.948 and 1. This high estimated survival rate is consistent with other bowhead life history data.

No abstract

No abstract