The question of whether migratory birds track a specific climatic niche by seasonal movements has important implications for understanding the evolution of migration, the factors affecting species’ distributions and the responses of migrants to climate change. Despite much research, previous studies of bird migration have produced mixed results. However, whether migrants track climate is only one half of the question, the other being why residents remain in the same geographic range year-round. We provide a literature overview and test the hypothesis of seasonal niche tracking by evaluating seasonal climatic niche overlap across 437 migratory and resident species from eight clades of passerine birds. Seasonal climatic niches were based on a new global dataset of breeding and non-breeding ranges. Overlap between climatic niches was quantified using ordination methods. We compared niche overlap of migratory species to two null expectations, 1) a scenario in which they do not migrate and 2) in comparison to the overlap experienced by closely related resident species, while controlling for breeding location and range size. Partly in accordance with the hypothesis of niche tracking, we found that the overlap of breeding vs. non-breeding climatic conditions in migratory species was greater than the overlap they would experience if they did not migrate. However, this was only true for migrants breeding outside the tropics and only relative to the overlap species would experience if they stayed in the breeding range year-round. In contrast to the hypothesis of niche tracking, migratory species experienced lower seasonal climatic niche overlap than resident species, with significant differences between tropical and non-tropical species. Our study suggests that in seasonal non-tropical environments migration away from the breeding range may serve to avoid seasonally harsh climate; however, different factors may drive seasonal movements in the climatically more stable tropical regions.

Range maps migratory birds:

The GeoMiB database was created specifically to produce the best possible range map for each species classified as at least partially non-resident (migratory or nomadic) in Eyres et al. (2017), by using the best source for seasonal range maps for each region and combining each species’ seasonal range map across regions. The distribution maps in the GeoMiB database are polygon range maps or atlas data (which were summarized into a 1° latitudinal-longitudinal grid and turned into polygon range maps), depending on the source. Sources were chosen as the best available information on species’ seasonal range maps for a region, so maps were compiled for most species from multiple sources. Each source was either downloaded from existing online databases or species maps were scanned from books, geo-referenced, and digitized into a geodatabase (Table 2). As some sources covered the same regions, we applied a priority algorithm that overlaid the source maps in a specific order and then deleted areas covered by higher-priority sources in the lower-priority source maps to derive a final species map (from highest to lowest priority: book sources, regional databases, global maps; Supplementary Table 2). This ensured that sources with higher priority were used for those regions rather than maps from sources with lower priority, unless no map from sources with higher priority was available, in which case lower-priority sources supplied the map. Therefore, maps for each species were pieced together from the best available sources for each region of the species’ occurrence. The GeoMiB database was compiled in PostgrSQL version 9.1/9.6 with PostGIS version 2.0/2.3.2. As most of the underlying data are copyrighted, the database cannot be made fully publicly accessible.

Peak breeding and wintering months:

For each species, we characterised the breeding and non-breeding climatic niches using seasonal distribution and climate data. Breeding time is species-specific so we determined the three peak breeding months for each species individually using information from the literature (del Hoyo et al., 2019, and others; see Table S1 for details). Where no information was available on the breeding months, these were chosen using information from con-generic species breeding in the same geographic region (31 of 437 species in the final analyses, for details see Table S1). The three non-breeding months for each species were defined as 6 months later than the breeding season, a somewhat arbitrary decision given the different degree of climatic seasonality and migratory timing in different regions and species, but chosen to be globally consistent across all species.

This dataset includes the non-breeding distributions of migratory species. These are as extent-of-occurrence polygons sampled into 1 degree grid. These were combined with breeding ranges (not uploaded here) from https://macroecology.ku.dk/resources/global_avian_distributional_database/

Peak breeding and non-breeding months were collected from the literature (Handbook of the bird online). Migratory classification comes from Eyres et al 2017 and is already deposited on the Senckenberg data repository: http://dx.doi.org/10.12761/SGN.2017.10058