Morphological drivers of avian differential migration patterns

Background/Question/Methods

Bird banding data collected at Hueston Woods Biological Station over 14 years of bi-weekly banding during fall and spring migration was used to examine morphological and migrational patterns. Ideal Randomized Trees were used to identify migrational communities, taxonomically specific migrational behaviors, migrational behavior under intra-specific competitive selection and environmental constraints on migration. An Ideal Randomized Tree is a novel statistical method where a conditional inference tree is informed by ensemble methods and validated on held out data.

Results/Conclusions

Species-specific Ideal Randomized Trees were not as informative in discerning differences between patterns that result from intra-specific competition and those resulting from environmental constraints as a single community tree where taxonomic information, and group proportionate variables were used in addition to absolute measurements as predictive variables.  Early and late migrating communities had distinctly different morphological drivers of arrival, as did certain taxonomic groupings. In spring, interspecific competition was more prevalent among early arriving communities. Arrival date during fall migration was driven more by environmental constraints on morphology. Important thresholds in fall were based on weight. In spring, most groups split in migrational pattern based on sex and age class.