A bird distribution model for ring recovery data: where do the European robins go?

For the study of migratory connectivity, birds have been individually marked by metal rings for more than 100 years. The resulting ring recovery data have been compiled in numerous bird migration atlases. However, estimation of what proportion of a particular population is migrating to which region is confounded by spatial heterogeneity in ring recovery probability. We present a product multinomial model that enables quantifying the continent-wide distribution of different bird populations during different seasons based on ring recovery data while accounting for spatial heterogeneity of ring recovery probability. We applied the model to an example data set of the European robin Erithacus rubecula. We assumed that ring recovery probability was equal between different groups of birds and that survival probability was constant. Simulated data indicate that violation of the assumption of constant survival did not affect our estimated bird distribution parameters but biased the estimates for recovery probability. Posterior predictive model checking indicated a good general model fit but also revealed lack of fit for a few groups of birds. This lack of fit may be due to between-group differences in the spatial distribution on smaller scales within regions. We found that 48% of the Scandinavian robins, but only 31% of the central European robins, wintered in northern Africa. The remaining parts of both populations wintered in southern and central Europe. Therefore, a substantial part of the Scandinavian population appears to leap over individuals from the central European population during migration. The model is applied to summary tables of numbers of ringed and recovered birds. This allows us to handle very large data sets as, for example, those presented in bird migration atlases.