Comparison of survival models using mark-recapture rates and age-at-death data for bottlenose dolphins, Tursiops truncatus, along the South Carolina coast

A variety of approaches have been used for modeling survival in wildlife populations. For cetaceans, age-specific survival has been derived by fitting functions to age-at-death data from strandings. Markre-capture models have also been applied to photo-identification (photo-id) data, to determine overall or age-class specific survival rates. Different modeling approaches each have underlying assumptions that, when violated, can introduce bias into the resultant survival estimates. Both photo-id and stranding/life-history studies of bottlenose dolphins have been conducted in South Carolina. This research provides an opportunity to apply and compare survivorship models using data from both stranded carcasses and photo-id surveys. A Bayesian framework was applied to age-at-death data collected from 255 stranded carcasses to create a continuous age-specific survival function using an adjusted form of Siler's competing-risk model. With the application of this model, a young-of-the-year (YOY) survival rate of 0.681 (SE = 0.027, 95% CI: 0.627-0.731) was obtained. This estimate was compared to an independent estimate of YOY survival derived from a Kaplan-Meier product-limit estimator (phi=0.812, SE = 0.064, 95% CI: 0.696-0.948) applied to photo-id data (N=38 surveys). A comparison of the two models find the survival estimate from the Siler model was significantly lower than that derived from photo-id data (p<0.05). These estimates were further compared to previously published estimates of other coastal Tursiops populations to find the Siler estimates reported in this study were the lowest. Conversely, for individuals older than one year, the Siler model produced significantly higher survival estimates (p<0.05) than those derived from a mark-recapture model applied to photo-id data. Sources of bias potentially exist within both of types of analyses, including an overestimation of survival due to high levels of permanent emigration (photo-id), underestimation of neonatal mortality due to unobserved births (photo-id) and unequal probability of carcass recovery (age-at-death). Therefore, the incorporation both types of models into future studies is suggested.