Incorporating site and year-specific deuterium ratios (δ2H) from precipitation into geographic assignments of a migratory bird

Abstract

The study of migratory connectivity is rapidly growing in ornithology, as is the technology used to measure it. While use of extrinsic markers, such as archival tags, is becoming more prevalent, for many small species the best tool available for tracking birds remains intrinsic markers, such as stable-hydrogen isotope ratios (δ2H). Many researchers have raised concerns that spatial and temporal environmental variation introduces a large amount of error into isotope-based assignments, limiting their utility. Here, using feathers, we sought to address these issues in developing δ2H base maps for assigning pied flycatchers Ficedula hypoleuca of known origin to 15 sites across the breeding range (approx. 4 020 800 km2). We evaluated the effects of including random site variation and year-specific precipitation δ2H (δ2HP) maps on assignments, compared to using mean annual growing season δ2Hp and no site effects. We found a positive correlation between feather δ2H (δ2HF) and mean annual δ2HP, resulting from large scale geographic variation. Repeatability of feather δ2H for individuals sampled in multiple years was strong overall, but variable among populations. Annual variation in δ2HP explained 21% of within individual variation in δ2HF. Neither year- nor site-specific methods improved assignment precision or accuracy. All three methods assigned flycatchers of unknown origin captured at an African overwintering site to similar breeding areas. However, methods using long-term means of δ2Hp assigned birds more precisely than year-specific methods. Our results suggest that annual variation in this system is primarily a result of food web or individual level processes and that random site effects are not strong enough to drastically impact accuracy. We conclude that improvements in isotope based geographic assignments will rely on the addition of prior information, such as relative abundance in a Bayesian framework, or additional intrinsic markers.