Résumé
Long-lived species show delayed maturity and generally skip breeding sites on a given year, causing difficulties in estimating demographic parameters. A novel multi-state capture recapture model (model G for general) is proposed to estimate survival and recruitment. Model G considers long and short periods of non-attendance at the breeding site. Model G is compared against a reduced model (model R) to test if pre-breeders skip the studied site a given year, potentially prospecting other breeding habitats. This model is applied to the vulnerable Cory's shearwaters seabird species. As for many species, natal philopatry, prebreeding attendance and prospecting play an important role for recruitment into the breeding population. Model G shows that all Cory's shearwaters come back to their natal colony, and attend the colony before recruiting into the breeding population. Moreover, model G is preferred against model R, highlighting that prospecting plays a potentially role in the recruitment processes. Return and recruitment probabilities show similar time variation, suggesting an environmental forcing. The maximum probability of the first breeding is reached at the age of seven years. During the first year the annual survival rate is 0.52 (IC95%[0.40, 0.64]) and pre-breeders and adult annual survival rates do not differ, being respectively 0.88, IC95%[0.83, 0.92] and 0.89, IC95%[0.88, 0.9]. Model G provides an improvement to estimate demographic parameters for long-lived species life cycle and an adequate framework to test the influence of covariates. Model G is structured by age and breeding status categories, allowing easy further population models.