Abstract
Several multistratum capture—recapture models were used to test various hypotheses about possible geographic and temporal variation in survival, movement, and recapture/resighting probabilities of 2399 adult Roseate Terns (Sterna dougallii) color—banded from 1988 to 1992 at the sites of the four largest breeding colonies of this species in the northeastern USA. Linear—logistic ultrastructural models also were developed to investigate possible correlates of geographic variation in movement probabilities. Based on goodness—of—fit tests and comparisons of Akaike's Information Criterion (AIC) values, the fully parameterized model (Model A) with time— and location—specific survival, movement, and capture probabilities, was selected as the most appropriate model for this metapopulation structure. With almost all movement accounted for, on average >90% of the surviving adults from each colony site returned to the same site the following year. Variations in movement probabilities were more closely associated with the identity of the destination colony site than with either the identity of the colony site of origin or the distance between colony sites. The average annual survival estimates (0.74—0.84) of terns from all four sites indicate a high rate of annual mortality relative to that of other species of marine birds.