Modelling survival and breeding dispersal to unobservable nest sites

Context Demographic parameters in wildlife populations are typically estimated by monitoring a limited number of individuals in observable sites and assuming that these are representative of the whole population. If individuals permanently disperse to unobservable breeding sites, recruitment and immature survival are expected to be negatively biased and breeding-site fidelity cannot be investigated. Aims To develop a method to obtain unbiased estimated of survival, recruitment and breeding dispersal when individuals can move to, or recruit in, unobservable sites. Methods We used the flexibility of multi-event capture-recapture models to estimate dispersal and recruitment to unobservable sites, merging observations made at two sites within the same breeding locations. We illustrated the model with data on little penguin (Eudyptula minor) breeding in artificial as well as in natural nests. Natural nests are unknown or inaccessible and birds in these sites remain unobservable. Encounters at beaches surrounding the colony suggested that marked animals can permanently move to unobservable nests. We built the multi-event model considering two possible states of the individuals (alive breeding in a nest box and alive in a natural nest) and three types of observations (encountered at a nest only, encountered at the beach only and encountered at both places). This model ensured that the breeding dispersal to unobservable places became estimable. Key results Results indicate that the estimated survival was 8% higher than when recaptures at artificial nests were analysed alone. Also, fidelity to artificial nests was 12% lower than to natural nests. This might reflect the greater availability of natural sites or, alternatively, a heterogeneity between these two types of nest. Conclusions We obtained an estimate of local survival of little penguins breeding at Penguin Island that incorporates the permanent migration to unobservable sites and found an asymmetric dispersion towards natural nests. Implication Our conclusions suggest a need for more careful treatment of data derived from artificial sites alone, as demographic parameters might be underestimated if animals prefer natural breeding sites or if they are in greater proportion compared with artificial ones. The analytical approach presented can be applied to many biological systems, when animals might move into inaccessible or unobservable breeding sites.