Estimation of age-specific survivorship when age is unknown

Although the concept of bird mortality (or survivorship) changing with age is not new in ecology, its estimation remains limited. Traditional methods to calculate age-specific survivorship in birds and other groups rely on the availability of data on individual ages. This demographic trait is usually calculated from mark-recapture techniques and is constrained to those individuals for which at least the birth year (right-censored) is known and, when possible, those for which the year of death has been recorded (uncensored). As expected, this reduces considerably the number of capture histories that can be used if a large proportion of the dataset consists of birds first captured as adults (left-censored) with no information on their birth and death years. Thus, the inference process requires a modeling structure that can deal with this joint probability of unknowns: age for left-censored individuals; and the age-specific mortality rate, while using the information provided by the capture histories for uncensored individuals. Moreover, a hierarchical structure is required when recapture probability is low and when juvenile mortality is different from adult mortality. Here we developed a hierarchical Bayesian approach that assumes a parametric form for the age-specific mortality rate and that allows the estimation of ages for left-censored individuals while allowing for different sources of uncertainty. We applied this model to estimate age-specific survivorship and detection probability for a colony of sooty terns at the Dry Tortugas National Park.