Changes in phenology and distribution are being widely reported for many migratory species in response to shifting environmental conditions. Understanding these changes and the situations in which they occur can be aided by understanding consistent individual differences in phenology and distribution and the situations in which consistency varies in strength or detectability.

Studies tracking the same individuals over consecutive years are increasingly reporting migratory timings to be a repeatable trait, suggesting that flexible individual responses to environmental conditions may contribute little to population-level changes in phenology and distribution. However, how this varies across species and sexes, across the annual cycle and in relation to study (tracking method, study design) and/or ecosystem characteristics is not yet clear.

Here, we take advantage of the growing number of publications in movement ecology to perform a phylogenetic multilevel meta-analysis of repeatability estimates for avian migratory timings to investigate these questions. Of 2,433 reviewed studies, 54 contained suitable information for meta-analysis, resulting in 177 effect sizes from 47 species.

Individual repeatability of avian migratory timings averaged 0.414 (95% confidence interval: 0.3–0.5) across landbirds, waterbirds and seabirds, suggesting consistent individual differences in migratory timings is a common feature of migratory systems. Timing of departure from the non-breeding grounds was more repeatable than timings of arrival at or departure from breeding grounds, suggesting that conditions encountered on migratory journeys and outcome of breeding attempts can influence individual variation.

Population-level shifts in phenology could arise through individual timings changing with environmental conditions and/or through shifts in the numbers of individuals with different timings. Our findings suggest that, in addition to identifying the conditions associated with individual variation in phenology, exploring the causes of between-individual variation will be key in predicting future rates and directions of changes in migratory timings. We, therefore, encourage researchers to report the within- and between- individual variance components underpinning the reported repeatability estimates to aid interpretation of migration behaviour. In addition, the lack of studies in the tropics means that levels of repeatability in less strongly seasonal environments are not yet clear.

The methods for collecting and analysing the data are presented in the manuscript and supporting information.

Here, we share the systematic search results, data, and code from Franklin et al. (2022). Individual repeatability of avian migration phenology: a systematic review and meta-analysis. Journal of Animal Ecology. You can also find all materials on GitHub at https://github.com/kirstyfranklin/Avian_migration_meta-analysis.

• WoS_tidy.csv: A tidied version of the Web of Science search results.
• scopus_tidy.csv: A tidied version of the Scopus search results.
• Rayyan_tidy_with_decisions.csv: A tidied version of the output from Rayyan abstract screening software. This includes the decisions made from abstract screening.
• Full_text_screening.csv: List of full texts screened, decision of whether to include or exclude, and reasons for excluding.
• meta-analysis_data-2022.csv: data extracted from the papers and used in the analyses. This can also be found in Table S1 in the Supporting Information.
• meta-analysis_descriptions.csv: Full description of the extracted data. This can also be found in the Supporting Information.
• Hackett.tre: Phylogenetic tree used in our analyses (provided by Benedikt Holtmann).
• Franklin_etal_Supporting_Information.html: Code for the exploration, processing and analyses in a user-friendly, self-contained knitted format. We recommend the reader to navigate this document if interested in the code.
• Franklin_etal_Supporting_Information.Rmd: R markdown file containing the processing and analyses code (to build the above .html file).

Missing values are indicated as "NA" in all datasets.